Vegetarian weight-loss trials: the average and the long-term gap
A meta-analysis found a modest average weight difference, with smaller estimates in trials lasting at least a year.

Illustrative food image, not a photograph of the study intervention.
What was actually pooled
Huang and colleagues pooled 12 randomized controlled trials involving 1,151 participants. Their searches covered PubMed, EMBASE, and UpToDate through September 22, 2014. The interventions had a median duration of 18 weeks, and the analysis used a random-effects model after assessing heterogeneity. The central outcome was net change in body weight, not a clinical endpoint such as diabetes incidence. Across the studies, vegetarian groups lost an average of 2.02 kilograms more than nonvegetarian groups, with a 95% confidence interval from 1.23 to 2.80 kilograms more loss. That is a between-group difference, not a promise that every participant lost that amount.
Huang RY, Huang CC, Hu FB, Chavarro JE. Journal of general internal medicine, 2016. PMID 26138004 · DOI
Vegan and lacto-ovo results were not identical
The vegan subgroup had a pooled weight difference of -2.52 kilograms, with a 95% confidence interval of -3.02 to -1.98. The lacto-ovo-vegetarian estimate was -1.48 kilograms, but its confidence interval ranged from -3.43 to 0.47. Although the abstract's prose describes benefit in both categories, the latter interval includes zero, so it is not a clear standalone statistical signal. These are subgroup summaries within a meta-analysis, not a direct randomized comparison establishing that vegan diets outperform lacto-ovo diets. They also do not isolate an extracted-oil restriction. A strict whole-food, no-oil approach is more specific than the intervention category named in this review.
Huang RY, Huang CC, Hu FB, Chavarro JE. Journal of general internal medicine, 2016. PMID 26138004 · DOI
Energy restriction was another moving part
Trials that included energy restriction had a pooled weight difference of -2.21 kilograms, with a 95% confidence interval of -3.31 to -1.12. Those without energy restriction had a difference of -1.66 kilograms, with a confidence interval of -2.85 to -0.48. The abstract describes the former as a greater reduction, but these subgroup estimates should not be treated as a randomized test of calorie restriction itself. Different trials can differ in other ways too. What the data do show is that a weight advantage appeared in both categories. They do not establish a single calorie target or identify one meal as the explanation for the pooled effect.
Huang RY, Huang CC, Hu FB, Chavarro JE. Journal of general internal medicine, 2016. PMID 26138004 · DOI
The longer-term estimate was smaller
For studies with follow-up shorter than one year, the pooled weight difference was -2.05 kilograms, with a 95% confidence interval of -2.85 to -1.25. In studies lasting at least a year, it was -1.13 kilograms, with a confidence interval of -2.04 to -0.21. This contrast places the 18-week median duration in perspective: the evidence base was weighted toward relatively short interventions. The authors explicitly called for more long-term trials of body-weight control. The review supports a modest average advantage relative to the comparison diets tested. It does not establish permanent weight maintenance or the effect of any particular recipe paired with this article.
Huang RY, Huang CC, Hu FB, Chavarro JE. Journal of general internal medicine, 2016. PMID 26138004 · DOI