Low quality corneal disease systematic reviews limit the amount of usable evidence for informing practice guidelines
Background: High-quality systematic reviews and meta-analyses (SRMAs) underpin trustworthy clinical practice guidelines. Cochrane Eyes and Vision US Project (CEV@US) maintains a database of SRMAs in eyes and vision to support the development and update of practice guidelines in ophthalmology and optometry in the United States and worldwide. Previous work by CEV@US has shown that a majority of SRMAs in many eye conditions are unreliable.
Objectives: To summarize the reliability of SRMAs in ocular surface diseases published between 2017 and 2022.
Methods: SRMAs of cornea-related interventions (2017-2022) excluding dry eye were searched through the CEV@US database. The reliability of SRMAs was assessed by one reviewer and verified by another based on a set of minimal criteria (Table 1). A single reviewer assessed adherence to SRMA reporting guidelines and registration.
Results: We assessed 221 SRMAs of cornea-related interventions, of which 87 (39%) were judged as reliable. Of the remaining 134 unreliable SRMAs, 27 (20%) failed to describe eligibility criteria, 83 (62%) used only one database or unrefined search strategies, 64 (48%) did not assess risk of bias of included studies, 50 (37%) did not conduct meta-analysis appropriately, and conclusions from 34 (25%) were discordant with their findings. Additionally, 26 (19%) SRMAs had other concerns: predominantly financial conflicts. A total of 96 (43%) of the 221 SRMAs stated that they followed relevant reporting guidelines, but 53/96 (55%) were judged as unreliable. Of 200 non-Cochrane SRMAs, 34 (17%) were prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO) or International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY), whereas all 21 Cochrane reviews were prospectively registered and published (Figure 1).
Conclusions: Nearly 60% of the published SRMAs in ocular surface disease over the past 5 years were unreliable, primarily because they failed to conduct comprehensive literature searches or assess the risk of bias of included studies. The methodology of future SRMAs should be well-planned, preregistered and clearly reported to produce trusted evidence and avoid research waste.
Patient, public and/or healthcare consumer involvement: Patients or healthcare consumers were not directly involved in this study. The findings, however, may indirectly contribute to them to produce more trusted evidence in healthcare.