Meta-DiSc

Meta-DiSc performs meta-analyses of diagnostic test accuracy (DTA) studies to estimate pooled measures of diagnostic performance and quantify heterogeneity across studies.

Key Features:

• Bivariate random effects model: Implements a bivariate random effects model to jointly synthesize sensitivity and specificity.
• Logit variance estimates: Calculates logit variance estimates for sensitivity and specificity.
• Bivariate I-squared: Computes the bivariate I-squared statistic to quantify heterogeneity in diagnostic accuracy.
• 95% prediction ellipse: Estimates the area of the 95% prediction ellipse on the ROC plane.
• Median odds ratios: Calculates median odds ratios for sensitivity and for specificity.
• Univariate random effects models: Provides univariate random effects models as an alternative when studies are limited or bivariate models do not converge.
• Subgroup analyses and meta-regression: Performs subgroup analyses and meta-regression to explore sources of heterogeneity.
• Graphical outputs: Generates forest plots and ROC plane plots for graphical description of study results.
• Pooling and summary measures: Pools sensitivity and specificity and estimates positive and negative likelihood ratios and the diagnostic odds ratio.
• sROC analysis: Performs summary ROC (sROC) curve analysis.
• Validation: Validates computational algorithms by comparing results with STATA/SAS and MetaDTA.

Scientific Applications:

• DTA meta-analysis: Synthesis of diagnostic accuracy data (sensitivity, specificity) across studies.
• Heterogeneity assessment: Quantification and investigation of heterogeneity in sensitivity and specificity using bivariate I-squared and meta-regression.
• Impact estimation: Estimation of diagnostic test impact in hypothetical populations given specified prevalence values.
• Small-study analysis: Alternative analysis for limited datasets or non-convergent bivariate models using univariate random effects.

Methodology:

Implements bivariate and univariate random effects models; computes logit variance estimates for sensitivity and specificity, bivariate I-squared, area of the 95% prediction ellipse, and median odds ratios; performs subgroup analyses and meta-regression; produces forest and ROC plane plots; pools sensitivity and specificity and estimates likelihood ratios, diagnostic odds ratio, and sROC curves; validated by comparison with STATA/SAS and MetaDTA.