Weekly well-day ketone checks predict DKA risk as well as twice-weekly checks

Design

Simulation of testing frequencies using regression and gradient-boosted tree models on pooled trial data

Population

1,410 adults with type 1 diabetes from the EASE 2 and EASE 3 trials

Primary outcome

Discrimination for DKA or severe ketosis in the following month

Effect

AUC 0.678 weekly vs 0.692 twice-weekly (P = 0.19); GBT 0.711 vs 0.719 (P = 0.38)

Earlier work showed that routine capillary ketone testing on well days, over a month, predicts diabetic ketoacidosis risk in the following month independently of clinical factors. This study used the EASE 2 and EASE 3 trial repository (1,410 adults with type 1 diabetes, the empagliflozin-as-adjunct trials) to simulate how often testing needed to be done to keep that accuracy.

Once-weekly testing was the lowest frequency that matched twice-weekly testing: area under the curve 0.678 vs 0.692 (P = 0.19) in the maximum-ketone model and 0.711 vs 0.719 (P = 0.38) in the machine-learning model. The authors suggest using existing strips before they expire.

This matters most for people with type 1 diabetes on an SGLT2 inhibitor, where ketosis can occur with near-normal glucose, and for anyone with a DKA history. Weekly testing is a burden people are more likely to sustain. Note the discrimination is modest — this stratifies risk, it does not diagnose impending DKA.

• For people with type 1 diabetes at DKA risk, a weekly well-day capillary ketone check is enough to stratify risk.
• Prioritise this for anyone on an SGLT2 inhibitor with type 1 diabetes, and anyone with previous DKA.
• Test on a well day, at a consistent time, and ask the patient to note the highest value each month.
• Rising well-day ketones are a prompt to review insulin doses, carbohydrate intake and SGLT2 use.
• Sick-day ketone testing rules are unchanged — test whenever unwell, whatever the glucose.

Why it matters

It makes routine ketone surveillance realistic enough that patients might actually keep doing it.

Don't overread it

The testing schedule was simulated in trial data, and an AUC near 0.7 stratifies risk rather than predicting individual DKA.

The statistics, in plain English

The area under the curve measures how well a test separates people who will and will not develop DKA: 0.5 is a coin toss, 1.0 is perfect. Values around 0.7 are modest. The non-significant P values mean weekly testing was not detectably worse than twice-weekly, not that the two are proven identical. This was a simulation within trial data, not a trial of testing schedules.

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.