Comparative reviews of diagnostic test accuracy in imaging research: evaluation of current practices.

Article type
Authors
Dehmoobad Sharifabadi A1, Leeflang M2, Treanor L1, Kraajipoel N3, Salameh JP4, Alabousi M5, Asraoui N1, Choo-Foo J1, Takwoingi Y6, Deeks J6, McInnes M4
1University of Ottawa
2University Medical Centre (UMC), Amsterdam
3Academic Medical Center
4Ottawa Hospital Research Institute
5McMaster University
6University Hospitals Birmingham NHS Foundation
Abstract
Background: Research comparing the diagnostic accuracy of two or more imaging tests is essential to determine optimal diagnostic pathways and downstream treatment decisions. High-quality comparative methods would include primary studies that apply the index tests to every study participant or randomly allocate participants to receive one of the index tests. If comparative imaging reviews largely rely on non-comparative primary studies, the conclusions regarding the comparative accuracy of index tests may be biased.

Objectives: The purpose of this methodological review was to determine the extent to which comparative imaging systematic reviews of diagnostic test accuracy (DTA) use primary studies with comparative or non-comparative designs.

Methods:
MEDLINE was used to identify DTA systematic reviews published in imaging journals between January 2000 and May 2018. Systematic reviews comparing at least two index tests (one of which was imaging-based); review characteristics were extracted. Study design and other characteristics of primary studies included in the systematic reviews were evaluated.

Results: One hundred three comparative imaging reviews were included; 11 (11%) included only comparative studies, 12 (11%) included only non-comparative primary studies, and 80 (78%) included both comparative and non-comparative primary studies. For reviews containing both comparative and non-comparative primary studies, the median proportion of non-comparative primary studies was 81% (IQR 57-90%). Of 92 reviews that included non-comparative primary studies, 86% did not recognize this as a limitation. Furthermore, among 4182 primary studies, 3438 (82%) were non-comparative and 744 (18%) were comparative in design.

Conclusions: Most primary studies included in comparative imaging reviews are non-comparative in design and awareness of the risk of bias associated with this is low. This may lead to incorrect conclusions about the relative accuracy of diagnostic tests and be counter-productive for informing guidelines and funding decisions about imaging tests.

Patient or healthcare consumer involvement: Medicine today relies on cost effectiveness analyses, by comparing DTA imaging measures as the cornerstone of medical diagnoses. The demand for accurate comparative data combined with minimal awareness of valid comparative study designs may lead to counter-productive research and inadequately supported clinical decisions. Using comparative accuracy imaging reviews with a high risk of bias to inform guidelines and funding decisions may have detrimental impacts on patients.