Why sleep sits in the Four Majors

It is 02:47 in a lot of houses. In some, a parent reaches before they are awake, looks at the line, decides whether to wait, treat, or hold their breath for ten more minutes. In others, the adult living with T1D reaches for their own phone, hears the partner stir and go still, and runs the same calculation. The line says the same thing in either case. By the morning, it tells the story: how long it took to come up, how much it overshot, how much sleep anyone in the house actually got. The body talks back honestly. The question is whether anyone got to listen.

For most of the last twenty years, sleep has been talked about in T1D as a wellness layer. Get more of it, try a sleep app, dim the screens. From the conversations I have in DAFNE, and from the doorway of my own home at 02:47, that framing has always undersold what is happening. The evidence we now have says sleep is metabolic. One short night drops peripheral insulin sensitivity by 14 to 21 percent in adults with T1D (Donga 2010, Diabetes Care). The fragmentation in T1D is not random; the alarms that protect against hypo are the same alarms that wake the house, and shorter, more broken sleep tracks with higher HbA1c and worse glycaemic variability across the pooled T1D literature.

The four parts

This guide is written to be read in order, but each part stands alone if you arrive needing one moment translated.

The week, not just the night

The Friday-night-to-Monday-morning shift is one of the most common patterns in the data. Bedtime drifts later across the weekend, the alarm comes earlier on Monday, and the Sunday-night-Monday-morning glucose trace looks different from the Tuesday-night-Wednesday-morning one. In our own house, Sunday evenings have always been the tell, and the Monday-morning lines tend to carry the difference into the afternoon. From the conversations I have in DAFNE, and from clinic visits with adolescents whose school start times are out of sync with their natural sleep timing, the shift is rarely about whether someone is sleeping enough hours total. It is about whether the hours land in the same window.

What the literature says

Across more than 100,000 adults followed in ten prospective cohort studies, both extremes of sleep duration raised the risk of incident type 2 diabetes, and difficulty maintaining sleep raised it more than short duration alone (Cappuccio 2010, Diabetes Care). The consensus from the American Academy of Sleep Medicine is that adults aged 18 to 60 should sleep at least seven hours per night on a regular basis. Beyond duration, the timing layer matters: the gap between work-day and free-day sleep timing (called “social jet lag”) is independently associated with insulin resistance and BMI, and chronic late chronotype tracks with worse glycaemic outcomes in T1D and T2D.

The T1D-specific picture

The T1D-specific synthesis arrives at the same direction with smaller effect sizes. Across the pooled T1D literature, shorter and more fragmented sleep tracks with higher HbA1c and worse glycaemic variability (Reutrakul 2016, Sleep Medicine). At the GNL platform, in approximately 500,000 patient-days from approximately 1,300 individuals living with T1D over more than ten years, the optimum sleep window for time-in-range was 6.5 to 7.5 hours; people sleeping less than 6.5 hours saw time-in-range fall by around 8 percentage points, and people whose sleep timing varied by more than 1.5 hours from night to night saw time-in-range fall by around 10.5 percentage points compared with the most regular sleepers.

Regularity beats duration

The point is not “everyone should sleep more”. Many readers cannot. The point is that the regularity of the sleep window is a lever, and that the lever is most powerful when AID is doing the overnight work and you can give the algorithm the same window every night to do its work in. For families, this maps onto bedtime routines that hold across the week, including weekends. For working adults, it maps onto a stable bedtime range plus or minus an hour, with the wake time as the anchor that travels well across days. For shift workers, the picture is harder; chronic shift work raises diabetes risk independently, and the question of whether shift patterns are negotiable or compensable is a conversation worth having with both an employer and a clinical team. On the population evidence the two levers are comparable; in the GNL Syno cohort the regularity gap edged duration by a few percentage points, so on this dataset regularity does win, but read it as “comparable plus a small lead” rather than a clean knockout.

Around eighty percent of the benefit from sleep regularity is captured by holding a stable window; the final twenty percent depends on chronotype, household, work, and biology that the literature has not yet measured well, so the picture will keep moving. If your data shows a clear weekend-vs-weekday split in your time-in-range, your CGM report is the most useful thing you can bring to your diabetes care team. They can review whether the pattern is sleep-driven, AID-overnight-target-driven, or both.

The morning after

You can feel it before the line shows it. The adult walking into work after four hours of sleep, knowing the lunchtime correction is going to behave differently from yesterday’s. The parent making breakfast for a child who slept badly, watching the post-meal rise climb to a number that did not climb yesterday. From the conversations I have in DAFNE and BERTIE, this is the most consistent observation people bring back to the room, and it is the observation that took the literature longest to catch up with.

For a long time the assumption was that sleep mattered for mood, weight, and general wellness, and that the metabolic effects, if any, were small and slow. The evidence we now have says the metabolic effect is fast and specific.

The trial spine

Donga’s group at Leiden ran the experiment that closed the question. One night of four hours’ sleep dropped peripheral insulin disposal by 14 to 21 percent the next day in adults with T1D (Donga 2010, Diabetes Care). The defect sat in muscle and fat, not the liver. One night.

The mechanism layer underneath that result is now well-mapped. Spiegel’s appetite axis, Cappuccio’s population direction across more than 100,000 adults, Cedernaes on the cellular signature in adipose and skeletal muscle: short or fragmented sleep is a metabolic event, not a wellness one. Grace carries the depth on each strand.

Why this matters in T1D

That picture changes how we should think about sleep in T1D. It is not a wellness layer; it is part of the substrate the rest of the system runs on. The 14 to 21 percent drop in next-day peripheral insulin sensitivity is the size of effect that, in the cohort GNL has assessed, shows up the day after as a stiffer post-meal rise and a more variable correction response. The hunger and appetite signals after a poor night are not a willpower problem; they are an endocrine output. In paediatric clinic the same biology explains why a child whose sleep has been broken by ear infections or by stress at school can suddenly run higher for a week, even when nothing has changed about their dose, their carb-counting, or their illness. Children and adolescents need more sleep than adults; AASM and ISPAD ranges sit higher by age band, and this is a useful conversation to have with the diabetes team alongside any insulin titration.

Why your morning glucose is higher than your overnight low

You go to bed at 6.5 mmol/L (117 mg/dL). You wake at 4am to a CGM line that has drifted up to 8.0 mmol/L (144 mg/dL), with no food, no correction, no obvious explanation. By 7am it is at 9.5 mmol/L (171 mg/dL), and the pre-breakfast bolus has more work to do than yesterday’s. From the conversations I have in DAFNE and BERTIE, and from the same conversation at home with my own daughter, this is the moment people most often blame on themselves: the late-evening snack, the basal that “must be wrong”, the discipline that “must have slipped”. It is rarely any of those things. The body is doing exactly what bodies do at dawn. The morning rise is not a personal failing.

The mechanism, in plain terms

The dawn phenomenon was first described by Schmidt and colleagues in 1981 as a night-to-morning rise in blood glucose without any food intake. Bolli and Gerich confirmed in 1984 that the same phenomenon occurred in both T1D and T2D, driven primarily by nocturnal growth-hormone pulses around 03:00 to 06:00 that accelerate hepatic glucose production and reduce peripheral insulin sensitivity (Porcellati and colleagues, 2013, Diabetes Care 30-year review). The healthy non-diabetic body pre-empts this with a small anticipatory insulin pulse just before dawn. T1D, by definition, cannot do that. T2D, in the early stages, can do less of it. The visible rise is the gap between what the body needs and what insulin replacement, on its current settings, can deliver.

How big is the rise, really

The size of that gap depends largely on overnight insulin. With older NPH or lente insulins that wane overnight, the apparent dawn rise can look like 3 to 5 mmol/L (50 to 100 mg/dL). With continuous subcutaneous insulin infusion at a single basal rate, or with the modern long-acting analogues (glargine U100, glargine U300, degludec), the true T1D dawn-phenomenon magnitude settles at around 1.4 to 1.7 mmol/L (25 to 30 mg/dL) pre-breakfast, rising to around 2.8 to 3.3 mmol/L (50 to 60 mg/dL) post-breakfast (Porcellati 2013, Diabetes Care). At the population scale, in non-insulin-treated T2D, the dawn phenomenon contributes around 0.4 percentage points to HbA1c on its own, and no oral antidiabetes agent in current use eliminates it.

The lever that moves it

The lever that most consistently moves the dawn rise is overnight insulin programming. On multiple daily injections, that conversation is about the timing and dose of basal insulin (often glargine U300 or degludec). On pump therapy, that conversation is about the overnight basal-rate profile: many people benefit from a programmed basal-rate increase running roughly 03:00 to 06:00. On automated insulin delivery, the system is doing this work every night, and many people benefit from a tighter overnight target. The MiniMed 780G AHCL pivotal trial ran a randomised within-person crossover between a 5.5 mmol/L (100 mg/dL) target and a 6.7 mmol/L (120 mg/dL) target across roughly 90 days each; the lower target produced higher time-in-range, with a modest rise in time below 3.9 mmol/L (Carlson 2022, Diabetes Technology and Therapeutics). Overnight target tightening is the AID lever this finding maps onto.

When the algorithm reaches first

The alarm goes at 02:47. Someone in the house wakes, but only just. The treatment box is in reach because everyone knows where it lives. From the doorway of my own home and from the conversations I have in clinic, the work that families and adults living with T1D are doing overnight is one of the most under-acknowledged forms of healthcare labour in the country. It is invisible in the morning, and the morning glucose number rarely captures it.

The pre-AID benchmark

Before continuous glucose monitoring and before automated insulin delivery, this work was uncountable. It is now countable. The Juvenile Diabetes Research Foundation’s CGM study group analysed 36,467 nights of unblinded sensor data in 176 children and adults with T1D on intensive treatment. Hypoglycaemia events occurred on 8.5 percent of nights, with a median duration of 53 minutes; on 23 percent of those nights, glucose stayed below 3.3 mmol/L (60 mg/dL) for at least two hours (JDRF CGM Study Group 2010, Diabetes Care). That is the pre-AID benchmark. Lower baseline HbA1c and any hypo during a baseline blinded CGM week were the strongest predictors of how often nocturnal hypo recurred. Pump or multiple daily injections did not change the rate.

What the algorithms changed

The algorithms changed where this hypo could be intercepted. Buckingham’s predictive low-glucose suspend trial in 81 children with T1D ran 3,420 randomised in-home crossover nights. On suspend-active nights, time below 3.9 mmol/L (70 mg/dL) overnight halved across both age groups, with no rise in morning ketones and no severe hypo or DKA across the trial (Buckingham 2015, Diabetes Care).

Modern advanced hybrid closed-loop has gone further. The MiniMed 780G pivotal trial in 157 adolescents and adults saw HbA1c fall from 7.5 to 7.0 percent, time-in-range rise from 68.8 to 74.5 percent, and time below 3.9 mmol/L (70 mg/dL) fall from 3.3 to 2.3 percent, with closed-loop time averaging 94.9 percent and zero severe hypos or DKA across the study phase (Carlson 2022, Diabetes Technology and Therapeutics). The same direction repeats across thirty-one papers on the MiniMed 780G in children, adolescents, and young adults.

The bidirectional loop, and what AID does not solve

What AID does not do is reach the people for whom it is not yet available. Provision in the UK is uneven; eligibility under NICE TA943 is real but slow; many families and adults are still on multiple daily injections or sensor-augmented pump without the closed-loop layer. Sleep is more fragmented in T1D than in non-diabetic peers, even on the most modern technology, because the alarms that protect against hypo are the same alarms that wake the house (Reutrakul 2016, Sleep Medicine). The bidirectional loop is real: poor glucose breaks sleep, broken sleep makes the next day’s glucose harder, which breaks the next night.

The other thing the night can be hiding

Obstructive sleep apnoea is more common in adult T1D than in non-diabetic peers (Banghoej and colleagues, 2017, Journal of Diabetes and its Complications, found around forty-six percent of a Danish adult T1D cohort had at least mild OSA on home polygraphy, with cardiac autonomic neuropathy the strongest predictor). The signs are loud snoring, witnessed pauses in breathing, daytime sleepiness, and unexplained morning rises that the algorithm cannot explain. If any of those describe a night in your house, that is a clinic conversation; STOP-BANG screening and an overnight study are the next steps your team can arrange.