Contingency management to reduce methamphetamine use and sexual risk among men who have sex with men: a randomized controlled trial

We enrolled MSM from Seattle, WA. Between June 2007 and August 2008, participants were recruited through referrals from an STI clinic; an HIV clinic; community-based organizations (CBOs); printed media; radio; community outreach; and, referral by peers participating in the study.

This study was initially designed to inform the design of a trial to test whether CM could prevent HIV acquisition among HIV-negative men who use methamphetamine. Therefore, from June 2007 to November 2007, we enrolled only HIV-negative men. Because of slow enrollment of HIV-negative men, in November 2007, we opened the study to HIV-positive men and revised the study objective to test whether a CM intervention could reduce HIV acquisition and transmission risk. Several studies illustrate that a considerable proportion of HIV-positive men use methamphetamine and that methamphetamine facilitates HIV transmission behaviors [9]. Additional inclusion criteria were: age ≥18 years; willingness to be randomized and provide locator information; and, no plans to move from the study catchment area within 6 months of enrollment. Behavioral eligibility criteria were ≥1 episodes of anal sex and ≥2 episodes of methamphetamine use in the month prior to screening. We did not require a methamphetamine-positive urine sample for enrollment as we thought this requirement would limit the utility of CM as a public health intervention in practice. We excluded participants who reported a mutually monogamous relationship with a partner of the same HIV status lasting ≥2 years and men who expressly asked for detoxification, counseling, or drug rehabilitation services. HIV status was the only eligibility criterion modified during the course of the study.

The study intervention was a 12-week CM program, adapted from previous studies, in which vouchers of escalating value were offered for consecutive urine samples that did not contain methamphetamine or crack or cocaine (herein referred to as cocaine) metabolites [15]. Initially, the intervention consisted of thrice-weekly drop-in urine testing visits for a total of 36 visits. Vouchers started at $2.50 for the first stimulant-free sample and increased by $1.25 for every consecutive stimulant-free sample thereafter up to a maximum of $10.00. Participants submitting three stimulant-free urine samples in a row earned a $10.00 bonus. The maximum payout for this program was $453.75, similar to the payout in other programs [17, 18]. When participants submitted a stimulant-containing sample, or missed a visit, no vouchers were issued and the value of the voucher for the subsequent stimulant-free sample was reset to $2.50. If a participant submitted a week of stimulant-free samples after submitting a sample containing stimulant metabolites, he returned to the voucher value prior to the stimulant-containing sample ("rapid reset") [19]. Vouchers were redeemable immediately upon accrual for pre-paid gift cards and goods and services; we never offered cash. All CM participants signed a contract delineating the expectations of the CM program [20]. Study personnel administering the CM intervention followed a simple, scripted protocol for the reporting of results of urine testing and administering vouchers. Such a protocol was used to avoid the provision of counseling around the results of the urine testing. The protocol, which required no formal training, was developed by one of us (SS) who has extensive expertise in delivering CM interventions.

All seven participants enrolled in the study while the above procedures were in place reported difficulty adhering to the intervention schedule, and only two attended ≥12 of their 36 visits. In response, in September 2007, we reduced the number of weekly urine testing visits to two (24 visits over the 12-week intervention) and increased the value of vouchers for the first stimulant-free urine sample to $7.50; other studies have employed a similar schedule [21]. As before, these vouchers increased by $1.25 for each consecutive stimulant-free sample to a maximum of $10.00. Additionally, we gave participants a $20.00 bonus for two consecutive stimulant-free samples. The maximum payout for this program was $476.25. We continued to withhold vouchers and reset voucher values to baseline for urines containing stimulants and for missed visits; however, to encourage participants to attend visits, we gave men submitting stimulant-positive samples a voucher worth $2.50.

Under the initial CM intervention schedule, drop-in urine-testing visits were available from 10:00 am to 6:00 pm on Tuesdays, Thursdays, and Saturdays; we offered extended hours for working participants. After the enrollment visit, we sent postcards or e-mails to all participants encouraging participation in the intervention. We reminded participants who did not attend urine-testing visits for the first week by phone or e-mail. We sent postcards, phoned, or e-mailed all CM participants again at the midpoint of the intervention period. Under the revised CM intervention schedule, drop-in urine testing visits were available from 10:00 am to 6:00 pm on Tuesdays and Saturdays with flexible hours for working participants. Postcard, phone, and e-mail reminder strategies remained the same.

We tested urine samples with the QuickScreen Pro Multi-Drug Screening Test (Phamatech, Inc., San Diego, CA), a point-of-care test used to qualitatively detect stimulant metabolites. For this assay, the estimated mean detection time in urine ranges from 43.6 to 66.9 hours for methamphetamine [22] and is 88.4 hours for benzoylecgonine, a cocaine metabolite [23]. We repeated 10% of all urine tests; none were discordant. Study staff monitored the collection of all urine samples and tested the samples immediately after their provision.

Participants randomized to both control and CM arms received a printed list of local counseling, treatment, and outreach services at baseline and at each study visit. Study staff offered all participants assistance accessing services. Control participants did not submit twice-weekly urine samples and did not receive vouchers during the first 12 weeks of the study.

The study randomized participants using block sizes of two, four, and eight varied randomly with a pseudo-random number generator, a deterministic process use to generate an effectively unpredictable sequence of numbers [24]. The randomization list was used to assemble sequentially numbered, sealed, opaque envelopes containing intervention arm assignments. The research coordinator and principal investigator were blinded to the randomization code. From June 2007 to April 2008, participants were randomized 1:1 to the intervention and control arms under a protocol funded by the National Institute on Drug Abuse. In April 2008, we received additional funding from Public Health--Seattle & King County (PHSKC) to deliver the CM intervention, but not to enroll additional control participants. Therefore, between April 2008 and August 2008, we randomly assigned participants 3:1 to the intervention and control arms. We conducted analyses with and without the men enrolled under the 3:1 randomization scheme. Results were similar and we present the results of analyses that included all randomized men in this manuscript. While uneven randomization schemes may affect a trial's statistical efficiency, they have no impact on a trial's validity [25]. Follow-up ended in February 2009.

The same study personnel administered the CM intervention and conducted the study visits. These personnel also performed data entry of the results of the urine testing and HIV/STI testing. The behavioral endpoints were automatically transferred from the ACASI to a database without the potential for modification. It was not possible to blind those administering the CM intervention to a participant's study arm.

The University of Washington institutional review board approved the study protocol. All participants provided written informed consent.

All study procedures took place at a large, public-transport accessible, community-based AIDS service organization. We screened participants for eligibility by phone and in person. All participants attended scheduled study visits every six weeks for six months. At each study visit, participants completed an audio computer-assisted self-interview (ACASI) that used a 6-week recall period and included questions about sexual behavior and substance use at the respondent- and partnership-level. Participants submitted urine samples for methamphetamine and cocaine metabolite testing at each of these visits.

At enrollment, 12 weeks, and 24 weeks, participants were tested for HIV/STI after completing the ACASI [26]. Participant-centered risk reduction counseling accompanied all testing [27], and condoms and lubricant were offered to all participants. Given the connections between substance use and sexual risk, HIV risk reduction counseling often included discussions of substance use.

The primary outcome was report of unprotected anal intercourse with a partner of unknown or discordant HIV status (non-concordant UAI) in the prior six weeks. The study included four secondary endpoints: the number of non-concordant UAI partners; results of methamphetamine urine testing; self-reported weekly or more frequent use of methamphetamine; and, self-reported use of >8 quarters (two grams) of methamphetamine. We chose non-concordant UAI as the primary outcome because this study was designed as a preliminary test of whether CM can be ultimately employed as an HIV prevention tool. We have previously shown this metric to be associated with HIV acquisition [28]. The secondary substance use outcomes represent intermediate variables in the causal chain in which methamphetamine use may facilitate the high risk sex that in turn leads to HIV/STI transmission.

We conducted a test-retest reliability study of the behavioral endpoints assessed in this trial [29]. From December 2006 to March 2007, we enrolled 102 MSM recruited from an STI clinic and an HIV clinic to complete the ACASI employed in this study. Of the 102 men who completed an initial ACASI, 98 (96%) returned to complete a second ACASI in a median of 4 days (range: 2-5). Participants took no more than 38 minutes to complete the initial ACASI. We calculated reliabilities using kappa statistics for binary outcomes and intraclass correlation coefficients by repeated measures analysis of variance (ANOVA) for continuous outcomes. The reliabilities of self-reported non-concordant UAI, weekly or more frequent use of methamphetamine, number of non-concordant UAI partners, and number of quarters of methamphetamine were 0.72 (95%CI: 0.47, 0.87), 0.88 (95%CI: 0.62, 0.99), 0.85 (95%CI: 0.78, 0.91), and 0.96 (95%CI: 0.94, 0.99), respectively.

The primary purpose of this trial was to gather data for the design of a phase IIb or III clinical trial. The study was originally designed to follow 60 individuals. Assuming that 14% of participants would report non-concordant UAI at baseline, this sample size was chosen to provide estimates of the proportion of men who report non-concordant UAI in the prior six weeks with an expected width of the 95% confidence interval (CI) of 0.085. We expected to lose 25% of our study population to follow-up and aimed to enroll 80 participants. Additional resources, however, allowed us to enroll 127 participants.

We used chi-squared tests and t-tests to compare retention at the 24-week study visit and CM intervention adherence, respectively, between groups of participants defined by selected characteristics. We used generalized estimating equations (GEE) to estimate the proportion of participants reporting non-concordant UAI in the prior six weeks at each study visit. Analyses estimating the effectiveness of CM were intention to treat, on the basis of a binomial regression models with log links, to calculate relative risks comparing the proportion of men reporting non-concordant UAI between the study groups at study visits during the 12-week intervention and 12-week follow-up periods [30]. Our analyses of secondary binary endpoints employed the same analytic methods. Finally, we used a GEE model with a log link and negative binomial errors to evaluate the intervention's effect, expressed as a rate ratio, on the number of non-concordant UAI partners. We defined statistical significance at the P < 0.05 level.

We pre-selected baseline covariates that we thought would most strongly predict our outcomes of interest for adjusted analyses, including predictors of retention at the 24-week study visit. Of these covariates, we adjusted for variables that resulted in at least a 10% change in the adjusted RR (aRR) compared to the unadjusted RR [31]. We chose such strict criteria because of our relatively small sample size and because several potential confounders differed between our intervention and control groups. All analyses included a covariate indicating the baseline value of the outcome of interest. Estimates of the proportion of men reporting non-concordant UAI at each visit and analyses comparing any non-concordant UAI and number of non-concordant UAI partners between CM and control were also adjusted for HIV status and use of other substances (inhaled nitrites, gamma-hydroxybutyrate, ecstasy, or erectile dysfunction medications) in the prior 6 weeks at baseline. Analysis of detection of methamphetamine use by urinalysis was also adjusted for stage of change for methamphetamine use at baseline. We note that duration of methamphetamine use and injection use of methamphetamine were both evaluated as potential confounders; their inclusion in models that already included the above variables produced similar results to analyses without these two predictors.

We assessed participants' stage of change for methamphetamine use by a single question [32, 33]: "Have you been trying to cut back or stop using methamphetamine?" This question was added after the first 19 participants were enrolled as it was not a variable we had initially planned to collect. The pattern of missingness, however, is missing completely at random (MCAR) which is unlikely to introduce bias to complete case analyses like the one presented in this manuscript [34].

Over our 15-month enrollment period, we screened 222 men for eligibility (14.8/month) and enrolled and randomized 127 men (8.5/month). Of these randomized participants, 107 (84%) attended the 24-week study visit; retention at 6-week assessments was similar for both groups (Figure 1). HIV-negative men were less likely to attend the 24-week study visit than HIV-positive men (74% v. 93%, P < 0.01) and men who reported methamphetamine use for ≥10 years were less likely to attend the 24-week study visit than men who reported methamphetamine use for <10 years (81% v. 94%, P = 0.05).

The largest difference in follow-up between the CM and control group was for the first visit after baseline (69% v. 77%). Of the 22 CM participants loss-to-follow-up at this visit, one was incarcerated, one was in a drug rehabilitation program, three withdrew from the study, and 17 could not be located for study visit attendance. Among the 13 control group participants who did not complete the first follow-up visit, one had died, two withdrew from the study, and 10 could not be located for study visit attendance.

Table 1 presents characteristics of the study population. The only statistically significant difference between the study populations at baseline was that between the distributions of race/ethnicity (P = 0.03).

Table 2 provides metrics of intervention adherence among CM participants. Participants who earned <$15,000 per year attended more visits than participants who earned ≥$15,000 per year (41% v. 20%, P = 0.04). Participants who reported using methamphetamine weekly or daily at baseline attended fewer visits than participants who reported less frequent use (31% v. 51%, P = 0.02) and participants whose urine contained methamphetamine at baseline attended fewer urine testing visits than participants whose urine did not contain methamphetamine (26% v. 43%, P = 0.02). Since the mean detection time of the methamphetamine assay is shorter than the interval between urine testing visits, the proportion of metabolite-free urine samples provided by CM participants during the intervention visits may be over-estimated.

At baseline, 6-weeks, 12-weeks, 18-weeks, and 24-weeks, the adjusted probabilities of reporting non-concordant UAI in the prior six weeks were 31.0% (95%CI: 27.6%, 34.3%), 19.7% (95%CI: 16.3%, 23.1%), 12.1% (95%CI: 9.9%, 14.4%), 9.1% (95%CI: 7.2%, 10.9%) and 10.2% (95%CI: 8.3%, 12.1%), respectively. Figure 2 presents the adjusted proportion of men reporting non-concordant UAI at each study visit by study arm. Comparing participants at study visits six weeks apart, participants at the later study visit were less likely to report non-concordant UAI in both the CM (aRR = 0.80; 95%CI: 0.64, 0.99) and control (aRR = 0.84; 95%CI: 0.71, 1.00) arms than participants at the earlier study visit.

CM and control participants were similarly likely to report non-concordant UAI during the 12-week intervention period. CM participants were less likely than control participants to report non-concordant UAI during the subsequent 12-week follow-up period, and also reported fewer non-concordant UAI partners than control participants during both the intervention and follow-up periods. None of these differences were statistically significant (Table 3).

One man in CM arm and one man in the control arm acquired pharyngeal gonoccocal infection (3.4 v. 4.5 per 100 person-years; P = 0.43). There were no incident syphilis, rectal gonococcal, or rectal chlamydial infections. Among the HIV-negative participants, one man in the CM arm and no men in the control arm acquired HIV (8.9 v. 0 per 100 person-years; P = 0.27).

During the 24 weeks of follow-up at the 6-week study visits attended by both CM and control participants, 28 (49%) control participants and 44 (63%) CM participants submitted ≥1 sample containing methamphetamine metabolites; 26 (46%) control participants and 23 (33%) CM participants submitted ≥1 sample containing cocaine metabolites; and, 43 (75%) control participants and 55 (79%) CM participants submitted ≥1 sample containing methamphetamine or cocaine metabolites.

CM and control participants were comparably likely to submit urine samples positive for methamphetamine at study visits during the intervention period (Table 4); however, during the follow-up period, CM participants were somewhat more likely to submit a urine sample containing methamphetamine (P = 0.11).

Comparing participants at study visits six weeks apart, control participants at the later study visit were less likely to report either weekly or daily methamphetamine use or using >8 quarters of methamphetamine in the prior six weeks than participants at the earlier study visit (aRR = 0.77; 95%CI: 0.68, 0.88 and aRR = 0.69; 95%CI: 0.56, 0.86, respectively); although the frequency and quantity of methamphetamine use declined over time among men in the CM arm, these changes were not as large with respect to frequency (aRR = 0.89; 95%CI: 0.83, 0.96) and not statistically significant with respect to quantity (aRR = 0.93; 95%CI: 0.85, 1.02). CM participants were significantly more likely than control participants to report weekly or daily methamphetamine use and to report using >8 quarters of methamphetamine at each study visit; these differences were statistically significant during the 12-week follow-up period (Table 4).

Accounting for baseline stage of change, control group participants were more likely than CM participants to report being in the action or maintenance stage of change for methamphetamine use during the intervention (74% v. 63%, P = 0.57) and follow-up periods (81% v. 66%, P = 0.18); however, these differences were not statistically significant. In the 6 weeks prior to baseline, 19% of control participants and 14% of CM participants participated in a substance abuse treatment program or peer support group (P = 0.45). After baseline, 26% of control participants and 21% of CM participants reported participating in such interventions (P = 0.38).