INDEPENDENT THINKING. WHOLE-FOOD LIVING.NO ANIMALS / NO OIL
THENANODIET
← Metabolism / Health Desk

Plant-based scores changed—and later diabetes risk changed with them

Three US cohorts examined changes in healthful and unhealthful plant-based indices, rather than a single baseline score.

Whole-food ingredients prepared without added oil

Illustrative food image, not a photograph of the study intervention.

The distinctive question was change

This study asked whether changes in plant-based dietary patterns were associated with later type 2 diabetes, rather than looking only at participants' starting diets. Researchers followed 76,530 women in the Nurses' Health Study, 81,569 women in Nurses' Health Study II and 34,468 men in the Health Professionals Follow-up Study. Every four years, they assessed an overall plant-based diet index, a healthful index and an unhealthful index. The design therefore adds a time dimension to the food-quality question. It remains prospective observational research: participants were not randomly assigned to improve, maintain or reduce their dietary scores over those four-year intervals.

Chen Z, Drouin-Chartier JP, Li Y, Baden MY et al.. Diabetes care, 2021. PMID 33441419 · DOI

Higher scores and subsequent risk

Researchers documented 12,627 type 2 diabetes cases over 2,955,350 person-years. Each 10% increment over four years in the overall plant-based index was associated with a hazard ratio of 0.93 for subsequent diabetes, with a 95% confidence interval of 0.91–0.95. For the healthful index, the corresponding hazard ratio was 0.91, with an interval of 0.87–0.95. These describe approximately 7–9% lower relative risk, as reported by the authors, rather than an absolute reduction of seven to nine percentage points. They also describe changes in an index, not simply adding a single food or adopting a vegan label.

Chen Z, Drouin-Chartier JP, Li Y, Baden MY et al.. Diabetes care, 2021. PMID 33441419 · DOI

Decreases and a useful null result

Compared with participants whose indices remained relatively stable, defined as within plus or minus 3%, those with the largest decreases, greater than 10%, had higher subsequent risk. The pooled hazard ratio was 1.12 for the overall index and 1.23 for the healthful index, with confidence intervals of 1.05–1.20 and 1.16–1.31, respectively. Changes in the unhealthful plant-based index were not associated with diabetes risk. That null result should remain visible: this particular analysis did not find that worsening or improving every plant-based score tracked risk in the same way. Nor does a null association prove that all foods represented by an index are equivalent.

Chen Z, Drouin-Chartier JP, Li Y, Baden MY et al.. Diabetes care, 2021. PMID 33441419 · DOI

Adjustment does not create randomisation

The models accounted for initial body mass index and initial and four-year changes in alcohol intake, smoking, physical activity and other factors. Weight changes accounted for 6.0–35.6% of the associations between changes in the overall and healthful indices and diabetes risk. That range is not evidence that weight explained everything, or that diet acted entirely independently of weight. Statistical adjustment helps frame the observed relationships but does not turn them into a randomised test. The defensible conclusion is the authors' association: improving overall and healthful plant-based scores tracked lower later diabetes risk, while declining scores tracked higher risk in these three US cohorts.

Chen Z, Drouin-Chartier JP, Li Y, Baden MY et al.. Diabetes care, 2021. PMID 33441419 · DOI