Identifying ineffective interventions from Cochrane reviews through the interpretation of the main Summary of Findings table

Article type
Authors
Franco JVA1, Vergara-Merino L2, Verdejo C2, Schuller-Martínez B2, Kirmayr M2, Kopitowski K3, Madrid E4, Meza N4, Pérez-Bracchiglione J4, Muscat Meng D5, Johansson M5
1Research Department ,Cochrane Associate Centre, Instituto Universitario Hospital Italiano
2Escuela de Medicina, Universidad de Valparaíso, Centro Asociado Cochrane Chile
3Family and Community Medicine Division, Hospital Italiano de Buenos Aires
4Interdisciplinary Centre for Health Studies (CIESAL), Cochrane Associate Centre, School of Medicine, Universidad de Valparaíso
5Cochrane Sustainable Healthcare
Abstract
Background:
Cochrane reviews (CR) of interventions summarise the evidence of the comparative effectiveness of interventions in healthcare by collating all available randomized trials and performing, when possible, meta-analysis. CR have highlighted the importance of the implementation of highly effective interventions (e.g. corticosteroids for women at risk of premature birth), however, their impact on de-implementation is unclear. Currently, there is not a comprehensive summary of recent high-quality research on interventions that might be candidates for de-implementation due to their lack of efficacy or their associated harms. Furthermore, more information is needed in terms of how those interventions can be defined (via the author’s conclusions or summary of findings [SoF] tables).
Objectives:
To identify CR in which the main SoF table described moderate-to-high quality evidence of no effect or suggested harm.
Methods:
We analysed CR interventions with SoF tables. We then identified those reviews in which: the estimates of at least one outcomes with moderate or high quality of evidence in the main SoF table included no effect (e.g. significant difference thresholds for dichotomous and continuous outcomes), and/or the estimates of adverse events suggested harm (e.g. beyond significant difference thresholds for dichotomous and continuous outcomes). We presented a descriptive summary of the characteristics of the patients, interventions, comparisons and outcomes (PICO) questions covered by these reviews. We also analysed the correspondence between the author’s conclusions and confidence in the outcomes of SoF tables.
Results
We screened 597 CR and found 36 reviews (see attached) with moderate or high-quality evidence of no beneficial and/or harmful effect for at least one outcome in the main SoF table. In five of these reviews, authors did not conclude that the interventions should potentially be deimplemented; in two reviews this was based on outcomes not presented in the main SoF table and in three reviews on uncertainties in the body of evidence. Our main limitations included that we did not assess secondary SoF tables and the variable report of SoF tables (not systematically assessing harms or clinically relevant outcomes). Additionally, since we included reviews with non-statistically significant measures, we might have missed reviews that reported statically significant but clinically irrelevant effects (that would also be candidates for de-implementation). We could not screen them systematically because thresholds for clinical significance vary substantially across reviews.
Conclusions
We provide a structured process to identify interventions that may be considered ineffective and/or harmful. Whereas we found limitations in our screening process, we believe that this process is sustainable and can serve as input for a knowledge translation (KT) strategy for the Sustainable Healthcare Field. We will pilot a Knowledge Translation package for these 36 reviews and present it to relevant stakeholders.
Patient involvement: no.