OBJECTIVES:

Retractions related to COVID-19 publications has highlighted how prevalent retractions of primary research can be. Whilst there is a responsibility for authors of primary research to correct errors or remove fraudulent studies, the responsibility of correcting errors in downstream research is less clear. To further understand the situation we identifyed case studies where evidence synthesis has included retracted articles or articles with subsequently identified errors, how it has been handled and propose potential solutions.

METHODS:

Identified case studies from work the abstract authors were undertaking where evidence synthesis ended up including primary studies with errors or retractions included.

RESULTS:

Case studies with different responses to errors in primary articles were identified: A typographical error in a primary study was identified in evidence synthesis used to support a NICE health technology assessment; authors of the primary study acknowledged the error but did not correct the publication meaning the error remained in the publication and subsequent meta-analyses. Retracted COVID-19 publications were acknowledged by authors of one published meta-analysis; authors updated analysis to remove the offending research. A primary study was retracted whilst the evidence synthesis publication was under peer-review; following publication authors were notified of the retraction and opted to retract rather than revise their analysis.

CONCLUSIONS:

Prevailing errors in evidence synthesis, despite retraction or correction of a primary study, could negatively influence healthcare decision making. Living systematic reviews or meta-analyses are one solution but may not always be practicable. To minimise the risk of perpetuating errors, authors of secondary analyses could commit to ensure data included in their analyses is accurate for a defined period after the initial analysis has been conducted, e.g., 2 years, after which the analyses could be presumed to be outdated. The defined period could be influenced by several factors including rate of new studies being published in the indication.