Moderation of treatment outcomes by polygenic risk for alcohol‐related traits in placebo‐controlled trials of topiramate

Abstract

Background

In two 12‐week, randomized, placebo‐controlled trials (RCTs) in individuals with alcohol use disorder (AUD), topiramate significantly reduced heavy drinking days (HDDs), and alcohol‐related problems. In a secondary analysis of those findings, we examined four broad measures of genetic risk—polygenic scores (PGS)—of problematic alcohol use (PAU), drinks per week (DPW), and time to relapse to any drinking (TR) and heavy drinking (THR) as moderators of topiramate's effect on HDDs and alcohol‐related problems.

Methods

We analyzed data from 285 individuals with AUD (65.6% male) of European‐like ancestry, who were treated with either topiramate (49.1%) or placebo (50.9%). All patients underwent genome‐wide array genotyping, and PGS were calculated using summary statistics from genome‐wide association studies of PAU, DPW, and TR and THR (two time‐to‐event outcomes among patients treated in AUD pharmacotherapy trials). We hypothesized an interaction effect in which greater genetic risk—particularly for PAU—would be associated with a greater therapeutic response to topiramate than placebo.

Results

As shown previously, topiramate significantly reduced both HDDs (odds ratio [OR] = 0.50, p < 0.001) and Short Index of Problems (SIP) scores (b = −3.04, p < 0.001) more than placebo. There were nonsignificant associations of higher PGS with more HDDs (OR = 1.17, 95% CI = 0.98–1.41, p = 0.091) and a greater reduction in HDDs in the topiramate group (OR = 0.80, 95% CI = 0.62–1.03, p = 0.089). There were also significant interaction effects with treatment on SIP score by PGS for PAU (b = −1.64, SE = 0.78, p = 0.033), TR (b = −2.16, SE = 0.72, p = 0.003), and TRH (b = −2.17, SE = 0.72, p = 0.003).

Conclusions

These findings provide proof of principle for the use of alcohol‐related PGS as moderators of the effects of topiramate for treating AUD. Larger RCTs of topiramate are needed to provide adequate statistical power to validate this pharmacogenetic approach to precision AUD treatment.

In a secondary analysis of data from two randomized controlled trials of topiramate in 285 individuals with alcohol use disorder (AUD), measures of genetic risk for problematic alcohol use (PAU) and time to relapse to any drinking (TR) and heavy drinking (THR) significantly moderated topiramate's effect on alcohol‐related problems. These findings support the use of alcohol‐related polygenic scores as moderators of the effects of topiramate for treating AUD and support the need for larger studies that use this approach.

INTRODUCTION

Although alcohol use disorder (AUD) is prevalent worldwide and associated with adverse health, economic, and social effects, only a small minority of individuals receive treatment for it (Cohen et al., 2007). Pharmacological treatments, a cornerstone of intervention for most psychiatric disorders, are particularly underutilized in AUD, with only 2.1% of patients diagnosed with a past‐year AUD having received such treatment in 2022 (SAMHSA, 2022).

Topiramate, first approved by the FDA in 1996 as an anticonvulsant, is also approved to prevent migraine and, in combination with phentermine, for weight loss. Although it is not approved to treat AUD, topiramate has demonstrated efficacy in promoting abstinence and reducing heavy drinking, where it yields small‐to‐medium effects (Blodgett et al., 2014; Fluyau et al., 2023). To enhance the drug's efficacy and reduce its adverse effect burden, recent efforts have focused on the use of pharmacogenetics to personalize topiramate treatment for AUD (Hartwell & Kranzler, 2019).

Evidence that a single‐nucleotide polymorphism (SNP) in GRIK1 (rs2832407), the gene that encodes the GluK1 subunit of the kainate receptor (Gryder & Rogawski, 2003; Kaminski et al., 2004), was associated with AUD (Kranzler et al., 2009) led to the hypothesis that the SNP moderates the response to topiramate treatment in individuals with the disorder. In the absence of a topiramate‐treated AUD patient sample suitable for identifying a genetic predictor of treatment response, we hypothesized that a variant associated with the disorder itself could serve as a proxy for biological risk and a biomarker for treatment response. To test the hypothesis, we conducted two 12‐week, randomized, placebo‐controlled clinical trials (RCTs) of topiramate (Kranzler et al., 2014; Kranzler, Morris, et al., 2021).

In the first study (N = 138) (Kranzler et al., 2014), topiramate‐treated patients reported significantly fewer drinking days and heavy drinking days (HDDs) than the placebo group. In the 122 self‐identified European‐American (EA) individuals, post hoc genotyping showed an interaction between genotype and medication group, such that rs2832407*C‐allele homozygotes treated with topiramate had the greatest reduction in HDDs, while rs2832407*A‐allele carriers showed no drug‐placebo difference. In a subsequent 12‐week RCT (Kranzler, Morris, et al., 2021), 170 EAs with AUD were randomly assigned to treatment with topiramate or placebo prospectively based on rs2832407 genotype (i.e., C‐allele homozygotes vs. A‐allele carriers). In this study, although topiramate reduced HDDs significantly more than placebo, the SNP did not moderate the therapeutic effect. Combining samples from the two studies to maximize the power for pharmacogenetic analysis (Kranzler, Hartwell, et al., 2021) showed a significant effect of topiramate on alcohol‐related outcomes but no moderating effect of the SNP.

We report here the results of a secondary analysis of the two topiramate trials in which we tested whether more comprehensive measures of genetic risk—that is, polygenic scores (PGS)—for four alcohol‐related traits moderated the effect of topiramate in reducing HDDs. To do so, we created PGS for problematic alcohol use (PAU), alcohol consumption (drinks per week [DPW]), and two treatment outcome measures: time to relapse to any drinking (TR) and time to relapse to heavy drinking (THR). See Figure S1 for an overview of the study and Table S1 for information on the genome‐wide association studies (GWAS) that served as discovery samples .

We hypothesized that the four PGS would be associated with greater reductions in HDDs in patients treated with topiramate. That is, genetic risk for PAU, DPW, and risk of relapse (TR and THR) would be associated with a better response to topiramate treatment. We conducted similar analyses with an alcohol‐related problem score as the outcome measure. We chose the PAU PGS as the primary moderator variable because it reflects genetic risk for a trait that is the focus of many treatment efforts and, with a discovery sample of >1 million individuals, provides good statistical power for the analysis. PAU is a compound trait that comprises (1) International Classification of Diseases, 9th revision (ICD‐9) and 10th revision (ICD‐10) diagnoses of AUD, (2) DSM‐IV alcohol dependence, and (3) alcohol‐related problems identified using questions 4–10 of the Alcohol Use Disorders Identification Test (i.e., the AUDIT Problem scale) (Zhou et al., 2023). The time‐to‐event measures were derived from GWAS of outcomes following treatment with naltrexone, acamprosate, or placebo. The mechanisms of action of these two drugs differ substantially from those of topiramate (Kranzler & Hartwell, 2023) and were derived from a sample of just over 1000 individuals.

MATERIALS AND METHODS

Target sample

The recruitment procedures, study criteria, treatment protocol, and outcome measures of the two topiramate studies in this analysis are described in the original publications (Kranzler et al., 2014; Kranzler, Morris, et al., 2021). Both studies used similar recruitment and treatment protocols. Briefly, patients were recruited through advertisements or clinical referrals, completed a telephone screening interview to determine eligibility, and, if eligible, were invited to attend an in‐person visit, where they gave informed consent and underwent medical and laboratory testing. Study drug was initiated at a dosage of 25 mg/day at bedtime and increased over 6 weeks to a maximum of 200 mg/day in two divided doses. The number of placebo capsules was similarly increased gradually. Patients also received medical management, a brief, supportive intervention provided by a nurse or physician with a goal of promoting medication adherence and reducing drinking (Pettinati et al., 2004). The sample from the first study (Kranzler et al., 2014) comprised 138 patients (62.3% male) of whom 67 (48.6%) were randomized to receive topiramate and 71 (51.4%) received placebo. In this study, patients were randomized to treatment condition based on age, sex, and pretreatment alcohol consumption. The sample from the second study (Kranzler, Morris, et al., 2021) comprised 170 patients (71.2% male) of whom 85 (50%) were randomized to each of the two treatment groups with stratification by treatment goal (abstinence vs. reduced drinking) and genotype group (rs2832407*CC vs. AA/AC).

At baseline, medication groups in each study were comparable on age, sex, and the frequency of drinking days and HDDs during the 90‐day pretreatment period. Study protocols were approved by the institutional review boards at all participating sites and patients gave written informed consent. The studies were registered on clinicaltrials.gov: NCT00626925 and NCT02371889.

We limited the analysis presented here to European‐like‐ancestry (EUR) individuals because the discovery GWAS were of individuals of EUR and the calculation of PGS for other ancestries was not expected to be as predictive of the target phenotype in non‐EUR groups. We also excluded individuals whose DNA sample did not yield genotype data adequate for analysis (n = 6), or who were missing a key variable in the analysis (n = 1) (Figure S2). We also excluded patients who, when enrolled in the clinical trials, did not self‐identify as EA (n = 16). This yielded a total of 285 individuals for the pharmacogenetic analysis, among whom there were no significant between‐treatment‐group age or sex differences (Table 1).

TABLE 1.

Demographics of the study samples individually and combined, by medication group.

	Placebo (N = 140)	Topiramate (N = 145)	Total sample (N = 285)
Study sample, n (%)
Kranzler et al. (2014)	59 (48.4)	63 (51.6)	122 (42.8)
Kranzler, Hartwell, et al. (2021)	81 (49.7)	82 (50.3)	163 (57.2)
Sex, n (% male)	93 (69.3)	94 (64.8)	187 (65.6)
Age (mean ± SD)	51.4 ± 9.4	51.4 ± 10.9	51.4 ± 10.1

Note: All comparisons within and between study groups are nonsignificant (p > 0.05).

Genotyping and imputation

Genotype data were filtered for individual call rates and excessive heterozygosity using PLINK v1.9 (Purcell et al., 2007), which resulted in 6 individuals being excluded. Variants were excluded if they had a call rate <98%, minor allele frequency <0.002, or deviated significantly from Hardy–Weinberg equilibrium (p‐value < 1 × 10⁻⁶). Imputation was done through the Michigan Imputation Server (Das et al., 2016) using the Haplotype Reference Consortium reference panel (McCarthy et al., 2016). To identify groups based on genetic similarity, we conducted a principal component analysis using PLINK v1.9 and compared the results with the 1000 Genome Project Phase 3 reference data (Auton et al., 2015). Specifically, we compared the genotype data of patients to those of 1337 individuals assigned to the European ancestry superpopulation, which consists of Utah residents with Northern and Western European ancestry, Toscani in Italy, Finnish in Finland, British in England and Scotland, and Iberian individuals in Spain, and assigned individuals to European ancestry based on their genetic similarities to the European population in the reference panel. Excluding individuals who did not self‐report European ancestry yielded a final target sample of 285 individuals.

Polygenic score calculation

We used PRS‐continuous shrinkage (PRS‐CS) software to generate PGS using summary statistics from GWAS of four alcohol‐related phenotypes—PAU (Zhou et al., 2023), DPW (Saunders et al., 2022) and two time‐to‐event treatment outcomes—TR and THR (Biernacka et al., 2021). We used PGS, rather than individual SNPs as moderators, because the latter account for very small proportions of phenotypic variance. PGS aggregates the effects of many SNPs to generate a meaningful estimate of an individual's genetic risk for a trait. As reviewed in Kember et al. (2024), PGS for substance use traits have shown utility in predicting the development or clinical progression of substance use disorders in clinical populations, enabling the identification of high‐risk individuals who might benefit from targeted interventions.

In calculating PGS, default values were chosen for all parameters (except phi—see below) using guidelines on the GitHub software page (https://github.com/getian107/PRScs). Because of their large sample sizes, we used the summary statistics for PAU and DPW to calculate the global shrinkage parameter phi, which reflects the overall polygenicity of the phenotypes. To account for their limited sample size, phi was set to a fixed value of 0.01 for the two time‐to‐event outcome variables. Because linkage disequilibrium varies across populations, we used a EUR reference panel from the 1000 Genomes Project, phase 3 to account for population‐specific patterns.

Statistical analysis

We used generalized linear models to examine the main and moderating effects of PGS on the efficacy of topiramate in reducing HDDs (i.e., change over time) and alcohol‐related problems as measured by the Short Index of Problems (SIP) score. HDDs were modeled using a binomial distribution and a logit link function to examine the probability of HDDs throughout the 12‐week treatment period. The number of HDDs served as the number of events and the number of days in the study as the number of trials. We used a normal distribution and identity link function for the analysis of the SIP score. We also used a robust (sandwich) covariance matrix to adjust standard errors. The models included effects for the pretreatment percentage of HDDs (or SIP score), medication group, PGS, and the interaction of PGS with the medication group to test each of the four alcohol‐related PGS: PAU, DPW, and the two time‐to‐event variables—TR and THR—as moderators of the differences in the treatment groups on the two outcome measures over time. The covariates tested and, when significantly associated with outcome included as effects, were study (Kranzler et al., 2014; Kranzler, Morris, et al., 2021), sex, and age. We used odds ratios (ORs) with 95% confidence intervals to quantify the effect size for HDDs and the unstandardized coefficient b for SIP score. Analyses were conducted in SPSS v29.

RESULTS

Table 2 shows the model results for HDDs. There was a significant overall difference in reported HDDs between the two studies (odds ratio [OR] = 1.34, p = 0.044), with topiramate‐ and placebo‐treated patients combined reporting more HDDs during the second study (Kranzler, Morris, et al., 2021) than the first (Kranzler et al., 2014). Age was negatively associated (OR = 0.98, p = 0.023) and the percentage of pretreatment HDDs (OR = 1.03, p < 0.001) positively associated with the number of HDDs during treatment. Sex was not associated with HDDs. Consistent with findings from the two primary studies, topiramate significantly reduced HDDs by about one‐half compared with placebo treatment (OR = 0.50, p < 0.001).

TABLE 2.

Model results for the number of heavy drinking days during the study period.

Predictors	Coefficient	Standard error	Odds ratio	95% CI	P‐value
Covariates
Study (2021a)	0.296	0.147	1.34	1.01–1.79	0.044
Age	−0.016	0.007	0.98	0.97–1.00	0.023
Sex (male)	−0.163	0.151	0.85	0.63–1.14	0.279
Pretreatment heavy drinking	0.025	0.003	1.03	1.02–1.03	<0.001
Medication group (MedGrp)	−0.696	0.147	0.50	0.37–0.67	<0.001
Problematic alcohol use (PAU)
PAU PGS	0.158	0.093	1.17	0.98–1.41	0.091
MedGrp × PAU PGS	−0.223	0.131	0.80	0.62–1.03	0.089
Drinks per week (DPW)
DPW PGS	0.137	0.113	1.15	0.92–1.43	0.23
MedGrp × DPW PGS	−0.077	0.159	0.93	0.68–1.27	0.63
Time to relapse to any drinking (TR)
TR PGS	0.062	0.097	1.06	0.88–1.29	0.52
MedGrp × TR PGS	−0.113	0.137	0.89	0.68–1.17	0.41
Time to relapse to heavy drinking (THR)
THR PGS	0.055	0.099	1.06	0.87–1.28	0.58
MedGrp × THR PGS	−0.109	0.138	0.90	0.68–1.17	0.43

Note: All PGS models adjusted for study, age, and pretreatment heavy drinking. For the study variable, results are shown for Kranzler, Morris, et al. (2021), with those from Kranzler et al. (2014) as the reference group. Pretreatment heavy drinking is the percentage of heavy drinking days during the 90‐day pretreatment period. For the variable medication group (MedGrp), placebo is the reference group.

Abbreviation: 95% CI, 95% confidence interval.

Although the conditional main effects of PGS did not reach statistical significance (ps = 0.23–0.58), all coefficients of the conditional main effects were positive (ORs ranged from 1.06 to 1.17), with higher PGS associated with more HDDs. Of the four PGS, the PAU PGS, hypothesized to be the strongest moderator, showed the largest conditional main effect (OR = 1.17, 95% CI = 0.98–1.41, p = 0.091) and interaction effect (OR = 0.80, 95% CI = 0.62–1.03, p = 0.089) on HDDs (Figure 1A). Similarly, the coefficients of all interactions of PGS with medication group, although nonsignificant, were negative (ORs from 0.80 to 0.93), with higher genetic risk associated with a greater reduction in HDDs with topiramate treatment than with placebo (Figures 2A, 3A, and 4A). Thus, compared to individuals with lower PGS, those with higher PGS consistently reported more HDDs and had greater reductions in HDDs with topiramate treatment than with placebo, though none of the interactions were significant.

FIGURE 1.

(A) Percent heavy drinking days as a function of treatment group, polygenic score for problematic alcohol use, and their interaction. The Y‐axis shows the weekly percentage of heavy drinking days during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for problematic alcohol use. The shaded areas are the 95% confidence intervals for the interaction effects. (B) Short index of problems score as a function of treatment group, polygenic score (PGS) for problematic alcohol use (PAU), and their interaction. The Y‐axis shows the SIP score during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for problematic alcohol use. The shaded areas are the 95% confidence intervals for the interaction effects. Treatment group × PAU PGS is significant at p = 0.033.

FIGURE 2.

(A) Percent heavy drinking days as a function of treatment group, polygenic score for drinks per week, and their interaction. The Y‐axis shows the weekly percentage of heavy drinking days during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for drinks per week. The shaded areas are the 95% confidence intervals for the interaction effects. (B) Short index of problems score as a function of treatment group, polygenic score for drinks per week, and their interaction. The Y‐axis shows the SIP score during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for drinks per week. The shaded areas are the 95% confidence intervals for the interaction effects.

FIGURE 3.

(A) Percent heavy drinking days as a function of treatment group, polygenic score for time to relapse to any drinking, and their interaction. The Y‐axis shows the weekly percentage of heavy drinking days during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for time to relapse to any drinking. The shaded areas are the 95% confidence intervals for the interaction effects. (B) Short index of problems score as a function of treatment group, polygenic score (PGS) for time to relapse to any drinking (TR), and their interaction. The Y‐axis shows the SIP score during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for TR. The shaded areas are the 95% confidence intervals for the interaction effects. Treatment group × TR PGS is significant at p = 0.003.

FIGURE 4.

(A) Percent heavy drinking days as a function of treatment group, polygenic score (PGS) for time to relapse to heavy drinking (TRH), and their interaction. The Y‐axis shows the weekly percentage of heavy drinking days during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for time to relapse to heavy drinking. The shaded areas are the 95% confidence intervals for the interaction effect. (B) Short index of problems score as a function of treatment group, polygenic Score (PGS) for time to relapse to heavy drinking (TRH). The Y‐axis shows the SIP score during the 12‐week treatment period for the placebo group (pink) and the topiramate group (blue). The X‐axis shows the log of the polygenic score for time to relapse to heavy drinking. The shaded areas are the 95% confidence intervals for the interaction effects. Treatment group × TRH PGS is significant at p = 0.003.

Table 3 shows the model results for SIP score. There was no significant difference in SIP scores between studies (p = 0.82) nor was there an association with age (p = 0.87). However, there was a significant sex difference (b = 2.16, SE = 0.79, p = 0.006), with males reporting greater alcohol‐related problems than females at the end of the study. Pretreatment SIP score was predictive of end‐of‐study SIP score (b = 0.57, SE = 0.05, p < 0.001) and topiramate significantly reduced SIP scores (b = −3.04, SE = 0.74, p < 0.001).

TABLE 3.

Model Results for the Short Index of Problems (SIP) Score During the Study Period.

Predictors	Coefficient	Standard error	95% CI	P‐value
Covariates
Study (2021a)	−0.171	0.744	−1.63 to 1.29	0.819
Age	0.006	0.038	−0.07 to 0.08	0.871
Sex (Male)	2.156	0.785	0.62 to 3.69	0.006
Pretreatment SIP	0.568	0.046	0.48 to 0.66	<0.001
Medication group (MedGrp)	−3.039	0.738	−4.49 to 1.59	<0.001
Problematic alcohol use (PAU)
PAU PGS	1.042	0.587	−0.11 to 2.19	0.076
MedGrp × PAU PGS	−1.643	0.770	−3.15 to −0.14	0.033
Drinks per week (DPW)
DPW PGS	0.940	0.527	−0.09 to 1.97	0.074
MedGrp × DPW PGS	−1.151	0.757	−2.63 to 0.33	0.128
Time to relapse (TR)
TR PGS	0.349	0.501	−0.65 to 1.35	0.492
MedGrp × TR PGS	−2.164	0.723	−3.58 to −0.75	0.003
Time to heavy drinking (THR)
THR PGS	0.339	0.501	−0.64 to 1.32	0.498
MedGrp × THR PGS	−2.172	0.723	−3.59 to −0.76	0.003

Note: All PGS models adjusted sex and pretreatment SIP score. For the study variable, results are shown for Kranzler, Morris, et al. (2021), with those from Kranzler et al. (2014) as the reference group. For the variable medication group (MedGrp), placebo is the reference group.

Abbreviation: 95% CI, 95% confidence interval.

None of the conditional main effects of the four PGSs were significantly associated with the change in SIP score, but all were positive, with higher genetic risk associated with greater alcohol‐related problems. There were statistically significant interaction effects on SIP score for treatment × PAU PGS (b = −1.64, SE = 0.77, p = 0.033) (Figure 1B), treatment × TR PGS (b = −2.16, SE = 0.72, p = 0.003) (Figure 3B), and treatment × THR PGS (b = −2.17, SE = 0.72, p = 0.003) (Figure 4B), with higher genetic risk for these traits associated with a greater reduction in SIP scores in topiramate‐treated than placebo‐treated patients. The treatment by DPW PGS interaction, though not significant, also had a negative effect estimate (b = −1.15, SE = 0.76, p = 0.13).

DISCUSSION

In this secondary analysis of data from two RCTs of topiramate for treating AUD, we examined the main and moderating effects of PGS for PAU, DPW, TR, and THR on the number of HDDs and the severity of alcohol‐related problems (i.e., SIP score). As reported in a prior analysis of the two studies combined (Kranzler, Hartwell, et al., 2021), there was a highly significant effect of topiramate treatment on HDDs. We also found a nonsignificant (p = 0.089) moderating effect of the PAU PGS on HDDs, consistent with our primary hypothesis that greater genetic risk for PAU would be associated with a greater reduction in HDDs by topiramate. Although the moderating effects on HDDs of the other three PGS that we examined—DPW, TR, and THR—were also nonsignificant, their interactions with the medication group were all in the expected direction, such that greater genetic risk for these alcohol‐related measures was associated with a greater reduction in HDDs by topiramate than placebo. Analyses of alcohol‐related problems, as measured by the SIP score, showed a significant interaction with treatment for three of the PGS: PAU, TR, and THR. Although the interaction of treatment with DPW PGS on SIP score was in the same direction as the other PGS, it was not statistically significant. Although the interaction effects of the PGS with topiramate on HDDs did not reach significance, topiramate reduced alcohol‐related problems more than placebo among individuals with greater genetic risk for alcohol‐related traits. These findings provide proof of principle for the use of PGS as an alternative to candidate SNPs as moderators of the response to pharmacotherapies for AUD. Treatment response is a complex trait influenced by many genetic variants of small effect (Manolio et al., 2009; Motsinger‐Reif et al., 2013). Thus, candidate SNPs, even if replicable, are unlikely to exert clinically meaningful moderating effects on treatment outcomes. Although pharmacokinetic effects of single genetic variants are well documented, for example, in the antidepressant treatment of major depression (Fabbri et al., 2018), most such variants are structural, often involving gene deletion or duplication, and are more readily linked to metabolic effects on drug exposure than SNPs. For pharmacodynamic effects, such as those examined here, the use of PGS may provide a valid alternative to candidate gene studies. For that potential to be realized, target samples must be large enough to yield adequate statistical power to detect an interaction of PGS with treatment on an alcohol‐related outcome. However, because RCTs are costly and time‐consuming, the growing availability of electronic health record (EHR) data may provide samples large enough to identify genetic moderators of alcohol pharmacotherapies.

In addition to the small target sample in the present study, there are other limitations that should be acknowledged. Despite no measured pretreatment differences across the two RCTs, combining them could have introduced unmeasured confounding unremedied by the inclusion of study as a covariate in the analyses. Because in Kranzler et al. (2014), patients were diagnosed with DSM‐IV alcohol dependence, while in Kranzler, Hartwell, et al. (2021), they were diagnosed with DSM‐5 AUD, we could not directly compare the diagnostic severity of the two samples. However, the samples were comparable on mean scores on the Short Index of Problems (SIP) (Miller & Tonigan, 1995), values that were consistent with mild‐to‐moderate AUD severity. The PGS for problematic alcohol use, time to relapse to first drink, and time to relapse to heavy drinking interacted with treatment group on SIP score but not HDDs. Whereas both outcome measures are self‐reported and retrospective, they are subject to recall bias and underreporting. Thus, the differential findings may reflect more accurate reporting of alcohol‐related problems, as these yielded more consistent evidence of moderation by PGS, perhaps because they were less error prone. Because neither study required patients to be abstinent, nor was abstinence a required goal of treatment, the observed effects may underestimate the therapeutic benefit that an abstinence‐oriented trial of topiramate might yield (Bujarski et al., 2013). Finally, as all patients included in the analysis were of European ancestry and most were male, the findings may not generalize to other population groups or female patients.

The study's strengths include the large discovery samples available for generating the PAU and DPW PGS. Second, the study's use of PGS based on genome‐wide genotyping is a more powerful approach to characterizing complex pharmacogenetic traits like treatment moderation (Kember et al., 2024; Torkamani et al., 2018). Finally, the topiramate studies, which comprised the target sample, used rigorous study designs and were characterized by high levels of treatment retention and medication adherence, all of which support the validity of the findings.

CONCLUSIONS

Successful identification and testing of pharmacogenetic effects of medications for treating AUD will likely depend upon using other methods than candidate gene studies. One promising approach is the use of PGS as indicators of genetic liability. This will require the selection of optimal discovery samples for generating PGS and ensuring that target samples are large enough to provide adequate statistical power. Efforts in this area have the promise of informing AUD pharmacotherapy by matching patients to the treatments most likely to reduce their drinking behavior and minimizing adverse pharmacological effects.

FUNDING INFORMATION

This study received support from the National Institute on Alcohol Abuse and Alcoholism grants AA03510, AA013736, AA023192, AA030056, and AA028292 (to RLK) and the Veterans Integrated Service Network 4 Mental Illness Research, Education, and Clinical Center.

CONFLICT OF INTEREST STATEMENT

Dr. Kranzler is a member of advisory boards for Clearmind Medicine and Lilly Pharmaceuticals; a consultant to Sobrera Pharmaceuticals and Altimmune; the recipient of research funding and medication supplies for an investigator‐initiated study from Alkermes; a member of the American Society of Clinical Psychopharmacology's Alcohol Clinical Trials Initiative, which was supported in the last three years by Alkermes, Dicerna, Ethypharm, Imbrium, Indivior, Kinnov, Lilly, Otsuka, and Pear; and an inventor on US provisional patent “Multi‐ancestry Genome‐wide Association Meta‐analysis of Buprenorphine Treatment Response.”

Supporting information

Figure S1

Figure S2

Table S1

Kranzler, H.R. , Jinwala, Z. , Davis, C.N. , Xu, H. , Biernacka, J.M. , Zhou, H. et al. (2025) Moderation of treatment outcomes by polygenic risk for alcohol‐related traits in placebo‐controlled trials of topiramate. Alcohol: Clinical and Experimental Research, 49, 1297–1305. Available from: 10.1111/acer.70052

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.