comparative epidemiology of dependence on tobacco, alcohol, controlled substances, and inhalants:...

8/13/2019 Comparative epidemiology of dependence on tobacco, alcohol, controlled substances, and inhalants: Basic finding…

1/25

Experimental and Clinical Psycho pharmacology1994 . Vol . 2 No. 3 244- 268 In th e publ ic domain

Comparative Epidemiology of Dependence on Tobacco,Alcohol, Controlled Substances, and Inhalants:Basic Findings From the National Comorbidity SurveyJames C. Anthony, Lynn A . W arne r, and Ronald C . Kessler

Studying prevalence of Diagnostic an d Statistical Manual (3rd ed., rev.,Am erican Psychiatr ic Associat ion , 1987) drug dependence among A mericans15-54 yea rs old, we foun d abo ut 1 in 4 (24%) had a history of tobaccodependence; about 1 in 7 (14%) had a history of alcohol dependence; an dabout 1 in 13 (7.5%) had a history of dependence on an inhalant or controlleddrug. About one th i rd of tobacco sm okers had developed tobacco dependen ceand abou t 15% of drink ers had become alcohol depend ent. Am ong users ofthe o ther drugs, about 15% had become dep endent. M any more Americansage 15-54 have been affected by dependence on psychoactive substances thanby o ther psychiatr ic d is turbances now accorded a h igher pr ior i ty in m e n t a lheal th service delivery systems, prev enti on, and sponsored research program s.

The aim of this article is to report basic descrip-tive findings from new research on the ep idemiol-ogy of drug dependence syndromes, conducted aspart of the National Comorbidity Survey (NCS). Inthis study, our research team secured a nat ional lyrepresentative sample and applied s tandardizeddiagnostic assessments in a way that allows directcomparisons across prevalence estimates an d cor-

James C . Anthony, Etiology Branch, Addiction R e-search Center, National Inst i tu te on Drug Abuse andJohns Hopkins University; Lynn A . W arner an d RonaldC. Kessler, Institute for Social Research and Depart-men t of Sociology, University of Michigan.The National Comorbidity Survey (NCS) is a collabo-rative epidemiologic investigation of the prevalence,causes, and consequences of psychiatric mo rbid ity andComorbidity in the United States. The NC S is supportedby U.S. Pub lic He alth Service G rants M H 46376 an dMH 49098 with supplem ental suppor t from the NationalIns t i tute on Drug Abuse and W. T. Grant FoundationGrant 90135190.Preparation of this article w as suppor ted by theNa t iona l Inst i tu te on Drug Abuse Addict ion ResearchCenter . W e acknowledge H . Chilcoat fo r valuable re -search assistance.Correspondence conce rning this article may be ad-dressed to James C . An thony , P. O. Box 5180, Ad dic tionResearch Center , National Ins t i tute on Drug Abuse,Baltimore, Maryland 21224. Electronic mail may be sentto [email protected]. jhu.edu.

relates of tobacco dependenc e, alcohol de pen-dence, and dependence on other psychoactivedrugs (Kessler et al., 1994).For this overview of the survey's findings, aprimary goal has been to answer tw o basic epide-miologic questions about drug dependence involv-ing tobacco, alcohol, controlled drugs such ascocaine, and inhala nts: Firs t , in the populationunder s tudy, what proportion of persons nowqualifies as a currently active or former case ofdrug dependence? Second, where are the affectedcases more likely to be found with in th e sociodemo-graphic s tructure of the s tud y population?In addition, population estimates presented inthis article shed light on the epidemiology ofdepend ence on tobacco, alcohol, and the followingindividual drugs and d rug groups: cannabis; heroin;cocaine; psychostimulants other than cocaine; an-algesic drugs; a dr ug group co nsisting of anxiolytic,sedative, and hypnotic drugs; psychedelic drugs;and inhalant drugs. The following population esti-mates ar e presented fo r each of these listed drugs,including tobacco and alcohol: (a ) l ifetime preva-lence of drug dependence, evalua ted in relation tocriteria published in the Diagnostic and StatisticalManual of Mental Disorders, Third Edition, Re-vised (DSM-III-R; Am erican Psychiatr ic Associa-tion, 1987); (b) l ifetime prevalence of extramedicaldrug use, defined to encompass illicit drug use aswell as patients taking prescribed med icines to get

244


2/25

COMPARATIVE EPIDEMIOLOGY OF D R U G S 245high, taking more than was prescribed, or takingmedicines for other reasons not intended by thedoctor; and (c) the proportion of extramedicalusers who had become drug dependent.Using estimates such as these, we seek to de-scribe the broad population experience with formsof psychoactive drug use that generally occurwithout scrutiny or control by prescribers, pharma-cists, or other health practitioners. Although con-ceding many reasons people might deny or under-report their illicit drug use or drug problems, wedraw attention to how often illicit drug use andsymptoms of drug use disorders are acknowledgedin survey research of this type. For example, on thebasis of confidential interviews conducted for theEpidemiologic Catchment Area (ECA) survey morethan 10 years ago, we found that one in three adultAmericans (30.5%) reported a history of recent orpast illicit drug use. On the basis of self-reportalone, 20% of these illicit drug users had a historyof dependence on controlled substances or arelated drug disorder. Not counting tobacco depen-dence, about one in six adult Americans (17%) t diagnostic criteria for either an alcohol ordr ug disorder, or both (Anthony & Helzer, 1991).These are substantial estimates that convey thepublic health significance of drug use and drugdependence in the United States, and they are fartoo large to be due to the type of exaggeration andoverreporting sometimes found in surveys of druguse in early adolescence (Johnston, O'Malley, &Bachman, 1992). If a correction could be made forunderreporting, these substantial estimates wouldbe even larger.

In the 14 years since the start of the ECAsurveys, the population's drug experience haschanged i n important ways, with passage through anow-subsiding epidemic of crack smoking andother cocaine use (Harrison, 1992; Kandel, 1991).The NCS chronicles results of these changes anddraws strength from some methodological refine-ments that were not part of the ECA researchplan: (a) a nat ionally representative sample of15-54-year-olds; (b) a more complete assessmentof extramedical drug use, applying measurementstrategies developed for the National HouseholdSurvey on Drug Abuse (NHSDA; U.S. Depart-ment of Health and Human Services [USDHHS],1993); (c) deliberate alignment of diagnostic crite-ri a and the measurement strategies used to assessdependence on alcohol, tobacco, and other drugs,

so as to allow comparisons across drug groups; and(d) more thorough adjustment for nonresponsebiases introduced by designated respondents whodeclined to be assessed, perhaps for reasons con-nected to alcohol or drug dependence.

The descriptive estimates presented in this over-view set the stage fo r ongoing research in which weare testing hypotheses about suspected determi-nants and consequences of drug dependence, in-cluding links between drug dependence and otherpsychiatr ic conditions such as anxiety and mooddisorders (Kessler, in press). These findings mayinterest pharmacologists and other scientists whoare concerned about the population's drug experi-ence outside the boundaries of laboratory andclinical research and practice. Those who study thereinforcing functions o f drug use and dependenceliability o f individual drugs m ay gain useful insightsby considering comparative aspects of the epidemi-ology of tobacco, alcohol, and other drug depen-dence, including epidemiologic evidence on thet ransit ion from a single occasion of drug usetoward the development of drug dependence, atopic of considerable interest within the clinicalan d research community (e.g., see Anthony, 1991;Glantz & Pickens, 1992; He nningfield, 1992). Inves-t igators also will find these population estimatesuseful as they seek to substantiate th e potentialpublic health significance o f their pharmacologicstudies or to analyze public policies. Finally, theseestimates m ay have value fo r primary care practitio-ners and family doctors, as well as psychologists,psychiatris ts , or other specialists who prescribepsychoactive drugs or who need to anticipate howfrequently the health status of their patients mightbe complicated by a history of dependence ontobacco, alcohol, or other drugs.

Method and MaterialsThe NCS was based on a stratified, multistage

area probability sample of persons 15 to 54 yearsold in the noninstitutionalized civilian populationin the 48 coterminous United States, including arepresentative sample o f students living in campusgroup housing. Fieldwork was carried out by theprofessional field staff of the Survey ResearchCenter at the Universi ty of Michigan betweenSeptember 14,1990 and February 6,1992. To allowmidcourse adjustments and adaptation to unantici-pated problems, the fieldwork was organized in


3/25

246 J. ANTHONY, L. W A R N E R , AND R. KESSLERrelation to time d release of six replica te sub-samples, each designed to be representa t ive of thestudy population. Overall response rate w as 82.4%,with a to tal of 8,098 p articipants . A more detaileddiscussion of the NCS sampling design and itsimplementation has been given by Kessler et al.(1994).

After sampling, 1 of the 158 specially trainedsurvey interviewers met with each designated re-spondent to adminis ter th e Composite Interna-t ional Diagnostic Interview (CIDI), as adapted forth e N CS to yield detailed inform ation about abroad range of psychiatric disorders, includingdrug dependence (Cot t ier et al., 1991; Robins etal., 1988; Wittchen, in press). These interviewerspar t ic ipated in a 7-day study-specific trainin g pro-gram in the use of the CIDI before beginningf ieldwork. They were trained to follow a protocolin tended to engage th e designated respondents 'in teres t in the survey and to reinforc e surveypar t ic ipa t ion; to secure a private location for theinterview, whic h most often w as wi thin the place ofresidence; to develop trust and rapport with therespondent and to obtain informed consent beforestar t ing the interview; to administer the surveyquestions, as worded, in a fixed sequence; and torecord each subject's responses in a precededresponse booklet. The interviewers were not givenspecial train ing in psychopathology or psychophar-macology, and they were not made aware of anyke y hypotheses under s tudy or of our researchteam's interest in the reinforcin g functions servedby drug use. The highly s t ructured and s tandard-ized interview schedule also was used to gatherinformat ion on suspected correlates and conse-quences of psychiatr ic disorders , includ ing educa-t ional a t ta inmen t , occupat ion , and o ther character -is t ics of des ignated respondents or thei rhouseholds.Because previous surveys have provided someevidence that survey nonrespo ndents have morepsychiatric disorders than respondents, a supple-me nta l nonresponse survey was conducted in tan-dem wi th th e main N CS survey. This w as done byselecting a random subsample o f designated respon-dents who initially were not interviewed eitherbecause of refusal o r (in rare cases) ina bility tocontact after many attempts. These persons wereasked to complete a short-form version of thediagnostic interview.

Assessment of Alcohol an d Other Drug UseRespondents were asked separate questions onthe i r use of alcoho lic beverages, tobacco, and theo the r individual drugs and drug groups lis ted

below. The survey questions on the frequency andrecency of tak ing controlled substances and inhal-ants were nearly identical to s tandardized assess-men t s developed for the N H S D A , now sponsoredby th e Office of Applied Studies within th e Sub-stance Abuse and Men tal Hea lth Services Ad min -istrat ion. These questions clarified our focus onillicit use of Schedule I drugs such as mar i j uana ,heroin, and LSD, as well as extramedical use ofcocaine and other drugs that can be obta inedthrough legitimate medical channels . They werephrased to encompass use of these drugs an dmedicines "o n your own, e i ther wi thou t your ow nprescription from a doctor, or in greater amountsor more often than prescribed, or for any reasonother than a doctor said you should take them(USDHHS, 1993). In addition, respondents wereasked w hethe r they had s tar ted to feel depen denton a drug w hi le tak ing it in accord wi th a doctor'sprescription. Ensuing survey questions coveredtopics such as age of onset, frequ enc y, and recencyof extramed ical drug use for each of the followingindividual drugs and drug groups, which wereadapted from NHSDA convent ions : hero in ; o theropioids and analgesics that can b e obtained throu ghmedica l channels (e.g. , morphine, propoxyphene,codeine); cannabis (marijuana, hashish, or both);psychedelic drugs (e.g. , LSD, peyote, mescaline);inha lant dr ugs (e.g., gasoline or ligh ter fluids,spray paints , amyl nitr i te, nitrous oxide); cocaine( inc luding crack cocaine and freebase); psy-chostimulants other than cocaine (e.g. , dextroam-phetamine, methamphetamine) ; and anxiolytic,sedative, an d hypnotic drugs (e.g., secobarbitaland diazepam, as well as more recently introducedcompounds such as flurazepam, a lprazolam, andt r iazolam) .Consistent with the NHSDA and EGA surveys,the NCS assessment strategy included a deta i ledverbal description of each drug group an d lists ofqualifying drugs tha t were read to each participan t,but i t did not include the NHSDA colored pill cardwith pictures o f different pharmaceut ical products .Fur thermore , the NCS interviewer read th e ques-tions and recorded each participant 's answers on apreceded response fo rm . This approach w as consis-


4/25

COMPARATIVE EPIDEMIOLOGY OF DRUGS 247tent with prior ECA surveys on drug dependencebut was in contrast with the NHSDA approach ofallowing the respondent to self-administer surveyquestions on drug use. When interviewers adminis-ter the questions, it is possible to reduce theimpact of low levels of literacy and re adin g achieve-ment among participants and to use interview skipouts and branching patterns that ca n shorten th einterview and distribute its coverage to otherimportant topics such as suspected risk factors anduse of mental health and other medical services.Although we acknowledge that some populationgroups m ay report m ore drug use when theyself-administer NHSDA questionnaires (Schober,Caces, Pergamit, & Brande n, 1992), we have madea direct comparison of NCS and 1991 NHSDAestimates, generally finding that the estimateswere very close to one another, an d that th eNHSDA estimate always was located within the95% confidence interval for the correspondingNCS estimate. W e return to this topic in theDiscussion section.NCS questions on the use of alcoholic beveragesfollowed a similar NHSDA format and elicitedinformation about age of onset, frequency, re-cency, and q uan tity of drinking. Respond ents alsowere asked whether they had consumed at least 12drinks in any single year of their lives.

Diagnostic Assessment of Alcohol and OtherDrug DependenceTh e CIDI diagnostic assessment of alcohol andother drug dependence for the NCS was based onDSM-III-R criteria, translated into standardizedsurvey questions for administration by a trained layinterviewer. As in the ECA program method usedfor D iagnostic Interview Schedule diagnoses (Ro b-ins, Helzer, Croughan, & Radcliff, 1985), eachparticipant's answers to the survey questions havebeen recorded an d converted to a machine-readable format, and a computer program ha s

been used to determine whether th e diagnosticcriteria have been met. As summ arized recently byWittchen (in press), World Health Organizationfield trials and other methodological studies haveprovided evidence that the CIDI assessments foralcohol and other drug dependence have accept-able levels of interrater reliability an d test-retestreliability, and generally are congruent with inde-pendently made standardized clinical diagnoses.

DSM-III-R criteria require evidence concern-ing nine manifestations of alcohol or other drugdependence grouped under the heading of Crite-rion A, modeled loosely after the original Edw ards-Gross concept for an alcohol dependence syn-drome (Edwards & Gross, 1976). The list of ninemanifestations covers a range of signs or symp-toms, such as those that occur in the context of adrug w ithdrawal syndrome, as well as behavioralmanifestations of drug dependence such as unsuc-cessful attempts to stop or cut down on drug use,and sustaine d use despite recognition that it isrelated to social, psychological, or physical prob-lems. To qualify for a DSM-III-R drug depen-dence diagnosis, a t least three of these nine Crite-rion A manifestations must be met. In addition,Criterion B requires that the disturbance haspersisted for at least one m onth or that presentingfeatures of drug dependence have appeared repeat-edly over a longer period of time. The CIDIincludes two or more survey items designed to tapthe domains represented by each of the nineCriterion A manifestations, as well as questionsconcerning Criterion B. The CIDI lifetime diagno-sis for alcohol or other drug dependence is notmade unless there is positive evidence that therespondent meets both Criterion A and C riterion B .This assessment of alcohol or other drug depen-dence was administered whenever participantsreported occasions of extramedical use of con-trolled substances or in hala nts in their lifetimes, orwhen they reported consuming 12 or more drinksin any one year. For the assessment, identicalstandardized questions were asked for alcohol,controlled substances, and inhalants. By holdingconstant both the diagnostic criteria and the man-ner in which the criteria were assessed, we soughtto reduce methodologic variation that otherwisemight distort comparisons between alcohol and theother drugs. It was not possible to control for thesedifferences in the ECA surveys (An thon y, 1991;Anthony & Helzer, 1991).

Two other methodologic contrasts between th eECA surveys and the NCS also should be men-tioned in relation to controlled substances. First, incontrast with the NCS, the ECA surveys did notcheck fo r drug dependence Diagnostic and Statisti-cal Manual of Mental Disorders, 3rd ed., DSM-III;Am erican P sychia tric Association, 1980) whe n aparticipant reported use of a medicine in accordwith a doctor's prescription, even if that use had


5/25

248 J. ANTHONY, L. W A R N E R , AND R . KESSLERled to feelings of depen dence . Second , the NC Sinc luded in halan ts whe n assessing drug dep en-dence (DSM-III-R), whereas th e EGA surveysdid not. As in the EGA surveys, dependence wasassessed whenever part icipants reported at leastseveral occasions of extramedical drug use, underthe assumption that even as few as six occasionsmight be sufficient fo r deve lopment of drug depen-dence, but that drug dependence would be ex-tremely rare or improbable among persons w hohad used the d rug no more tha n several t imes.

Assessment of Tobacco Use and DependenceWhe n the NC S fieldwork started, there we reinsufficient funds to allow N CS assessment oftobacco use and tobacco dependence d urin g an

already lengthy interview. Midway through field-work, supplemental funding and interview t imebecame avai lable for inclusion of a CIDI sectionon DSM-III-R tobacco dependence designed tobe paral lel to the CIDI assessment fo r DSM-III-Rdependence on alcohol and other drugs, but alsoincluding a few standardized quest ions on behav-iors specific to tobacco smoking. Because of theNCS replicate sampling plan, it was possible toadminister the tobacco assessment to a representa-tive subsample of NCS participants, consisting of4,414 persons, or 55% of the total NCS sample.This assessment of tobacco dependence in-cluded quest ions about dai ly tobacco smoking butnot about more infrequent or i rregula r smoking.Special analyses of the 1991 NHSDA data havebeen completed to fi l l this gap of informatio n. The1991 N HSD A was conducted midw ay throug h theNCS fieldwork, with a nationwid e probabilitysamp le of persons 12 years of age and old er andwith survey assessments of drug use already de-scribed in this art icle. The NHSD A survey ques-tions ascertain whether tobacco cigarettes weresmoked, even on a single occasion, so that theresulting estimates for tobacco use correspond tothe NCS est imates on alcohol use on at least oneoccasion and extramedical use of other drugs on atleast one occasion. To conform with the NCS, theNHSDA analyses were restricted to 15-54-year-olds.1

Quality Control Measures During FieldworkThe interviewers were mo nitored by 18 regionalsupervisors responsible for editing completed in ter-

views before they w ere forwarded to the na t iona lfield office. In addi t ion , cent ra l field office staffreviewed in terview s as soon as they were receivedfrom supervisors. Whenever errors were found orimpor tant information w as missing, t he interviewassignments were sent back to the field for resolu-t ion, and the respondents were recontacted toclarify their answers.

Analysis ProceduresWe present survey-based populat ion est imatesfo r the lifetim e prevalence of extramedical druguse and the lifetime prevalence of drug depen-dence in relat ion to alcohol , tobacco, an d the otherindividual drugs or drug grou ps previously l isted.Each prevalence estimate for drug use is a propor-

tion in which th e numerator consists of the esti-ma t e d numbe r of persons w ho have had at leastone occasion of extramedical drug use in theirlifetimes, whereas th e de nomi na t or is the totalstudy population. Each population prevalence esti-mate for drug dependence has the same denomin a-tor, but the num erato r is the est imated n um ber ofpersons w ho qualify for the CIDI lifetime diagno-sis for drug dependence according to DSM-III-Rcriteria. In addit ion, w e report fo r each individualdrug or drug group est imated proport ions of drugusers in the s tudy popula t ion who had developeddrug dependence. In concept , these proport ionsmay be regarded as estimates of the lifetimeprevalence of drug dependen ce among users in thestudy popu lat ion. Algebra can be used to show th ateach proport ion is equ al to the lifetime prevalenceof drug dependence in the study popula t ion, di -vided by the lifetime prevalence of drug use in thestudy population. Because we had to rely onNHSDA estimates for tobacco use, it was neces-sary to apply the algebraic method when weestimated the proportion of tobacco smokers whohad developed tobacco dep endence, an d stan dar derrors have not been estimated for these tobaccoestimates.We also present est imates for the strength ofassociat ion between drug dependence and plau-sible de te rminants of drug dependence, includingfixed characteristics such as birth year (age) and

These NHS DA analyses were conducted by HowardChilcoat at the Etiology Branch of the Na tiona l Inst itu teon Drug Abuse, Addiction Research Center.


6/25

COMPARATIVE EPIDEMIOLOGY OF DR UGS 249sex, as well as potentially modifiable characteris-tics such as employment status; Table 1 gives afrequency distribution fo r each variable consid-ered in this analysis on the basis of unweightedNCS sample data. To index the strength of associa-tion between drug dependence and each variablelisted in Table 1, we have produced an estimate fo rthe odds ratio. When a particular characteristichas no association with being a currently or for-merly active case of drug dependence, the oddsratio estimate will b e 1.0, o r indist inguishable from1.0 wi thin th e l imits of survey precision. A n oddsratio above 1.0 signals a positive association,whereas an odds ratio between 0.0 and 1.0 signalsan inverse association (Fleiss, 1981). The oddsrat ios for this analysis have been estimated usinglogistic regression models and the Statistical Analy-si s System's PROC LOGISTIC (SAS Institute,1988).

Aside from th e unweighted sample data given inTable 1, all results reported in this article ar ebased on conventional procedures fo r analysis ofcomplex sample survey data. We have used weightsto compensate for variation in sample selectionprobabil i t ies as well as poststratification adjus t-ment factors that compensate fo r survey nonre-sponse as well as other potential sources of surveyerror. Corresponding weights and poststratifica-tion adjustment factors also have been taken intoaccount in the 1991 NHSDA estimates for tobaccouse reported hereinafter.Because of the complex sample design andweighting, standard errors of proportions wereestimated using the Taylor series linearizationmethod (Woodruff & Causey, 1976). The PSRA-TIO program in the OSIRIS IV statistical analysisand data management package was used to makethese calculations (University of Michigan, 1981).Standard errors of odds ratios were estimatedusing the method of Balanced Repeated Replica-t ion in 44 design-based balanced subsamples (Kish& Frankel, 1970). These analytic procedures havebeen described in more detail by Kessler et al .(1994).

ResultsHow Many 15-54- Year-Old Americans HaveDeveloped Drug Dependence?

Table 2 shows l ifetime prevalence estimates anda standard error for each estimate on the basis of

CIDI interviews administered to the 15-54-year-old NCS study population between late 1990 andearly 1992. According to Table 2 (see column 2),an estimated 24.1% of this study population haddeveloped tobacco dependence (±1.0%), whereas14.1% had developed alcohol dependence(±0.7%), and 7.5% (±0.4%) had developed depen-dence on at least one of the controlled substancesor inhalant drugs listed in Table 2 . Thus, in rankorder, a history of tobacco dependence appearedmost frequently in this study population, affectingabout 1 in 4 persons. Alcohol dependence was nextmost prevalent, having affected about 1 in 7persons. A history of dependence on other drugsfol lowed, in aggregate having affected about 1 in13 persons.N ot counting tobacco and alcohol, cannabisaccounted for more dependence than any otherdrug or drug group: In the NCS study population,4.2% qualif ied for the l ifetime diagnosis o f canna-bis dependence (Table 2, row 4, column 2). Depen-dence on cocaine, including crack cocaine, wasnext in rank: An estimated 2.7% of the 15-54-year-old study population had developed cocaine depen-dence. Prevalence estimates fo r only tw o otherdrug categories were above 1.0%: Dependence onpsychost imulants other than cocaine (e.g., amphet-amines) was 1.7%, and dependence upon anxio-lytic, sedative, or hypnotic drugs was 1.2% (Table2, row 7 , column 2 ).

Within the Study Population, Where W as DrugDependence Found?Drug dependence was not distributed randomlywithin the study population; some groups were

affected more than others. This can be seen inTable 3 in which logistic regression was used toproduce odds ratio estimates that show the strengthof association between drug dependence andvar ious selected prevalence correlates such as agean d sex.

Sociodemographic Variation:Sex, Age, an d Race-EthnicityMen were somewhat more likely than women tohave been affected by dependence on alcohol andby dependence on controlled substances or inhal-ants, but not by tobacco dependence. When we

compared the odds of dependence for men versus


7/25

Table 1Description of National Comorbidity Survey Sample in Relationto Selected Chara cteristicsCharacteris t ic

SexMaleFemaleAge15-2425-34 5̂45 +

RaceWhiteBlackHispanicOtherE mploymentWorkingStuden tHomemakerOtherEducat ion (i n years)0-111213-1516+Income (in thousands of dollars)00-1920-3435-6970+Household compositionLive aloneLive with spouseLive with otherLive with parent

Mari ta l s ta tusMarried or cohabi t ingSeparated, widowed, or divorcedNever marriedReligionProtestantCathol icN o preferenceOtherRegionNortheas tMidwestSouthWestUrbanicityMetropol i tanOther urbanN o n u r b a n

M en

8681,2111,1286402,93142 73621273,0295256

2877391,2289479339639931,364527688

2,0258143202,051

5031,2932,0061,063301477

7229931,352780

1,7021,277868

Women

9001,4151,1148223,15358437 11433,010552483206

7351,4511,185880

1,3811,0381,358474510

2,3195868362,3597501,1422,4691,187292303

8311,0841,5298071,8861,452913

Total3,8474,2511,7682,6262,2421,4626,0841,0117332706,0391,077489

4931,4742,6792,1321,8132,3442,0312,7221,0011,1984,3441,4001,1564,4101,2532,4354,4752,2505937801,5532,0772,8811,5873,5882,7291,781

Note. Unweighted sample data from th e National Comorbidity Survey in the coterminous Uni tedStates, 1990-1992.


8/25

COMPARATIVE EPIDEMIOLOGY OF D R U G S 251Table 2Estimated Prevalence of Extramedical Use and Dependence in Total StudyPopulation an d Lifetime Dependence Among Users

Proportionwith a history ofdependenceDrug categories

Tobacco3AlcoholOther drugsCannabisCocaineStimulantAnxiolytics, etc.bAnalgesicsPsychedelicsHeroinInhalants

P24.114.17.54.22.71.71.20.70.50.40.3

SE1.00.70.40.30.20.30.20.10.10.10.1

Proportionwith a history ofextramedical u seP

75.6.91.551.046.316.215.312.79.710.61.56.8

SE0.60.51.01.10.60.70.50.50.60.20.4

Dependenceamongextramedical usersP

31.915.414.79.116.711.29.27.54.923.13.7

SE

0.70.70.71.51.61.11.00.75.61.4Note. Weighted est imates from th e Nation al Comorbidity Survey data gathered in 1990-1992 fo rpersons 15-54 years ol d n = 8,098). Dash ind icates data n ot es t imated . P = E st imated prevalencepropor t ion.a« = 4,414. bAnxiolytics, sedatives, an d hypnotic drugs, grouped.

the odds of dependence for wom en, the odds rat iowas 2.81 fo r alcohol dependence (95% confidenceinterval [CI] = 2.29, 3.44) and 1.62 for controlledsubstances or inhalants (95% CI = 1.24, 2.13). Incontrast, th e observed odds ratio w as 1.18 fo rtobacco dependence, no more than slightly greaterthan th e odds ratio value (1.0) that is expectedunder the null hypothesis of no association. More-over, the 95% CI for the association betweentobacco d epend ence and sex had a span from 0.99to 1.40, trapping the nul l value of 1.0. Thus, theevidence is balanced toward a male excess in theprevalence of dependence on alcohol , control ledsubstances, or inhalants within this study popula-tion, but not toward a m ale excess in prevalence oftobacco dependence.W ith respect to age, a histo ry of tobacco depe n-denc e was least comm on amo ng 15-24-year-olds,wh ereas a history of dependence on alcohol , con-trolled substances or inh alan ts was least commonamong 45-54-year-olds. As shown in Table 3, theodds of tobacco dependence among 15-24-year-olds were about one-half the odds of tobaccodependence among 45-54-year-olds (odds rat io[OR] = 0.48; 95% CI = 0.35,0.64). How ever, com-pared with 45-54-year-olds, tobacco dependencewas no more common among 25-34-year-olds(OR = 0.96) or 35-44-year-olds (OR = 1.0).

In contrast , a history of alcohol dependence w asleast common am ong 45-54-year-olds. By compari-son wi th these older adults, the 15-24-year-oldswere an est imated 1.41 t imes more likely to qualifyfo r a l ifet ime a lcohol dependence diagnosi s(OR = 1.41; 95% CI = 1.08,1.84); the correspond-ing odd s ratio was 1.65 for the 25-34-year-olds(95% CI = 1.27, 2.16) and for the 35-44-year-olds(95%CI = 1.35,2.02).A history of dependence on control led sub-stances or inhalants was most likely to be foundamong young adults, and was least likely to befound amo ng 45-54-year-olds. By comparison with45-54-year-olds, th e est imated odds of depen-dence on these drugs were 2.64 t imes greateramong 15-24-year-olds, 3.50 times greater among25-34-year-olds, and 3.08 times greater among35-44-year-olds.

Compared with White Americans, th e African-American segment of the study populat ion was lesslikely to have a history of tobacco dependence(OR = 0.59), alcohol dependence (OR = 0.35), ordependence on other drugs (OR = 0.54); HispanicAm ericans also were less likely than W hi te Am eri -cans to have a history of tobacco depend ence, butthis was not the case for alcohol dependence p > .05) o r dependence on other drugs p > .05).These inverse associat ions between drug depen-


9/25

Table 3Demographic Correlates of Tobacco, Alcohol, an d Other Drug Dependence

TobaccoCharacteris t ic

SexMaleFemaleAge15-2425-3435-4445 +RaceWhiteBlackHispanicOther

E mploymentWorkingStudentHomemakerOtherEducat ion (i n years)0-111213-1516 +Income (in thousands of dollars)00-1920-3435-6970+Household compositionLive aloneLive wi th spouseLive wi th o the rLive with parentMarital statusMarried or cohabi t ingSeparated, widowed, divorcedNever marriedReligionProtes tantCathol icN o preferenceOtherRegionNortheas tMidwestSouthWestUrbanici tyMetropol i tanOther urbanNonurban

OR1.181.000.48*0.961.001.001.000.45*0.59*0.611.000.45*1.53*1.81*1.68*1.85*1.47*1.001.361.44*1.241.001.001.050.870.39*2.14*2.12*1.000.770.70*1.000.75

1.051.041.000.930.790.851.00

95% CI0.99,

0.35,0.70,0.80,

0.33,0.38,0.36,

0.30,1.08,1.31,1.14,1.34,1.04,

0.99,1.05,0.91,

0.80,0.59,0.28,1.79,1.59,

0.57,0.50,0.46,

0.79,0.76,0.69,0.58,0.59,

1.40

0.641.331.25

0.600.921.05

0.672.182.482.462.542.08

1.861.981.69

1.381.280.562.572.84

1.040.991.25

1.421.421.251.071.22

AlcoholOR

2.81*1.001.41*1.65*1.65*1.001.000.35*1.000.591.000.72*0.72*2.39*1.53*1.45*1.36*1.001.59*1.271.231.001.000.49*0.52*0.43*0.981.311.000.60*0.56*1.000.77

1.36*1.34*1.001.51*1.001.081.00

95% CI2.29,

1.08,1.27,1.35,

0.25,0.72,0.24,

0.51,0.56,1.72,1.23,1.14,1.05,

1.17,0.91,0.87,

0.38,0.36,0.32,0.80,0.98,

0.45,0.42,0.54,

1.01,1.03,1.07,0.74,0.82,

3.44

1.842.162.02

0.501.391.42

0.990.923.331.911.841.76

2.181.781.72

0.640.740.591.201.75

0.800.751.10

1.821.752.141.361.41

Other DrugsOR

1.62*1.002.64*3.50*3.08*1.001.000.54*0.860.531.000.691.59*3.31*1.50*1.47*1.321.002.11*1.431.211.001.000.59*0.670.44*1.081.371.000.51*0.46*1.000.821.250.861.001.79*1.92*1.72*1.00

95%1.24,

1.28,1.92,1.64,

0.37,0.56,0.20,

0.46,1.00,2.10,1.04,1.08,0.89,

1.35,0.85,0.72,

0.41,0.44,0.28,0.75,0.86,

0.36,0.33,0.48,0.88,0.63,1.31,1.33,1.18,

CI2.13

5.456.385.79

0.791.341.42

1.022.515.212.161.981.94

3.312.412.05

0.851.020.691.542.17

0.710.651.391.781.162.462.772.51

Note. Dashes indica te data no t estimated. O R = odds ratio; C I = conf idence interval .*p


10/25


11/25

254 J . ANT HO NY , L . W A R NE R , AND R . KESSLERMarital Status and Religious Preference

In thi s s tudy popula t ion, there was a t endencyfor a history of tobacco dependence to be morecommon among persons who w ere current ly mar-ried or living with par tners as if married (see married in Table 3) and among formerly m a r -ried individuals (separated, divorced, or widowed) ,an d less common among persons w ho were nevermarr ied ; thi s was not the case fo r dependence onalcohol or other drugs (Table 3). On the otherhand, there were fairly consistent associat ionsinvolving rel igious preference, wi th a history ofdependence found m ore f requent ly among per-sons w ho professed no rel igious preference, and bycomparison, least f requent ly among Cathol ics(OR = 0.70 for tobacco; OR = 0.56 for alcohol;O R = 0.46 fo r other drugs). In addit ion, Protes-tants had lower prevalence of dependence onalcohol (OR = 0.60) and control led substan ces orinha lants (OR = 0.51) in a co ntrast to personswith no rel igious preference, but the odds rat ioest imate fo r tobacco dependence among Protes-tan ts w as closer to the null va lue of 1.0 and theassociation was not statistically significant by con-vent iona l s tandards (OR = 0 .77;p > .05).

Location of ResidenceFor dependence on control led substances or

inhalants, there were nonrandom distributions inrelat ion to both region of the country and urban-nonurban location of residence. Compared withresidents of states in the South, individuals livingin th e West were found to be an estimated 1.79times more likely to have developed dependenceon these drugs. Residents of metropoli tan areaswere most likely to have a history of dependenceon control led substances or inha lants (OR = 1.92),fol lowed by res idents of other urban a reas(OR = 1.72), and residents of no nu rba n areaswere least likely to have been drug d ependent .

Alcohol dependence also was associated withregion of the country, but n ot with met ropol i t an o rother urban envi ronments . In a comparison withresidents of the South, the odds of alcohol depen-dence were about 50% greater among personsliving in the West (O R =1.51) and about 35%greater among persons l iving in the Northeas t(OR = 1.36) or in the Midwest (OR = 1.34).In contrast with dependence o n alcohol o r other

drugs, tobacco dependence w as di s t r ibuted essen-tially a t r a ndom i n re la t ion to region of the c oun t ryand urbanicity. All of the tobacco odds rat ioscorresponding to region and u r b a n and nonurba nresidence were close to the nul l va l ue of 1.0 andth e associated co nfidence intervals ha d spans frombelow 1.0 to above 1.0 (Tab le 3).Drug Use and the Transitionto Drug Dependence

Lifetime prevalence proport ions f or drug depen-dence in the study population are determined inpart by how many persons have t ried each type o fdr ug at least once and survived to be interviewed,and in par t by how m a n y o f these surviving d rugusers had proceeded to become drug dependent .For example, considerably fewer members of thestudy popula t ion had t ried tobacco than alcohol(75.6% l ifetime prevalence fo r tobacco use vs.91.5% fo r alcoho l use, a s shown in Table 2, column4). However, dependence was more likely to occuramong tobacco smokers than among alcohol drin k-ers: Of the tobacco smok ers, 31.9% had developedtobacco dependence, whereas only 15.4% of thealcohol d r i nke rs had developed alcohol depen-dence (Table 2, colum n 6). In consequence, withinth e total study popula t ion, the lifetime prevalenceof tobacco dependence (24.1%) was greater thanth e l ifetime prevalence of alcohol dependence(14.1%), even though more persons had consumedalcohol (91.5%) th an had smoked tobacco (75.6%).W h e n control led substances and inha lants wereconsidered as a single group , the data showed th atslightly more than o ne half of the stud y popu lat ionhad take n one or m ore of these drugs for extramedi-ca l reasons (51.0%) an d 14.7% of the users haddeveloped dependence on at least one of the listeddrugs (Table 2, row 3, c o l umns 4 and 6). None t he -less, o ne m ight expect considerable variat ion in theprevalence of ext ramedica l drug use and in thet ransit ion from drug use to drug dependence ,across in dividu al drugs a nd drug groups.After alcohol and tobacco, cannabis was thenext most frequently used drug listed in Table 2,but it ranked low in relat ion to our index ofdependence among users (Table 2, column 6).Within th e s tudy popula t ion, a n estimated 46.3%had used cannabis a t least once, but only 9.1% o fth e users had developed cannabis dependence:For every user with a history of cannabis depen-


12/25

COMPARATIVE EPIDEMIOLOGY OF D R U G S 255dence, there were 10 users who had not becomedependent. y comparison, an estimated 16.2% ofthe study population had tried cocaine at leastonce, and 16.7% of them had qualified as cocainedependent: For each cocaine user with a history ofcocaine dependence, there were 5 users who hadnot become dependent (Table 2, columns 4 and 6).

An estimated 15.3% had used psychostimulantsother than cocaine (e.g., amphetamines) for extra-medical reasons, and 11.2% of these users hadprogressed to develop dependence on these drugs.Corresponding estimates for the anxiolytic, seda-tive, or hypnotic drugs were 12.7% and 9.2%,respectively. An estimated 9.7% of the study popu-lation had used analgesic drugs fo r extramedicalreasons; 7.5% of these analgesic users had becomedependent (Table 3, columns 4 and 6).

An estimated 10.6% of the 15-54-year-old studypopulation reported using psychedelic drugs atleast once, 6.8% reported use of inhalants, and1.5% reported using heroin. An estimated 4.9% ofthe psychedelics users qualified for the depen-dence diagnosis. Among inhalant users, an esti-mated 3.7% qualified as dependent. By compari-son, among heroin users in this sample, 23.1% hadbecome dependent (Table 2).

Age and Drug Dependence by Drug GroupTo some extent, age-associated variation in drug

dependence that was observed in our logisticregression analyses should be understood in rela-tion to differences in prevalence of drug use, inaddition to other factors. For example, a history oftobacco use was less common among 15-24-year-olds as compared with older age groups (see Table4 and Figure 1); this by itself might be sufficient toaccount fo r lower l ife t ime prevalence of tobaccodependence in the comparison of young versusolder persons. However, the NCS data also high-l ighted the importance of age-related differencesin the transition from tobacco use to tobaccodependence. In analyses that considered smokersonly, we found that 15-24-year-old smokers wereless likely than older smokers to have developedtobacco dependence (Table 4 and Figure 1).

For two groups of medical ly prescribed drugs,namely, the analgesics and the anxiolytic, sedative,and hypnotic drugs, the survey-based estimates forl ifetime prevalence o f dependence were higher forolder age groups versus the youngest age group,

despite generally lower prevalence o f extramedicaluse among older adults. This was true for lifetimeprevalence of dependence among extramedicalusers of these drugs as well as for l ifetime preva-lence of dependence among al l persons (Table 4an d Figure 1).A substantially different pattern was observedfo r marijuana, cocaine, psychostimulants otherthan cocaine, and the psychedelic drug group,which showed comparatively lower l ifetime preva-lence of dependence in the oldest age group, alongwith generally lower l ifetime prevalence of extra-medical drug use (Table 4 and Figure 1).

Alcohol offered a unique profile, with 45-54-year-olds having a relatively high l ifetime preva-lence of alcohol use (93.1%) but a relatively lowprevalence of alcohol dependence (10.1%) as com-pared with the other age groups. Alcohol also isnotewor thy because the l ifetime prevalence ofdependence among drinkers was about 10% forpersons 45-54 years of age, considerably less thanestimates of about 16% that were observed for allthree younger age groups (Table 4 and Figure 1).

Drug Dependence by Drug Group and by SexIn relation to lifetime prevalence of drug depen-

dence and l ifetime prevalence of extramedical use,th e rank ordering o f individual drugs and druggroups w as generally identical for men and women(Table 5 and Figures 2 and 3). The exception canbe seen in relation to a sex difference in theranking of psychedelic drugs, which had beentaken by 14.1% of men as compared with 7.2% ofwomen.

In respect to 6 out of 10 individual drugs or druggroups, the lifetime prevalence estimates fo r depen-dence among users were roughly similar for menand women. For example, slightly more than 30%of the tobacco smokers had developed tobaccodependence, and slightly more than 22% of theheroin users had developed heroin dependence—regardless o f sex. Minimal male-female differencesin the prevalence of dependence among users wereobserved for cocaine, analgesics, hallucinogens,and inhalants (Table 5 and Figure 4).In contrast, fo r alcohol a nd cannabis, male userswere somewhat more likely than female users tohave developed dependence (Table 5 and Figure4). Furthermore, wi th respect to the drug groupt ha t included a nxiolytics, sedatives, and hypnotics,


13/25

256 J . A N T H O N Y , L . W A R N E R , AND R . KESSLERTable 4Estimated Prevalence Proportion (P) of Extramedical Use and Dependence inTotal Study Population and Lifetime Dependence Among Users by Age

Propor t ion witha history ofextramedical use

Drug and age (years)Tobacco3

15-2425-3435-4445 +TotalAlcohol15-2425-3435-4445+TotalCannab i s15-2425-3435-4445 +TotalCocaine15-2425-3435-4445 +TotalStimulants15-2425-3435-4445 +TotalAnxiolytics, etc.b15-2425-3435-4445 +TotalAnalgesics15-2425-3435-4445 +TotalPsychedelics15-2425-3435-4445 +Total

P64.476.480.182.775.682.595.094.993.191.536.561.652.125.546.310.626.917.24.416.211.520.318.77.215.38.616.016.08.012.7

10.912.09.15.39.78.314.912.83.510.6

SE1.10.81.31.40.61.10.70.80.80.52.11.81.61.91.11.01.61.40.70.61.11.11.10.80.70.91.20.91.00.51.00.80.80.90.50.71.11.00.60.6

Proportion witha history ofdependenceP

15.226.527.227.224.113.615.615.610.114.15.65.04.40.84.22.64.22.60.52.71.62.81.40.51.70.21.11.91.61.20.20.81.00.90.70.70.70.50.020.5

SE1.72.11.52.31.01.11.11.01.00.70.90.50.70.40.30.60.40.40.30.20.40.70.30.20.30.10.30.40.60.20.10.10.20.50.10.20.20.10.020.1

Dependenceamong users

P23.634.734.032.931.916.516.416.410.715.415.38.18.53.19.124.515.515.311.816.713.513.97.76.511.22.16.811.720.39.21.66.8

11.516.37.58.84.53.80.64.9

SE—————1.31.11.01.10.72.30.71.31.50.74.81.92.46.01.52.93.21.72.01.61.12.02.46.91.10.71.02.97.61.02.51.31.10.60.7


14/25

COMPARATIVE EPIDEMIOLOGY OF DRUGS 257Table 4 continued)

Proportion witha history of

extramedical useDrug and age (years)

Heroin15-2425-3435-4445 +Total

Inhalants15-2425-3435-4445 +Total

Any drug group015-2425-3435-4445 +Total

P0.71.72.70.71.58.19.95.71.86.8

42.864.756.431.551.0

SE0.30.30.50.30.21.00.80.70.50.41.81.71.61.81.0


0.10.30.80.10.40.60.10.10.040.37.49.58.52.97.5

SE0.10.10.40.050.10.30.10.10.040.11.10.90.90.80.4

Dependenceamong usersP

20.115.031.810.723.17.91.52.62.23.7

17.314.714.99.214.7

SE10.16.010.77.75.63.80.71.42.31.42.31.21.52.50.7

Note. Weighted estimates from th e Na t iona l Co morbidi ty Survey data gathered in 1990-1992 fo rpersons 15-54-years-old n = 8,098). Dashes indicate data no t estimated.aTobacco use estimates f rom t he 1991 Na tion al Household Survey on Drug Use, as described in theAssessment of Alcohol a nd Other Drug U se sect ion, n = 4,414. Depe ndence among users estimatedby th e algebraic method described in the Analysis Procedures section; standard errors notestimated.bAnxiolytics, sedatives, an d hypno tic drugs, grouped.c Any drug group refers to the aggregate category comprising th e controlled substances an dinha lan t drugs, but not alcohol or tobacco.female users were som ewhat more likely than maleusers to have developed dependence. For ex-ample, an estimated 14% of men (±0.8%) and11.5% of women (±0.8%) reported extramedicaluse of at least one anxiolytic, sedative, or hypnoticdrug. Among these extramedical users, an esti-mated 6.6% of the m en (±1 .0%) and 12.3% of thewomen (±2.2%) had developed dependence onthis class of drug s (Table 5) .

DiscussionComparison of Prevalence EstimatesOn the basis of general consensus, a DSM-III-R task panel on drug dependence decided tomodify the Edwards-Gross alcohol dependenceconcept and to adopt this m odification as a unifiedconstruct that would apply to all psychoactivedrugs (Kosten & Kosten, 1991; Kosten, Rounsa-ville, Babour, Spitzer, & Williams, 1987). As aresult of this development, for the first time in an

epidemiologic field survey we have been able tohold constant the diagnostic criteria for drugdependence as they are applied to users of to-bacco, alcohol, controlled substances, an d inhal-ants and to reduce marked between-drug differ-ences in diagnostic assessments fo r drugdependence. In doing so, we found that a substan-tial proportion of the 15-54-year-old population inthe Uni ted States— about 1 in 13 persons or7.5%—has been affected by dependence on con-trolled substances or inhalants . By comparison,almost twice as man y had developed alcohol depen-dence, about 14%, and ab out three time s as ma ny,24.1%, had developed tobacco dependence atsome time in life up to the time of the survey. Th eextent to which t he na tion's health is compromisedby this history of drug dependence is likely to be atopic of investigation for m any years.To place the p otential health and social burden sof drug dependence in context with the burden ofother psychiatric morbidity, it may be useful to


15/25


16/25

COMPARATIVE EPIDEMIOLOGY OF D RU G S 259Table 5Estimated Prevalence Proportion P) of Extramedical Use and Dependence in Total Study Population andLifetime Dependence Among Users by Sex

Proportion witha history ofextramedical useDrug

Tobacco3AlcoholCannabisCocaineStimulantsAnxiolytics, etc.bAnalgesicsPsychedelicsHeroinInhalantsDrug group0

P78.393.551.719.518.414.011.614.12.2

9.455.8

SE0.80.51.30.90.80.80.70.80.20.71.3

M en WomenProportion witha history of

dependenceP

25.620.16.23.51.81.00.80.70.50.49.2

SE1.41.00.60.40.40.10.20.20.20.20.7

Dependenceamongmale usersP

32.721.412.018.09.76.66.75.022.34.116.4

SE—1.01.11.91.91.01.61.16.51.71.2

Proportion witha history ofextramedical useP

73.189.641.012.912.211.57.97.20.94.346.4

SE0.70.71.50.70.80.80.50.60.20.41.5


22.68.22.31.91.61.40.70.30.20.15.9

SE1.30.70.30.30.30.30.10.10.10.10.5

Dependenceamongfemale usersP

30.99.25.514.913.312.38.64.725.22.712.6

SE—0.80.72.02.02.21.41.212.92.01.0

Note. We ighted es t imates from the N at ion al Com orbidi ty Survey data gathe red in 1990-1992 for persons 15-54-years-old n =8,098). Dashes indicate data not estimated.aTobacco u se es t imates from th e 1991 Nat iona l Household Survey on Drug Use, as descr ibed in t he Assessment o f Alcohol an d OtherDrug U se section, n = 4,414. Dependence amo ng users es t imated by the algebraic method descr ibed in the Analysis Procedures section;standard errors not estimated.bAnxio ytics, sedat ives , and hypn ot ic drugs , grouped.c Drug group" refers to the aggregate category com pris ing the control led substances and inha lant drugs , but not alcohol or tobacco.

death certificates, wh ich in ma ny places have beenth e main sources of epidemiologic data. In addi-tion, it generally is necessary to take these threeinborn variables into account before assessingevidence about potentially modifiable risk factorsor condit ions that can be leveraged fo r preventionand control of disease. Moreover, birth year, sex,and race are fixed characteristics that cannot bemodified by changes in con dit ions such as diseaseexperience or toxic exposures.In contrast , when we turn to the epidemiologicstudy of potent ial ly mod ifiable condit ions such aseducat ional at tainment , employment status, andaccess to material weal th, w e face the possibilitythat these characterist ics mig ht influence the occur-rence of disease and also might be influenced bydisease. It is in this context that th e l imitat ions ofcross-sectional survey data and the advantages ofprospective or long itudinal data become mostappare nt . Reasoning along these l ines, we caut ionagainst th e causal interpretat ion of cross-sectionalsurvey resul ts , and we note that lifetime preva-lence ratios or odds ratios from l i fet ime prevalenceanalyses typically cannot be inte rpre ted as risk

ratios or relative risk estimates (Kramer, VonKorff, & Kessler, 1980).Even though they should not be given causalinter preta tion , the cross-sectional associations andodds ratios of the type estim ated for this study canpoint out segments of the populat ion where cur-rently active or former cases of drug dependenceare more or less likely to be found. In addit ion tohighlighting the location of these cases within thepopula t ion, cross-sectional odds ratios serve toindex th e strength of each observed association,whether causal or noncausal . Together with ourbasic prevalence estimates, th e observed patternsof association (or lack thereof) represent thebeginning steps for a comparat ive epidemiology oftobacco dependence, alcohol dependence, anddependence on controlled substances.Scanning across odds ratio estimates from ouranalyses, one can find many commonalities andsome noteworthy differences in the associationswith dependence on tobacco, alcohol , and otherdrugs. Ma ny of the observed associations areconsistent w ith prior research, recen t ly summ a-rized by us and by others (e.g., Anthony, 1991;


17/25


18/25

COMPARATIVE EPIDEMIOLOGY OF D R U G S 261Anthony & Helzer, 1991; Anthony & Helzer, inpress; Hawkins, Catalano, & Miller, 1992; Kandel,1991). For example, sex (being a man) had amoderate degree of association with alcohol depen-dence and other drug dependence but not wi thtobacco dependence. This finding converges withother evidence showing an increasing health bur-den of tobacco smoking among women, and per-haps a declining burden among men (e.g., Adams,Gfroerer, & Rouse, 1989; Johnston et al., 1992).

Another comparative difference w as observed i nthe associations with age, especially in the contrastof 45-54-year-olds wi th 15-24-year-olds. The 45-54-year-olds were more likely to have a history ofcurrent or past tobacco dependence; they were lesslikely to have a history of dependence on alcoholor on other drugs. However, age-specific preva-lence estimates fo r individual controlled sub-stances prompt a slightly different conclusion abouttw o classes of drugs that are available by prescrip-tion. Namely, compared with 15-24-year-olds, theoldest adult extramedical drug users under studywere more likely to have become dependent onanalgesic drugs and on anxiolytic, sedative, orhypnot ic drugs, which ar e used widely fo r legiti-mate medical reasons. This aspect of the 45-54-year-old population's drug experience differs fromit s experience wi th alcohol, Schedule I drugs suchas heroin, and inhalant drugs.

These relationships between age and drug depen-dence, generally consistent with prior survey find-ings in the United States, may be understandableas a reflection of variation in drug availability orlevel of drug involvement across different birth

cohorts, periods, or segments of the life span(Kandel, 1991). Alternately, there are two generalepidemiologic observations that m ay clarify under-lying issues of interpretation.

The first general observation involves the forceof drug-related mortality, which can have an impor-tant but nearly hidden impact on age-specificestimates from cross-sectional assessments. At thetime of assessment, each age group under studyconsists of survivors from one or more originalbirth cohorts (e.g., live births in a given year, or in agiven span of years). At some point, mortalityassociated with drug dependence starts to contrib-ute to the loss of drug-dependent persons fromeach birth cohort. Although the force of drug-related mortality can be in operation for all ages ofdrug users, accumulated attrition due to depen-dence-related deaths becomes an increasingly im-portant factor i n successive years of adulthood.In this respect, alcohol provides a good illustra-tion. A s shown in Figure 1, the lower lifetimeprevalence of dependence observed among theoldest drinkers might be due partly to a prematuremortali ty that is secondary to alcohol dependence.

There also was a sharp drop in prevalence ofheroin dependence across the two oldest agegroups surveyed: 2.7% for 35-44-year-olds versus0.7% for 45-54-year-olds. Among 2,242 partici-pants age 35-44 years in the NCS sample, therewere 64 heroin users, but among 1,462 participantsage 45-54 years, there were only 13 heroin users.Based on the experience of these heroin users, wewere able to estimate that almost one third of the35-44-year-old heroin users in the study popula-

Figure 1 opposite). Drug-specific estimates for the lifetime prevalence of drugdependence (first row of graphs), lifetime prevalence of extramedical d rug use (secondrow of graphs), an d lifetime prevalence of drug dependence am ong extramedical drugusers (third row of graphs), based on estimates presented in Table 4. Weightedestimates based on standardized assessment of 8,098 Americans 15-54-years-old, wh owere selected by probability sampling and interviewed for the National ComorbiditySurvey, 1990-1992. Th e scale for individual graphs has been tailored for each set ofestimates, with variation across rows and within rows. Error bars have been used to showthe precision (standard error) of each estimate whenever possible. No error bar hasbeen drawn when the standard error was quite small (i.e., half t he height of the diamondsymbol used to depict the point estimate) or in the case o f tobacco dependence amongusers (see Analysis Procedures section). ASH = anxiolytic, sedative, and hypnotic drugs,grouped (e.g., secobarbital and diazepam, as well as more recently introducedcompounds such as flurazepam, alprazolam, an d triazolam); Any drug group" =aggregate category comprising the controlled substances an d inhalant drugs, but notalcohol o r tobacco; t = change in scale.


19/25

26 2 J . ANTHONY, L . W A R N E R , A N D R . KESSLER

TobaccoAlcoholCannabisCocaine including crackStimuants other than cocaineAnxiolytics sedatives hypnoticsAnalgesic drugsPsychedelic drugsHeroinInhalants

Males

aine ^pnotics Q

»•»

.4 0

Females

oeoo0 .3

0 .2

0 .1

30 o 30P er c en t ag e

Figure 2 . Est imated l ifetime prevalence of drug depen-dence, by sex, based on estimates presented in Table 5.

tion had a history of heroin dependence, but wehad too few users to produce a correspondinges t imate for the 45-54-year-old hero in users. Inth e context of this discussion, it is informative thatwithin this group of 13 people who were 45-54-years-old and who had used heroin at some pointin th eir lives, there w as only on e user with a historyof heroin dependence. We speculate that prema-ture mortality associated with heroin dependencemight account for these sharp differences in theobserved heroin experience of these two olderadult age groups within our s tudy popula tion.The second general epidemiologic observat iont ha t meri ts considerat ion when studying age anddrug dependence relates current age to the transi-t ions from adolescence through age 34, whichrecent evidence pegs as the main period of risk forini t iat ing drug use and d rug depend ence (e.g., seeA nthony , 1991; Kandel , Murphy, & Karus, 1985).At the time of assessment in 1990-1992, the15-24-year-olds had just started to pass thro ughthis main risk period. The importance of this factmight be seen most clearly in relation to theanalgesics and th e anxiolytic-sedative-hypnotic druggroup. Fo r these drugs, the prop ortion of extra-medical users age 15-24 years who had developeda history of dependence w as low, relative to theother age groups (Figure 1). This may reflect thatthese bir th cohorts have just entered the high-riskperiod, and with passing t ime, their experience

may prove to be m ore l ike the experience of theirelders .In relat ion to these N CS findings, tw o otherassociat ions deserve special comm ent:African Americans an d drug use. There is a

widespread popular belief that African Americansare especial ly vulnerable to drug dependence,perhaps because of their overrepresentat ion incertain clinica l samples. However, co nsistent withevidence from other recent epidemiologic surveys,the NCS estimates indicate that tobacco depen-dence, alcohol dependence, and dependence onother drugs a re more common among Wh i te non-Hispanic Americans than among African Ameri -cans (e.g., see Anth ony & Helze r, 1991; K an de l,1991; Lillie-Blanton, Anthony, & Schuster, 1993).Residence in a nonurban area. Alcohol an dtobacco are wide ly ava i lable throughout the Uni t e dStates, but the illicit availability of control ledsubstances seems to vary considerably. Perhapsreflecting these different pat terns of availability,residents of nonurban a reas were a lower preva-lence group fo r dependence on control led sub-stances but not for dependence on tobacco oralcohol (see Table 3). Analogously, the ECAsurvey in the Durham-Piedmont popula t ion ofN or th Carol ina found som ewh at lower prevalenceof drug di sorders among inhabi tan t s of rural areascompared with those living in urban a reas (An-thony & Helzer, 1991).

Males Fema lesAlcohol O

Cannabis <Cocaine including crackStimuants other than cocaineAnxiolytics sedatives hypnoticsAnalgesic drugsPsychedelic drugsHeroinInhalants

100

—o—o—o—0»

o-

A

-o-e

-c•0

-c o«0

0

0 100P e r c e n t a g eFigure 3 . Est imated l ifetime prevalence of ext ramedi-ca l drug use , by sex, based on est imates presented inTable 5 .


20/25

COMPARATIVE EPIDEMIOLOGY OF DRUGS 263Males

Tobacco O-HeroinAlcoholCocaine including crackCannabisStimuants other than cocaineAnxiolytics sedatives hypnoticsAnalgesic drugsPsychedelic drugsInhalants

Females e

—-e

40 40P er c en t ag eFigure 4 . Estimated lifetime prevalence of drug depen-dence among extramedical drug users, by sex, based onestimates presented in Table 5.Transition From Drug Use to Drug Dependence

When EGA results were published, some read-ers were surprised to find that m any of the individu-al s who had used heroin or other controlledsubstances on more than five occasions had notdeveloped drug dependence or other drug disor-der, even though prior research on Vietn am veter-ans ha d indicated as much. Fo r example, amongheroin users in the EGA sample, only 44% hadbecome a case of heroin dependence or heroinabuse as defined by DSM-III criteria. The corre-sponding estimates were close to 20% for extra-medical users of cannabis, psychostimulants otherthan cocaine, and anxiolytics or sedative-hypnoticdrugs (Anthony & Trinkoff, 1989). Similarly, Hel-zer (1985) and Robins (1993) have reported thatmost Vietnam veterans wh o used heroin and otheropioid drugs in Vietn am did not become depen-dent takin g these drugs while overseas.Although n ot directly comp arable to these EG Aestimates based on the experience of individualswho had used drugs on more than five occasions,the NCS estimates also suggest that a large major-ity of persons w ho hav e initiated extramedical druguse have not proceeded to develop drug depen-dence. Even for drugs known for their dependenceliability (e.g., tobacco, cocaine, heroin), th e propor-tion of drug users who were found to have becomedepend ent was in a range from 20-40%. Consider-ing these prevalence estimates on the transition

from drug use to drug dependence, it is noteworthythat NCS interviewers were able to develop trustand rapport sufficient to elicit reports of illegaldrug-taking behavior from many participants. W especulate that m any participants who acknowledgepast illicit behavior also may be willing to reportpersonal problems and other symptoms of drugdependence that they have experienced. At thesame time, we acknowledge a strong possibilitythat some drug-dependent participants did notreport on these problems with completeness ortotal accuracy, which would tend to attenuate theproportion of dependent persons among extramedi-cal drug users.On conceptual grounds, it can be argued thatth e transition from drug use to d rug dependence inthe pop ulation is determined partly by the reinforc-ing functions served by drug-taking and associatedbehaviors, with a linkage back to profiles of drugactivity discovered through laboratory research(e.g., see Schuster, 1989; Thompson, 1981). Inaddition, this transition seems to be determ ined inpart by other factors, some linked to the reinforc-ing functions of drug use, an d some not so readilystudied inside the laboratory (e.g., see Brady, 1989;Glantz & Pickens, 1992; Schuster, 1989). In theory,the array of these interrelated factors includesrelative drug availability an d opportunities for useof different drugs as well as their cost; patterns an dfrequencies of drug us e that differ across drugs;different profiles of vulnerabilities of individualswhose extramedical use starts with on e drug versusanother, as well as both formal a nd informal socialcontrols and sanctions against drug use or in itsfavor, which might be exercised either withinintimate social fields such as the family or work-place or by larger units of social organization.Considered al l together, the array of theoreticallyplausible determinants of the transition from druguse to drug dependence runs a span from th emicroscopic (e.g., the dopamine receptor) throughthe macroscopic (e.g., social norms for or againstdrug use; intern ation al drug-control policies).A better und erstan ding of these sources ofvariation, as well as attention to methodologicalfeatures of cross-sectional survey research, wouldhelp u s account for the rank ordering of individualdrugs and drug groups in relation to the prevalenceof extramedical drug use and in relation to theproportion of extramedical users who were foundto have developed dependence. It is of consider-


21/25

264 J . ANTHON Y, L . WA RN ER, AND R. KESSLERable interest that the ra nk ordering by prevalenceof extramedical drug use was qu i te s imi la r for bothmen and wome n, differing only by the higherprevalence of psychedel ic dru g use among men.T he ran k ordering in re la t ion to t ransi t ions fromdrug use to drug dependence was not the same asthat seen fo r prevalence of extramedical drug use.In addi t ion, th e ranking o f drugs i n relat ion to theuse-to-depend ence t ransi t ion showed some varia-tion across male and fema le drug users and acrossage groups. For bo th m en and w omen , and for al lbut the oldest age group of drug users, tobacco andheroin were top ranked; psychedel ic drugs andinha lants were at the bo t t om. There were male-female differences in lifetime prevalence of depen-dence among extramedical drug users only fo ralcohol and cannabis . A n est imated 21.4% of themale alcohol drinke rs had developed dependence,compared with a n estimated 9.2% of female d r i nk-ers. Corresponding male and female est imates fo rcannabis were 12% and 5.5%, respectively. It isnotewor thy that alcohol and cannabis had higherrank among men, whereas the anxiolytics-seda-t ives-hypnotic drug group w as higher ranked fo rwomen (Table 5).Notwiths tanding these general observat ions,some at tent ion should be given to the fact t ha t15-24-year-old drug users had a comp arat ivelyhigh l i fet ime prevalence of drug dependence inconnect ion with cocaine and alcohol and withSchedule I drugs such as mari juan a. To i l lustrate ,for cocaine, almost 25% of the 15-24-year-oldusers had developed dependen ce. By comparison,only 15% of the 25^4-year-old cocaine users haddeveloped dependence. To the extent that the15-24-year-olds have many rem aining years a t risk,thei r already high value m ay become even higher,but for a possibly com pensat ing influence. Nam ely,it has been observed that early onset illicit drugusers (e.g., those w ho ini t i a te illicit drug use beforeage 17) seem to be at increased risk fo r developingdrug problems compared wi th drug users with alater start (e.g., af ter age 17), even whe n statisticaladjustments are ma de fo r differences in dura t ionof drug use (e.g., see Anthony & Petronis, 1993;Robins & Przybeck, 1985). It follows that thecumulat ive occurrence of drug dependence at firstmight be especial ly high among 15-24-year-olddrug users, but then might decline as the lower riskexperience of later onset drug users is added tothis birth coho rt 's total dru g experience. This is an

empirical ques t ion tha t deserves cont inued s tudy,including future analyses of the NCS data on theexperience of individual birth cohorts born since1935.

Future Directions fo r Researchand Other ImplicationsFutu re direct ion s for research on these topicscan be guided by careful considerat ion of thepresent study's deficiencies. Foremost among thesel imitations is a cross-sectional study design thathas placed heavy reliance on retrospective self-report methods, constraining scientific inferenceabout risk and risk factors in relat ion to d rugdependence . Given unbounded resources , wewould have l iked to m a k e a prospective investiga-

tion of each d rug's users, start ing well before dr uguse had begun, with periodic observat ions sus-ta ined through per iods of risk fo r drug depen-dence and other drug-related hazards, includingth e risk of drug- induced dea th. Ins tead, fo r costconta inment , as in the ECA program, th e a n n u a lNHS DA , and oth er large-scale epidemiologic sur-veys, we have sampled the pop ulat ion's experiencecross-sectionally an d have measured retrospec-tively. This approach leaves out the experiences ofpeople w ho have died, as well as those w ho failedto recall and report their drug involvement withaccuracy. Thus, it is useful to remember tha t onone side l i fet ime prevalence est im ates based onself-reports are he mme d in by the seriousness ofdeath , on the ot her side b y long-forgotten or casualdrug involvement tha t doesn' t seem w orth mention-ing at the t ime of assessment. To the extent thatsome conditions may be associated with consider-able risk of death (e.g., heroin dependence in theUnited States), this commonly used study designactually can yield an unde rc ount of morbidi ty:M a ny seriously affected persons have died. To theextent that other condit ions may be regarded asinconsequential and point less to mention (e.g.,taking a puff on someone else's mari juana ciga-ret te without inh al ing ), the same approach canyield an overcount of morbidi ty: Many mildlyaffected persons are neglected.A s noted previously in this section, cross-sectional survey designs also can lead to mi s i n t e r -preted time relationships. For example, in thisstudy, it appears that l ow educa t iona l achievementmight signal an increased risk of alcohol depen-


22/25

COMPARATIVE EPIDEMIOLOGY OF DRUGS 265dence. In fact, the cross-sectional evidence of thisstudy does not clarify whether alcohol dependenceis a risk factor for low educational achievement,whether lo w education is a risk factor fo r alcoholdependence, whether the relationship is recipro-cal, or whether unmeasured antecedents mightexplain the observed associations between educa-tion and alcohol dependence. In this instance, weare fortunate to have recently published evidencefrom a prospective study, which found excess riskof alcohol disorders among adults who had notreceived high school or college diplomas comparedwith college graduates (Crum, Bucholz, Helzer, &Anthony, 1992; Crum, Helzer, & Anthony, 1993).Of course, whether cross-sectional or prospective,the evidence from a single observational studydoes not always lead to clear inferences aboutcause and effect; our study is no exception. As inexper imenta l research, systematic replication isessential, and ultimately causal inferences must bebased on judgments about the available evidence.

Limitations of more secondary importance in-clude a restriction of the sampling f rame to nonin-st i tut ionalized residents who could be sampledfrom ident if ied dwelling units. Anthony an dTrinkofT (1989) and Anthony and Helzer (1991)have discussed and demonstrated how overallprevalence rates for drug dependence and relatedconditions change very little when institutionalresidents (and by extension, the homeless) areincluded within th e survey sampling frame. How-ever, it should be acknowledged that this overallgeneralization might not hold for population sub-groups with especially high rates of drug-relatedincarceration (e.g., young men of African-Americanheritage).

Current interest in hair analysis and other bioas-says for illicit drug use raises a legitimate questionabout interview assessments of drug dependence(e.g., see Kidwell , 1992). Denned in terms ofDSM-III-R diagnostic criteria and case defini-tions, drug dependence is a psychiatric disturbancefo r which recent drug use is but one indicator.Except in unusual circumstances, these diagnosticcri ter ia for drug dependence can be assessed onlyby means of interviews or examinations, eitherwith designated respondents themselves or within formants fo r these respondents. Given the size ofthe NCS sample and restricted resources, it wasnot possible to interview informants.

Against this background of study limitations, it is

important to focus on the two epidemiologic ques-tions for which this type of cross-sectional fieldstudy is indispensable: (a) In the population, whatproportion of persons has been affected by drugdependence? and (b) Comparing subgroups, wherein the population are cases of drug dependencemore likely to be found? Although the answers tothese questions do not constitute definitive evi-dence with regard to cause and effect relationshipsor mechanisms of action, these answers have animportant public health value. As mentioned inour introduction, these answers can be of use tomembers of the scientific community and can serveto guide program and policy decisions.In this respect, the most important implicationsof this study's results m ay concern it s quantitativefindings about the prevalence and location of drugdependence within the population, and the transi-tion from drug use to drug dependence, nowassessed with epidemiologic estimates on the basisof an NCS sample designed to generalize to a largesegment of the American population. The NCSdraws attention to the relative frequency of depen-dence on tobacco, alcohol, controlled substancesand inhalants, disclosing that m an y more Ameri-cans age 15-54 have been affected by drug depen-dence than by other psychiatric disturbances nowaccorded a higher priority in mental health servicedelivery systems, prevention, and government-sponsored research programs. NCS findings onprevalence correlates add to a growing body ofevidence that African Americans do not seem tobe more vulnerable to drug dependence by virtueof their race, and these findings also point towardsome population segments such as homemakers, inwhich excess drug dependence has been suspectedbut never demonstrated clearly. Finally, the NCSresults highlight an increasingly well-documentedobservation that many drug users, perhaps a vastmajori ty, do not seem to make a transition to drugdependence. For them, instead, drug use lacks themajor complications associated with clinically de-fined syndromes of drug dependence. In a time ofincreasing concern about government expendi-tures for health and health care reform, the distinc-tion between drug use and drug dependence de-serves greater consideration, with commensurateallocation of resources in the direction of drugdependence syndromes that affect public healthand society all too commonly.


23/25

266 J. ANTHONY, L . W A R N E R , A N D R . KESSLERReferences

Adams, E. H., Gfroerer, J. C., & Rouse, B. A. (1989).Epidemiology of substance abuse including alcoholan d cigarette smoking. Annals of the New York Acad-em y of Sciences, 562, 14-20.American Psychiatric Association. (1980). Diagnostican d statistical manual of mental disorders 3rd. ed.).Washington, DC: Author.

American Psychiatric Association. (1987). Diagnostican d statistical manual of mental disorders 3rded., rev.).Washington, D C : Au tho r .

Anthony, J. C. (1991). The epidemiology of drug addic-tion. In N. S. Miller (Ed.), Comprehensive handbook ofdrug an d alcohol addiction (pp. 55-86). New York:Marcel Dekker.

Anthony, J. C., & Helzer, J. E. (1991). Syndromes ofdrug abuse and dependence. In L. N. Robins & D. A .Regier (Eds.), Psychiatric disorders in America (pp.116-154). N ew York: Free Press.

Anthony , J . C. , & Helzer, J. E. (in press). Epidemiologyof drug dependency. In M. Tsaung, M. Tohen, & G.Zahner (Eds.), Textbook of psychiatric epidemiology.New York: Wiley.

Anthony, J . C ., & Petronis, K. R. (1993). Early-onset druguse and risk of later drug problems. Manuscr ip t submit-ted for publicat ion .

An thony , J. C., & Trinkoff , A. M. (1989). United statesepidemiologic data on drug use and abuse: How arethey relevant to testing abuse liability of drugs? InM. W. Fischman & N. K. Mello (Eds.), Testing fo rabuse liability of drugs in humans ( NI DA ResearchMonograph No. 92, pp. 241-266). Rockville, M D :National Ins t i tu te of Drug Abuse.

Bland, R. C., Orn , H., & N e w m a n , S. C. (1988). L i fe t imeprevalence of psychiatric disorders in Edmonton . ActaPsychiatric Scandinavica, 77 (Suppl. 338), 24-32.

Brady, J. V. (1989) Issues in hum an drug abuse liabilitytesting: Overview and prospects for the future. InM. W. Fischman & N. K. Mello (Eds.), Testing fo rabuse liability of drugs in humans (NIDA ResearchMonograph No. 92, pp. 357-370). Rockville, M D:Nat iona l Institu te of Drug Abuse.

Breslau, N ., Fenn, N., & Peterson, E. L. (1993). Ear lysmoking in i t iat ion and n icotine dependence in acohort of young adults. Drug and Alcohol D ependence,33 , 129-137.

Canino, G . J., Bird, H . R., Shrout, P . E., Rubio-Stipec,M., Bravo, M., Ma rtinez, R . Sesman, M ., & Guevara,L. M. (1987). The prevalence of specific psychiatric

disorders i n Puerto Rico. Archives of General Psychia-try, 44, 727-735.

Compton, W . M., Helzer, J. E., Hwu, H.-G., Yeh, E.-K.,McEvoy, L., Tipp, J. E., & Spitznagel, E. L. (1991).N ew methods in c ross-cultural psychiatry: Psychia trici l lness in Taiwan and the United States. AmericanJournal of Psychiatry, 148, 1697-1704.

Cottier, L., Robins, L. N ., Gran t , B. , E laine, J., Towle, I .,Wit tchen , H.-U., Sartorius, N ., Burke, J. , Regier, D .,Helzer, J., & Janca, A. (1991). The CIDI-core sub-stance abuse and dependence questions: Cross cul-t u ra l an d nosological issues. British Journal of Psychia-try, 159, 653-658.

C r u m , R. M.

comparative epidemiology of dependence on tobacco, alcohol, controlled substances, and inhalants:...

Documents