Blog
Sleep Is a Diabetes Treatment. Treat It Like One.
We have a habit of sorting health into separate drawers. Sleep goes in the wellness drawer with hydration and step counts. Diabetes goes in the medical drawer with insulin, A1c, and clinic visits. For people who take insulin, those two drawers are one and the same. Sleep is not adjacent to diabetes. It is part of it. The evidence says so, and it's time the tools did too.
The evidence runs both directions
This is a relationship, not a one-way street, and the honest framing is that it's associational and bidirectional rather than cleanly causal in both directions. Most of the studies are small, so I'll state the samples.
Poor sleep makes the morning harder. Across 81 people with type 2 diabetes, poor sleep quality was associated with a substantially larger dawn phenomenon (26.5 ± 13.1 vs 14.4 ± 12.8 mg/dL, P=0.001), a far higher prevalence of it (78.4% vs 18.2%), and higher A1c (7.3% vs 7.0%, P=0.032), with sleep quality and dawn magnitude correlated at r=0.494 (Huang Y et al., Int J Endocrinol 2017;2017:4578973).
Unstable glucose makes for unstable sleep.
In 22 adults with type 1 diabetes over 243 nights, the variability of sleep efficiency tracked the variability of mean glucose almost in lockstep (r=0.8), and nights with disordered sleep structure were associated with lower time in range and more hypoglycemia (Botella-Serrano M et al., Front Endocrinol 2023;14:998881).
And sleep changes how the body uses insulin. The literature on sleep and insulin sensitivity is consistent in direction: restrict sleep and insulin sensitivity falls; extend it and sensitivity tends to improve (see Reutrakul and Van Cauter; Diabetes Therapy 2015, DOI 10.1007/s13300-015-0141-z). I'm citing the review rather than any single number, because these vary and the direction is the durable part.
None of this proves a clean causal loop in both directions. But it repeatedly establishes, across study types, that sleep and glucose are entangled.
You cannot treat one well while ignoring the other.
People With Diabetes already know this
They don't need the citations. They live it. In our 2026 research, the patients we call High-Concern Strugglers are the most likely to use sleep as the yardstick for how their diabetes is doing — when they ask themselves whether they're managing well, sleep is the answer they reach for (DQ&A PWD Segmentation and JTBD Report, January 2026, Slides 7 and 56). And across all patient groups, the number-one overnight priority is avoiding a nighttime low, with a preferred wake-up glucose of around 100 mg/dL (Slide 59).
So patients are already doing the thing the drawers tell them not to do. They are treating sleep as a signal of diabetes. They judge their condition by how they slept and when they woke up.
The lived experience is ahead of the system.
Treating it like a treatment
Here is the gap, and it's the one that matters: patients treat sleep as a signal. Nothing yet treats it as something you can act on. You can measure your sleep. You can measure your overnight glucose. You can know, with real precision, that last night was bad and why. And then the night is over, and you repeat the experience.
Approaching sleep like a diabetes treatment means doing something in the window, not just reporting on it afterward. That is what sleep-only automation is for — automated insulin delivery that works while you sleep and steps back when you wake, built for people on pens who manage their days well and have never had anything to protect their nights.
I'll end on the discipline this whole arc has tried to hold: there is no published evidence that Luna improves sleep, or deep sleep, or next-day insulin sensitivity. That is the biggest thing I refuse to claim, and I won't, because we haven't earned it. What I'll claim is the shape of the problem. Sleep and glucose are one system. Patients already treat sleep as a sign of diabetes. The aim of what we're building is to make it treatable.
Sleep is a diabetes treatment. It's time the tools treated it like one.
Nights protected. Days yours.