Oral Topiramate Reduces the Consequences of Drinking and Improves theQuality of Life of Alcohol-Dependent Individuals:  A Randomized Controlled Trial FREE

Topiramate, a sulfamate-substituted fructopyranose derivative, is efficaciousat both reducing craving and heavy drinking, and improving abstinence amongalcohol-dependent individuals.¹ Conceptually,topiramate's efficacy might be due to its contemporaneous actions at 2 neuronalsystems that, combined, reduce mesocorticolimbic dopamine activity, a crucialmechanism by which alcohol exerts its rewarding effects.² Topiramatefacilitates the inhibitory neurotransmitter γ-aminobutyric acid on anonbenzodiazepine receptor,³ thus decreasingthe extracellular release of dopamine in the midbrain.⁴ Additionally,topiramate might suppress mesocorticolimbic dopamine activity by antagonizingthe excitatory effects of glutamate receptors of the α-amino-3-hydroxy-5-methylisoxazole-4-propionicacid (AMPA) and kainate types on these dopamine neurons.^2- 8 Topiramate'santiglutaminergic effects may be more pronounced in chronic alcoholics, comparedwith nonpathological drinkers, because they have enhanced binding sites ofboth AMPA and kainate receptors.⁶

Despite these exciting neuropharmacological effects of topiramate ondrinking behavior, an important question pertaining to its overall clinicaleffectiveness remains. Does topiramate's efficacy at improving drinking outcomesresult in an appreciable improvement in quality of life or a reduction inthe harmful psychosocial consequences of alcohol? This question reflects thefact that the deleterious psychosocial consequences of pathological drinkingon social, occupational, or recreational activities are a defining characteristicof alcohol dependence syndrome.⁹ Indeed, thepersistence of these harmful consequences of pathological drinking is criticalto the concept of alcoholism as a chronic disease.¹⁰ Aspharmacotherapy is targeted at the narrow focus of reducing the "symptom"of drinking, formal psychotherapies are often coadministered to reduce theharmful psychosocial consequences of pathological drinking. Yet, it is unknownwhether it is sufficient to reduce the harmful psychosocial consequences ofalcohol dependence simply by producing a marked reduction in drinking. Furthermore,studies that have yoked formal psychotherapy with pharmacotherapy are unableto address this question properly because the treatment effects observed fromthe "dose" of psychotherapy are at least equivalent to, and may be largerthan, those resulting from the "dose" of pharmacotherapy. The present studyprovides a unique opportunity to address the question of whether drinkingreductions can, at least in the short term, bring about significant reductionsin the harmful psychosocial consequences of drinking. In this case, the pharmacotherapywas not coupled with psychotherapy but rather with a compliance-enhancementintervention to increase observance with taking the medication.

In this study we examine whether pharmacotherapy with topiramate, comparedwith placebo, in alcohol-dependent individuals receiving a standardized medicationcompliance treatment, is associated with a reduction in the harmful psychosocialconsequences of drinking and an improvement in quality of life. We also examinethe relationship between changes in psychosocial function and heavy drinking.

We enrolled 150 men and women who had been diagnosed with alcohol dependenceaccording to the DSM-IV¹¹ andwho were current drinkers. Subjects were 21 to 65 years old, scored 8 or higheron the Alcohol Use Disorders Identification Test^12- 14 (anassessment of personal and social harm consequent to alcohol consumption),and drank 21 or more (women) and 35 or more (men) standard alcohol drinksper week during the 90 days prior to enrollment. One standard drink was definedas 0.35 L of beer, 0.15 L of wine, or 0.04 L of 80-proof liquor. Enrolledsubjects had a negative urine toxicological screen for narcotics, amphetamines,cannabinoids, or sedative-hypnotics at enrollment. We excluded individualswith a current DSM-IV Axis 1 diagnosis other thanalcohol or nicotine dependence, with clinically significant alcohol-withdrawalsymptoms or physical abnormalities, or who were compelled to receive alcoholtreatment. Subjects were also excluded if they had received alcohol treatmentwithin 30 days of recruitment or were pregnant or lactating. Abstinence atstudy entry was not an enrollment criterion but a treatment goal.

We received ethical approval for this study from the institutional reviewboard at The University of Texas Health Science Center at San Antonio. Werecruited individuals who responded to a newspaper or radio advertisementseeking participants for an alcohol treatment study between December 29, 1998,and April 11, 2001.

At week 0 (baseline), after providing written informed consent, we establishedthat the subjects were physically healthy from their medical history, physicalexamination results, electrocardiogram, and hematological and biochemicaltest results. Women took a urine pregnancy test to confirm that they werenot pregnant. We assessed the overall severity of alcohol dependence withmeasures of that component on the Clinical Global Impression Scale (CGI-S).¹⁵ The CGI-S is a 7-point scale that ranges from 1 (reportedlyabstinent and not seeking alcohol or "not addicted") to 7 (reportedly drinkingmore and constantly seeking alcohol or "extremely, severely addicted"). Wemeasured the psychosocial impact of drinking on multiple domains of generallifestyle with the Quality of Life Enjoyment and Satisfaction Questionnaire(Q-LES-Q).¹⁶ The Q-LES-Q is composed of 93items, each containing responses on a 5-point scale that ranges from 1 ("notat all or never") to 5 ("all the time"). Ninety-one of these items were groupedinto 8 summary scales.¹⁶ Five of the 8 summaryscales were scored for all subjects and included physical health/activities(13 items; maximum score = 65); subjective feelings (14 items; maximum score= 70); leisure time activities (6 items; maximum score = 30); social relationships(11 items; maximum score = 55); and general activities (14 items; maximumscore = 70). We scored the other 3 summary scales when applicable, and theseincluded work (13 items; maximum score = 65); household duties (10 items;maximum score = 50); and school/coursework (10 items; maximum score = 50).We measured satisfaction with medication and overall life satisfaction andcontentment on individual item scales, each with a maximum score of 5. Wemeasured the specific and direct harmful consequences of drinking using theDrinker Inventory of Consequences (DrInC).¹⁷ Eachof the 50 DrInC items was assessed on a 3-point scale ranging from "neveror once or a few times" to "daily or almost every day." These 50 DrInC itemswere further subdivided into 6 subscales: physical consequences (8 items;maximum score = 24); intrapersonal consequences (8 items; maximum score =24); social responsibility consequences (7 items; maximum score = 21); interpersonalconsequences (10 items; maximum score = 30); impulse control consequences(12 items; maximum score = 36); and control items (5 items; maximum score= 15); as well as a total consequences scale (45 items; maximum score = 135).The total consequences scale was the sum of all subscales other than the controlitems subscale. A lower score denoted a decreased adverse consequence of drinkingas compared with a higher score. The control items subscale was the sum of5 reverse-scored validity items. We also measured self-reported drinking,past 90 days, using the timeline follow-back method.¹⁸

We enrolled eligible participants at the beginning of week 1. From week1 through week 12, we assessed participants weekly on the CGI-S. On a secondCGI item, Clinical Global Impression–Change (CGI-C) scale, we quantifiedthe change in global functioning on a 7-point scale ranging from 1 ("verymuch improved") to 7 ("very much worse"). We assessed participants every 3weeks on the Q-LES-Q and the DrInC. We measured self-reported drinking usingthe timeline follow-back method, and performed weekly safety checks that includedan assessment of vital signs (ie, blood pressure, pulse, and temperature),weight, and breath alcohol concentration each week. Additional health andsafety checks, including electrocardiograms; adverse events; concomitant medications;withdrawal symptoms; hematological, biochemical, pregnancy, and drug tests;and physical examinations, were performed at scheduled intervals from week1 through week 12.¹

Topiramate and matching placebo tablets were provided by Ortho-McNeilPharmaceutical, Inc, Raritan, NJ. From the beginning of week 1 through week12, escalating doses of matching placebo or topiramate (up to 300 mg/d) wereprovided according to the schedule described previously¹ asan adjunct to weekly brief behavioral compliance enhancement treatment (BBCET).The maximum topiramate dose was administered between week 8 and week 12. Subjectsadministered either topiramate or placebo received an identical number oftablets. Medication was dispensed in blister packs labeled with identification,study and visit numbers, and date. Medication packs returned at each weeklyvisit, along with the calendar-based, pill-taking schedule, were used to calculatethe pill count.

A standard minimum psychosocial adherence enhancement procedure, BBCETemphasizes that medication compliance is critical to changing the alcoholic'sdrinking behavior. Minimal interventions, such as the brief advice of Edwardset al,¹⁹ have proven to be effective and beneficialtreatments for alcoholism. We modeled our BBCET on the clinical managementcondition in the National Institute of Mental Health collaborative depressiontrial, which was used as an adjunct to the medication condition.^20- 21 Trainednurse practitioners performed BBCET weekly for 12 weeks using a standardizedmanual.²² Nurses' adherence to the protocolfor delivering BBCET was monitored by the same physician (N.A.-D.) for theduration of the study.

Our 4 outcome measures of psychosocial functioning over the 12-weektrial period were as follows: CGI-S, a Global Clinicians' Rating of the severityof alcohol dependence (completed weekly); CGI-C, a Global Clinicians' Ratingof improvement in psychosocial functioning from baseline (completed weekly,except at week 0); Q-LES-Q, an overall measure of psychosocial functioningand quality of life (completed at weeks 0, 3, 6, 9, and 12); and DrInC, self-ratingsof the specific and direct harmful consequences of drinking (completed atweeks 0, 3, 6, 9, and 12).

Data management was conducted according to the Food and Drug Administrationguidelines of Good Clinical Practice. Data quality (including double-dataentry) was supervised by a master's-level database coordinator and statistician.Individual subject plots were checked for unusual values and completeness.Outcome measure values were validated as correct against the case records.Data were analyzed using Statistical Analysis System version 8.1 (SAS InstituteInc, Cary, NC).²³

We defined 2 binary responses based on the CGI-S and CGI-C scales. Onthe CGI-S, an individual without a clinically significant addiction was definedas scoring either 1 (reportedly abstinent and not seeking alcohol) or 2 (reportedlydrinking less and occasionally seeking alcohol); individuals with higher scoreswere considered to have a clinically significant severity of alcohol dependence.Similarly, on the CGI-C scale an individual was defined as having a clinicallysignificant improvement if he or she scored 1 (very much improved) or 2 (muchimproved); individuals with higher scores were considered not to have experienceda clinically significant improvement. Treatment effects were estimated asthe odds ratio (OR) of topiramate vs placebo for these 2 longitudinal binaryresponses. Odds ratios measured the relative likelihood for achieving theseresponses between topiramate and placebo. The generalized estimating equationapproach²³ was used to account for the correlationof observations within individuals with autoregressive structure as the workingcorrelation matrix, as implemented by SAS PROC-GENMOD software. Analyses wereadjusted for subject baseline characteristics including age, sex, age at onset,and drinks per day during the 90-day period prior to enrollment.

We categorized Q-LES-Q summary scales as low and high scores. This categorizationwas based on the different percentiles of the maximum possible scores, rangingfrom the 50th to the 90th percentiles. For a study week, if a subject hada score higher than the given percentile of the maximum score, he or she wasdesignated to have a high score; otherwise, it was determined that the scorewas low. Pearlstein et al²⁴ suggested thata score of 70 or higher represents a "normal" quality of life. We observedthat increasingly stringent cutoff points demonstrated the contrasts betweenthe treatment groups more effectively (Figure1), so we chose the 90th percentile of the maximum score as thefinal discrimination criterion to do hypothesis testing. "High" scores wereconsidered to be indicative of improvement. We used the generalized estimatingequation approach to do repeated-measures testing on the summary scales ofphysical health/activities, subjective feelings, leisure time activities,social relationships, general activities, work, household duties, satisfactionwith medication, and overall life satisfaction and contentment. We used autoregressivewithin-subject correlation for determining robust estimates. Treatment effectswere ORs comparing the likelihood of high vs low scores in the topiramateand placebo groups. We adjusted all models for the baseline summary score,age, and sex. Using the 90th percentile of the maximum possible summary scoreas the cutoff point, we tested whether the ORs comparing the topiramate vsplacebo groups were equal to 1, with P values and95% confidence intervals (CIs). We also estimated ORs and the corresponding P values and 95% CIs at the last visit using logistic regression.We used exact logistic regression for testing the OR of satisfaction withmedication at the last visit, due to quasi-complete separation. We could notdo hypothesis testing on the summary scale of school/coursework owing to thesmall sample size.

For the analysis of the DrInC subscales, we applied logarithmic transformationsto the subscales plus 1 to enhance the normality of the residuals. We usedgeneral linear models and repeated-measures models with interaction of timeand treatment group to model slopes that estimated the change in DrInC scoreper unit increase in the study period (ie, every 3 weeks) within the topiramateand placebo groups. We used SAS PROC MIXED software²³ andadjusted for the intrasubject correlation by using a compound symmetry covariancestructure based on the Akaike Information Criterion and the Schwarz BayesianCriterion.²⁵ We estimated robust parameters.We contrasted the slopes of change in the scores per unit increase in thestudy period between the topiramate and placebo groups. We also used analysisof covariance to estimate and contrast the means of the topiramate and placebogroups at the 12th week for each subscale. All models were adjusted for thebaseline score, age, and sex.

Finally, we characterized the relationship between our psychosocialmeasures and drinking behavior for the topiramate group. Our purpose was toexamine whether the topiramate-associated reductions in heavy drinking wereassociated with predictable improvements in clinical condition and qualityof life, and a reduction in the harmful consequences of alcohol consumption.For the drinking component, we calculated the percentage of heavy drinkingdays (PHDD), defined as the days for which the number of drinks was 5 or greaterfor men and 4 or greater for women, divided by the number of study days. Wethen divided the PHDD metric into 4 fixed quartiles (0-25, 25-50, 50-75, and75-100). For the CGI-S and CGI-C scales, we calculated the number of subjectsin each drinking quartile who were categorized as "reportedly abstinent andnot seeking alcohol" or "significantly improved," respectively. For the Q-LES-Qscale, we calculated the number of subjects with "high" scores (ie, improvement)in each drinking quartile. Since the DrInC scale is a continuous variable,we displayed its trend across time alongside that of PHDD. Type I error wascontrolled by conducting only planned analyses.

In each group, 75 subjects received treatment (Figure 2). We have shown that at baseline the topiramate vs placebogroups had similar past-90-days baseline drinking levels (mean = 9.59, SD= 7.01 vs mean = 8.85, SD = 4.42 drinks/day); age (mean = 41.51, SD = 8.75vs mean = 42.05, SD = 8.83 years); sex distribution (30.6% vs 26.6% were women);ethnic distribution (61.3% vs 66.6% were white); social class (34.6% vs 37.3%were from social class 1); severity of addiction; and age at onset of problemdrinking (48% vs 44% were early-onset alcoholics).¹ Atstudy end, participants who received topiramate compared with those on placebohad significantly superior improvement on all drinking outcomes, including27% fewer heavy drinking days (P<.001).¹ These improvements in self-reported drinking werecorroborated by the objective biochemical marker of transient alcohol consumption,plasma γ-glutamyl transferase.¹ A meanof 83.0 (SD, 4.9) tablets were taken in the topiramate group, compared with82.0 (SD, 4.3) in placebo group. The adverse events reported more frequentlyfor topiramate compared with placebo recipients were dizziness (28.0% vs 10.7%; P = .01); paresthesia (57.3% vs 18.7%; P<.001); psychomotor slowing (26.7% vs 12.0%; P = .02); memory or concentration impairment (18.7% vs 5.3%; P = .01); and weight loss (54.7% vs 26.7%; P = .001). No serious adverse events occurred.¹

The CGI-S and CGI-C scores decreased over time, and after we adjustedfor subject characteristics at baseline, the estimated OR of topiramate vsplacebo was 2.63 for being "reportedly abstinent and not seeking alcohol,"with 95% CI from 1.52 to 4.53 (P = .001), and 2.17for being "significantly improved," with 95% CI from 1.16 to 2.60 (P = .01).

On the Q-LES-Q, the odds of obtaining "high" scores for the topiramatecompared with the placebo group increased with rising percentiles of the maximumpossible subscale score over the study period (Figure 1). Thus, increasingly strict criteria for improved qualityof life were accompanied by a corresponding rise in the Q-LES-Q scores inthe topiramate compared with the placebo group. For all Q-LES-Q summary scales,the estimated odds of "high" scores based on the 90th percentile cutoff pointwere higher in the topiramate compared with the placebo group, and achievedstatistical significance on the subscales of physical health/activities (3.32;95% CI, 1.48-7.46; P = .004); subjective feelings(3.57; 95% CI, 1.77-7.22; P<.001); general activities(2.37; 95% CI, 1.07-5.25; P = .03); and overall lifesatisfaction and contentment (2.81; 95% CI, 1.21-4.29; P = .01) across the trial period; and at study end (Table 1). For the repeated-measures analyses, the ORs ranged from1.5 to 8.7, and for the endpoint analyses, the ORs ranged from 1.6 to 6.8.For the satisfaction with medication summary scales, there was a quasi-completeseparation of data points at the last visit; only 1 subject had a "high" score(5.6%) in the placebo group and 15 subjects had a "high" score (44.1%) inthe topiramate group. Thus, we used exact logistic regression to test thenull hypothesis. We were unable to model school/coursework data because ofthe reduced sample size, as only a few of our subjects were attending schoolor participating in ongoing academic activities (5 in the placebo group and6 in the topiramate group at study end). Nevertheless, at study end, therewas 1 subject in the topiramate group with a "high" score and 0 in the placebogroup.

Table 2 presents a reductionfor both groups on all the DrInC subscales at study end, with topiramate-associateddecreases being significantly greater than those for placebo.

For descriptive purposes, we explored the relationship between PHDDand CGI-C, overall life satisfaction on the Q-LES-Q, and total DrInC scalesover the study period. Figure 3 showsthat between baseline and study end there was a shift in the number of subjectsfrom the higher quartiles of PHDD to the lowest quartile, and a correspondingincrease in the number of individuals who were "significantly improved" inthe lowest quartile. Similarly, Figure 4 showsa shift in the number of subjects from the higher quartiles of PHDD to thelowest, and a corresponding increase in the number of individuals who had"high" scores on the Q-LES-Q in the lowest PHDD quartile. Figure 5 is consistent with these findings, showing an almost parallelrelationship between the reduction in the harmful consequences of drinkingand the reported PHDD across the study period. On aggregate,Figures 3 to 5show that there was a dramatic trend in the reduction of heavy drinking betweenbaseline and study end for recipients of topiramate. Further, there was apredictable and apparently direct relationship between the reduction in PHDDand an improvement in the individuals' clinical condition and quality of life,as well as a reduction in the harmful consequences of alcohol consumption.

Our results show that topiramate is more effective than placebo at improvingthe quality of life and overall clinical condition and at reducing the severityof addiction and harmful consequences of heavy drinking. Topiramate's effectat improving psychosocial functioning was robust, with an increasing trendtoward better outcomes as treatment progressed. Strikingly, these reductionsin the harmful consequences of drinking displayed a similar trend to the reductionin the PHDD. Thus, as heavy drinking was reduced, more individuals experiencedan improvement in psychosocial functioning. We did not do a formal correlationaltest between psychosocial functioning and heavy drinking as these variablesare obviously colinear. Further, due to the colinearity of these variables,it is not possible for us to attribute any cause-and-effect prediction tothe relationship between psychosocial improvements and reduced heavy drinking.It is tempting to speculate that the reductions in heavy drinking might explaina substantial part of this effect. This is because the principal target ofthe "minimum" behavioral intervention was to enhance compliance with medicationrather than to target any antecedents, triggers, or perpetuating psychosocialfactors associated with the heavy drinking. Nevertheless, the possibilitythat the "minimum" behavioral intervention promulgated some ongoing psychosocialimprovement cannot be discounted entirely because such brief treatments havebeen shown previously to be effective treatment for alcoholism in a varietyof settings.^26- 27 Also, we didnot have an appropriate "control" group—that is, participants who receivedthe "minimum" intervention alone (without adjunctive placebo medication);therefore, the independent effects of our "minimum" intervention cannot beteased out from the effects of the medication alone.

We propose that topiramate's effectiveness at improving psychosocialfunctioning in alcohol-dependent individuals is not medication specific; suchan effect could also be expected for other pharmacotherapies if they had similarefficacy. For instance, findings from a large open-label trial²⁸ suggestthat psychosocial improvements among alcoholics receiving acamprosate occurirrespective of the nature of psychosocial support. Topiramate has, however,not been compared directly with other putative therapeutic medications fortreating alcoholism; its comparative efficacy is unknown. Nevertheless, basedon the limited data from this single clinical trial, the effect size for topiramate'streatment efficacy suggested that it is at least as efficacious as other promisingmedications such as naltrexone or acamprosate for treating alcoholism.²⁹ If this is the case, the central issue with pharmacotherapyresearch might not be whether it is coupled with a "minimum" or "maximum"psychosocial intervention, but rather whether the combined treatment producesa substantial decrease in drinking. That is, the harm of the alcohol dependencesyndrome can be lessened during treatment irrespective of the type of psychosocialintervention and pharmacotherapy, so long as it is effective at reducing heavydrinking. The important implication here is that even though abstinence isthe "gold standard" of alcoholism treatment and was the goal of this study,a harm-reduction strategy based on reducing heavy drinking may be a worthwhiletreatment goal, especially if the patient will not or cannot become abstinent.Indeed, a recent influential review of the literature has suggested that harmreduction strategies might be as beneficial as abstinence-oriented approaches,particularly if treatments are directed to accommodate the preferences andneeds of the individual or target populations.³⁰ Theconcern often raised with this harm-reduction approach is that alcoholicswho drink even at "social" levels (1 and 2 drinks per day for women and men,respectively) might be at greater risk of relapse than abstinent individuals³¹ as they are being exposed continuously to the mostpowerful drinking cue of all, actual alcohol consumption.^32- 33 Therefore,when the medication is stopped, relapse to a pattern of heavy drinking willbe quite likely. In gainsay, these arguments are based on the controversialpremise of "controlled" drinking studies; the corresponding analogy here wouldbe that as the delivery of pharmacotherapy is relatively short, such as afew weeks or months, individuals who do not become abstinent by study endmight be at an increased risk of relapse. The appropriateness of such an analogyto pharmacotherapy studies can be questioned in 2 ways. First, there is insufficientknowledge about the long-term outcomes of individuals who complete such studieswith varying levels of drinking outcome, such as from abstinence to socialdrinking, by study end. Second, the more apt scenario might be to consideran individual who is drinking at social levels at the end of a pharmacotherapystudy as being in "partial remission" but not "cured." Thus, the pertinentquestion becomes whether this "partial remission" could be maintained or abstinenceultimately achieved if efficacious medication is prescribed long-term or indefinitely.That is, instead of short-term pharmacotherapy, if it is accepted that alcoholismis a chronic disorder requiring continuous and possibly lifetime pharmacologicaltreatment, in the same manner as the provision of insulin to a diabetic isessential, could pharmacotherapy not simply be provided as long-term treatmentto maintain clinical improvement? Also, will the psychosocial gains in generallifestyle and well-being obtained by pharmacotherapy-driven, long-term maintenanceof social drinking levels operate to provide time for the alcoholic to heal"naturally" and ultimately become abstinent as he or she becomes more ableto be productive in society? Finally, what individual or group characteristicsdetermine those subjects who might do just as well with medication-maintainedsocial drinking compared with those who remain abstinent?^30,34 Researchis needed in this understudied area of alcoholic pharmacotherapy to provideanswers to these questions. Not only would this provide pharmacological knowledgeon the long-term management of alcoholism, but it might help to bridge thegap between research and practice, where patients are often seen for prolongedperiods and often over a lifetime.

One caveat of our study was that it was not designed to provide forpost-treatment follow-up, so it is not known how long these psychosocial improvementswere sustained. What can be said, however, is that even when a "minimum" complianceenhancement intervention is coupled with effective pharmacotherapy, thereare important gains in psychosocial functioning during treatment. Our combinedregimen of topiramate and BBCET not only treats the "symptom" of drinkingbut is effective at improving psychosocial functioning during treatment thatencompasses the alcohol dependence syndrome. We note that the treatment effectsdescribed in this trial may only be possible with moderately dependent alcoholics,and may not generalize to those with the most severe and entrenched form ofthe disease.

In summary, our results continue to provide evidence that topiramateis a safe and effective treatment for alcohol dependence syndrome, and thatpharmacotherapies that reduce heavy drinking can significantly improve qualityof life and reduce the consequences of alcohol consumption.

Correspondence: Bankole A. Johnson, MD, PhD, Department of Psychiatry,The University of Texas Health Science Center at San Antonio, 7703 Floyd CurlDr, Mail Stop 7792, San Antonio, TX 78229-3900 (bjohnson@uthscsa.edu).

Submitted for publication September 3, 2003; final revision receivedMarch 12, 2004; accepted March 15, 2004.

We are grateful to Ortho-McNeil Pharmaceutical, Inc, for providing medicationand a research grant in partial support of this investigator-initiated project.Additional support was provided by funding from the Division of Alcohol andDrug Addiction, Department of Psychiatry at The University of Texas HealthScience Center at San Antonio. We thank the National Institute on AlcoholAbuse and Alcoholism for its support of Prof Bankole A. Johnson (grants AA10522-08 and 12964-01) and Asst Prof Nassima Ait-Daoud (grant K23 AA 00329-01).

We deeply appreciate the skilled technical assistance of the staff atthe South Texas Addiction Research and Technology Center, Department of Psychiatry,The University of Texas Health Science Center at San Antonio. We also aregrateful to Deanne Hargita, MPA, for her excellent support of database functions.We are indebted to Eva Jenkins-Mendoza, BS, for her outstanding service asproject coordinator. We also thank Robert H. Cormier, Jr, BA, for his assistancewith manuscript preparation.