Preventing Hypoglycaemia: Basal Review, IOB-Aware Corrections, and Sick-Day Rules

Four levers, in the order they tend to bite

The four most common drivers of recurrent hypoglycaemia tend to surface in a recognisable order across years of clinic conversation. The persistent overnight low usually means basal is too high, not the bedtime snack is too small. The post-meal stack means a correction was layered on top of insulin still working. The sick-day low means a panicked insulin reduction crashed the system at the wrong moment. The exercise hypo means the late-drop window after sport was missed. None of these are failures; they are patterns the body shows up with, and each has a conversation with the diabetes team that names what to look at first.

Lever 1: Basal review, the overnight pattern

The most common prevention conversation in adult clinic, and the one that takes longest to land, is the basal review for the persistent overnight low. The pattern is recognisable: a flat or slowly-descending CGM line from midnight to 4am, hitting 3.5 to 3.7 mmol/L (63 to 67 mg/dL), recovering as dawn cortisol comes in, a wake-up at 5.8 to 6.5 with no memory of anything happening overnight. The instinct is to add a bedtime snack. The basal-led answer is that the snack is treating a symptom of an over-delivered basal, and the snack tends to set up the next morning’s high.

The trial spine that supports the basal-led answer comes from the algorithmic protection literature. ASPIRE In-Home (Bergenstal 2013, NEJM) showed that simply suspending insulin delivery overnight when glucose hit a threshold cut the area under the nocturnal hypo curve by around 38 percent, with HbA1c essentially unchanged. The implication is direct: the overnight insulin was doing more work than the body needed for stretches of the night. The clinic conversation that follows from this is to bring two weeks of overnight CGM reports to the next appointment and ask the team to look at the basal pattern first, before any conversation about snacks. On a pump or AID system, the basal rate or the algorithm’s overnight target is the lever; on long-acting insulin, the bedtime dose is the lever and the team’s adjustment is in 10 to 15 percent steps over a few weeks.

Lever 2: IOB-aware corrections, the post-meal stack

The post-correction hypo, almost without exception in adult clinic, traces back to a correction layered on top of insulin that was still working. The technical name is insulin stacking. The mechanism is that bolus insulin keeps acting for three to four hours after delivery (longer on some systems, depending on the algorithm’s insulin action time setting); a glucose reading two hours after a meal that still looks high is, ninety percent of the time, simply showing the meal absorption running ahead of the insulin curve, which would have closed the gap in the next ninety minutes without any further dose. A correction at that two-hour mark adds insulin to insulin still on board, and the resulting low arrives at the four-hour mark.

The lever is the insulin-on-board number itself, surfaced on every modern pump and on most bolus-calculator apps. Every GNL Insulin On Board guide page (the cluster hub at /insulin-on-board-t1d/) treats the IOB read as the gate before any correction. The protocol is to either wait until the IOB on screen has dropped to a value the team has agreed (often below 1 to 2 units, depending on body weight and insulin sensitivity), or to halve the calculated correction. The conversation in clinic is to agree the wait window, not to memorise a number off the internet. On AID systems, the algorithm does much of this work in the background; on hybrid closed-loop where the user can still bolus on top, the IOB read still matters, just as it does on a standard pump. The correction-dose figures the GNL platform surfaces in the Hyper Treatment Explorer are population-average estimates at a user’s TDD, not personalised doses; people apply their own correction factor agreed with their own diabetes team.

Lever 3: Sick-day rules, where insulin is never stopped

The sick-day hypo usually arrives because the basic rule was reversed. The rule is that insulin is never stopped during illness, even when the person is not eating; the dose may need adjusting, but the insulin keeps flowing. Stopping basal entirely because a child is off their food is the route to ketones and DKA, not to hypo prevention. The right move is the opposite: keep basal going, test blood ketones if glucose is high or the person is unwell, and adjust the next bolus by ketone band rather than by appetite. The international anchor for paediatric sick-day management is ISPAD 2024 Chapter 13; the UK adult anchor is the Diabetes UK adult sick-day guidance.

Where sick-day rules do drive hypo is at the other end of the curve: the family who has bolused for a meal the child then refuses, or the person on AID whose system has been ramping up automatic corrections through a viral illness only for the appetite to come back suddenly. In the first case the response is fast carbohydrate matched to the bolus already given; in the second case the conversation with the diabetes team is about whether the activity-mode setting on the AID system needs flipping during the recovery phase. A written sick-day plan, agreed with the team in advance and kept somewhere everyone in the household knows about, is the prevention layer that matters most. Asking for one before the next bug arrives is the practical close.

Lever 4: Exercise timing, the two windows

Exercise produces hypoglycaemia in two distinct windows, and the second is the one most people miss. The first window is the immediate one, zero to two hours after the activity, when working muscle is pulling glucose out of the blood at an elevated rate and any insulin on board is being potentiated by the muscle’s heightened sensitivity. The second window is the late one, six to twelve hours after the activity, often arriving overnight as a slow descent through the small hours. The DirecNet paediatric exercise study (Tsalikian 2006, Journal of Pediatrics) ran the controlled comparison: stopping basal during 75 minutes of afternoon aerobic exercise dropped the hypo rate from 43 percent to 16 percent during and shortly after the activity. The late-window evidence comes from the bedtime-snack and overnight-carb studies (Kalergis 2003, Campbell 2015), which show that the muscle’s glycogen replenishment work continues for hours after the gym door has closed.

The lever for the immediate window is either a pre-exercise carbohydrate top-up or a basal reduction (or both); the choice depends on the activity type, the body weight, and the IOB on board at the start. The GNL Exercise Planner explorer surfaces population-average carbohydrate suggestions per body weight band, capped at 60 kg of body weight to match the platform’s carb-cap rule. The lever for the late window is usually a smaller meal-time bolus at the post-exercise meal, or a temporary basal reduction overnight on a pump or AID system; the diabetes team is the route to the personal numbers. A useful clinic ask is for a written exercise plan that names both windows, because the late-window plan is the one most adults with T1D have never been given explicitly.

The AID-era picture, briefly

The Karges 2024 DPV registry (Lancet Diabetes Endocrinol) is the largest contemporary real-world dataset on AID safety in young people, with 13,922 children and adolescents tracked across hybrid closed-loop and open-loop sensor-augmented pump care. Hypoglycaemic coma dropped by around a third on hybrid closed-loop (incidence rate ratio 0.68, 95 percent CI 0.48 to 0.97); overall severe hypoglycaemia rates were essentially unchanged. The CLOuD 48-month extension (Ware 2024, Diabetes Care) showed the same pattern holds at four years of use. The implication for prevention is that AID does most of the work for the basal-review and the IOB-aware-correction levers in the background, and most of the remaining hypo prevention conversation in clinic is about exercise timing and sick-day rules. The technology has not eliminated the conversation; it has narrowed it.