Tree nuts in diabetes trials: keep the effect size small
Twelve trials found modest glucose-marker changes, with short duration and poor study quality limiting confidence.

Illustrative food image, not a photograph of the study intervention.
What the nut trials actually compared
The tree-nut review brought together randomized dietary trials lasting at least three weeks in people with diabetes. The eligible comparisons emphasized tree nuts in one diet and used an isocaloric diet without tree nuts as the control. Twelve trials, involving 450 participants, met the review criteria. The investigators examined HbA1c, fasting glucose, fasting insulin and HOMA-IR. This is important context for interpreting the result: the evidence concerns diets compared at equal energy, not an experiment that simply added nuts without regard to the rest of the diet. The median amount used in the included trials was 56 grams per day.
Viguiliouk E, Kendall CW, Blanco Mejia S, Cozma AI et al.. PloS one, 2014. PMID 25076495 · DOI
A statistically detectable but modest difference
Compared with control diets, the nut-emphasizing diets lowered HbA1c by 0.07 percentage points on average. The 95% confidence interval ranged from a reduction of 0.10 to 0.03 percentage points, with P equal to 0.0003. Fasting glucose was lower by 0.15 mmol/l, with a confidence interval from 0.27 to 0.02 mmol/l lower and P equal to 0.03. These are the actual pooled differences reported in the abstract. Describing them as large improvements would discard the scale of the evidence. The result supports a small average marker difference in these trial comparisons, rather than a sweeping claim about diabetes control.
Viguiliouk E, Kendall CW, Blanco Mejia S, Cozma AI et al.. PloS one, 2014. PMID 25076495 · DOI
The insulin findings were not significant
The pooled analysis did not find significant treatment effects on fasting insulin or HOMA-IR. Although the direction of those estimates favoured tree nuts, the abstract explicitly separates that direction from a statistically significant treatment effect. This makes the four measured outcomes a mixed set rather than four confirmed benefits. The HbA1c and fasting-glucose findings should not be used to fill in an insulin result that the analysis did not establish. The review's own limitations also matter here: most trials were short and of poor quality. Those weaknesses qualify the positive marker findings as well as the outcomes that were not significant.
Viguiliouk E, Kendall CW, Blanco Mejia S, Cozma AI et al.. PloS one, 2014. PMID 25076495 · DOI
A median trial amount is not a prescription
The 56 grams per day figure is the median trial amount, not an individualized recommendation derived from this review. The authors called for longer, higher-quality trials, including research focused on replacing high-GI carbohydrates with nuts. That proposed focus is a future research need, not proof that such a replacement explains every included result. The PubMed disclosure statement also lists author relationships with nut-industry organizations, including research support from the Almond Board of California and the International Tree Nut Council. Those disclosures belong alongside the short-duration and quality limitations when weighing this small pooled effect on measured glucose-control markers in diabetes.
Viguiliouk E, Kendall CW, Blanco Mejia S, Cozma AI et al.. PloS one, 2014. PMID 25076495 · DOI