The Co-occurrence of DSM-IV Alcohol Abusein DSM-IV Alcohol Dependence:  Results of the National Epidemiologic Survey on Alcohol and RelatedConditions on Heterogeneity That Differ by Population Subgroup FREE

When the signs and symptoms of 2 mental disorders seem related, questionsoften arise about their relationship to each other. This has been the casefor alcohol abuse and alcohol dependence since DSM-III introducedthe division of alcoholism into these 2 disorders. Previous questions aboutalcohol abuse and dependence have included whether abuse should be droppedfrom the nomenclature,¹ whether abuse usuallydevelops into dependence,^2- 6 andwhether abuse and dependence are interchangeable.^7- 9 Thesequestions have largely been answered in the negative. However, a questionnot yet addressed is the degree to which alcohol dependence occurs with orwithout abuse. This question is important for several reasons. First, in geneticsresearch, the presence or absence of alcohol abuse among individuals withdependence may introduce unrecognized heterogeneity into the diagnostic phenotypeof alcohol dependence. Second, the occurrence of alcohol dependence with andwithout abuse may affect prevalence estimates of dependence in surveys thatuse assessment of alcohol abuse as a screening method for alcohol dependence.This has been done in several recent surveys, including the National Instituteof Mental Health National Comorbidity Survey Replication, its follow-up survey,the adolescent version of the survey, and the World Mental Health comorbiditysurveys.^10- 11 In all of thesestudies, symptoms of alcohol dependence are not covered among individualswho screen negative for DSM-IV alcohol abuse, potentiallyleading to underestimation of dependence if it occurs commonly without abuse.Third, case recognition in clinical practice may be affected by use of alcoholabuse as a screening method for alcohol dependence, the method found in awidely used screening procedure for mental disorders in primary care and othermedical settings.^12- 13 In thisscreening procedure, patients who do not report symptoms of alcohol abuseare not asked about alcohol dependence, potentially leading to unrecognizedcases of dependence if patients (or subgroups of patients) experience alcoholdependence without symptoms of abuse.

The conceptual background of the definitions of DSM-IV alcohol abuse and dependence supports examination of this issue becauseit is inconsistent with the final structure of abuse and dependence in DSM-IV. The DSM-IV division betweenalcohol abuse and dependence¹⁴ was based¹⁵ on the theoretical formulation of the alcohol dependencesyndrome (ADS) of Edwards and Gross.¹⁶ TheADS was described as a combination of physiologic and psychological processesreflecting impaired control over drinking, constituting what is now considereda "complex" disorder. Consistent with this, DSM-IV alcoholdependence criteria reflect both physiologic and psychological symptoms. TheADS was defined as one "axis" of alcohol problems, differentiated from anotheraxis consisting of problems such as alcohol-related injuries or social orlegal problems.¹⁷ This second axis correspondsto DSM-IV alcohol abuse. Importantly, the 2 axesin the "biaxial" ADS distinction were not considered "orthogonal."¹⁷ Instead, they were defined as different types ofalcohol-related problems that would co-occur in some but not all cases. Thebiaxial¹⁷ distinction provided the basis forseparating the dependence and abuse criteria in DSM-III-R and DSM-IV. However, DSM-III-R and DSM-IV departed from the ADS conceptby creating a hierarchical structure between alcohol dependence and abuse,leaving alcohol abuse undiagnosed in individuals meeting criteria for alcoholdependence. Because of this, little is known about the extent to which alcoholabuse and dependence co-occur, or whether knowledge about dependence is lostby failing to note the co-occurrence of abuse. Given current interest in betterphenotypes for etiologic research, addressing a potentially common form ofheterogeneity is important. Furthermore, a clear understanding of the relationshipbetween abuse and dependence is needed if one disorder is to be used as ascreening method for the other.

Many clinicians consider abuse an expected prodromal stage of dependence.If this were an accurate concept, there would be little reason for a separate DSM-IV alcohol abuse category at all. Furthermore, dependencewould seldom occur without abuse, and abuse would therefore be a reasonablescreening method for dependence. While this clinical view was supported bya retrospective study of patients and volunteers,¹⁸ sucha study is vulnerable to problems of recall and nonrepresentativeness.¹⁹ In contrast, prospective studies of representativesamples that eliminate these problems consistently show that most untreatedindividuals initially diagnosed as having alcohol abuse (DSM-III-R or DSM-IV) do not progress to alcoholdependence 1 to 15 years later.^{2- 3,5,20- 21} Thesestudies, while important, do not address the extent to which dependence occurswith or without abuse, and whether this produces meaningful heterogeneityin dependence. If DSM-IV alcohol dependence occurswithout abuse in an appreciable proportion of cases, then attention to theimplications is warranted.

Earlier national data suggested that a substantial minority of individualswith current DSM-IV alcohol dependence did not alsohave abuse.²² However, these earlier data arenow 10 years old, and the earlier report was not focused on this question,attracting little attention. Therefore, using new national data, we addressedthese questions: (1) What is the prevalence of DSM-IV alcoholdependence with and without DSM-IV alcohol abusein the general population, and what proportion of DSM-IV alcohol dependence cases are not accompanied by abuse? (2) Does theco-occurrence of abuse with dependence differ in sex-, race-, and age-specificsubgroups of the population?

Subjects were participants in the National Epidemiologic Survey on Alcoholand Related Conditions (NESARC), a nationally representative face-to-facesurvey of 43 093 respondents 18 years and older conducted by the NationalInstitute on Alcohol Abuse and Alcoholism in 2001 to 2002. For this analysis,we included all NESARC respondents except American Indians, who constituteda distinct group too small to analyze, leaving the present sample of 42 392.The target population of the NESARC was the civilian noninstitutionalizedpopulation residing in the United States, including Alaska, Hawaii, and theDistrict of Columbia. African Americans and Hispanics were oversampled, aswere young adults. The NESARC also included a group-quarters sampling frame.Details of the sampling frame are provided elsewhere.^23- 24 Theoverall survey response rate was 81%. The NESARC sample was weighted to adjustfor probabilities of selection of a sample housing unit or housing unit equivalent,nonresponse at the household and person levels, the selection of 1 personper household, and oversampling of young adults. Once weighted, the data wereadjusted to be representative of the US population on a variety of sociodemographicvariables, including region, age, sex, race, and ethnicity based on the 2000decennial census. Of the subjects in this report, 47.9% were male. White subjectscomposed 72.4% of the sample, while 11.3% were African American, 11.8% Hispanic,and 4.5% Asian. By age, 21.9% were aged 18 to 29 years, 30.9% 30 to 44 years,31.0% 45 to 64 years, and 16.3% 65 years or older.

The presence of DSM-IV alcohol abuse and dependencewas assessed with the Alcohol Use Disorder and Associated Disabilities InterviewSchedule–DSM-IV Version (AUDADIS-IV) from theNational Institute on Alcohol Abuse and Alcoholism,²⁵ afully structured diagnostic interview for nonclinician interviewers. The AUDADIS-IVincludes an extensive list of symptom questions that operationalize DSM-IV criteria for alcohol abuse and dependence. The DSM-IV alcohol abuse criteria were rated independentlyof whether dependence was present, allowing identification of alcohol-dependentindividuals with and without abuse. The high reliability and validity of theAUDADIS alcohol dependence diagnosis has been demonstrated in numerous clinicaland general population studies in the United States and abroad.^{8- 9,26- 36} Thereliability of DSM-IV alcohol abuse diagnosis isalso adequate when determined nonhierarchically (independently of dependence),^26- 27,37 as done herein.

In this study, we addressed both current (last 12 months) and lifetimedependence. A variable was also created to indicate whether participants hadbeen in treatment for alcohol problems in the previous 12 months. This providedparallel information about abuse among individuals treated in a variety ofsettings, including alcohol and drug inpatient and outpatient settings, mentalhealth settings, offices of outpatient physicians, offices of other healthcare practitioners, and human service agencies.

Approximately 1800 professional interviewers from the US Bureau of theCensus administered the AUDADIS-IV by means of laptop computer–assistedsoftware with built-in skip logic and consistency checks.³⁸ Theinterviewers had an average of 5 years' experience on census and other health-relatednational surveys. All interviewers completed 10 days of training. This trainingwas standardized across the Census Bureau's 12 regional offices through centralizedtraining sessions under the direction of the National Institute on AlcoholAbuse and Alcoholism and the census headquarters staff. For quality controlpurposes, regional supervisors recontacted a random 10% of all respondentsand asked a subset of the interview questions to verify the accuracy of theinterviewer's performance. This careful process showed that the interviewersperformed at a high level, as indicated by the high reliability of the instrument.²⁹ In the very few cases when the accuracy of the interviewswas uncertain, the interview data were discarded and the interview was repeatedby a supervising interviewer.

The prevalence of DSM-IV alcohol dependencewith and without abuse is shown in percentages weighted for characteristicsof the sample design. Thus, the figures are representative of the US generalpopulation. In addition, the percentage of dependence cases without abuseis presented for the total sample and by sex-, race-, and age-specific subgroupsof the population.

For the full sample, the total prevalence of current (last 12 months) DSM-IV alcohol dependence was 3.80%. When broken down bythe presence of abuse, the prevalence of alcohol dependence with abuse was2.52% and the prevalence of dependence without abuse was 1.28%. Thus, aboutone third (1.28/3.80) of those with current diagnoses of alcohol dependencedid not also have alcohol abuse.

The results for current disorders by race/ethnicity and age groups arepresented separately for men and women in Table 1. Among all men (Table1), the prevalence of dependence cases with and without abuse was3.80 and 1.55, respectively; thus, 29.0% of DSM-IV alcoholdependence cases among men did not also have abuse. Among white men, the prevalenceof dependence cases with and without abuse was 4.01 and 1.39, respectively,indicating that 25.7% of DSM-IV alcohol dependencecases did not have abuse. When minorities were examined, the proportion ofalcohol dependence without abuse was much higher. Among all African Americanmen, 43.0% of those with alcohol dependence did not have abuse, with the highestproportion (52.0%) occurring among those aged 45 to 64 years. Among Hispanicmen, 39.5% of those diagnosed as having DSM-IV alcoholdependence did not have abuse. Estimates became unstable in the oldest groups,especially among minorities, because of the small proportion of dependentsubjects in the oldest groups.

The overall prevalence of dependence among women was 2.28%. The prevalenceof DSM-IV alcohol dependence with and without abusewas 1.23% and 1.05%, respectively (Table1). Thus, among women, 46.1% of current DSM-IV alcohol dependence cases did not also have abuse. Among white women,the prevalence of dependence cases with and without abuse was very similarto the total: 1.33 and 1.04, respectively. Among female minorities, alcoholdependence without abuse was somewhat more common. Among African Americanwomen, 48.5% of cases of dependence did not have abuse, whereas among Hispanicwomen, 55.2% of those diagnosed as having DSM-IV alcoholdependence did not have abuse. The numbers among female Asians were very smallbut suggested a similar pattern. Again, estimates were unstable in the oldestgroups because of the small number of dependent subjects.

Among the 231 subjects in the sample who had received any treatmentfor alcohol problems in the previous 12 months, the results were different.Of these, 69.0% met full criteria for current DSM-IV alcoholdependence with abuse, whereas only 6.1% met criteria for current DSM-IV alcohol dependence without abuse, and thus, only 8.1% had currentdependence without abuse. Specific population subgroups are not shown becauseof small numbers. These numbers are consistent with a clinical concept thatalcohol dependence rarely occurs without abuse, but they differ from the resultspresented in the more representative groups. The small proportion of individualswith current dependence who were in treatment is consistent with the previousepidemiologic literature.

When lifetime DSM-IV dependence was considered,the prevalence for the full sample was 12.48%, consisting of 1.73% withoutabuse and 10.75% with abuse. Thus, 13.86% of the lifetime cases of DSM-IV alcohol dependence did not have accompanying DSM-IV alcohol abuse. Table 2 givesthe results for lifetime disorders for men and women, respectively.

For lifetime DSM-IV alcohol dependence diagnosesamong men (Table 2), the prevalenceof dependence cases with and without abuse was very similar to that in thetotal sample: 15.44% and 1.74%, respectively. Thus, 10.1% of DSM-IV alcohol dependence cases among men did not also have abuse.This proportion was lower among white men, among whom only 8.1% of lifetime DSM-IV alcohol dependence cases did not have abuse. Amongminorities, the proportion of lifetime dependence cases without abuse washigher. Among all African American men, 16.1% of dependence cases did nothave abuse, with the highest proportion (25.0%) occurring among those aged18 to 29 years. Among Hispanic males, 19.2% of those diagnosed as having lifetime DSM-IV alcohol dependence did not have abuse, with a similarproportion (22.2%) noted for Asian men.

The proportion of lifetime cases of dependence without abuse was higheramong women than men (Table 2).Among all women, 22.1% of all DSM-IV alcohol dependencecases did not have accompanying abuse. The proportion was lowest among whitewomen (20.4%) and somewhat higher among African American (24.8%) and Hispanicwomen (29.1%).

This is the first study, to our knowledge, to focus on the general populationprevalence of current and lifetime DSM-IV alcoholdependence with or without alcohol abuse. The large, nationally representativesample and careful measurement in the NESARC made it possible to examine animportant potential source of heterogeneity in DSM-IV alcoholdependence, namely, the presence or absence of DSM-IV alcoholabuse. The results of this study clearly demonstrate that a substantial proportionof current DSM-IV alcohol dependence cases occurwithout accompanying abuse. Furthermore, this phenomenon occurs much morecommonly in the population groups typically understudied and underserved foralcohol use disorders: women and disadvantaged race/ethnicity minority populations.Results for the subset of subjects in treatment contrasted with the full sample,illustrating the risks inherent in generalizing about common disorders fromsamples of clinical cases. When lifetime disorders were considered, about14% of the dependence cases did not have accompanying abuse. The proportionwas higher among women in all race/ethnicity groups and among male minorities.

The study findings are consistent with the original biaxial conceptualizationof alcohol use disorders that served as the basis for the DSM-IV distinction between dependence and abuse symptoms. This conceptionplaced alcohol dependence on one axis and related problems or consequencesof drinking on another.¹⁷ The findings supportthe idea that conditions on these 2 axes co-occur in some but not all cases,information that is lost with DSM-IV because of theexclusion of abuse when dependence is present. The findings are also consistentwith those from a large national survey conducted 10 years earlier²² and thus seem robust to time trends or sampling differences.

Different biopsychosocial processes may give rise to the symptoms ofalcohol dependence and alcohol abuse. For example, genes affecting alcoholreward, craving, or withdrawal (characterizing dependence) may differ fromgenes affecting novelty-seeking or behavioral undercontrol (characterizingabuse). Given that a meaningful subset of dependence cases did not have accompanyingabuse, this type of heterogeneity should be explored. Furthermore, the heterogeneitywas not evenly distributed across major subgroups of the population. Rather,it was higher among women than men and among minorities. Thus, if between-studyresults differ on the relationship of candidate genes to alcohol dependence,sample differences in sex or ethnic composition leading to variability inthe prevalence of abuse among dependent cases could be examined as one potentialexplanation of the inconsistencies.

With advances in genetics research, the need to improve the level ofinformation offered by phenotypes has become increasingly clear. Biological"endophenotypes" are increasingly seen as offering promise.^39- 40 However,identification of useful endophenotypes requires that their links with clinicaldisease be well established.⁴¹ Reduced heterogeneityin disease indicators should facilitate the search for endophenotypes of alcoholdependence as well as for the genes underlying disease. The presence or absenceof abuse symptoms among individuals with alcohol dependence may be one suchsource of heterogeneity in the clinical phenotype.

In general population surveys of DSM-IV alcoholdependence, use of DSM-IV alcohol abuse as a screeningtool for dependence might be advantageous if it were effective. However, becausea sizable proportion of alcohol dependence cases do not also have symptomsof abuse, the use of DSM-IV alcohol abuse as a screeningmethod will result in missed cases and an underestimation of the prevalenceof DSM-IV alcohol dependence. Because the underestimationwill vary by sex and by race/ethnicity, inferences about the relationshipof dependence to important characteristics such as comorbidity (eg, majordepression, which is more common among women) are likely to be altered. Theseissues are also likely to affect time trend data and longitudinal studies,especially when time trends in prevalence, age at onset, or longitudinal coursediffer by sex or race/ethnicity. Statistical modeling cannot be counted onto adjust for loss of information on alcohol dependence due to this methodof screening in such studies. Furthermore, as the underestimation falls mostheavily on women and minorities, decision makers may be misled about the needfor treatment and prevention programs for these already underserved groups.Future general population studies that involve assessment of alcohol use disordersshould test screening techniques in appropriate samples⁴² beforeapplying the techniques in full-scale surveys.

A widely used screening instrument for mental disorder in primary care¹² and its variants¹³ usesthe DSM-IV alcohol abuse criteria as a method toscreen for DSM-IV alcohol dependence. Any increasein screening for alcohol use disorders in medical settings represents importantprogress in the potential for disease prevention. However, our results indicatethat use of abuse symptoms to screen for dependence will result in failureto detect female and minority patients with alcohol dependence, perpetuatingan ongoing oversight in primary medical settings.^43- 46 Giventhat women and minorities are at increased risk of liver damage from heavydrinking⁴⁷ and cirrhosis,⁴⁸ detectionand intervention for alcohol problems in these groups in medical settingsare especially important.

Correspondence: Bridget F. Grant, PhD, PhD, Laboratory of Epidemiologyand Biometry, Room 3077, Division of Intramural Clinical and Biological Research,National Institute on Alcohol Abuse and Alcoholism, National Institutes ofHealth, Mail Stop 9304, 5635 Fishers Ln, Bethesda, MD 20892-9304 (bgrant@willco.niaaa.nih.gov)

Submitted for publication October 14, 2003; final revision receivedMarch 11, 2004; accepted March 16, 2004.

The NESARC is sponsored by the National Institute on Alcohol Abuse andAlcoholism, with supplemental support from the National Institute on DrugAbuse. This study was supported by National Institute on Alcohol Abuse andAlcoholism grants R01AA08159 and K05AA00161, and by the New York State PsychiatricInstitute, New York (Dr Hasin).

The views and opinions expressed in this report are those of the authorsand should not be construed to represent the views of any of the sponsoringorganizations, agencies, or the US government.

We thank Valerie Richmond, MA, for manuscript preparation and editorialassistance.