SGLT2 Inhibitor Timing in Decompensated Heart Failure: Starting It Before He's Eating Again
He qualifies for the drug his guidelines call the fourth pillar of his regimen, and the visit that could finally start it may be the best chance he gets. He also hasn't kept down half a meal tray in three days — the exact circumstance under which this drug has been linked to a rare, easy-to-miss complication.
W.C., a 63-year-old man who still drives a produce delivery route three mornings a week, noticed his work boots fitting tighter about ten days before he came in — the kind of ordinary-seeming change he'd shrugged off twice before, and this time couldn't, once climbing two flights to a walk-up delivery left him stopping on the landing to catch his breath. He has ischemic cardiomyopathy from an anterior MI four years ago, treated then with a stent to his LAD, and has carried an ejection fraction in the low 30s since; his outpatient regimen — carvedilol, sacubitril/valsartan, spironolactone — had kept him out of the hospital until now. Type 2 diabetes has been part of his history for just over a decade, managed on metformin and glipizide with an A1c that has run in the low 7s, not tightly controlled but not neglected either. He presented with a ten-pound weight gain, orthopnea requiring three pillows, and bibasilar crackles to the mid-lung fields; admission labs showed a BNP sharply above his last outpatient value and a creatinine only marginally above his own norm.
Three days of IV furosemide have done real work — four kilograms off, crackles down to the bases, jugular venous pressure no longer visible above the clavicle at 45 degrees — but he isn't through the door yet: trace edema persists at both ankles, and he's still on the drip rather than converted to an oral dose. The harder problem, and the one actually dividing the team, is what to do about the fourth pillar of his heart-failure regimen. He has never been started on an SGLT2 inhibitor, and by every major guideline he now qualifies for one; the question is whether today, hospital day three, is the day to start it. He is also eating less than half of what's on his tray — nausea he attributes to "hospital food," though the team suspects at least some of it is the same gut congestion that produced his tight boots. That detail turns out to matter more than it looks like it should: reduced oral intake is one of the specific, named circumstances under which this drug class has been linked to a rare but serious complication — ketoacidosis without the usual warning sign of a high blood sugar to make anyone suspect it.
At the bedside, hospital day three
Start it today. He is close to the patient EMPULSE was built around — hospitalized for acute heart failure, randomized a median of three days into admission once clinically stable — and empagliflozin produced a clinical-benefit win ratio of 1.36 against placebo in that trial, with part of that advantage showing up early, driven partly by decongestion the drug adds on top of whatever the loop diuretic is already doing. He's three days in, stabilizing, and not on insulin. This isn't the fragile first-hour patient.
There's a practical argument that matters as much as the mechanistic one. If this doesn't get written today, the real odds it gets started at all drop substantially. Outpatient follow-up after a heart-failure discharge is inconsistent, and a fourth-pillar drug that never gets added because everyone assumed "the next visit" isn't a safer plan — it's a plan that fails quietly.
The EMPULSE population and this patient aren't the same case, and the difference is exactly what matters here. He's eaten less than half his meals for three days running, in a man with type 2 diabetes on a sulfonylurea. That's not a hypothetical risk profile — reduced oral intake is one of the specific precipitants named across the euglycemic diabetic ketoacidosis, or eDKA, case literature, and the mechanism is direct: this drug lowers insulin secretion by lowering glucose independent of carbohydrate intake, and when intake is already down, that combination can tip a well-compensated diabetic into eDKA without ever producing the high blood sugar that would normally raise suspicion.
And I'd rather lean on the larger trial than EMPULSE alone. DAPA-ACT HF-TIMI 68 — more than four times EMPULSE's enrollment — didn't meet its own primary endpoint on dapagliflozin against placebo, and it recorded more symptomatic hypotension and more worsening kidney function on the drug than on placebo. That's not a reason to abandon the class. It's a reason not to treat "in-hospital" as one uniformly safe window regardless of what else is happening to this particular patient right now.
I don't think this is actually a choice between today and "whenever outpatient follow-up happens" — that sets up the two least useful options and picks between them. Hold it until he's reliably taking in at least half his meals for a full day, off the drip, and check a point-of-care ketone level before the first dose and again the next morning. If today is day three and he's still under fifty percent, that's tomorrow's decision, not next month's. And before he leaves, someone needs to spend five minutes teaching him the same rule surgeons already give patients on this drug before procedures — hold it if you're sick, vomiting, or not eating for more than a day, restart once you are. That's the piece that actually protects him after this admission, not just during it.
Agreed: not started today. Empagliflozin will be started once he sustains at least half his meal intake for a full 24 hours off the IV diuretic drip, with a point-of-care ketone check before the first dose and a second check the following morning. Glipizide stays held until intake recovers. Discharge teaching will include explicit sick-day guidance — hold the SGLT2 inhibitor during vomiting, prolonged poor intake, or acute illness, and restart once eating normally.
Not agreed, and the reason the plan carries a genuine open question rather than a settled one:
He starts empagliflozin in-house, under monitoring, exactly as the cardiologist wanted — just a day or two later than "today."
The threshold-based plan quietly becomes the outpatient-dependent prescription the cardiologist warned against, and the team has no agreed answer for how hard to push a same-week follow-up to make sure it still happens.
Cardiology and pharmacology left the room with different confidence in how that split will resolve. Nobody set a formal deadline; the endocrinologist's threshold stands as the working rule, revisited at tomorrow's rounds regardless of which way his appetite has moved.