Table 8Point estimates and individual 95% confidence intervals with alternative meta-analysis methods

Summary
TPR (%)
Short humerus
Summary
TPR (%)
Short femur
Summary
JTPR (%)
Summary
FPR (%)
Short humerus
Summary
FPR (%)
Short femur
Summary
JFPR (%)
Analyses in the 11 paired studies
Separate meta-analyses of the two tests
 a. Bivariate, normal likelihood36.9 (29.3, 45.1)36.0 (26.0, 47.3)NA5.0 (3.5, 7.1)7.6 (5.5, 10.3)NA
 b. Bivariate, binomial likelihood37.9 (27.7, 50.3)35.4 (23.1, 49.5)NA4.8 (3.0, 7.4)7.4 (5.0, 10.7)NA
Joint meta-analyses of the two tests
 c. Normal likelihood, using within-study correlation37.3 (29.9, 45.5)37.5 (27.9, 48.1)30.9 (23.0, 40.2)4.9 (3.4, 7.0)7.6 (5.6, 10.2)3.0 (2.2, 4.1)
 d. Normal likelihood, ignoring within-study correlation37.5 (29.4, 46.5)36.4 (26.1, 48.0)27.5 (19.5, 37.2)4.9 (3.4, 7.0)7.3 (5.3, 10.0)2.8 (2.0, 4.0)
 e. Multinomial likelihood (uses within-study correlations)35.3 (26.9, 41.8)35.0 (22.4, 46.2)26.1 (16.6, 34.0)4.9 (2.8, 7.5)7.3 (4.6, 10.5)2.7 (1.6, 4.2)
 f. Binomial likelihood (ignores within-study correlation)34.6 (20.3, 44.2)35.9 (20.5, 50.4)26.8 (11.3, 39.2)4.8 (2.9, 7.7)7.3 (4.6, 11.5)2.8 (1.7, 4.4)
Analyses in all 30 studies
Separate meta-analyses of the two tests
 a. Bivariate, normal likelihood36.9 (29.3, 45.1)35.8 (29.1, 43.1)NA5.0 (3.5, 7.1)7.6 (6.1, 9.4)NA
 b. Bivariate, binomial likelihood38.0 (27.8, 50.3)35.2 (27.8, 43.3)NA4.8 (3.0, 7.4)7.4 (5.8, 9.3)NA
Joint meta-analyses of the two tests
 c. Normal likelihood, using within-study correlation34.2 (27.2, 42.0)37.1 (30.6, 44.2)29.8 (23.1, 37.5)4.9 (3.5, 6.7)7.6 (6.2, 9.4)3.0 (2.3, 3.8)
 d. Normal likelihood, ignoring within-study correlation35.9 (28.9, 43.6)35.8 (29.0, 43.3)26.7 (20.4, 34.1)5.0 (3.6, 7.0)7.5 (6.1, 9.3)2.9 (2.2, 3.9)
 e. Multinomial likelihood (uses within-study correlations)34.7 (27.9, 42.5)34.8 (28.7, 41.9)25.2 (18.9, 32.4)4.9 (3.1, 7.4)7.4 (6.0, 9.2)2.8 (1.9, 3.9)
 f. Binomial likelihood (ignores within-study correlation)35.9 (27.7, 45.1)34.9 (28.0, 42.7)27.5 (19.2, 37.2)4.9 (3.1, 7.4)7.4 (5.8, 9.3)2.8 (1.8, 4.1)

FPR = false-positive rate (1−specificity); REML = restricted maximum likelihood; TPR = true-positive rate (sensitivity); JFPR/JTPR = joint-[false|true]-positive rate. The joint meta-analyses of the two tests use the unstructured variant of the between-study covariance matrix as per equation (24).

From: Analysis of the Example

Cover of Evaluating Practices and Developing Tools for Comparative Effectiveness Reviews of Diagnostic Test Accuracy
Evaluating Practices and Developing Tools for Comparative Effectiveness Reviews of Diagnostic Test Accuracy: Methods for the Joint Meta-Analysis of Multiple Tests [Internet].
Trikalinos TA, Hoaglin DC, Small KM, et al.

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.