An Off-Label Request in Type 1 Diabetes, and a History That Changes the Answer
The drug class carries a real, trial-confirmed ketoacidosis risk in type 1 diabetes — but her own chart holds a detail that determines whether that risk is abstract or concentrated, and nobody has asked her about it yet.
Priya N., a 27-year-old woman, keeps four beehives in her apartment building's shared courtyard, a hobby she took up during a rough stretch after her diagnosis and now treats with the same disciplined attention she brings to her insulin pump settings. She has had type 1 diabetes since age 14, is on a hybrid closed-loop pump system with continuous glucose monitoring, and her A1c has run a reasonable 7.4% for the past year — but she's gained nearly nine kilograms since starting intensive insulin therapy in her early twenties, and her glucose trace shows real day-to-day variability she describes as exhausting to manage around her work schedule as a courtroom stenographer. She read about SGLT2 inhibitors online, brought it up directly, and wants to know why it isn't simply part of her regimen already.
What her chart also shows, and what she doesn't raise unprompted, is two documented episodes in the past eighteen months where her care team noted intentional under-dosing of mealtime insulin during particularly demanding trial weeks — a pattern she described at the time as "just cutting corners," not restrictive eating, but one her endocrinologist flagged for follow-up regardless. That history matters directly to the drug she's asking for: SGLT2 inhibitors lower glucose independent of insulin, which means a patient who is intermittently running relatively insulin-deficient can look reassuringly normal on a glucose meter while ketones climb underneath — the defining, dangerous feature of euglycemic diabetic ketoacidosis, and the reason this drug class carries a explicit labeled warning against use in type 1 diabetes that it does not carry in type 2.
Weighing a real benefit against a risk that hides behind a normal number
I'd decline this, at least today. DEPICT and inTandem, the two major trials of SGLT2 inhibitors added to insulin in type 1 diabetes, both found several-fold increases in diabetic ketoacidosis — and a meaningful share of it was euglycemic, meaning her glucose meter would look reassuring right up until it wasn't. No SGLT2 inhibitor carries a US type 1 indication. What she's describing is real and frustrating, but it's a quality-of-life problem, and DKA is not a symptom I'm willing to trade a quality-of-life gain for.
I'd read those same trials less as a blanket warning and more as a risk that's genuinely modifiable. The DKA signal wasn't spread evenly across every enrolled patient — it concentrated around insulin under-dosing, reduced carbohydrate intake, and intercurrent illness. Structured mitigation — ketone meter training, explicit sick-day rules, starting at the lowest studied dose — changes an individual patient's actual risk, not just the population average the trial reports.
I'm not saying dismiss the DKA data; I'm saying the data itself points to exactly which patients carry the concentrated risk, which means it's answerable rather than unknowable — we just need to know whether she's one of them.
She is, or she might be, and that's the part neither of you has said out loud yet. Her chart has two documented episodes of intentional insulin under-dosing in the past eighteen months — exactly the substrate that turns "modifiable risk" into a real one. Adding a drug that lowers glucose independent of insulin, in a patient who has already shown a pattern of running relatively insulin-deficient by choice under stress, isn't the same decision as offering it to a type 1 patient without that history.
This doesn't have to be a flat no, and I don't think it should be — but it has to be a different conversation first. Before we talk about SGLT2 inhibitors, I want an honest conversation with her about those two episodes, what was actually happening during those trial weeks, and whether that pattern is still active. That conversation changes what a safe answer even looks like.
Agreed: SGLT2i decision deferred to a dedicated follow-up visit, behavioral health referral placed, and Priya given an honest, direct explanation of why the two documented episodes changed the conversation rather than being treated as settled history. She was not upset by this — by her own account, she'd wondered if it would come up.
Not agreed: what threshold of reassurance at follow-up would make the endocrinologist comfortable prescribing. The pharmacologist and primary care physician both indicated they'd be satisfied by a clean behavioral-health assessment and several weeks of unremarkable CGM data; the endocrinologist wanted a longer observation window given the trial safety signal, a difference not resolved at this visit.