Spin bias in clinical trials of cannabis derivatives and their synthetic analogues: a meta-epidemiologic study

Session Type
Communicating evidence including misinformation and research transparency
Pereira GC1, Prates G1, Medina M1, Ferreira C1, Latorraca COC2, Pacheco RL2, Martimbianco ALC3, Riera R4
1Universidade Federal de São Paulo, Brazil
2Centro Universitário São Camilo / Hospital Sírio-Libanês, Brazil
3Universidade Metropolitana de Santos / Hospital Sírio-Libanês, Brazil
4Hospital Sírio-Libanês / Universidade Federal de São Paulo / Cochrane Affiliate Rio de Janeiro, Brazil

Background: numerous trials have been conducted to investigate the effects of cannabis derivatives and their synthetic analogues for different conditions, including pain, nausea and vomiting, eating and anxiety disorders, depression, Alzheimer's disease, multiple sclerosis, epilepsy, among others. Against the setting of a promising novel intervention, investigators may tend to emphasise benefits when interpreting the findings of trials on cannabinoids, which supports the rationale for assessing the frequency of spin bias in these studies. Spin bias occurs, for example, when the results of a study are misreported, with overestimation of positive findings and underestimation of negative ones [1]
Objectives: to estimate the frequency and conduct a qualitative analysis of spin bias in publications of clinical trials about the therapeutic use of cannabis derivatives and their synthetic analogues for any medical condition.
Methods: meta-epidemiologic study [2] conducted in the Discipline of Evidence-Based Medicine, Escola Paulista de Medicina, Universidade Federal de São Paulo (Unifesp), Brazil and supported by academic scholarship from Fundação de Apoio à Pesquisa do Estado de São Paulo (FAPESP, number 2019/20691-4).
Results: 65 publications with at least one efficacy primary outcome were included. The results analysis for the primary outcome indicated statistically significant effects in 44.6% (29/65) of the publications, and 70.7% (45/65) of the conclusions were considered favorable to the intervention. Among the 36 publications that found statistically nonsignificant results for the primary outcome, 44.4% (16/36) presented conclusions favorable to or recommended the intervention, which represents spin bias. Qualitative analysis of the 16 studies with spin bias showed selective outcomes reporting (elevating secondary outcomes that had positive results or reporting only subgroup results), deviations from the planned statistical analysis, and failure to consider or report uncertainty in the estimates of treatment effects.
Conclusions: spin bias were found in 44.4% of the publications of trials assessing the therapeutic use of cannabis derivatives and their synthetic analogues, and presenting statistically nonsignificant results.
Patient, public and/or healthcare consumer involvement: when not observed by readers, spin bias can lead to misconduct in clinical practice through the use of interventions that are not efficacious or whose efficacy is uncertain.