Systematic Review and Meta-Analysis of Studies Evaluating Diagnostic Test Accuracy: A Practical Review for Clinical Researchers-Part I. General Guidance and Tips

Abstract

In the field of diagnostic test accuracy (DTA), the use of systematic review and meta-analyses is steadily increasing. By means of objective evaluation of all available primary studies, these two processes generate an evidence-based systematic summary regarding a specific research topic. The methodology for systematic review and meta-analysis in DTA studies differs from that in therapeutic/interventional studies, and its content is still evolving. Here we review the overall process from a practical standpoint, which may serve as a reference for those who implement these methods.

Keywords: Diagnostic test accuracy; Meta-analysis; Systematic review.

Fig. 1. Diagram of study process and frame of research questions.

DOR = diagnostic odds ratio, FN = false negative, FP = false positive, HCC = hepatocellular carcinoma, HSROC = hierarchical summary receiver operating characteristic, SROC = summary receiver operating characteristic, TN = true negative, TP = true positive

Fig. 2. Process to select literature according to Preferred Reporting Items of Systematic Reviews and Meta-Analyses guideline.

Fig. 3. Templates for presenting results of QUADAS-2 assessment for assessing quality of studies.

We can present results of QUADAS-2 assessments in tabular form (A) or in graphics (B). QUADAS = Quality Assessment of Diagnostic Accuracy Studies

Fig. 4. Description of reconstructing diagnostic 2-by-2 table.

FN = false negative, FP = false positive, TN = true negative, TP = true positive

Fig. 5. Graphs used in meta-analysis of diagnostic test accuracy studies.

A. Coupled forest plots. B. Summary receiver operating curve (SROC) plot. Open circle (o) represents false positive rate (x-coordinate) and sensitivity (y-coordinate) of individual studies. Size of bubbles reflects precision of estimate. C. SROC curve. CI = confidence interval

Fig. 6. Funnel plot to assess publication bias.

A. Funnel plot with Egger's regression line. Each dot represents primary study. X-axis shows study result (i.e., diagnostic odds ratio [DOR]) and y-axis represents study size (i.e., standard error of study result). Empty region, to be filled with results of relatively small studies with negative results or small effect size, makes plot asymmetrical. Asymmetry of plot would indicate that such studies may not ever have been published, thus raising possibility of publication bias being presented as review result. B. Deeks funnel plot. In Deeks funnel plot, x-axis is natural logarithm of DOR and y-axis is 1/√ effective sample size (ESS). According to Deeks et al. (35), it is preferred method for meta-analysis of diagnostic test accuracy studies owing to its high statistical power.