The focus of this brief review is on the four drug products approved by the FDA in the United States for the treatment of alcohol dependence. (Although other medications are approved for specific use in treating acute alcohol withdrawal symptoms, that indication is outside the scope of this brief review.) The applications for three of these products were supported by efficacy data from clinical trials reviewed by the FDA. The basic design of the trials and analytic approaches to the data that supported marketing approval are described below and summarized in . The fourth drug, disulfiram, was approved in 1951, before the statutory requirement that drugs be demonstrated to be effective before marketing. The 1969 determination that disulfiram was effective was made during the Drug Efficacy Study Implementation process, based on literature review by a panel convened by the National Academy of Sciences/National Research Council. A larger emphasis in this overview has been placed on Vivitrol, a long-acting injectable form of naltrexone, since it was the most recent approval and therefore best reflects the current FDA approach to this area of drug development.
Design of Studies Supporting Regulatory Approvals of Alcoholism Treatment Medications in the United States
The 1994 approval of oral naltrexone tablets, marketed as ReVia, for the treatment of alcohol dependence, was based on two investigator-initiated studies. The FDA did not review either protocol prospectively, and one specific primary endpoint was not required. The study population in both trials consisted of alcohol-dependent patients who were abstinent at study entry. In one of the studies, a minimum of 10 days of abstinence was emphasized in the review. A variety of analyses were considered, including the proportion of subjects maintaining abstinence and the proportion completing the study without relapse to heavy drinking over the 12-week observation period. Both studies are described in the published medical literature (Volpicelli et al, 1992
; O'Malley et al, 1992
) but it should be noted that the FDA analyzed the raw data and had access to some additional data not included in the published reports.
The application for acamprosate, marketed as Campral, approved in 2004, was supported by three pivotal trials, one lasting 3 months and two lasting roughly a year each. In addition, safety data and summaries of efficacy results were provided for six additional 6-month studies and three 1-year studies. All of these studies were designed and completed outside of the United States before the initial interactions occurred between the commercial sponsor and the FDA; therefore, the FDA did not provide input on study design or analysis. A single 6-month study in the United States was also conducted; this study did not demonstrate a significant advantage for acamprosate over placebo.
The protocols for two of the pivotal studies did not clearly specify a primary analysis, and various analyses were performed. The third pre-specified a survival analysis. The sponsor proposed that an analysis of the mean number of cumulative days of abstinence be considered the primary endpoint. However, the method of ascertainment of the number of drinking days was not sufficiently systematic to allow for precise counting of the number of days drinking or not drinking. The common endpoint applied (retrospectively) to all three studies was the percentage of patients remaining continuously abstinent throughout treatment, because this seemed to be credibly determined and represented a clear clinical benefit. In the three European pivotal efficacy studies, subjects randomized to acamprosate were more likely than subjects randomized to placebo to be assessed by the clinician as having been abstinent http://www.accessdata.fda.gov/drugsatfda_docs/nda/2004/21-431_Campral.cfm
The 2006 approval of depot (long-acting injectable) naltrexone, marketed as Vivitrol, was based on efficacy data from a single study, using a regulatory pathway that permitted the application to rest, in part, on the FDA's previous finding of efficacy for oral naltrexone (ie, 505(b) (2) submission). The previous Agency finding of efficacy for oral naltrexone was based, as noted above, on two studies in recently detoxified alcohol-dependent subjects abstinent at study entry. Notably, the labeling for oral naltrexone does not describe this feature of the studies or stipulate that patients should be abstinent at treatment initiation. The single study submitted in support of efficacy was a 6-month placebo-controlled study in alcohol-dependent outpatients who were not required to refrain from drinking between screening and study drug initiation, although some did. In designing the study, the sponsor had discussed with the FDA the possibility of defining patterns of drinking behavior short of complete abstinence as treatment success. Although it has also been noted that the consumption of limited quantities of alcohol may actually confer a health benefit in patients without alcohol use problems, conventional wisdom has long held that sustained, controlled drinking at low levels has not seemed feasible for the alcohol-dependent patient. However, to explore the possibility that this feature of alcoholism may be amenable to treatment (perhaps pharmacologic treatment in particular), this study was designed to evaluate the event rate of heavy drinking. A heavy drinking event was defined as a day in which a male patient consumed at least five standard drinks and a female patient consumed at least four standard drinks (one standard drink is equivalent to 12
oz of 5% beer, 5
oz of 12–17% wine, 3
oz of fortified wine, or 1.5
oz of 80 proof liquor). The event rate was to be calculated over the duration of treatment (ie, over 24 weeks or up to the time of treatment discontinuation). A number of secondary analyses of drinking patterns were also pre-specified, including analyses of the proportion of patients who were treatment responders (responder analyses) as requested by the FDA.
The protocol specified an analysis that aggregated the drinking data by treatment group rather than evaluating the percentage of patients in each treatment group who were considered ‘treatment responders'. However, the development program for this product took place when outcome measures for alcoholism treatment trials were in evolution. During the course of development, the Agency became aware of information developed by NIAAA that provided empirical support for the selection of a responder definition based on drinking patterns. The FDA's understanding of the information was that individuals who drank occasionally, but never heavily, during the observation or survey period, were at very low risk of experiencing the adverse social and occupational consequences of alcohol use, even if those individuals had a history of alcohol problems. Therefore, a pattern of drinking that included ‘no heavy drinking days' could be considered a ‘non-risky' drinking pattern, even in alcohol-dependent patients. The FDA believed that the proportion of patients achieving and maintaining complete abstinence from alcohol drinking remained an endpoint of indisputable significance, but noted that the evidence suggested that the proportion of patients able to maintain a non-risky drinking pattern would be of similar significance. Although an analysis based on alcohol consumption provides only indirect evidence of treatment benefit (true clinical benefit would need to capture the impact of treatment on how patients actually feel and function, and improved psychosocial, occupational, or physical well-being is often not measured and is unlikely to be captured in a 6-month study), this responder definition (ie, no heavy drinking days over the study period) appeared to be an appropriate surrogate for clinical benefit. Therefore, FDA review of the application gave greater attention to the analyses that were originally designated as secondary.
As a result of the lack of prior information about the effect of naltrexone in subjects drinking at baseline, FDA reviewers also examined (using both responder analyses and event rate analysis) the effect of depot naltrexone on the subgroups of patients who were abstinent before treatment initiation, compared with the patients who were drinking.
Only 9% of study participants did not drink alcohol during the 7-day lead-in period before treatment initiation. In this subgroup, the applicant's analysis showed that treatment with depot naltrexone was associated with a very substantial reduction in the event rate of heavy drinking. Even with fewer than 20 subjects per group, statistical significance was reached in this subgroup. In contrast, only a nonsignificant reduction in heavy drinking, using the applicant's analysis, was seen in the 91% of patients who consumed alcohol in the week before treatment initiation.
The FDA's approach to the responder analysis demonstrated a similar phenomenon. For the responder analysis, patients who did not drink heavily from the end of the ‘grace period' to the end of the treatment period were operationally defined as responders. FDA allowed for a grace period in the analysis, during which patients who sampled alcohol would not be counted as treatment non-responders for two reasons: to allow for engagement in treatment, and to address the hypothesis that naltrexone works by reducing the reward experienced when alcoholics consume alcohol, thus decreasing the potential for a ‘lapse' to turn into a ‘relapse.' In the FDA's analysis, in the study population as a whole, few subjects achieved a complete absence of any heavy drinking days, and no effect of naltrexone treatment was apparent when the rates were inspected. In contrast, the success of the very small subgroup who were abstinent at baseline was impressive and a clear treatment effect was observed http://www.accessdata.fda.gov/drugsatfda_docs/nda/2006/021897_toc
Therefore, the study drug seemed to be effective only in the small subgroup who did not drink during the lead-in period. Taking into consideration the lack of any prior evidence that naltrexone has an effect on drinking in patients who are drinking at treatment initiation, the FDA determined that efficacy had been demonstrated only in the subgroup of patients who have been able to refrain from drinking for at least a brief period (several days) at the time of treatment initiation.
Going forward, the FDA generally views alcoholism as a chronic disorder and alcoholism treatment drugs as maintenance treatments. The current approach of chronically administered drugs for chronic conditions, in general, is that the default observation period is 12 weeks (after any grace period, titration, and so on). However, if there is a reason that the effect cannot be observed in that time period, the studies should be longer. For example, obesity studies are 1 year in duration http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/
. The FDA has been asking for 6-month studies for alcoholism treatment because there are data indicating that drinking patterns over 3 months may not be stable or representative of future experience (Zweben and Cisler, 2003
). It is also reported that periods of abstinence are quite common among alcohol-dependent individuals: periods of abstinence lasting at least 3 months were reported by 62.3% of a heterogeneous group of alcohol-dependent individuals (many not treatment seeking) through retrospective/historical interviews (Schuckit et al, 1997