Ketone Monitors

How accurate are ketone monitors?

Updated

This site gives general information, not medical advice. Talk to your doctor or diabetes team before changing how you monitor or treat any condition.

We have not tested any device. This page reports what published studies and regulators found.

Abbott’s ketone sensor (2021 feasibility study)

Twelve healthy adults on low-carbohydrate diets each wore three ketone sensors for 14 days, giving 3,132 pairs of sensor and finger-prick readings (Alva et al., 2021):

  • 82.4% of sensor readings were within 0.225 mmol/L or 20% of the finger-prick value, and 91.4% within 0.3 mmol/L or 30%.
  • Below 1.5 mmol/L, the average difference was 0.129 mmol/L. At 1.5 mmol/L and above, the mean absolute relative difference was 14.4%.
  • The signal changed by 2.1% over 14 days on a single calibration. Five sensors did not collect valid data because of device issues.

Abbott Libre Duo 10 Day (FDA review)

The FDA reviewed six clinical studies with more than 600 participants aged 2 and over, and found the device tracked clinically meaningful differences in ketone levels and identified raised ketones before DKA (FDA). 84.1% of adult sensors and 68.8% of children’s sensors lasted the full 10 days (Abbott). Detailed accuracy figures for the commercial sensor were not in the FDA or Abbott announcements.

SiBio KS1 (2026 randomized trial)

Seven healthy volunteers wore SiBio’s KS1 for 14 days while taking ketone drinks or a placebo. The sensor detected the rise in ketones after the drinks, but daily readings declined across the 14 days in both groups, which the authors attribute to sensor drift. The study did not compare the sensor with blood tests, so it could not measure accuracy directly (Kjær et al., 2026).

Blood, breath and urine tests

  • Blood meters: accurate up to about 3.0 mmol/L, with falling precision above about 5 mmol/L (Dhatariya, 2017).
  • Breath meters: moderate correlation with blood BHB; sensor cross-sensitivity and drift limit accuracy (Fante et al., 2025). One study found a significant link between breath and blood readings in adults but not in children (Huang et al.).
  • Urine strips: don’t detect BHB at all, can understate early DKA and stay positive as it resolves (Dhatariya, 2017).

Reading accuracy figures

Sensor studies usually report the share of readings within a set distance of a reference test (for example within 0.225 mmol/L or 20%), or the mean absolute relative difference (MARD): the average gap as a percentage. Lower MARD and higher “within” percentages mean closer agreement. A 2025 review says sensor accuracy, calibration and clinical validation are still challenges for CKMs (Fante et al., 2025).

Sources