Diagnostic systematic review is a vital step in the evaluation of diagnostic technologies. In many applications, it involves pooling pairs of sensitivity and specificity of a dichotomized diagnostic test from multiple studies. We propose a composite likelihood (CL) method for bivariate meta-analysis in diagnostic systematic reviews. This method provides an alternative way to make inference on diagnostic measures such as sensitivity, specificity, likelihood ratios, and diagnostic odds ratio. Its main advantages over the standard likelihood method are the avoidance of the nonconvergence problem, which is nontrivial when the number of studies is relatively small, the computational simplicity, and some robustness to model misspecifications. Simulation studies show that the CL method maintains high relative efficiency compared to that of the standard likelihood method. We illustrate our method in a diagnostic review of the performance of contemporary diagnostic imaging technologies for detecting metastases in patients with melanoma.
SackettDLHaynesRB. The architecture of diagnostic research. BMJ2002; 324: 539–541.
2.
KnottnerusJAMurisJW. Assessment of the accuracy of diagnostic tests: the cross-sectional study. J Clin Epidemiol2003; 56: 1118–1128.
3.
HonestHKhanK. Reporting of measures of accuracy in systematic reviews of diagnostic literature. BMC Health Serv Res2002; 2: 4–4.
4.
ReitsmaJBGlasASRutjesAWS, et al.Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol2005; 58: 982–990.
5.
Ma X, Nie L, Cole SR, et al. Statistical methods for meta-analysis of diagnostic tests: an overview and tutorial. Stat Methods Med Res 2016; 25: 1596–1619.
6.
MosesLEShapiroDLittenbergB. Combining independent studies of a diagnostic test into a summary roc curve: data-analytic approaches and some additional considerations. Stat Med1993; 12: 1293–1316.
7.
IrwigLMacaskillPGlasziouP, et al.Meta-analytic methods for diagnostic test accuracy. J Clin Epidemiol1995; 48: 119–130.
8.
RutterCMGatsonisCA. A hierarchical regression approach to meta-analysis of diagnostic test accuracy evaluations. Stat Med2001; 20: 2865–2884.
9.
LittenbergBMosesLE. Estimating diagnostic accuracy from multiple conflicting reports. Med Decis Making1993; 13: 313–313.
10.
WalterS. Properties of the summary receiver operating characteristic (SROC) curve for diagnostic test data. Stat Med2002; 21: 1237–1256.
11.
ArendsLHamzaTVan HouwelingenJ, et al.Bivariate random effects meta-analysis of ROC curves. Med Decis Making2008; 28: 621–621.
12.
RutterCGatsonisC. Regression methods for meta-analysis of diagnostic test data. Acad Radiol1995; 2: S48–S48.
13.
Van HouwelingenHCZwindermanKHStijnenT. A bivariate approach to meta-analysis. Stat Med1993; 12: 2273–2284.
14.
Van HouwelingenHCArendsLRStijnenT. Advanced methods in meta-analysis: multivariate approach and meta-regression. Stat Med2002; 21: 589–624.
15.
ChuHColeSR. Bivariate meta-analysis of sensitivity and specificity with sparse data: a generalized linear mixed model approach. J Clin Epidemiol2006; 59: 1331–1331.
16.
HarbordRMDeeksJJEggerM, et al.A unification of models for meta-analysis of diagnostic accuracy studies. Biostatistics2007; 8: 239–251.
17.
ChuHGuoHZhouY. Bivariate random effects meta-analysis of diagnostic studies using generalized linear mixed models. Med Decis Making2010; 30: 499–508.
18.
HamzaTHvan HouwelingenHCStijnenT. The binomial distribution of meta-analysis was preferred to model within-study variability. J Clin Epidemiol2008; 61: 41–51.
19.
HamzaTHReitsmaJBStijnenT. Meta-analysis of diagnostic studies: a comparison of random intercept, normal-normal, and binomial-normal bivariate summary ROC approaches. Med Decis Making2008; 28: 639–649.
20.
FournierDASkaugHJAnchetaJ, et al.AD Model Builder: using automatic differentiation for statistical inference of highly parameterized complex nonlinear models. Optim Methods Softw2012; 27: 233–249.
VarinCReidNFirthD. An overview of composite likelihood methods. Stat Sin2011; 21: 5–42.
23.
LiangKYZegerSL. Longitudinal data analysis using generalized linear models. Biometrika1986; 73: 13–22.
24.
LinD. Cox regression analysis of multivariate failure time data: the marginal approach. Stat Med1994; 13: 2233–2247.
25.
SimelDLBossuytPMM. Differences between univariate and bivariate models for summarizing diagnostic accuracy may not be large. J Clin Epidemiol2009; 62: 1292–1300.
26.
GlasASLijmerJGPrinsMH, et al.The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol2003; 56: 1129–1135.
27.
PinheiroJCBatesDM. Approximations to the log-likelihood function in the nonlinear mixed-effects model. J Comput Graph Stat1995; 4: 12–35.
28.
BesagJ. Spatial interaction and the statistical analysis of lattice systems. J R Stat Soc Ser B (Methodol)1974; 36: 192–236.
29.
CoxDReidN. A note on pseudolikelihood constructed from marginal densities. Biometrika2004; 91: 729–737.
30.
KentJT. Robust properties of likelihood ratio tests. Biometrika1982; 69: 19–27.
31.
WhiteH. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica J Econometric Soc1980; 48: 817–838.
32.
BroströmGHolmbergH. Generalized linear models with clustered data: fixed and random effects models. Comput Stat Data Anal2011; 55: 3123–3134.
33.
JerantAFJohnsonJTSheridanC, et al.Early detection and treatment of skin cancer. Am Fam Physician2000; 62: 357–386.
34.
WurmEMSoyerHP. Scanning for melanoma. Sun2010; 5: 6–6.
35.
XingYBronsteinYRossMI, et al.Contemporary diagnostic imaging modalities for the staging and surveillance of melanoma patients: a meta-analysis. J Natl Cancer Inst2011; 103: 129–142.
36.
VarinCVidoniP. A note on composite likelihood inference and model selection. Biometrika2005; 92: 519–528.
37.
GaoXSongPXK. Composite likelihood Bayesian information criteria for model selection in high-dimensional data. J Am Stat Assoc2010; 105: 1531–1540.
38.
ShapiroSSWilkMB. An analysis of variance test for normality (complete samples). Biometrika1965; 52: 591–611.
39.
BartlettMS. Properties of sufficiency and statistical tests. Proc R Soc Lond Ser A Math Phys Sci1937; 160: 268–282.
40.
DouglasJ. Confidence regions for parameter pairs. Am Statistician1993; 47: 43–45.
41.
ChuHNieLChenY, et al.Bivariate random effects models for meta-analysis of comparative studies with binary outcomes: methods for the absolute risk difference and relative risk. Stat Methods Med Res2012; 21: 621–633.
42.
Chen Y, Chu H, Luo S, et al. Bayesian analysis on meta-analysis of case-control studies accounting for within-study correlation. Stat Methods Med Res. 2015; 24: 836–855.
43.
Chen Y, Luo S, Chu H, et al. An empirical Bayes method for multivariate meta-analysis with application in clinical trials. Commun Stat Theory Methods 2014; 43: 3536–3551.
44.
ChenYLuoSChuH, et al.Bayesian inference on risk differences: an application to multivariate meta-analysis of adverse events in clinical trials. Stat Biopharmaceut Res2013; 23: 1042–1053.
