Mealtime Insulin: Predictability, not Perfection

Why mealtime insulin is structurally hard

If mealtime feels harder than basal or correction work, that is not a skill failure. It is a structural one. In people without T1D, insulin is released straight into the portal vein, reaching the liver first and damping the post-meal rise before it ever shows up in the bloodstream. Injected or pumped insulin does not get that route. It sits in the subcutaneous tissue and absorbs slowly, reaching the liver later and at lower concentration than the body would have managed on its own. So even when the dose is right, timing and meal composition can produce spikes, late drops, or both.

The teaching I lean on most heavily here, in DAFNE clinics and in conversations with families at Birmingham Women’s and Children’s, is that the carb-ratio you have been given is doing more work than people realise. It was tuned on usual mixed meals; it already covers the meal-pattern most people eat most of the week. When the meal stays inside that pattern, the ratio holds. When the meal steps outside it, the ratio is not broken. The meal has changed.

The 130 percent ICR insight

The single most useful number in mealtime dosing is one that almost no one has been told. The insulin-to-carbohydrate ratio set with your team is a mixed-meal ratio, not a carbohydrate-only ratio. Worked out on usual meals, around 40 to 55 percent carbohydrate by energy with 10 to 20 percent protein and 20 to 40 percent fat, it covers approximately 130 percent of the insulin a carbohydrate-only meal would need (Pemberton 2021, Pediatric Diabetes). That single observation explains three things at once. Fruit-only meals matched gram-for-gram with the ratio go low afterwards. Pizza, fry-ups, and cheesy pasta run high for hours despite perfect counting. Usual mixed meals work fine, because that is what the ratio was designed for.

The practical implication is not that the ratio is wrong. It is that the ratio is calibrated to a window. Carb-only meals undershoot the window, so they need less insulin. Fat-and-protein-heavy meals overshoot it, so they need more insulin and a different dose shape. Most meals fall in the middle, and the ratio looks after them.

Pre-bolus timing, the lever you can move

Of every single move you can make at mealtime, the cheapest one to start with is also the most reliable. Bolusing around 20 minutes before the first bite of a usual mixed meal, on a standard rapid analogue, is the dose-timing change that most consistently lowers the post-meal spike (Slattery 2018, Diabetic Medicine). The mechanism is the timing mismatch the previous section described: the food arrives in minutes; the analogue insulin starts in tens of minutes. Closing the gap before the meal starts is the closest you can get to giving the insulin a fair chance.

Two practical points the figure does not say loud enough. First, the optimal pre-bolus window is not 20 minutes for everyone, every meal, every starting glucose. It is 20 minutes for a typical adult with rapid analogue insulin (Humalog, NovoRapid, Apidra) starting from a glucose in the target range. If the starting glucose is low, the window shrinks; if the starting glucose is high, it lengthens. Second, the ultra-rapid analogues (Lyumjev, Fiasp) absorb faster from the depot than the older rapids (Leohr 2021, Clinical Pharmacokinetics), so they need a shorter pre-bolus, sometimes none at all. The principle holds, the number changes with the analogue.

The fat-and-protein meal, the curve that does not behave

The lasagne, the cheese-laden pasta, the takeaway curry, the slow-cooked Sunday roast. These meals are where the standard ratio and the standard pre-bolus timing both miss, in the same way, every time. Glucose rises later, lower, and longer. The first hour after eating looks fine. The first two hours look fine. Then around hour three the line starts rising and does not come back down for another two or three hours. People who have lived with this know the pattern; people who have not can spend years correcting the second-half rise and finding the corrections themselves do not land cleanly either.

The dose response that fits this curve has two parts. There is more insulin overall, because fat and protein contribute to glucose late in the digestive process; and there is a different dose shape, because a single up-front bolus would land long before the second peak arrives. On a pump, that is a split bolus, often around 50 percent up front and 50 percent extended over two to three hours. On injections, that is a smaller dose at the start and a top-up around hour two. The percentages from Pemberton 2021 give a population-average starting point: roughly plus 30 to 35 percent on top of the usual carb-only insulin for a high-fat-and-carb meal, and roughly plus 50 percent for a high-fat-and-protein-and-carb meal. Variability across people is genuinely wide, minus 17 to plus 124 percent in the source studies, so these numbers are starting points to refine with CGM and the diabetes care team, never instructions to follow.

The dosing matrix, in plain English

Five meal patterns cover most of what comes off a real kitchen counter. The same carb-ratio you have been given handles the middle three; the carb-only and the high-fat-and-protein cases are where the adjustment lever moves. Tap each pattern for the practical detail.

The matrix gives a population-average starting point. The actual numbers belong to you, your insulin, your activity, and the meal in front of you. Consistency is what makes these numbers reliable. The more your meals stay inside the usual mixed-meal pattern most of the week, the more the ratio holds when you ask it to. Variability in the meal is variability in the dose.

What I do at home, what we teach in DAFNE

I live with T1D myself. The first move that mattered for me was the pre-bolus on breakfast: 20 minutes ahead of the porridge bowl, on the train, took the morning spike from routine to occasional. Nothing about the dose changed, only the timing. The second move was the split bolus on lasagne; around plus 40 percent on the carb dose, half up front and half over two and a half hours, is what landed for me. The principle is general, the numbers are mine.

What we teach in structured education (DAFNE, BERTIE-online, the BWC DYNAMIC and GAME programmes) is the same shape: count the carbs accurately, pre-bolus on the rapid analogue, watch the CGM, refine with your team.

Strategies that lower the post-meal spike

The pre-bolus is the headline. Half a dozen smaller moves stack with it and add up to a meaningfully calmer post-meal hour. None of them require new equipment; all of them are clinic conversations worth having if you have not already.

What changes on AID, and what does not

Hybrid closed-loop systems take over the basal layer and the small in-between corrections, and they leave the meal bolus largely with you. The carb count, the pre-bolus timing, and the dose-shape decision for fat-and-protein meals are still yours. The places AID struggles most are the places your input matters most: the high-fat-and-protein meal that asks for an extended bolus the algorithm cannot infer from carbs alone, and the underestimated carb count that runs the post-meal high for hours. The systems quieten the rest of the day so there is more attention left for the meals.

References

The three meal patterns, in depth

The hub gives the framework. The three parts that follow take each meal pattern in turn, with the keystone diagrams that have been in use across the GNL community for years. Read in sequence if it is the first time through; otherwise, jump to the part that matches the meal in front of you tonight.