Skip to main content

Table 3 Median estimates and 95% CI (in brackets) for parameter estimates using different methods

From: A trivariate meta-analysis of diagnostic studies accounting for prevalence and non-evaluable subjects: re-evaluation of the meta-analysis of coronary CT angiography studies

Method

 

Sensitivity

 

Specificity

 

Prevalence

 

PPV

TGLMM

 

98.0 (96.7,99.3)

 

87.5 (82.7,92.3)

 

47.8 (37.9,57.7)

 

87.8 (83.3,92.3)

Model 1

 

98.0 (96.7,99.3)

 

87.4 (82.5, 92.3)

 

49.3 (38.9,59.7)

 

88.4 (84,92.7)

Model 2

 

98.1 (96.9,99.3)

 

75.9 (69.3,82.5)

 

47.8 (37.9,57.8)

 

78.9 (71.9,85.9)

Model 3

 

91.7 (88.1,95.4)

 

89 (85.4,92.7)

 

47.8 (37.9,57.7)

 

88.4 (84.1,92.7)

Intent-to-diagnose

 

91.7 (88.1,95.3)

 

76.2 (69.7,82.6)

 

47.9 (37.9,57.9)

 

78 (70.2,85.7)

Method

 

NPV

 

LR+

 

LR −

 

AUC

TGLMM

 

97.9 (96.4,99.5)

 

7.8 (4.8,10.9)

 

0.02 (0.01,0.04)

 

0.99 (0.96,1)

Model 1

 

97.8 (96.1,99.4)

 

7.8 (4.8,10.9)

 

0.02 (0.01,0.04)

 

0.99 (0.96,1)

Model 2

 

97.8 (96.2, 99.4)

 

4.1 (2.9,5.2)

 

0.02 (0.01,0.04)

 

0.98 (0.97,1)

Model 3

 

92.1 (88.4,95.8)

 

8.4 (5.5,11.3)

 

0.09 (0.05,0.14)

 

0.96 (0.93,0.99)

Intent-to-diagnose

 

90.9 (86.4,95.5)

 

3.8 (2.7,5.0)

 

0.11 (0.06,0.16)

 

0.93 (0.89,0.96)

  1. “TGLMM” stands for the extended TGLMM. Model 1 excludes non-evaluable subjects, Model 2 takes non-evaluable subjects as index test positives, Model 3 takes non-evaluable subjects as index test negatives and the intent-to-diagnose approach takes non-evaluable subjects as fasle positives and false negatives. Positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), negative likelihood ratio (LR −) and area under the curve (AUC) are summerized.