In this randomized controlled trial of CM, we enrolled 127 non-treatment-seeking MSM over 15 months and successfully collected follow-up data on over 80% of participants. These findings suggest that a randomized controlled trial that specifically seeks to enroll this population is acceptable and feasible. However, less than half of our participants were HIV-negative and restricting enrollment to HIV-negative MSM who use methamphetamine, the population of interest for a trial using HIV as an outcome, would be challenging. Unlike HIV-negative men, HIV-positive men are often enrolled in ongoing medical care and case management. HIV-positive men may also be more likely to recognize their methamphetamine use as problematic with some attributing their HIV infection to their methamphetamine use . In addition, the success of peer recruitment (over 50% of HIV-negative participants were recruited through peers) over place-based recruitment strategies that have been successful in other HIV prevention trials  indicates that HIV-negative MSM who use methamphetamine may not share the same social and geographic spaces as HIV-negative men who do not use methamphetamine. HIV-negative men may have also been uncomfortable participating in a trial conducted at an organization well known for its services for HIV-positive individuals.
Unfortunately, our findings related to CM are discouraging. First, a relatively small proportion of men consistently provided urine specimens for the intervention meaning that actual exposure to the intervention among men assigned to CM was quite limited. Several factors may contribute to CM participants receiving such a low "dose" of the intervention, including the location, the magnitude of the incentives, time limitations, and monitored collection of urine samples. Anecdotally, many men reported not attending urine-testing visits because they knew their urine would test positive. This anecdotal evidence is supported by our data. Out of all of the CM visits attended, a median of 85% of urine samples were free of stimulant metabolites. Second, and most importantly, our findings suggest that CM is very unlikely to be effective as a stand-alone intervention among MSM. While the 95% CI around the intervention-period aRR was very wide and does not rule out the possibility that CM could reduce methamphetamine use during the period of actual intervention, our follow-up period aRR and 95% CI exclude the possibility that CM would have a large, sustained benefit on methamphetamine use as a stand-alone intervention in a population like the one we studied. In addition, the control group reported larger declines in self-reported use over study follow-up than the CM arm. Although measures of sexual risk among our CM participants were similar to or lower than those among control participants, it is unclear how CM might result in lower or similar levels of sexual risk if it increases methamphetamine use.
We are aware of six studies evaluating CM among individuals who use methamphetamine in outpatient substance abuse treatment [15, 21, 37–40]. Five of those trials found that, compared to other psychosocial interventions, CM leads to the submission of more stimulant-free urine samples and an increased duration of abstinence from stimulants during the CM intervention period . None of the three studies that followed participants beyond the intervention period reported differential effects of CM compared to other treatments, but all three studies showed that all participants experienced sustained reductions in methamphetamine use for up to 8 months after ceasing to receive CM. Among MSM seeking outpatient substance abuse treatment, CM has been shown to be comparable in decreasing methamphetamine use and receptive UAI as other psychosocial interventions . However, none of the previously published randomized trials included a no-treatment/minimal intervention control arm making it difficult to determine whether the absence of differences between CM and other interventions reflects comparable levels of efficacy or an absence of effect with all interventions.
Our study differs in four important ways from previous studies of CM. First, while we did not exclude participants who were receiving other substance use interventions, participants were not enrolled because they were seeking drug treatment. Second, we compared CM to a minimal intervention control arm. Third, CM was employed as a stand-alone intervention. Finally, CM participants were only expected to submit urines samples two times per week rather than three times, as has been done in some, but not all studies of CM . We cannot say whether CM's failure to reduce methamphetamine use in our study reflects a lack of effect among persons who are not expressly seeking drug treatment, CM's ineffectiveness as a stand-alone intervention, the ineffectiveness of our CM schedule, or a more general lack of efficacy of CM that would have been consistently observed in other trials had those studies included no-treatment/minimal intervention control groups.
We found that self-reported methamphetamine use declined among control participants but remained relatively stable among CM participants. We cannot readily explain this effect. We do not believe that CM participants used vouchers to obtain methamphetamine, since the difference in drug use by study arm was only evident after CM recipients ceased to receive vouchers. Likewise, we do not believe our findings simply reflect more accurate reporting of methamphetamine use by CM participants since we found that CM recipients were also more likely to test positive for methamphetamine. Participants in the two study arms reported similar use of outside treatment and support services, making it unlikely that differential use of other treatments affected our results. It is possible that participation in the CM arm may have put participants in contact with other participants and CBO clients in ways that facilitated access to and use of methamphetamine. Similar phenomena have been reported in other studies [42, 43]. Also, discontinuation of twice-weekly CM visits, which may have provided a source of support for CM participants, may have led to heavier methamphetamine use among CM participants compared to controls.
Similar to a study of CM and other psychosocial interventions in the substance abuse treatment setting , we found that sexual risk declined in both of our study arms. However, in the case of the current study, our comparison arm was a minimal intervention control group. People who volunteer for studies may be more motivated to change their behavior than those who do not, participants may enroll during a period of high-risk activity and decrease their risk independent of intervention effects, and simply measuring a behavior may change a participant's behavior or reports of that behavior . Each of these explanations may contribute to why sexual risk in our study, as in past studies, declined in both control and CM arms. Had we not used a minimal intervention control group and only measured participants in the CM intervention, we might have erroneously concluded that CM reduced sexual risk. This finding emphasizes that no-treatment/minimal intervention control groups are essential to rigorous intervention evaluation.
We note that the characteristics of our study population and the outcomes among CM participants were generally similar to those in a San Francisco public health program designed to reduce methamphetamine use among MSM (PROP) from December 2003 to December 2005. Our screening (14.8/month in the current study and 10.3/month in PROP) and enrollment rates (8.4/month and 7.4/month), attendance at CM visits (37% and 41%), and incentives earned (24% and 31% of the total possible) were similar . Compared to PROP participants , participants in the present study were more likely to have used methamphetamine for >10 years (50% v. 38%), but were less likely to report weekly or more frequent methamphetamine use (64% v. 86%) and to be HIV-positive (55% v. 78%). More than 80% of men in both populations reported methamphetamine use with sex and 54% of participants reported injecting methamphetamine at enrollment. These similarities demonstrate that our trial probably closely replicated what might occur in a non-research setting.
Our study has several important limitations. First, we did not power the current study to detect differences between study groups based on methamphetamine use or sexual risk. Instead, this study was powered to provide precise estimates of the proportion of methamphetamine-using MSM who report non-concordant UAI. Despite this limitation, the precision of our post-intervention period aRR rules out the potential of CM to effect a large, sustained reduction in methamphetamine use as a stand-alone intervention, at least in the population we followed. Second, randomization did not provide us with study arms balanced with respect several pertinent variables. In particular, persons assigned to the CM arm were more likely to use methamphetamine at baseline, and higher levels of drug use have been associated with inferior responses to CM in prior studies . It is possible that our results are subject to confounding bias stemming from imbalances in variables that we did not measure. Third, we altered the CM intervention after beginning the study. These modifications may have not provided sufficient incentive for abstinence from methamphetamine use . Finally, our measures of self-reported sexual behavior and substance use are subject to social desirability bias.