A systematic review on compliance with QUADAS-2 application guideline

ID: 

132

Session: 

Poster session 5

Date: 

Thursday 27 October 2016 - 10:30 to 11:00

Location: 

All authors in correct order:

Zheng Q1, De Souza NN1, Shi L1, Chan ESY1
1 Cochrane Singapore, Singapore
Presenting author and contact person

Presenting author:

Qishi ZHENG

Contact person:

Abstract text
Background: QUADAS-2 is recommended for use in diagnostic accuracy (DTA) systematic reviews to evaluate the risk of bias and applicability of the primary studies. To ensure appropriate application, a four-phased approach has been recommended. However, in practice reviewers may use the standard signalling questions to assess the study quality directly without adopting the four-phase approach. This may bias the conclusion of quality assessment.

Objectives: To investigate the compliance with the QUADAS-2 application guideline on the four-phase approach in DTA systematic reviews.

Methods: We performed a comprehensive literature search in MEDLINE, Embase and Cochrane Library to identify eligible DTA systematic reviews using QUADAS-2. Reviews including less than 10 studies and protocols were excluded. Data on four key domains were extracted together with the information of QUADAS-2 implementation. The compliance rates were quantified and compared between Cochrane and non-Cochrane systematic reviews.

Results: We identified 94 eligible studies, 18 of which were Cochrane DTA systematic reviews. Around 61.7% of the index tests were subjective assessments, of which 58.6% were diagnostic imaging tests. In contrast, most of the objective index tests were laboratory tests (75.0%). As suggested in the QUADAS-2 guideline, appropriate review-specific tailoring is essential for objective index tests as some of the signalling questions are not applicable. However, only 38.9% of the studies modified the appraisal questions accordingly. Furthermore, 37.2% of the studies failed to report a clear rating guideline. Generally Cochrane DTA reviews performed better compared to non-Cochrane reviews, in terms of clear description of rating guideline (prevalence ratio (PR) 1.57, 95% CI 1.22 to 2.03) and appropriate tailoring (PR 1.85, 95% CI 1.22 to 2.03).

Conclusions: Although QUADAS-2 has been published for over five years, many reviewers still use its default version and do not adhere to the recommended four-phase approach. There is need for a greater awareness of appropriate implementation of QUADAS-2 amongst systematic reviewers, journal editors and peer reviewers.