Design: Systematic review and meta-analysis of 10 randomised trials
Population: 2,827 people with diabetes
Primary outcome: Sleep quality and HbA1c
Effect: Sleep quality MD −1.47 (−1.98 to −0.96); HbA1c MD −0.35% (−0.44 to −0.25); I² 89–94%

A meta-analysis in Medicine pooled ten randomised trials (2,827 participants, four countries) comparing cognitive behavioural therapy for sleep with usual care or health education in people with diabetes.

CBT improved sleep quality scores (mean difference −1.47 points) and HbA1c (−0.35 percentage points). Insomnia severity improved by about 2 points, but that confidence interval crossed zero. BMI did not change. Heterogeneity was very high for every outcome except BMI, with I² of 88–94%.

The direction is consistent with what is known about CBT for insomnia in general. The size of the HbA1c effect is less certain: with such variable trials, the pooled number averages very different interventions and populations.

For a patient with diabetes and chronic insomnia, CBT for insomnia is a reasonable first-line option before hypnotics. Do not present it to patients as a glucose-lowering treatment.

Ask about insomnia at the diabetes review; it is common and often unmentioned
Consider CBT for insomnia, including digital or group formats, before a hypnotic
Do not count on sleep therapy to lower HbA1c by a fixed amount
Screen for sleep apnoea before attributing poor sleep to insomnia alone

Why it matters

Poor sleep is a treatable part of diabetes care, not just a complaint to note.

The statistics, in plain English

I² above 75% means the trials disagreed far more than chance would explain, so the average effect may not apply to any single setting. The insomnia severity estimate's interval runs from −4.24 to +0.12, which includes no effect, so that result is uncertain despite the reported P value.