Diagnostic test accuracy studies assess a diagnostic test’s performance against a reference standard. In this review, we explore and compare statistical methods used in meta-analyses of diagnostic test accuracy studies. Specifically, we evaluate two frequentist methods – split component synthesis (SCS) and bivariate model (BM) – alongside two Bayesian approaches: Bayesian hierarchical summary receiver operating characteristic (BHSROC) and Bayesian bivariate model (BBM). We also include their latent class variants (LC-BHSROC and LC-BBM). Using a meta-analysis of various multiplex nucleic acid amplification tests (NAATs/PCRs) against Campylobacter spp. as a case study we illustrate the practical applications of these methods. The reference standard was culture, and due to differences in cut-off values and primers among the NAAT/PCR brands, substantial heterogeneity was anticipated. Our findings reveal that the BM and BBM methods tend to estimate higher sensitivities than the other approaches, even when the number of studies is substantial, and heterogeneity is moderate – as observed in this case study. In such scenario, the SCS method or the BHSROC model may offer more robust and reliable outcomes. While our review is based on a real-life meta-analysis rather than simulations, it offers practical insights into the strengths and limitations of these statistical approaches for diagnostic test accuracy studies.