A Cost Analysis of Long Term Antibiotic Prophylaxis forSpontaneous Bacterial Peritonitis in Cirrhosis

Ananya Das, M.D., D.M.Division of Gastroenterology, University Hospitals of Cleveland, Cleveland, Ohio

Objective:The aim of this study was to perform a costanalysis of different strategies of long term antibioticprophylaxis for spontaneous bacterial peritonitis (SBP)in patients with cirrhosis and ascites. The study involveda cost analysis using a decision analysis model and pa-tients with cirrhosis and ascites who are at risk fordeveloping SBP.Methods: Two different strategies ofantibiotic prophylaxis were compared with a “no pro-phylaxis” strategy in patients with cirrhosis and ascitesusing a decision analysis model. In strategy I, antibioticprophylaxis was administered in all patients with cir-rhosis and ascites and in strategy II, patients were strat-ified into a low risk and a high risk group on the basis ofserum bilirubin and ascitic fluid protein levels; onlypatients in the high risk group received antibiotic pro-phylaxis. The cost per patient treated for 1 yr was theoutcome measure compared in the different strategies.Clinical input probabilities and ranges used were ob-tained by searching the MEDLINE database for Englishlanguage articles. Cost estimates were obtained from auniversity hospital setting. Cost analysis was done with asocietal perspective, and only direct costs were takeninto account. Sensitivity analyses were performed toevaluate the effect of variations in the key clinical prob-abilities and cost estimates ranging from a best case to aworst case scenario on the outcome measure.Results:The estimated cost per patient treated in strategy I,strategy II, and strategy III (the strategy of “no prophy-laxis”) were $1311, $1123, and $3509, respectively. Overa broad range of clinical and cost variables, the strategyof risk stratification and restriction of antibiotic prophy-laxis to the subgroup of patients with cirrhosis andascites who were at high risk for SBP (as identified byserum bilirubin >2.5 mg/dl and ascitic fluid protein <1g/dl) was the most favored strategy. However, when thecost of prophylaxis was low or the probability of aprimary episode of SBP was at the lower end of therange reported in the literature, administering antibioticprophylaxis to all patients with cirrhosis and ascites wasthe least costly strategy.Conclusion: This cost analysisindicates that antibiotic prophylaxis particularly whenrestricted to a subgroup of patients who, by simple lab-

oratory parameters, are identified to be at high risk forSBP, is very cost-effective in the prevention of SBP inpatients with cirrhosis and ascites. (Am J Gastroenterol1998;93:1895–1900. © 1998 by Am. Coll. of Gastroen-terology)

INTRODUCTION

Spontaneous bacterial peritonitis (SBP) is a frequent andsevere complication in patients with cirrhosis and ascites(1). The pathogenesis of SBP involves translocation ofenteric flora across the gut mucosa into mesenteric lymphat-ics, with eventual bacteremia and seeding of opsonin-defi-cient ascites (2). This hypothesis is supported by studies inwhich selective intestinal decontamination with differentabsorbable and nonabsorbable antibiotics against aerobicGram-negative organisms have been effective in preventingmost episodes of SBP (3–7). However, it is important toknow whether antibiotic prophylaxis for SBP is cost-effec-tive, particularly so because none of these studies has shownthat antibiotic prophylaxis for SBP improves survival inpatients with cirrhosis and ascites. In addition, although riskfactors (elevated serum bilirubin and low ascitic fluid pro-tein) for developing an episode of SBP are well identified(8), it is not known whether restricting antibiotic prophy-laxis to a subgroup of patients with cirrhosis and ascites whoare at high risk for SBP improves cost-effectiveness.

MATERIALS AND METHODS

A decision analysis model for antibiotic prophylaxisagainst SBP in patients with cirrhosis and ascites was con-structed using standard methods as described elsewhere (9).The decision tree (Fig. 1) was analyzed by using a decisionanalysis software (DATA 3.0, TreeAge Softwares, Inc.)

Strategies compared

Three strategies were compared. In the first strategy(strategy I) antibiotic prophylaxis was given to all patients;in the second strategy (strategy II), patients were stratifiedinto high risk (serum bilirubin.2.5 mg/dl and ascitic fluidtotal protein,1 g/dl) and low risk categories, depending onserum bilirubin and ascitic fluid protein levels. All patientsReceived Oct. 16, 1997; accepted June 5, 1998.

THE AMERICAN JOURNAL OF GASTROENTEROLOGY Vol. 93, No. 10, 1998Copyright © 1998 by Am. Coll. of Gastroenterology ISSN 0002-9270/98/$19.00Published by Elsevier Science Inc. PII S0002-9270(98)00426-2

1895

in the high risk category received antibiotic prophylaxis,whereas patients in the low risk group received none. In thethird strategy (strategy III), none of the patients receivedantibiotic prophylaxis.

Outcome measure

The cost per patient treated for 1 yr was the outcomemeasure compared in the various strategies. Calculation ofthe outcome measure was based on a hypothetical cohort ofpatients with cirrhosis and ascites. Based on the availableliterature (8), it was assumed that$50% of patients withcirrhosis and ascites would have high risk factors for devel-oping SBP and also that all patients with SBP would betreated as inpatients. Analysis was limited to 1 year becauseknowledge regarding the natural history of SBP is limited tothis time frame.

Probability and costs

Clinical probability values and ranges used for analysiswere obtained by searching the MEDLINE database forEnglish language articles and also by reviewing bibliogra-phies of selected articles (Table 1). Cost estimates wereobtained from a university hospital setting and cost analysis

was done with a societal perspective. In this analysis, anattempt was made to use approximated actual cost ratherthan hospital charges or reimbursements (10). Initially max-imum hospital charges in a university hospital setting wascalculated for a prototype patient with cirrhosis and asciteswho had been admitted to the hospital for treatment of anepisode of SBP. The charges include initial CBC and bloodchemistry, ascitic fluid diagnostic tap, fluid analysis, micro-biology, chest x-ray, limited ultrasound of the abdomen, IVantibiotics for 5 days, inpatient charges for 5 days, dailyCBC, alternate day blood chemistries, follow-up diagnosticparacentesis, fluid analysis, and professional charges. Theaverage hospital charge for a hospital admission for treat-ment of SBP was US $13,300. This is comparable to thatreported in other studies (11). For the baseline analysis, thehospital charges were multiplied by 0.6 (the average oper-ating cost/charge ratio for the state of Ohio) to approximatethe actual costs involved. A sensitivity analysis was alsodone by varying the cost/charge ratio from 0.45 to 0.9 toreflect the widespread variation in the cost/charge ratio fromstate to state. The cost/charge ratio was obtained from theHealth Care Financing Administration cost reports (12). The

FIG. 1. Decision tree for different strategies for antibiotic prophylaxis for SBP. The decision tree identifies decision alternatives and their clinical outcomes.The square at the root of the tree indicates a decision node, circles indicate chance branch points and triangles indicate outcomes.

1896 DAS AJG – Vol. 93, No. 10, 1998

cost of the antibiotics to be used for prophylaxis wereobtained from a survey of the local pharmacies.

Sensitivity analyses

Sensitivity analyses were done to evaluate the effect ofvariations in key clinical probabilities and cost estimatesranging from best case to worst case scenario on the out-come measures.

RESULTS

Baseline analysis

In the base case analysis with a 30% 1-yr probability ofa primary episode of SBP and a 70% probability of arecurrent episode of SBP without antibiotic prophylaxis, theestimated costs of treatment per patient over 1 yr are shownin Fig. 2. The lowest cost per patient treated was in strategyII, closely followed by strategy I. Cost per patient treatedwas highest in the no-prophylaxis strategy (strategy III).

Sensitivity analyses

In one-way sensitivity analyses, variations in different clin-ical variables (i.e., cost of hospitalizations, proportion of pa-tients with cirrhosis and ascites who were at high risk fordeveloping SBP, probability of mortality from an episode ofSBP, probability of recurrence of SBP after a primary episodewith or without antibiotic prophylaxis) within the ranges re-ported in the literature did not generally affect the conclusionsof the base case analysis. However, when the probability of aprimary episode of SBP without prophylaxis was,0.07, thestrategy of no-prophylaxis was the preferred strategy; whenthis probability was between 0.07 and 0.1, strategy I wasmarginally more cost-effective than strategy II (Fig. 3). Also,

when the cost of antibiotic prophylaxis was less than US $150(for example, when trimethoprim-sulfamethoxazole or once aweek ciprofloxacin was used for prophylaxis), strategy I be-came the preferred strategy (Fig. 4).

When the two variables of cost of hospitalization fortreatment of an episode of SBP and cost of antibiotic pro-phylaxis were varied simultaneously over a broad range, theconclusion of the baseline analysis was robust enough tohold except at the lower limit of the range of cost ofprophylaxis considered for this analysis, when strategy Iwas the most favored option (Fig. 5).

DISCUSSION

An estimated 900,000 people in the US have cirrhosis ofthe liver (13); in 1989, cirrhosis was the ninth most frequent

FIG. 2. Bar diagram showing cost per patient treated over 1 yr indifferent strategies.

TABLE 1Clinical and Cost Variables Used in the Analysis

Variable Baseline (Range) Reference

Clinical probability1-Yr probability of primary episode of SBP

Without antibiotic prophylaxis 30 (8–50)%High risk group 50% 8Low risk group 8% 8With antibiotic prophylaxis 3 (0–5)% 5, 6

1-Yr probability of recurrent episode of SBP 4, 16Without antibiotic prophylaxis 70 (50–80)%With antibiotic prophylaxis 20 (10–30)%

Incidence of side effects with long term prophylaxis 5 (0–10)% 23Compliance with long term antibiotic therapy 22

Without side effect 85 (50–100)%With side effect 50 (0–100)%

Cirrhotic with ascites at high risk for SBP 50% 1, 2, 8Mortality of SBP 30 (20–50)% 17Cost estimates $Hospitalization cost per episode of SBP 7,850 (6,000–12,000)(including cost of diagnostic work up, pharmacy, nursing, physician services, etc.)Cost of antibiotic prophylaxis over 1 yr 600 (35–800)*

* The wide variation in this cost variable reflects the variations in the cost of different antibiotics and also in the different regimens used for prophylaxis.Quinolone-based regimens are uniformly more costly than trimethoprim-sulfamethoxazole–based regimens.

AJG – October 1998 COST OF ANTIBIOTICS FOR PERITONITIS IN CIRRHOSIS 1897

cause of death in the US and appeared as the first listeddiagnosis in 72,232 hospitalizations (14). The reportedprobability of a primary episode of SBP without antibioticprophylaxis in patients with cirrhosis and ascites varies inthe literature from 8% to 50% (8, 15). Those patients who

already have had an episode of SBP are uniquely susceptibleto recurrent episodes; the 1-yr and 2-yr probability of re-currence are as high as 69% and 74%, respectively (16). Theaverage mortality from SBP is 20–50% (17). In view of thevast magnitude of the morbidity and mortality from SBP and

FIG. 3. One-way sensitivity analysis showing the effect of variations of probability of SBP without antibiotic prophylaxis on cost per patient over 1 yr indifferent strategies. This analysis shows that when the probability of SBP without antibiotic prophylaxis is,0.07, strategy III is the preferred strategy; whenthis probability is between 0.07 and 0.1, strategy I is the most cost-effective. In all other situations, strategy II is the best option.

FIG. 4. One-way sensitivity analysis showing the effect of variations of cost of antibiotic prophylaxis on cost per patient over 1 yr in different strategies.This analysis shows that strategy II is the optimum strategy except when the cost of antibiotic falls below the threshold value of US $150 when strategyofadministering antibiotic prophylaxis to all patients becomes the most cost-efficient option.

1898 DAS AJG – Vol. 93, No. 10, 1998

the financial burden associated with it, any safe and cost-effective prophylactic intervention warrants critical reviewfor incorporation into the routine care of patients with cir-rhosis.

The pathogenesis of SBP is not entirely known. The mostaccepted hypothesis based on clinical observation and ani-mal studies suggests translocation of intestinal bacteriaacross the gut mucosa into mesenteric lymph nodes, withsubsequent bacteremia and seeding of the opsonin-deficientascitic fluid, resulting in peritonitis (18, 19). Because mostof the causative organisms are Gram-negative gut flora, oralantibiotic prophylaxis directed toward selective intestinaldecontamination by inhibiting the aerobic Gram-negativeintestinal flora while preserving the anerobic flora has beenshown in several randomized clinical trials to decrease theincidence of the primary episode and the recurrence of SBP(3–7, 20). However, it is important to note that none of thesestudies has shown any improvement in survival, inasmuchas antibiotic prophylaxis is unlikely to change the naturalhistory of the basic disease.

In this cost analysis, long term antibiotic prophylaxis forSBP is highly cost-effective in the management of patientswith cirrhosis and ascites. Not all patients with cirrhosis andascites are at the same risk for developing SBP. Patientswith ascitic fluid protein concentration,1 g/dl and serumbilirubin level .2.5 mg/dl are at higher risk for SBP (8). Inthis analysis, in most situations it was less expensive toidentify the patients with cirrhosis and ascites who are at

high risk for developing SBP and to restrict antibiotic pro-phylaxis to this group. Also, these patients can be readilyidentified by simple laboratory tests that are part of theroutine work-up of patients with chronic liver disease andascites. However, from this cost analysis, it also seemsprudent to treat all patients with cirrhosis and ascites whenthe cost of antibiotic prophylaxis is relatively low, whichwould be likely when either trimethoprim-sulfamethoxazoleor weekly ciprofloxacin is used for prophylaxis.

Recently, two other cost analyses of antibiotic prophy-laxis against SBP have been published (11, 21). In both ofthese studies, the use of prophylactic antibiotics for theprevention of SBP was found to be a cost-saving strategy. Incomparison with these studies, in the present analysis, thefactors of noncompliance and side effects were incorporatedinto the decision tree. Understandably, in the actual clinicalsetting both these factors will have significant impact on theoverall efficiency of any preventive strategy. In none ofthese studies, a strategy of risk stratification was directlycompared with a strategy of universal prophylaxis. In thepresent analysis, the strategy of risk stratification and re-stricting antibiotic prophylaxis was almost always the mostpreferred strategy. Also, in these two studies, hospitalcharges rather than actual costs were used as cost estimates.In our analysis, to better approximate the actual costs, costto charge ratio was used for cost estimates.

Several limitations of this cost-effective analysis exist.Overall, this decision analysis model was designed to favor

FIG. 5. Two-way sensitivity analysis showing the effect of simultaneous variations of the cost of antibiotic prophylaxis (along the x-axis) and cost ofhospitalization (along the y axis) on the choice of the preferred strategy. This analysis shows over a broad range of possibilities, strategy II is thepreferredstrategy except at the lower range of cost of antibiotic prophylaxis when the strategy of antibiotic prophylaxis in all patients (strategy I) is the most favoredstrategy.

AJG – October 1998 COST OF ANTIBIOTICS FOR PERITONITIS IN CIRRHOSIS 1899

prophylactic antibiotic administration in patients with cir-rhosis and ascites, in particular because of our assumptionsof a high compliance rate and a low incidence of sideeffects. Although definitive data from large trials are notavailable, data from smaller trials suggest high compliancerates for long term antibiotic prophylaxis in this group ofpatients which is comparable to the experience in patientswith granulocytopenia (4, 7, 22). Noncompliance wouldincrease the incidence of primary and recurrent episodes ofSBP and would add to the cost per patient treated in thepreventive strategies. Side effects of long term antibioticprophylaxis using trimethoprim-sulfamethoxazole and nor-floxacin are reported in the range of 0–10% (23). In patientswith AIDS on trimethoprim-sulfamethoxazole for PCP pro-phylaxis, bone marrow suppression because of tri-methoprim-sulfamethoxazole is a major concern often ne-cessitating discontinuation of therapy. However, in thestudy by Singhet al. (7), none of the patients on long termprophylaxis with trimethoprim-sulfamethoxaole for SBP de-veloped any adverse hematological side effects. Also, froma review of current literature, the issue of resistant patho-gens with long term antibiotic prophylaxis does not seem tobe a major clinical problem (4, 24, 25).

Because this analysis was limited to one year, the possi-bility of more than one recurrence of SBP was not taken intoaccount. Presumably, in a patient who already had an epi-sode of SBP, worsening hepatic function with time wouldlead to increasingly frequent episodes of SBP without an-tibiotic prophylaxis; in that event, the cost efficacy of pre-ventive strategies would increase.

Finally in the present analysis, only the direct cost ofdiagnosis, treatment and prophylaxis of SBP were taken into account. Indirect costs (for example, loss of work time)and costs related to other frequent and associated compli-cations (GI hemorrhage, hepatic encephalopathy) were notconsidered. It was presumed that all patients with SBP weretreated as inpatients, although, arguably, a few minor epi-sodes of SBP could be treated with outpatient antibiotictherapy and close follow-up.

In conclusion, this cost analysis suggests that long termantibiotic prophylaxis for SBP in patients with cirrhosis andascites is cost-effective. A strategy of risk stratification inwhich antibiotic prophylaxis is restricted to a subgroup ofpatients, who by simple laboratory parameters are identifiedto be at high risk for SBP, further improves cost-effective-ness.

Reprint requests and correspondence: Ananya Das, M.D., D.M., 80West Center Street, Apt #402, Akron, OH 44308.

REFERENCES

1. Rimola A. Infections in liver disease. In: McIntyre N, Benhamou JP,Bircher J, eds. Oxford textbook of clinical hepatology. Oxford: OxfordMedical Press, 1991:1272–84.

2. Runyon BA. Pathogenesis and diagnosis of spontaneous bacterialperitonitis. In: Rodes J, Arroyo V, eds. Therapy in liver diseases.Barcelona: Ediciones DOYMA S.A., 1992:388–96.

3. Rimola A, Bory F, Teres J, et al. Oral non-absorbable antibioticsprevent infections in cirrhosis with gastrointestinal hemorrhage. Hepa-tology 1985;5:463–7.

4. Gines P, Rimola A, Planas R, et al. Norfloxacin prevents spontaneousbacterial peritonitis recurrence in cirrhosis: Results of a double-blind,placebo-controlled trial. Hepatology 1990;12:716–24.

5. Soriano G, Guarner C, Teixido M, et al. Selective intestinal decon-tamination prevents spontaneous bacterial peritonitis. Gastroenterol-ogy 1991;100:477–81.

6. Rolachon A, Cordier L, Bacq Y, et al. Ciprofloxacin and long termprevention of spontaneous bacterial peritonitis: Results of a prospec-tive controlled trial. Hepatology 1995;22:1171–4.

7. Singh N, Gayoski T, Yu VL, et al. Trimethoprim-sulfamethoxazole forthe prevention of spontaneous bacterial peritonitis in cirrhosis: Arandomized trial. Ann Intern Med 1995;122:595–8.

8. Andreu M, Sola R, Sitges-Serra A, et al. Risk factors for spontaneousbacterial peritonitis in cirrhotic patients with ascites. Gastroenterology1993;104:1133–8.

9. Sox HC. Medical decision making. Stoneham, MA: Butterworth-Hei-nemann, 1988.

10. Finkler SA. The distinction between cost and charge. Ann Intern Med1982;96:102–9.

11. Younossi ZM, McHutchinson JG, Ganiats TG. An economic analysisof norfloxacin prophylaxis against spontaneous bacterial peritonitis.J Hepatology 1997;27:295–8.

12. The Federal Registrar 1994, Vol. 59, No. 102, Friday, May 27, pp.27838–9.

13. Doufour MC, Stinson FS. Trends in cirrhosis morbidity and mortality:United States, 1979–88. Semin Liver Dis 1993;13:109–25.

14. Centers for Disease Control and Prevention. Deaths and hospitaliza-tions from chronic liver disease and cirrhosis—United States, 1980–1989. MMWR 1993;41:969–73.

15. Llach J, Rimola A, Miquel N, et al. Incidence and predictive factors offirst episode of spontaneous bacterial peritonitis in cirrhosis withascites: Relevance of ascitic fluid protein concentration. Hepatology1992;16:724–7.

16. Tito L, Rimola A, Gines P, et al. Recurrence of spontaneous bacterialperitonitis in cirrhosis: Frequency and predictive factors. Hepatology1988;8:27–31.

17. Runyon BA. Spontaneous bacterial peritonitis: An explosion of infor-mation. Hepatology 1988;8:171–5.

18. Llovet JM, Bartoli R, Planas R, et al. Bacterial translocation in cir-rhotic rats. Its role in the development of spontaneous bacterial peri-tonitis. Gut 1994;35:1648–52.

19. Runyon BA, Borizo M, Young S, et al. Effect of selective boweldecontamination with norfloxacin on spontaneous bacterial peritonitis,translocation and survival in an animal model of cirrhosis. Hepatology1995;21:1717–24.

20. Soriano G, Guarner C, Thomas A, et al. Norfloxacin prevents bacterialinfection in cirrhosis and gastrointestinal hemorrhage. Gastroenterol-ogy 1992;103:1267–72.

21. Inadomi J, Sonnenberg A. Cost-analysis of prophylactic antibiotics inspontaneous bacterial peritonitis. Gastroenterology 1997;113:1289–94.

22. Bow EJ, Rayner E, Louie TJ. Comparison of norfloxacin with cotri-moxazole for infection prophylaxis in acute leukemia. The trade off forreduced gram negative sepsis. Am J Med 1988;84:847–54.

23. McEvoy GK. AHFS Drug Information 96. Bethesda: American Soci-ety of Health System Pharmacists, 1996:539, 589.

24. Karp JE, Merz WG, Hendriksen C, et al. Oral norfloxacin for preven-tion of Gram negative bacterial infections in patients with acute leu-kemia and granulocytopenia: A randomized, double-blind, placebo-controlled trial. Ann Intern Med 1987;106:1–7.

25. Johnson PC, Ericsson CD, Morgan DR, et al. Lack of emergence ofresistant fecal flora during successful prophylaxis of traveller’s diar-rhea with norfloxacin. Antimicrob Agents Chemother 1986;3:671–4.

1900 DAS AJG – Vol. 93, No. 10, 1998