An A1c of 6.1% and a Glucose Log That Tells a Different Story
His A1c looks excellent. His own glucose log and two real hypoglycemic episodes say otherwise — and in end-stage renal disease, that gap between the number and the reality is a known, documented problem, not a fluke.
Arthur N., a 67-year-old man, has played chess at the same public library table every Wednesday afternoon for as long as anyone at the branch can remember, a ritual he's kept up even on dialysis days by scheduling his sessions around it. He has end-stage renal disease from longstanding diabetic nephropathy, on hemodialysis three times weekly for the past two years, and type 2 diabetes managed on basal insulin. His most recent A1c came back at 6.1%, a number that would ordinarily suggest excellent control — but his home glucose log tells a different story, with values swinging widely between 60 and 240 mg/dL over the same period, including two symptomatic hypoglycemic episodes he reported to the dialysis unit last month.
That discrepancy isn't a data error; it's a known limitation of A1c specifically in end-stage renal disease. A1c reflects average glucose exposure over red blood cell lifespan, and dialysis patients often have shortened red cell survival plus erythropoietin-stimulated new cell production, both of which can artificially lower the measured value regardless of his true average glucose. Reading his 6.1% at face value, in other words, risks concluding his control is better than it actually is — and adjusting his insulin dose based on that number alone could make his real hypoglycemia problem worse, not better.
A number that looks reassuring and a log that isn't
I'd keep anchoring to A1c, reading his 6.1% as likely lower than his true average given the known ESRD bias, rather than switching to a different marker altogether. A1c has the largest, most familiar evidence base of any glycemic measure — a less-established alternative could carry its own unknown biases we haven't fully characterized yet.
I'd switch to glycated albumin for him specifically. Freedman’s group, working in the dialysis population, validated it as tracking actual glycemic exposure more accurately than A1c in exactly this population, since it isn't confounded by red-cell lifespan or erythropoietin the way his A1c clearly is right now — his own albumin is normal, so the limitation glycated albumin carries in malnourished patients doesn't apply to him.
I understand the appeal of staying with a familiar marker, but "familiar" isn't the same as "accurate for this specific patient" — his own numbers are the clearest demonstration of that his chart could show.
I don't think either averaged marker actually solves the problem in front of us. He's already had two real hypoglycemic episodes, and neither A1c nor glycated albumin tells us when those are happening relative to his dialysis sessions. Real-time continuous glucose monitoring, now increasingly validated even in ESRD patients, gives us that directly.
I'd start CGM regardless of which averaged marker we ultimately trust more for long-term trend-tracking. For a patient having real, symptomatic hypoglycemia, knowing exactly when it's happening — intradialytic, post-dialysis, overnight — matters more right now than resolving which surrogate lab value is theoretically more accurate.
Basal insulin reduced today given his documented hypoglycemia, CGM started to directly track his glucose pattern around dialysis sessions, and glycated albumin added to his lab panel for longer-term trend comparison against A1c. Arthur was relieved the team took his home log seriously rather than reassuring him based on the A1c alone.
Not agreed: whether A1c should be dropped from his panel entirely going forward, or kept as a secondary reference point despite its known unreliability in his case. The nephrologist favored dropping it; the primary care physician wanted it retained for continuity with his longitudinal record and comparison across his care team. Left as a documentation preference, not a clinical dispute affecting his actual treatment.