Post-treatment Effects of Topiramate on Alcohol-Related Outcomes: A Combined Analysis of Two Placebo-Controlled Trials

Abstract

BACKGROUND:

Topiramate reduces drinking and alcohol-related problems and is increasingly being used to treat alcohol use disorder (AUD). In a randomized controlled trial (RCT) of topiramate, rs2832407, a single nucleotide polymorphism (SNP) in the GRIK1 gene moderated topiramate’s effects (Study 1). However, a second RCT (Study 2) did not replicate the SNP’s moderating effect during treatment. The current analysis combines data from these two studies to examine topiramate’s effects on alcohol-related outcomes and on its pharmacogenetic moderation during a 6-month post-treatment period.

MATERIAL AND METHODS:

This analysis includes 308 individuals with problematic alcohol use (67% male; mean age=51.1; topiramate: 49%, placebo: 51%). It uses generalized linear mixed models to examine changes in self-reported alcohol consumption and alcohol-related problems and concentrations of γ-glutamyltransferase. The report combines published 3- and 6-month follow-up data from Study 1 with similar, unpublished data from Study 2.

RESULTS:

The overall multivariate medication effects on outcomes during 3- and 6-month follow-up were not significant (p=0.08 and p=0.26, respectively). The moderating effect of the SNP on primary treatment outcomes was also not significant during either follow-up period (p=0.13 and p=0.16, respectively). However, during the 3-month post-treatment period, drinks per day was significantly lower in the topiramate group than the placebo group in the rs2832407*CC-genotype group.

CONCLUSIONS:

The robust effects of topiramate on alcohol-related outcomes during treatment diminish substantially once the medication is discontinued. Research is needed both to determine the optimal treatment duration and to identify clinically useful pharmacogenetic moderators of topiramate for treating AUD.

www.clinicaltrials.gov registration: NCT00626925 and NCT02371889

INTRODUCTION

Topiramate has multiple pharmacological actions, including blocking voltage-dependent sodium channels, increasing agonist activity at some GABA-A receptor subtypes, antagonizing AMPA and kainate glutamate receptors, and inhibiting the enzyme carbonic anhydrase.¹ It is approved by the Food and Drug Administration as an anticonvulsant, to prevent migraine, and (in combination with phentermine) to promote weight loss. In individuals being treated for heavy drinking is also efficacious in reducing drinking days and heavy drinking days^2,3 and is increasingly being used off-label to treat alcohol use disorder (AUD).⁴ This approach is supported by treatment guidelines that recommend prescribing topiramate to individuals with moderate or severe AUD, either as a first-line⁵ or a second-line⁶ treatment. Despite its growing use in treating AUD, most studies of topiramate to treat the disorder have followed participants only during the active treatment period, so there is little information on whether its beneficial effects persist after the medication is discontinued.

We conducted two 12-week, placebo-controlled, randomized clinical trials (RCTs) of topiramate^3,7 that examined: 1) its efficacy and tolerability in treating heavy drinking and 2) the moderating effect on topiramate’s efficacy of a single nucleotide polymorphism (rs2832407) in the gene encoding a kainate receptor subunit. Both studies included 3- and 6-month post-treatment follow-up visits.

In Study 1 (N=138 heavy drinkers),⁷ topiramate-treated patients reported fewer drinking days (DDs) and heavy drinking days (HDDs) than the placebo group. Among the European-ancestry (EA) subsample (n=122), rs2832407*C-allele homozygotes (42% of the sample) treated with topiramate reduced the number of HDDs more than those treated with placebo, while neither the heterozygotes (53% of the sample) nor the rs2832407*A-allele homozygotes (15% of the sample) showed a drug-placebo difference. The effect of topiramate on HDDs seen during treatment in rs2832407*C-allele homozygotes persisted through a 6-month follow-up period.⁸

In Study 2,³ 170 EA patients with AUD were stratified using a 2-level rs2832407 genotype variable (comprising 36% C-allele homozygotes and 64% A-allele carriers), chosen a priori due to the small number of rs2832407*A-allele homozygotes in Study 1. Patients were randomized within genotype to receive topiramate or placebo. Although topiramate reduced HDDs significantly more than placebo, rs2832407 did not moderate the therapeutic effect. A combined analysis of the two topiramate studies,⁹ which provided greater statistical power to test the moderator hypothesis, confirmed the lack of an effect of rs2832407 on topiramate’s efficacy in reducing alcohol consumption during the treatment period.

The goal of the current study was to extend the findings of Kranzler et al.⁹ by examining the combined RCT data during 3- and 6-month follow-up periods. Specifically, the previously unpublished 3- and 6-month follow-up data from Study 2 are reported here in combination with the published post-treatment findings from Study 1.⁸ This report also includes a meaure of drinks per day, which augments the report on post-treatment outcomes in Study 1.⁸ We hypothesized that combining the two studies would provide greater statistical power than either individual study and would better demonstrate both the main effects of topiramate and the moderating effects of rs2832407 during the 3- and 6-month follow-up periods.

MATERIALS AND METHODS

Patients and Procedures

CONSORT diagrams and a full description of the procedures for both studies are provided in the primary publications.^3,7 In brief, 308 individuals with problematic alcohol use were recruited to participate in one of two, 12-week RCTs (Study 1, n=138; Study 2, n=170) in which they received up to 200 mg/day of topiramate or placebo. Patients were enrolled and treated at UConn Health in Farmington, CT (Study 1: n=76) or at the University of Pennsylvania Perelman School of Medicine (Study 1: n=62, Study 2: n=164, total n=226) or the Corporal Michael J. Crescenz Veterans Affairs Medical Center (Study 2: n=6) Philadelphia. Study procedures and consent documents were approved by the institutional review boards at the 3 participating sites and patients gave written informed consent to participate. The studies were registered on clinicaltrials.gov: NCT00626925 and NCT02371889.

Study criteria

Inclusion criteria were age 18-65 (Study 1) or 18-70 (Study 2), current heavy drinking (≥24 drinks per week for men, ≥18 drinks per week for women), ability to read English at an 8^th grade level, and, if a woman of childbearing potential, use of a reliable birth control method. All but one patient in Study 1 had a DSM-IV diagnosis of alcohol dependence (n=127 or 92.0%) or alcohol abuse (n=10 or 7.2%)¹⁰ and a goal of reduced drinking. All patients in Study 2 were EA, had a current DSM-5 AUD diagnosis¹¹ and a goal of either reduced drinking or abstinence, and were randomly assigned to treatment group based on rs2832407 genotype. Exclusion criteria for both studies were a current, clinically significant physical or psychiatric diagnosis, including drug dependence (other than nicotine), or a clinical condition that warranted abstinence from alcohol (e.g., a recent history of alcohol-related gastritis), as abstinence was not required for participation. Patients who were mandated to receive alcohol treatment were not enrolled in either study.

Procedures

Both studies recruited patients through advertisements, clinical referrals, or medical record screening. An initial telephone screening interview was followed by an in-person visit, where patients who were potentially study-eligible gave informed consent, provided a medical and psychiatric history, and underwent a physical examination and clinical laboratory testing.

The treatment protocols in the two studies were similar. For the first 6 weeks, patients attended weekly visits and gradually increased the medication from 25 mg/day (or one placebo capsule) to a maximum of 100 mg (or two placebo capsules) twice daily. During the last 6 weeks of treatment, the frequency of visits was reduced to bi-weekly. At each treatment visit, a nurse delivered Medical Management,¹² a structured form of counseling in which patients are urged at each visit to adhere to the medication regimen by emphasizing that the medication can only have beneficial effects if consumed regularly. Patients were also counseled to reduce their drinking and either increase abstinent days or, if they chose a goal of abstinence, to stop drinking or remain abstinent. At the final treatment visit, patients were scheduled for a 3-month follow-up visit at which time a 6-month follow-up visit was scheduled. The studies differed on the following features: a) Study 1 recruited only patients who sought to reduce their drinking, while Study 2 recruited patients whose treatment goal was either reduced drinking or abstinence; b) Study 1 recruited patients of all population groups, while Study 2 recruited only EAs; c) Study 1 did not genotype patients prior to randomization, while Study 2 randomized patients within rs2832407-genotype group.

Measures

We used the Timeline Follow-Back (TLFB) method¹³ at baseline and at each treatment and post-treatment visit to measure the quantity and frequency of alcohol consumption. We attempted to contact and evaluate all patients at all scheduled time points, regardless of whether they completed treatment. Patients for whom more than 90 days had elapsed from their prior visit (including those who missed an assessment) were asked to report on their drinking since the last assessment. We used the Short Index of Problems (SIP),¹⁴ a 15-item questionnaire, at all timepoints to obtain alcohol-related problem scores (from 0-45) over the preceding 3 months. The biomarker, γ-glutamyltransferase (GGT), was measured at baseline, at the end of the treatment period, and at each follow-up visit. Patients were paid $50 to complete the assessments at the end of treatment and each follow-up visit.

Statistical Analysis

A multivariate model using the generalized linear mixed procedure in SAS v9.4 was used to test for treatment group differences on five outcome variables simultaneously for each of four time periods. All available data were estimated using restricted maximum likelihood estimation, under the assumation that missing data were missing at random. The five outcomes assessed were: 1) drinks per day (DPD; calculated as the total number of drinks divided by the number of days in the time period), 2) percentage of heavy drinking days (PHDD; calculated as the percent of days in the time period on which women consumed ≥4 drinks and men consumed ≥5 drinks), 3) percentage of days abstinent (PDA), 4) natural log-transformed γ-glutamyltransferase concentration (lnGGTP), and 5) the Short Index of Problems (SIP) score. The four time periods were: 1) the 90 days pretreatment (reflected in the baseline measures), 2) the 12 weeks of treatment, 3) the first 3 months of post-treatment follow-up, and 4) the second 3 months of post-treatment follow-up. An overall multivariate test for group differences on the five outcomes and individual tests for each of the outcomes were performed separately at each time point.

Additional models tested the moderating effect of rs2832407 (C-allele homozygotes vs. heterozygotes + A-allele homozygotes) only among EA patients (n=292). In addition to treatment group, genotype group, and their interaction effect, an additional fixed effect included in all models was study variable (Study 1 or Study 2). Pretreatment drinking measures were also included as covariates in the respective models for the treatment period and the 3- and 6-month post-treatment follow-up periods. We included data from the pretreatment and treatment periods, which were reported previously,^3,7,9 to model changes over time. An unstructured covariance matrix for random effects was used to account for the correlation among the outcome variables. We report both unadjusted p-values and Hsu-Nelson simulated p-values adjusted for multiplicity.¹⁵ The alpha level for statistical significance was set at 0.05. To calculate effect size, the percentage of variance explained for an effect (i.e., either the difference between treatment groups or an interaction with genotype) was obtained for each outcome and converted to a standardized mean difference.¹⁶ With 308 individuals in the combined sample, two groups, and four timepoints we had power of 0.80 to detect a small effect size (Cohen’s f=0.18).

RESULTS

The sample of 308 patients included 138 patients from Study 1 and 170 patients from Study 2. Detailed descriptions of the study samples can be found in Kranzler et al. 2014 (Study 1)⁷ and Kranzler et al. 2021 (Study 2).³ The majority of the sample (n=207, 67%) was male, with an average age of 51.1 years (SD=10.3) and nearly all (n=292, 95%) were of European ancestry. Approximately half of the patients (n=152, 49%) were randomly assigned to receive topiramate and half placebo (n=156, 51%). Observed allele frequencies for rs2832407 were 63.3% for the major allele (C) and 37.7% for the minor (A) allele, and accordingly, we compared C-allele homozygotes (n=112, 38%) with A-allele carriers (n=180, 62%).

As reported previously, there was a high rate of medication adherence in the two RCTs. Specifically, in Study 1, placebo patients ingested a mean of 91.1% of daily doses, while topiramate patients consumed a mean of 89.4% of daily doses.⁷ In Study 2, the adherence rates were 94.3% and 90.0%, respectively.³ There was also a high rate of post-treatment follow-up across the two studies, with 274 (89.0%) patients providing data at 3 months and 265 (86.0%) at 6 months. This included 123 (89.1%) patients at 3 months and 118 (85.5%) at 6 months in Study 1⁸ and 151 (88.8%) patients at 3 months and 147 (86.5%) at 6 months Study 2.

Table 1 shows the main effects of treatment group in the multivariate models for each of the different time periods. At baseline, the treatment groups were similar on all measures, consistent with random assignment. During treatment, there was a significant overall multivariate between-group difference (F_5,266 = 6.96, p<0.001). Analysis of the individual alcohol-related outcomes showed that after adjustment for multiple comparisons the topiramate-treated group did significantly better than the placebo group on all measures, except PDA. As can be seen in the table, the size of the effect of topiramate treatment ranged from 0.292 (reflecting a small effect, i.e., increased PDA) to −0.522 (reflecting a medium effect, i.e., decreased PHDD).

Table 1:

Estimated Treatment Effect with Unadjusted and Adjusted P-Values (N=308)

Time Point Outcome Measure	Effect*	Standard Error	Effect Size^	Unadjusted P-Value	Adjusted P-Value
Baseline
Overall					0.63
Drinks per Day	0.409	0.311	0.144	0.19	0.61
% Days Abstinent	1.188	2.140	0.057	0.58	0.98
% Heavy Drinking Days	0.948	2.893	0.031	0.74	1.00
lnGGTP	0.009	0.103	0.013	0.93	1.00
SIP	−0.267	0.971	−0.031	0.78	1.00
Treatment Period
Overall					<0.001
Drinks per Day	−0.810	0.210	−0.434	<0.001	<0.001
% Days Abstinent	6.625	2.722	0.292	0.016	0.066
% Heavy Drinking Days	−13.567	3.028	−0.522	<0.001	<0.001
lnGGTP	−0144	0.050	−0.395	0.005	<0.020
SIP	−2.992	0.749	−0.463	<0.001	<0.001
3-Month Follow-Up Period
Overall					0.083
Drinks per Day	−0.227	0.244	−0.125	0.35	0.82
% Days Abstinent	0.128	3.489	0.008	0.97	1.00
% Heavy Drinking Days	−2.457	3.845	−0.091	0.52	0.95
lnGGTP	0.056	0.063	0.133	0.38	0.85
SIP	−1.753	0.721	−0.300	0.016	0.063
6-Month Follow-Up Period
Overall					0.26
Drinks per Day	−0.07	0.256	−0.106	0.42	0.89
% Days Abstinent	0.679	3.719	0.020	0.86	1.00
% Heavy Drinking Days	−4.539	3.770	−0.154	0.23	0.64
lnGGTP	−0.006	0.065	0.028	0.92	1.00
SIP	−1.743	0.802	−0.299	0.031	0.12

^*Mean difference between Topiramate and Placebo groups (TOP – PLA)

^{^}Standardized mean difference

lnGGTP=natural log of γ-glutamyltransferase concentration; SIP=Short Index of Problems score

At the 3-month follow-up, the overall multivariate test was not significant (F_5,253 = 1.97, p=0.083), though on 4 of the 5 outcomes the direction of effect favored topiramate. During this period, the only individual outcome measure that was nominally better in the topiramate group was SIP score (p=0.016) (see Figure 1 and Supplementary Figures 1–4), an effect that was not significant after adjustment for multiple comparisons (p=0.063). Similarly, at the 6-month follow-up, although the groups did not differ significantly overall (multivariate F_5,240 = 1.31, p=0.26), on all 5 outcome measures the direction of effects favored topiramate. Although during this time the SIP score was nominally lower in the topiramate group (0.031), the effect did not survive correction for multiple comparisons (p=0.12). Effect sizes during the 3- and 6-month follow-up periods were all ≤0.300, which was seen for reductions in the SIP score.

Figure 1:

Mean (±SEM) SIP Score by Treatment Group and Time Point

Table 2 shows the results of the multivariate models testing the moderating effect of rs2832407 on treatment outcomes. At baseline, there were no significant interaction effects of treatment with genotype group. During treatment, although the overall multivariate interaction effect was not significant (F_5,257 = 1.83, p=0.11), the direction of effects on all 5 outcome measures favored topiramate only in the CC genotype group. Further, as shown in Figure 2, during treatment there was a significant univariate effect on DPD (unadjusted p=0.008; adjusted p=0.036), with topiramate-treated patients reporting fewer DPD than those receiving placebo, but only in the CC-genotype group. Supplementary Figures 5–8 show effects across the study periods for the other outcome measures. As can be seen in Figure S5, during treatment, PDA was nominally higher among topiramate-treated patients than controls only in the CC-genotype group (unadjusted p=0.034), but the effect did not survive adjustment for multiple comparisons.

Table 2:

Estimated Treatment by Genotype Interaction Effect with Unadjusted and Adjusted P-Values (n=292 individuals of European ancestry)

Time Point Outcome	Effect*	Standard Error	Effect Size^	Unadjusted P-Value	Adjusted P-Value
Baseline
Overall					0.95
Drinks per Day	−0.283	0.676	−0.047	0.68	0.99
% Days Abstinent	2.553	4.863	0.054	0.60	0.98
% Heavy Drinking Days	−4.564	6.740	−0.086	0.50	0.95
lnGGTP	−0.051	0.180	−0.004	0.78	1.00
SIP	−1.588	2.066	−0.084	0.44	0.92
Treatment Period
Overall					0.11
Drinks per Day	−1.187	0.445	−0.339	0.008	0.036
% Days Abstinent	11.455	5.362	0.250	0.034	0.13
% Heavy Drinking Days	−10.682	6.615	−0.227	0.11	0.36
lnGGTP	−0.145	0.110	−0.145	0.19	0.54
SIP	0.374	1.545	0.033	0.81	1.00
3-Month Follow-Up Period
Overall					0.13
Drinks per Day	−1.271	0.453	−0.335	0.005	0.024
% Days Abstinent	6.948	6.184	0.130	0.26	0.709
% Heavy Drinking Days	−14.752	6.998	−0.273	0.036	0.154
lnGGTP	−0.241	0.210	−0.164	0.25	0.695
SIP	−0.688	1.449	−0.047	0.64	0.990
6-Month Follow-Up Period
Overall					0.16
Drinks per Day	−1.275	0.563	−0.314	0.025	0.099
% Days Abstinent	8.600	7.938	0.147	0.28	0.72
% Heavy Drinking Days	−16.546	8.217	−0.293	0.045	0.16
lnGGTP	−0.078	0.131	0.026	0.55	0.97
SIP	0.813	1.666	0.042	0.63	0.99

^*Difference in Difference [(CC Top – CC Pla) – (AA/AC Top – AA/AC Pla)]

^{^}Standardized mean difference

lnGGTP=natural log of γ-glutamyltransferase concentration; SIP=Short Index of Problems score

Figure 2:

Mean (±SEM) Drinks per Day by Genotype Group, Treatment Group, and Time Period

During the 3-month follow-up period, although the overall multivariate interaction effect was not statistically significant (F_5,209 = 1.72, p=0.13), the direction of effects for 4 of the 5 outcome measures favored topiramate only in the CC-genotype group (see also Supplementary Figures 5–8). During this period, there was a significantly greater reduction in DPD patients treated with topiramate than those receiving placebo only in the CC-genotype (p=0.005), an interaction effect that survived multiple testing correction (adjusted p=0.024) (see Figure 2). The topiramate group also showed a greater reduction in PHDD than placebo only in the CC-genotype group (p=0.036), though this effect did not survive multiple testing correction.

The overall multivariate interaction effect at the 6-month follow-up was not statistically significant (F_5,233 = 1.62, p=0.16), though the direction of effect for 4 of 5 outcomes favored topiramate over placebo group only in the CC genotype group. During this period, although there were effects of topiramate treatment among C-allele homozygotes but not the AC/AA genotype group on DPD (unadjusted p=0.025) and PHDD (unadjusted p=0.045), neither remained significant after correction for multiple comparisons.

As can be seen in Table 2, the largest effect sizes in the moderator analyses were for DPD, which ranged from −0.339 during treatment to −0.314 during the 6-month follow-up period, a small-to-medium effect reflecting a greater reduction in the topiramate group than placebo, but only in patients with the CC genotype.

DISCUSSION

This study of the post-treatment effects of topiramate among individuals with problematic alcohol use showed beneficial effects of the active medication on 4 of 5 alcohol-related outcomes during the 12-week treatment period, with a medium effect size for PHDD. However, these effects did not persist during the 3-month and 6-month post-treatment follow-up period. The within-treatment effects are consistent with, but extend prior reported findings from the individual studies.^3,7 Although Study 1⁸ showed topiramate-treated patients to have persistently lower SIP scores than placebo-treated patients throughout the 6-month follow-up period, that effect was not replicated here in the combined samples.

During the 3-month post-treatment follow-up period, both treatment groups reported significant improvements over those seen during treatment on all self-reported outcomes. Although the effect seen during the 6-month follow-up period was not significant, the self-reported post-treatment outcomes remained significantly improved relative to baseline values. Thus, non-study-related factors may have affected the course of drinking and alcohol-related variables during treatment, with their effects becoming more evident once treatment abated. Further research on the factors that contribute to the effects observed during post-treatment follow-up is needed to enhance the persistence of beneficial treatment effects of topiramate, an important consideration for any treatment, pharmacological or otherwise.

A combined analysis of the within-treatment effects from Study 1 and Study 2⁹ showed no multivariate moderating effect of rs2832407 on topiramate’s efficacy in improving alcohol-related outcomes. However, the present combined analysis, which used a different analytic approach than the prior report⁹ to model the post-treatment findings, yielded a univariate effect on DPD that was significant after correction for multiple comparisons. Specifically, topiramate-treated patients in the CC-genotype group reported fewer DPD than those receiving placebo. Although the combined analyses did not show robust post-treatment effects of topiramate treatment in rs2832407*C-homozygotes, the overall direction of effects was consistent with a moderating effect of rs2832407 on topiramate treatment response. Nonetheless, the SNP’s moderating effects do not appear to warrant its use in clinical care of patients with AUD.

Although treatment response is partially genetically determined,¹⁷ it is a complex trait influenced by multiple genetic variants of small effect.^18,19 Thus, it is unlikely that a single, non-coding SNP could serve clinically as a moderator of treatment response among individuals with AUD. The findings reported here, together with negative findings on the moderating effect of naltrexone treatment by a polymorphism in the mu-opioid receptor gene,²⁰ suggest that the use of candidate SNPs is not clinically useful in predicting the response to AUD pharmacotherapy. Thus, efforts to identify pharmacogenetic moderators for AUD should focus on genome-wide analyses,¹⁹ from which polygenic risk measures can be derived, as these could prove to be clinically useful in selecting medications to treat AUD.

These findings are similar to those reported previously for oral naltrexone,^21–23 whose beneficial effects during treatment diminished once the treatment was discontinued. There was also a post-treatment reduction in the effects of extended-release naltrexone (XR-NTX) on drinking outcomes in a study of individuals experiencing homelessness and AUD.²⁴ An exception to this pattern of findings was seen in a 24-week RCT in which individuals with AUD and post-traumatic stress disorder were randomly assigned to receive naltrexone or placebo combined with either prolonged exposure therapy or supportive counseling.²⁵ Participants in all 4 treatment groups in that study had large decreases in the percentage of days drinking (PDD), with naltrexone treatment associated with a significantly lower PDD than placebo. At 6 months post-treatment, although all 4 groups showed increases in PDD, the prolonged exposure therapy plus naltrexone group had the smallest increase in PDD. Similarly, in a study of acamprosate for treating alcohol dependence,²⁶ the reduced risk of return to drinking seen during the 48-week treatment period was sustained for an additional 48 weeks after the study medication was discontinued. Thus, the medication, concomitant treatments, and the patient population being treated appear to be important factors affecting the persistence of pharmacotherapeutic effects in AUD. Nonetheless, these findings suggest that AUD is similar to major depression in being a chronic, relapsing disorder for which at least 6 months of treatment is recommended to limit the risk of relapse.²⁷

Post-treatment follow-up assessments are important for evaluating the longer-term clinical implications of treatment, but generally are not included in trials because they add to the study’s costs and lengthen its duration. These factors may be particularly burdensome to drug development efforts given the call for longer treatment trials by the U.S. Food and Drug Administration (FDA)²⁸ and the European Medicines Agency (EMA).²⁹ Specifically, the FDA now requires that pivotal studies supporting the approval of “a medication for treating alcoholism” be at least 6 months in duration and the EMA requires that “confirmatory trials for treating alcohol dependence” be at least 12 months in duration. Although longer treatment trials ensure a longer exposure to the medication, thereby potentially enhancing the persistence of treatment effects, the question of the durability of treatment effects following medication discontinuation has important clinical implications.

The current findings should be considered in the context of the study’s limitations. Although, overall, the moderating effect of rs2832407 does not appear to be clinically significant, the sample – despite being combined from two RCTs – is comparatively small. In addition, sample attrition diminished the power available to detect effects during follow-up, particularly for GGTP, which unlike self-reported drinking measures, cannot be obtained via telephone. Further, the post-treatment follow-up battery did not include a detailed assessment of other factors (e.g., continued engagement in other forms of treatment in both treatment groups) that could have reduced the effects of medication treatment over time. Thus, a study in a substantially larger sample could yield findings that support the utility of rs2832407 as a biomarker for use in selecting AUD patients most likely to respond to topiramate treatment.

Study strengths include a high rate of medication adherence and treatment completion,⁷ with a high-rate of follow-up during the post-treatment period, which support the internal validity of the findings. Combined analysis of the findings from the two studies augmented the total sample size, which provided statistical power adequate to detect small effects.

In summary, it appears that although 12 weeks of treatment with topiramate has robust beneficial effects on multiple alcohol-related measures, medication effects wane during the 6 months after treatment is discontinued. Studies of topiramate longer than 12 weeks are needed to define the optimal duration of treatment with the medication and future research should examine factors that contribute to the maintenance of treatment gains following the cessation of pharmacological treatment.