Diagnostic Test Meta-analysis (SROC)
mada/metandi.Pool multiple diagnostic-accuracy studies: pooled sensitivity and specificity (random effects), diagnostic odds ratio DOR, and an SROC curve showing the threshold effect. Just enter each study's 2×2 counts. Computed locally in your browser; data are not uploaded.
① 2×2 counts per study
One study per row, four columns in order: TP (true positive) FP (false positive) FN (false negative) TN (true negative). At least 3 studies.
How to use & methodology
Why can't diagnostic meta-analysis just pool AUC?
In diagnostic studies, sensitivity and specificity are negatively correlated because of differing thresholds (the threshold effect), so pooling them separately is biased. They should be modelled together and shown with an SROC curve, or with a bivariate random-effects model. This tool gives a random-effects pooled point plus an SROC curve.
How do I read the SROC curve?
It traces accuracy across thresholds in the '1−specificity vs sensitivity' space — the closer to the top-left, the better. Study points scatter around the curve; the closer the curve and pooled operating point are to the top-left, the better the overall diagnostic performance.
What does a high I² mean?
I² reflects between-study heterogeneity. It is common in diagnostic meta-analysis (differing populations, thresholds, reference standards). When I² is high, interpret the pooled values cautiously and consider subgroups or meta-regression (which need more specialised models).
Does it differ much from the gold-standard model?
Moses-Littenberg is a classic closed-form method — intuitive and easy to compute — but less robust than the bivariate/HSROC model when studies are few or the threshold effect is strong. For key conclusions, cross-check with R's mada/metandi; this tool suits a quick overview.