[Skip to Navigation]
Views 551
Citations 0
Comment & Response
March 1, 2021

Methodologic and Reporting Issues in Published Meta-analysis

Author Affiliations
  • 1Department of Brain and Behavioral Sciences, University of Pavia, Pavia, Italy
  • 2Doctoral School “Evidence-Based Assessment and Psychological Interventions,” Babeș-Bolyai University, Cluj-Napoca, Romania
  • 3Department of Clinical Psychology and Psychotherapy, Babeș-Bolyai University, Cluj-Napoca, Romania
JAMA Pediatr. 2021;175(6):643-644. doi:10.1001/jamapediatrics.2021.0010

To the Editor Fraguas et al1 report a meta-analysis of school-based antibullying interventions for prevention of bullying. We have several concerns regarding the methodology and reporting.

Almost all included trials (85%) evaluated universal prevention and should have ideally included a measure of risk reduction, such as incident or recurrent cases.2 The authors did not report any such outcomes. Effects were transformed from standardized mean differences to numbers needed to treat.3 For continuous outcomes (the majority in the meta-analysis), the transformation requires defining a threshold for response,3 an aspect that was not reported. Moreover, outcomes were averaged across heterogeneous categories combining instruments of variable psychometric properties and indexing very diverse constructs. For example, the “mental health problems” category included depression and social anxiety, but also empathy and self-efficacy, among others. The adjudication of instruments to outcome categories was sometimes contradictory, eg, empathy measurements were tallied in one instance at “mental health problems” and in another at “attitudes discouraging bullying”1 (eTable 2 in the Supplement). Predictably, effect estimates for these categories had very large heterogeneity, further limiting clinical relevance.

Add or change institution