Design: systematic review and meta-analysis of randomised controlled trials, settings analysed separately, PROSPERO registered
Population: 483 critically ill and 137 long-term care patients with hyperglycaemia receiving tube feeding
Primary outcome: glycaemic outcomes — mean blood glucose in critical care, HbA1c in long-term care
Effect: mean glucose −0.54 mmol/L (95% CI −0.78 to −0.31), moderate certainty; HbA1c −0.92% (95% CI −1.72 to −0.13), low certainty

This systematic review pooled randomised trials comparing diabetes-specific nutritional formulas against standard formulas for tube feeding, handling critical care and long-term care separately. Five trials with 483 critically ill patients and three with 137 long-term care patients met the criteria.

In critical care, the specific formulas lowered mean blood glucose by 0.54 mmol/L (95% CI −0.78 to −0.31), at moderate certainty and with low heterogeneity. Measures of glucose spread also improved, but at I-squared of 98% and low certainty, which makes those estimates close to uninterpretable. In long-term care, HbA1c fell by 0.92 percentage points (95% CI −1.72 to −0.13), again at low certainty and I-squared 85%.

Everything else — infections, length of stay, mortality, anything a patient would notice — was either not reported, inconsistent, or not significant. So the finding is a glycaemic one only. Whether a mean glucose 0.54 mmol/L lower in an intensive care unit translates into anything is unknown, and the trials were not built to answer it.

Reserve the question for patients whose enteral feeding is genuinely driving hyperglycaemia, not for everyone with diabetes who needs a tube.
Insulin titration against the feed remains the main lever; formula choice is an adjunct to it.
Diabetes-specific formulas cost substantially more in Indian hospitals — weigh a 0.54 mmol/L mean difference against that.
Record why a formula was changed, so the effect on the glucose chart can actually be read afterwards.

Why it matters

Formula choice is often made by the dietetics service without the treating physician seeing the evidence behind it, and the evidence turns out to be thinner than the marketing.

The statistics, in plain English

Certainty grading is doing the work here. The mean glucose result carries moderate certainty with I-squared 9%, meaning the trials broadly agreed — that one you can rely on. The variability and HbA1c results carry low certainty with I-squared of 98% and 85%, meaning the trials disagreed almost completely, and the pooled number is an average of incompatible results rather than a best estimate. The HbA1c confidence interval reaching to −0.13 also means the true effect could be trivially small.