Diagnostic Test Accuracy Meta-Analysis: A Practical Guide to Hierarchical Models

Background: Accurate evaluation of diagnostic tests is essential to guide clinical decision-making, particularly in surgical practice. Systematic reviews and meta-analyses of diagnostic test accuracy (DTA) are key for evidence synthesis; however, traditional approaches, including univariate pooling or simplified summary ROC (SROC) models such as the Moses–Littenberg method, often yield biased and clinically misleading estimates.Methods: An extensive literature review was conducted to synthesize current evidence on hierarchical random-effects models for DTA meta-analysis. Based on this framework, a simulated dataset was generated, and a comprehensive meta-analysis was performed with a didactic and explanatory focus. The analysis illustrates key methodological concepts, interpretation of model outputs, and the use of complementary tools, including likelihood ratios, scattergrams, meta-regression, publication bias assessment, and outlier detection.Results: The traditional meta-analysis performed with MetaDisc, applying the DerSimonian–Laird and Moses–Littenberg methods, produced separated sensitivity and specificity pooled estimates with artificially narrow confidence intervals and a symmetric, theoretical SROC curve extrapolated beyond the observed data range, ignoring threshold variability and underestimating between-study heterogeneity. In contrast, the hierarchical random-effects model provided more realistic and clinically interpretable estimates. Joint modeling of sensitivity and specificity revealed substantial between-study variability, a strong negative correlation consistent with a threshold effect, and wider, asymmetric confidence intervals that accurately reflected uncertainty. Influence diagnostics identified outliers and highly influential studies that distorted the traditional model results.Conclusions: Promoting the correct application and interpretation of hierarchical models in DTA meta-analyses is essential to ensure high-quality, reliable, and scientifically robust evidence.