SGLT2 Inhibitor Initiation at an eGFR of 23
A single patient with non-diabetic proteinuric CKD, declining steadily for over a year, sits at an eGFR the pivotal SGLT2-inhibitor trials only barely reached. The disagreement is whether today is still early enough to start the drug, or whether her numbers have already moved past the point where starting helps more than it alarms.
Dana K., a 61-year-old woman, spends three afternoons a week picking her grandchildren up from school, a routine she rearranged her whole retirement around after thirty-one years managing a middle-school cafeteria. She has carried a hypertension diagnosis for two decades, reasonably controlled on losartan and amlodipine, and has never had diabetes — her most recent HbA1c was 5.6%. Nephrology has followed her for four years for what her chart calls presumed hypertensive nephrosclerosis: no biopsy was ever pursued, since her urine albumin-to-creatinine ratio and slow, steady decline fit the picture well enough that a tissue diagnosis wasn't judged likely to change management. That decline has stopped being slow. Over the past fourteen months, four consecutive lab draws have tracked her eGFR down from 34 to 23, with her UACR rising alongside it to 480 mg/g — a consistent trend, not a single alarming number, and one that has continued despite losartan already titrated to the maximum tolerated dose.
EMPA-KIDNEY is the trial the team keeps returning to, because it is the one that actually enrolled patients this far down: unlike DAPA-CKD's eGFR floor of 25, EMPA-KIDNEY extended enrollment to eGFR 20, and 54% of its participants — a clear majority, and like Dana — had no diabetes at all. The relative risk reduction for the primary composite outcome held across eGFR subgroups down to that floor. What the trial's own numbers also show is that SGLT2 inhibition produces an early, expected drop in eGFR in the first few weeks, a hemodynamic effect of afferent arteriolar vasoconstriction rather than genuine nephron loss. The magnitude of that drop is where the argument actually sits, because the two numbers usually quoted are not the same number: the typical dip runs under 10%, which on Dana's 23 means a redraw somewhere near 21, while the greater than 30% decline that guidelines name is a threshold for stopping to investigate rather than an expected effect — in CREDENCE it occurred in 0.5% of treated patients. On her 23, though, that rare version lands near 16, and 16 is a number that changes what a clinic conversation is about regardless of the reversible mechanism behind it.
Nephrology clinic, reviewing the fourth consecutive decline
Start empagliflozin today. EMPA-KIDNEY enrolled patients down to an eGFR of 20 specifically to answer whether benefit persists this far into disease, and the relative risk reduction for the composite kidney-and-cardiovascular outcome held across every eGFR subgroup down to that floor — it did not attenuate meaningfully as baseline function fell. More than half of that trial's population — 54% — had no diabetes, which is exactly why it exists as a separate, later trial from DAPA-CKD: to test the drug in patients like Dana specifically, not just as an extension of diabetic-kidney-disease data.
If her eGFR had first crossed 25 last week, I wouldn't frame this as urgent — but it hasn't; it's been falling in a straight line for over a year, and every month spent confirming what's already confirmed is a month of RAASi-plus-SGLT2i combination benefit she doesn't get back.
I'm not disputing the subgroup data — EMPA-KIDNEY genuinely does support her population. What concerns me is what happens to Dana specifically in the first month. I want to be precise about the size of it, because the 30% figure gets quoted as though it were the expected dip and it isn't — it's the threshold at which you stop and look for another cause, and it's rare. The typical dip would take her from 23 to about 21, purely from afferent arteriolar vasoconstriction rather than any real loss of nephron mass. My concern is that she has less room underneath her than most patients this argument gets made about: even an ordinary dip puts her within a few points of where dialysis-access planning conversations start, and I'd want us prepared for what that number will look like before it appears, not surprised by it.
Trend data over 14 months is real evidence, I agree — my hesitation isn't about re-confirming the decline itself, it's about whether her remaining runway before RRT is long enough for a slower future decline to actually translate into meaningfully more dialysis-free time, against genuine new risks at her stage: hyperkalemia on top of losartan, genital mycotic infection, volume depletion.
I've drawn four of these labs myself over the past year, and I'd frame the dip differently: it isn't a new uncertainty layered onto an already-uncertain trend — it's a known, described, time-limited pharmacologic signature landing on top of a trend that was never actually in question. Delaying to watch one more redraw before starting doesn't get us better information than we already have; it gets us the same information a month later, on a drug that needs time on board to matter.
Start now, with an explicit plan: repeat basic metabolic panel at two and four weeks, expect and don't overreact to a drop into the high teens if her potassium and volume status stay where they are today, and reassess the real trajectory — not the first post-initiation number — against her own baseline curve.
Agreed: empagliflozin started today, losartan unchanged, with basic metabolic panel rechecked at two and four weeks and an explicit expectation set in the chart that an eGFR drop into the high teens on that recheck is anticipated, not a new event requiring the drug to be stopped.
Not fully agreed: the clinical pharmacologist's underlying question — how much dialysis-free time this actually buys someone already this far into disease — has no clean answer from EMPA-KIDNEY's own trial duration, which wasn't designed to resolve it at the individual-patient level. That uncertainty is being carried forward as a documented open question for her next visit, not treated as resolved by today's decision to start.