Plant-based diet scores: quality changes the question
Healthy and unhealthy plant-based indices pointed in different directions in US cohorts.

Illustrative food image, not a photograph of the study intervention.
A plant label is not the scoring system
Satija and colleagues studied three large US cohorts using several plant-based diet indices. The overall index rewarded plant foods and scored animal foods in the opposite direction. The healthful index instead rewarded selected plant-food groups, including whole grains, fruits, vegetables, nuts, and legumes, while reverse-scoring less-healthful plant groups and animal foods. The unhealthful index reversed the emphasis between the two plant-food categories. These are research scoring rules, not diagnoses of a person’s diet. Importantly, they are not a simple comparison of everyone who identifies as vegan with everyone who does not.
Ambika Satija, Shilpa N Bhupathiraju, Eric B Rimm, Donna Spiegelman et al.. PLoS medicine, 2016. PMID 27299701 · DOI
The direction depended on food quality
The investigators documented 16,162 incident type 2 diabetes cases. After adjustment including body mass index, higher healthful-index scores were associated with lower incidence, whereas higher unhealthful-index scores were associated with higher incidence. Comparing extreme deciles, the reported hazard ratios were 0.66 for the healthful index and 1.16 for the unhealthful index. These are relative associations between groups, not individual forecasts. Presenting only the word plant-based would lose the main distinction the investigators set out to examine: the categories of plant foods counted in the score changed the relationship observed in the data.
Ambika Satija, Shilpa N Bhupathiraju, Eric B Rimm, Donna Spiegelman et al.. PLoS medicine, 2016. PMID 27299701 · DOI
An index cannot isolate one ingredient
The study assessed diet repeatedly using food-frequency questionnaires, rather than assigning menus. The authors explicitly noted possible measurement error and residual or unmeasured confounding. A score also combines multiple foods, so its association cannot identify one ingredient as the cause. Because both specialized indices reverse-scored animal foods, their contrast should not be recast as an experiment proving that any particular animal food must be replaced by any particular plant food. The design supports an association between dietary patterns and later diagnoses. It does not establish a clinical effect for each item within a scoring category.
Ambika Satija, Shilpa N Bhupathiraju, Eric B Rimm, Donna Spiegelman et al.. PLoS medicine, 2016. PMID 27299701 · DOI
Make the distinction useful at the store
A more useful shopping question than whether a package carries a plant-based claim is what role it would play in your usual meals. You might compare a meal organized around beans and grains with one organized around sweet drinks and desserts, without pretending to calculate a validated research score in the aisle. The paper offers a vocabulary for discussing pattern quality, not a moral ranking of eaters or a ban on an occasional food. Keep its endpoint in view: incident diabetes in cohorts. It was not a trial of your shopping list, a diabetes-treatment protocol, or a prediction that a single substitution guarantees a result.
Ambika Satija, Shilpa N Bhupathiraju, Eric B Rimm, Donna Spiegelman et al.. PLoS medicine, 2016. PMID 27299701 · DOI