Continuous glucose monitoring in surgical and ICU patients agreed well with point-of-care tests

Design

Prospective observational accuracy study

Population

58 adults with diabetes having surgery or ICU admission

Primary outcome

Median absolute relative difference between CGM and point-of-care glucose

Effect

MARD 10.9% (5.1 to 19.4%); 98% of paired readings in error grid zones A and B

This prospective observational study compared a continuous glucose monitor with point-of-care glucose measurements in 58 people with diabetes needing glucose-lowering medication who had surgery or ICU admission. Mean age was 71.

The median absolute relative difference was 10.9%, and 1,186 of 1,212 paired readings (98%) fell in clinically acceptable error grid zones. Bland-Altman analysis gave a mean difference of -1 mM (95% limits of agreement -5.6 to 3.4 mM). Agreement was worse during hypoperfusion, by 4.5% (95% CI 2.3 to 6.7).

No hypoglycaemia (below 4 mM) occurred, so accuracy in the low range, the dangerous one, could not be tested. The limits of agreement are wide, so a single reading can differ from a point-of-care value by several millimoles per litre. The device is a useful addition rather than a replacement, particularly in shock.

• Continuous monitoring tracked point-of-care values well in most paired readings.
• Confirm with a point-of-care or laboratory glucose when readings drive insulin changes, especially in shock or poor perfusion.
• Do not rely on it for hypoglycaemia detection; the study had no low values to test.
• Check local regulatory approval for in-hospital use of the device.

Why it matters

Hospital glucose care still depends on intermittent fingerprick testing, and this shows that a sensor can follow it reasonably in sick patients.

Don't overread it

The study tested accuracy only, in 58 patients with no hypoglycaemia, and did not test whether using the sensor improves outcomes.

The statistics, in plain English

The limits of agreement (-5.6 to 3.4 mM) show where 95% of the differences fell, and that range is wide enough to change a dosing decision. A MARD of 10.9% is good overall, but averages hide the occasional large error.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.