Commentary on Garbutt et al. (2014): Can we predict who benefits from naltrexone in the treatment of alcohol dependence?

Abstract

Naltrexone, an opioid antagonist, was approved by the U.S. Food and Drug Administration in 1994 to treat alcohol dependence. It has subsequently been widely investigated as a treatment for alcohol dependence and heavy drinking using a variety of study designs. The most comprehensive meta-analysis of naltrexone’s efficacy showed that it exerts a modest effect on multiple drinking outcomes, the greatest effect being a 17% reduction in the risk of relapse to heavy drinking compared to placebo (1). Multiple studies, beginning shortly after the medication’s approval and continuing to the present, have sought to identify and validate moderators of naltrexone’s efficacy. These studies selected certain clinical or genetic features to identify alcohol-dependent individuals who were most likely to show a robust treatment response. In this issue of the journal, Garbutt et al. (2) systematically review 622 studies in an effort to discern factors that moderate the therapeutic response to the use of naltrexone in alcohol treatment. The authors focused on moderating variables for which at least two published studies were identified. The moderators considered included a family history of alcohol dependence; pretreatment craving for alcohol; the Asn40Asp single nucleotide polymorphism (rs1799971) in OPRM1, which encodes the mu-opioid receptor; sex; pretreatment drinking patterns; psychopathology; alcoholism subtype; and sweetliking/disliking. The review applied established criteria to assess the strength of the evidence by rating the studies on risk of bias, directness of the link between the intervention and an important health outcome, consistency of the findings, and precision of the estimated effect. Garbutt and colleagues conclude that, despite promising findings of moderation, most notably for a family history of alcoholism or the presence of the rs1799971*Asp40 allele, the literature is inadequate to recommend a personalized treatment approach to the use of naltrexone to treat alcohol dependence. It is hard to disagree with the authors in view of the evidence that they have marshaled: only 28 of the 622 studies reviewed met the requirement of being either a placebo-controlled study of naltrexone with analysis of a moderator variable or of naltrexone treatment in which groups with and without a moderating feature were compared. The small fraction of naltrexone studies that were suitable for inclusion in the meta-analysis underscores the limits of the literature. Of perhaps greater concern, though, is the observation that no prospective study has yet been published that evaluated a moderator of naltrexone response as its primary objective. Two prospective studies, as yet unpublished, but listed on www.clinicaltrials.gov, are not included in the analysis. These studies, in which alcohol-dependent patients are being allocated randomly within Asn40Asp genotype group to treatment with naltrexone or placebo, will be an important addition to the growing literature on this potential moderator of naltrexone treatment response. Not surprisingly, Garbutt and colleagues concluded that the strength of evidence for moderation of the response to naltrexone treatment of alcohol dependence was low. This sober appraisal of the literature is an important contribution to the effort to personalize the use of naltrexone to treat alcohol dependence, as it highlights the need for more studies, particularly prospective studies that are adequately powered to detect moderation. The same careful consideration that Garbutt and colleagues argue for in evaluating the use of naltrexone should characterize any effort to identify moderators of alcohol use disorder treatment. This includes, for example, pharmacogenetic studies of ondansetron (3,4) and topiramate (5), both of which yielded promising initial results, but require prospective validation before the use of these genetic moderators can be recommended for clinical use. It is also important not to forget that the large-scale effort embodied by Project MATCH−the primary intention of which was to identify clinical and psychosocial predictors of response to different psychotherapies (6)–produced little evidence of moderation despite the large sample and prospective design that characterized it (7,8). Thus, although large, prospective studies are a necessary feature, they are not sufficient to demonstrate the efficacy of a moderator approach in alcohol treatment. Of equal importance to this effort is the care with which the hypothesized moderators are selected. Candidates for moderation should be based firmly on empirical evidence, rather than being of primarily theoretical interest, and they must be measured reliably. Further, moderator analyses require attention to potential confounders (e.g., population heterogeneity). Although there is great potential value in the personalized treatment of alcohol use disorder (9), the challenge in achieving that potential will be to identify novel candidate moderators and test them in large-scale, well controlled clinical trials, particularly during a time of scarce research resources.