Design: Systematic review and meta-analysis of 23 RCTs
Population: 3,055 women with gestational diabetes
Primary outcome: Glycaemic control and pregnancy outcomes
Effect: Fasting glucose MD −0.26 (−0.42 to −0.09), 95% PI −0.93 to 0.42; I² >90%

This meta-analysis pooled 23 randomised trials covering 3,055 women with gestational diabetes, comparing telemedicine-supported care with usual care. Pooled estimates favoured telemedicine for fasting glucose (mean difference −0.26 mmol/L), two-hour post-meal glucose (−0.50) and gestational weight gain (−0.78 kg), and it was associated with fewer caesareans, less macrosomia and less neonatal hypoglycaemia.

The caveats are large. Heterogeneity for the glucose and weight outcomes exceeded 90%, and the prediction intervals for each crossed zero, meaning a future trial could plausibly show no benefit. HbA1c, preterm birth and NICU admission did not differ. GRADE certainty was low to very low throughout.

For services stretched by distance or clinic capacity — a common situation in India — remote glucose review is a reasonable supplement. It should sit on top of antenatal visits, not replace them.

Consider remote glucose-log review to supplement, not replace, antenatal visits in gestational diabetes.
Programmes differed widely between trials; define what is reviewed, by whom and how often.
HbA1c, preterm birth and NICU admission did not improve in the pooled data.
Most useful where travel or clinic capacity limits in-person follow-up.

Why it matters

It tempers enthusiasm for replacing clinic visits with apps on the strength of a pooled average.

The statistics, in plain English

A prediction interval estimates where the result of the next similar trial would fall. When it crosses zero despite a significant pooled average, the trials disagree so much that benefit in any one setting is uncertain.