Low-GI diets and HbA1c: a small average with substantial variation
A synthesis of 29 trial comparisons found a modest HbA1c advantage alongside uneven trial results.

Illustrative food image, not a photograph of the study intervention.
A pattern comparison, not a carbohydrate ban
This review asked whether lower glycaemic index or glycaemic load dietary patterns improved diabetes markers compared with higher-GI or higher-GL diets. It brought together 29 randomized trial comparisons involving 1,617 participants. Eligible trials lasted at least three weeks, so the question was broader than the response to a single meal. The participants had type 1 or type 2 diabetes, although the evidence predominantly concerned middle-aged adults with overweight or obesity and moderately controlled type 2 diabetes. These details define the population behind the headline; they do not establish an identical response for every person or every carbohydrate-containing food.
Chiavaroli L, Lee D, Ahmed A, Cheung A et al.. BMJ (Clinical research ed.), 2021. PMID 34348965 · DOI
Read the HbA1c difference in its own units
The pooled HbA1c difference favored lower-GI or lower-GL patterns by 0.31 percentage points. Its 95% confidence interval ranged from 0.19 to 0.42 percentage points lower than the comparison diets. That is a difference between dietary groups, not a claim that HbA1c fell by 31%, and not a remission rate. HbA1c was the review's primary outcome, which makes it the clearest place to start when interpreting the findings. The authors rated certainty for this result as high. Their conclusion described small improvements in established treatment targets rather than evidence that dietary patterns eliminated diabetes or replaced its treatment.
Chiavaroli L, Lee D, Ahmed A, Cheung A et al.. BMJ (Clinical research ed.), 2021. PMID 34348965 · DOI
The pooled result was not uniform
The HbA1c analysis had substantial statistical heterogeneity, with I-squared of 75%. That variation belongs beside the average effect rather than in a footnote: the trials did not all produce the same result. The review also found reductions in fasting glucose, LDL cholesterol, non-HDL cholesterol, apo B, triglycerides, body weight, BMI and C-reactive protein. It did not find reductions in blood insulin, HDL cholesterol, waist circumference or diastolic blood pressure. These different findings argue against describing the intervention as an across-the-board metabolic improvement. Certainty was moderate for most secondary outcomes, with imprecision the main reason for downgrading.
Chiavaroli L, Lee D, Ahmed A, Cheung A et al.. BMJ (Clinical research ed.), 2021. PMID 34348965 · DOI
An addition to concurrent treatment
Most participants were receiving glucose-lowering drugs or insulin. The authors therefore framed the dietary results as improvements beyond concurrent treatment, not as proof that treatment could be withdrawn. The synthesis measured glycaemic and other risk markers; its abstract does not report a reduction in clinical complications or a diabetes-remission endpoint. It also discloses extensive author relationships with food-industry and other organizations, including a fellowship jointly funded by the Canadian Sugar Institute. Those disclosures are relevant context, not grounds to erase the randomized evidence. The defensible takeaway remains a modest average advantage for a dietary pattern in the populations actually studied.
Chiavaroli L, Lee D, Ahmed A, Cheung A et al.. BMJ (Clinical research ed.), 2021. PMID 34348965 · DOI