Meta-Analysis of Diagnostic Test Accuracy

Diagnostic test accuracy meta-analysis aims to systematically combine evidence from multiple diagnostic studies to evaluate how well a test identifies individuals with and without a target condition. This session introduces the principles of systematic reviews and meta-analyses of diagnostic accuracy, beginning with the \(2 \times 2\) contingency table, which classifies test results into true positives, false positives, true negatives, and false negatives. Students will learn to interpret key diagnostic performance measures, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and positive and negative likelihood ratios. Particular attention will be given to the distinction between sensitivity and specificity, which describe test performance among individuals with and without the target condition, and predictive values, which are influenced by disease prevalence. Students will also explore how prevalence and patient spectrum can affect the interpretation and applicability of diagnostic accuracy estimates across different clinical settings.

The second component focuses on methodological challenges in synthesizing diagnostic accuracy results. Students will examine the assessment of risk of bias and applicability concerns using the QUADAS-2 tool, as well as the implications of variation in diagnostic thresholds across studies. Different thresholds can produce different sensitivity and specificity combinations, creating a trade-off to consider when comparing diagnostic tests. To address this complexity, the session introduces the summary receiver operating characteristic (SROC) curve, which summarizes the relationship between sensitivity and specificity across studies. Students will also learn the basic principles of bivariate models and hierarchical summary receiver operating characteristic (HSROC) models, which account for the relationship between sensitivity and specificity and variation across studies. These methods provide a statistical framework for synthesizing diagnostic accuracy while recognizing that study results may differ because of threshold choices, populations, and methodological characteristics.

The final component emphasizes interpreting pooled diagnostic accuracy estimates and their uncertainty. Students will learn to assess summary sensitivity and specificity alongside confidence regions or intervals, the SROC curve, and the methodological quality of the included studies. They will also recognize that high pooled sensitivity or specificity does not necessarily mean a test performs equally well across populations or settings. The session concludes by distinguishing diagnostic accuracy meta-analysis from meta-analysis of intervention effects: diagnostic accuracy reviews evaluate a test’s ability to classify or identify a condition, whereas intervention meta-analyses typically estimate the effects of treatments or exposures on outcomes. By the end of this session, students should be able to explain the principal measures of diagnostic accuracy, describe the roles of QUADAS-2, SROC, bivariate, and HSROC methods, and interpret pooled findings cautiously in light of uncertainty, prevalence, diagnostic thresholds, and study quality.