A systematic review and meta-analysis pooled 10 studies, randomised and observational, in adults with type 2 diabetes, with follow-up from 8 weeks to 24 months. People with primary polycythaemia, end-stage kidney disease or heart failure were excluded. The question was how far empagliflozin, dapagliflozin and canagliflozin move haemoglobin and haematocrit, and whether that movement matters.
Across three randomised trials totalling 171 patients, haematocrit rose by 2.29 percentage points against placebo (95% CI 1.48 to 3.09) and haemoglobin by 0.51 g/dL (CI 0.28 to 0.74). In real-world data covering 9,646 patients the prevalence of polycythaemia rose from 2.4% to 9.7%, with severe cases in 1.4%. The changes appeared early after starting and reversed on stopping. Evidence on clots is thin: a single observational study of 100 patients with erythrocytosis reported a 10% event rate.
The rise in haematocrit is well known and is probably part of why these drugs work, so this is not a reason to avoid them. But a near one-in-ten prevalence of polycythaemia in real-world use is higher than most clinicians would guess, and the patients who matter are identifiable in advance: those starting with a high-normal haematocrit, smokers, people living at altitude, and those with obstructive sleep apnoea or on testosterone replacement. A full blood count at baseline and at review costs almost nothing and answers the question.
- Record a baseline haemoglobin and haematocrit before starting an SGLT2 inhibitor, and repeat at the 3–6 month review.
- Ask about smoking, obstructive sleep apnoea and testosterone therapy — these stack with the drug effect.
- In a patient with a rising haematocrit, check hydration first; part of the change is volume, not only red cell production.
- Do not stop an SGLT2 inhibitor for a haematocrit rise alone in an asymptomatic patient whose value is still in range — recheck instead.
- If erythrocytosis is marked and persists, remember the change reverses on withdrawal, so a supervised trial off the drug is diagnostic.
The statistics, in plain English
The randomised estimate looks precise — haematocrit up 2.29 points, CI 1.48 to 3.09 — but rests on only 171 patients in three trials, so it is a confident-looking number on a small evidence base. The real-world jump from 2.4% to 9.7% comes from observational data with no comparison group, so part of that rise may reflect who was prescribed the drug rather than what the drug did. The thrombosis figure, 10% in 100 patients, is one study and should be read as a reason to look, not as a risk estimate. Note also an inconsistency in the paper itself: the methods exclude people with end-stage kidney disease and heart failure, while the conclusion mentions them — apply the findings to the population actually studied.
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.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for diabetes & endocrinology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free