Background: Designs of studies examining dietary patterns (DPs) carry different limitations including short durations and conditions not reflective of the real world in controlled trials (CTs) and confounders associated with dietary choices and health outcomes in observational studies (OSs). Examining both study designs may allow for a more accurate understanding of the impact of DPs. Objective: To compare the direction of findings, effect sizes and certainty of evidence (COE) between CTs and OSs examining the relationship between a vegan DP and body mass index (BMI).
Methods: An overview of systematic reviews (SRs) was conducted by healthcare consumers by searching five databases for SRs published from 2018 to January 2023 that examined adults in the general population who followed a vegan, compared with a nonvegetarian, DP. SRs could include primary studies with any study design. If COE was not assessed in included SRs, authors assessed COE for CTs and OSs separately using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) method.
Results: A total of 524 abstracts were screened, 73 full SRs were reviewed, and 5 SRs examined the impact of a vegan diet on BMI. Three SRs included 11 CTs ranging from 12 to 26 weeks in duration and concluded that a vegan diet reduced BMI compared with no intervention or another intervention diet. However, nearly all included CTs demonstrated some risk of bias and sample sizes were relatively small, resulting in low COE. Two SRs focused on OSs, summarized 67 cross-sectional and cohort studies, many reflecting several years of follow-up, and described low risk of bias in individual studies. COE was not marked down for any domain but was low owing to study design. Pooled mean effect size from OSs (-1.72 kg/m2) was similar to the range in mean reduction found in CTs comparing a vegan diet with no intervention (no pooled data; mean reduction from -1.03 to -3.90 kg/m2). Conclusion: Interpreting the full body of evidence when examining the impact of a DP revealed consistent findings across study designs and led to improved confidence in findings. Methodologists may consider upgrading OSs for plausible confounding if results are consistent with CTs.