Genetic Epidemiology of Alcohol-Induced Blackouts

Background  Alcohol-induced blackouts (ie, periods of anterograde amnesia) have received limited recent research attention.

Objective  To examine the genetic epidemiology of lifetime blackouts and having had 3 or more blackouts in a year, including analyses controlling for the frequency of intoxication.

Design, Setting, and Participants  Members of the young adult Australian Twin Register, a volunteer twin panel born between January 1, 1964, and December 31, 1971, were initially registered with the panel as children by their parents between 1980 and 1982. They underwent structured psychiatric telephone interviews from February 1996 through September 2000. The current sample contains 2324 monozygotic and dizygotic twin pairs (mean [SD] age 29.9 [2.5] years) for whom both twins' responses were coded for blackout questions and for frequency of intoxication.

Main Outcome Measure  Data on lifetime blackouts and having had 3 or more blackouts in a year were collected within an examination of the genetic epidemiology of alcoholism.

Results  A lifetime history of blackouts was reported by 39.3% of women and 52.4% of men; 11.4% of women and 20.9% of men reported having had 3 or more blackouts in a year. The heritability of lifetime blackouts was 52.5% and that of having had 3 or more blackouts in a year was 57.8%. Models that controlled for frequency of intoxication found evidence of substantial genetic contribution unique to risk for the blackouts and a significant component of genetic risk shared with frequency of intoxication.

Conclusions  The finding of a substantial genetic contribution to liability for alcohol-induced blackouts including a component of genetic loading shared with frequency of intoxication may offer important additional avenues to investigate susceptibility to alcohol-related problems.

Alcohol-induced blackouts (ie, periods of anterograde amnesia) are reportedby most, but not all, alcohol-dependent individuals.^1-4 Blackoutsalso have been found to be a fairly common consequence of heavy alcohol useamong those who are not alcohol dependent.^5-10 Incollege students, the risk for blackouts has been reported to increase withbinge drinking frequency.¹⁰ Factors associatedwith risk for blackouts in other samples include drinking on an empty stomach,^1,2 gulping drinks,^1,2 benders,^1,2 drinking when fatigued,² priorbrain insult or injury,¹ and alcohol consumptionvariously operationalized as quantity,^1,2,6,8,9 frequency,⁸ and frequency of intoxication.^4,5

One group attempted¹¹ to elicit blackoutsin a controlled setting in 10 subjects who were recruited based on self-reportof being healthy and able to drink heavily and quickly (8 were diagnosed asbeing alcoholic; 5 reported a history of blackouts). Their protocol beganwith 2 days of inpatient clinical assessment. On day 3, subjects fasted for5 hours and then drank 16 to 18 oz of bourbon over a 4-hour period. Memorytesting, begun 1 hour after the start of drinking, continued for 4 hours.The next day additional testing was conducted to determine whether the subjectscould remember stimuli from the prior day's testing. Of the 10 subjects, 5displayed deficits in short-term memory for both neutral and emotional stimulithat persisted to the next day. Their blood alcohol concentrations rose morequickly and reached significantly higher peaks than those of subjects whodid not experience memory loss. The most intriguing observation was that the5 subjects in whom blackouts were induced in the laboratory setting were thesame 5 individuals who had reported having had prior blackouts. These dataprovide evidence for substantial variation in susceptibility to alcohol-inducedblackouts that is perceptible even among individuals who drink heavily.

The current investigation used data collected in recently completedtelephone interviews of the Australian Twin Register^12,13 toexamine whether there was a significant genetic contribution to the variancein susceptibility to alcohol-induced blackouts. The assessment included questionsregarding lifetime history of alcohol-induced blackouts and having had 3 ormore blackouts in a single year. Because these analyses were conducted withdata from a genetically informative twin sample, we were able to estimateadditive genetic, shared (familial) environmental, and nonshared (individual-specific)environmental contributions to the risks for these blackout measures. Becausealcohol consumption has been reported to be moderately heritable, additionalanalyses were undertaken to determine whether the genetic contribution tothe blackout measures remained significant with control for alcohol consumption.We chose to use the respondents' frequency of intoxication for this purposeas prior reports have found this consumption measure to be associated withincreased risk for blackouts. Our hypothesis was that a significant residualgenetic component to risk for each blackout measure would be observed evenafter controlling for the frequency of intoxication during the heaviest drinkingperiod.

The study's methods have been described in detail elsewhere.^12,13 A summary is provided below.

Subjects were members of the young adult cohort of the Australian TwinRegister, a volunteer panel of twins born between January 1, 1964, and December31, 1971. Almost all were registered with the panel between 1980 and 1982by their parents in response to approaches through either school systems ormass media appeals. All data reported herein are from a comprehensive diagnostictelephone assessment completed from February 1996 through September 2000 bytrained lay interviewers. Prior to the interview, verbal consent was obtainedaccording to the terms of approval from the institutional review boards ofWashington University School of Medicine, St Louis, Mo, and Queensland Instituteof Medical Research, Brisbane, Australia. Of 4010 pairs that were located,interviews were completed with both members of 2765 pairs (69% pairwise responserate) and with 1 member of an additional 735 pairs (78% individual responserate).¹² The most common reasons for nonparticipationincluded refusal by twin, incapacitation and/or death, and lack of availablecontact information. Singletons were excluded from analyses as were pairsin which one or both twins reported never having used alcohol. The currentsample included the remaining 2324 monozygotic (MZ) and dizygotic (DZ) same-sexpairs (mean [SD] age, 29.9 [2.5] years) in which both twins had responsescoded for the blackout questions. The sample included 542 MZ female, 432 MZmale, 418 DZ female, 373 DZ male, and 559 DZ opposite-sex pairs.

A standardized psychiatric diagnostic assessment, an adaptation of theSemi-Structured Assessment for the Genetics of Alcoholism,¹⁴ wasadministered by telephone. The interview enabled lifetime DSM-IV¹⁵ diagnoses of psychiatric disordersto be made and included nondiagnostic sections as well. The current analysesfocused on questions that assessed alcohol-induced blackouts: "Have you everhad blackouts when you didn't pass out while drinking—that is, you drankenough so that the next day you couldn't remember things you had said or done?"Those responding affirmatively were asked: "Have you ever had 3 or more blackoutsin a 12-month period?" Those who reported never having had at least 5 drinksin a 24-hour period or who denied both having at least 1 drink monthly for6 consecutive months and ever having been drunk were skipped out of the alcoholdiagnostic section without having been asked the blackout questions. Theserespondents were assumed never to have had a blackout and were coded as such.

The interview included several questions that assessed various aspectsof alcohol consumption. We chose to control for respondents' frequency ofintoxication in analyses because of this measure's previously reported associationwith risk for blackouts. Frequency of intoxication was assessed as follows:"Think about the period in your life lasting at least 12 months when you weredrinking the most. During that period, how often did you get drunk, that is,your speech would be slurred or you would be unsteady on your feet?" The 10response categories were collapsed into the following 6 based on responsefrequency and distribution: (1) less than 3 days per year, (2) 3 to 12 daysper year, (3) 2 to 3 days per month, (4) 1 day per week, (5) 2 days per week,and (6) 3 or more days per week. Onset and recency information available forthis measure and for 3 or more blackouts in a year (but not lifetime blackouts)revealed that these periods overlapped for 95.6% of respondents.

Primary statistical analyses were performed using SAS¹⁶ andthe Stata Statistical Software: Release 6.0.¹⁷ An αlevel of .05 was required for statistical significance. Stata's Huber-Whiterobust variance estimator option was used to obtain 95% confidence intervals(CIs) adjusted for the statistical nonindependence of twin-pair observations.Lifetime prevalence of blackouts and of having had 3 or more blackouts ina year was examined by sex. Logistic regression was used to evaluate the relationshipof each blackout measure with lifetime alcohol dependence in women and inmen; the relationship of each blackout measurement with frequency of intoxicationwas similarly evaluated.

The genetic and environmental contributions to twin-pair resemblancewere estimated using a normal liability threshold model.^18,19 Forbinary or ordinal variables, the estimates reflect the resemblance of pairmembers in terms of their liability to display a phenotype. Liability is assumedto have an underlying continuous and normal distribution in the populationwith the phenotype expressed in individuals whose risk exceeds a superimposedthreshold (estimated from the data).

The twin model assumes that individual differences arise from additivegenetic, shared (familial) environmental, and nonshared (individual-specific)environmental sources. The degree to which variance arises from each of thesesources is determined based on their expected effects on twin-pair resemblance.Thus, while members of an MZ pair are genetically identical, DZ twins will,on average, share half of their genes. Shared environmental influences areassumed not to vary on the basis of zygosity. Based on these assumptions,structural equations are developed (cov_MZ = v_G +v_SE;cov_DZ = ½ V_G + V_SE)that allow estimation of genetic and shared environmental contributions tovariance in liability to blackouts. (cov indicates covariance; v, variance; G, additive genetic;and SE, shared environmental.) Nonsharedenvironmental (NSE) effects will contribute only to differences within pairsand are estimated as v_NSE = (1 − v_G − v_SE).With inclusion of both female and male same-sex pairs, separate parametervalues can be estimated for women and men. Furthermore, the inclusion of datafrom DZ opposite-sex pairs enables estimation of a sex-specific additive geneticcontribution representing the effects of a set of genes specific to eithersex (in the current report's univariate models, the contribution of a male-specificeffect was determined).

For all genetic analyses, models were fitted using the structural equationmodeling program MX,²⁰ which uses maximum likelihoodestimation to calculate parameter values and likelihood-based 95% CIs. Thefit of models was compared using the goodness-of-fit χ² testand the Akaike Information Criterion values to select the model that bestcombined goodness-of-fit and parsimony. When the fit of 2 models did not differsignificantly, the model with fewer parameters was considered to be the moreparsimonious. The current report provides only the results of the best-fittingmodels for each set of analyses; more detailed results including those ofother models are available from the corresponding author on request. Separateunivariate analyses were undertaken using contingency table data for lifetimeblackouts, having had 3 or more blackouts in a year, and frequency of intoxication.

Because genetic factors were found to affect the risk for frequencyof intoxication, 2 methods were then used to control for its effects on thevariance for each blackout variable. The first approach fitted bivariate Choleskydecomposition models to data on frequency of intoxication and each blackoutvariable in turn. In these models, additive genetic, shared environmental,and nonshared environmental contributions to the variance of each variablewere estimated and the variance of the blackout variable was further decomposedinto unique components and those also loading on frequency of intoxication.This approach explores possible shared contributions to the liability in multiplemeasures arising from 1 or more latent factors. It takes advantage of thefact that the cross-correlation between frequency of intoxication in one twinand a blackout variable assessed in the co-twin is expected to vary as a functionof the degree to which these 2 variables share genetic vs shared environmentalinfluences. For example, a 2:1 ratio of MZ/DZ twin pair cross-correlationsis predicted if common genetic, but not common shared environmental, factorsare contributing to risk. Furthermore, if genetic influences on 2 variablescompletely overlap, the genetic contribution to the correlation of the 2 variablesshould be the geometric mean of the genetic variances of the 2 variables;the extent to which the estimated genetic contribution is attenuated fromthe geometric mean provides information about the extent to which there isonly partial overlap of genetic influences, allowing decomposition of thetotal genetic variance in the blackout variable into a component specificto blackouts and a component that reflects genetic influences associated withthe frequency of intoxication.²¹ The full model,in which all parameter values were allowed to vary between women and men,was again first calculated. Reduced models that dropped terms and constrainedestimates between women and men were then calculated with model fit comparedby the likelihood ratio χ² test.

In the second approach, genetic models were fitted in which the normalliability threshold model for the blackout variable was modified to controlfor the regression of the blackout variable on the reported frequency of intoxication.For these analyses, frequency of intoxication was recoded as 5 binary dummycontrol variables (with the reference group being <3 days per year), toallow for possible nonlinearities in the relationship between frequency ofintoxication and risk of blackouts. We modeled jointly the probit regressionof each blackout variable on the frequency of intoxication variables, andthe genetic and environmental contributions to the residual variance in theblackout variable. In this approach, the threshold value for each individualis defined by a probit regression equation, in which the intercept correspondsto the sex-specific mean threshold, and the β coefficients estimate theprobit regression of the dependent variable on the control variables. Thus,an individual with a reported high frequency of intoxication will be morelikely to experience blackouts, corresponding to a lower threshold on theliability distribution. This process tests for residual genetic and environmentalcontributions to variance in risk for the blackout variable, controlling forthe regression of the risk for that blackout variable on the frequency ofintoxication. It, thus, removes any effects related to the frequency of intoxicationon the risk for the blackout variables without directly estimating the degreeof shared underlying liability. The fit of the full model and that of reducedmodels was again compared using the likelihood ratio χ² test.

A lifetime history of blackouts was reported by 39.3% of women and 52.4%of men (odds ratio [OR], 1.71; 95%CI, 1.50-1.93). Having had 3 or more blackoutsin a year was reported by 11.4% of women and 20.9% of men (OR, 2.06; 95%CI,1.73-2.45).

In both women and men, having had a blackout was significantly associatedwith a lifetime diagnosis of alcohol dependence (Table 1) and similar, stronger associations were seen with havinghad 3 or more blackouts in a year. Blackouts also were common among non–alcohol-dependentindividuals with prevalence rates of 31.5% in woman and 40.3% in men. Thosereporting having had a blackout also had more frequent drinking to intoxicationduring their heaviest drinking year (ORs ranging from 2.51 [95% CI, 2.06-3.05]for 3-12 days per year to 16.37 [95% CI, 12.29-21.81] for ≥3 days per week).A similar pattern was noted for having had 3 or more blackouts in a year (ORsranging from 4.89 [95% CI, 2.67-8.96] for 3-12 days per year to 78.31 [95%CI, 43.90-139.70] for ≥3 days per week).

Separate univariate genetic models were fitted for lifetime blackouts,having had 3 or more blackouts in a year, and frequency of intoxication. Thebest-fitting model for each variable (Table2), obtained by dropping or constraining individual terms and determiningwhether the change in model fit was significant, retained only terms representingadditive genetic and nonshared environmental contributions to variance which,in each case, could be constrained to be equal across gender. The heritabilityestimates for these variables (with 95% CIs) were as follows: lifetime blackouts53% (95% CI, 45%-60%), 3 or more blackouts in a year 58% (95% CI, 48%-67%),and frequency of intoxication 43% (95% CI, 37%-48%).

Genetic and environmental contributions for the blackout variables aresubdivided in Table 3 into effectsthat were associated with differences in frequency of intoxication and effectsthat seemed to be specific to blackouts based on the results of fitting bivariategenetic models. The best-fitting models did not retain any shared environmentalterms and constrained all parameters to be equal in women and men. Evidencewas found for additive genetic and nonshared environmental contributions tothe variance in the blackout variables that also loaded on frequency of intoxicationand similar contributions specific to blackouts. For lifetime blackouts, thecomponents that also loaded on frequency of intoxication represented 59.6%[0.31/(0.31 + 0.21)] of the total additive genetic variance and 4.2% [0.02/(0.02+ 0.46)] of the total nonshared environmental variance. The best-fitting modelfor having had 3 or more blackouts in a year reduced to a similar form withan estimated 61.0% of the additive genetic and 14.6% of the nonshared environmentalvariance found to be shared with those for frequency of intoxication.

In analyses that controlled for the regression of the risk for lifetimeblackouts on the frequency of intoxication, there was again no evidence ofa significant shared environmental component nor of significant variationin parameters estimates by sex. The best-fitting model, thus, contained onlya residual additive genetic component that was somewhat attenuated (heritability= 37% [95% CI, 28%-46%]) and a nonshared environmental term. The correspondingmodel for having had 3 or more blackouts in a year could be similarly reducedyielding a residual heritability estimate of 41% (95% CI, 29%-53%).

The current investigation found evidence of a substantial genetic contributionto the risk for lifetime blackouts and for having had 3 or more blackoutsin a year (respective heritability estimates of 53% and 58%). Our sample'shigh alcohol consumption made it well suited to detect a trait such as blackoutsensitivity that is dependent on alcohol exposure for its expression. Giventhe current availability of animal models (eg, spatial working memory tasks)²² appropriate for the study of alcohol-induced blackouts,it is conceivable that the genetic loci underlying our findings could be readilyidentified.

The analyses controlling for the frequency of intoxication provide auseful frame of reference for consideration of potential pathways by whichgenetic influences on alcohol-induced blackouts may be mediated. The bivariateCholesky model-fitting approach is generally more conservative in its estimationof the additive genetic variance for the blackout variables that is not sharedwith frequency of intoxication with assignment to shared variance, for example,of that arising through the effects of a third variable on both risk of blackoutsand drinking to intoxication. The probit regression method more directly estimatesthe residual genetic and environmental contributions to the risk for blackoutswhen the effects of frequency of alcohol intoxication have been controlledstatistically. These 2 approaches confirm significant genetic contributionsto the risk of blackouts that are not explained by (probit regression approach)and not correlated with (bivariate model) genetic effects on drinking to intoxication.Thus, our results suggest the involvement of multiple factors in the susceptibilityto blackouts, some of which also contribute risk for frequency of alcoholintoxication. We will first consider potential sources of genetic risk forblackouts that are not shared with those for frequency of intoxication.

The genetic contribution unique to blackout risk most likely arisesfrom genes whose products mediate alcohol's effects on hippocampal neurotransmission.^23,24 Blackouts appear to result from thefollowing 2 actions of alcohol²³: (1) potentiationof γ-aminobutyric acidα (GABA_A)–mediated inhibitionand (2) antagonism at excitatory N-methyl-D-aspartate(NMDA) glutamate receptors. An investigation²⁵ inwhich subanesthetic doses of a benzodiazepine, midazolam hydrochloride, andketamine hydrochloride were given to humans found that their coadministrationresulted in a significantly greater memory deficit than that seen with eitherdrug alone, suggesting that these actions could operate in tandem.

Both GABA_A agonists²⁶ andbenzodiazepines²⁷ have been shown to causeperiods of anterograde amnesia in a dose-dependent fashion. The degree towhich alcohol has direct effects on hippocampal GABA_A receptorsremains controversial^28-30;however, evidence suggests that several routes exist by which indirect effectsmay be mediated (neuroactive steroids,^31,32 theseptohippocampal pathway,³³ and nicotinic acetylcholine[nACh] receptors).^34-36 Neuroactivesteroid levels increase markedly after alcohol ingestion and are known tolead to enhanced GABA_A transmission.^31,32 Whenadministered systemically to rats, neuroactive steroids produce a spatialreference memory impairment similar to that seen with alcohol.³² Directinfusion into the medial septum of either alcohol or GABA_A agonistsresults in impairment on memory tasks similar to that seen with systemic alcohol.³³ The current at α4β2 nACh receptors atpresynaptic and postsynaptic locations on GABA_A neurons, potentlyenhanced by alcohol, results in increased GABA release.^34-36 Thus,genes whose products have either direct or indirect involvement in GABA_A-mediated transmission could be contributing genetic risk specificto blackouts. Respective examples in mice include a point mutation introduced^37,38 into the GABA_A receptor's α-1subunit that resulted in elimination of the amnestic effects of benzodiazepinesand an α4β2 nACh receptor polymorphism found³⁹ tohave greater alcohol responsiveness.

The administration of NMDA antagonists have been reported to cause periodsof anterograde amnesia in humans²⁵ and rodents.⁴⁰ Impairment in both spatial and nonspatial hippocampus-dependentmemory has been observed in NMDA receptor subunit-1 knockout mice.^41,42 Transgenic mice overexpressing theNMDA receptor subunit-2B demonstrated superior learning and memory in contextualand cued fear conditional tasks.⁴³ The degreeto which alcohol inhibits NMDA receptors varies with their subunit composition⁴⁴ and is reduced by altering subunit 1.⁴⁵ Thus,polymorphisms that affect the structure of NMDA receptors,^41,42,45 theirsubunit composition,⁴⁴ or the degree to whichvarious receptor types are expressed⁴³ couldbe contributing genetic risk specific to blackouts.

Genetic factors contributing to individual differences in alcohol metabolismare one likely source of the risk shared between blackouts and frequency ofintoxication. A significant relationship has been reported⁴⁶ betweenalcohol consumption, variously assessed across multiple time points in anolder cohort of Australian twins, and genotype at the loci for alcohol dehydrogenase(ADH) isozymes ADH1B and ADH1C. The results were in a pattern that suggestedgreater alcohol consumption was associated with slower metabolism of alcoholto acetaldehyde. The phenotype, maximum 24-hour alcohol consumption, has beenreported to display evidence for linkage close to the ADH loci.⁴⁷ The finding of Goodwin et al¹¹ that the blood alcohol concentrations of those inwhom blackouts were induced rose more rapidly and reached higher peaks suggeststhat these subjects may have had slower alcohol metabolism and perhaps morerapid absorption. However, another report⁴⁸ foundthat genotype at ADH1B and ADH1C loci made only a limited contribution to postalcohol challenge testblood alcohol concentrations. Gastric emptying speed^49,50 andgastric ADH activity⁵¹ also contribute to individualdifferences in alcohol metabolism and, thus, could be a source of shared geneticvariance.

One major source of shared genetic risk not related to alcohol metabolismis likely to result from variation in the level of response to the sedativeeffects of alcohol. Specifically, those in whom alcohol is less sedating areable to drink more and, while doing so, will be predisposed to experiencea blackout rather than just passing out. Interest in this area (eg, lowerlevel of response also has been implicated as a risk factor for alcohol dependence)⁵² has led to the selective breeding of mice and ratstrains on the basis of their level of response. Alcohol, at a dose that leadsto observable behavioral changes, was found³⁰ tosignificantly enhance the peak amplitude of hippocampal GABA_A inhibitorypostsynaptic currents in highly sensitive mice and rats while causing no significantchange in the inhibitory postsynaptic currents of the low sensitivity strains.Thus, genes that regulate GABA_A receptor sensitivity to alcoholappear to play a significant role in the sedative response. However, as withother alcohol-related phenotypes, level of alcohol response is a polygenictrait for which a number of contributing loci have already been identifiedin rodents.^53,54

When interpreting our results, a number of methodological limitationsmust be considered. Reliance on self-report may have led to bias due to false-negativeresponses in some participants (eg, those who drink alone) because awarenessof a blackout generally requires either that someone else be present to retainand report memory for events or that something memorable transpire (eg, theindividual was injured or has no recollection of having returned home). Itis unclear how those whose blackouts were due to a combination of drugs andalcohol responded in our assessment. Although some error could have been introducedby our coding as negative for blackouts those who were skipped out beforethe blackout questions because of their very limited alcohol use, a more conservativereanalysis of our data (coding these individuals as missing) only slightlydecreased the heritability estimates observed. It is possible that the totalnumber of lifetime blackouts would have been a better dependent variable,but it was not included in our assessment. Similarly, the frequency of severeintoxication, if available, might have better controlled for alcohol consumption.In terms of temporal association, the best choice for a control variable mighthave been the quantity of alcohol consumed to induce a blackout, an interesting,but not realistically obtainable, alternative. Finally, the generalizabilityof our results to other less heavily drinking populations remains to be established.

The finding of a substantial genetic contribution to blackout risk offersan important clue to the underlying susceptibility to alcohol-related problems.Given the great societal costs of alcohol misuse,⁵⁵ weare hopeful that our work will motivate attempts to identify in animals thegenes underlying our findings. Additional investigations could then be undertakento determine the degree to which this work is pertinent to alcoholism riskin humans. Even if blackout susceptibility is only indirectly associated withalcoholism risk, it may prove to be associated with greater levels of otheralcohol-related risk-taking behavior, victimization, and susceptibility toperiods of anterograde amnesia associated with other drugs (eg, benzodiazepines).

Corresponding author and reprints: Elliot C. Nelson, MD, MissouriAlcoholism Research Center, Department of Psychiatry, Washington UniversitySchool of Medicine, 40 N Kingshighway, Suite 1, St Louis, MO 63108 (e-mail: nelsone@psychiatry.wustl.edu).

Submitted for publication September 4, 2002; final revision receivedJune 9, 2003; accepted July 2, 2003.

This study was supported in part by grants AA00277 (Dr Nelson), AA07728,AA10249, AA11998, and AA13321 (Dr Health), DA00272 and DA12854 (Dr Madden),and AA13326 (Dr Martin) from the National Institutes of Health, Bethesda,Md, and grants 941177 and 981351 from the Australian National Health and MedicalResearch Council, Canberra (Dr Martin).