J-DOIT3 re-analysis finds no clear lower HbA1c floor when hypoglycaemia is rare

Design

Post-hoc landmark analysis of a randomised trial

Population

2,540 Japanese adults with type 2 diabetes followed for a median 8.5 years

Primary outcome

Risk of myocardial infarction, stroke and death by updated mean risk-factor levels

Effect

No clear lower HbA1c threshold for MI or death; LDL below 80 mg/dL linked to lowest stroke risk

J-DOIT3 randomised 2,542 Japanese adults with type 2 diabetes to intensive or conventional control of glucose, blood pressure and lipids, and followed them for a median of 8.5 years. Severe hypoglycaemia was rare in both arms: seven events in the intensive group and four in the conventional group. This secondary analysis took a landmark at one year and related updated mean HbA1c, systolic pressure and cholesterol to later outcomes.

For glucose, the authors found no clear lower HbA1c threshold below which the risk of myocardial infarction or death began to climb again, and suggest that a target below 6.5% might minimise mortality where severe hypoglycaemia is uncommon. For lipids, LDL cholesterol below 80 mg/dL was linked to the lowest stroke risk.

This is a post-hoc, observational look inside a trial, not a test of a lower target. Patients who reached low HbA1c values may differ in many ways from those who did not. The abstract does not report hazard ratios, so the size of any gradient cannot be judged from it.

The practical message is narrower than the headline: the old fear of tight control rested partly on hypoglycaemia, and with agents that rarely cause it, the case for loosening targets in well people is weaker. Individualised targets set by current guidelines have not changed.

• Record severe hypoglycaemia at every review; it is what made tight targets dangerous in older trials.
• Where a patient reaches a low HbA1c on agents that rarely cause hypoglycaemia, there is no reason here to de-intensify.
• Do not push sulfonylureas or insulin harder to chase a sub-6.5% target on the strength of this analysis.
• Keep LDL cholesterol in view alongside glucose; stroke risk tracked LDL in this cohort.

Why it matters

It questions the assumption that very low HbA1c is harmful in itself rather than through the drugs used to reach it.

Don't overread it

This is a post-hoc association inside a trial; it did not randomise anyone to a target below 6.5%.

The statistics, in plain English

A landmark analysis starts the clock one year in and compares people by the values they had reached. Because those values were not randomised, any link with outcomes is an association and can be confounded by health, adherence or frailty.

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.