Accurately estimating the size and prevalence of hidden populations is a persistent challenge in social science research. This study systematically evaluates the relative bias of various sampling and estimation methods by leveraging a Bayesian framework that integrates data from multiple studies across different contexts. We formalize population features, sampling strategies, estimands, and estimation techniques. The Bayesian model allows for the pooling of data from eight studies on human trafficking prevalence, supported by the Prevalence Reduction Innovation Forum (PRIF), to assess the systematic bias of different estimation approaches. We compare these methods against benchmark estimates, such as those derived from the Horvitz-Thompson estimator on proportional samples. Our analysis, implemented using the hiddenmeta R package, provides insights into the accuracy and generalizability of existing estimation techniques for hidden populations. This research contributes to advancing methodological rigor in prevalence estimation, offering a comparative assessment of estimation biases across diverse real-world applications.