Comments on Ritter et al.'s report of a comparative trial of LAAM and methadone maintenance.

Sir—Ritter et al .’s [1] claims their results show ‘LAAM maintenance has equivalent efficacy to methadone’ and ‘both treatments were equally effective in retaining clients in maintenance treatment’ are misleading. Their retention results actually favour methadone and their study was too small to identify confidently a clinically significant difference between the two maintenance drugs. The triallists randomly assigned 101 primary care maintenance patients who had been receiving methadone for at least 8 weeks to either continue with methadone or to receive l-alpha acetylmethadol (LAAM). Treatment and assessment were unblinded with doses determined by the subjects’ general practitioner. After 12 months, 35 of the 52 methadone subjects and 28 of the 49 LAAM subjects were still receiving their assigned treatment. Unfortunately, Ritter et al . did not use these data in their analysis. Rather, they excluded three methadone and five buprenorphine subjects who either left or withdrew before receiving their first dose. This violates the principle of analysis by intention to treat under which all randomized patients should be accounted for and analysed in their assigned groups [2]. The violation biases retention in favour of buprenorphine. The triallists also classified as retained in treatment an additional nine LAAM patients who had switched back to methadone. As methadone subjects did not have a similar opportunity to change maintenance drugs, this also biased the published retention results in favour of buprenorphine. This is why Ritter et al .’s results suggest that methadone subjects tended to be more likely to leave in the first year of treatment RR 1.12 (95% CI, RR 0.95–1.47, P = 0.14) and that buprenorphine reduced the absolute risk of leaving by 12.7% (95% CI, ARD 4.0%–29.3%). When retention in assigned treatment data from all randomized subjects are considered, methadone subjects were 85% as likely to leave treatment as LAAM subjects (95% CI, RR 0.62–1.15, P = 0.29). Using the absolute reduction in the risk of subjects’ leaving maintenance prematurely (10.2%; 95% CI, ARD 8.7%–29.0%), it is necessary to treat 10 patients for a year with methadone to prevent one premature loss that would have occurred with LAAM. While these effects are clinically important, they are not statistically significant. The width of the confidence intervals suggests this was because trial lacked statistical power. This was confirmed by a series of power calculations. If one treatment actually increased by 20% the proportion of subjects retained for a year, a study of 100 subjects would have only an 18% chance of identifying a statistically significant difference in retention (experimental 0.72, control 0.60, a = 0.05). The trial also lacked the power to identify statistically significant differences in the four self-reported outcomes: the proportion of subjects using heroin in the previous month; cost of heroin per day; days used heroin in the last 28; and changes in the OTI Q score, (Table 1). At least 450 subjects are needed to be confident that a study would detect clinically significant differences between methadone and buprenorphine. Indeed, even this estimate is optimistic as the assumed differences in outcome are greater than would be expected [3]. Ritter et al . were wrong to claim their failure to identify a statistically significant difference in outcome provides evidence that methadone and LAAM are equivalent. Rather, it was almost certainly caused by their study’s lack of statistical power. To the contrary, when properly analysed, their data suggest retention was better in methadone maintenance. This is consistent with the results of comparisons of methadone and LAAM conducted in maintenance clinics [3]. Table 1 The number of subjects required to detect differences in the five outcomes; b = 0.80, a = 0.05.