Soy protein and LDL: a small pooled effect, carefully counted
The title mentions 46 studies, but 41 supplied LDL data.

Illustrative food image, not a photograph of the study intervention.
Count the studies behind the title
The 2019 soy-protein meta-analysis examined 46 controlled trials identified by the US Food and Drug Administration. Of those, 43 supplied data suitable for meta-analysis, and 41 contributed LDL-cholesterol results. Keeping those counts separate avoids implying that all 46 studies supplied every outcome. The investigators compared soy protein with non-soy protein in adults, not soy consumers with nonconsumers followed for decades. This is intervention evidence about circulating cholesterol, and it should remain distinct from observational claims about cancer, longevity, or overall dietary identity.
Blanco Mejia S, Messina M, Li SS, Viguiliouk E et al.. The Journal of nutrition, 2019. PMID 31006811 · DOI
The size of the measured effect
At a median soy-protein dose of 25 grams daily and a median follow-up of six weeks, the pooled LDL difference was 4.76 mg/dL lower with soy protein. The 95% confidence interval ranged from 6.71 to 2.80 mg/dL lower. The authors characterized the reduction as approximately 3% to 4%. These are two ways of describing the pooled lipid result, not separate benefits to add together. Neither estimate says that every person eating a soy food will experience the same absolute or relative LDL change.
Blanco Mejia S, Messina M, Li SS, Viguiliouk E et al.. The Journal of nutrition, 2019. PMID 31006811 · DOI
Protein dose is not food weight
The reported 25 grams refers to soy protein, not 25 grams of tofu, edamame, or a finished dish. Converting the trial's median exposure into a food portion would require additional compositional information and attention to the individual interventions. The pooled comparison also matters: soy protein was assessed against non-soy protein. It does not automatically estimate the consequence of simply adding extra food without changing anything else. A useful reading therefore retains both the protein unit and the comparator, rather than reducing the paper to a serving-size slogan.
Blanco Mejia S, Messina M, Li SS, Viguiliouk E et al.. The Journal of nutrition, 2019. PMID 31006811 · DOI
Do not borrow a broader claim
The study arose in the context of an FDA reassessment of a soy heart-health claim. That historical setting does not verify the current legal status of any labeling rule, and this article makes no claim about today's regulation. More importantly, an LDL meta-analysis is not itself a trial of cardiovascular events. The result supports a modest cholesterol-lowering effect of the tested soy-protein comparisons over generally short follow-up. It does not establish an effect for an isolated isoflavone supplement, a specific recipe, or replacing prescribed lipid treatment with soy foods.
Blanco Mejia S, Messina M, Li SS, Viguiliouk E et al.. The Journal of nutrition, 2019. PMID 31006811 · DOI