Unobserved heterogeneity in the effect of a randomized-controlled prevention trial among men who have sex with men
D. Smolenski, J.M. Wilkerson, B.R.S. Rosser
University of Minnesota School of Public Health, Division of Epidemiology and Community Health, Minneapolis, United States
Background: The Men’s INTernet Study randomized trial identified a short-term, but no long-term effect in reducing high-risk sexual behavior. In this investigation, we considered the impact of unobserved heterogeneity in both the treatment and control arms to determine if there were subgroups that varied as a function of change over time.
Methods: In 2006, 650 MSM were randomized to complete an online intervention or a standard questionnaire. The primary outcome was the number of male unprotected anal sex partners (UAIMP) over the three months prior to study enrollment and at each of the 3-, 6-, 9-, and 12-month follow-ups. We estimated iterative latent classes of the five time-point growth curves using a semi-continuous model to account for zero inflation. We tested both linear and quadratic growth curves.
Results: The quadratic growth curves produced better fit to the data. Model estimation yielded three comparable classes in each exposure group. Comparison between treatment and control identified that the high-risk (10% of sample) and low-risk (64% of sample) curves had equivalent parameterization. The moderate risk curve (26% of sample) had an equivalent intercept, but a steeper linear slope for the treatment group. Figure 1 displays the growth curves.
Conclusions: We identified a longitudinal effect of treatment for men who began the trial with a moderate number of UAIMP. The observed reductions in the other two classes were equivalent for both the intervention and control arms. Results suggest targeting online HIV prevention interventions for MSM based on moderate frequency of risk behavior.