Reporting and Assessing the Quality of Diagnostic Accuracy Studies for Cervical Cancer Screening and Management

Abstract

Objective: We adapted the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool for studies of cervical cancer screening and management and used the adapted tool to evaluate the quality of studies included in a systematic review supporting the 2019 Risk-Based Management Consensus Guidelines.

Methods: We evaluated the quality of all studies included in our systematic review for postcolposcopy (n = 5) and posttreatment (n = 23) surveillance using QUADAS-2 criteria. Subsequently, we adapted signaling questions to indications of cervical cancer screening and management. An iterative process was carried out to evaluate interrater agreement between 2 study authors (M.A.C. and N.W.). Discrepant ratings were discussed, and criteria were adapted accordingly. We also evaluated the influence of study quality on risk estimates and between study variation using stratified subgroup meta-analyses.

Results: Twelve signaling questions for bias assessment that were adapted to or newly developed for cervical cancer screening and management are described here. Interrater agreement on bias assessment increased from 70% to 83% during the adaptation process. Detailed assessment of bias and applicability showed that all studies on postcolposcopy management and 90% of studies on posttreatment management had high risk of bias in at least 1 domain. Most commonly, high risk of bias was observed for the patient selection domain, indicating the heterogeneity of study designs and clinical practice in reported studies.

Conclusions: The adapted QUADAS-2 will have broad application for researchers, evidence evaluators, and journals who are interested in designing, conducting, evaluating, and publishing studies for cervical cancer screening and management.

FIGURE 1

Components of quality assessment for diagnostic accuracy studies. The assessment domains include patient selection, index test, reference standard, and flow and timing. The patient selection domain addresses potential sources of bias in the selection of patients included in the study (i.e., the study population). The study population is sampled from a larger source population (the population that is eligible based on study inclusion criteria). The target population is the external population to which results are intended to inform. The index test domain addresses potential sources of bias for the assay/test under investigation in the study. A simple schematic shows possible outcomes of positive or negative index test results. The reference standard domain addresses potential sources of bias in the ascertainment and/or measurement of study outcomes. A simple schematic shows possible outcomes of positive or negative results for the reference standard. Together, the results from the index test and reference standard can be included in a 2 × 2 table to evaluate the diagnostic accuracy of an index test. Flow and timing addresses potential sources of bias in the flow and timing of procedures carried out in the study.