The development of DSM-III-R alcohol dependence in male and female relatives of alcohol-dependentprobands and control subjects is shown. A significant effect of sex and relativevs control status (hazard ratio, 2.73; P < .001[for both variables]) is seen.
The development of any coaggregating DSM-III-R disorder (alcohol dependence, nonnicotine drugdependence, antisocial personality disorder, mood disorder, or anxiety disorder)in male and female relatives of alcohol-dependent probands and control subjectsis shown. A significant effect of sex (hazard ratio, 1.55; P < .001) and relative vs control status (hazard ratio,1.58; P < .001) is seen.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Nurnberger JI, Wiegand R, Bucholz K, et al. A Family Study of Alcohol DependenceCoaggregation of Multiple Disorders in Relatives of Alcohol-DependentProbands. Arch Gen Psychiatry. 2004;61(12):1246–1256. doi:10.1001/archpsyc.61.12.1246
Copyright 2004 American Medical Association. All Rights Reserved.Applicable FARS/DFARS Restrictions Apply to Government Use.2004
Alcohol dependence tends to aggregate within families. We analyzed data
from the family collection of the Collaborative Study on the Genetics of Alcoholism
to quantify familial aggregation using several different criterion sets. We
also assessed the aggregation of other psychiatric disorders in the same sample
to identify areas of possible shared genetic vulnerability.
Age-corrected lifetime morbid risk was estimated in adult first-degree
relatives of affected probands and control subjects for selected disorders.
Diagnostic data were gathered by semistructured interview (the Semi-Structured
Assessment for the Genetics of Alcoholism), family history, and medical records.
Rates of illness were corrected by validating interview and family history
reports against senior clinicians’ all sources best estimate diagnoses.
Sex, ethnicity, comorbidity, cohort effects, and site of ascertainment were
also taken into account.
Including data from 8296 relatives of alcoholic probands and 1654 controls,
we report lifetime risk rates of 28.8% and 14.4% for DSM-IV alcohol dependence in relatives of probands and controls, respectively;
respective rates were 37.0% and 20.5% for the less stringent DSM-III-R alcohol dependence, 20.9% and 9.7% for any DSM-III-R diagnosis of nonalcohol nonnicotine substance dependence,
and 8.1% and 5.2% for antisocial personality disorder. Rates of specific substance
dependence were markedly increased in relatives of alcohol-dependent probands
for cocaine, marijuana, opiates, sedatives, stimulants, and tobacco. Aggregation
was also seen for panic disorder, obsessive-compulsive disorder, posttraumatic
stress disorder, and major depression.
The risk of alcohol dependence in relatives of probands compared with
controls is increased about 2-fold. The aggregation of antisocial personality
disorder, drug dependence, anxiety disorders, and mood disorders suggests
common mechanisms for these disorders and alcohol dependence within some families.
These data suggest new phenotypes for molecular genetic studies and alternative
strategies for studying the heterogeneity of alcohol dependence.
Alcohol consumption and alcohol dependence are influenced by geneticfactors in humans and experimental animals. The heritability of alcohol consumptionis estimated at 35% to 40% in twin studies.1-3 Twinstudies of alcohol dependence have generally shown a ratio of monozygoticconcordance–dizygotic concordance of about 2:1, with differences inthe absolute value of the concordance related to diagnostic criteria.4-8 Astudy of 133 adoptees was performed by Goodwin and Schulsinger9;adopted-away sons of alcoholics had an 18% rate of alcohol dependence (similarto that in sons of alcoholics raised in their biological families) and adopted-awaysons of control subjects had a 5% rate of alcohol dependence. Bohman10 studied 2324 adoptees, finding 39% alcohol dependencein sons of alcoholic fathers and 20% alcohol dependence in sons of controls;the comparable figures for daughters were 20% and 6%. Family studies werereviewed by Cotton11 in 1979, including dataon 6251 relatives of alcoholics and 4083 relatives of controls. Rates of illnessin fathers in the 2 groups were 27% and 5%, respectively; respective ratesin mothers were 5% and 1%.
Characteristically, women have had lower rates of alcohol dependencethan men, in epidemiologic and family and twin studies.12,13 Thequestion arises whether risk for alcohol dependence is passed on comparablyby female and male relatives in families with cases of alcohol dependence.Kaij and Dock14 found that grandsons of alcoholicshad equal risk whether they were the son of a son of an alcoholic or the sonof a daughter of an alcoholic. Likewise, Cloninger et al15 reportedas much alcohol dependence in relatives of female alcoholics as in relativesof male alcoholics. This suggests that the 2 sexes are equivalent in geneticload for alcohol dependence and that differential expression of the illnessin the 2 sexes is related to nongenetic factors.
We may distinguish between comorbidity (disorders occurring togetherin an individual) and coaggregation (disorders occurring together in families).Certain psychiatric disorders have been reported to be more prevalent in personswith alcohol dependence compared with controls (comorbid disorders), includingchild conduct disorder and adult antisocial personality disorder (ASPD),10,16 depression,17-20 andanxiety disorders.18,21 Otherdisorders have been noted in relatives (coaggregating disorders), includingdrug abuse or dependence,22 somatization infemale relatives,15 and attention-deficit/hyperactivitydisorder in juvenile offspring.23
Our goal was to perform a family study using modern diagnostic criteriaand a structured assessment for multiple disorders on Axis I and ASPD on AxisII. The purpose was to identify disorders aggregating in relatives of personswith alcohol dependence and, thus, specify areas of possible shared geneticvulnerability factors. We studied 1269 probands with alcohol dependence ascertainedthrough treatment facilities without regard to family history (a subset ofhigh-density families were later studied for genetic linkage12,24).They and their relatives are compared with the relatives of 232 probands ascertainedat 6 centers to represent a population control sample. In all, 8296 adultfirst-degree relatives of alcohol-dependent probands were compared with 1654adult controls.
The Collaborative Study on the Genetics of Alcoholism began in 1989with the participation of 6 centers (Indiana University School of Medicine;University of Iowa; University of Connecticut; State University of New YorkHealth Sciences Center at Brooklyn; University of California, San Diego; andWashington University). A semistructured interview (the Semi-Structured Assessmentfor the Genetics of Alcoholism [SSAGA]25,26)was designed to assess alcohol dependence by multiple criteria and other majorpsychiatric disorders. The study was approved by institutional review boardsat each site. Probands with alcohol dependence by DSM-III-R criteria (American Psychiatric Association, 1987) and definite alcoholismby the criteria of Feighner et al27 were systematicallyascertained from consecutive admissions to treatment facilities and invitedto participate in the study. This ascertainment method was specifically designedto support a family study. Inclusion criteria included the availability of4 first-degree relatives, at least 2 of whom were living in the catchmentarea of one of the participating sites. Probands and relatives were requiredto be English speaking. Exclusion criteria included habitual intravenous druguse (>30 times in a lifetime or within 6 months of ascertainment), known humanimmunodeficiency virus–positive status, or a terminal illness not relatedto alcoholism. These families were designated as stage 1 and form the dataset for the analyses in this report. A subset of families (designated as stage2) contained at least 2 additional first-degree relatives with alcohol dependence;members of these families were included in analyses of genetic linkage andelectrophysiological features (Rice et al28 providea summary). Control families were identified from various source populations,including motor vehicle registrants, dental clinic attendees, and parentsof college students. Such families were required to have spouses and 3 childrenolder than 13 years willing to participate. Because controls were selectedto represent a subset of the general population, they were not excluded ifthey met the criteria for alcohol dependence or any other psychiatric disorder.Probands, spouses, and first-degree relatives of those aged 18 years and olderwere then invited to participate (children and adolescents in these familieswere assessed using age-appropriate instruments and will be the subject ofa separate report). A comparison of demographic characteristics in the relativeand control groups is shown in Table 1.Differences in age, sex, and ethnicity were accounted for in the multivariateanalyses presented later.
Participants provided informed consent and were personally interviewedwith the SSAGA and the Family History Assessment Module (FHAM). Medical recordswere sought for those with a history of psychiatric treatment. Certain disorders,including posttraumatic stress disorder (PTSD), nicotine dependence, and attention-deficit/hyperactivitydisorder, were only assessed using a revised version of the SSAGA preparedin 1997 and, thus, the denominators for these disorders are smaller than forthe other disorders. All diagnoses were assessed on a lifetime basis. Disordersjudged to be organic (ie, direct effects of substance use) were not includedin these analyses. Original analyses separated control probands from controlrelatives; later analyses combined control probands and control relatives(referred to generically as “controls” from this point on), andresults were essentially the same. Interview instruments and procedures werethe same for relatives of alcohol-dependent probands and controls. Conferencecalls and continuously updated procedure manuals were used to standardizeinterview technique and scoring norms across sites.
Assessment folders were developed for each subject, including the interviewand information from relatives and medical records, where available. A subsetof subjects (1929 of 9950 subjects, or 19.4% of the sample studied herein)were diagnosed by best estimate procedures. This included all members of thestage 2 families included in the genetic linkage studies and a subset of controls.Probands with alcohol dependence were also diagnosed by best estimate procedures,but their data are not included herein because of their ascertainment as affectedpersons. The best estimate process involved senior clinician assessment (J.I.N.,S.O., T.R., M.S., L.K., T.P., L.B., S.K., and V.H.) of all diagnostic materialto assign lifetime diagnoses, blind to proband or relative status. The bestestimate process was then used to validate information from the SSAGA andFHAM. Considering the best estimate as the gold standard, diagnoses basedon algorithmic extraction of information from the SSAGA were classified astrue positive, true negative, false positive, and false negative. These functionswere used to adjust rates of diagnoses in interviewed subjects not includedin the best estimate process (4324 of 9718 subjects, or 44.5%), using thefollowing function: Total number of affected subjects = [SSAGA-positivesubjects × (true positives/total positives)] + [SSAGA-negativesubjects × (false negatives/total negatives)].
By using a similar procedure, information from the FHAM was validated,to assign diagnoses to subjects who did not participate in the SSAGA interview(3497 of 9718 subjects, or 36.0%). As an example, 75.9% of subjects implicatedas having DSM-III-R alcohol dependence by 1 first-degreerelative in fact were assigned that diagnosis in the best estimate process,87.7% of subjects implicated by 2 relatives received that diagnosis by bestestimate, and 93.7% of subjects implicated by 3 relatives received that diagnosisby best estimate. In this way, the results of the best estimate process weregeneralized to all first-degree relatives of alcohol-dependent probands andcontrol probands.
Because the total subject group (and the group of best-estimated subjects)included many more relatives of alcoholics than controls (8296/9718), thevalidation of SSAGA and FHAM was differentially reflective of relationshipsamong self-report, relatives’ reports, and clinician judgment in familiesof an alcoholic proband. However, we have tested the premise that informationfrom relatives is related to self-report (by SSAGA) similarly in familiesof alcohol-dependent probands and controls. In fact, comparison of the proportionof SSAGA-affected subjects among groups of relatives and controls separatedby number of implications of illness by family members did not differ significantlyfor any condition used in the analysis, with a single exception: subjectswith 3 or more implications of DSM-IV alcohol dependenceby family members were more likely to be affected by SSAGA among relativesof alcohol-dependent probands (66%) than among controls (37%). Because onlya few controls were in this category, correction for this difference wouldreduce the estimated prevalence of alcohol dependence in controls only from14.4% to 14.0%. The diagnostic criteria are DSM-III-R unlessotherwise noted. We have observed the standard DSM conventionthat subjects diagnosed as having substance dependence are not also diagnosedas having substance abuse, even though most would meet the symptomatic criteriafor abuse as well.
Age correction was performed for 6 disorders: alcohol dependence by DSM-III-R and DSM-IV, major depression,mania, drug dependence (any diagnosis of nonalcohol nonnicotine substancedependence), and ASPD. For these disorders, data from all directly interviewedaffected subjects in the data set were used to generate an age-of-onset function.The group of subjects at risk was then divided by decade, and the size ofthe unaffected portion of each group adjusted by the proportion of affectedsubjects, with onset by the median of that decade, producing the age-adjustednumber of subjects at risk (a modified Stromgren procedure, as in the studiesby Johnson and Leeman29 and Gershon and colleagues30). The age of onset was considered to be the yearthat a subject met the full criteria for the disorder. For ASPD (presumedto be a continuous trait that begins, by definition, during childhood), theage of onset was considered to be the age at the time of the first symptomand, thus, age-corrected data for this group of adult subjects were identicalto the raw data. We also studied Kaplan-Meier survival curves using the Coxproportional hazards regression model for these disorders (ie, alcohol dependence,major depression, mania, drug dependence, and ASPD). Rates of illness in relativesof alcohol-dependent probands were compared with rates of illness in controlsusing the χ2 test. Relative risk (RR) estimates were calculatedwith their 95% confidence intervals (CIs).
A multivariate logistic regression model was used to account for theeffects of sex, ethnicity, site of ascertainment, birth cohort, comorbidityin the proband, and comorbid alcoholism in the relative. Cohort effects wereassessed by dividing relatives into groups by decade of birth. Familial effects(the effect of the variable number of first-degree relatives in families)were controlled using a random effects odds ratio method and a marginal oddsratio method.
Table 2 shows the prevalence ofvarious psychiatric disorders in first-degree relatives of probands with CollaborativeStudy on the Genetics of Alcoholism–determined alcoholism (DSM-III-R alcohol dependence plus Feighner et al27 definitealcoholism) and in the sample of controls. Many disorders seem to clusterin families with an alcohol-dependent proband, including alcohol dependenceitself (by 4 criterion systems), other forms of substance dependence, ASPD,several anxiety disorders, major depression, and dysthymia. Diagnoses thatdid not seem to cluster in relatives include anorexia, bulimia, mania, andseveral forms of substance abuse, including DSM-IV alcoholabuse. Somatization disorder was too infrequently diagnosed for an accuratecomparison. Attention-deficit/hyperactivity disorder was assessed only inthose subjects interviewed after 1997, and that subgroup does not show a significantaggregation. Table 3 corrects the rawdata for all interviewed subjects, based on the SSAGA false-negative and false-positiverates (see the “Methods” section). This correction lowers therate of DSM-IV alcohol abuse and increases the rateof DSM-IV alcohol dependence, among other changes.Versions of Table 2 and Table 3 in which probands are diagnosed by DSM-IV or International Classification of Diseases, 10thRevision (ICD-10) are available at the following Web site (http://ipr.iupui.edu/coga/research.html); in general, RRs are quite comparable to those presented herein, andthere is no case in which a disorder shows significant familial aggregationin one version of the table but not others. Controlling for the effect offamily size did not change the pattern of familial aggregation (data not shown).Marginal odds ratios were significant for each disorder in Table 2, with P < .05.
For certain disorders, we were able to systematically include informationfrom relatives in assigning diagnoses, because detailed information was providedin the FHAM. These included alcohol dependence, drug dependence, mania, depression,and ASPD. These data have also been age corrected. Results are presented in Table 4. Adjusted rates of DSM-III-R alcohol dependence are 37.0% and20.5% in relatives of probandsand controls, respectively (RR, 1.8; 95% CI, 1.6-2.0). Comparable figuresfor DSM-IV alcohol dependence are 28.8% and 14.4%(RR, 2.0; 95% CI, 1.8-2.3); for any form of DSM-III-R nonalcoholnonnicotine substance dependence, 20.9% and 9.7% (RR, 2.2; 95% CI, 1.9-2.5);for primary major depression, 19.5% and 18.0% (RR, 1.1; 95% CI, 1.0-1.2);for mania, 1.2% and 0.9% (RR, 1.3; 95% CI, 0.7-2.3); and for ASPD, 8.1% and5.2% (RR, 1.6; 95% CI, 1.3-2.0). These prevalence figures represent the bestestimate of rates of illness for the full complement of 8296 adult first-degreerelatives of alcohol-dependent probands and 1654 adult controls. Relativerisk estimates in siblings alone (data not shown) were generally comparableto those in all first-degree relatives; the RR for ASPD in siblings of 2.0(95% CI, 1.5-2.7) was somewhat higher than that in first-degree relativesgenerally. Cox proportional hazards regression ratios for these disorders(interviewed relatives only) are 2.9 (P < .001)for DSM-III-R alcohol dependence, 3.2 (P < .001) for DSM-IV alcoholdependence, 3.2 (P < .001) for drugdependence, 1.1 (P=.08) for major depression, 1.7(P=.16) for mania, and 2.5 (P < .001) for ASPD.
Prevalence estimates were also determined separately by sex (data notshown). Generally, similar patterns are seen in RRs for male and female relatives,with an exception being an RR of 1.8 (95% CI, 1.4-2.4) for ASPD in men comparedwith an RR of 1.2 (95% CI, 0.8-1.8) in women. As expected, the absolute prevalencerates in male relatives exceed those in female relatives for DSM-III-R alcohol dependence (48.0% vs 26.3%), DSM-IV alcohol dependence (37.1% vs 20.7%), drug dependence (26.0% vs 15.9%),and ASPD (11.7% vs 4.6%). Prevalence rates in female relatives exceed thosein male relatives for depression (22.8% vs 16.0%). Prevalence rates for alcoholdependence for each sex are plotted by age in Figure 1; significant effects of sex, and study vs control status,are seen. Rates of alcohol dependence in female relatives of alcohol-dependentprobands are not different from rates in male controls. Prevalence rates forall coaggregating disorders are shown by sex in Figure 2; significant effects of sex, and study vs control status,are seen. In addition to having a higher chance of having any of the coaggregatingdisorders, relatives have more disorders (mean ± SD, 1.3 ± 1.2vs 0.6 ± 0.9 for men [Wilcoxon z = 12.6, P < .001]; and mean ± SD,0.9 ± 1.1 vs 0.4 ± 0.7 for women [Wilcoxon z = 10.7, P < .001]).
Noting the relatively high prevalence of ASPD in the control group,we examined symptom counts for antisocial behavior in relatives and controlsfor 29 variables (to assess whether the ASPD-diagnosed controls were as symptomaticas the ASPD-diagnosed relatives of alcohol-dependent probands). Semi-StructuredAssessment for the Genetics of Alcoholism–positive relatives of probands(n = 325) showed a mean ± SD of 18.2% ± 9.2%of possible symptoms and SSAGA-positive controls (n = 37) showeda mean ± SD of 21.3% ± 8.6% of possiblesymptoms. Semi-Structured Assessment for the Genetics of Alcoholism–negativerelatives of probands showed a mean ± SD of 11.9% ± 3.1%of symptoms and SSAGA-negative controls showed a mean ± SDof 11.0% ± 3.2% of symptoms. Subjects with false-negativeSSAGA results (compared with best estimate results in the same person) showed14.8%±6.5% of symptoms, whereas subjects with true-negative SSAGA resultsshowed 11.5%±3.0% of symptoms. If we make the conservative assumptionthat SSAGA-negative subjects with a score of 18% or more are actually falsenegatives (only 2.3% of true negatives have that many symptoms), the prevalenceestimate of 3.6% in controls using raw data (Table 2) would increase to 5.4%; this tends to support our finalestimate of 5.3% in controls.
Because cohort effects have been reported for alcoholism and depression,28,31-33 wecontrolled our data for cohort effects (linear and quadratic) by dividingsubjects into decade of birth (Table 5).We also controlled for sex, ethnicity, ascertainment site, comorbidity inprobands, and comorbidity in relatives. These analyses include results forinterviewed subjects only (because the estimates for all relatives includeprojected data). Most disorders studied continue to show aggregation withthese effects accounted for, including alcohol dependence (by DSM-III-R, DSM-IV, Feighner et al,27 and ICD-10), alcohol abuse(by DSM-III-R), all forms of substance dependenceexcept for opiate dependence, ASPD, major depression, obsessive-compulsivedisorder, panic disorder, and PTSD. Thus, these disorders are found in increasedrates in relatives of alcoholic probands, independent of whether the probandhas the disorder or whether the relative has comorbid alcoholism. If we rerunthe analysis without controlling for comorbidity (data not shown), the categoriesof any anxiety disorder and opiate dependence are significantly aggregatedas well, suggesting that in these cases it is a comorbid disorder (alcoholdependence plus any anxiety disorder or alcohol dependence plus opiate dependence)that is familial. The correction for comorbidity applied herein is a conservativeone, testing the hypothesis that the pure form of one (noncomorbid) disorderis related to the pure form of the other. This underestimates the common geneticvariance represented by familial comorbid illness, which is substantial.
Alcohol dependence diagnosed by any of 4 criterion sets is clearly familial.The risk ratio we report herein is somewhat lower than has been reported inother family studies (those reviewed by Nurnberger and Gershon34 andMerikangas and Risch35). This is primarilyrelated to our control values, which range (Table3) from 10.76% (ICD-10) to 21.88% (Feighneret al27). Relative risk increases modestly,progressing from the more inclusive criteria of Feighner et al to the morerestrictive criteria of ICD-10. The high rates incontrols are also seen in the raw data (15.93% for DSM-III-R and 16.92% for Feighner et al), and are slightly increased by agecorrection, correction for SSAGA false negatives and false positives, andinclusion of uninterviewed relatives. Of course, these are lifetime diagnosesand quite often do not represent an active drinking problem. The best approximationswe have for the true rate of DSM-IV alcohol dependencein controls come from the data summarized in Table 4 (14.4% in all relatives). This is comparable to the SSAGA-correctedrate of 13.56% in Table 3 (our bestestimate of true prevalence in interviewed relatives only).
It may be argued that the present set of controls has been selectedfor health, in that the participation of spouse and children were required.However, the rate of DSM-III-R alcohol dependencewe report in controls, based on direct interview (17.3% [Table 4]), is similar to the rate reported by Kessler et al (14.1%)in the National Comorbidity Survey.20 Ratesof major depression, drug dependence, mania, and ASPD (17.5%, 7.6%, 0.5%,and 3.3%, respectively) are also generally comparable to National ComorbiditySurvey rates (17.1%, 7.5%, 1.6%, and 3.5%, respectively). The alcohol-dependentprobands, with 4 participating first-degree relatives, may also be relativelyhealthy because they represent affected persons with some family ties. Inthat regard, the substantially increased rates of illness in family membersare even more remarkable.
We may also ask about the true RR of alcohol dependence in relativesof alcoholic probands compared with controls. It seems that it is safe (andconservative) to estimate an RR of about 2. Higher rates are seen with ICD-10 and DSM-IV criteria in Table 2, but they are decreased when best estimateprocedures are used. Estimates of RR in siblings alone (the Risch λ)also give an estimate of about 2 (data not shown).
Alcohol abuse (DSM-IV) does not cluster infamilies of alcohol-dependent probands. Semi-Structured Assessment for theGenetics of Alcoholism–corrected rates are 4.0% and 3.5% for relativesof probands and controls, respectively, for the DSM-III-R definition of abuse, and 12.6% and 11.2%, respectively, for the DSM-IV definition. Cocaine abuse, opiate abuse, sedativeabuse, and stimulant abuse also show no difference between familial groupsin SSAGA-corrected rates. Rates of abuse are low in relation to rates of dependence.In controls, corrected data show 6-fold more subjects with DSM-III-R alcohol dependence than abuse; the comparable value forDSM-IV alcohol dependence and abuse is 1.2-fold; for cocaine,2-fold; for marijuana, 3-fold; for opiates, 2-fold; for sedatives, 2-fold;and for stimulants, 2-fold. In relatives of alcohol-dependent probands, thevalues are generally higher (eg, 2-fold for DSM-IV and 9-fold for DSM-III-R alcohol dependencevs alcohol abuse). It may be that abuse is underdiagnosed using the presentcriteria, that dependence is overly diagnosed, or that both are true. On theother hand, it may be more accurate to think of these disorders as truly independentof each other. These issues have also been discussed by Grant and colleagues.36-40
In contrast to abuse, all forms of nonalcohol substance dependence showaggregation in relatives of alcoholic probands, including cocaine (RR, 3.1),marijuana (RR, 1.8), opiates (RR, 2.5), sedatives (RR, 2.0), stimulants (RR,2.7), and tobacco (RR, 2.2). In fact, the RR for any form of drug dependenceexcluding tobacco is 2.3, which is equal to or greater than the RR for alcoholdependence by any definition. This is consistent with studies showing evidencefor a generalized genetic predisposition to substance dependence41-44 aswell as specific factors related to alcohol dependence. Support for specificgenetic factors for substance dependence other than alcohol would requireprobands with other forms of substance dependence (which is beyond the scopeof this study).
The excess of ASPD diagnoses in relatives of alcoholic probands is consistentwith many previous studies. The prevalence of ASPD in controls in this studyis relatively high. Estimates in these data vary from 3.3% in interviewedrelatives to 5.2% in all relatives to 6.2%, applying corrections for SSAGAfalse negatives and false positives. However, estimates of ASPD in relativesof alcoholic probands are consistently higher (7.1%, 8.1%, and 8.8%, respectively).
Individual anxiety disorders that remain modestly but significantlyaggregated after controlling for multiple factors include obsessive-compulsivedisorder, panic disorder, and PTSD. The rates of some of these disorders inrelatives were determined as part of previous reports,45,46 anddo not seem substantially different in the present (expanded) data set. Theexcess of anxiety disorders in relatives cannot, in general, be explainedby the presence of an anxiety disorder in the proband (except for the categoryof any anxiety disorder). Previous studies of PTSD have shown ambiguous resultsin the assessment of the familial relationship with alcohol dependence.47,48
There is a modest excess of major depression (odds ratio, 1.35) in relativesof alcoholics after controlling for multiple factors. In a separate analysis,the rate of comorbid major depression was not elevated in alcoholic probands,although the rate of secondary depression (depression in the context of heavydrinking or other organic precipitants) was elevated.49 Secondarydepression also seemed to be elevated in relatives of alcoholic probands inthat study, and comorbid alcoholism and depression aggregated in families.There is no evidence for aggregation of mania in relatives of alcoholic probands(Table 4). Comorbid mania does seemto be elevated in alcoholic subjects themselves in the Collaborative Studyon the Genetics of Alcoholism data set.50 Comorbidalcoholism and mania are also more likely to appear in relatives of comorbid(alcohol-dependent and manic) probands.50
Attention-deficit/hyperactivity disorder23 andbulimia51 have been reported in some previousstudies to be related to alcohol dependence. We cannot confirm a relationshipin this population.
The addition of family history data on uninterviewed relatives resultedin minor adjustments in prevalence estimates (the exception being DSM-IV alcohol dependence in controls), suggesting that the diagnosticprofile of uninterviewed relatives was fairly similar to that of the interviewedrelatives.
An excess of men among subjects diagnosed as having externalizing disordersand an excess of women among those diagnosed as having depression would beexpected from previous studies.
Variation in rates of illness by ascertainment site may reflect differentsources for controls at different sites (eg, motor vehicle records vs dentalclinics). Our multivariate analysis considered site as a confounding variableand still revealed significant coaggregation of multiple disorders.
Family studies by their nature include environmental effects, genetic-environmentalinteractions, and genetic effects. One complex effect is that of assortativemating, which is known to occur in families with cases of alcohol dependence.52,53 These effects may never be completelycontrolled for; analysis of sibling pairs, however, eliminates the effectof multigenerational assortative mating and shows generally similar resultsto analysis of all first-degree relatives.
Family studies may suggest new phenotypes for genetic linkage and associationstudies. It would be useful to consider ASPD or the combination of ASPD andalcohol dependence as a genetic phenotype. Linkage studies of habitual smokingin this sample have been reported,54 as havelinkage studies of alcoholism and/or depression.49 Becauseevidence in the present analysis suggests that anxiety disorders (specifically,panic disorder, PTSD, and any anxiety disorder) aggregate in relatives ofalcoholics independent of comorbidity, it would seem useful to test anxietyas a possible alternate phenotype within families with alcohol dependence.
Another, and perhaps more general, role of family studies is to definethe familial/genetic relationships between disorders. A disorder more commonin relatives than controls may share specific genetic vulnerability factorswith the illness in the proband. In combination with twin studies, we maythink of the genetic spectrum of alcohol dependence as including not onlyASPD but also multiple forms of drug dependence and some forms of depressiveand anxiety disorders. This familial coaggregation is distinct in origin andsignificance from comorbidity (multiple disorders in the same person), whichmay result from secondary effects of one disorder on another. Coaggregationin families is more likely to represent shared genetic variance.
Correspondence: John I. Nurnberger, Jr,MD, PhD, Institute of Psychiatric Research, Indiana University School of Medicine,791 Union Dr, Indianapolis, IN 46202-4887 (email@example.com).
Submitted for Publication: August 12, 2003;final revision received April 23, 2004; accepted April 30, 2004.
Funding/Support: This national collaborativestudy is supported by grant U10AA08403 from the National Institute on AlcoholAbuse and Alcoholism, Bethesda, Md.
Previous Presentation: This study was presentedin part at the Research Society for Alcoholism meeting; June 29, 2002; SanFrancisco, Calif, and June 23, 2003; Ft Lauderdale, Fla.
Acknowledgment: The Collaborative Study onthe Genetics of Alcoholism (principal investigator: H. Begleiter, coprincipalinvestigators: L. Bierut, H. Edenberg, V. Hesselbrock, B. Porjesz)includes9 different centers where data collection, analysis, and storage occur. The9 sites and principal investigators and coinvestigators are as follows: Universityof Connecticut, Hartford (V. Hesselbrock); Indiana University School of Medicine,Indianapolis (H. Edenberg; J. Nurnberger, Jr; P. M. Conneally; and T. Foroud);University of Iowa, Iowa City (R. Crowe and S. Kuperman); State Universityof New York Health Sciences Center at Brooklyn (B. Porjesz and H. Begleiter);Washington University, St Louis, Mo (L. Bierut, J. Rice, and A. Goate); Universityof California at San Diego (M. Schuckit); Howard University, Washington, DC(R. Taylor); Rutgers University, Piscataway, NJ (J. Tischfield); and SouthwestFoundation, San Antonio, Tex (L. Almasy). Lisa Neuhold serves as the NationalInstitute on Alcohol Abuse and Alcoholism staff collaborator. This nationalcollaborative study is supported by National Institutes of Health grant U10AA08403from the National Institute on Alcohol Abuse and Alcoholism.
In memory of Dr Reich, coprincipal investigator of the CollaborativeStudy on the Genetics of Alcoholism since its inception and one of the foundersof modern psychiatric genetics, we acknowledge his immeasurable and fundamentalscientific contributions to the Collaborative Study on the Genetics of Alcoholismand the field.