Skip to main content

Table 2 Simulation results under MAR assumption

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

Model

TGLMM

Model 1

Model2

Model3

Intent-to-diagnose

Estimate

Bias%

meanSE

CP

Bias%

meanSE

CP

Bias%

meanSE

CP

Bias%

meanSE

CP

Bias%

meanSE

CP

ω m1 m0 =0.1

Se

−0.3

0.041

0.94

−0.3

0.041

0.94

4.6

0.036

0.81

−12.6

0.037

0.33

−12.6

0.036

0.33

Sp

−0.1

0.017

0.93

−0.1

0.017

0.93

−11.9

0.018

0

1.1

0.015

0.84

−11.9

0.017

0

Prev

0.8

0.034

0.93

0.8

0.034

0.93

0.8

0.034

0.93

0.8

0.034

0.93

0.8

0.034

0.93

PPV

−0.1

0.046

0.94

−0.3

0.046

0.94

−22.6

0.047

0.08

−0.9

0.046

0.94

−29

0.049

0.01

NPV

−0.1

0.018

0.93

−0.1

0.018

0.93

−0.2

0.018

0.93

−2.9

0.020

0.81

−4.6

0.022

0.59

LR+

1.6

1.188

0.92

1.6

1.189

0.93

−49.2

0.307

0

−0.5

1.160

0.92

−57.6

0.271

0

LR −

0.9

0.044

0.94

0.9

0.044

0.94

1.5

0.044

0.94

27.9

0.039

0.33

46.8

0.045

0.04

ω m1 =0.1,ω m0 =0.2

Se

−0.1

0.041

0.94

−0.1

0.041

0.94

4.7

0.036

0.80

−12.3

0.036

0.34

−12.3 0.036

0.34

 

Sp

−0.1

0.017

0.94

−0.1

0.017

0.94

−22.3

0.017

0

2.2

0.014

0.62

−22.3

0.017

0

Prev

0.4

0.034

0.93

9.6

0.036

0.90

0.4

0.034

0.93

0.4

0.034

0.93

0.4

0.034

0.93

PPV

−0.3

0.046

0.93

3.1

0.044

0.88

−36

0.047

0

2.7

0.044

0.89

−42.1

0.047

0

NPV

−0.1

0.018

0.94

−1.3

0.020

0.93

−1.4

0.020

0.92

−2.7

0.020

0.83

−6.3

0.024

0.36

LR+

1.4

1.195

0.94

1.4

1.194

0.94

−65.1

0.159

0

12.3

1.312

0.95

−70.8

0.147

0

LR −

0.6

0.044

0.93

0.6

0.044

0.93

14.7

0.050

0.85

26.1

0.038

0.39

66.1

0.051

0

ω m1 =0.2,ω m0 =0.1

Se

-0.1

0.023

0.93

-0.1

0.023

0.93

8.7

0.018

0.12

-21

0.020

0

-21

0.019

0

Sp

0

0.009

0.93

0

0.009

0.93

-10.6

0.009

0

1.1

0.008

0.74

-10.6

0.009

0

Prev

0

0.018

0.93

-8.4

0.017

0.72

0

0.017

0.91

0

0.017

0.91

0

0.0168

0.89

PPV

-0.1

0.025

0.93

-3.7

0.027

0.83

-19.1

0.025

0

-4

0.026

0.8

-30.6

0.025

0

NPV

0

0.010

0.92

1.1

0.009

0.76

1.1

0.009

0.74

-4.6

0.011

0.05

-6.2

0.012

0

LR+

0.3

0.655

0.93

0.3

0.653

0.93

-44.1

0.196

0

-11.7

0.570

0.62

-59.3

0.154

0

LR −

0.3

0.025

0.93

0.3

0.025

0.93

-10.8

0.022

0.62

47.4

0.021

0

66.7

0.024

0

  1. Three scenarios are studied: equal or unequal missing probabilities for the diseased and non-diseased groups. Bias in percentage(Bias%), mean standard error (meanSE) and 95% confidence interval coverage probability (CP) are summarized for estimates of sensitivity (Se), specificity (Sp), prevalence (Prev), positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+) and negative likelihood ratio (LR −) from different models. “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 false positives and false negatives.