Background Gonorrheal infection occurs at multiple anatomical sites as a result of different types of sex. Assuming anal sex, oral sex, rimming, and kissing transmit infection leads to seven possible routes of transmission. Recent models of gonorrheal infection have shown that the route-specific transmission probabilities cannot be directly estimated from currently available data. Here, we have illustrated how theoretical models can be used to inform epidemiological study designs aimed at estimating these transmission probabilities. This methodology that we call ‘model-based study design’ informs 1) necessary sample sizes, 2) which variables need to be measured, and 3) how sensitive resulting estimates are to the analytical model misspecification.

Methods We simulated cohorts of high risk MSM over 2 years, where every three months, each man completes a sexual behavior questionnaire and has gonorrheal testing at all sites. Cohorts were simulated under many of conditions, such as measuring different variables, different levels of under and over reporting of sex acts, and different patterns of sexual behavior in the population. The simulated data were analyzed in a Bayesian framework where prior knowledge of the joint prevalence of single-site and multi-site gonorrheal infection was integrated into the analysis using the Stan programming language. Outcomes included coverage of true transmission probabilities, bias, and uncertainty in route-specific transmission probabilities.

Results Under ideal conditions, we have shown that route-specific gonorrheal transmission probabilities can be estimated from study designs similar to ongoing CDC projects. However, we also found that failure to measure heterogeneity in sexual behavior, a high preponderance of very high-risk behavior, and systemic under-reporting of certain sex acts (but not random recall bias) significantly limit the power of cohort studies regardless of the design.

Conclusion Model-based study design provides a general method for the design, analysis, and evaluation of studies for complex parameters that cannot be estimated directly from available data.

Disclosure No significant relationships.