Metformin at eGFR 24: A Number That May Not Be Real
A progressive renal decline and a week of poor intake land on the same visit, and the disagreement isn't really about whether metformin is safe below eGFR 30 — it's about whether today's number can be trusted enough to have that argument at all.
Warren T., a 74-year-old man, has spent most of the last four years restoring a 1967 Airstream trailer in his driveway, and the plan — once the plumbing and the axles are both trustworthy — is a cross-country trip with his wife to see their grandchildren in Oregon. He has type 2 diabetes of fourteen years' duration, well controlled for most of that time on metformin 1000mg twice daily, plus longstanding hypertension on lisinopril and hyperlipidemia on atorvastatin; his last three A1c values have run 6.8–7.1%, and he has never been hospitalized for a diabetes-related complication. He mentions, almost as an aside while the nurse draws his labs, that he's had about a week of loose stools and a poor appetite after what he assumed was a bad batch of gas-station coffee on a supply run for Airstream parts — nothing that kept him from working on the trailer, but enough that he's been drinking less water than usual.
Today's basic metabolic panel puts his eGFR at 24, and the number by itself would be concerning enough — well below the 30 mL/min/1.73m² threshold his metformin labeling still names — but reading it against his own chart makes it harder to interpret cleanly. His eGFR was 42 a year ago and 33 six months ago: a real, progressive downward trend, the kind that argues for a genuine chronic decline rather than a one-time lab artifact. But this visit also carries a plausible acute layer on top of that trend — a week of reduced intake and diarrheal losses is exactly the kind of transient volume depletion that can drop a measured eGFR further than a patient's true, rehydrated baseline would show. The number in front of the team this morning may not be reliably his chronic number at all, which means the question isn't only whether to continue metformin below 30 — it's whether today's 24 is even the right number to be having that argument about.
At the follow-up visit, reading a number that may not be real
Stop the metformin today. The FDA's own 2016 label revision, the one that walked the contraindication back from a serum-creatinine cutoff to an eGFR-based one, still draws its line at 30 — and he's at 24. Lactic acidosis from metformin accumulation is rare, but when it happens in a man his age it doesn't present as a nuisance; it presents in the ICU. He has safe alternatives. I don't think a fourteen-year track record of good control is a reason to keep him on a drug that's now, by the numbers in front of us, outside its own labeled boundary.
I'd push back on how much weight the label itself should carry here. The Cochrane review by Salpeter and colleagues pooled decades of observational data, including patients who met the old contraindication criteria, and found no cases of lactic acidosis clearly attributable to metformin at rates any higher than in patients without renal impairment. The 2016 revision was itself an acknowledgment that the original cutoff was more conservative than the evidence underneath it.
And the alternative isn't risk-free either — stopping metformin in a man this well controlled usually means starting insulin, and insulin clearance itself declines with renal function. Hypoglycemia in a 74-year-old on a new insulin regimen is a common, documented harm, not a theoretical one, and it's the kind of harm we'd be trading the theoretical one for.
I don't think we're actually ready to have either version of this argument yet. Look at the trend: 42 a year ago, 33 six months ago, 24 today, and now overlay a week of poor intake and diarrheal losses on top of it. That's a textbook setup for a volume-depletion dip riding on top of a real chronic decline — which means today's 24 may understate his actual rehydrated baseline, and neither of you has a chronic number to defend a position about yet.
This doesn't require resolving whose reading of the acidosis literature is right. Hold the metformin today, on the acute-illness grounds alone — that's standard sick-day guidance regardless of his renal trend — encourage oral rehydration, and recheck his renal function in one to two weeks once he's back to baseline intake. If he settles back toward the low 30s, this is a different, more defensible conversation about continuing at a reduced dose. If he holds at the mid-20s even rehydrated, then the disagreement above is the real one to have, with a number we can actually trust.
Agreed within minutes once framed as a sick-day hold rather than a discontinuation decision: metformin paused, oral rehydration encouraged, renal function rechecked in ten to fourteen days. All three voices treated this as compatible with their own position rather than a compromise between them — the nephrologist still expects the recheck to confirm real progressive CKD; the pharmacologist still expects it to settle back toward the low 30s.
Not agreed, and deliberately left open rather than smoothed over: what happens if the recheck lands exactly where the chronic trend line already predicted, in the mid-20s with no acute illness left to blame. The nephrologist and pharmacologist both confirmed they'd be back at the same disagreement, with the same arguments, and no new information to settle it — a genuine unresolved question about the site's own historical labeling versus its own contradicting evidence, one this visit's rehydration plan was never going to answer either way.