Benefit-And-Cost Analysis of MedicalInterventions: The Case of Cimetidine and

Peptic Ulcer Disease

September 1981

NTIS order #PB82-118910

LIBRARYOFFICE OF TECHNOLOGY ASSESSMENT CASE STUDY #11

THE IMPLICATIONS OF

COST-EFFECTIVENESSANALYSIS OF

MEDICAL TECHNOLOGY

SEPTEMBER 1981

BACKGROUND PAPER #2: CASE STUDIES OFMEDICAL TECHNOLOGIES

CASE STUDY #n: BENEFIT-AND-COST ANALYSIS OFMEDICAL INTERVENTIONS: THE CASE OF CIMETIDINE

AND PEPTIC ULCER DISEASE

Harvey V. Fineberg, M. D., Ph. D.Associate Professor

and

Laurie A. Pearlman, A. B.Research Analyst

Harvard School of Public Health, Boston, Mass.

OTA Background Papers are documents that contain information believed to beuseful to various parties. The information under-girds formal OTA assessments or isan outcome of internal exploratory planning and evaluation. The material is usuallynot of immediate policy interest such as is contained in an OTA Report or TechnicalMemorandum, nor does it present options for Congress to consider.

CONGRESS OF THE UNITED STATES

Office of Technology AssessmentWashington D C 20510

Library of Congress Catalog Card Number 80-600161

For sale by the Superintendent of Documents,U.S. Government Printing Office, Washington, D.C. 20402

Foreword

This case study is one of 17 studies comprising Background Paper #2 for OTA’sassessment, The implications of Cost-Effectiveness A Analysis of Medical Technology.That assessment analyzes the feasibility, implications, and value of using cost-effec-tiveness and cost-benefit analysis (CEA/CBA) in health care decisionmaking. The ma-jor, policy-oriented report of the assessment was published in August 1980. In additionto Background Paper #2, there are four other background papers being published inconjunction with the assessment: 1 ) a document which addresses methodologicalissues and reviews the CEA/CBA literature, published in September 1980; 2 ) a casestudy of the efficacy and cost-effectiveness of psychotherapy, published in October1980; 3 ) a case study of four-common diagnostic X-ray procedures, to be published insummer 198 1; and 4 ) a review of international experience in managing medical tech-nology, published in October 1980. Another related report was published inSeptember of 1979: A Review of Selected Federal Vaccine and Immunization Policies.

The case studies in Background Paper #2: Case Studies of Medical Technologiesa r-e being published individually. They were commissioned by OTA both to provideinformation on the specific technologies and to gain lessons that could be applied tothe broader policy aspects of the use of CEA/CBA. Several of the studies were specifi-cally requested by the Senate Committee on Finance.

Drafts of each case study were reviewed by OTA staff; by members of the ad-visory panel to the overall assessment, chaired by Dr. John Hogness; by members ofthe Health Program Advisory Committee, chaired by Dr. Frederick Robbins; and bynumerous other experts i n clinical medicine, health policy, Government, and econom-ics. We are grateful for their assistance. However, responsibility for the case studies re-mains with the authors.

Director

Ill

Advisory Panel on The Implications ofCost-Effectiveness Analysis of Medical Technology

John R. Hogness, Panel ChairmanPresident, Association of Academic Health Centers

Stuart H. AltmanDeanFlorence Heller SChoolBrandeis University

Sheldon LeonardManagerRegulatory AffairsGeneral Electric Co.

Barbara J. McNeilDepartment of RadiologyPeter Bent Brigham Hospital

Robert H. MoserExecutive Vice PresidentAmerican College of Physicians

Frederick MostellerChairmanDepartment of BiostatisticsHarvard University

Robert M. SigmondAdvisor on Hospital AffairsBlue Cross and Blue Shield Associations

Jane Sisk WillemsVA ScholarVeterans Administration

OTA Staff for Background Paper #2

Joyce C. Lashof, Assistant Director, OTAHealth and Life Sciences Division

H. David Banta, Health Program Manager

Clyde J. Behney, Project Director

Kerry Britten Kemp, * EditorVirginia Cwalina, Research Assistant

Shirley Ann Gayheart, SecretaryNancy L. Kenney, Secretary

Martha Finney, * Assistant Editor

Other Contributing Staff

Bryan R. Luce Lawrence Miike Michael A. RiddioughLeonard Saxe Chester Strobe]*

OTA Publishing Staff

John C. Holmes, Publishing Officer

John Bergling Kathie S. Boss Debra M. Datcher Joe Henson

● OTA contract personnel.

Preface

This case study is one of 17 that compriseBackground Paper #2 to the OTA project on thelmplications of Cost-Effectiveness Analysis ofMedical Technology.* The overall project wasrequested by the Senate Committee on Laborand Human Resources. In all, 19 case studies oftechnological applications were commissionedas part of that project. Three of the 19 were spe-cifically requested by the Senate Committee onFinance: psychotherapy, which was issued sepa-rately as Background Paper #3; diagnostic X-ray, which will be issued as Background Paper#.5; and respiratory therapies, which will be in-cluded as part of this series. The other 16 casestudies were selected by OTA staff.

In order to select those 16 case studies, OTA,in consultation with the advisory panel to theoverall project, developed a set of selectioncriteria. Those criteria were designed to ensurethat

●

●

●

●

●

●

●

as a group the case studies would provide:

examples of types of technologies by func-tion (preventive, diagnostic, therapeutic,and rehabilitative);examples of types of technologies by physi-cal nature (drugs, devices, and procedures);examples of technologies in different stagesof development and diffusion (new, emerg-ing, and established);examples from different areas of medicine(such as general medical practice, pedi-atrics, radiology, and surgery);examples addressing medical problems thatare important because of their high fre-quency or significant impacts (such asCost );examples of technologies with associatedhigh costs either because of high volume(for low-cost technologies) or high individ-ual costs;examples that could provide informativematerial relating to the broader policy andmethodological issues of cost-effectivenessor cost-benefit analysis (CEA/CBA); and

● examples with sufficient evaluable litera-ture.

On the basis of these criteria and recommen-dations by panel members and other experts,OTA staff selected the other case studies. These16 plus the respiratory therapy case study re-quested by the Finance Committee make up the17 studies in this background paper.

All case studies were commissioned by OTAand performed under contract by experts in aca-demia. They are authored studies. OTA sub-jected each case study to an extensive reviewprocess. Initial drafts of cases were reviewed byOTA staff and by members of the advisorypanel to the project. Comments were providedto authors, along with OTA’s suggestions forrevisions. Subsequent drafts were sent by OTAto numerous experts for review and comment.Each case was seen by at least 20, and some by40 or more, outside reviewers. These reviewerswere from relevant Government agencies, pro-fessional societies, consumer and public interestgroups, medical practice, and academic med-icine. Academicians such as economists and de-cision analysts also reviewed the cases. In all,over 400 separate individuals or organizationsreviewed one or more case studies. Although allthese reviewers cannot be acknowledged indi-vidually, OTA is very grateful for their com-ments and advice. In addition, the authors ofthe case studies themselves often sent drafts toreviewers and incorporated their comments.

The case studies were selected and designed tofulfill two functions. The first, and primary,purpose was to provide OTA with specific in-formation that could be used in formulatinggeneral conclusions regarding the feasibility andimplications of applying CEA/CBA in healthcare. By examining the 19 cases as a group andlooking for common problems or strengths inthe techniques of CEA/CBA, OTA was able tobetter analyze the potential contribution thatthese techniques might make to the managementof medical technologies and health care costsand quality. The second function of the caseswas to provide useful information on the spe-cific technologies covered. However, this wasnot the major intent of the cases, and

Some of the case studies are formal CEAS orCBAs; most are not. Some are primarily con-cerned with analysis of costs; others are moreconcerned with analysis of efficacy or effec-tiveness. Some, such as the study on end-stagerenal disease, examine the role that formalanalysis of costs and benefits can play in policyformulation. Others, such as the one on breastcancer surgery, illustrate how influences otherthan costs can determine the patterns of use of atechnology. In other words, each looks at eval-uation of the costs and the benefits of medicaltechnologies from a slightly different perspec-

tive. The reader is encouraged to read this studyin the context of the overall assessment’s objec-tives in order to gain a feeling for the potentialrole that CEA/CBA can or cannot play in healthcare and to better understand the difficulties andcomplexities involved in applying CEA/CBA tospecific medical technologies.

The 17 case studies comprising BackgroundPaper #2 (short titles) and their authors are:

Artificial Heart: Deborah P. Lubeck and John P.Bunker

Automated Multichannel Chemistry Analyzers:Milton C. Weinstein and Laurie A. Pearlman

Bone Marrow Transplants: Stuart O. Schweitz-er and C. C. Scalzi

Breast Cancer Surgery: Karen Schachter andDuncan Neuhauser

Cardiac Radionuclide Imaging: William B.Stason and Eric Fortess

Cervical Cancer Screening: Bryan R. Luce

Colon Cancer Screening: David M. Eddy”’CT Scanning: Judith L. WagnerElective Hysterectomy: Carol Korenbrot, Ann

B. Flood, Michael Higgins,” Noralou RO O S,and John P. Bunker

End-Stage Renal Disease: Richard A. RettigGastrointestinal Endoscopy: Jonathan A. Show-

stack and Steven A. SchroederNeonatal Intensive Care: Peter Budetti, Peggy

McManus, Nancy Barrand, and Lu AnnHeinen

Nurse Practitioners: Lauren LeRoy and SharonSolkowitz

Orthopedic Joint Prosthetic Implants: Judith D.Bentkover and Philip G. Drew

Periodontal Disease Interventions: Richard M.Scheffler and Sheldon Rovin

Selected Respiratory Therapies: Richard M.Scheffler and Morgan Delaney

These studies will be available for sale by theSuperintendent of Documents, U.S. Govern-ment Printing Office, Washington, D.C. 20402.Call OTA’s Publishing Office (224-8996) foravailability and ordering information.

Case Study #11

Benefit-and-Cost Analysis ofMedical Interventions: The Case of

Cimetidine and Peptic Ulcer Disease

Harvey V. Fineberg, M. D., Ph.D.Associate Professor

and

Laurie A. Pearlman, A.B.Research Analyst

Harvard School of Public HealthBoston, Mass.

AUTHORS’ ACKNOWLEDGMENTS

A number of people aided us at various stages in the preparation of this report.For their valuable advice, we would like to thank Dr. Morton I, Grossman, Center forUlcer Research and Education, Veterans Administration Center, Los Angeles; Dr.Thomas P. Al my, Dartmouth Medical School; Dr. Marvin H. Sleisenger, VA HospitalMedical Service, San Francisco; Dr. Benjamin A. Barnes, Center for the Analysis ofHealth Practices, Harvard School of Public Health; Dr. Janet Elashoff, Center forUlcer Research and Education, VA Center, LOS Angeles; Dr. Morton Paterson, Smith-Kline Corp.; Dr. Morris Olitsky, Robinson Associates, Inc.; Dr. George von Haunal-ter, Dr. Virginia Chandler, Stanford Research Institute; Dr. Richard Monson, Har-vard School of Public Health; and Dr. Murray C. Wylie, University of MichiganSchool of Public Health. Dr. Grossman also kindly shared with us a manuscript inpress. Several researchers described to us their work in progress. We are especiallygrateful to Dr. Wylie and to Dr. ]ohn F. Geweke of the University ofpartment of Economics. We thank Dr. Leighton Read of the Center forHealth Practices for introducing us to useful literature on ulcer disease.

We also appreciate the cooperation of Mr. Charles Dennison, Mr.berg, and Ms. Ethel Black of the National Center for Health Statistics,

Wisconsin De-the Analysis of

Donald Green-who promptly

responded to our numerous requests for information. Dr. Joan Standard of the Foodand Drug Administration was also very helpful.

By naming people who have advised and assisted us, we do not mean to shiftresponsibility for any errors or misinterpretations that may appear in this case study.

Contents

PageSummary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 3The Benefit-and-Cost Model for Medical

Interventions. ..,.... . . . . . . . . . . . . . . . . . 3Peptic Ulcer Disease. . . . . . . . . . . . . . . . . . . . . 4Cimetidine. . . . . . . . . . . . . . . . . . . . . . . . . . . 4The Benefit-and-Cost Model Applied to

Cimetidine. . . . . . . . . . . . . . . . . . . . . . . . . . 5Review of Benefit-and-Cost Analyses of

Cimetidine. . . . . . . . . . . . . . . . . . . . . . . . . . . 6Suggestions for Further Research . . . . . . . . . . . 6

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6The Benefit-and-Cost Model for Medical

Interventions . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Peptic Ulcer Disease . . . . . . . . . . . . . . . . . . . . . . 10

Definition and Etiology . . . . . . . . . . . . . . . . . . 10Symptoms and Diagnosis . . . . . . . . . . . . . . . . 11Treatment and Natural History . . . . . . . . . . . . 11Epidemiologic Patterns . . . . . . . . . . . . . . . . . . 13Cost of Illness . . . . . . . . . . . . . . . . . . . . . . . . . 18Summary . . . . . . . . . . . . . ........ . . . . . . 22

Cimetidine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23Physiologic Rationale and Development . . . . . 23Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

The Benefit-and-Cost Model Applied toCimetidine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Elements in the Analysis . . . . . . . . . . . . . . . . . 25Clinical Effects . . . . . . . . . . . . . . . . . . . . . . . . . 26Health System Effects . . . . . . . . . . . . . . . . . . . 39Outcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

Review of Benefit-and-CostAnalyses ofCimetidine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Available Analyses . . . . . . ........ . . . . . . 49The Study by Robinson Associates, Inc. . . . . . 49Guidelines for Review of Health Care

Cost Analyses . . . . . . . . . . . . . . . . . . . . . . . 57Suggestions for Further Research. . . . . . . . . . . . . 58References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

LIST OF TABLESTable No. Page

1. Selected Contemporary InterventionsUsed in Peptic Ulcer Disease . . . . . . . . . . . . . 7

2. Number of Deaths in the United States WithUlcer Disease as the Primary Cause, 1960-79. 15

3. Mortality Rates in the United States forDeaths Due to Ulcer Disease, 1953-78 . . . . . . 16

4. Number of U.S. Hospital Discharges WithUlcer Disease Diagnoses, 1966-78 . . . . . . . . . 16

5. Ulcer Disease in the United States Accordingto the Health Interview Survey, NCHS . . . . . 17

6. Costs of Ulcer Disease in 1975 as Estimatedby NCDD and SRI . . . . . . . . . . . . . . . . . . . . 19

7. Costs of Ulcer Disease in 1975 and 1977as Estimated by SRI. . . . . . . . . . . . . . . . . . . . 22

Table No. Page8.

9.

10.

11.

12<

1 3

14.

15.

16.

17.

18.

19.

20.

NCHS and CPHA Data on Number ofSelected Surgical Procedures in theUnited States, 1966-78. . . . . . . . . . . . . . . . . .Components and Measures of Componentsin a Benefit-and-Cost Analysis of Cimetidine.Short-Term, Double-Blind, Placebo-Controlled Studies of Cimetidine: Effect onDuodenal Ulcer Healing . . . . . . . . . . . . . . . .Short-Term, Double-Blind, Placebo-Controlled Studies of Cimetidine: Effect onGastric Ulcer Healing . . . . . . . . . . . . . . . . . .Short-Term, Double-Blind, Placebo-Controlled Studies of Cimetidine: Effect onDuodenal Ulcer Pain Relief . . . . . . . . . . . . . .Short-Term, Double-Blind, Placebo-Controlled Studies of Cimetidine: Effect onGastric Ulcer Pain Relief. . . . . . . . . . . . . . . .Controlled Trials of MaintenanceCimetidine in Peptic Ulcer. . . . . . . . . . . . . . .Results of Treatment With Two HypotheticalSubpopulations of Ulcer Patients: Type AMore Resistant to Treatment and Proneto Relapse Than Type B. . . . . . . . . . . . . . . . .Short-Term, Double-Blind, Placebo-Controlled Studies of Cimetidine: Effect onAntacid Consumption. . . . . . . . . . . . . . . . . .Number of Selected Surgical Procedures inthe United States, 1966-78 . . . . . . . . . . . . . . .Proportion of Patients With First-ListedDiagnosis of Ulcer Disease Having Surgery,1966-78 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Number and Rate of All and SelectedAbdominal Surgical Procedures in theUnited States, 1970-78. . . . . . . . . . . . . . . . . .Percentage Cost Savings Estimated From

22

26

29

31

32

33

36

37

40

41

43

43

Robinson Associates Study . . . . . . . . . . . . . . 56

LIST OF FIGURES

Figure No. Pagel. Benefit-and-Cost Model for Medical

L

2.

3.

4.

5.

6.

7.

Interventions . . . . . . . . . . . . . . . . . . . . . . . . . 8Number of Deaths in the United States WithUlcer Disease as the Primary Cause, 1966-79. . 15U.S. Hospital Discharges With Ulcer DiseaseDiagnoses, All Sites, 1966-78 . . . . . . . . . . . . . 17Paradigm Decision Tree: Cimetidine andAlternative lntervention Strategies . . . . . . . . . 26NCHS Data on Number of Selected SurgicalProcedures in the United States, 1966-78. ,... 42CPHA Data on Number of Selected SurgicalProcedures in the United States, 1966-78. . . . . 42Number of Deaths in the United StatesFrom Ulcer Disease, 1976-79 . . . . . . . . . . . . . . 46

Case Study #11:

Benefit= and-Cost Analysis ofMedical Interventions: The Case of

Cimetidine and Peptic Ulcer Disease

Harvey V. Fineberg, M. D., Ph. D.Associate Professor

and

Laurie A. Pearlman, A.B.Research Analyst

Harvard School of Public HealthBoston, Mass.

SUMMARY

Introduction

This case study presents a conceptual modelfor assessing the benefits and costs of medicaltechnology, and uses this model as a frameworkfor analyzing the benefits and costs of cimeti-dine in the treatment of peptic ulcer disease.

The body of the study is organized into threemajor parts: 1) a description of the benefit-and-cost model; 2) a selective description of clinicalfeatures, epidemiologic patterns, and costs ofpeptic ulcer disease; and 3) a review of the de-velopment, dissemination, health benefits, andresource costs of cimetidine. The study endswith a critique of one major analysis of cimeti-dine’s costs and benefits and some suggestionsfor further research.

The Benefit-and-Cost Model forMedical Interventions

The benefit-and-cost model stresses that anevaluation of medical technology must apply toan identifiable patient population and a specifichealth intervention. An intervention may have adiagnostic or a therapeutic purpose. A patient

population may be defined in terms of diagnos-tic category, a clinical sign or symptom, a riskfactor, or a complication of disease.

The model posits two principal classes of ef-fects: clinical effects and health system effects.The specific components of each depend on thepopulation and intervention of interest. Clinicaleffects and health system effects interact andlead to an outcome, expressed in terms of healthstatus and resource costs.

The components of the model apply to bothcost-effectiveness analyses (CEAs) and benefit-cost analyses (BCAs), but the two analytic ap-proaches have distinct purposes and measuresome components in different ways. The modelcan also serve as a basis for identifying thestructural components of a decision analysisthat compares alternative medical interven-tions.

The model and a set of guidelines for reviewof health care benefit-and-cost analyses are usedto organize and guide our discussion of the costsand benefits of cimetidine in peptic ulcerdisease.

4 ● Background Paper #.2: Case Studies of Medical Technlologies

Peptic Ulcer Disease

Ulcers probably have multiple causes, butgastric acid and pepsin appear to be necessary

ingredients. Epigastric pain (pain in the uppermiddle abdomen) is often a prominent symptomof peptic ulcers, but the clinical presentation isvariable. Furthermore, typical ulcer symptomsmay be caused by conditions other than ulcers.A definite diagnosis requires direct visualizationby endoscopy or radiographic imaging of theulcer. Specific treatments of ulcer disease aredirected at reducing the presence or effects ofgastric acid.

Ulcer disease is a chronic condition withspontaneous remissions and recurrences. Ratesof complications and mortality from ulcers arerelatively low, Excessive mortality appears to bepresent only in the first year or so followingdiagnosis. Little reliable information existsabout the natural history of ulcer disease in thegeneral population.

Peptic ulcer is a common condition that af-fects millions of Americans at some time duringtheir lives. The best available epidemiologicevidence suggests that about 250,000 Americansdevelop new peptic ulcers each year. New duo-denal ulcers are more than four times as com-mon as new gastric ulcers. Some studies havefound that the incidence of duodenal ulcer risesgradually with age; others have found that it re-mains fairly constant above age 35. Above age40, the incidence of gastric ulcer appears to risemore dramatically than the incidence of duo-denal ulcer. Duodenal and gastric ulcers are epi-demiologically distinct. Several lines of clinicaland epidemiologic evidence suggest that overthe past 20 years the occurrence of new ulcershas declined, or ulcer disease is generally lesssevere than it was at one time, or both.

The basis for some estimates of the costs ofulcer disease and the benefits of treatment is theHealth Interview Survey of the National Centerfor Health Statistics (NCHS). Results of theHealth Interview Survey, based on self-reportedconditions in a household survey, however, ap-pear to overestimate the occurrence and conse-quences of ulcer disease.

We estimate that the costs of ulcer disease in1975 were approximately $2 billion. Just underhalf of this total was due to health care expend-itures (direct costs), and the remainder was dueto productivity losses from morbidity and mor-tality (indirect costs). Our estimate is based on areview of two independent analyses of the costsof ulcer disease, one by the National Commis-sion on Digestive Diseases (NCDD) and theother by the Stanford Research Institute (SRI).The NCDD and SRI estimates of the total costsof ulcer disease in 1975, $1.3 billion and $2.6billion, respectively, differ by approximately

$1.3 billion. The NCDD and SRI estimates of di-rect costs differ by approximately $400 million,a difference that reflects differences in the twostudies’ methods and differences in their detailedassumptions and procedures. Their indirect costestimates differ by approximately $900 million,a difference that reflects differences in thestudies’ projected morbidity losses. SRI’S in-direct cost estimate, the higher one, is based ondata from the Health Interview Survey, which isan inflated indicator of disease-specific morbidi-ty. In both the NCDD and SRI studies, esti-mated indirect costs are based on a rather lowdiscount rate—2.5 percent. Use of a smaller dis-count rate increases the present value of futureearnings, thereby increasing apparent costs ofillness due to morbidity and premature death.

In addition to estimating costs for 1975, SRIprojected an estimate of peptic ulcer costs in1977. Because of unwarranted assumptions ofgrowth in the morbidity of ulcer disease and useof more expensive resources, the problem ofoverestimated costs is compounded for 1977.

Cimetidine

Cimetidine represents a new class of hista-mine antagonists, called H2-receptor antago-nists, which block stimulation of gastric acidsecretion. The product was developed after ex-tensive research by the Smith Kline & Frenchpharmaceutical firm and is marketed under theregistered brand name Tagamet®.

The Food and Drug Administration (FDA)approved the use of cimetidine for up to 8 weeksby patients with duodenal ulcer disease or hy-

persecretory conditions such as Zollinger-Elli-son syndrome in August 1977. Use of cimetidinespread rapidly. Since March of 1978, the drughas been prescribed in approximately 60 percentof ambulatory visits for ulcer disease eachmonth. In 1978, a conservatively estimated 1.5million to 2 million U.S. ambulatory patientswith ulcers and other symptoms of gastric acid-ity were treated with cimetidine. Worldwidesales to hospitals and pharmacies in 1979 prob-ably exceeded $400 million.

The Benefit- and-Cost ModelApplied to Cimetidine

Organized according to the benefit-and-costmodel presented earlier, this part of the casestudy describes available information about theeffects of cimetidine. It deals separately withcimetidine’s clinical effects, its health system ef-fects, and its potential impact on outcome.

Numerous controlled studies of patients withduodenal ulcer confirm that cimetidine pro-motes healing and provides faster and morecomplete pain relief than placebo. Less con-clusive evidence suggests the drug may be moreeffective than placebo for patients with gastriculcer. An intense antacid program appears to beabout as effective as cimetidine for patients withduodenal ulcers, but more evidence of this isstill needed. Clinical studies have also shownthat relief of symptoms is not a reliable indi-cator of healing. In general, European studieshave found more favorable results with cimeti-dine than have U.S. trials.

Cimetidine used for up to 2 months appearsto be a relatively safe drug. Most known side ef-fects are minor or reversible; however, recentlyreported changes in the bacterial flora of thestomach and endocrinologic effects may bemore significant. Available studies of mainte-nance cimetidine for periods up to 1 year do notalter the current assessment of the drug’s rela-tive safety. As is the case with any new drug,possible long-term consequences of cimetidine’suse are not known.

Compared to an intense course of antacids,cimetidine is about equally effective and morerisky, but less troublesome to patients with

duodenal ulcer. Cimetidine plus a moderateamount of antacid. costs no more than a thera-peutically equivalent course of intense antacidtherapy. Experts now differ in their recommen-dations for initial therapy of duodenal ulcer,some favoring cimetidine and others antacids. Areasonable approach is to select therapy basedon each individual patient’s preferences andpersonality.

Compared to placebo, maintenance treatmentwith cimetidine for as long as 1 year signifi-cantly reduces the chance of ulcer recurrenceduring the treatment period. Once cimetidinetreatment is discontinued, patients appear torelapse at the same rate as they would havewithout maintenance treatment. We are awareof no controlled trials comparing maintenancecimetidine to treatments other than placebo.There is little empirical evidence either thatcimetidine prevents future complications ofulcer disease or that cessation of cimetidine pro-motes complications. At present, FDA is consid-ering approval of cimetidine for use longer than8 weeks in patients with duodenal ulcers whoare at high risk for surgery.

In European trials, but not in U.S. studies,cimetidine-treated patients tend to consume lessantacid than placebo-treated patients. Very lim-ited empirical data are currently available onthe possible effects of cimetidine on use of othermedication, on diagnostic tests, and on physi-cian visits. There are several studies under waythat may shed light on these matters.

Data from NCHS show that in 1978, the firstfull calendar year after cimetidine was intro-duced in the United States, there was an unex-pectedly sharp decline in the rates of surgery forulcer disease. This drop occurred against abackground of falling rates of surgery and hos-pitalization for ulcer disease over the previousdecade. Although other explanations of thelarge drop in surgery for ulcer disease in 1978are possible, the widespread use of cimetidinemay have been a contributing factor.

There is little evidence of any effect of cime-tidine on mortality from ulcer disease. In oneshort-term trial and one maintenance study, pa-tients treated with cimetidine lost significantly

fewer days of work than patients taking pla-cebo, but no controlled study has comparedwork loss among patients receiving different ef-fective treatments.

Review of Benefit- and-Cost Analysesof Cimetidine

Several analyses of the resource cost implica-tions of cimetidine have been undertaken in thepast few years. One major study, by RobinsonAssociates, Inc., estimated that if cimetidinehad been used in 80 percent of duodenal ulcerpatients, 1977 cost; for duodenal ulcer disease inthe United States would have been reduced by$645 million. This conclusion rests on subjectiveestimates provided by selected physician expertsof the clinical and health system effects ofcimetidine, and on independent estimates of thecosts of duodenal ulcer disease based in part onthe costs of peptic ulcer disease in 1977 pro-jected by SRI.

Our critical review of the Robinson Associ-ates study focuses on the following five areas:1) the accuracy of the clinical and health systemeffects projected by their physician experts,2) the relation between a percentage reduction inhealth services devoted to ulcer disease and sav-ings in health resources, 3) the accuracy of theestimated total costs of all duodenal ulcer dis-ease used as a baseline for percentage savings,4) the applicability of projected percentage ef-fects to the total population of patients withduodenal ulcer disease, and 5) the validity of the

INTRODUCTION

This case study has both a specific and a gen-eral objective. The specific objective is to assessavailable evidence about the benefits and costsof cimetidine, a recently introduced pharma-ceutical agent, in the treatment of peptic ulcerdisease. The general objective is to present anapproach to the evaluation of medical technol-ogy that emphasize:; salient features of both thepatient population and the medical interventionof interest. The specific purpose serves thegeneral one —we present our analysis of cimeti-

methods used to compute average percentage ef-fects due to cimetidine. We question some of theassumptions and methods used in each of theseareas. Aside from the fundamental issue of pos-sible inaccuracy in the physician estimates, webelieve the Robinson Associates study over-states expected savings by twofold to threefold.Potential bias introduced by the selection ofphysician informants would increase the magni-tude of the overestimate.

Despite our criticisms of the study by Robin-son Associates, available data and analyses sup-port the belief that cimetidine currently savesmore health resources than it costs. Whetherfurther studies will affirm this conclusion ornew developments will alter cimetidine’s costeffectiveness are empirical questions for thefuture.

Suggestions for Further Research

If the object of analysis is to help inform clini-cians and health policy decisionmakers aboutthe efficient use of resources in the care of pa-tients with ulcer disease, then the most helpfulapproach would be to do a CEA of alternativeinterventions oriented to particular groups ofpatients, comparing incremental clinical bene-fits to marginal resource costs. Rather than enu-merate the resource and health implications fora cross-section of the population in a singleyear, the analysis might equally or more useful-ly focus on a cohort of patients and project ef-fects over their lifetimes.

dine and ulcer disease as an application of thegeneral model for benefit-and-cost evaluation. ’

Peptic ulcer is a logical choice for this kind ofevaluation for several reasons. As a diagnosticcategory, it comprises several anatomically andepidemiologically distinct entities, but these aresufficiently related to make peptic ulcer a validdiagnosis. This common medical problem has a

‘We use the term “benetit-dnd-c(wt analysl~ ‘ t[) enc[)m pass bothcc)st-ettect iveness and C(WI -bend it (or bend It -cost I analyses.

Case Study #11 Benefit-and-Cost Analysis of Medical lnterventions: The Case of Cimetidine and Peptic Ulcer Disease ● 7

highly variable clinical course and an evolvingpattern of clinical expression and occurrence.These features help demonstrate that a carefulassessment of a disease can be as important tothe evaluation of technology as is a comprehen-sive understanding of the technology itself.

Our selection of cimetidine emerged gradual-ly. Initially, we wanted to use an assessment ofpeptic ulcer disease as a backdrop for reviewingthe costs and benefits of a number of diagnostictests and therapeutic interventions such as thoselisted in table 1. (In addition to these contem-porary interventions, the variety of clinical ap-proaches to ulcer disease over the past centuryconstitutes a rich history for anyone interestedin the progress and byways of medical science(85).) It soon became evident that we couldeither review several interventions superficially

Table 1 .—Selected Contemporary InterventionsUsed in Peptic Ulcer Diseasea

—.- .—Diagnostic interventions

—

Imaging● Air-barium, double-contrast radiographic studies● Fiberoptic endoscopy

Physiologic function tests● Gastric secretory testing

Therapeutic interventionsMedical

● Antacids● Anticholinergics● Cimetidine

Surgical● Partial gastrectomy● Truncal vagotomy, with antrectomy or drainage

procedure● Highly selective vagotomy

.—aDeSCrlptlOnS of Some of these lnterv~nt;ons are prov!ded later lri~h Is st Udy

—.

or analyze one in detail. We elected the lattercourse, believing it would produce a morecoherent exposition of the general model.

We selected cimetidine for several reasons.First, it is a recent innovation that was dissemi-nated rapidly. Second, as a chemical entity, thedrug cimetidine does not evolve technically (un-like, for example, endoscopy) and its effects arerelatively independent of the skill of the clini-cian (unlike, for example, surgery’s). Since thereare fewer such complications related to the tech-nology, we can appreciate more readily thecomplexity introduced by features of the dis-ease. Finally, the clinical effects of cimetidinehave been studied extensively, and its costs andbenefits have been and continue to be formallyassessed.

The body of this case study is organized intothree main parts. First, we present a briefdescription of a general benefit-and-cost modelfor evaluating medical interventions. Second,we describe pertinent clinical and epidemiologicfeatures of peptic ulcer disease, and summarizeseveral cost-of-illness studies of the disease.Third, we review the development, dissemina-tion, health benefits, and resource costs ofcimetidine. As a framework for the analysis ofcimetidine, we use the general benefit-and-costmodel for evaluating medical interventions anda set of questions provided in a section of thiscase study entitled “Guidelines for Review ofHealth Care Benefit-and-Cost Analyses. ” Weoffer a critique of one major analysis of cimeti-dine’s costs and benefits and end with sugges-tions for further research.

THE BENEFIT-AND-COST MODEL FOR MEDICAL INTERVENTIONS

Every assessment of the benefits and costs ofmedical care should apply to an identifiable pa-tient population and a specific health interven-tion. The ultimate objective of a benefit-and-cost assessment is to measure the effects that aspecific intervention has on the health outcomeof those patients and on resource consumption.Implicit in this objective is a societal perspec-tive. The health and resource outcomes resultfrom the intervention’s direct and induced ef-

fects on the clinical well-being of patients andon other components of the health system.These relations and interactions are summarizedin the benefit-and-cost model shown in figure 1.

The principal components of the model are asfollows: 1) population, 2) intervention, 3) clin-ical effects, 4) health system effects, and 5) out-come. The population may be delineated interms of a particular diagnosis or pathologic en-

8 ● Background Paper #2: Case Studies of Medical Technologies

Figure 1 .—Benefit-and-Cost Model for Medical Interventions

tity (e.g., peptic ulcer), a risk factor (e.g.,cigarette smoking), a clinical sign or symptom(e.g., dyspepsia), or a complication of disease(e.g., gastrointestinal hemorrhage). Interven-tions are of two broad types: tests, which aremeant to produce information about the clinicalstatus of the patient; and treatments, which areintended to alter the development or course ofdisease. 2 Clinical effects include any physical orpsychological changes that may alter the healthstatus of the patient; these effects may be shortor long term. Health system effects include allchanges in the methods and means of medicalcare that are consequent to the initial interven-tion. The health outcome is reflected in mortali-ty and morbidity, i.e., in the length and qualityof life. The resource outcome, resource costsand savings, pertains to net effects on socialresource consumption.

‘The distinction between tests and treatments is useful analyti-cally, but not absolute, since, albeit rarely, a therapeutic trial mayalso have a diagnostic intent.

The general framework of the model appliesto any intervention and patient population. Thedetailed components under clinical and healthsystem effects, however, will vary with the par-ticular disease and intervention being consid-ered. Thus, for example, if we were analyzingan intervention that might affect chronic diseasein the elderly (e.g., a prevention or treatmentfor senile dementia), we would want to considernursing home use explicit] y under health systemeffects. In general, the components identified forclinical effects and for health system effectsshould be: pertinent to the disease and interven-tion; complete, in that all important effects areconsidered; and mutually exclusive, so that asingle effect is not counted twice. They shouldalso be components for which readily availableand accurate measures can be obtained. The va-lidity and feasibility of a cost-effectiveness or abenefit-cost evaluation depend on the extent towhich the analytic components conform to thesecriteria.

According to the model, an intervention itselfmay alter a patient’s clinical status, effectchanges in the health system, and consume re-sources. Clinical effects include both the ad-vantages and risks of care. The direct clinical ef-fects of a test are typically limited to side effectsand complications, but a test can also alter clini-cal status by inducing changes in the health sys-tem, primarily by altering the choice of therapy.A treatment is intended to have direct clinicaleffects, but can also alter subsequent use ofdiagnostic procedures (a health system effect)by changing the course of the disease.

Clinical effects and health system effects caninteract in both directions. As illustrated in themodel, interactions among the various healthsystem components may also occur. Changes ina patient’s clinical status are likely to alter thefuture course of medical care for the patient;and shifts in the medication, hospitalization,surgery, or other care given to the patient arelikely to affect clinical status.

Although the model is premised on the appli-cation of a particular intervention for a par-ticular disease, health system effects may not belimited to the target disease entity. For ex-ample, if an intervention reduces the number ofphysician visits for a particular disease, it couldalter the number of diagnostic tests and amountof medication employed for other diseaseproblems.

Health outcome typically includes mortalitymeasures, such as number of deaths, age-ad-justed death rates, or years of life lost. It also in-cludes morbidity measures, such as quality-of-life or health-status indexes. Morbidity andmortality may also be combined into a unitarymeasure, such as quality-adjusted life years(152) or another multiattribute utility scale (87).As indicated in the model, morbidity and mor-tality also have direct implications for pro-ductivity and hence for social resource con-sumption.

The benefit-and-cost model for a particularpopulation and intervention suggests the com-plexity of undertaking a comprehensive assess-ment of either all uses of a single intervention orall interventions for a particular population.

Consider two interventions, endoscopy andcimetidine, and the population of patients withduodenal ulcer. Both interventions are used insome patients with duodenal ulcer; each is usedindependently of the other in some patients withduodenal ulcer; and both interventions are alsoused, singly or together, in some patientswithout duodenal ulcer. Moreover, neither in-tervention is used in some patients with duo-denal ulcer. Compound these partial overlapswith additional interventions, add variations inthe particular populations for which data areavailable, and the magnitude of the problembegins to become apparent.

The benefit-and-cost framework outlined hereis applicable to both BCA, or cost-benefit anal-ysis (CBA), and CEA. A BCA assesses the netvalue of an intervention by summing all effectson a common scale. Typically, both resourceexpenditures and health outcome effects areassigned monetary values. A variety of meansto measure the resource value of health benefitshave been proposed; the most widely used is ex-pected productivity loss based on discountedfuture earnings at the age of death or disability(31,89). Thus, a BCA converts decreased deathsand disabilities into increases in productivity,and treats them as the indirect benefits of ahealth intervention. These indirect benefits areadded to any direct savings in health resourceconsumption (the direct benefits) to yield a netvalue.

In the cost-effectiveness approach, the aim isto measure the efficiency with which an inter-vention achieves health benefits. The questionsaddressed in CEA are: 1) What is the most effi-cient way to achieve a particular health benefit?or 2) Given specified available resources, whatintervention strategy offers the greatest gain inhealth benefit? Answering these questions re-quires the commensuration of different types ofbenefits, such as morbidity and mortality, butpermits benefits to be measured in their own,nonmonetary terms. A cost-effectiveness ap-proach is more likely than a benefit-cost ap-proach to preserve a sense of intangible healthcare benefits, which in the latter are typicallynoted and left unassessed. Although CEA maybe more suitable for comparing alternative in-

terventions for a particular disease, however,BCA is necessary for comparing health carewith other socially desirable uses of resources.

If one chooses to develop a decision analysiscomparing the use of a particular interventionwith its alternatives, the benefit-and-cost modelcan serve as a useful basis for identifying perti-nent structural components: chance events (e.g.,important results of the principal and subse-quent interventions), choices (e.g., use of otherhealth system components), and outcomes (e.g.,net benefits and costs). The decision-analyticapproach is a prescriptive model for choosingamong alternative treatment strategies. Even iftechnically correct in all assumptions and com-putations, a decision analysis does not neces-sarily predict the management strategies em-ployed by physicians. In estimating the cost ef-fectiveness of a given intervention, it is equally,if not more, important to apply a descriptivemodel (i. e., to base estimates on changes inmanagement strategy that occur in practice).The distinction between prescriptive and de-scriptive assessment is analogous to the differen-tiation between efficacy (effects under ideal con-

PEPTIC ULCER DISEASE

Definition and Etiology

A peptic ulcer is a crater that extends throughthe full thickness of the mucosa (mucous mem-brane) of the stomach or duodenum (the first orproximal portion of the small intestine). Thepathologic appearance of benign gastric(stomach) and duodenal ulcers is similar; bothare believed to be related to too much stomachacid and pepsin for the level of mucosalresistance (82) .3 Although the presence ofstomach acid is necessary for ulcers to develop,the level of acid is often normal in patients withulcer disease; these patients presumably haveimpaired tissue resistance. Sturdevant andWalsh (140) list 17 factors other than excessivegastric acidity that may predict increased—..—

‘l)(’p\tn\ ,Ir(’ dIgw[ IL’(’ (,n.ynlc~ th.lt .lre .1[ tlve onlv in the prc\-t,n( t’ ()1 <I( ICI }{(’nc t’, i t i~ d I t tic u I t t () wpa r.] tc t ho role> of ~cid ~ndp(p~ln I n t h(’ p.ltho~(,n(,~ts (It ulcer.

ditions) and effectiveness (effects under averageconditions) noted in reports from OTA (112).

Any benefit-and-cost analysis encounters nu-merous conceptual and practical difficulties.These range from the presence of uncertaintyand the lack of reliable information to questionsof measurement and methods of aggregationover persons and time, to value judgments. Sys-tematic reviews of methodologic issues in CEAand BCA in health have been presented by otherauthors (149,151).

Following descriptions of peptic ulcer diseaseand cimetidine in the next two parts of this casestudy, we present an analysis of the costs andbenefits of cimetidine in peptic ulcer disease,using the general benefit-and-cost model de-scribed above. Later in the case study, we pre-sent a set of guidelines in the form of questionsto be used in reviewing benefit-and-cost anal-yses in health care. These guidelines presumefamiliarity with the basic assumptions and ap-proaches in BCA and CEA. We believe they arehelpful for review of benefit-and-cost analysesof cimetidine such as that presented in the nextto the last part of this case study.

likelihood of developing duodenal ulcer. Thesefactors include sex, age, blood type, a fewdiseases, and habits such as smoking and drink-ing coffee. Despite the popular notion of the“high anxiety, ulcer-prone person, ” psy-chological stress and personality factors havenot been shown conclusively to be related to thedevelopment of ulcers,

The gastrointestinal tract is a continuousorgan, and there is a continuum in the anatomiclocation of ulcers in the stomach and duo-denum. For unknown reasons, peptic ulcersshow a predilection for areas at or near rnucosaljunctions (81). Since gastric and duodenal ulcersappear to differ in generic and other features,there are reasons to consider them separately inclinical and epidemiologic studies. Often, how-ever, they are considered together, and the situ-ation is further complicated by the frequent oc-

currence of new duodenal ulcer in patients whopreviously had gastric ulcer (17).

Symptoms and Diagnosis

Both duodenal and gastric ulcers produce ab-dominal pain in the patient, typically in theepigastric region (upper middle abdomen). Lessoften, they produce nausea and vomiting. Usu-ally, the pain is relieved by food, but in somepatients, food may exacerbate pain. Most pa-tients with epigastric pain do not have ulcers; aDanish study found that 68 percent of men and83 percent of women with epigastric pain didnot have ulcers (cited by 140). Some patients de-velop painless ulcers and have bleeding or per-foration as the first manifestation of ulcer dis-ease (119,140).

The specific diagnosis of peptic ulcer dependsprimarily on imaging examinations, with eitherbarium X-rays or more direct fiberoptic endos-copy. Fully flexible fiberoptic gastroscopes wereintroduced in 1958 (77). Numerous technical im-provements made since have enhanced the flexi-bility, ease of control, and clinical usefulnessof these instruments (10). Endoscopists haveformed their own professional society (theAmerican Society for Gastrointestinal Endos-copy), and the endoscopic procedure is widelyused. Radiographic examination of the stomachhas been improved in recent years by the use ofan air-barium, double-contrast technique in-volving high-density barium sulfate, efferves-cent tablets to distend the stomach and simethi-cone to break up small air bubbles (96).

Acid secretion and other tests play a second-ary role in diagnosis, except in occasional pa-tients, such as those whose ulcer is caused bygastrinoma (a gastrin-secreting tumor that pro-duces the Zollinger-Ellison syndrome of severeulcers, intractable pain, and diarrhea).

Treatment and Natural History

The treatment of peptic ulcer disease is in-tended to relieve symptoms, promote healing,and prevent recurrences and complications(140). Gastric acid is the focus of contemporaryspecific treatment for peptic ulcer—reducingacid secretion by pharmacologic or surgical

means, neutralizing acid with antacids, or in-creasing tissue defenses against acid.4 S o m ephysicians begin treatment on the basis of clin-ical symptoms without pursuing a definitivediagnosis (140). A U.S. patient who is diag-nosed as having a new peptic ulcer will typicallybe told to eat a regular, nutritious diet and toavoid aspirin, alcohol, cigarettes, and coffee.Specific medication might include antacids orcimetidine and possibly anticholinergic drugs(drugs that block the passage of impulsesthrough the parasympathetic nerves).

Surgery is normally reserved for patients withrecalcitrant symptoms, frequent relapse, orcomplications such as bleeding, perforation, orobstruction. A large variety of surgical proce-dures has been advanced over the past century,and there is considerable difference of opinionabout the optimal timing and selection of anelective surgical procedure for patients withpeptic ulcer disease (44,73,111,124). Highlyselective vagotomy has been advocated recently(78). This procedure entails transection of onlythose nerve fibers that supply the lower esopha-gus and body of the stomach; the nerve supplyto the remainder of the stomach and to other ab-dominal organs is left intact. Proponents ofhighly selective vagotomy believe that this pro-cedure obviates some unpleasant side effects ofstandard vagotomy (cutting of the vagus nerve).The surgical procedure is technically demand-ing, however, and its comparative effectivenessin the hands of many different surgeons remainsto be shown. Cochran, et al. (30) have describedthe complexity of evaluation and requirementsfor adequate assessment of any surgical treat-ment for ulcer disease.

Over the years, an enormous variety of non-surgical therapeutic regimens has been em-ployed to treat peptic ulcers. An example is diet:Leube introduced a starvation regimen in 1876;Lenhartz recommended frequent small feedingsin 1906; and Sippy proposed a bland diet in1915, variations of which remained popular for-

many years (84, 119). Now dietary restrictionsare believed to play no role in the managementof peptic ulcers (119,140), Despite the demon-strated ineffectiveness of diet in the treatment ofulcers, special diets are still widely prescribed(153). The plethora of unsubstantiated, but tra-ditional and trusted, treatments led one author-ity to exclaim in the late 1960’s: “Few conditionsprovide such a splendid opportunity for practic-ing 19th century medicine in the second half ofthe 20th century as gastric ulcer” (37). The1960’s witnessed the introduction, spread, anddecline of gastric freezing, a nonsurgical treat-ment intended to reduce stomach acid and pro-mote healing. Such treatment was eventuallyproven to be ineffective and occasionally harm-ful. Some clinicians have also used X-ray ther-apy to treat ulcer disease in selected patients,and renal failure has been reported as one latecomplication of such therapy (143).

The reasons for such diverse treatments, andparticularly, for the extended use of some in-effective approaches, rest partly in the expres-sion and natural history of ulcer disease. First,as mentioned earlier, the cardinal symptom ofulcer disease is stomach pain; so subjective anexpression of illness as stomach pain may re-spond to suggestion or placebo. Second, ulcersoften heal spontaneously; thus, any apparentsuccess with treatment should be compared tothe natural rate of healing. Finally, ulcer diseasetends to be chronic, with recurrences and remis-sions; effective short-term treatment may ormay not alter the long-term outlook.

The subjective nature of ulcer disease and itsvariable course suggest that evaluations of treat-ment must be controlled carefully for bias, pref-erably with double-blind randomization, Onthis score, the state of clinical assessments ofpeptic ulcer disease appears to be improving.Chalmers, et al. (25) reviewed studies of pepticulcer treatments published in a leading gastro-enterology journal and found that more than sopercent of the therapeutic trials published after1976 had a randomized, controlled design, com-pared to 30 percent or fewer of those publishedbetween 1970 and 1974. In addition, improvedendoscopic methods now permit a more defini-

tive diagnosis to be established in patients in-cluded in clinical trials.

Assessment of long-term results of any inter-vention in ulcer disease requires comparison tothe natural history of the disease. Ideally, thenatural history of peptic ulcer disease would bedefined through long-term followup of a rep-resentative sample of patients with ulcer dis-ease. As discussed below, however, availableinformation about the natural course of ulcerdisease is fragmentary.

Fry (57) reported a 5- to 15-year followup of212 patients with ulcer disease diagnosed be-tween 1948 and 1957 in his general practice inLondon. He found that symptoms tended to re-cur and worsen for the first 5 to 10 years, andthen usually diminished, irrespective of treat-ment. Sixteen percent of patients with duodenalulcer and 18 percent with gastric ulcer requiredsurgery. Complications of bleeding occurred in14 percent and complications of perforation in 6percent. Only one patient died from causes re-lated to ulcer.

Krause (91) found similarly low mortalityfrom ulcer in 371 Swedish patients with duo-denal ulcer followed for 25 to 35 years. In astudy based on a 50-percent random sampleof all patients with duodenal ulcer diagnosed be-tween 1963 and 1968 in the population of500,000 persons living in Copenhagen County,Denmark, Bonnevie (20) found a significant ad-ditional mortality risk in the first year followingdiagnosis of ulcer, but not thereafter. Griebe, etal. (66) interviewed 154 patients living; inCopenhagen in 1976 who had developed duo-denal ulcer disease in 1963. One hundred andtwenty patients (78 percent) had been treatedmedically; nearly half of these patients wereasymptomatic, and approximately 16 percentstill had severe symptoms. Thirty-four patientshad been treated surgically, but their clinicalstatus is not described further.

A Veterans Administration (VA) study fol-lowed more than 600 patients with gastric ulcerdiagnosed in 16 hospitals during a 7-year period(69). More than 75 percent of the patients ex-perienced ulcer healing with medical treatment

within 12 weeks, but 42 percent of these patientshad one or more recurrences in the following 2years. Patients who failed to heal initially wereassigned randomly to further medical or sur-gical treatment. Two years later, a higher pro-portion of patients in the the surgical groupwere alive and free of symptoms and recur-rence, but the differences between the surgicaland medical groups were not statistically signifi-cant. Expressed as a proportion of incidence peryear among all patients, complications of hem-orrhage occurred in 2.5 percent, obstruction in1.2 percent, and perforation in 0.6 percent.

For several reasons, the available data on thenatural history of ulcer disease are unsatisfying.The data come from different geographic 1oca-tions and cover different time periods and dif-ferent mixes of patients with duodenal andgastric ulcer. Patients received various treat-ments (and differing proportions were offeredsurgery), and results reflect the history undervaried treatments rather than a natural historyof the disease. Rates of complication and deathdue to ulcer are low and difficult to assess inrelatively small cohort studies; Bonnevie’sanalysis (20) is exceptional in specifying the at-tributable mortality risk from newly developedulcer disease. Finally, such studies of the clinicalcourse of disease are necessarily dated. If thecourse of ulcer disease is changing over time,data from previous patient cohorts may not ap-ply today.

Epidemiologic Patterns

Ulcer disease is a common medical problem,but has apparently become less common overthe past 20 years. Here we summarize estimatesof the present incidence and prevalence of ulcerdisease and describe the basis for the conclusionthat ulcer disease is occurring less frequently.We conclude this section with comments di-rected specifically to the Health Interview Sur-vey conducted by NCHS, since although its re-sults are used in several estimates of the costs ofulcer disease and the benefits of intervention, webelieve the Health Interview Survey overesti-mates the prevalence of ulcer disease.

Several aspects of the definition of diseaseand of data collection limit our ability to com-

pare results from different studies of the oc-currence of ulcer disease in the United States to-day. Any effort to assess the incidence and prev-alence of ulcer disease is necessarily restricted toa particular place and time. Insofar as there aregeographic variations and shifts in the diseaseover time, projections to other countries and tothe present data are uncertain.

In addition, different studies define the prev-alence and incidence of this chronic and recur-rent disease differently. Some (e.g., 35,36,57)define prevalence to mean the “period preva-lence, ” or the number of patients who sufferfrom ulcer disease during a given time period;others (e. g., 105) use prevalence to mean the“lifetime prevalence, ” or the proportion of pa-tients who have ever had an ulcer. Incidencemay be taken to mean the proportion of a pop-ulation at risk that first develops ulcers in agiven time period (e.g., 18,19) or the percentagethat develops either a new or recurrent activeulcer crater during a given time period (e. g.,147). The methods employed in different studiesto detect disease also vary, ranging from the useof autopsy results, through review of clinicalrecords, to the use of questionnaire surveys.

On the basis of a number of epidemiologicstudies, some experts estimate that the currentincidence of new cases of duodenal ulcer in theUnited States is about 200,000 per year and thatthe incidence of new cases of gastric ulcer isabout one-fourth that (140).

Bonnevie (17, 18, 19) reported several com-prehensive surveys of duodenal and gastriculcer disease occurring between 1963 and 1968in Copenhagen County, Denmark (an area with500,000 inhabitants). Defining incidence as newulcer disease and basing the diagnosis on reviewof hospital records, he estimated the annual in-cidence of duodenal ulcer per 1,000 persons age15 and over to be about 1.8 for men, 0.8 forwomen, and 1.3 overall (18). The annual in-cidence of gastric ulcer alone per 1,000 inhabi-tants age 15 and over he estimated to be approx-imately 0.3 for both men and women (19). Healso found that duodenal and gastric ulcers oc-cur in the same patient much more often thanwould be expected by chance if the two typesoccurred independently (17). Bonnevie (18,19)

cites earlier population surveys conducted inEngland, Scotland, Norway, and Denmark thatfound incidence of duodenal ulcer rangingfrom 0.38 to 2.70 per 1,000 inhabitants age 15and over and incidence of gastric ulcer rangingfrom 0.1 to 1.14 per 1,000 inhabitants age 15and over.

A mail survey of Massachusetts physiciansconducted in 1967 and 1968 found the incidencerates of reported duodenal ulcer of 1,000 per-sons age 25 and over to be approximately 2.9per year for men and 1.5 per year for women(105). In the same study, physicians reportedthe incidence of gastric ulcer per 1,000 personsage 25 and over to be approximately 0.35 peryear.

Different epidemiologic studies have foundvarying patterns of age-specific incidence ofduodenal and gastric ulcer. In general, the in-cidence of duodenal ulcer appears to rise grad-ually with age or to remain essentially constantabove age 35, and the incidence of gastric ulcerappears to rise more dramatically above age 40.Bonnevie (18) found the age-specific incidencerate of duodenal ulcer to increase gradually inboth sexes to a maximum of 3 per 1,000 inhabi-tants between age 75 to 79. Gastric ulcer showeda more dramatic rise in incidence above age 40,peaking at a level of about 1 per 1,000 for menage 60 to 64 and for women above age 70 (19).Among Massachusetts physicians surveyed inthe late 1960’s, the incidence of duodenal ulcerin both sexes appeared to increase up to age 25to 34, and then to remain fairly constant; gastriculcer in male physicians continued rising to apeak at age 65 to 74 (105). Fry’s review of hispatient experience showed duodenal ulcersreaching their peak incidence in both sexes inthe decade 1930-39 and gastric ulcers reachingtheir peak some 20 years later in the decade1950-59 (57).

The aforementioned incidence figures are sub-stantially lower than the rates found in recenthousehold interview surveys conducted byNCHS (35,36). After we review evidence con-cerning the prevalence and changes in the oc-currence of ulcer disease during the past 20years, we will discuss NCHS’s Health Interview

Survey (which is based on household inter-views) in more detail.

According to traditional medical lore, 1 U.S.male in 10 will develop a duodenal ulcer by age55. As pointed out by Mendeloff (104), this easi-ly remembered figure is based on projectionsmade by Ivy in 1946 (84). A number of autopsystudies in Britain and elsewhere (cited by 104)confirmed this figure. It may be argued that in-sofar as the stress of illness can provoke ulcera-tion, autopsy results may be misleading for thepopulation at large. However, the previously

cited survey of Massachusetts physicians (105)also found that approximately 10 percent ofmale physicians age 65 through 74 at some timehad duodenal ulcer. The current level is a matterof conjecture, because the lifelong prevalencerate of ulcers ultimately depends on age-specificincidence rates, and these rates appear to bedeclining,

Ulcer disease appears to have been occurringless frequently or less severely, or both, over thepast 20 years. This conclusion derives from sev-eral lines of clinical and epidemiologic evidence.These include overall declines in rates of mor-tality and hospitalization due to ulcer disease,and, especially, several age-cohort analyses ofthe incidence and mortality of ulcer disease.

Susser (141,142) deduced from age-specificmortality rates between 1900 and 1960 thatthere was a decrease in risk of ulcer disease ineach successive age cohort, producing a rise inthe mean age of patients. This decline was cor-roborated by a cohort analysis conducted byMonson and MacMahon in their survey of Mas-sachusetts physicians (105). Monson and Mac-Mahon found the age-specific risk of developingulcer disease among physicians born between1922 and 1932 to be much lower than the ratefor those born in the preceding 20 years. Astudy of British physicians found a 40-percentdecrease in the incidence of duodenal ulcerdisease between 1947 and 1965 (103).

U.S. mortality from ulcer disease has declinedsteadily since the early 1960’s (see table 2 andfig. 2). The age-adjusted mortality rate droppedby two-thirds in 1977 from its 1962 peak level

Case Study #11 Benefit-and-Cost Analysis of Medical Interventions: The Case of Cimetidine and Peptic Ulcer Disease ● 15

Table 2.–Number of Deaths in the United States With Ulcer Disease as the Primary Cause, 1960-79—

Peptic, siteYear Gastric Duodenal unspecified Gastrojejunal Total1960 . . . . . . . . . . . . . 5,707 5,653 — 322 11,6821963 . . . . . . . . . . . 6,330 5,831 — 244 12,4051966 ......, . . . . . . 5,599 4,722 — 197 10,5181968 . . . . . . . . . . . . . 3,829 4,413 1,218 721 10,1811969 . . . . . . . . . . . . . 3,719 4,381 1,212 798 10,1101970 . . . . . . . . . . . . . 3,502 3,916 1,189 739 9,3461971 . . . . . . . . . . . . . 3,385 3,680 1,055 700 8,8201972 . . . . . . . . . . . . . 3,274 3,510 1,132 756 8,6721973 . . . . . . . . . . . . . 3,289 3,385 1,014 765 8,4531974 ..., . . . . . . . . . 3,050 3,048 971 751 7,8201975 . . . . . . . . . . . . . 2,900 2,920 923 710 7,4531976 . . . . . . . . . . . . . 2,834 2,686 908 698 7,1261977 . . . . . . . . . . . . . 2,669 2,452 779 662 6,5621978 . . . . . . . . . . . . . — — — — 5,550’1979 . . . . . . . . . . . . . — — — — 5,560b

aprellmtnary flgure5, extrapolated from a10-PercentsamPlebprehmlnary~gure5 extrapolated froma IO percent sampleoverthe first 6monthsof 1979

SOURCE National Center for Health Statlsttcs, D!vlslonof Vital Statlstlcs, Hyattsvflle, Md

Figure 2.— Deaths in the United States With UlcerDisease as the Primary Cause, 1966-79

6,000

5,000t

1965 1970 1975 1980Year

Note: 1978 and 1979 figures are preliminary.

SOURCE Based on data from the National Center for Health Stat{ stfcs, DIvI-slon of V!tal Statlstlcs Hyatt svllle Md

(see table 3), Hospitalizations for ulcer diseasehave also declined steadily in both the UnitedStates (see table 4 and fig. 3) and Great Britain(21). The drop in U.S. hospitalizations appearsmainly due to a fall in admissions for duodenalulcer, whereas the drop in Great Britain is duemore to declining admissions for gastric ulcer.Mendeloff (104) reported a sO- percent declinein the number of diagnoses of duodenal ulcerbetween 1960 and 1972 among an apparentlyconstant population in the U.S. armed forces.Data from a large U.S. manufacturing companyshowed a 56-percent drop in episodes of disabil-ity due to duodenal ulcer and a 68-percent dropin episodes of disability due to gastric ulcer be-tween 1960 and 1970 among male employees(3).

Some of these trends might be explained bythe advent of a dramatic and continuing im-provement in the prevention and care for ulcerdisease during the past 20 years, but no likelycandidate representing this can be found (21).The data are consistent with a shift in the spec-trum of ulcer disease toward less severe forms, apossibility posited by Mendeloff (104). Such ashift may accompany what appears to be thesimplest explanation: Ulcers are occurring lessfrequently than they did previously. The rea-

16Ž Background Paper #2 Case Studies of Medical Technologies

Table 3.—Mortality Rates in the United Statesfor Deaths Due to Ulcer Disease 1953-78

——Age-ad lusted rate per

Year 100,000 populatationb

— — — — — —1953 . . . . . . . . . . . . . .1958 . . . . . . . . . . . . .1960 . . . . . . . . . . . . . .1961 . . . . . . . . . . . . . .1962 . . . . . . . . . . . . . .1963 . . . . . . . . . . . . . .1964 . . . . . . . . . . . . . .1965 . . . . . . . . . . . . . .1966 . . . . . . . . . . . . . .1967 . . . . . . . . . . . . . .1968 . . . . . . . . . . . . . .1969 . . . . . . . . . . . . . .1970 . . . . . . . . . . . . . .1971 . . . . . . . . . . . . . .1972 . . . . . . . . . . . . . .1973 . . . . . . . . . . . . . .1974 . . . . . . . . . . . . . .1975 . . . . . . . . . . . . . .1976 . . . . . . . . . . . . . .1977 . . . . . . . . . . . . . .1978 . . . . . . . . . . . . . .

5.15.35.25.25.45.24.64.34.23.93.73.63.23.02.92.72.42.22.11.8—

Crude rate

aRates~hown lncludegastrlc, duodenal, andpeptlc ulcer (siteunspeclfled)bAdJusted to 1940 population, the standard population used by the National

Center for Health Statlstlcs

SOURCE National Center for Health Statlstlcs, Dlvlslon of Vital Statlstlcs,Hyattsvdle, Md

sons for the apparently declining incidence andseverity of ulcers are matters for speculation,

The data on ulcer disease from the Health In-terview Survey of NCHS warrant separate con-sideration for three reasons. First, the Health ln-terview Survey data are gathered in a uniquemanner; they are based on self-reported condi-tions in household interviews. Second, es-timates of disease incidence and consequencesobtained from Health Interview Survey data aresubstantially greater than those obtained fromother sources, including other NCHS sourcesand epidemiologic studies such as those de-scribed above; also, estimates from the HealthInterview Survey show little change between theyears 1968 and 1975. Finally, the survey datadeserve special attention, because they are usedto estimate some of the costs and benefits oftreatment for ulcer disease that we review laterin this case study.

Household surveys of chronic digestivediseases in the United States were conducted in

Table 4.—Number of U.S. Hospital Discharges With Ulcer Disease Diagnoses, 1966-78

Year GastricPeptic, site

Duodenal unspecified Subtotal a Gastrojejunal— — - . . — . —— — — — — — - - - - - - - - - - T o t a lb - . .Ulcer as first-listed diagnosis1966 . . . . . . 166,100 345,200 — 511,3001970 . . . . . . 89,200 273,500 68,300 431,0001971 . . . . . . 94,100 251,400 68,600 414,1001972 . . . . . . 99,300 241,400 81,200 421,9001973 . . . . . . 102,900 227,100 68,100 398,1001974 . . . . . . 101,500 239,800 75,300 416,6001975 . . . . . . 101,500 224,100 77,000 402,6001976 . . . . . . 103,400 194,000 81,100 378,5001978 . . . . . . 105,100 166,300 81,900 353,300

Ulcer as a listed diagnosis1966 . . . . . . 223,800 464,300 — 688,1001970 . . . . . . 127,200 384,200 108,900 620,3001971 . . . . . . 137,200 358,600 110,800 606,6001972 . . . . . . 147,300 362,300 131,800 631,4001973 . . . . . . 149,800 339,900 123,700 613,4001974 . . . . . . 156,400 360,200 136,100 652,7001975 . . . . . . 158,400 336,200 150,300 644,9001976 . . . . . . 160,700 302,300 158,300 621,3001977 . . . . . . 173,000 285,900 159,300 618,2001978 . . . . . . 165,400 279,400 184,600 629,400

—afncludes gastric, duodenal, and peptic ulcer (Site unspecified)blncludes gastric, duodenal, pept(c (site unspecified), and 9aStr0Je]LInal ulcer

SOURCE. National Center for Health Slatistlcs, National Hospital Discharge Survey, Hyattsvllle. Md

14,7007,4006,6007,4007,2008,7009,1006,9007,200

17,5009,1009,1009,5009,600

11,20012,40010,7008,700

10,800

526,000438,400420,700429,300405,300425,300411,700385,400360,500

705,600629,400615,700640,900623,000663,900657,300632,000626,900640,200

Case Study #11: Benefit-and-Cost Analysis of Medical Interventions, The Case of Cimetidine and Peptic Ulcer Disease ● 17

Figure 3.—U.S. Hospital Discharges With UlcerDisease Diagnoses, All Sites, 1966.78

750

700

650

600

550

500

450

400

350

1965 1970 1975

Year

SOURCE Based on data from the National Center for Health Statistics, Na-Iional Hospital Discharge Survey Hyattsville Md

by NCHS 1968 and 1975 (35,36). The surveysconsisted of questions asked at a sample ofhouseholds designed to represent the civilian,noninstitutionalized U.S. population. SelectedHealth Interview Survey results pertaining toulcer disease are summarized in table 5. Theprojected incidence of new ulcers based on theHealth Interview Survey, approximately600,000 cases per year, is more than double thatbased on other epidemiologic evidence de-scribed earlier. People interviewed at homereported approximately 7 million physicianvisits for ulcers in 1975, nearly triple the 2.5million physician visits for ulcer disease that

Table 5.—Ulcer Disease in the United StatesAccording to the Health Interview Survey, NCHSa

Measure 1968 1975 1978

Number of conditions (in 000’s) 3,360 3,955 3,778Prevalence per 1,000 personsb. 17.2 18.9 17.7Incidence per 1,000 persons c. 3.0 2.9 –Ever hospitalized for ulcer

disease. . . . . . . . . . . . . . . . . . 40.60/0 38.30/0 —Ever had surgery for ulcer

disease. . . . . . . . . . . . . . . . . . 6.90/o 8.1% —

Currently under M.D. care. . . . . 61.1 0/0 65.40/o —M.D. visits in past 12 months:

0. . . . . . . . . . . . . . . . . . . . . . . 32.4% 36.1% —

1 1 7 . 1 % 17.8% —2 to 4 : : : : : : : : : : : : : : : : : : : : 23.70/o 26.30/0 —5 or more. . . . . . . . . . . . . . . . . 18.20/o 15.80/o —Unknown . . . . . . . . . . . . . . . . 4.60/0 4.00/0 —

Number of bed-disability daysd:0. . . . . . . . . . . . . . . . . . . . . . . . 74.5 74.5 —1 to 3, ., . . . . . . . . . . . ... , . . — 7.4 –4 to 7 ...., . . . . . . . . . . . . . . . — 5.2 –8 to 14 . . . . . . . . . . . . . . . . . . . 4.9 4.4 —15 to 30 . . . . . . . . . . . . . . . . . . 3.6 3.6 –31 or more. . . . . . . . . . . . . . . . 3.4 2.2 –

alnClude~ ~astrlc duodenal, pep~lc (site unspeclfled), and 9astrolelunal utcerbcondltlon repor~ed as hav(ng been present at some time durln9 the Year Prior

to Interviewcon set reported as wlthln year prior to lntervlewdA bed.dlsablllty day IS a day In which a person StayS In bed for all or most of

the day because of ulcer

SOURCE National Center for Health Stat! stlcs, Dwislon of Health InterwewStatlstlcs, Hyattsvllle, Md

year reported in the NCHS National Ambula-tory Medical Care Survey (34). In contrast toother epidemiologic evidence for the decliningincidence of ulcers, the Health Interview Surveyresults show little change, with even a slightlyincreased prevalence between 1968 and 1975.

These discrepancies may derive from severalsources. Most likely, more people report havingulcers in the Health Interview Survey than ac-tually have them. Some individuals withoutmedical training may think of any stomachtrouble as “ulcers” and use the specific medicalterm more broadly than is clinically correct. In1975, more than 36 percent of the people whoreported having ulcers in the previous year didnot see a doctor for that reason. (The propor-tion with newly reported ulcers who were self-diagnosed is not given. ) Many of those who didsee a doctor may have been treated on the basisof symptoms without a definite diagnosis. TheHealth Interview Survey may be an accuratesummary of what the noninstitutionalized pub-lic reports, but that is not the same as an ac-

18 ● Background Paper #2: Case Studies of Medical Technologies

curate epidemiologic assessment of a diseaseproblem.

Cost of Illness

Studies of the cost of peptic ulcer disease areamong the earliest efforts by economists toassess the costs of individual diseases (14).Beginning with the very first studies, a basicdistinction was drawn between direct costs(health system expenditures to prevent, diag-nose, and treat the disease) and indirect costs(economic losses due to morbidity and mortali-ty). Most economic studies measure the indirectcosts of illness in terms of loss of productivity

due to disability from disease and loss of futureproductivit y due to premature death.

The same basic categories of direct and in-direct costs continue to be used in contemporaryeconomic analyses of the cost of illness (31,114).Most researchers take an aggregate approach tomeasuring direct costs of disease, using datafrom third-party payers, NCHS, and other hos-pital and physician surveys, and estimatingtotal expenditures for a given disease populationin a given time period, usually 1 year; we willreturn to these methods shortly. First, however,we will mention patient-specific alternatives tomeasuring direct costs of illness.

One alternative is to trace over time expend-itures for a cohort of patients with a particulardisease. As far as we know, no such studies ofulcer disease have been published, but at leastone study now under way at the University ofWisconsin may produce useful information ofthis sort (58). We comment on this stud y b yWeisbrod and Geweke in our discussion ofcimetidine. Such cohort studies have the ad-vantage of being patient-specific and may showrelationships between interventions and expend-itures at one point in time and subsequent clin-ical courses and health expenditures. Cohortcost studies thus could complement other re-search, possibly as a part of longitudinal studiesof the clinical course of disease.

A second patient-specific approach is to studythe cost of treating episodes of illness. Duodenalulcer disease was one of eight medical condi-tions studied in this way by Scitovsky and Mc-

Call (128), These investigators defined anepisode of duodenal ulcer illness as a 6-monthperiod beginning with the date of diagnosis ofduodenal ulcer. They assessed the cost of treat-ing episodes of duodenal ulcer disease fornonhospitalized patients treated at the Palo AltoMedical Clinic in 1964 (35 patients) and in 1971(27 patients). In constant dollar terms, th e

overall cost of treating ambulatory patientswith duodenal ulcer declined slightly (but notsignificantly) between 1964 and 1971. Theaverage number of physician visits per patientduring the defined 6-month episode of illness fellfrom 4.7 in 1964 to 3.8 in 1971. The averagenumber of X-rays also declined slightly. Thesedecreases were nearly offset by increasedexpenditures for drugs.

Patient-specific studies are very useful formany purposes, but they are not intended toprovide a cross-sectional view of all costs for allpatients with ulcer disease in a given timeperiod. Providin g such a view is the aim ofstudies that take an aggregate approach to es-timating direct costs of disease.

In two recent studies of the cost of ulcerdisease discussed below, the indirect costs ofulcer disease were measured by using the“human capital” approach of estimating lossesin productivity attributable to the disease. Anumber of philosophical objections have beenraised to the “human capital/lost productivity”approach to valuing lives, e.g., productivitymeasures omit consideration of pain and suffer-ing. Alternative methods for valuing life, suchas a “willingness-to-pay” approach, have beenused (1), but not as often as the human capitalapproach. Over the past 20 years, the sophis-tication of lost productivit y estimates has in-creased considerably, and now may include thediscountin g of future earnings, the adjustmentof future earnings for productivity gains, ad-justments for labor force participating rates,and calculation of productivity loss for peopleperforming unpaid housework (31), In additionto the sophistication of analysis, a second majordifference between recent studies of the cost ofulcer disease and the earliest studies 20 yearsago is the greater amount of information nowavailable about the prevalence, distribution,

and health consequences of the disease. Asdiscussed in the previous section of this casestudy, however, uncertainty about the evolvingepidemiology of ulcer disease is a major sourceof discrepancies in contemporary estimates ofthe cost of the disease.

One of the two recent analyses of the cost ofulcer disease that we will now discuss wasundertaken as part of the NCDD assessment ofthe socioeconomic impact of digestive diseases(4). The other analysis was prepared at SRIunder contract with Smith Kline & French Lab-oratories by Von Haunalter and Chandler(146). Both the NCDD and SRI studies es-timated the cost of ulcer disease in 1975, and wefocus primarily on those figures. In addition,the SRI study projected estimates for 1977; theseserved as the basis for a major cost-effectivenessstudy of cimetidine (the study by RobinsonAssociates (121)) that is reviewed in anotherpart of this case study.

The costs of peptic ulcer disease in 1975, asestimated by NCDD and SRI, are summarizedin table 6. The total cost (direct and indirect)estimated by NCDD is approximately $1.3billion; the estimate by SRI is approximately$2.6 billion.6 Table 6 also shows a “midpointestimate” of approximately $2 billion. Webelieve $2 billion to be a defensible overall costestimate, for reasons we shall explain. Pepticulcers accounted for less than 1 percent of totalcosts of all illness in 1975 (114), and, accordingto NCDD figures, health system expendituresfor ulcer were approximately 9 percent of healthexpenditures for all digestive diseases in 1975.Of the total $2 billion costs for ulcer disease,just under half are attributed to health systemcosts (direct costs); the rest are attributed to lostproductivity due to premature mortality and tomorbidity (indirect costs).

A comparison of the NCDD and SRI esti-mates by cost category reveals that the discrep-ancy between them is largely due to differencesin the indirect costs attributed to morbidity (see

‘Smith Kllne & French markets c]metlcilne.*Peptic ulcer cilsease was t~n ly IIne 01 n umert~u~ digestive

diseases t(~r which NCDt) dcvelt~ped cost estimates. The auth(lrs(~t the NCDD rep(~rt ( 4 ) +tate c]ea r] v that they c(~n>i (]er the] rest I ma tes to be con scrva t I ve.

Table 6.—Costs of Ulcer Disease in 1975 asEstimated by NCDD and SRI

(millions of dollars)

ApproximateNCDD SRI midpoint

estimates estimates estimates

Direct costsHospitalization . . . . . $501 $ 803 $652Physician visits . . . . . 123 240 182Drugs . . . . . . . . . . . . .

1

100Nursing home care . . 102a 11

}108a

Other professional 3

Subtotal . . . . . . . . . $726 $1,157 $942Indirect costs b

Mortality. . . . . . . . . . $369 $ 357 $ 369cMorbidity . . . . . . . . . . 179d 1,116 648

Subtotal . . . . . . . . . 548 1,473 1,017

Total ... ... ... .$1,274 $2,630 $1,959

aT’hl~ ~um represents tfle total for drugs, nursing homes, and other PrOfeSS~Onalcosts Figures were not broken down further

bFuture earnings discounted at 25 PercentcThe fllgher NcDD figure IS adopted for reasons explalned In the textdThls figure ,s imputed from Information supplled In the NCDD rePort

SOURCES NCL)D estimates: T P Almy, et al , “Report of the Workgroup on theSocloeconomlc Impact of Dlgestlve Diseases of the Subcommitteeon Epldemlology and Impact. ” In Report to fhe Congress of theUrr/fed .Sf.sfes of fhe Nat/ona/ Comm/ss/on on Llges(we Diseases,1979 (4)SRI estimates: G Von Haunalter and V V Chandler, Cost of U/cerD/sease In Me Un/fed Sfafes, 1977 (146)

table 6). In addition, SRI’s estimates of directcosts for hospital and physician services arenotably higher than NCDD’s (see table 6).Closer examination of the sources of thesediscrepanices in direct cost estimates revealsvariation in the two studies’ analytic methods,as well as shortcomings in data needed for suchcost estimates.

Medical care costs attributable to a particular, disease may be estimated in two ways: 1) by a

“top-down” approach that begins with total ex-penses for all disease and imputes to a particulardisease the proportion of costs equal to the pro-portion of total units of service used by patientswho have the disease; or 2) by a “bottom-up”approach that prices and sums the units of serv-ice consumed by patients who have a particulardisease. Each approach has its strong points,and ideally, the two would corroborate eachother. In general, the top-down approach issimpler; by definition, the sum of all top-downestimates for each disease equals the totalexpenditures for all disease. Theoretically, thesame would be true for bottom-up calculations,but such calculations are typically undertaken

for a single disease only, and potential in-consistencies between known total expendituresfor all disease and the sum of disease-by-diseaseexpenditures bottom-up calculations remainuntested in typical bottom-up calculations.

NCDD and SRI both used a bottom-up ap-proach to estimate hospital costs due to ulcerdisease, but differed in the detailed assumptionsthey employed. NCDD began with the numberof hospital days for each ulcer diagnosis ob-tained by the Hospital Discharge Survey ofNCHS, and multiplied that number by theaverage charge per hospital day, The averagewas obtained from Blue Cross/Blue Shieldfigures for Federal workers and from medicaredata for patients over 65 years of age. SRI alsobegan with NCHS figures on numbers of dis-charges, but it used a more complicated calcula-tion that involved an estimated proportion ofsurgical and nonsurgical cases from the Com-mission on Professional and Hospital Activities(CPHA), an allocation to hospitals of differentsizes based in part on American Hospital Asso-ciation data, and estimated daily costs based oninformation from disparate sources combined inan unspecified manner. The end result of SRI’Scalculation was an estimate of hospital costs($803 million) that is approximately 60 percentlarger than NCDD’s estimate ($501 million).Further exploration of the discrepancy betweenthe two figures would require more detailsabout the calculation:; than was provided ineither report. Interestingly, and usefully, SRIalso applied a top-down cross-check using esti-mated hospital expenditures for 1975 and theproportion of ulcer hospital days to total hos-pital days, and came up with an estimated costof $738 million, reasonably close to our $652mill ion midpoint estimate for this cost compo-nent.

To estimate the cost of physician services,NCDD used a top-down approach, multiplyingthe cost of all physician services for fiscal year1975 by the proportion of total visits attribut-able to ulcer. SRI used a bottom-up approach,multiplying units of service (computed separate-ly for initial and followup visits) by unit costs,estimated on the basis of multiple sources. SRI’sestimate for physician visits for ulcer disease

($240 million) is approximately double that ob-tained by NCDD ($123 million) and, if correct,would imply that a physician visit for ulcerdisease is twice as expensive as a typical physi-cian visit. Although this seems unlikely, it is im-possible to judge the difference in the cost ofphysician visits without a more comprehensiveanalysis. We settled on a $182 million midpointestimate of the cost of physicians’ services as areasonable compromise.

Estimates for remaining direct costs are com-parable in the NCDD and SRI studies. We haveimputed the NCDD figure of $102 million froma more global estimate for selected digestivediseases that included ulcer disease and wasadopted by NCDD (113). Summing the abovecomponents for each report, we find that theestimated direct costs presented in the tworeports differ by more than $400 million:NCDD, $726 million; SRI, $1,157 million. Ourfinal midpoint estimate is $942 million.

Indirect cost estimates for ulcer mortality lossare straightforward. NCDD and SRI used iden-tical methods to estimate lost future earningsfrom death due to ulcer. The small difference inthe two studies’ figures for mortality loss(NCDD, $369 million; SRI, $357 million) isp r e s u m a b l y due to the fact that SRI usedsmaller, preliminary mortality figures (6,840deaths) rather than the final NCHS figures(7,245 deaths) that NCDD used. We haveadopted NCDD’s $369 million estimate.

The very large difference in the NCDD andSRI studies’ estimated ulcer morbidity costs

(NCDD, $179 million; SRI, $1,116 million)stems from several sources. Most important,SRI attributed to ulcer disease morbidity asestimated in the Health Interview Survey cond-ucted by NCHS in 1975. As discussed earlier,the Health Interview Survey estimates are basedon the responses of people interviewed at homewho say they have had an ulcer at some timeduring the past year. These estimates are in-consistent with other evidence for the decliningprevalence of ulcer disease, and they almostsurely overestimate morbidity due to the dis-ease. Furthermore, SRI assumed that the eco-nomic effects of work loss are distributed by age

in the same way the disease is distributed. Sinceolder patients tend to lose more days of workand earn less per day, the assumption of uni-form effects inflates the actual productivity loss.This flaw is acknowledged in SRI’s report, butno correction or sensitivity analysis is offered.Ulcer patients who continue to work might havelower productivity, and this effect, also omittedfrom SRI’s calculations, would increase the ac-tual loss of productivity due to ulcer disease andtend to offset the effect of the assumption aboutage distribution.

NCDD considered and expressly rejectedusing data from the Health Interview Survey,because “there were also serious questionsraised by experts in digestive diseases about thevalidity of the self-reported diagnosis-specificmorbidity information” (4). Instead, NCDD ac-cepted a more global estimate of morbidity lossdue to 15 different digestive diseases, includingliver disease, gallbladder disease, and hernia(113). The NCDD figure for morbidity loss dueto ulcer disease shown in table 6 is approximate-ly 6 percent of that total, a percentage equal tothe ratio of the mortality cost for ulcer com-pared to that for all 15 diseases.7 The NCDDreport also refers to data collected for an earlierreview of the medical and socioeconomic im-portance of digestive disease, published byAlmy and his coworkers (3). That earlierpublication included data on absenteeism due todigestive disease at a large northeastern U.S.manufacturing company during the 13-yearperiod from 1959 to 1972. In persons whomissed 3 or more days of work during that peri-od due to 1 of the 15 digestive diseases coveredby the NCDD morbidity estimate, more than 20percent of days lost were attributed to ulcer. Wedo not propose translating such figures, ob-tained over a 13-year period from one largefirm, to a national estimate of days of work lostin a later year. However, if ulcer disease doesaccount for 20 percent of the total morbiditycosts assigned to the 15 digestive diseases in theNCDD report, the NCDD morbidity figure

would be very close to our midpoint estimate of$648 million.

The magnitude of indirect cost estimates isvery sensitive to the rate at which future costsare discounted. Both NCDD and SRI discountedfuture earnings at 2.5 percent, although NCDDalso presents some alternative calculations at a10-percent discount rate. Economists agreemore on the appropriateness of discountingthan on the appropriate rate to employ, but 2.5percent is at the very low, end of the spectrum.The smaller the discount rate, the higher thepresent value of future earnings and the higherthe apparent indirect cost of illness. For exam-ple, the “present” value of lifetime earnings for a32-year-old man in 1975 was $148,195 at a 2.5-percent discount rate and $176,882 at a IO-per-cent discount rate (4). This is not a differentialpoint between the NCDD and SRI analyses,since they both used the same low discount rate,but the reader should be aware of the large dif-ference a change in the discount rate can makeand be wary of unadjusted comparisons be-tween these and other cost-of-illness studies thatmay use different discount rates. In addition,“present” values are usually expressed in termsof dollars in the base year. Estimates discountedto different base years will differ in part becauseof inflation and are directly comparable in re-source cost terms only if adjusted into constantdollars.

In the SRI analysis, Von Haunalter andChandler extrapolated their estimated costs to1977 (expressed in 1977 dollars) by assumingvariably inflated rates for different unit costs ofmedical care and a 2-percent annual increase inthe number of persons with ulcer disease (seetable 7). The presumed 2-percent annual in-crease, based on responses to the Health Inter-view Survey in 1968 and 1975, is contrary to allother indicators of the changing epidemiologyof ulcer disease; it is also contradicted by sub-sequent preliminary data obtained in the 1978Health Interview Survey (see table 5, p. 17). Thepresumed growing population with ulcers is alsotreated by Von Haunalter and Chandler ashaving the identical age distribution and spec-trum of disease severity as assumed for thepopulation in 1975. Their assumptions about

22 ● Background Paper #2 Case Studies of Medical Technologies

Table 7.—Costs of Ulcer Disease in 1975 and 1977 asEstimated by SRI (millions of dollars)a

— —1975 1977

Direct costsHospital care. . . . . . . $ 803 — $1,072 —Physicians . . . . . . . . . 240 — 283 –Drugs . . . . . . . . . . . . . “loo — 113 —Nursing home care . . 11 — 15 —Other professional . . 3 — 3 –

Subtotal . . . . . . . . . $1,157 4 4 0 / o $ 1 , 4 8 6 460/0

Indirect costsb

Mortality. . . . . . . . . . . 357 — 408 —Morbidity . . . . . . . . . . 1,116 — 1,330 —

Subtotal . . . . . . . . . 1,473 56% 1,738 54%

Total . . . . . . . . . . $2,630 1000/0 $3,224 1000/0

aFl~Ur~~ fOr 1975 and 1977 are expressed In terms of dollars In the re’

spectlve base years, not In constant dollars

SOURCE Adapted from G Von Haunalter and V V Chandler, Cos/ of U/cerD}sease In rhe Un//ed Stales, 1977(1 46)

these characteristics of the presumed growingulcer population are questionable for reasonsdiscussed in the section of this case study aboveon epidemiologic patterns of peptic ulcerdisease. Von Haunalter and Chandler’s projec-tion of the population with ulcers to 1977 thuscompounds the problem of overestimation ofthe costs of ulcer disease, particularly indirectcosts. Some specific estimates that feed intodirect costs are also Inappropriately projectedupward. For example, the percentage of hospi-talized patients undergoing surgery for ulcerdisease is presumed in the analysis to increasebetween 1975 and 1977. According to data fromboth NCHS and CPHA (42), however, the per-centage actually decreased (see table 8).

Summary

We may summarize the salient clinical andepidemiologic features of peptic ulcer disease asfollows,

Ulcers probably have multiple causes, butgastric acid and pepsin appear to be necessaryingredients. Epigastric pain is often a prominentsymptom of peptic ulcers, but the clinical pres-entation is variable. Furthermore, typical ulcersymptoms may be caused by conditions otherthan ulcers. A definite diagnosis requires directvisualization by endoscopy or radiographicimaging of the ulcer. Specific treatment of ulcerdisease is directed at reducing the presence or ef-fects of gastric acid.

Ulcer disease is a chronic condition withspontaneous remissions and recurrences. Ratesof complication and mortality from ulcers arerelatively low. Excessive mortality appears to bepresent only in the first year or so followingdiagnosis. Little reliable information existsabout the natural history of ulcer disease in thegeneral population.

Peptic ulcer is a common condition that af-fects millions of Americans at some time duringtheir lives. The best available epidemiologicevidence suggests that about 250,000 Americansdevelop new peptic ulcers each year. New duo-denal ulcers are more than four times as com-mon as new gastic ulcers. Some studies havefound that the incidence of duodenal ulcer risesgradually with age, others have found that it re-mains fairly constant after age 35. After age 40,

Table 8.—NCHS and CPHA Data on Number of Selected Surgical Procedures (partial Gastrectomy, Vagotomy)in the United States, 1966-78

.— — —.———Partial gastrectomy –

——Vagotomy Total—

Year NCHS CPHA NCHS CPHA NCHS C P H A —

—1966 . . . . . . . . . . 74,500

—.—.—— 61,200 — 135,500

1970 . . . . . . . . . . 55,800—

59,000 62,800 30,000 118,600 89,0001971 . . . . . . . . . — 57,000 — 28,000 85,0001972 . . . . . . . . . . 63,300

—52,000 59,300 24,000 122,600 76,000

1973 . . . . . . . . . . — 52,000 — 27,000 79,0001975 . . . . . . . . . .

—53,300 45,000 52,800 23,000 106,100 68,000

1976 . . . . . . . . . . 54,200 45,000 48,300 16,000 102,500 71,0001977 . . . . . . . . . 51,100 37,000 45,500 19,000 96,600 56,0001978 . . . . . . . . . . 39,700 29,200 29,200 17,000 68,900 46,000

—. ——————————sOURCES fVCHS data: National Centfr for Health Statistics, National Hospital D!scharge Survey, Hyaltsvllle, Md

— —

CPHA data: Commlsslon 01 Professional and Hospital Acttvltles data compiled by J D Elashoff and M I Grossman, 1960 (42)

the incidence of gastric ulcer appears to risemore dramatically than the incidence of duo-denal ulcer. Duodenal and gastric ulcers are epi-demiologically distinct. Several lines of clinicaland epidemiologic evidence suggest that overthe past 20 years the occurrence of new ulcershas declined, or ulcer disease has becomegenerally less severe than it was at one time, orboth.

The basis for some estimates of the costs ofulcer disease and benefits of treatment is theHealth Interview Survey of NCHS. Results ofthe survey, based on self-reported conditions ina household survey, however, appear to over-estimate the occurrence and consequences ofulcer disease.

We estimate that the costs of ulcer disease in1975 were approximately $2 billion (see table 6,p. 19). Just under half of this total was due tohealth care expenditures (direct costs), and theremainder was due to productivity losses frommorbidity and mortality (indirect costs). ourestimate is based on a review of two independ-ent analyses of the costs of ulcer disease, oneprepared by NCDD (4) and the other by SRI(146). The NCDD and SRI studies estimates Ofthe total costs of ulcer disease in 1975, $1.3billion and $2.6 billion, respectively, differ byapproximately $1.3 billion. The approximately$400 million difference between the two studies’direct cost estimates (NCDD, $726 million; SRI,$1,157 million) arises in part from differencesin the studies’ methodologies (top-down v.bottom-up calculations) and in part from dif-ferences in their more detailed assumptions andprocedures. The approximately $900 milliondifference in the two studies’ indirect cost esti-mates (NCDD, $548 million; SRI, $1,473 mil-lion) reflects differences in the two studies’ pro-jected morbidity losses. The higher estimate is

CIMETI DINEPhysiologic Rationale andDevelopment

The major physiologic stimulant to acidsecretions in humans is the ingestion of food,

based on data from the Health Interview Sur-vey, which is an inflated indicator of disease-specific morbidity. In both the NCDD and SRIstudies, estimates of indirect costs are based onthe relatively low discount rate of 2.5 percent,although the NCDD report also supplied esti-mates based on a lo-percent discount rate. SRIalso projected an estimate of peptic ulcer costsin 1977. Because of unwarranted assumptions ofgrowth in the morbidity of ulcer disease and theuse of more expensive resources, the problem ofoverestimated costs is compounded for 1977.

We have briefly noted alternatives to cross-sectional expenditure assessments of the directcosts of illness, including tracking patient co-horts over time and measuring costs of treatingepisodes of illness. Results of a study of the lat-ter type in one setting found that an episode ofduodenal ulcer disease cost approximately thesame to treat in 1971 as in 1964, in constantdollar terms.

The human capita] approach is the principalmethod used to assess indirect costs of illness. Ingeneral, lost productivity is measured as thepresent value of discounted stream of futureearnings. Use of a smaller discount rate in-creases the present value of future earnings,thereby increasing apparent costs of illness dueto morbidity and premature death.

The next four parts of this case study dealwith the evaluation of cimetidine in peptic ulcerdisease. We begin with background on the de-velopment and dissemination of cimetidine.Then we explore current understanding of thecosts and benefits of this drug in the treatmentof peptic ulcers, using the general benefit-and-cost model as a framework. Finally, we critiquea major report on the costs and benefits of cime-tidine and offer a few suggestions for furtherresearch.

but three chemical substances in the body arealso known to stimulate acid secretion in thestomach: acetylcholine, gastrin, and histamine.The first two are clearly involved in the physio-logic release of acid; histamine appears to

24 Ž Background Paper #2: Case Studies of Medical Technologies

potentate the action of other acid stimulants(136). Even before the physiologic role of hista-mine was well understood, some researchersbelieved that a drug that would block histaminestimulation of gastric acid would be of greatvalue in the treatment of ulcer disease. Conven-tional antihistamines have no effect on his-tamine receptors in the stomach.

By the mid-1960’s, J. W. Black and his col-leagues at the British laboratory of Smith Kline& French Laboratories (the pharmaceutical divi-sion of SmithKline Corp. ) had set up a researchprogram to develop new kinds of histamineblockers. Their strategy in this effort was simi-lar to that which had led to their successfuldevelopment of beta-adrenergic blockers, name-ly, systematic chemical manipulation to createnullifiers of the parent drug’s effects (136). Theyreported their first successful effort in 1972 (12).

This work demonstrated the existence of anew class of histamine receptors, designated H2-receptors, which were distinct from the classicH1-receptors. The new histamine antagonist,called burimamide, was very effective in sup-pressing stomach acid production, not only inresponse to histamine, but from other stimulias well, Burimamide had to be injected to be ef-fective, and it was supplanted by an orally ac-tive Hz-receptor antagonist, metiamide. Thisdrug was used in human trials in 1974, but wasabandoned when it was found to cause granu-locytopenia and agranulocytosis (growth offewer white blood cells than normal) and atleast one fatality (24,45). The chemical metia-mide possessed a thiourea side chain that wasbelieved to be the offending component, and itwas replaced by a cyanoguanidine chain. Theresult of this chemical manipulation was thethird H1-receptor antagonist, cimetidine (SKFTagamet®). 8 Along the way to the discovery of.—— ————-.

“The structural similarities of histamine and H, antagonists areevident from their chemical structures.— — — —

StructureCompoundname Imidazole ring Side Chain (R)

Histamine A{HN + N -C H2 CH2NH2

Burimamide Hr{HN ~ N -CH2CH2CH2CH2CH2CH2CH2NHCNHCH,CH, 1{x

sMeteamide HN + N -CH2SCH2CH2NHCNCH ,

Lti, R !+Cimetidine wHN @ N -CH 2SCH 2CH 2NHCNHCH ,

N-C = N

cimetidine, Black and his colleagues developed,tested, and rejected more than 700 differentcompounds (135).

Cimetidine powerfully inhibits all phases ofgastric acid production. The drug not only in-terferes with histamine-stimulated acid output,but also inhibits the effects of gastric andacetylcholine. Preliminary studies found that a300 mg oral dose of cimetidine reduced noc-turnal and basal acid secretion by 90 to 95 per-cent (74,97). Cimetidine also lowers the acidresponse to food by 70 percent, a reductiontwice that ach(46,120).

Diffusion

After prelimleased for use

eved by anticholinergic agents

nary trials, cimetidine was re-in Great Britain in November

1976. FDA recognized the clinical promise ofcimetidine and rated it 1A, FDA’s highest clas-sification, meaning the drug is a new molecularentity believed to represent a major therapeuticadvance over other drugs (47). FDA approvedcimetidine on August 16, 1977 (50), for up to 8weeks use in patients with duodenal ulcer dis-ease and in patients with hypersecretory condi-tions such as Zollinger-Ellison syndrome, sys-tematic mastocytosis, and multiple endocrineadenomas. Although cimetidine is not yet offi-cially approved in the United States for longerterm use, some physicians use it for more than 8weeks in patients with duodenal ulcer (7).

The use of cimetidine spread rapidly in U.S.clinical practice. By the beginning of 1978,private practitioners prescribed cimetidine inapproximately 40 percent of ambulatory visitsmade by patients with duodenal ulcers, ByMarch 1978, the proportion of such visits result-ing in a prescription for cimetidine reached ap-proximately 60 percent, where it has remainedsince that time. A substantial fraction of totalambulatory use of the drug, perhaps as much ashalf, is for patients with acid reflux, gastritis,gastric ulcer, or problems other than duodenalulcer. 9

“Estimates based on figures published by the National DiseaseTherapeutic Index, which obtains reports of clinical practices froma sample of private practitioners.

Case Study #11 Benefit-arid-Cost Analysis of Medical Interventions. The Case of Cimetidine and Peptic Ulcer Disease ● 25

The amount of cimetidine used by hospital-ized patients is more difficult to estimate direct-ly, but such use is widespread. In addition to be-ing used in patients with ulcer disease, the drughas been used to prevent and treat stress-relatedgastritis and bleeding (39,100). Many physiciansare probably using cimetidine in seriously ill pa-tients who are susceptible to gastrointestinalhemorrhage (48). A recent randomized trial ofpatients hospitalized in an intensive care unitfound that regular antacid therapy is more effec-tive than cimetidine in preventing gastrointes-tinal bleeding (117).

From a commercial viewpoint, cimetidine is aspectacularly successful product. In 1977, cime-tidine was marketed in 65 countries; in 1978, itwas sold in 90 countries. In 1977, SmithKlinereported sales of gastrointestinal drugs of $90.5million; in 1978, sales of these drugs were $315million. In its 1978 annual report, SmithKlinestated that Tagamet® was its most importantsingle product (135). Worldwide sales to hospi-tals and pharmacies in 1979 probably exceeded$400 million. This translates into a rough esti-mate of 10 million patients per year consuming

cimetidine worldwide .10 A conservatively esti-mated 1.5 million to 2 million ambulatory U.S.patients with ulcer disease and ulcer-like symp-toms were treated with cimetidine in 1978.11

Cimetidine has thus become one of the mostwidely used pharmaceuticals in the world in aremarkably short time. Part of the reason forthis success rests in the widespread prevalenceof ulcer disease and ulcer-like symptoms. Smith-KIine pioneered and persevered in developingand marketing a new class of pharmaceuticalagents. Furthermore, as discussed in the nextpart of this case study, a substantial number ofcontrolled trials attest to the effectiveness andrelative safety of this drug in the treatment ofulcer disease.

10This estimate follows from high and low estimates, as shown:High estimate Low estimate

Retail value of sales $500 million $300 million– Retail price per tablet $0.25 $030= Number of tablets (A) 2 billion 1 billionWeekly number of tablets 28 28

x Mean weeks of treatment 5 6= Number of tablets patient (B) 140 168Estimated number of patients (A/B) 14,300,000 6,000,000

11This estimate is based on a conservative projection of the num-bers of Americans living at home who report ulcer disease in 1968and 1975 and the proportion who visit their physicians for thisproblem at least once in the year (36).

THE BENEFIT-AND-COST MODEL APPLIED TO CIMETIDINE

Elements in the Analysis

The major components of the benefit-and-cost model presented earlier in this case studyare shown in table 9. Listed under each compo-nent are a number of measures pertinent tocimetidine. Ideally, benefit and cost estimateswould be made separately for each class of pa-tients that might be treated with cimetidine. Thebasis for separating groups of patients could bedemographic features (e.g., age, race, sex), clin-ical diagnosis (e. g., duodenal ulcer, gastriculcer, Zollinger-Ellison syndrome), or stage ofdisease (e.g., number of days since diagnosis,previous treatments, complications).

Figure 4 is a paradigm decision-tree that dis-plays the sequence of decisions and chanceevents that follow from the initial choice of in-tervention in a particular group of patients with

ulcer disease. Clearly, the model requires a verylarge amount of data. It is not possible in thisreview to discuss potential sources of data forevery estimate that follows each choice of strat-egy. Rather, we select for discussion the majorelements of information required by the modeland the available evidence.

Our primary emphasis is on patients withduodenal ulcer, the most common form of ulcerdisease; we discuss patients with gastric ulcer inless detail. In addition to being used in these pa-tients, cimetidine is sometimes used in patientswith gastrinoma.

12 The traditional treatment forpatients with gastrinomas includes gastrectomy

I zThlS is a ~asl rin-secret ing tumor, usual ]y located In the Pan-

creas. It causes the Zoll inger-Ell ison syndrome of severe ulcers, in-tractable pain, and diarrhea. First described In 1955, the Zollinger-Ellison syndrome has been recorded in more than 2,000 cases in theliterature (101 }.

26 ● Background Paper #2: Case Studies of Medical Technologies

Table 9.—Components and Measures ofComponents in a Benefit-and-Cost

Analysis of Cimetidine

Clinical effectsShort-term● Heal ing● Relief of symptoms● Side effects and

adherence● Complications● RecurrenceLong-term. Recurrence● Side effects and

adherence● Complications

Health system effectsMedication● Antacids● Anticholinergics. Diet. OtherDiagnostic testsŽ Laboratory● Monitoring chemistries. Imaging

—X-ray—Endoscopy

● Physiologic function● Gastric acidPhysician visits

● Number/time periodHospitalization● Number/time period● DurationSurgery● Number/time period● T y p eOther● Non-M. D. provider visits● Nursing care

OutcomeHealth. Mortality

—Number of deaths—Age-adjusted mortality—Years of life lost

● Morbidity—Days and severity of

pain— Days of disability

Resource costs and savings● Days of work lost

—Premature death—Temporary and

permanent disability● Cimetidine purchase● Implications of health

system effects

(excision of the whole or part of the stomach)time of surgery for the primary tumor, butcimetidine has been employed successfully as analternative to gastrectomy in these patients(99,138). Because of the rarity of gastrinoma asa cause of ulcer disease, the costs and benefits ofthe use of cimetidine in patients with this diseaseare not significant from a societal viewpoint.Since the clinical value of cimetidine fornonulcer disease such as dyspepsia (94) and up-per gastrointestinal hemorrhage (41) is outsidethe scope of this report, we do not address itbelow. We limit our focus to elements of thecost effectiveness of cimetidine in peptic ulcerdisease and do not attempt a global assessmentof the value of this drug.

Clinical Effects

No treatment for duodenal ulcer has beensubjected to as many randomized, controlled,double-blind studies as cimetidine has (68).These studies of cimetidine vary in their metho-dological stringency and completeness. In a re-view of the quality of 10 published, random-ized, controlled trials of H2 antagonists (includ-

Figure 4.—Paradigm Decision Tree: Cimetidine and Alternative Intervention Strategies

Patient population: Initial treatment Patient adherence Clinical effect: Clinical effect:Disease/stage/ I decision

Ito treatment I Healing (in a given I Pain relief in a

demographic featuresI

regimen (spectrum)I

time period)I spectrum

given time periodI (spectrum)

❑ : Decision node (matters of choice)O: Chance node (probabilistic events)

Figure 4.— Paradigm Decision Tree: Cimetidine and Alternative Intervention Strategies (Continued)

Clinical effect: Clinical effect: Health system effect: Health system effect: Health system effect:

Treatment side Disease course and Change in use of other Number of physician Tests used (yes/no

effects (spectrum) complications

(short and long I (short and long

medication (by type v is i ts / t ime per iod

I o f m e d i c a t i o n r e l a t e d

choice for each test

I I in each time period)

term).. term) 1 to change in clinical 1 I

Health system effect: I Health system effect: I Outcome. Health I Outcome: Resource

Hospitalization (in each Surgery (by type,I I

(for each strategy)I

costs (net, for each

time period) i n each time period) strategy)

I I I❑ Decision node (matters of choice)O Chance node (probabilistic events)

28 ● Background Paper #2: Case Studies of Medical Technologies

ing several on metiamide), Chalmers, et al. (25)rated only one “poor” —a record that comparedquite favorably with Chalmers, et al. ’s assess-ment of clinical trials of other treatments forulcer disease.

There is one important methodological differ-ence between the controlled studies of ulcer dis-ease done in the 1970’s and those done earlier. Inthe more recent studies, fiberoptic endoscopyreplaced gastrointestinal X-rays as the meansused to verify the presence and healing of ulcers.This direct visual confirmation of ulcer statuscan reduce diagnostic errors and consequentvariability in experimental results. As a result,endoscopy-controlled studies may be more like-ly to find statistically significant differences inthe clinical effectiveness of various treatments,

In addition to controlled studies, several sym-posia have been devoted to cimetidine (22,52,150), and a number of review articles have ap-peared in major medical journals (e.g., 48,126).This work has provided reliable informationthat can contribute to estimates of clinical bene-fits and risks in a CEA, but a number of impor-tant areas of uncertainty remain.

Short-Term Clinical Effects

HEALING

At least 10 double-blind, placebo-controlledstudies examining the short-term clinical effectsof cimetidine in patients with duodenal ulcershave been published in the English language.Together these 10 studies (see table 10) providecompelling evidence that cimetidine promoteshealing of duodenal ulcers. Overall, the rate ofhealing in 4 to 6 weeks among cimetidine-treated patients was approximately 70 percent,almost twice the level achieved by placebo-treated patients (36 percent). Similar resultswere obtained in a half-dozen additional studiesconducted in France, West Germany, Italy, andSpain (7).

One notable exception to the almost uniform-ly significant findings of cimetidine’s superiorityover placebo is the large, multicenter U.S. studyby Binder, et al. (11). Among outpatients assess-ed at the end of 4 and 6 weeks (57 on cimetidine;54 on placebo), no significant differences were

observed in the proportions healed (67 and 56percent, respectively). [t is evident that thestatistical conclusions from this study are dif-ferent from the others not because of worse per-formance of cimetidine, but because of a sub-stantially higher rate of healing within the pla-cebo group.

It is possible that the patients in the U.S. trialdiffered from those in the European studieseither because of differences in the naturalhistory of the disease in different countries orbecause the U.S. subjects tended to be at a dif-ferent stage of illness. For example, some of theearlier European studies were restricted to pa-tients who were considered candidates for sur-gery. The importance of criteria for patientselection and evaluation, as well as possiblevariation in the course of disease in differentcountries, was stressed in a Swiss study thatfound a very high proportion of patients withpeptic ulcer healing under placebo treatment(125).

It is possible that the discrepant results arepartly related to differences in antacid consump-tion. With one exception (108), all the con-trolled studies permitted ad libitum antacids forall patients. Patients in the European studieswere usually provided tablet antacids, whichare less potent than the type of liquid antacidused in the U.S. study (81,106). Overall, theU.S. patients consumed more antacid than theirEuropean counterparts. More to the point,among the subjects in the U.S. study, placebo-treated patients whose ulcers healed consumedmore antacid than those whose ulcers did notheal. (Mean antacid consumption was 12 per-cent higher among inpatients whose ulcer healedand 112 percent higher among outpatients thanin those whose ulcers failed to heal; differencesin median antacid consumption were 68 and 21percent, respectively. )

This raises the possibility that a partialtherapeutic effect was realized in the placebogroup in the U.S. study. Underlying this pos-sibility is the assumption that antacids promoteulcer healing. Antacids have been shown in atleast two endoscopy-controlled studies to have agreater effect than placebo on healing of duo-denal ulcer (93,116). One, a study by Lam, et al.

Case Study #11 Benefit-and-Cost Analysis of Medical Interventions. The Casc of Cimetidine and Peptic Ulcer Disease ● 2 9

Table 10.—Short-Term, Double-Blind, Placebo-Controlled Studies of Cimetidine:Effect on Duodenal Ulcer Healing

Study

D1

D2

D3

D4

D5 ‘

D6

D7

D8

D9

D10

CimetidineInvestigator a daily doseyear/country (grams)— .— — — . . — .Bank, et al. (6) 1.2

1976/South 1.6Africa

Bardhan, et al. (8)1979/UnitedKingdom

Binder. et-al. (11) ‘ -

1978/UnitedStates

Placebo

Duration Number of Number/o/.(weeks) patients b healed

6 19 (19) 8 (420/o)

14 46 (50) 13 (28°/0)

2

12 2 (inpatient) 49 (53) 18 (37%)

2 (outpat ient ) 27 7 (260/. )4 (outpatient) 27 (103) 13 (48°/0)6 (outpatient) 27 17 (63%)

Cimetidine v.

Cimetidineplacebo;

significantNumber of Number/o/o differencepatients b healed (ps0.05)

8 (8) 7 (86%) Yes (p< 0.01)

11 (11) 9 (82°/0)

70 (78) 43 (61 %) Yes (p< 0.001)

64 (72) 45 (70%)

43 (45) 24 (56%) Yes (p< 0.05)

26

}

12 (46%) Yes (p< 0.05)28 (107) 16 (570/’) No29 2 2 ( 7 6 % ) N o

— — B-lack wood, et al. (13) 1.6 - 6 12 (NA) d 3 (25%) 11 (NA)d 9(82%) Yes (p< 0.025)

1976/UnitedKingdomC

. —Bodemar and

Walan(15) 0.8 6 14 (15) 2(1 4°/0)15 (15) 12 (80°/0) Yes (p< 0.001)

19761 Sweden 1.2 15 (15) 14 (93°/0)

Gray, et al.West Germany 1 4 20 (20) 5 (20°/0) 20 (20) 17 (850/o) Yes (p< 0.0005)

Hetzel, et al, (76) 1.2 ‘- 6 42 (44) 16 (380/o) 43 (44) 36 (840A) Yes (p< 0,001)1978/Australia

Moshal, et al. (108) -

1977/South 0.8 6 19 (21) 8 (42°/0)I

‘g ( 4 0 ) 14 (74°/0)Africa Yes (p< 0.05)

1.2 17 11 (65%) —

Northfield and 6 21

)

4 (19°/0) 21 13 (620/o)Blackwood (1 10) 1.6

(NA)d (NA)d Yes (p< 0.05)

1977/UnitedKingdom e 12 15 4 (27%) 17 15 (88°/0)

Semb, et al. (129)1977/Norway 1.2 4 20 (20) 12 (60°/0) 20 (22) 17 (850/o) No

aNumber~ [n parentheses refer to references listed at the end of this case study ‘NA Not availablebN “mbers , “ p’rerl~heses are numbers enterln9 study elncludes Patients from study D4cpatlents Included In sfudy D9

SOURCE Modlf!ed after K D Bardhan C(metldlne (n Duodenal Ulceration “ 1978 (7), and D H. Wlnshlp, “Clmetldlne In the Treatment of Duodenal Ulcer,’” 1978 (154)

(93) found that 20 (77 percent) of 26 Chinese pa- cidental patients (93). The second study, bytients experienced ulcer healing after 4 weeks’ Peterson, et al. (116) in the United States, usedtreatment with aluminum-magnesium antacid higher doses of liquid aluminum-magnesium an-tablets (25 mEq per dose; seven doses per day). tacid (150 mEq per dose; seven doses per day).Only 8 of 24 patients treated with placebo ex- In this double-blind trial, 28 (78 percent) of 36perienced healing at the end of 4 weeks, a sig- patients on high-dose antacids experienced heal-nificantly lower fraction (p < 0.005). Relatively ing ulcers at 28 days, compared to 17 (45 per-low doses of antacids were effective in the Lam cent) of 38 patients on placebo (p < 0.005). Astudy; Chinese patients have similar parietal cell recent British review, prompted by the lessermasses and lower acid production than Oc- reliance on antacids by British clinicians com-

30 ● Background Paper #2. Case Studies of Medical Technologies

pared to Americans, concluded that antacidswere effective in promoting healing of duodenalulcer (106).

Notice that the rates of healing with antacidsin the Lam (93) and Peterson (116) studies (77and 78 percent, respectively) are much higherthan the healing rate (56 percent) in the placebogroup using ad libitum antacids in the multi-center U.S. trial of cimetidine (11). In fact, thehealing rates with antacids in the Lam and Pe-terson studies were even higher than the rate ofhealing with cimetidine (67 percent) in the U.S.multicenter trial (11)

This observation leads to an important ques-tion: Is cimetidine more effective than a con-certed antacid program in promoting ulcer heal-ing? The question is important because a CEAshould seek to compare the incremental effectsof competitive alternatives with one another, aswell as with a do-nothing strategy.

It is statistically unsound and maybe mislead-ing to compare selected groups from differentstudies. Fortunately, at least one randomized,double-blind study has compared cimetidinewith intensive antacid therapy in patients withduodenal ulcers (80). This multicenter trialfound that in 15 (52 percent) of 29 patients tak-ing antacids seven times daily, and in 40 (62 per-cent) of 65 patients taking cimetidine, ulcershealed after 4 weeks .13 The rate of healing in pa-tients taking cimetidine was not significantlybetter than the rate in patients taking antacid (p> 0.1), and the authors concluded that “800 and1,200 mg of cimetidine daily produced duodenalulcer healing and pain relief equivalent to 210ml of Al-Mg antacid daily” (80).

This conclusion should be qualified. Conven-tional tests of significance, as employed by theseinvestigators, are concerned with the risk offalsely rejecting the null hypothesis of “no dif-ference” between treatments (the a or type Ierror). In the Ippoliti study, the observed dif-ference did not justify a conclusion to reject thehypothesis of “no difference” at a 95-percent

IJThe Cimetidine patients were divided into two groups with dif-ferent dosage regimens: 33 patients received 1,200 mg daily and 21(64 percent) experienced healing; 32 patients received 800 mg dailyand 19 (59 percent) experience healing.

level of confidence. However, also of concern isthe complementary error, namely the failure toreject the null hypothesis when in fact a differ-ence in treatment outcomes in present (the B ortype 11 error) (49). This error, which may beclinically important, has been overlookedfrequently in trials of the treatment of duo-denal ulcer (27), as well as in other medical re-search (54).

We have estimated that if cimetidine trulyhealed 10 percent more ulcers than did antacids(62 V. 52 percent, the findings of the Ippolitistudy), then, given the number of patients in thetrial, there was less than one chance in three thatthe investigators would have found a statistical-ly significant difference.14 This would argue fora more tentative clinical conclusion. It argues aswell for more extensive research on the ques-tions of the relative clinical effectiveness ofcimetidine and antacids. 15

Several double-blind randomized trials havecompared cimetidine to placebo in patients withgastric ulcer. These are summarized in table 11.(A number of additional reports of interim re-sults (150) and studies without endoscopic as-sessment of healing (95) are excluded. ) Two ofthe European trials— one by Bader, et al. (5), theother by Frost, et al. (56)-—found a statisticallysignificant improvement in healing with cimeti-dine at 4 and 6 weeks, respectively. However,this finding was not borne out in the trials byCiclitira, et al. (28) and Dyck, et al. (40). Thelatter trials did tend to favor cimetidine (14 perc-ent more patients healed at 4 weeks in the Cicli-tira study (28) and 19 percent more at 6 weeks inthe Dyck study (40)), but these differences werenot statistically significant. The point madeabove concerning the chance of B-error appliesto the interpretation of these studies as well.Also pertinent is the earlier discussion of thetendency of U.S. patients to consume greater

ItMore Preclse]y, given the stated assumptions, the power of theexperiment (1 -~) is estimated to be 0.68.

‘51ppoliti has conducted a second, unpublished randomized trialof patients with duodenal ulcers, treating 65 patients withcimetidine and 62 patients with an intense antacid regimen. In thisstudy, the proportion showing healed ulcers at 4 and 6 weeks wasvirtually identical in the two groups. At 4 weeks, the proportionswith healed ulcers were 62 percent for patients taking cimetidineand 66 percent for patients taking antacids; at 6 weeks, the propor-tions were 85 percent and 84 percent, respectively (67).

Case Study #11. Benefit-and-Cost Analysis of Medical lnterventions: The Case of Cimetidine and Peptic Ulcer Disease ● 31

Table 11 .—Short-Term, Double-Blind, Placebo-Controlled Studies of Cimetidine:Effect on Gastric Ulcer Healing

Cimetidine v.

Placebo Cimetidineplacebo;

Cimetidine significantInvestigatora daily dose Duration Number of Number/O/O Number of Numberl% difference

Study year/country (grams) (weeks) patients b healed patients b healed (p<0.05)

G1 Bader, et al. (5)19771 France 1 4 27 10 (37°/0) 26 18 (690/o) Yes (p <0.02)

G2 Ciclitira, et al. (28)1977 UnitedKingdom 1 4 25 13 (52%) 35 23 (66°/0) NO

G3 Dyck, et al. (40)19781United States 1.2 2 28 4 (14%) 29 7 (24%) NO

G4 Frost, et al. (56)1977/United Kingdom 1 6 22 6 (270/o) 23 18(78°/0) Yes (p< 0.002)

aNumber~ in parentheses refer to references hsted at the end of this case study

SOURCE Based on J W Freston, “Clmetldlne In the Treatment of Gastric Ulcer R@vlew and Commentav “ 1978 (55), and H R Wulff and S J Rune, “A Comparison ofStudies on the Treatment of Gastric Ulceration With C!met!dlne, 1978 (156)

amounts of antacid, which hinders comparisonamong studies done in the United States andEurope.

The effectiveness of antacids alone in thehealing of gastric ulcer is debatable, with con-trolled studies reaching conflicting conclusions(55). One multicenter, randomized study of pa-tients with gastric ulcer in the United Statescompared three treatment regimens: cimetidinealone, antacid alone, and cimetidine plus ant-acid (43). This study found no significant differ-ences in healing among these three groups at 12days or 6 weeks.

16 N. control group taking pla-cebo only was included, apparently because ofethical concerns about withholding a potentiallyeffective treatment (i. e., antacids) from all pa-tients (55). This omission, subsequently la-mented by at least some of the investigators(53), leaves open the question of whether treat-ment with either cimetidine or antacids is supe-rior to placebo in patients with gastric ulcer.

In summary, cimetidine has been shown con-clusively to promote healing of duodenal ulcer,and some evidence suggests it is more effective

than placebo in patients with gastric ulcer .17 Ingeneral, European studies have found more fa-vorable results with cimetidine than have U.S.trials. In patients with gastric ulcers, cimetidinehas not been shown convincingly to be more ef-fective than an intense course of antacids.Whether cimetidine is more effective than an in-tense antacid program in healing duodenalulcers is still open to question. Of course, pro-motion of healing is only one aspect of short-term clinical performance (see table 9, p. 26).

PAIN RELIEF

Seven of the 10 randomized, controlled stud-ies of duodenal ulcer listed in table 10 also com-pared cimetidine to placebo in terms of painrelief. Those findings are summarized in table12. Comparison across studies is complicated bythe variety and subjectivity of measures em-ployed. These measures include frequency ofpainful days and nights, number of pain-freeweeks, severity of pain, proportion of asympto-matic patients, and days of treatment requiredto achieve symptom relief. An additional com-plication arises because the time frame for meas-

‘OAfter 12 days, 16 percent of 67 patients taking antacids, 20 per-cent of 71 patients taking clmetldine, and 25 percent ot 65 patientstaking both experienced ulcer healing: after 6 weeks, the resultswere, respect ivei y, 61 percent (JI 62 pat]ents, 59 percent (If 68 pa-tients, and 70 percent (lt 60 patient+.

“In addition to cimetidine and antacids, three other drugs havebeen shown in controlled clinical trials to promote healing ofulcers better than a placebo: colloidal bismuth, carbenoxolone,and trimipramine (139). None is now used for this purpose in theUnited States.

32 ● Background Paper #2 Case .Studies of Medical Technolologies

Table 12.—Short-Term, Double-Blind,Placebo-Controlled Studies of Cimetidine:

Effect on Duodenal Ulcer Pain Relief

Study a Summary of results

D1

D2

D3

D4D5

D6

07Da

D9D10

Cimetidine group asymptomatic for mean of 4 outof 6 weeks; placebo group free of symptoms fora mean of 2.4 out of 6 weeks; in other words,the cimetidine group had a 44% reduction inthe mean number of weeks with some pain.(Statistical significance not reported.)

Cimetidine group experienced a 34% reduction indays with pain and a 36% reduction in nightswith pain compared to the placebo group. Dif-ferences in the frequency of pain were sig-nificant in each of the 4 weeks of the study (p< 0.005).

Inpatients: significantly more patients takingcimetidine than taking placebo had day andnight pain relief at the end of 3 days (61 v. 30%;p < 0.01); difference not significant at the endof 1 week (69 v. 55%).

Outpatients: significantly more patients takingcimetidine had day and night pain relief at theend of 1 week (45 v. 31 O/.; p < 0.05); differencenot significant after the first week.

Not reported.Cimetidine group had “significantly lower pain

score” (p range < 0.02 to < 0.1) both day andnight during the first 3 weeks (method of meas-urement and computation not fully explained).After 3 weeks, cimetidine group had “signifi-cantly” (p < 0.1) less day pain in weeks 5and 6.

Cimetidine group had significantly more pain-freedays each week of the study (p < 0.01).

Not reported.At the end of 6 weeks, cimetidine group had sig-

nificantly more asymptomatic patients (780/’)than did the placebo group (47°/0) (p < 0.025).

Not reported.Cimetidine group required significantly fewer days

to achieve symptom relief (8.1 ± 9.9 [S. D.]) thandid the placebo group (20.6 ± 9.1 [S.D.]) (p‹0.01).

asee table 10 for lnvestlgator, year, and country

urement varies from study to study. Nonethe-less, these studies fairly consistently repor tgreater pain relief with cimetidine than with pla-cebo. Again, the multicenter U.S. trial (11)shows less striking differences than the othertrials.

The Ippoliti study (80) comparing cimetidineto antacid found prompt symptom relief withboth treatments (55 percent of cimetidine-treated patients and 58 percent of antacid-treated patients became asymptomatic by day).

At the end of 4 weeks’ treatment, 63 percent ofpatients taking antacids and 80 percent of thosetaking cimetidine were asymptomatic, a differ-ence that was not statistically significant (p >0.1).

Placebo-controlled studies of patients withgastric ulcer varied in the extent to whichcimetidine-treated groups experienced morerapid or complete pain relief than those treatedwith antacids (see table 13). The Englert study(43) comparing antacids and cimetidine foundsimilar symptom response in all groups.

Investigators differ in their conclusions aboutthe correspondence between ulcer healing andpain relief. Several investigators report a poorcorrelation between healing and symptom relief(e.g., 6,62,108), and others say the correlation isgood (e.g., 8,61,80). Part of the reason for dif-fering assessments may be a difference in whatvarious investigators consider a good or a poorcorrelation. For example, Bardhan (8) found theassociation between healing and symptom reliefto be significantly different from what wouldhave been expected to occur by chance. On theother hand, the same data show that the ability

of pain relief to predict ulcer healing is not verystrong. 18

As in our discussion of B-error above, thispoints out the important distinction between theinterpretation of results based on statisticalcriteria and that based on clinical criteria:Results that fail a test of statistical significancemay still be clinically meaningful; conversely,statisticall y significant differences may not beparticularl y meaningful clinically. The degree ofassociation between healing and pain relief ispertinent to a cost-effectiveness assessment,because a patient’s decision to return to normalactivity depends at least as much on symptomsas on the physical repair of the ulcer. If esti-mates of a drug’s comparative effectiveness inreturnin g patients to work are based primarily

on healing, they may be misleading insofar as

‘“The probability (If ulcer healin g given symptc)m improvementIS 50 60 = 0.77, and the pr(~babi] I t y of nonhea] ing given noimprovement In symptc~ms is 42/72 = 0..58, These are n[)t ~lar.ticularly large predictive values. In the Ippoliti study (80), the cor-resp~~nding values are a bit lt]wer in the first case (42 ’60 = O 70)and somewhat h]gher in the second (18 20 = 0,90),

Table 13.—Short-Term, Double-Blind,Placebo.Controlled Studies of Cimetidine:

Effect on Gastric Ulcer Pain Relief

Study a Summary of results

G1 Patients taking cimetidine tended to have morerapid and greater relief of pain, but differenceswere not statistically significant.

G2 Cimetidine-based group had significantly fewerattacks of pain during each week of the study.

G3 No systematic or significant differences in theseverity or frequency of pain at 2 weeks or at 6weeks between the cimetidine-treated andplacebo-treated groups.

G4 Group taking cimetidine had fewer days of pain,but differences were not statisticallysignificant.

aSee table 11 for Investigator year and country

pain relief and healing do not correspond to oneanother, Almy (2) suggests a further possibilityif a patient returns to work after symptoms haveremitted but before healing has occurred: Work-days gained might be lost later on owing to lateconsequences of unhealed ulcer.

In summary, evidence from most controlled,double-blind studies suggests that cimetidinepromotes faster and more complete pain reliefthan does a placebo in duodenal ulcer, but notnecessarily in gastric ulcer. An intense antacidprogram appears to be about as effective ascimetidine, but more evidence on this questionis needed. The correspondence between healingand symptom relief is imperfect: The associa-tion is not random, but relief of symptoms is nota reliable clinical predictor of healing.

SAFETY AND ADHERENCE

No pharmacologic agent is perfectly safe. Adrug’s side effects depend on its toxicity, thedosage and duration of administration, and theindividual susceptibility of the patient. The im-portance of side effects of any one treatmentshould be judged in relation to the severity ofthe disease being treated and the risks of alter-native interventions.

Before turning to cimetidine, let us brieflyconsider the alternative of antacid therapy as abaseline. Unlike cimetidine, the Al-Mg antacidsuspensions usually prescribed are, for the most

part, not systemically absorbed. The most com-mon adverse side effect of these antacids is diar-rhea, which is related to the dose of magnesiumsalts. In studies with intense antacid regimens,27 percent (80) and 36 percent (43) of patientstaking antacids developed diarrhea. In thePeterson study (116) comparing antacids with aplacebo, 66 percent of the antacid group and 21percent of the placebo group were switched forat least 7 days to an alternative medicationbecause of diarrhea. Mild diarrhea may not bevery important medically, but this effect, alongwith the need for frequent administration, doesdiscourage patient adherence to high-dose an-tacid regimens. Aluminum salts bind phosphateions, and this may produce hypophosphatemiain patients who have intestinal malabsorptionproblems. This rare consequence may becountered by selecting a different type of an-tacid or giving phosphate supplements. 19

Cimetidine in short-term use has been associ-ated with a wide range of side effects. Themanufacturer instituted a formal, postmarkedsurveillance system that covered 9,907 ambula-tory patients and found a total of 577 adverseevents in 442 patients (4.4 percent of all pa-tients) (59). Only a fraction of the adverseevents were believed to be attributable to cimeti-dine; for example, 30 of 254 adverse gastrointes-tinal events occurred in circumstances thatstrongly suggested an association with cimeti-dine. No deaths were attributed to use of thedrug. An extensive review by Kruss and Littmanin 1978 (92) of publications, manufacturers’files, and submissions to FDA concluded thatcimetidine was safe enough to be used in pa-tients with duodenal ulcer disease for up to 8weeks, and indeed, this is the use currently ap-proved by FDA. A high proportion of patientsdevelops clinically insignificant elevations inserum creatinine which resolves promptly withcessation of therapy (92). Gynecomastia (ex-cessive development of breast tissue in males)has been reported in under 1 percent of patientson short-term treatment; the incidence increases

‘“A ditterent arr~y (Jt metabolic p r o b l e m s m a y Itlllow use ofcalclurn cdrbt~ndte antacid (which 1s abs~~rbed systemical [y ), but+1 nce this is nc~t usual]}’ presc rlbed hy physician+ in the UnitedState<. we will not c~~nslder It I urt ht’r

34 Ž Background Paper #2: Case Studies of Medical Technologies

with longer term use (72). Mental confusion(102), reversible hepatitis (145), and severalcases of severe allergic reaction (33) have alsobeen reported. Agranulocytosis, the principalproblem with cimetidine’s predecessor, meti-amide, has been reported to occur transientlywith cimetidine (32,90). In addition, at least onefatality due to aplastic anemia has been reportedin association with cimetidine (26).

Recent bacteriological studies of the gastricjuice of patients before and after 4 weeks’ treat-ment with cimetidine found major increases intotal bacterial counts, including large numbersof fecal-type organisms, following treatment(122), This effect, noted after short-term use,would be of greater concern with long-term useof cimetidine, as explained below. The principaldeterminant of gastric flora in humans is theacidity of the stomach, and increases in fecaltypes of bacteria are found in the stomachs ofachlorhydric patients such as those with per-nicious anemia (38). Patients with perniciousanemia have long been recognized to have a sig-nificantly increased risk of developing gastriccancer (107). Possibly, the increased incidenceof cancer is related to the metabolic activity offecal-type bacteria that reduce nitrates and maylead to the development of carcinogenic N-ni-troso compounds in the stomach (123). At pres-ent, however, this long-term risk of cimetidine(or any agent that chronically reduces gastricacidity) is speculative.

In the past year, new evidence has accumu-lated concerning the effects of cimetidine on thereproductive function in males. Earlier animalstudies had clearly shown an antiandrogenic ef-fect of large doses of cimetidine administered for6 to 12 months (92). During the past year, atleast one case of reversible impotence has beenattributed to cimetidine (155). In addition, astudy of seven patients with ulcer disease, duo-denitis, gastritis, or esophagitis found a 30- per-cent reduction in mean sperm count after 9weeks of cimetidine treatment; the luteinizinghormone response to luteinizing hormone-releasing factor was also reduced (144).20 This

“’The m e a n iperrn c~mnts were 134.3 mllli(>n per ml bet(>recvmet tdine and 94.0 mlllif)n per ml at ter treat men t. The in ves-t ig~t{lrs state that the reduct Ion wa~ 43 percent, but this ~]ver-

study included no control group of ill patientsnot taking cimetidine. Sperm counts remainedwithin the wide fertile range, but the antiandro-genic side effects of cimetidine should be eval-uated further. 21

Thus, cimetidine used for up to 2 months ap-pears to be a relatively safe drug, but reportedincreases and shifts in gastric flora and en-docrinologic effects are disturbing. Cimetidineis more risky than antacids, but less trouble-some to the patient. The more extensively adrug is used, the more difficult it is to impute acausal relation to sporadically reported side ef-fects or case fatalities. On the other hand, trulyassociated but rare side effects can affect sub-stantial numbers of patients if a drug is verywidely prescribed, as is cimetidine. One’s at-titude toward the safety of cimetidine dependsin part on the weight placed on the possibility ofunanticipated and remotely occurring side ef-fects such as those that have occurred with othermedications like diethylstilbestrol (75).

COMPLICATIONS

The major complications of ulcer disease arebleeding from the base of the ulcer, obstructiondue to swelling or fibrosis, perforation throughthe intestinal wall into the peritoneal cavity,and penetration into the pancreatic bed. Asnoted previously, these complications are rela-tively uncommon and rarely occur as the initialmanifestation of ulcer disease.

The principal question of interest here iswhether short-term use of cimetidine alters thelikelihood of near-term complications. SeveralBritish investigators have reported patients whodeveloped perforation of peptic ulcers shortlyafter the cessation of cimetidine therapy(60,148). An increased risk of perforation fol-lowing cimetidine therapy is not substantiatedby controlled studies comparing longer term useof cimetidine and placebo following an initialcourse of cimetidine. These studies, discussed inthe section on long-term clinical effects below,

statement is apparently based on dividing the difference in meansperm count (40. 3 million per ml3) by the final, rather than the in-itial, count,

‘ ‘Additional studies are underw~y, accc~rding to FDA (5 I I.

assess whether maintenance doses of cimetidinecan reduce the likelihood of ulcer recurrence,

ULCER RECURRENCE FOLLOWING SURGERY

Cimetidine has been used to treat ulcers thatrecur following surgery for ulcer disease. We areaware of two randomized, controlled trials ofcimetidine’s effectiveness in preventing ulcera-tion after surgery (71,88). These studies reacheddifferent conclusions. In Britain, Kennedy andSpencer (88) compared cimetidine with placeboin patients who had undergone one of a varietyof surgical procedures (including gastrectomyand vagotomy with and without a drainage pro-cedure) and who, after surgery, had developedulcers at various locations (stomach, duo-denum, or jejunum). The 12 patients treatedwith 1 g of cimetidine daily did not show signifi-cantly more healing at 6 weeks than the 12 pa-tients treated with placebo.

A more recent study in West Germany (71)was restricted to patients who had undergonepartial gastrectomy and developed ulcers at ornear the site of the surgery. After 4 weeks oftreatment, ulcers had healed in six of seven pa-tients treated with 1 g of cimetidine daily, butnone of the eight treated with placebo had heal-ed (difference significant, p < 0.01). After 8weeks, all seven cimetidine-treated patients, butonly one of eight placebo-treated patients hadhealed (difference significant, p < 0.01). The in-cidence of relapse after cessation of cimetidineand effects of maintenance on preventing recur-rence after surgery have not yet been reported incontrolled trials.

RECOMMENDATIONS FOR TREATING NEWLYDIAGNOSED, UNCOMPLICATED ULCER

Gastroenterologists differ in their recom-mended treatment for patients with newly diag-nosed, uncomplicated duodenal ulcers. Some,stressing comparable rates of healing, long ex-perience with antacids, and uncertainties atten-ding any recently introduced drug, recommendan initial trial of intense antacid therapy (140).Others, impressed with cimetidine’s per-formance and concerned about lack of patientadherence to an antacid regimen, prefer to usecimetidine (98).

The choice between antacids and cimetidine isclearly closely balanced. Rather than adopt ei-ther approach exclusively, a conscientious clini-cian might better weigh the choice for each pa-tient individually, taking account of presentuncertainties as well as each patient’s personali-ty and preferences. For example, patients varyin their willingness to persevere with antacids inthe face of mild to moderately uncomfortableside effects. In addition, patients, as well as doc-tors, vary in their attitudes toward known andunknown risks. Thus, a young man trying tostart a family would surely view possible anti-androgenic effects differently than would awoman or elderly man.

As times goes on, new evidence may reducepresent uncertainties about the comparativebenefits and risks of cimetidine. Individual pa-tient characteristics and values might still makethe preferred treatment different for differentpatients who are all classified in the same gen-eral diagnostic category.

Long-Term Clinical Effects

The use of cimetidine beyond the short-termtreatment of ulcer may take two forms: 1 ) inter-mittent administration if symptoms or ulcera-tions recur, and 2) maintenance treatment withthe aim of preventing ulcer recurrence.

Cimetidine is probably very commonly usedfor intermittent treatment of ulcers (7), but weare aware of no controlled studies comparingcimetidine to alternative approaches. One study(64) suggests that with cimetidine, healing of asecond ulcer is slower than healing of an initialulcer. In 25 patients with recurrent ulcers, 5 2percent healed after 4 weeks of treatment withcimetidine compared to 76 percent who hadhealed within 4 weeks after diagnosis of theirfirst ulcer. Interpretation of the results of thisstudy, however, is clouded by differences in theinitial treatment history of these patients andambiguity in the report. For example, the 25 pa-tients with recurrent ulcers included a majoritywhose first ulcers had been treated with cimeti -dine and others whose first ulcers had healedspontaneously. In addition, most of the 25 pa-tients had been maintained on placebo, but an

36 Ž Background Paper #2: Case Studies of Medical Technologies

unspecified number (between 1 and 7) had beenmaintained on low-dose cimetidine.

We are aware of no studies comparing main-tenance cimetidine to maintenance antacids.Perhaps it has been assumed that few patientswould adhere to long-term treatment with effec-tive doses of antacids. Grossman (67) suggeststhat this might be reconsidered in light of the re-cent study by Lam, et al. (93), who found thatrelatively small doses of antacids given in theform of tablets were effective in promoting thehealing of duodenal ulcers.

Most research on long-term use has comparedmaintenance doses of cimetidine to placebo(with antacids ad libitum). These studies formthe basis for the following discussion.

ULCER RECURRENCE

Table 14 summarizes the results of six double-blind controlled studies, published in English,comparing maintenance cimetidine to placebofor periods ranging from 80 days to 1 year. Pa-tients in these studies were given 400 or 800 mg

of cimetidine daily. Investigators consistentlyreport a statistically significant reduction insymptoms and recurrent ulceration during theperiod of treatment in the cimetidine-treatedgroup compared to those given placebo. Theconsistency of results is particularly strikinggiven the range of criteria used to selectpatients —with some studies including patientswith recently treated new ulcers (e. g., 70),others limited to chronically ill patients (e.g.,16), and others restricted to patients consideredcandidates for surgery (e. g., 64).

The conclusions from these studies are rein-forced by a recent review by Burland, et al. (23).These authors compiled results from 15 double-blind maintenance trials, either completed or inprogress, involving 695 patients. Overall, thenumber developing recurrent ulcers while tak-ing placebo appears to be twice that observedwith maintenance cimetidine. Approximately 10percent of patients treated with placebo and 50percent of cimetidine-treated patients remainedin remission during 12 months of treatment.

Table 14.—Controlled Trials of Maintenance Cimetidine in Peptic Ulcer

Ulcer recurrenceSymptomatic relapse (by endoscopy)

Initial Duration DifferenceInvestigator (premaint.) of maint. Maint. Number Number/% Number Number/% significant

Study year/country t reatment (months) t reatment ana lyzed relapse analyzed recur (p <0.05)M1 Bardhan, et al. (9) 53 cimetid. 6 P bid 31 18 (58%) 27 20 (74%) Yes

1979/United 7 other C 400 mg 29 4 (14%) (p< 0.005)Kingdom bid

M2 Blackwood, et al.recurrence

(13) Not 6 P hs 24 12 (50%) 24 21 (88%) Yes1976/United specified C 800 mg hs 21 8 (380/. ) 21 5 (250/.) (p< 0.0005)Kingdom

M3 Bodemar & Walanrecurrence

(16)/ 65 cimetid. 12 P bid 36 30 (83%) 36 38 (83%) Yes1978/United 3 other C 400 mg 32 12 (38%) 32 6 (38%) (P< 0.0005)Kingdom bid

M4 Gray, et al.(64)1978/United 52 cimetid. 6 P hs 30 24 (80%) 29 24 [83%) YesKingdom 8 other C 400 mghs 26 11(42%) 22 7 (32%)

M5 Gudmand-Hbyer, Not 12 P bid 25 20(80%) — Yeset al. (70)

—specified C 400 mg 26 3(12%) – (p< 0,1301)

19781 Denmark—

bidM6 Hetzel, et al. (78) Not 2 2/3 P bid 31 10 (32%)c — — Yes

Australia specified 14 (4570) —- .

—C 400 mg 36 0

bid

aNumber~ in parentheses refer to references listed at the end of this case studybp = pla~eboi c = clrnetldlne; bld = twice daily; hs = at bed!lrne.CThe repo~ states in the same paragraph both that 10 patients on placebo suffered relapse and that 45 perCer)t of those on placebo had relapsed

Results were identical with 400 and 800 mg ofcimetidine daily.

The performance of different maintenanceregimens may depend on the initial treatmentreceived by patients. Those whose ulcers havehealed initially with placebo, antacids, or cime-tidine may differ in their susceptibility to recur-rence. Consider the possibility that patientswith newly developed ulcers fall into two clini-cally indistinguishable subpopulations, one(type A) being more resistant to treatment andprone to relapse than the other (type B). Nowconsider the hypothetical experimental situationillustrated in table 15. Seventy patients in eachof two groups are assigned randomly to initialtreatment with cimetidine or with antacids. Ineach group of 70, 40 are type A and 30 are typeB. Both treatments produce healing in 75 per-cent of type A patients, but cimetidine is twiceas effective as antacids (67 v. 33 percent) in pro-ducing healing in type B patients.

After initial treatment, only those patientswho have healed are followed for possible re-lapse. (In the case of maintenance studies, onlypatients initially healed are tested with mainte-nance therapy. ) Assuming that a given propor-tion (one-third) of all type A patients and that alarger proportion (one-half) of all type B pa-tients both relapse within 6 months, then cimeti-dine-treated patients will appear to be moreprone to relapse (40 v. 38 percent). This can betrue even when, as shown in the last column oftable 15, cimetidine results in a greater fractionof the initial population of patients remainingasymptomatic.

Empirical evidence consistent with such anadverse selection of patients whose ulcers healinitially with cimetidine may be found in theresults collected by Burland, et al. (23). Amongpatients treated with maintenance placebo, 245(50 percent) of 290 patients initially treated withcimetidine developed symptomatic re-ulcera-tion, compared to 9 (30 percent) of 30 patientsinitially treated with placebo (difference signifi-cant, p c 0.05). On the other hand, there maynot be an adverse selection of patients who healfollowing cimetidine treatment as compared tothose who heal after antacid treatment. Ippolitifollowed patients with duodenal ulcer who hadbeen assigned randomly to treatment for up to 6weeks with a concerted antacid program or withcimetidine. 22 Among those whose ulcers healed,the rate of recurrence at 6 months (as deter-mined by endoscopic examination at 3 and 6months) was 54 percent among the 41 patientswho had been treated with cimetidine and 60percent among the 35 patients who had beentreated with antacids.

Following cessation of treatment, patientswho had been taking cimetidine begin to relapseat the same rate as the initial rate of relapseamong patients who were treated with mainte-nance placebo. This important finding is dem-onstrated in the study by Gudmand-Høyer, etal. (70). Once it is discontinued, maintenancecimetidine appears neither to accelerate recur-rence nor to effect any more permanent cure.

“Unpublished study (79).

Table 15.—Results of Treatment With Two Hypothetical Subpopulations of Ulcer Patients:Type A More Resistant to Treatment and Prone to Relapse Than Type B

Starting populationInitial treatment of patients

Cimetidine - Total = 70Type A = 40Type B = 30

Antacids Total = 70Type A = 40Type B = 30

Response to initialtreatment: Numberhealed (o/o of those

entered)

Total = 50 (71 O/. )Type A = 30 (75°/0)Type B = 20 (67%)

Total = 40 (570/. )Type A = 30 (750/. )Type B = 10 (33%)

Relapse in 6 monthsafter treatment

discontinued: Number Number remainingrelapsed (o/o of initially asymptomatic (0/0 of

healed) starting population)

Total = 20 (400/. ) 30 (43°/0)Type A = 10 (33°/0) 20 (50°/0)Type B = 10 (500/. ) 10 (33°/0)

Total = 15 (380/. ) 25 (36°/0)Type A = 10 (330/. ) 20 (50°/0)Type B = 5 (500/. ) 5(1 70/o)

1 L----- --- -_–--- –..–-.––_ -.._. _....__

38 ● Background Paper #2 Case Studies of Medical Technologies

The implications drawn from these studies areless consistent than are the findings. To some in-vestigators, the high relapse rate after cessationof treatment “suggests that prolonged cimeti-dine therapy is necessary to retain most patientsin remission” (76). Others, cognizant of thepotential unknown risks in treatment for longerthan 12 months, wonder whether “those pa-tients would have been better off if surgery hadbeen advised at a much earlier stage. A year hasbeen wasted in which they have been taking tab-lets daily when the end-result was surgery afterall” (Wulff, quoted in 150). But not all patientswho relapse become candidates for surgery, anda key question is the likelihood of their healingwithout surgery. Even if maintenance cimeti-dine accomplished nothing more than a l-yeardelay in surgery, it might be worthwhile forsome patients to defer the small, but definite,mortality risk from surgery in favor of the risksof cimetidine for 12 months. Indeed, as a pa-tient’s surgical risk increases, cimetidine’s un-known consequences become more acceptable(29).

In summary, compared to placebo, mainte-nance treatment with cimetidine significantlyreduces the chance of ulcer recurrence. Oncecimetidine is discontinued, patients begin torelapse at the same rate as they would havewithout maintenance treatment. We found nocontrolled studies of maintenance cimetidinecomparing alternative treatments other thanplacebo and no published reports studying peri-ods longer than 1 year of maintenance therapy.

SAFETY

Long-term studies with cimetidine turn up noimportant new side effects other than thosementioned in relation to short-term treatment.The incidence of gynecornastia may be as highas 4 percent in patients treated for 2 months to 1year (131). Presumably, changes in the bacterialflora of the stomach found after 4 weeks ofcimetidine treatment would persist with long-term therapy (122). As discussed earlier, thisraises the possibility that patients takingmaintenance cimetidine might have an increasedrisk of developing gastric cancer. Experienced

clinicians express concern that rare, but severe,side effects may not be evident in relativelysmall controlled trials and that risk of toxicity isgreatly magnified if treatment continues for pro-longed periods of time (70).

COMPLICATIONS

Available controlled trials tell us very littleabout possible effects of cimetidine on long-term complications of ulcer disease (hemor-rhage, obstruction, perforation, and penetra-tion). The reasons rest mainly in the nature ofthe disease, absence of reliable estimates ofbaseline rates (no randomly selected populationof ulcer patients has been followed over manyyears), and the comparative rarity of severecomplications, believed to be not more than afew percent per year following initial diagnosis(140).

As stressed by Grossman (70), the size of astudy needed to detect clinically relevantchanges in complication rates would be enor-mous. If, as he posits, 5 per 1,000 recurrencesresult in perforation, and we wanted to detect ata 0.05 significance level a treatment that wouldhalve the rate of recurrence, we would needmore than 15,000 patients in each of two ex-perimental groups to have a 90-percent chanceof finding that difference (49).

It may be that insofar as a treatment such ascimetidine therapy can reduce or delay recur-rence, it will reduce or delay complications.However, insofar as patients at higher risk ofcomplications are also more resistant to treat-ment that delays recurrence, reductions in com-plication rates will be less than reductions inrecurrence. The reasons are analogous to the ad-verse-selection bias hypothesized above. Thereis little convincing evidence that cessation ofcimetidine treatment can promote complica-tions (see earlier discussion), and likewise, thereare no convincing data from clinical trials thatcimetidine reduces complications. Given the sizeof studies that would be required, it seems un-likely that compelling evidence on this questionwill be forthcoming from controlled clinicaltrials.

PENDING APPROVAL BY FDA

At the present time, FDA is considering ap-proval of cimetidine for use longer than 8 weeksin patients with duodenal ulcer disease. Its ad-visory committee reportedly recommended inOctober 1979 the approval of maintenancecimetidine for patients who are at “high risk” forsurgery (50). This probably includes both pa-tients who are more likely to require surgeryand patients who are less likely to survive sur-gery. We understand that final decisions on thisquestion, as well as revised limits on the ap-proved duration of treatment, are not yet for-mulated. The principal drawback to longer termuse is the risk of unknown side effects. Avail-able evidence supports the effectiveness of cime-tidine in delaying recurrence. Physicians andpatient attitudes toward the unknown risks ofcimetidine will vary, but for those with relative-ly large and tangible risks from surgery,cimetidine is likely to be judged as a less danger-ous course.

FDA has not yet approved cimetidine for usein patients with gastric ulcer. This is apparentlyrelated to the conflicting evidence about the ef-ficacy of cimetidine for gastric ulcer (see table11, p. 31) and to a more general policy concernabout the possible role of nitrosoaminated com-pounds in the development of cancer.

Health System Effects

Empirical data on the health system effects ofcimetidine are more sparse than available in-formation about clinical effects. Some pertinentinformation is available, and several studies arein progress that may shed more light on these ef-fects, but at the present time, available evidenceis suggestive rather than conclusive. As dis-cussed in the next part of this case study, thelack of empirical evidence to inform estimates ofcimetidine’s health system effects seriouslyhandicaps available benefit-and-cost analyses.

Medication

Eight of the 20 placebo-controlled studies ofcimetidine shown in table 16 (D1, D2, D3, D5,D6, G2, G3, M3) compared antacid consump-

tion among patients in the experimental andcontrol groups. Five (D2, D5, D6, G2, M3) ofthe eight studies were conducted in Europe andone (D1) in South Africa. In these six studies,cimetidine-treated patients consumed between47 and 84 percent less antacid. In the remainingtwo studies (D3 and G3), both done in theUnited States, differences were less marked, andthere were no consistent trends toward de-creased antacid consumption among cimetidine-treated patients.

Possible effects of cimetidine use on the con-sumption of other drugs have not been reported.

Diagnostic Tests

Insofar as persistent or recurrent ulcer symp-toms lead physicians to perform diagnostictests, and insofar as cimetidine reduces or delayssymptoms, the drug could result in fewer diag-nostic tests if used without the constraints ofcontrolled triail protocols. It is important to bearin mind that cimetidine’s effectiveness in long-term use has been tested against placebo, butnot, to our knowledge, against an antacid pro-gram or other regimen. To the extent that cime-tidine produces biochemical or other abnormal-ities that physicians choose to evaluate further,it could increase the number of laboratory testsperformed.

In addition, if physicians felt obliged to screenfor unlikely but potentially serious side effects,such as granulocytopenia, the number of diag-nostic tests in patients treated with cimetidinecould increase. The presence and extent of thesedifferent effects are currently matters for spec-ulation.

Physician Visits

The range of potential effects posited for diag-nostic tests applies as well to physician visits.Secondary induced effects are also possible: Ifphysicians are visited less often for a principalproblem of ulcer disease, then less medicationand fewer procedures may be used for a lesstroublesome problem, such as mild to moderatejoint pain, that in itself might not prompt a per-son to seek medical care.

40 ● Background Paper #2: Case Studies of Medical Technologies

Hospitalization and Surgery

Costs of hospitalization and surgery are thelargest single component of medical expendi-tures for ulcer disease (see table 6, p. 19). Hence,the effects of an intervention such as cimetidineon the rates of hospitalization and surgery areparticularly important to a CEA.

Because, as discussed earlier, cimetidine wasdisseminated widely in a short period of time, itseems possible that its effects might be reflectedin global trends of hospitalization and surgery.Data we have compiled and analyzed fromNCHS do indicate an unexpectedly sharp de-cline in surgery in the first calendar year (1978)following the introduction of cimetidine in theUnited States. According to data from CPHAcompiled by Elashoff and Grossman (42), how-ever, the decline was less precipitous. Hospital-related effects that are linked more directly tothe use of cimetidine may emerge in the next fewyears from several studies currently in progress,which we will describe briefly.

Table 16.—Short-Term, Double-Blind,Placebo-Controlled Studies of Cimetidine:

Effect on Antacid Consumption

Antacid consumption ofcimetidine group compared to

Study a Time period that of placebo group

D1 6 weeks 83% reductionD2 4 weeks “Significantly” fewer tabletsb

03 First weekc Inpatients: no differencesOutpatients: 40% reduction

D4 — Not reportedD5 6 weeks 84% reductionD6 4 weeks 47% reductionD7 — Not reportedD8 — No antacids permittedD9 — Not reportedD10 — Not reportedG1 — Not reportedG2 4 weeks 61% reductionG3 2 weeks No significant differences

6 weeksG4 — Not reportedM l — Not reportedM2 — Not reportedM3 12 months 700/. reduction (approximately)M4 — Not reportedM5 — Not reportedM6 — Not reported

asee tables 10, I 1, and 14 for Investigator, year, and cOuntrYbscatter PIOIS of antacid consumption presented, no numbers provided or sta.

tlstlcal tests reported.c Not reported for later weeks of study

Two principal sources of nationwide hospitaldata are the Hospital Discharge Survey ofNCHS and the Hospital Record Study ofCPHA. Data from both sources are used in thisanalysis. Unless otherwise stated, HospitalRecord Study data are taken from a review byElashoff and Grossman (42). Hospital DischargeSurvey data were obtained directly from NCHSand then compiled for this case study.

Information from NCHS and CPHA is not inperfect agreement. Estimates in both theHospital Record Study of CPHA and the Hos-pital Discharge Survey of NCHS are based onsamples of non-Federal, short-term hospitaldischarges, stratified by hospital size and loca-tion. The fraction of records sampled is inverse-ly proportional to hospital size, so that theoverall probability of selecting a particulardischarge is approximately the same for eachclass of hospital size. Both sources estimatedischarges, not patients, so multiple admissionsfor an individual patient are indistinguishablefrom one-time-only admissions.

One major distinction between the twosources is the difference in parent populations ofhospitals. For the Hospital Discharge Survey,NCHS selects a representative sample from allU.S. hospitals. CPHA draws its sample for theHospital Record Study from the more than 750hospitals in its parent file. These hospitals com-prise approximately 13 percent of all U.S. hos-pitals, but they account for nearly 40 percent ofall hospital discharges. Thus, large hospitals areoverrepresented in the CPHA parent file. Thesubset selected for the Hospital Record Studydata is chosen to represent the size distributionfor all U.S. hospitals, but the extent to whichany bias is introduced by the inclusion or exclu-sion of U.S. hospitals in the CPHA parent set isnot well defined .23

Table 4 (p. 16) showed NCHS Hospital Dis-charge Survey data for peptic ulcer disease forthe years 1966 and 1970 through 1978. The first-

z lone ind ica kc)r t~f tfle represen ta t iveness of Hospital Rec~)rd-Study based figures is a comparison of Hospital Record Study es-timated deaths from peptic ulcer and total counts (not pr(~jrcti(~ns)tabulated by the NCHS Division c)f Vital Statistics. Between 1970and 1978, Hospital Rec~~rd Study estimates of annual deaths frompeptic u]cer were 92 to 114 percent of U, S. \’itd] Statistics counts(42).

Case Study #11 Benefit-and-Cost Analysis of Medical Interventions: The Case of Cimetidine and Peptic Ulcer Disease ● 41

listed diagnoses count those discharges forwhich one of the ulcer diseases was listed as theprimary diagnosis. These data, plotted in figure3 (p. 17), show a distinct downward trend overthe past decade for hospitalization of patientswhose first-listed diagnosis was ulcer disease.The CPHA Hospital Record Survey data show asimilar and significant downward trend, with aneven greater difference in the total decline from1970 to 1978 (42).24 According to both sets ofdata, the number of hospitalizations for ulcerdisease in 1978 is approximately what would beexpected by extrapolating the trend establishedthrough 1977.

Furthermore, both sets of data confirm thathospitalizations during the 1970’s have declinedfor duodenal ulcer and remained constant forgastric ulcer. Between 1970 and 1978, the ratioof hospitalizations for duodenal ulcer to thosefor gastric ulcer declined by 37 percent ac-cording to the Hospital Record Study and by 49percent according to the Hospital DischargeSurvey. This is further evidence for the epidemi-ologic distinction between duodenal and gastriculcer. The figures are incomplete because a rela-tively small number of diagnoses categorized as“peptic ulcer—site unspecified” is omitted, buttheir inclusion would not alter the trendsindicated.

Data from CPHA’s Hospital Record Studycan also be used to estimate the number of ad-

“Elashott and Grossman def]ne a trend as significant if: 1 ) theSpearman rank (}rder correlation is significant (p< O.OS), and 2)the difference between the 1970 and 1978 values exceeds the size ofthe Q5-percent c(~ntidence interval tor the med]an value (42).

missions for uncomplicated ulcer disease andthose for ulcer disease associated with hemor-rhage or perforation (42). Between 1970 and1978, uncomplicated duodenal ulcer admissionsdeclined by 46 percent; admissions for hemor-rhage declined by 37 percent; and admissionsfor perforation declined 24 percent, though fail-ing to reach statistical significance because ofthe small number of admissions for perforation.Uncomplicated gastric ulcer admissions showeda small, but significant, decline (18 percent). Noclear trend emerged for complicated gastriculcer admissions.

The number of surgical procedures for ulcerdisease has shown a decline during the 1970’sthat roughly parallels that for hospitalizations.Both NCHS and CPHA collect data on surgicalprocedures, but neither routinely relates oper-ations to discharge diagnoses. The principle sur-gical procedures used for ulcer disease are par-tial gastrectomy (excision of part of the stom-ach) and vagotomy (cutting of the vagus nerve),with pyloroplasty (enlargement of the pyloriccanal) or other drainage procedure (134). Thenumbers of these procedures performed duringselected years from 1966 through 1978 areshown in table 17 (NCHS data). During thisperiod, pyloroplasty and drainage procedureswere almost invariably performed in associationwith vagotomy. Virtually all vagotomies wereprobably undertaken for the treatment of ulcerdisease. Presumably, the great majority of par-tial gastrectomies were also done for ulcerdisease, but there are also several less commonindications for partial gastrectomy (e.g., gastriccarcinoma and trauma).

Table 17.—Number of Selected Surgical Procedures(Partial Gastrectomy, Vagotomy, Pyloroplasty and Drainage”) in the United States, 1966-78

Year

1966 . . . . . . . . . . . .1970 . . . . . . . . . . . .1972 ........, . . .1975. , ... , . . . . . .1976 . . . . . . . . . . . .1977 .., . . . . . . . .1978 . . . . . . . . . . . .

PyloroplastyPartial gastrectomy Vagotomy and drainage

74,500 61,000 56,80055,800 62,800 45,50063,300 59,300 42,00053,300 52,800 38,50054,200 48,300 31,20051,100 45,500 26,30039,700 29,200 20,600

aulcer dl~ea~e ,~ by far the most common Indlcatlon for these surgical procedures Over the time period shown PY1oroPlastY

and drainage were almost Invanably associated with vagotomy

SOURCE National Center for Health Statmtlcs, National Hospital Discharge Survey, Hyattsville, Md

Thus, the sum of partial gastrectomies andvagotomies can serve as a reasonable proxy forthe number of surgical operations done for pep-tic ulcer disease. Summing these two proceduresmay double count some patients who undergosurgery, because a patient who receives bothpartial gastrectomy and vagotomy is recordedunder both procedures. Despite the possibilityof some double counting, the trend over time inthe total of these two procedures would remaina useful index.

Estimates for the number of partial gastrec-tomies and vagotomies from NCHS tend to behigher than estimates from CPHA, but datafrom both sources show a distinct downwardtrend over time in the number of operations (seetable 8, p. 22). An acceleration (or deceleration)of this downward trend in surgery following theadvent of cimetidine might be ascribable to theintroduction of this new, widely used medica-tion.

We tested whether the number of surgicalprocedures performed in 1978 was differentfrom that which would be predicted by theprevious trend in the following way. First, wefitted a least-squares, linear regression line tothe surgical data available through 1977. Thepredicted number of procedures in 1978 is basedon a direct extension of the regression line. Thisis shown in figures 5 and 6, respectively, for theNCHS and CPHA surgery data. The NCHS es-timates of surgery are consistently higher thanthe CPHA estimates, but the rates of decline(slopes of the regression lines) are quite similar,within one standard error of each other .25 Theplots also show the 95- percent confidence inter-val about this regression line for individualestimates in each year for which data areavailable.

According to the NCHS data (figure 5), therate of surgery in 1978 is significantly (p< 0.01)below the rate that would have been predictedon the basis of the trend through 1977. The dropin 1978 is less striking in the CPHA data (figure6), but even here there is only about a 1 0 -percent chance that the estimated amount of

“Thr ~l(~pc (){ the regre~~lon line tor the’ NCHS data ts 0.33cII;that for the CPHA data is 0.3Q4s.

Figure 5.— NCHS Data on Number of SelectedSurgical Procedures (Partial Gastrectomy and

Vagotomy) in the United States, 1966.78

140

60,1965 1970 1975 1980

● Best fit computed by least-squares method. Confidence intervalsshown for curve fit to years 1966-77,

SOURCE Based on data from the Nat fonal Center for Health Stat[s ICS HyattsV1/1~ Md

Figure 6.—CPHA Data on Number of SelectedSurgical Procedures (Partial Gastrectomy and

Vagotomy) in the United States, 1966-78

40,000 JI 1 1 1 I E I I I I I J, --

1970 1975 1980

● Best fit computed by least-squares method. Confidence Intervalsshown for curve fit to years 1966-77.

SOURCE Based on data from the Comnllsslon on Professional and +ospltalActlvttles compiled by J D Ela Shoff and M I Grossman 1981)(42)

surgery in 1978 is in line with the precedingtrend. The number of surgical procedures in1978 was approximately 11,000 fewer than thepredicted number based on CPHA data, and26,000 fewer than the predicted number basedon NCHS data.

Further evidence for the apparent excessivedrop in surgery in 1978 compared to earlieryears comes from a comparison of the numberof surgical procedures as a proportion of hos-pitalizations for ulcer disease for various years(see table 18). Each year from 1966 to 1977, thenumber of operations was between 25 and 29percent of hospital admissions; in 1978, thenumber of operations for ulcer disease was 19percent of hospitalizations. Roughly speaking,for the decade before 1978, more than one infour patients hospitalized for peptic ulcer re-ceived surgery; in 1978, the proportion droppedbelow one in five.

If an unexpected decline in ulcer surgery did,in fact, occur in 1978, the question is, why? Didcimetidine play any role in the apparent drop?Are there other plausible explanations? Whatsorts of data might be obtained that couldanswer these questions?

Table 19 shows numbers and rates of surgicalprocedures for all abdominal surgery and forselected abdominal surgical procedures forselected years from 1970 through 1978. Therewas no general decline in abdominal surgeryover these years. Only surgery for peptic ulcerdisease shows a marked decline in 1978 com-pared with the rates in earlier years. Somesurgery for peptic ulcer is elective, and onemight imagine that part of the decline could berelated to a newly emerging, more cautious at-titude toward elective operations. This mightoccur, for example, as a result of more patientsseeking second opinions or of greater cost-con-

Table 18.— Proportion of Patients With First-Listed Diagnosisof Ulcer Disease Having Surgery, 1966-78

BA Number of patients

Number of surgical discharged withYear procedures diagnosis of ulcerb A/B— — .1 9 6 6 : : 135,500 526,000 0.2581970. ...., . . . . 118,600 438,000 0.2711 9 7 2 , 122,600 429,300 0.2861975 . . . . . . . . . . . . 106,100 411,700 0.2581976 . . . . 102,500 385,400 0.2671977 .., . . . . . . . 96,600 385,400 0.2511 9 7 8 68,900 360,400 0.191

al ~~1 udes part Ial gast rectom~ and va90t0mYblncludes gastric duodenal, gastrole)unal and peptic ulcer (site unspeclfled)

SOURCE Nat(onal Center for Health Statlstlcs, National Hosp/tal Dwcharge Survey, Hyattsvtlle, Md

Table 19.— Number and Rate of All and Selected Abdominal Surgical Procedures in the United States, 1970-78

Partial gastrectomyAll abdominal surgery and vagotomy Appendectomy Cholecystectomy a Herniorrhaphy b

—.Year Number RateC Number RateC Number RateC Number RateC Number RateC

—— —..1 9 7 0 2,440,000 122 119,000 6 325,000 16 367,000 18 496,000 251975 . . . . 2,894,000 138 106,000 5 319,000 15 442,000 21 549,000 261 9 7 6 . . . . , . , 2,809,000 133 102,000 5 306,000 14 442,000 21 507,000 241 9 7 7 2,937,000 139 97,000 4 342,000 16 446,000 21 533,000 251 9 7 8 2,830,000 132 69,000 3 299,000 14 432,000 20 510,000 24

a~urglca~ remo~~l o~~he gal I blad(erbsurglcal repa[r of a herniac Rates shown are per 10000 poputatlon

SOURCE Based on dala from the National Center for Health Stat[stlcs, National Hospital Discharge Survey, Hyatt svllle, Md

sciousness on the part of physicians. However,we find no parallel decline between 1977 and1978 in other abdominal surgery, such as her-niorrhaphy (surgical repair of a hernia), whichis probably more frequently elective than issurgery for ulcer disease.

A dramatic change in the criteria used todecide on surgery or use of a different type ofsurgery for patients with ulcer disease might ac-count for some decline. To our knowledge,however, neither the recognized indications forsurgery nor the types of operations havechanged dramatically in the past few years.Also we know of no changes in the standardcoding for operative procedures in 1978 thatmight account for the observed decline. Diag-nostic advances, such as fiberoptic endoscopy,may provide greater- assurance of benignity of aslowly healing gastric ulcer and thus avert somesurgery that would have been performedpreviously. Even if present, however, such ef-fects seem very unlikely to reach the propor-tions of the evident decline in 1978.

Results from at least one of the maintenancetrials comparing cimetidine with placebo sup-port the possibility that the decline in surgery in1978 is related to the availability of cimetidine.In a year of maintenance treatment, Bodemarand Walan (16) found that 1 patient in 32 whoreceived cimetidine and 15 in 36 who receivedplacebo underwent surgery because they hadtwo recurrences or because of severe symptomsat the first recurrence (difference significant,p < 0.0005). 26 Thus, one possible explanationfor the decline in surgery for ulcer disease in1978 is that the dramatic growth in the use ofcimetidine enabled more patients to be treatedsuccessfully medically. If cimetidine wereresponsible, the effect could be temporary. Pa-

‘@In a second maintena ace study that rept~rted sur~lcal e x -perience, p[wsible effects t~t cimetldine <~n ~urgery are obscured bythe practice of treating “placebo tal I ures ” with a L (~urse (~t L i met i-dine rather than surgery; th \ ettectwi a remissi{~n in m,]st ‘placebofai lures” durtng the 6 months [~t the study (M ~ In thi~ ~tudy, 30patients wew treated in]t ially with Imaintenance p]acebt~, 24 re-l a p s e d , al! t~t wht~m were t h e n trc>a ted with c I met i dine, and ~

underwent surgery wit h In t Ilt, ~-m on t h period L)} t hc \t udv. Ot the26 patients in it ial]y t rea tetl wtth rn~ in tenanct’ t inlet Idlnt>, 7 r~,-]apsed; 1 received surgery then ~nd ~ rm elved a w>c(~nci ct~urw of

(higher dcwe) c imetidine: 2 { t t h(wc ~ undt’rwwnt >urgc,rv wlthln the6-m(~nth perl(xi t~t the ~tudy

tients who were scheduled for an elective opera-tion may have decided with their physicians todelay surgery in order to try the new drug. Sincepatients appear to relapse at the same ratefollowing cessation of cimetidine, the decline insurgery might be followed by a compensatoryrebound, especially if more reports (e. g., 121)suggest increased risks or adverse side effectswith long-term use of cimetidine.

To date, only circumstantial evidence andargument by exclusion can make the case for therole of cimetidine in decreased rates of surgery.However, more direct evidence may be forth-coming from several sources. Murray Wylie ofthe University of Michigan is engaged in adetailed analysis of CPHA data on patients hos-pitalized with ulcer disease .27 Wylie has data onall patients discharged with a diagnosis of ulcerdisease from a cohort of 790 hospitals that par-ticipated continuously in the CPHA data systemfrom January 1974 through October 1978. Al-though the data do not specifically identify pa-tients who did and did not receive cimetidine, heis able to examine surgical rates on a month-to-month basis. Thus, he can test the corre-spondence between any accelerated decline insurgery and the introduction of cimetidine in theUnited States in August 1977.

Wylie’s preliminary impression is that the fre-quency of surgery began to drop even a fewmonths before the release of cimetidine. Hespeculates that this might be attributable to adelay in elective surgery in anticipation of thenew medication. It would be very informativeto compare changes in rates of surgery separate-ly for uncomplicated cases (presumably ad-mitted because of pain) and for those with hem-orrhage or perforation. If cimetidine is reducingsurgery by effecting a medical remission afterpatients are hospitalized, the largest drop insurgery as a proportion of admissions should befor patients hospitalized because of pain.

Wylie will also be able to analyze his dataseparately for surgical and medical admissions,including length of stay. This is of particular in-terest to a cost-effectiveness assessment of the

‘‘\$’e are gr<]tt’t ul t{) l’r(~tt’sw)r \4 }IIIL 1(, J \lIJ r lrl~ t }1 I> (l~,w rl p! It,n

{~t h t~ W( lrl, i n pr{~~r(’~~.

Case Study #11: Benefit-and-Cost Analysis of Medical Interventions: The Case of Cimetidine and Peptic Ulcer Disease ● 45

number of days of hospital care that might besaved by cimetidine. Presumably, some fractionof the reduction in surgery that might be at-tributable to cimetidine is due to patients notbeing hospitalized, and another fraction is dueto hospitalized patients being treated medicallyonly. (The large drop in surgery in 1978 com-pared with the drop in hospitalizations suggeststhat the latter fraction may be the larger. ) Onthe average, surgical lengths of stay would beexpected to be longer than medical, and a shiftfrom surgical to medical care in a hospitalshould typically produce a reduction in hospitaldays. If a very large number of patients who areconsidered potential candidates for surgery arefirst treated medically, however, any failures onthe medical regimen would then undergo sur-gery after a delay, and this could add to theaverage length of stay for patients. In addition,successful medical treatment with cimetidinemight or might not take longer than a medicalregimen without the drug. A few points of datawould be preferable to a lot of speculation.

Another approach to assessing cimetidine’seffects on the health system has been undertakenby Professors Burton Weisbrod and JohnGeweke at the University of Wisconsin .28 Theyare analyzing patient records developed for ac-counting purposes by the Texas medicaid pro-gram. Weisbrod and Geweke aim first to recon-struct medicaid expense records on a patient-by-patient basis for all patients with a diagnosis ofulcer disease. They have identified 1,206 pa-tients with ulcers in a sample that begins inJanuary 1976 and will extend to August 1979.These investigators have conducted a pilotstudy with 81 patients randomly selected fromthis population, 36 with and 45 without a his-tory of cimetidine use. Their intent is to com-pare the health and expenditure history (includ-ing nearly 50 categories of various expenses forhospitalization, physicians, drugs, nursinghomes, etc. ) for patients treated with and with-out cimetidine.

Weisbrod and Geweke recognize some in-herent limitations in the available data. For ex-ample, approximately one-fifth of the medicaid

‘F\\’c’ are gratet UI to [’r(lfes~t~r Cewcke t(lr \harlng intt~rmatl(lnabc~u t t hcl r \t udy t or t h 1+ reptlrt

claims do not include the patient’s diagnosis;medicaid patients over 65 years old are alsocovered by medicare, about which the investi-gators have no information; and the data do notinclude information concerning patient status atdischarge or work history. The major difficultyWeisbrod and Geweke face, however, is con-trolling for selectivity bias in those patients whodo receive cimetidine. Their approach is tostratify patients according to demographic fac-tors and clinical history. Although it will be im-possible to overcome the aforementioned bar-riers completely, Weisbrod and Geweke’s studypromises to be the first large-scale, patient-based study providing data on the direct and in-duced health system effects of cimetidine. Thedata base can also be used to describe the diffu-sion of the drug in a given patient populationand the pattern of present use by medical practi-tioners in one State.

We are aware of a few additional studies ofthe health system effects of cimetidine that arein more preliminary stages of development. Atleast one of these involves a health maintenanceorganization (HMO); if the HMO’s populationis sufficiently stable, it may be a particularlyvaluable setting for study. Ideally, one wouldseek results from a long-term, randomized, con-trolled study of patients with ulcer disease whoare or are not treated with cimetidine, but forethical and practical reasons, such a study isunlikely to materialize.

In summary, hospitalization and surgery forpeptic ulcer disease have both declined signifi-cantly during the past decade. The decline inhospital admissions for 1978 is consistent withearlier trends. However, the fall in surgical pro-cedures for ulcer disease in 1978 is unexpectedlylarge, amounting to 11,000 to 26,000 fewer pro-cedures in 1978 than would be expected from thetrend leading up to that year. The introductionand widespread use of cimetidine is one plausi-ble explanation for this unexpected decline.More specific information from studies in prog-ress, including a month-by-month tracing ofsurgical rates and a comparison of healthresources used by patients who did and did notreceive cimetidine, would help strengthen orrefute this inference.

46 ● Background Paper #2: Case Studies of Medical Technologies

Outcome

The outcome effects of cimetidine are conse-quences of its clinical and health system effects.We have already discussed how clinical andhealth system effects interest and lead to the twocomponents of outcome: health status and re-source costs. In this section, we present addi-tional empirical evidence about cimetidine’spossible effect on outcome.

The available empirical evidence plus meth-odologic and other considerations raised in alater section entitled “Guidelines for Review ofHealth Care Benefit-and-Cost Analyses” serveas a basis for our review of published analysesof cimetidine’s benefits and costs in the next partof this case study.

Health Status

As we have already discussed, it is convenientand usual to think of health status in terms ofmortality and morbidity.

MORTALITY

We are aware of no empirical studies of theeffects of cimetidine on mortality from pepticulcer disease. This is not surprising, becausemortality from this disease is relatively low. Acontrolled cohort study would require enor-mous numbers of patients, for reasons presentedearlier in the discussion of possible effects of ci-metidine on complication rates. One mightargue that insofar as cimetidine delays or sup-plants surgical intervention and attendant sur-gical mortality and delays the development ofcomplications, it will forestall some deaths.However, it is conceivable that patients whowould naturally develop the more virulent com-plications of peptic ulcer might benefit less fromcimetidine or that complications followingcessation of the drug would be more severe thanthey might have been had cimetidine not beenadministered. Any of these circumstanceswould counter potential improvements in sur-vival related to cimetidine. In addition, anysevere and unanticipated side effects from long-term use would further compromise cimetidine’sbeneficial effects on mortality.

Table 3 (p. 16) shows that mortality fromulcer disease has been declining steadily over thepast 15 years. Figure 7 shows NCHS ulcer mor-tality statistics on a quarterly basis from 1976 tomid-1979. It shows both a continuing down-ward trend and a seasonal variation in mor-tality, No unexpected mortality reduction fol-lowing the introduction of cimetidine in August1977 is evident. If cimetidine has saved lives ofulcer patients, the lives saved are too few tohave a substantial effect on overall mortality todate. Of course, these figures are mute on thequestion of whether even more widespread andconsistent use of cimetidine might demonstrablydelay or prevent deaths from ulcer disease in thefuture.

In summary, there are some reasons to be-lieve cimetidine might have beneficial effects onulcer mortality and other reasons to doubt it. Ifcimetidine did have a small beneficial effect onmortality, it would be very difficult to detect incontrolled cohort studies or from national mor-tality trends.

Figure 7.— Number of Deaths in the United StatesFrom Ulcer Disease (Gastric, Duodenal, and Peptic—

Site Unspecified), 1976-79

January- April- July- October-March June September December

NOTE: 1978 and 1979 figures extrapolated from a 10-percentsample.

SOURCE Based on data from the Natlooal Center for Health Statistics, Dlvlslonof Vital Stat jstics, Iiyattsvllle, Md

MORBIDITY

From the perspective of BCA, in which thereis an effort to translate morbidity into socialresource costs, an important consideration is theeffect of cimetidine on disability and days lostfrom work. Cimetidine produces more promptand consistent relief from ulcer pain than doesplacebo. In the short-term treatment of pepticulcers, it is reasonable to expect that faster heal-ing and pain relief can mean earlier return towork. This potential benefit may be reduced in-sofar as doctors prescribe and patients follow“rest at home” for a set number of days or weeksfollowing diagnosis of a new ulcer, irrespectiveof the promptness of symptom remission. Thatpolicy would be reasonable, for example, ifclinicians believed that patients returning to thestress of work with unhealed ulcers would bemore likely to develop bleeding or other com-plications of ulcer disease.

A number of the randomized clinical trials ofcimetidine in the United States included a spe-cial protocol to assess time lost from work(118). A preliminary report presented results in64 outpatients, 37 treated with cimetidine and27 with placebo. (Many of the 217 potentialsubjects were disqualified because of uncertainemployment status, a problem that is being rec-tified with a revised protocol. ) Among the pa-tients analyzed, there was a striking tendency tobe absent full time or to work full time. Com-pared to the number of days lost from workduring the week prior to treatment, the groupreceiving cimetidine averaged significantly moredays of work in weeks one, two, and four(p< 0.001) and in week six (P< 0.05) followingthe initiation of treatment. This report isnotable not only for its results, but because itrepresents an admirable effort to collect datapertinent to the economic consequences of amedical practice in the context of a controlledclinical trial.

One of the trials comparing maintenance ci-metidine with placebo also reported on the workexperience of patients (15). During the year ofthe study, 1 of 32 patients taking cimetidine didnot report to work for 79 days, and 23 of 26 pa-tients taking placebo did not report to work fora total of 1,405 days because of symptoms.

Thus, the cimetidine-treated patients reported towork an average of approximately 36 more daysper patient during the year of the study (dif-ference significant, p < 0.001).

The effectiveness of cimetidine compared toother treatments, such as antacids, in enablingpatients to return to work is not addressed inany of the controlled trials we have reviewed.

Resource Costs

The economic implications of an interventionsuch as cimetidine include the costs of the in-tervention itself, the resource costs and savingsrelated to induced effects on the health caresystem, and indirect effects on productivity re-lated to change in mortality and morbidity. Inthis section, we offer a few observations on thedirect costs of cimetidine compared to alter-natives. We defer consideration of the resourcevalue attached to the induced and indirect ef-fects of cimetidine until the next part of thisstudy, in which we review some of the benefit-and-cost analyses that have been carried out.

The daily cost of cimetidine is less than thedaily cost of antacid in doses that have beenshown to be as effective in promoting the heal-ing of newly discovered duodenal ulcers (80).The retail cost of cimetidine is approximately$0.25 to $0.30 per 300-mg tablet.29 Assumingconsumption of four tablets daily, the daily costof cimetidine is $1.00 to $1.20.

Antacids vary in their compositions, neu-tralizing capacities, and costs (115). Two of themore popular blends of aluminum hydroxideand magnesium hydroxide are Maalox® andMylanta II®. 30 The latter was the antacid used inthe studies by Peterson, et al. (116) and Ippoliti,et al. (80). Mylanta II® has approximately 50-percent more neutralizing capacity than thesame quantity of Maalox® and costs approx-imately $3.80 per 12-ounce bottle compared to$1.80 for Maalox®.31

-“’Ba+wi (In inlc~rmati(ln pr{>vicjecj b} tour Bostf}n-area d r u gst(~res a nci ct~n~i~ten t w’i t h est im a tw (}} the man utacturer.

‘OMylanta 1 1 c(~ntains, in addltit~n to antacid, t h e det(~arningsi I ictlne a ncl a n t Itla t ulen t ~imeth ic(~nc.

‘lC~~st e~timatcs bawd on average charge> at tour B(wt(>n-areadrug St (}res.

If we assume administration of the sameamount of antacid as used in the studies citedabove (seven daily doses, each with approx-imately 120 mEq of buffering capacity), thedaily cost would be approximately $1.58 forMaalox® (seven 45-ml doses) and $2.22 forMylanta® (seven 30-ml doses). If patients whoare prescribed cimetidine consume three or fouradditional doses of antacid daily, their medica-tion costs would still be comparable to those ofpatients who follow an intense antacid regimen.Thus, a typical patient can expect to pay nomore, and possibly somewhat less, for ci-metidine than for a therapeutically equivalentcourse of popular, brand-name antacids .32

Summary

Organized according to the benefit-and-costmodel for medical interventions presented earli-er, this part of our case study has describedavailable information about the effects of cime-tidine—its clinical effects, its health system ef-fects, and its potential impact on outcome.

Numerous controlled studies of patients withduodenal ulcer confirm that cimetidine pro-motes healing and provides faster and morecomplete pain relief than placebo. Less con-clusive evidence suggests the drug may be moreeffective than placebo for patients with gastriculcer. An intense antacid program appears to beabout as effective as cimetidine for patients withduodenal ulcer, but more evidence on this mat-ter is needed. Clinical studies have also shownthat relief of symptoms is not a reliable in-dicator of healing. In general, European studieshave found more favorable results withcimetidine than have’ U.S. trials.

Cimetidine used for up to 2 months appearsto be a relatively safe drug. Most known side ef-fects are minor or reversible, but recently re-ported changes in gastric flora and endocrino-logic effects are disturbing. Available studies of

“It may be pos~ible to tind less expensive, generic brands (Jt an-tacids with equivalent neutralizing capacity, but the costs (ItMaalox ” and Mylanta ” are no more than thcwe CJ[ m~wt otherbrands (~f aluminum-magnesium antacids (63), Antacids are ~va]]-ablc wit h(~ut a presc rip t itm, and their use d(w>s nc~t nccessart 1 y en-tail the ctwt tlt a phys]clan visit, but in this dlscussl(~n we haveassumed that a high -cl(w .lnt~cid regimen (Ir clmet i dine would bc~consumed on I y tol Ic)wi ng a ph ys]cia n \ ad~r ict’.

maintenance cimetidine do not alter this assess-ment. As with any new drug, uncertainty existsas to possible long-term consequences of thedrug’s use.

Compared to an intense course of antacids,cimetidine is comparably effective, more risky,and less troublesome to the patient with duo-denal ulcer. Cimetidine plus a moderate amountof antacids costs no more than a therapeuticallyequivalent course of intense antacid therapy,Experts now differ in their recommendations forinitial therapy of duodenal ulcer, some favoringcimetidine and others antacids. A reasonableapproach is to select therapy based on each pa-tient’s preferences and personality.

Compared to placebo, maintenance treatmentwith cimetidine as long as 1 year significantlyreduces the chance of ulcer recurrence. Oncecimetidine is discontinued, patients appear torelapse at the same rate as they would havewithout maintenance treatment. We are awareof no controlled trials comparing maintenancecimetidine to treatments other than placebo.There is little empirical evidence either thatcimetidine prevents future complications ofulcer disease or that cessation of cimetidine pro-motes complications. At present, FDA is con-sidering approval of cimetidine for use longerthan 8 weeks in patients with duodenal ulcerswho are at high risk for surgery.

In European trials, but not in U.S. studies,cimetidine-treated patients tend to consume lessantacid than placebo-treated patients. Verylimited empirical data are currently available onthe possible effects of cimetidine on use of othermedication, on diagnostic tests, or on physicianvisits. Several studies are underway that mayshed light on these matters.

Data we have compiled from NCHS show anunexpectedly sharp decline in the rates of sur-gery for ulcer disease in 1978, the first full calen-dar year after the introduction of cimetidine.This drop occurred against a background of fall-ing rates of surgery and hospitalization for ulcerdisease over the previous decade. Other expla-nations are possible, but the widespread use ofcimetidine may have contributed to the mag-

nitude of the 1978 decline in surgery for ulcer lost significantly fewer days of work than pa-disease. tients taking placebo, but no controlled study

There is little evidence of any effect oncompares work loss among patients on differ-

cimetidine on mortality from ulcer disease. Inent, effective treatments.

several studies, patients treated with cimetidine

— REVIEW OF BENEFIT-AND-COST ANALYSES OF CIMETIDINE

Available Analyses

We are aware of two analyses of the social re-source implications of cimetidine (109, 121).Both were sponsored by Smith Kline & FrenchLaboratories, through their office of cost-benefitstudies. One study, published by the Nether-lands Economic Institute in February 1977 (109),analyzed the possible effect of cimetidine on theDutch economy. That analysis estimated that ifcimetidine had been used by half of all ulcer pa-tients in the Netherlands, the potential savingswould have been $23 million, or 21 percent ofthe estimated $111 million total costs of ulcerdisease in 1975, We will not comment on thisstudy, because most of the issues it raises arealso raised in the second study, and the latter isa more recent analysis which focuses on theUnited States.

The second study, entitled The lmpact ofCimetidine on the National Cost of D u o d e n a lUlcers (121), was conducted by Robinson Asso-ciates, Inc., a marketing research and manage-ment consulting organization located in Penn-sylvania. Aelusions ofalong withpresented in

summary of the methods and con-the Robinson Associates study,our critique of the study, are

the section below.

The Study by Robinson Associates, Inc.

Summary of Methods and Conclusions

The Robinson Associates study (121) esti-mated that if cimetidine had been used in 80 per-cent of duodenal ulcer patients in the UnitedStates, 1977 national health care costs for duo-denal ulcer disease would have been reduced by$645 million (29 percent of that study’s esti-mated total expenditures for duodenal ulcer).An estimated $271 million would have been

saved in medical care costs. The estimated $271million savings is the net result of a $34 millionincrease in drug costs, offset ninefold by $305million in savings in other expense categories;the bulk of the $305 million medical care savingsis from estimated reductions in hospital care($258 million) and surgeons’ fees ($30 million).In addition to the $271 million net savings inmedical expenditures, the study estimated that$373 million would have been gained from in-creased productivity—$329 million (88 percent)from decreased morbidity, and the remainderfrom decreased mortality.

The Robinson Associates analysis was basedon two types of estimates. First, physician ex-perts were asked in late 1977 to estimate thelikely clinical and health system effects ofcimetidine compared to traditional therapy forduodenal ulcer patients. Then, applying costfigures derived principally from SRI’s assess-ment of the costs of ulcer disease in the UnitedStates (146), 33 Robinson Associates estimatedthe potential savings in 1977 due to the averagepredicted changes in health status and medicalcare. Summing the results for each cost cat-egory, the analysis yielded the conclusions sum-marized in the preceding paragraph.

A detailed reconstruction of the RobinsonAssociates analysis is beyond the scope of thisreview. Below we provide a description of cer-tain methodologic features of the analysis as abasis for our comments in the critique thatfollows. First, we consider the expert estimatesof the clinical and health system effects of cime-tidine; then, we consider the conversion of theseestimates into projected annual savings.

50 ● Background Paper #2. Case Studies of Medical Technologies

Twenty-three physicians who were familiarwith cimetidine served as expert consultants forthe Robinson Associates analysis. Each wasasked in an interview about five specified ulcer-patient types—ranging from a patient withtypical symptoms and newly diagnosed ulcer toa patient hospitalized with the complication ofbleeding but not requiring immediate surgery—which were intended to represent a spectrum ofseverity of illness in patients with duodenalulcer. The 23 physicians first estimated the pro-portion of all ulcer patients represented by eachtype. Then they described what they believed tobe the usual treatment for each patient type andany changes in that treatment that would bemade in a program that would include cimeti-dine. Finally, they estimated clinical and healthsystem effects (including such elements as re-currence, physician visits, hospitalizations,surgery, diagnostic tests, missed work, com-plications, and mortality) for each type of pa-tient treated without cimetidine and with up to 8weeks of cimetidine. The physicians were notasked to estimate dollar costs for any postulatedeffects.

The methods that were used in the RobinsonAssociates study to convert estimated clinicaland health system effects into dollar savingswere as follows. As a baseline for estimating thecosts of duodenal ulcers, Robinson Associatesused estimates of the costs of peptic ulcer diseasein 1977 developed by SRI (146). Within eachdirect and indirect cost category, the analystsestimated from secondary sources the fractionof peptic ulcer costs attributable to duodenalulcer disease (with the remainder attributable togastric ulcer).

To estimate the proportion of costs thatwould be saved by cimetidine, the analysts thenproceeded as follows. First, they assigned pricesto each component of each cost category. Forexample, in the category of hospital costs, theyestablished the cost per hospital day for patientswho do not have surgery and the cost per hos-pitalization for patients who do undergosurgery. Next, the analysts combined these costcomponents with physician estimates of themanagement and course of patients with andwithout cimetidine. For example, the hospital

costs for each type of patient were calculated asthe sum of costs for the proportion of those pa-tients given surgery plus costs for those notgiven surgery, as follows:

proportion of patients of this type who are hospitalizedx [(proportion of patients hospitalized without surgery

x number of nonsurgical hospitalizations per yearx average length of stay for nonsurgical patientsx cost per day)+ (proportion of patients hospitalized with surgeryx cost per surgical admission)]

= total annual hospital cost per patient of this type

Similar calculations yielded an estimate of theannual costs in each cost category for each typeof patient with and without cimetidine.

Next, the percentage difference in costs ineach cost category due to cimetidine was calcu-lated for each type of patient. The percentagechange for each type of patient was thenweighted by the proportion of all ulcer patientsestimated by the physicians to be represented bythat type. This yielded the percentage change ina given cost category for an “average” patienttreated with cimetidine rather than traditionaltherapy.

The percentage change in cost per “average”patient was further adjusted to reflect: theposited extent of use of cimetidine. The physi-cian consultants predicted that an average of 80percent of the five patient types would betreated with cimetidine 2 to 3 years after thetime of the interviews. This figure, used as thebasis for dollar projections in RobinsonAssociates study’s conclusion, is a composite ofthe estimated extent of use (ranging from 73 to93 percent) for each of the five types of patients.

Next, the percentage change in cost for eachcost category, as adjusted for the proportion ofpatients using cimetidine, was multiplied bythe costs of duodenal ulcer disease assigned tothat category. This provided a dollar estimate ofsavings in each category. Summing the dollarestimates over all cost categories yielded thetotal projected savings attributable to the use ofcimetidine by a given proportion of patients.

In short, the Robinson Associates analysisused prices of the components of medical careonly as a basis for estimating the percentage

change in cost due to cimetidine. Once derived,these percentages were applied to independentassessments of the cost burden of all duodenalulcer disease to estimate dollar savings.

Critique

The cost-and-benefit analysis that RobinsonAssociates prepared for Smith Kline & Frenchhas many positive attributes. First, the studyrepresents the kind of serious analysis of theeconomic effects of a new drug that is importantand valuable. If society is to attend to both theeconomic and clinical implications of medicalinterventions, careful analyses of costs andbenefits are essential. Second, we believe theanalysts selected appropriate categories ofresource costs to assess. Their direct cost com-ponents correspond roughly to the hea l thsystem effects outlined in the benefit-and-costmodel we presented earlier in this case study.Their translation of mortality and morbiditycomponents into indirect costs is appropriatefor a resource cost analysis. Third, their ap-proach of comparing estimated net resource ef-fects of cimetidine to resource use withoutcimetidine is a reasonable one. Fourth, theirmethod of obtaining physician estimates ofclinical and health system effects was an im-aginative one, and it required no guesses aboutcosts from clinicians. Finally, the report pro-vides sufficient detail about its methods andassumptions to allow the reader to reach in-dependent conclusions.

We believe this report deserves scrutiny, be-cause, to our knowledge, it is the most com-prehensive analysis of the resource implicationsof cimetidine in the United States. As athorough economic assessment of a recently in-troduced drug, the study may serve as a modelfor future evaluations of other emerging medicalpractices. In the discussion of the study thatfollows, we have attempted to examine theanalysis carefully in light of the benefit-and-costmodel and data presented earlier and the guide -lines for review of benefit-and-cost analyses thatare presented in the next section of this casestudy.

We believe the Robinson Associates studysubstantially overestimates expected savings

from cimetidine. The accuracy of the estimatedsavings attributable to cimetidine in the Robin-son Associates study depends on at least fivefeatures: 1) the accuracy of the clinical andhealth system effects projected by their physi-cian experts; 2) the relation between a percent-age reduction in health services devoted to ulcerdisease and savings in health resources; 3) theaccuracy of the estimated total costs of allduodenal ulcer disease used as a baseline forpercentage savings; 4) the applicability of pro-jected percentage effects to the total populationof patients with duodenal ulcer disease; and 5)the validity of the methods used to computeaverage percentage effects due to cimetidine. Wequestion some of the assumptions and methodsused in each of these five areas.

Let us consider first the physician experts’opinions of the clinical courses of patients withand without cimetidine. The mean of theseestimates is intended to represent an unbiasedestimate of the course of duodenal ulcer diseaseusing conventional treatment, and an unbiasedestimate of the effects of cimetidine. An unbias-ed estimate of the former is best achieved byphysicians of varied specialty backgrounds whotogether treat the full range of patients withulcer disease, An unbiased estimate of the latterrequires both knowledge of cimetidine’s clinicaleffects and a neutral attitude toward the drug.

The 23 physicians whose opinions form thebasis of the Robinson Associates study were allgastroenterologist-researchers who had partici-pated in early clinical trials of cimetidine andwhose participation in this study was solicitedby Smith Kline & French (121). This selection,the authors state, ensured informed opinionabout the potential effects of cimetidine—but itdoes not ensure individual objectivity or abalanced range of views. Of 32 physicians con-tacted by Smith Kline & French to participate inthe study, 4 refused either because they were toobusy or for unknown reasons.34 It is possible

“Four others were disqualified or unavailable because of exten-sive travel. One of the 24 physicians who agreed to participate wasnot interviewed because of illness (121).

that researchers who were less enthusiasticabout the drug were less eager to express theirviews when contacted by the manufacturer.

To enhance the credibility of subjective physi-cian estimates, Robinson Associates cite aDanish study (66) that compared observed ex-perience in 154 patients over 13 years with phy-sician estimates of some of the long-term con-sequences of ulcer disease (proportion treatedsurgically and proportion of medically treatedpatients with varying degrees of symptoms).The Danish study found that the mean estimatesof 143 physicians corresponded fairly closely topatient experience. The Danish investigatorsinterviewed a wide range of general practi-tioners, medical specialists, and surgeons to ob-tain their mean estimates. These investigatorsalso noted that there were some systematicbiases that tended to balance one another. Forexample, the 65 general practitioners in theDanish study estimated that 15 percent of pa-tients would undergo operations for ulcers, andthe 50 surgeons predicted 27 percent; theobserved proportion was 22 percent. Thus, thisstudy suggests the importance of using a broad-ly based sample to achieve unbiased meanestimates. Just as surgeons’ estimates alonemight not accurately represent surgery experi-ence, a group of research gastroenterologistsseems unlikely to represent a fair cross-sectionof physician experience with and expectationsfor patients who have ulcer disease.

The effects estimated by the physician con-sultants in the Robinson Associates study variedwidely. The projected cost consequences ofusing cimetidine in 100 percent of duodenalulcer patients ranged from a savings of 67 per-cent based on one physician’s estimates to an in-creased expenditure of 40 percent based onanother’s estimates. Seven of the physiciansprojected effects that yielded net losses or smallsavings (of less than 10 percent), while eightphysicians projected effects that led to savingsof 40 percent or more. If the selection wasbiased in favor of physicians at the “optimistic”end of the spectrum of clinical and health sys-tem effects of cimetidine, the mean cost savingsestimate will be similarly biased.

Cost savings in the Robinson Associatesstudy are estimated as a proportion of the totalcosts of duodenal ulcer disease. Two aspects ofthe Robinson Associates calculations deservecomment, and we expand on these points be-low. First, a given percentage reduction inhealth services requirements for a particulardisease probably does not convert directly to anequivalent proportion of health resourcesavings. Second, we believe that the baselinecosts of duodenal ulcer disease employed byRobinson Associates are too large, primarilybecause of an inflated indirect cost estimate.

An implicit assumption in applying a percent-age cost reduction to the health system expend-itures for ulcer disease is that savings will berealized in direct proportion to the decreaseduse of medical services. For example, if hospitaldays decline by 10 percent, then 10 percent ofresources devoted to hospital care are assumedto be saved. This calculation uses average costsper hospital day rather than marginal costs ofthe last 10 percent of hospital days. To the ex-tent that fixed and semivariable costs contributeto the cost of a hospital day, the marginal sav-ings from reducing a given fraction of hospitaldays will be less than the average cost of thosedays.35 The remaining fixed cost componentswill simply be redistributed over the remaininghospitalized patients, Thus, the direct conver-sion of percentage reduction in hospital days topercentage savings in resource costs of hospitalcare may be questioned.36 Short-term resourcesavings might even be less than the avertedmarginal costs, insofar as available supply ofhospital resources induces other demand.37 Ifhospital beds previously occupied by patientswith ulcer disease are filled by other patients(without ulcer disease) who previously wouldnot have been hospitalized, then potential sav-ings would be eroded further.

“Fixed c[~sts are independent ot the volume t~f services. Semi-variab]e costs are a t unct i~)n (>f both time and volume of services.

“A related problem is the trequent use of charges as pr[~xies forresource costs t~t care. Charges rel]ect average rather than mar-ginal costs, and tor a variety of reasons, charges for particul~lr ser-vices may differ trom their average rew~urce c[~sts.

“The notion ot hospital bed supply creating demand for mc~rehospital services, called Rt)emer’s Law, was t)riginally proposed 20yedrs ago ( 130).

The total costs of duodenal ulcer disease usedby Robinson Associates are based on the esti-mated costs of peptic ulcer developed by SRI(146). SRI’s estimate is substantially higher thananother recent, independent estimate of the costof ulcer disease by NCDD (4), and, as we dis-cussed earlier, we believe a more correct figurelies between the two. If Robinson Associateshad based their projected savings from cimeti-dine use on the costs of ulcer disease asestimated by NCDD, making no other changesin their analysis, the resulting estimated savingswould have been over 50 percent less. Use ofNCDD’s cost figures, without altering any otherassumption or calculation used in the RobinsonAssociates study, would have produced an es-timated savings of only $307 million, in contrastto the $645 million savings projected on thebasis of SRI’s figures. Use of our midpoint cal-culation developed in the section of this casestudy on the cost of peptic ulcer disease yieldsestimated savings of only $476 million.

Another important source of misestimation inthe Robinson Associates study is the assumptionthat the five patient types represent the fullrange of patients with duodenal ulcer disease.The most severely ill type of patient included inthe Robinson Associates study is one who ishospitalized and bleeding but not in need of im-mediate surgery. Thus, the study omits patientswho have very severe bleeding or other life-threatening complications of ulcer disease suchas perforation. According to CPHA data (42),nearly 6 percent of patients hospitalized forduodenal ulcer disease in 1977 had perforation,and 28 percent had bleeding. The number of ex-cluded patients who require prompt surgerymay be estimated conservatively to include 90percent of patients with perforation (or 5 per-cent of hospitalized patients) and between 10and 20 percent of patients admitted for bleeding(or an additional 4 percent of hospitalized pa-tients). Thus, approximately 9 percent ofhospitalized patients, all of whom receivesurgery, are excluded from the range of patientsin this study.

The omission of these patients from theRobinson Associates study has substantial con-sequences for the study’s cost estimates. For ex-

ample, consider the area of hospital costs alone.Assuming traditional therapy, Robinson As-sociates estimate total hospital costs to be $732million. At 80-percent cimetidine use, they es-timate savings in hospital costs to be $258 mil-lion, a 35-percent reduction from hospital costswith traditional therapy. According to the SRIfigures that served as a baseline for the Robin-son Associates estimates, nearly 72 percent ofhospital costs for ulcer patients in 1977 were dueto the estimated 20 percent of hospitalized pa-tients who underwent surgery (146). Assumingthe excluded patients, who are most severely ill,were responsible only for a proportionate shareof costs for surgical cases, the proportion oftotal hospital costs for duodenal ulcer diseasedevoted to these patients would be approx-imately 32 percent, and the dollar amount de-voted to their care would be $237 million .38

Although the expert consultants were notasked about this group of most severely ill pa-tients, we think that cimetidine would not havebeen expected to alter the acute management ofmore than a small fraction of them. Assumingthat 80-percent cimetidine use would have beenestimated to save as much as 15 percent (ap-proximately $36 million) of the hospital costsfor these patients, and then applying the propor-tion of savings estimated for “all” duodenalulcer patients in the Robinson Associates studyto the hospital costs attributable only to the in-cluded patients, we compute the savings in hos-pital care to be $209 million rather than $258million. 39

‘pProp(~rtlc~n of hospital costs due to surgical care (~t excludedpatients = proportion of surgical cases excluded x prc)pc>rt](>n oftotal costs due to surgical cases: 0.324 = (0.09 0.20) x 0,72.Dollar amf~unt devoted to hospital care of excluded patients =proporticln of hospital costs due to excluded patients x t[>tal hos-pita] costs: $237 mi]]ion = 0.324 X $732 mi]]i{}n,

“Let :Cs estimated hospital costs due to all pat]ents with

traditional therapyCE ~ estimated hospital costs due tcl excluded patients with

traditi<>nal therapycl ~ estimated h(}spital costs due t o included patients with

traditional therapySE estimated hosplta] savings from all patients with

80-percent cimetldine usePSE ~ estimated pr[lp(~rt ion [~t costs saved by 80-percent

cimet idlne use a t tribu table to excluded patientsPSI ~ estimated prtlporti(~n ot costs saved by 80-percent

clmetid]ne use a t tribu table t o Included patients

54 ● Background Paper #2: Case Studies of Medical Technologies

Thus, the incomplete spectrum of patients in-cluded in the study produces an overestimate insavings of nearly $50 million in the area ofhospital costs alone. The exclusion of the mostseverely ill patients also incurs additional, ifsmaller, overestimates of savings in other costareas. If cimetidine reduces the fraction of pa-tients who reach the most severely ill category,however, then this source of overestimationwould be reduced proportionately. The authorsof the Robinson Associates report repeatedlypoint out that their projected percentage savingsare unaffected by changes in estimated baselinecosts for ulcer disease. As we have just seenabove, however, the percentage savings cal-culated in the study are quite sensitive to the in-clusion or exclusion of different types of pa-tients with duodenal ulcer.

Two other sets of assumptions in the Robin-son Associates study also affect the calculatedsavings. The first of these is the relative dollarvalues assigned to the components of each costcategory (e.g., how much less expensive is ahospital stay for nonsurgical than for surgicalpatients?). The second set of assumptions is theestimated proportion of all included patients ineach of the five patient types. The data underly-ing these assumptions can be expected to varyover some range. To accommodate such varia-tion, one could, for example, estimate a con-fidence interval about physician estimates of theproportion of patients of each type. 40 T h eRobinson Associates study would have beenstrengthened by explicit sensitivity analysis,testing the effect on the conclusions ofsystematic alteration of key assumptions.

(continued from p. 53)Given:c - $732 million (from Robinson Associates study)C E - $237 million (from preceding footnote)CI - C – CE = $(732 – 237 million) = $495 million (by

definition, C = CI + CE)PSE - 0.15 (assumption; see text)P S I - 0.35 (from Robinson Associates study)Then:S - CE x psE + CI x PS I

- ($237 million) (0.15) + ($495 million) (0.35)= $209 million.

d’JT~iS is separate from the question of bias in the mean estimate,i.e., whether a group of gastroenterologist-researchers wouldperceive the world of ulcer patients to be made up of as high a pro-portion of “initial diagnosis patients” (type 1) as would a group ofgeneral practitioners or less specialized internists.

The method used by Robinson Associates tocompute the expected reduction in costs causedby cimetidine use has another subtle, but po-tent, effect on their estimate. Assume for themoment that the interviewed physicians didconstitute a representative sample of informedopinion about the effects of cimetidine. It wouldbe desirable, then, for the overall estimated per-centage reduction in costs to be a statistically

unbiased measure of individually perceived per-centage reductions. Take a simplified case. Ifphysician A provides estimates of cimetidine’seffects that produce a 70-percent decrease inresource consumption, and physician B pro-vides estimates that produce a 50-percentdecrease, we would like the overall estimatedreduction to be midway between the two, or 60percent. Since we presumably trust each physi-cian’s judgment equally, each perceived percent-age reduction should contribute equally to theoverall estimate of percentage reduction. How-ever, the method used by Robinson Associatesto compute percentage reduction in costs has theeffect of placing greater weight on the percent-age reduction estimates of physicians whoperceive ulcer disease as more severe and requir-ing higher levels of resources.

Mathematically speaking, this distortion oc-curs because the ratio of estimated means is notthe same as the mean of estimated ratios. To seehow this distortion can arise, again considertwo simplified examples. First, physician A andphysician B are asked about the consequences of. -ulcer disease with and without cimetidine for agiven type of patient. The effects are translatedinto various categories of resource cost, such ashospital care. Physician A estimates effects thatlead to a total annual cost of $1,000 withoutcimetidine and $500 with cimetidine use. Physi-cian B estimates effects that lead to a cost of$100 without cimetidine and $50 with the drug.In each case, the estimated percentage reductionis 50 percent . Proceeding as RobinsonAssociates did, we can compute an average costwithout cimetidine and an average cost withcimetidine.

$1,000 + $100average cost without = = $550cimetidine 2

$500 + $50average cost with = = $275cimetidine 2

Then the “average” percentage reduction at-tributed to cimetidine, as computed by Robin-son Associates, would be the difference betweenthese average costs divided by the cost withoutcimetidine, or:

$550 – $275 = 0 . 5 0$550

In this case, both physicians projected the samepercentage reduction, and the calculated percen-tage reduction agrees with both of them. So far,this approach appears sound.

Now consider the following variation. Physi-cian A estimates effects that cost $1,000 withoutcimetidine and $400 with cimetidine, a 60-per-cent reduction in costs. Physician B estimates ef-fects that lead to a cost of $100 without cimeti-dine and $60 with the drug, a 40-percent re-duction. The average estimated reduction is:

0.60 + 0.40 = 0.502

or 50 percent. Calculated by the method ofRobinson Associates, the percentage reductionIS:

(51,000 + $100) – ($400 + $60)

$1,000 + $100

$1,100 – $460=

$1,100

$640= — = 0 . 5 8

$1,100

Thus, the calculated reduction of 58 percent ismuch closer to the perceived reduction of physi-cian A, who viewed ulcer disease in this type ofpatient as more severe and costly than didphysician B.41

~ i sY ~b{)] iC~ ] Iv the ~1 I [~~(,nce beth,een the rncttl {~cj in effect used

by Rc&lnwln A’s~tlcidtes to calculdk’ a ‘ mean” percentage costreduction and the mean of the percentage reducti(>ns estimated by

the phy’>lcl.ln+ c a n be expres>ed a~ lt~ll[~w~—— number t~t physician expert%

;Cl ; c(wt ca]cu Iated }rom physlclan t est tmates ok effectsWI th tradl tlt~ndl tred tment ~ wlth(w t clmet]dlne)

cc, = C(M t cd ]CU [a ted [ r(~m phv~lclan I est I ma tes of et[ec ts—W. I t h d ] ] pat len t~ reccl v ln~ c i met i d I ne treatment

Then :TC1 _ ~~1 = c{)~t ~al,lng+ ~d]~ul~ted tr~,rn ~hy,~lclan I estlma te~

TCI – CCI = percentage c(wt reduc t](~n calculated trorn physicianTCI i e~t t ma te~

The estimation method used by RobinsonAssociates confounds the estimate of perceivedeffects of cimetidine, on the one hand, withvariability in the perceived severity and overallmanagement of ulcer disease, on the other. If theRobinson Associates study had taken the meanof physician estimates with traditional therapyas the baseline from which percentage reduc-tions were calculated, there might be a strongercase for an approach like that used. However,the calculated percentage reductions were ap-plied to an independently determined baselinecost. This reinforces the argument for seekingan unbiased measure of expected percentagereduction, namely the mean of the physicians’percentage estimates.

The practical consequences of this distortionare substantial. A series of bar graphs providedin the Robinson Associates report (their tables34 through 43, pp. 54-63) shows percentagechanges in cost based on the estimates of each ofthe 23 physician informants. A separate figurein the report depicts the distribution of physi-cian percentage estimates for each of nine costcategories and overall costs. On the basis of thebar graph for the distribution of physicianestimates of overall cost savings, we calculatethat the mean of the estimated percentage reduc-tions by the 23 physicians was approximately 24percent (see table 20). By contrast, the “mean”shown in the Robinson Associates bar graphand used in thereduction of 34dine use).

This suggestsreduction used

analysis was a percentage costpercent (at 100-percent cimeti-

that the “average” percentagein the Robinson Associates

Robinson Associates estimates “mean” percentage cost savings asan “average of total costs”:

i = 1 x 100

1=1The mean of percentage savings estimated by physicians is:

(1)

(2)n

In general, eqn. 1 # eqn. 2, although the two may give the sameresult in exceptional circumstances.

Table 20.—Percentage Cost Savings EstimatedFrom Robinson Associates Study

Physician number Percentage savingsa

1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68%2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506 487: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 478 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429 36

10: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 3511. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2812. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2713. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1914. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1315. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1216. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1117. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.518. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.519. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o21. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 9b

22. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 2 4b

23. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 4 0b

Total . . . . . . . . . . . . . . . . . . . . . . . . . . 562%Mean percentage cost savings = 562/23=24.4%

aGauged from the height of the bar graph for each physlclan In RobinsonAssoclatestable43

bEstlmated c os t i n c r e as e

SOURCE Robinson Associates, Inc, The /mpacfof C/met/d/neon the Naf/ona/

CosfofDuoderra/U/cers, 1978(121)

analysis was approximately 42 percent largerthan the mean of the estimated percentagereductions provided by the physician experts.42

(Similar discrepancies of varying degree arefound with each category of cost shown in theRobinson Associates bar graphs.) Basing overallpercentage cost savings on the mean of percent-age reductions estimated by the 23 physicianswould reduce the projected savings at 80-per-cent cimetidine use by approximatel y $ 1 9 0million, from $645 million to approximately$455 million.

In summary, we believe the Robinson Associ-ates analysis substantially overestimates ex-pected savings from cimetidine. Considering theexaggerated baseline costs of ulcer diseaseassumed in the analysis,trum of patients included,troduced by the method

42 0.34 –0.24= 0.416

0.24

the incomplete spec-and the distortion in-of calculating mean

percentage reduction in costs we believe theestimated $645 million savings are probably twoto three times too large. Potential bias intro-duced by the selection of physician informantswould increase the magnitude of that over-estimate.

Despite our criticisms of the Robinson Associ-ates study, we believe its basic thrust is prob-ably correct. Cimetidine does appear to savemore medical resources than it costs. The $305million savings in medical costs that RobinsonAssociates estimates from the use of cimetidineare approximately nine times the estimated $34million direct costs of the drug. Thus, even if thedrug costs were tripled and the estimated sav-ings reduced by two-thirds, use of cimetidinewould still appear to be an economically soundinvestment. Also, the estimated savings inhealth resources omit potential gains in produc-tivity from use of the drug.

The emerging empirical evidence cited in thesection of this case study on health system ef-fects supports the belief that use of cimetidineprobably saves medical resources. In the comingyears, more evidence will probably accumulateabout the costs, risks, and benefits of cimetidinecompared to alternatives. We may learn, for ex-ample, about newly recognized adverse effectsof the drug or about rebound in the number ofulcer patients undergoing surgery or about thedevelopment of safer, equally effective and ac-ceptable treatments. The comparative cost effec-tiveness of cimetidine for patients with ulcerdisease in the long run is a matter of continuing

empirical study,

Our discussion of the Robinson Associatesstudy illustrates some of the difficulties of de-signing and conducting economic analyses ofnewly introduced medical practices. The workof the Robinson Associates analysts must beviewed in the context of the information avail-able at the time it was done. The RobinsonAssociates study was undertaken before therewas widespread clinical experience with cimeti-dine, and the analysts faced a dearth of em-pirical findings relating directl y to resourcecosts. Given the information available at thattime, the analysts might have considered the fol-lowing procedure. First, define prototypical pa-

-

tients that represent the full range of patientswith ulcer disease, including the most severelyill. Second, obtain from a broad, representativegroup of physicians baseline estimates of thecourse of disease using traditional therapy.Third, check how closely these estimates corre-spond to other estimates of the total cost ofduodenal ulcer disease, examine criticallyassumptions that underlie the estimates of thehealth system effects in each major cost cate-gory, and reach consensus estimates. Fourth,present to physicians familiar with cimetidinethe consensus-estimated clinical courses for eachpatient type with traditional therapy; ask themto assume the consensus represents actual pa-tient experience; and then ask them to estimatewhat changes, if any, would follow from the in-troduction of cimetidine. Finally, calculate themean of the estimated percentage cost reduc-tions and apply it to appropriately estimatedcosts of illness.

Guidelines for Review of Health CareCost Analyses

Presented below are guidelines in the form ofa series of questions that may aid in the designand review of cost-benefit and cost-effectivenessstudies. These guidelines cover matters ofdefinition and purpose, analytic methods, andconclusions. They are presented here in conciseform and presume familiarity with the rationaleand basic components of benefit-and-costanalyses in health care (see, e.g., 89).

Objectives of the Analysis

1.

2.

What is the purpose of the analysis? Is ita) to assess the optimal management of in-dividual patients with a particular clinicalcondition; b) to measure the clinical andeconomic importance of particular clinicalproblems; c) to compare alternative strate-gies for addressing a particular health prob-lem in a particular population; d) to com-pare alternative investments in health pro-grams; or e) to compare health and othersocial resource investments?Are the interests and potential biases of theanalyst and client acknowledged? Are meas-ures taken to guard against potential bias?

Specification of the Problem

1.

2.

3.

Is the population of interest appropriatelydefined (e.g., a population with a particulardiagnosis, or having a particular clinicalsymptom, or undergoing a particular test ortreatment)? Is the population consistentlydefined throughout the analysis?Does the analysis specify the interventionsof interest and address them consistentlythroughout the analysis?Are the conditions of use (e.g., ideal v.average) specified and consistently treated inthe analysis?

Methods of Analysis

1.

2.

3.

4.

5.

6.

7.

Are the analytic methods selected appropri-ate to the objectives of analysis? (e. g., CEAv. BCA, use of decision-analytic framework,etc. )Is the time frame of analysis appropriate tothe objectives (e. g., is patient lifetime a moresuitable focus than a cross-section of pa-tients for a limited time?)Are clinical effects and other benefits appro-priately specified? Are the methods of as-sessment explained? Are incremental bene-fits the basis for analysis?Are cost estimates complete and appro-priately categorized? Has double countingbeen avoided? Are induced costs and savingsconsidered? Are marginal resources costs thebasis for analysis? Are methods fully ex-plained?Are benefits and costs aggregated properlyacross the population and intervention of in-terest? Is the analysis restricted to a few usesof multipurpose intervention?Are benefits and costs appropriately aggre-gated over time? Is discounting employed? Isthe discount rate appropriate?Are projected effects justified? Are theestimates based on empirical data or opin-ion? Are uncertainties recognized? Are thesources of all estimates clearly explained?Are estimates unbiased? Are assumptionsacknowledged, fully exposed, and justified?Are estimates based on evidence from thesame population and intervention that arethe subjects of analysis? Are extrapolationsand interpolations reasonable?

8.

9.

10.

Are underlying trends in disease distribution Conclusionsand severity taken into account? Are other 1.pertinent population trends assessed?Are technological changes and evolution of 2.practices taken into account?Are the distributions of benefits and costsimportant and are they considered?

Are the conclusions consistent with the anal-ysis and appropriately qualified?Are conclusions robust? Have assumptionsand key uncertainties been subjected to asensitivity analysis?

SUGGESTIONS FOR FURTHER RESEARCH

Our suggestions for further research arecognizant of broadened clinical experience withcimetidine in recent years and newly emergingempirical evidence of health system effects ofthe drug. Any analysis must take account of theshifting epidemiologic pattern of ulcer disease.We believe that sound CEAs comparing effectsof different strategies over an ulcer patient’slifetime would provide valuable guidance tomedical decisions for patients with ulcer disease.

As discussed earlier in this case study, anumber of well-controlled clinical trials haveassessed the clinical effects of cimetidine in thepast few years. These trials have not dealt withevery important clinical comparison (e.g.,maintenance antacids v. cimetidine), but theyhave provided a much sounder empirical basefor projecting some of the clinical consequencesof the drug. Since clinical use of cimetidine is sowidespread, a broad group of clinicians couldbe consulted for subjective estimates of likelyconsequences in areas where clinical trials arelacking. It would be possible to begin by send-ing participating physicians a summary of em-pirical clinical findings with cimetidine. Asystematic method, such as a Delphi process,could then be used to reach group consensus onkey probabilities .4:

There is still little empirical information onhealth system effects of cimetidine to serve as abasis for estimating resource costs and savings.Any analysis of the health system effects of anintervention in ulcer disease must take as a

43The Delphi method is an iterative process for reaching groupconsensus that attempts to separate the reasons for an opinionfrom the authority of the opinionholder. It has been employed inthe analysis of health care decisions to obtain estimates of prob-abilities from a group of medical experts (127).

baseline the epidemiologic trends in the diseasediscussed in this case study. As mentionedpreviously, the rate of surgery for ulcer diseaseshows a steeper decline in 1978 than would bepredicted from the previous trend. Confirma-tion of this drop and additional evidence linkingit to cimetidine would provide a sounder basisfor projecting direct cost savings in one area,and for attributing such savings to the use ofcimetidine. Additional evidence of health sys-tem effects of cimetidine may emerge from theongoing research that we have described.

In further analyses of cimetidine, one fun-damental concern must be the time frame of theanalysis. Calculations based on a single year,for example, will overlook the important dis-tinction between avoidance of surgery and delayin surgery. From the evidence we have cited inthis case study, there is a good reason to believethat a year of maintenance cimetidine imparts adelay in inevitable surgery rather than a long-lasting cure of ulcer disease,

More generally, with diseases such as pepticulcer, which are chronic, and with interventionssuch as cimetidine, whose long-term effects maybe very important, a benefit-and-cost analysismight best focus on a cohort of patients, project-ing effects over their lifetimes. Rather than at-tempt to enumerate all resource implications fora cross-section of the population in a singleyear, it might be more helpful to estimate thepresent-value lifetime resource costs and healtheffects for a given population of ulcer patients,Then, on the basis of available research evi-dence and subjective clinical judgments, onecould estimate the consequences for a given typeof patient of pursuing different managementstrategies. This approach would allow compari-

son of marginal costs with incremental benefitsfor each shift in strategy. Focusing on each pa-tient’s lifetime rather than on a given calendaryear would highlight, among other things, un-certainties in the long-term side effects of use ofa relatively new agent such as cimetidine.

The prototype decision tree shown in figure 4(P P. 26-27) could serve as a starting point forsuch an analysis. At least one group of research-ers, in Switzerland, has begun to take a de-cision-analytic approach to the lifetime conse-quences of duodenal ulcer treatments (137).Their published analysis is restricted to aspecific 50-year-old male patient, and one mightfault their figures for not discounting futurecosts and benefits, but these researchers do ap-proach the problem in an appropriate way.

REFERENCES

1.

2.

3.

4.

5.

6.

7.

Acton, J. P., Evaluating Public Programs ToSave Lives–The Case of Heart Attacks, Randreport R-950-RC (Santa Monica, Calif.: RandCorp., 1973).Almy, T. P., Dartmouth Medical School, Han-over, N. H., personal communication, 1980.Almy, T. P., et al., “Panel I: Prevalence andSignificance of Digestive Disease, ” Gastroenty.68:1351, 1975.

“Report of the Workgroup on the So-cioeconomic Impact of Digestive Diseases ofthe Subcommittee on Epidemiology and Im-pact, ” in Report to the Congress of the UnitedStates of the National Commission on Diges-tive Diseases, VOI. 4, pt. 4, DHEW publicationNo. (NIH) 79-1885 (Bethesda, Md.: NationalInstitutes of Health, 1979).Bader, J. P., et al., “Treatment of Gastric Ulcerby Cimetidine: A Multicentre Trial, ” in Cime-tidine —Proceedings of t h e S e c o n d Interna-tional Symposium on Histamine Hz-ReceptorAntagonists, W. L. Burland and M. A. Sim-kins (eds. ) (Amsterdam-Oxford: Excerpta Med-ica, 1977).Bank, S., et al., “Histamine H,-Receptor An-tagonists in the Treatment of DuodenalUlcers, ” S. Afr. Med. }. 50:1781, 1976.Bardhan, K. D., “Cimetidine in Duodenal Ul-aeration, ” i n Cimetidine: The WestminsterHospital Symposium, 1978, C. Wastell and P.

If the object of analysis is to help inform clini-cians and health policy decisionmakers aboutthe efficient use of resources in the care of pa-tients with ulcer disease, then the most helpfulapproach would be to do a CEA, oriented toparticular groups of patients, comparing incre-mental clinical benefits, in terms of morbidityand mortality, to marginal resource costs. Anassessment of the cost effectiveness of differentmanagement strategies in all patients with ulcerdisease would be an enormous undertaking, butanalyses do not have to be global to be validand useful. As long as their limitations areunderstood and taken into account, such studiescan aid both health policy formulation andmedical decisionmaking.

8.

9.

10.

11.

12.

13.

14.

15.

16.

Lance (eds. ) (New York: Churchill Livingstone,1978).Bardhan, K. D., et al., “Comparison of TwoDoses of Cimetidine and Placebo in the Treat-ment of Duodenal Ulcer: A Multicentre Trial, ”Gut 20:68, 1979.

“Double-Blind Comparison of Cimeti-dine a;d Placebo in the Maintenance of Heal-ing of Chronic Duodenal Ulceration, ” G u t20:158, 1979.Belber, J. P. “Gastroscopy a n d Duodenosco-py, ” in Gastrointestinal Disease, vol. 1, 2d cd.,M. H. Sleisenger and J. S. Fordtran (eds. ) (Phil-adelphia: W. B. Saunders Co., 1978).Binder, H. J,, et al., “Cimetidine in the Treat-ment of Duodenal Ulcer: A Multicenter Dou-ble-Blind Study, ” Gastroenty. 74:380, 1978.Black, J. W., et al., “Definition and Antag-onism of Histamine Hz-Receptors, ” Nature,236:385, 1972.Blackwood, W. S., et al., “Cimetidine in Duo-denal Ulcer: Controlled Trial, ” Lancet 2:174,1976.Blumenthal, I. S., Research and the Ulcer Prob-lem, Rand report R-336-RC (Santa Monica,Calif,: Rand Corp., June 1959).Bodemar, G., and Walan, A., “Cimetidine inthe Treatment of Active Duodenal and Prepy -loric Ulcers,” Lancet 2:161, 1976.

, “Maintenance Treatment of Recurrent

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27,

28,

29.

30<

P e p t i c U l c e r b y Cimetidine,” Lancet 1:403,1978.Bonnevie, O. E., “The Incidence in Copenha-gen County of Gastric and Duodenal Ulcersinthe Same Patient,” Sc, IGastr. 10:529, 1975.

, “The Incidence of Duodenal Ulcer inC o p e n h a g e n C o u n t y , ” SC. 1. Gastr. 10:385,1975.

, “The Incidence of Gastric Ulcer inCopenhagen County,” SC. J. Gastr-. 10:231,1975.

“Survivali nPepticUlcer,” Gastroerzty.75:1055,1978Brown, R.C., etal., “HospitalA dmissionsforPeptic Ulcer Xring 1958-1972;’ Br. Med. 1.1:35,1976.Burland, W. L . , and Simkins, M. A. (eds.LCimetidine: P~oceedings of the Secorzdlnterna-tional Symposium on Histurnine Hz-ReceptorA~I tago)lists (Amsterdam-Oxford: ExcerptaMedica, 1977).Burland, W. L. , e t a l . , “The Longer TermTreatment of Duodenal Ulcer With Cimeti-dine, ” in Cin~etidi)~e: The Westminster Hos-pital Symposium, 1978 , C. Wastell and P.Lance (eds. ) (New York: Churchill Livingston,1978) .

“Reversal of Metiamide-InducedAgran;locytosis During Treatment With Cime-tidine, ” L.uncet 2:1085, 1975.Chalmers, T. C., et al., “Randomized Con-trolled Trials in Gastroenterology With Par-ticular Attention to Duodenal Ulcer, ” Report tot}le Congress of t}ze U~zited States of the Nation-al Commission on Di.gestit~e Diseases, vol. 4 ,pt. 2B, DHEW publication No. (NIH) 79-1885(Bethesda, Md.: National Institutes of Health,1979).Chang, H. K., and Morrison, S. L., “Bone-Marrow Suppression Associated With Cimeti-dine” (letter), Ann, lnt. Med. 91:580, 1979.Christensen, E;., et al., “Treatment of DuodenalUlcer: Randomized Clinical Trials of a Decade(1964 to 1974),” Gastroenty. 73:1170, 1977.Ciclitira, P. J., et al., “A Controlled Trial ofCimetidine in the Treatment of Gastric Ulcer, ”in Cirtletidine. Proceedirlgs of the Secorld lrlter-nutiorlal Syrnvosium on Histamine H2-Recep-tor Antagonists, edited by W. L. Burland andM. A. Simkins (Amsterdam-Oxford: ExcerptaMedica, 1977).“Cimetidine for Ever (and Ever and Ever. ..)?”(editorial), Br Med. ]. 1:1435, 1978.Cochran, W. G., et al., “Experiments in Surgi-cal Treatmen ! of Duodenal Ulcer, ” in Costs,

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

Risks, and Benefits of Surgery, J. P. Bunker, etal. (eds. ) (New York: Oxford University Press,1977).~ooper, B. S., and Rice, D. P., “The EconomicCost of Illness Revisited, ” Soc. Sec. Buil. 39:21,1976,DeGalocsy, C., and Van Ypersele de Strihou,c., “Pancytopenia With Cimetidine” (letter),Ann. lnt, Med. 90:274, 1979.Delaunois, L . , “Hypersensitivity to Cimeti-dine” (letter), N. Erzg. }. Med. 300:1216, 1979.Department of Health, Education, and Wel-fare; Public Health Service, The .Vatiorza/Ambulatory Medical Care Survey: 1975 Sum-mary, United States, January-December 1975,Vital and Health Statistics Series 13, No. 33,DHEW publication No. (PHS) 78-1784 (Hyatts-ville, Md.: National Center for Health Statis-tics, January 1978).

, Health Services and Mental Health Ad-ministration, Acute Conditions: Incidence andAssociated Disability, United States, ]Llly 1970-)une 2972, Vital and Health Statistics Series 10,No. 82, DHEW publication No. (HSM) 73-1508(Hyattsville, Md,: National Center for HealthStatistics, April 1973).

, Office of Health Research, Statistics,and Technology, Prevalence of SelectedChronic Digestive Conditions, United States,1%’s, Vital and Health Statistics Series 10, No.123, DHEW publication No. (PHS) 79-1558(Hyattsville, Md.: National Center for HealthStatistics, July 1979).Doll, R., “Introductory Remarks, ” in A Syrn-posiurn on Carbenoxolone Sodium, M. Robsonand F. M. Sullivan (eds. ) (London: Butter-worths, 1968).Drasar, B. S., et al., “Studies on the IntestinalFlora, I: The Bacterial Flora of the Gastrointes-tinal Tract in Healthy and Achlorhydric Per-sons, ” Gastroenty. 56:71, 1969.Dunn, D. H., et al., “The Treatment of Hemor-rhagic Gastritis With the Hz-Blocking Antihis-t a m i n e , Cimetidine, ” Gastroenty. 72:1053,1977.Dyck, W. P., et al., “Cimetidine and Placebo inthe Treatment of Benign Gastric Ulcer A Mul-ticenter Double-Blind Study, ” Gastroenty.74:410, 1978.Dykes, P. W., et al., “Treatment of Upper Gas-trointestinal Hemorrhage With Cimetidine, ” inCimetidine: Proceedings of the Second Interna-tional Symposium on Histamine H2-,?eceptorAntagonists, W. L. Burland and M. A. Sim-

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

kins(eds.) (Amsterdam-Oxford: Excerpta Meal-ica, 1977).Elashoff, J. D., and Grossman, M. I., “Trendsin Hospital Admissions and Death Rates forPeptic Ulcer in the United States From 1970 to1978, ” Gastroe~zty. 78:280, 1980.Englert, E., Jr., et al., “Cimetidine, Antacid,and Hospitalization in the Treatment of BenignGastric Ulcer: A Multicenter Double-BlindStudy, ” Gastroe/zty. 74:416, 1978.Farris, J. M., and Greenburg, A. G., “TruncalVagotomy With Drainage for the ElectiveTreatment of Duodenal Ulcer Disease, ” in Co) Z-trol~ersy ill Surgery, R. L. Varco and J. P.Delaney (eds. ) (Philadelphia: W. B. SaundersCo., 1976).Feldman, E. J., and Isenberg, J. I., “Effects ofMetiamide on Gastric Acid Hypersecretion,Steatorrhea, and Bone-Marrow Function in aPatient With Systemic Mastocytosis, ” N. E/lg.]. Med. 295:1178, 1976.Feldman, M., et al., “Effect of Low-Dose Pro-pantheline on Food-Stimulated Gastric AcidSecretion: Comparison With an ‘Optimal Effec-tive Dose’ and Interaction With Cimetidine, ”N. E?I~. ]. Med. 297:1427, 1977.Finkel, M. J., “The FDA’s Classification Systemfor New Drugs: An Evaluation of TherapeuticGain, ” N. E)lg. ]. Med. 302:181, 1980.Finkelstein, W., and Isselbacher, K. J., “Cime-tidine, ” N. ErI~. ]. Med. 299:992, 1978.Fleiss, J. L., Statistical Methods for Rates andProportio}ls (New York: John Wiley & Sons,1973).Food and Drug Administration, Rockville,Md., personal communication, 1979.

Rockville, Md., personal communica-tion, 1980.Fordtran, J. S., and Grossman, M. I., “ThirdSymposium on Histamine H,-Receptor An-tagonists: Clinical Results With Cimetidine, ”Gastroerlty. 74:431, 1978.Fordtran, J. S., et al., “Round Table Discussionon Gastric Ulcer, ” Gastroenty. 74:431, 1978.Freiman, J. A., et al., “The Importance of Beta,Type II Error, and Sample Size in the Interpre-tation of the Randomized Control Trial, ” N.EIIg. /. Med. 299:690, 1978.Freston, J. W., “Cimetidine in the Treatment ofGastric Ulcer: Review and Commentary, ” Gas-troetlty. 74:426, 1978.Frost, F., et al., “Cimetidine in Patients WithGastric Ulcer: A Multicentre Controlled Trial, ”Br. Med. ]. 2:795, 1977.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

Fry, I., “Peptic Ulcer: A Profile, ” Br. Med. ).2:809, 1964.Geweke, J. F., Department of Economics, Uni-versity of Wisconsin, Madison, Wis., personalcommunication, 1979.Gifford, L. M., et al., Medical Affairs Depart-ment, Smith Kline & French Laboratories,Philadelphia, “ ‘Tagamet’ Post-Market Outpa-tient Surveillance Program: Interim Report onPhase I,” unpublished report.Gill, M. J., and Saunders, J. B., “Perforation ofChronic Peptic Ulcers After Cimetidine” (let-ter), Br. Med. ]. 2:1149, 1977.Gillespie, G., et al., “Short-Term and Mainte-nance Cimetidine Treatment in Severe Duoden-al Ulceration, ” in Cimetidine: Proceedings oftl~e Seco~zd lnternatio~zal Symposium on Hista-tni?ze Hz-Receptor Antagonists, W. L. Burlandand M. A. Simkins (eds. ) (Amsterdam-Oxford:Excerpta Medica, 1977).Gillies, R. R., et al., “Controlled Comparisonof Two Dosage Regimens of Cimetidine inDuodenal Ulcer: A Multicenter Study, ” Gas-troe~lty. 74;396, 1978.Gray, G. M., “Peptic Ulcer Disease, ” in Medi-cine, E. Rubenstein and D. D. Federman (eds. )(New York: Scientific American, October1979).Gray, G. R., et al., “Long-Term Cimetidine inthe Management of Severe Duodenal UlcerDyspepsia, ” Gastroenty. 74:397, 1978.

, “Oral Cimetidine in Severe DuodenalUlceration, ” L.ancet 1:4, 1977.Griebe, J., “Long-Term Prognosis of DuodenalUlcer: Follow-Up Study and Survey of Doctors’Estimates, ” Br. Med. ). 2:1572, 1977.Grossman, M. I., Center for Ulcer Researchand Education, Veterans Administration Cen-ter, Los Angeles, personal communication,1980.

“Closing Remarks (Cimetidine Sym-podium’), ” Gastroenty. 74:487, 1978.

“Resume and Comment. Chapter 10,Veterans Administration Co-Operative Studyon Gastric Ulcer, ” Gastroenty. 61:635, 1971.Gudmand-H@yer, E., et al., “Prophylactic Ef-fect of Cimetidine in Duodenal Ulcer Disease, ”Br. Med. ]. 1:1095, 1978.Gugler, R., et al., “Cimetidine for AnastomoticUlcers After Partial Gastrectomy: A Random-ized Clinical Trial, ” N. Eng. J. Med. 301:1077,1979.Hall, W. H., “Breast Changes in Males onCimetidine, ” N. Eng. ). Med. 295:841, 1976.

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

Hallenbeck, G. A., “Proximal Gastric Vagot-omy Without Drainage for Duodenal Ulcer, ” inControversy in Surgery, R. L. Varco and J. P.Delmey (eds. ) (Philadelphia: W. B. SaundersCo., 1976).Henn, R. M., et al., “Inhibition of Gastric AcidSecretion by Cirnetidine in Patients With Duo-denal Ulcer,” N. f%g. ]. Med. 293:371, 1975.Herbst, A. L., et al., “Adenocarcinoma of theVagina: Association of Maternal StilbestrolTherapy With Tumor Appearance in YoungWomen, ” N. Eng. ]. Med. 284:878, 1971.Hetzel, D. J., et al., “Cimetidine Treatment ofDuodenal Ulceration: Short-Term Clinical Tri-al and Maintenance Study, ” Gastroenty.74:389, 1978.Hirschowitz, B. I., et al., “Demonstration of aNew Gastroscope, the ‘Fiberscope’, ” Gastr-o-enty. 35:50, 1958.Hollender, L. F., and Marrie, A., Highly Selec-tive Vagotorn;j (New York: Masson PublishingU. S. A., Inc., 1979).Ippoliti, A . F., UCLA Center for Ulcer Re-search, Los Angeles, Calif., personal communi-cation, 1980.Ippoliti, A. F., et al., “Cimetidine Versus Inten-sive Antacid Therapy for Duodenal Ulcer: AMulticenter Trial, ” Gastroenty. 74:393, 1978.Isenberg, J . I., “Therapy of Peptic Ulcer, ”), A.M.A. 233:540, 1975.

“Pathogenesis of Peptic Ulcer, ” in Gas-tr-ointe;tinal Diseases, vol. I, 2d cd., M. H.Sleisinger and J. S. Fordtran (eds. ) (Philadel-phia: W. B, Saunders Co., 1978).

, “Round Table Discussion on DuodenalUlcer,” Gastrc~enty. 74:407, 1978.Ivy, A. C., “The Problem of Peptic Ulcer, ”]. A.M.A. 1321053,1946.

Peptic Ulcer (Philadelphia: Blakistonco., 1950).Jones, F. A., et al., Peptic Ulcer Healing: Re-cent Studies on Car-benoxo/one (Baltimore,Md.: University Park Press, 1978).Keeney, R. L., and Raiffa, H., Decision WithMultiple Objectives: Preferences and ValueTradeoffs (New York: John Wiley and Sons,1976).Kennedy, T., and Spencer, A., “Cimetidine forRecurrent Ulcer After Vagotomy or Gastrec-tomy: A Rardomised Controlled Trial, ” Br.Med. ]. 1:1242, 1978.Klarman, H. E,, “Application of Cost-BenefitAnalysis to the Health Services and the SpecialCase of Technologic Innovation, ” lnt, ]. Hlth.Serv. 4:325, 1974.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

105.

Klotz, S. A., and Kay, B. F., “Cimetidine andAgranulocytosis” (letter), Ann. Znt. Med.88:579, 1978.Krause, U., “Long-Term Results of Medicaland Surgical Treatments of Peptic Ulcer, ” ActaClzir. Sc. suppl. 310, 1963.Kruss, D. M., and Littman, A., “Safety ofCimetidine,” Gastroenty. 74:478, 1978Lam, S. K., et al., “Treatment of DuodenalUlcer With Antacid and Sulpiride: A Double-Blind Controlled Study, ” Gastroenty. 76:315,1979.Lance, P., et al., “Non-Ulcer Dyspepsia, Duo-denitis, and Cimetidine, ” in Cirnetidine: TheWestminster Hospital Symposium, 1978, C.Wastell and P. Lance (eds. ) (New York: Chur-chill Livingstone, 1978).Landecker, K. D., et al., “Cimetidine and theGastric Ulcer: A Double-Blind Controlled Tri-al, ” Med. ]. Aust. 2:43, 1979.Laufer, I., “Assessment of the Accuracy ofDouble-Contrast Gastroduodenal Racliology, ”Gastroenty. 74:874, 1976.Longstreth, G. F., et al., “Cimetidine Suppres-sion of Nocturnal Gastric Secretion i n ActiveDuodenal Ulcer, ” N . Eng. ]. Med. 294:801,1976.McCarthy, D. M., “Peptic Ulcer: Antacids orCimetidine?” Hospital Practice 14:52, 1979.

“Report on the United States Experi-ence W’ith Cimetidine in Zollinger-Elli:;on Syn-drome and Other Hypersecretory States, ”Gastroenty. 74:453, 1978.McDougall, B. R. D., et al., “Histamine H,-Receptor Antagonists in the Prophylaxis andControl of Acute Gastrointestinal l-laemor-rhage in Liver Disease, ” in Cimetidine: Pro-ceedings of the Second international Symposiu-m o)? Histamine Hz-Receptor Antagonists, W .L. Burland and M. A. Simkins (eds. ) (Amster-dam-Oxford: Excerpta Medica, 1977).McGuigan, J. E., “The Zollinger-Ellison Syn-drome, ” in Gastrointestinal Disease, vol. 1, 2dcd., M. H. Sleisenger and J. S. Fordtran (eds. )(Philadelphia: W. B. Saunders Co., 1978).McMillen, M. A., et al., “Cimetidine and Men-tal Confusion, ” N. Eng. ]. Med. 293:284, 1978.Meade, T. W., et al., “Recent History of Is-chaemic Heart Disease and Duodenal Ulcer inDoctors, ” Br Med. ]. 3:701, 1968.Mendeloff, A. I., “What Has Been Happeningto Duodenal Ulcer?” Gastroenty, 67:1020,1974.Monson, R. R., and MacMahon, B., “Peptic

106.

107.

108.

109.

110.

111.

112.

113.

114.

115.

116.

117.

118.

Ulcer in Massachusetts Physicians, ” N. Eng. ].Med. 281:11, 1969.Morris, T., and Rhodes, J., “Progress Report:Antacids and Peptic Ulcer, A Reappraisal, ”Gut 20:538, 1979.Mosbech, J., and Videbaek, A., “MortalityFrom and Risk of Gastric Carcinoma AmongPatients With Pernicious Anaemia, ” Br. Med.]. 2:390, 1950.Moshal, M. G., et al., “Treatment of DuodenalUlcers W i t h Cimetidine, ” S. Afr. ). M e d .52:760, 1977.Netherlands Economic Institute, Present Costof Peptic Ulceration to the Dutch Economy andPossible Zmpact of Cimetidine on This Cost(Rotterdam: Netherlands Economic Institute,1977).Northfield, T. C., and Blackwood, W. S.,“Controlled Clinical Trial of Cimetidine forDuodenal Ulcer, ” in Cimetidine: Proceedingsof the Second International Symposium on His-tamine Hz-Receptor Antagonists, W. L. Bur-land and M. A. Simkins (eds. ) (Amsterdam-Oxford: Excerpta Medica, 1977).Ochsner, A., “Treatment of Peptic Ulcer Dise-ase, ” in Controversy in Surgery, R. L. Varcoand J. P. Delaney (eds. ) (Philadelphia: W. B.Saunders Co., 1976).Office of Technology Assessment, U.S. Con-gress, Assessing the Efficacy and Safety ofMedical Technologies (Washington, D. C.:U.S. Government Printing Office, 1978).Paringer, L., and Berk, A., Cost of illness andDisease, Fiscal Year 1975 (Washington, D. C.:Georgetown University, Public Services Lab-oratory, 1977).Paringer, L., et al., Economic Cost of Illness,Fiscal Year 197’5, Report BIA (Washington,D. C.: Georgetown University, Public ServicesLaboratory, 1977).Peterson, W. L., and Fordtran, J. S., “Reduc-tion in Gastric Acidity, ” in GastrointestinalDisease, vol. I, 2d cd., M. H. Sleisenger and J.S. Fordtran (eds. ) (Philadelphia: W. B. Saun-ders Co., 1978).Peterson, W. L., et al., “Healing of DuodenalUlcer With an Antacid Regimen, ” N. Eng. ).Med. 297:341, 1977.Priebe, H. J., et al., “Antacid Versus Cimeti-dine in Preventing Acute GastrointestinalBleeding: A Randomized Trial in 75 CriticallyIll Patients, ” N. Eng. ]. Med. 302:426, 1980.Ricardo-Campbell, R., et al., “PreliminaryMethodology for Controlled Cost-BenefitStudies of Drug Impact: The Effect of Cimeti-

119.

120.

121.

122.

123.

124.

125.

126.

127.

128.

129.

130.

131.

dine on Days of Work Lost in a Short-TermTriaI in Duodenal Ulcer, ” J. Clin, Gastroenty.2:37, 1980.Richardson, C. T., “Gastric Ulcer, ” in Gastro-intestinal Disease, vol. 1 , 2d ed., M. H.Sleisenger and J. S. Fordtran (eds. ) (Philadel-phia: W. B. Saunders Co., 1978).Richardson, C. T., et al., “The Effect of Cimeti-dine, a New Histamine H,-Receptor Antago-nist, on Meal-Stimulated Acid Secretion,Serum Gastrin, and Gastric Emptying in Pa-tients With Duodenal Ulcer, ” Gastroenty.71:19, 1976.Robinson Associates, Inc., The impact o fCimetidine on the National Cost of DuodenalUlcers (Bryn Mawr, Pa.: Robinson Associates,Inc., 1978).Ruddell, W. S. J., et al., “Effect of Cimetidineon the Gastric Bacterial Flora, ” Lancet 1:672,1980.

“Pathogenesis of Gastric Cancer inPernici~us Anaemia, ” L.ancet 2:521, 1978.Sawyers, J. L., “Truncal Vagotomy and An-trectomy as the Preferred Operation for Treat-ment of Duodenal Ulcer, ” in Controversy inSurgery, R. L. Varco and J. P. Delaney (eds. )(Philadelphia: W. B. Saunders Co., 1976).Scheurer, U., et al., “Gastric and DuodenalUlcer Healing Under Placebo Treatment, ” Gas-troenty. 72:838, 1977.Schlippert, W., “Cimetidine: H z - R e c e p t o rBlockage in Gastrointestinal Disease, ” Arch.Int. Med. 138:1257, 1978.Schoenbaum, S. C., et al., “The Swine Influen-za Decision,” N. Eng. J. Med. 295:759, 1976.Scitovsky f A. A. and McCall, N., “Changes inthe Costs of Treatment of Selected Illnesses,1951 -1964-1971,” Palo Alto Medical ResearchFoundation, DHEW publication No. (HRA)77-3161 (Rockville, Md.: National Center forHealth Services Research, 1976).Semb, L. S., et al., “A Double-Blind Multi-center Comparative Study of Cimetidine andPlacebo in Short-Term Treatment of ActiveDuodenal Ulceration, “ in Cimetidine: Proceed-ings of the Second international Symposium onHistamine Hz-Receptor Antagonists, W. L.Burland and M. A. Simkins (eds. ) (Amster-dam-Oxford: Excerpta Medica, 1977).Shain, M., and Roemer, M. I., “Hospital CostsRelate to the Supply of Beds, ” Mod. Hosp.92:71, 1959.Sharpe, P. C. and Hawkins, B. W., “Efficacyand Safety of Cimetidine: Long-Term Treat-ment With Cimetidine, ” in Cimetidine: Pro-

132.

133.

134.

135.

136.

137.

138.

139.

140.

141.

142.

143.

144.

145.

ceedings of the Second International Symposi-um on Histamine Hz-Receptor Antagonists,W. L. Burland and M. A. Simkins (eds. ) (Am-sterdam-Oxford: Excerpta Medica, 1977).SK&F Lab Co., The History and Developmentof Hi-Receptor Antagonists (Carolina, PuertoRico: 1979).Sleisenger, M. H., and Fordtran, J. S. (eds. ),Gastrointestinal Disease (Philadelphia: W. B.Saunders Co., 1978).Smith, M. P., ‘ Decline in Duodenal Ulcer Sur-gery, ” ]. A.M.A. 237:987, 1977.SmithKline Corp., 1O-K Reports filed with theU.S. Securities and Exchange Commission,Washington, DC., 1977 and 1978.Soil, A. H., and Walsh, J. H., “Regulation ofGastric Acid Secretion, ” An??. Rev. Physiol.41:35, 1979.Sonnenberg, A. and Hefti, M., “The Costs ofPostsurgical Syndromes (Based on the Exampleof Duodenal Ulcer Treatment ),” Clinics in Gas-troenty. 8:235, 1979.Stadil, F., and Stage, J. G., “Cimetidine and theZollinger-Ellisoa (Z-E) Syndrome, ” in Cimeti-

, dine: The Westminster Hospital Syrnposiunl,\ 1978, C. Wastell and P. Lance (eds. ) (New

York: Churchill Livingstone, 1978).Sturdevant, R. A. L., “How Should Results ofControlled Trials Affect Clinical Practice?”(editorial), Gas/roenty. 73:1179, 1977.Sturdevant r R. A. L., and Walsh, J. H., “Duo-denal Ulcer, ” in Gastrointestinal Disease, vol.1, 2d cd., M. H. Sleisenger and J. S. Fordtran(eds, ) (Philadelphia: W. B. Saunders Co.,1978).Susser, M. , “C,~uses of Peptic Ulcer: A Selec-t ive Epidemiologic Review, /. Chro)z. Dis.20:435, 1967.Susser, M . , and Stein, Z., “Civilization andPeptic Ulcer, ” Lancet, 1 :115, 1962.Thompson, P. L., et al., “Late Radiation Ne-phritis After Gastric X-Irradiation for PepticUlcer,” Q. ]. Mcd, 40:145, 1971.Van Thiel, D. H., et al., “Hypothalamic-Pitui-tary-Gonadal Dysfunction in Men Using Cime-tidine, ” N. Eng. ), Med. 300:1012, 1979.Villeneuve, J. P., and Warner, H. A., “Cimeti-dine Hepatitis, ” Gastroetzty. 77:143, 1979.

146.

147,

148.

149.

150.

151.

152.

153.

154.

155.

156.

157.

Von Haunalter, G., and Chandler, V. V., Costof Ulcer Disease in the United States [MenloPark, Calif.: Stanford Research Institute,1977).Walker, C. O., “Chronic Duodenal Ulcer,” inGastrointestinal Disease, M. H. Sleiseng;er andJ. S. Fordtran (eds. ) (Philadelphia: W. B Saun-ders Co., 1973).Wallace, W. A., et al. “Perforation of C’hronicPeptic Ulcers After Cimetidine, ” Br. A4ed. ].2:865, 1977.Warner, K., et al., “Cost-Benefit and Cost-Effectiveness Analysis in Health Care: Metho-dology and Literature Review, ” draft reportsubmitted to the Office of Technology Assess-ment, U.S. Congress, Washington, D. C., Oc-tober 1979.Wastell, C., and Lance, P. (eds. ), Cimctidine:The West) ninster Hospital Symposium, 1978(New York: Churchill Livingstone, 1978).Weinstein, M. C., “Economic Evaluation ofMedical Procedures and Technologies: Prog-ress, Problems, and Prospects, ” in hledicalTechnology, NCHSR Research ProceedingsSeries, DHEW Publication No. (PHS) 79-3254(Hyattsville, Md.: National Center for HealthStatistics, 1979).Weinstein, M. C., and Stason, W. B., Hyper-tensio)~: A Policy Perspective ( C a m b r i d g e ,Mass.: Harvard University Pressr 1976).Welsh, J. D., “Diet Therapy of Peptic UlcerDisease,” Gastroenty. 72:740, 1977.Winship, D. H., “Cimetidine in the Treatmentof Duodenal Ulcer, ” Gastroenty. 74:402, 1978.Wolfe , M. M. , “Impotence on Cim[?tidineTreatment” ( le t ter ) , N, E)lg. ]. Med 303:94,1979.Wulff, H. R., and Rune, S. J., “A Comparisonof Studies on the Treatment of Gastric LJlcera-t ion With Cimetidine, ” in Cimetidi/ze: T h eWestminster Hospital Symposium, 1978, C.Wastell and P. Lance (eds. ) (New York: Chur-chill Livingstone, 1978).Zeckhauser, R., and Shepard, D., “Where Nowfor Saving Lives?” Law & Conternp. Prob.40:5, 1976.

* U, S (X)V ERNMENT PRINTING OFFICE 1981 341 -844/ 1018