Nutrition and type 1 diabetes:
The Plate, the Count, the Curve
Everything in one place. Read the plain version with Jude, earn your way into the evidence with Grace, then the full picture with John. Food is not the enemy. Stop wherever you have enough.
How we teach: three rules, borrowed from Taleb
You earn each level by showing you understand it, not by scrolling past it. We only teach what we would use on ourselves and the people we love.
Understanding beats memory and luck, so the checks reshuffle every time you retry. A pass means you got it, not that you guessed it. And we teach you to tell a trend (signal) from one reading (noise).
We give you the scaffolding and get out of your way. Roam where your curiosity leads, go as deep as you want, and ask Grace anything. We will not teach a bird how to fly.
Want to understand your own plates and your own meals, in your units? Ask Grace, then take it to your care team and your dietitian.
One page, three depths
This guide compounds: each layer rests on the one beneath it. Read Jude’s plain version, then pass a short understanding check to open Grace, then another to open John. You can roam freely within a layer; you cannot skip ahead a layer, because the next one would not make sense and you would be standing on a gap.
The whole thing, in plain words
Food is not the enemy. It is the partner that insulin works with, and the easiest way to make the partnership work is to make the food predictable. A reliable plate beats a perfect one. The balanced plate is the default: half the plate vegetables, a quarter carbohydrate, a quarter protein. When the food repeats, the glucose response repeats, and the insulin decisions get easier. The plate is not for the day you cook from scratch with a scale on the counter; it is for the ordinary day when you do not.
The three parts of a meal do different things to your glucose, and it helps to know which is which. Carbohydrate is the part that lifts glucose fastest, so it is the part that needs counting. Protein and fat lift glucose too, but slower and later, in the second half of the meal, and you can taste their effect most on a creamy curry or a slice of pizza that keeps you raised for hours.
Counting carbohydrate is the core skill, and the good news is that it does not have to be perfect to work well. Getting within about 10 g of the true amount is good enough for most meals; chasing the exact gram is effort that does not pay off. The dose itself comes from a ratio, a personal number agreed with your care team that says how many grams of carbohydrate one unit of insulin covers for you. None of this is personal advice; your ratio, your timing and the shape of your plate are set with your diabetes care team and your dietitian.
A default, not a rule. Birthdays and restaurants do not bend to it; the plate is for the ordinary days, so you can spend your wiggle room on the days that ask for it. The carbohydrate quarter is the part that needs counting.
Does this match the life of the person living it? The kitchen at 6pm is a place of tiredness and changed plans, not a laboratory. A reliable plate and a count that is close enough are kinder, and they hold up on the days a perfect rule would not survive. Food that brings guilt is harder to live with than food that brings a number you can read and learn from.The Pemberton lens, lived recognisability, one of the four GNL appraisal lenses.
The numbers underneath
How accurate counting needs to be
Counting carbohydrate at home does not need to be perfect. Smart 2009, a paediatric crossover trial, gave the same insulin dose for meals of 50, 60 and 70 g of carbohydrate and found no meaningful difference in glucose over 2.5 hours when the count was within roughly 10 g of the target.1 A later analysis found the failure boundary: a 20 g error produced a meaningfully higher chance of a hypo when the dose was calibrated for the larger meal.1 In adults, real-world counting error sits at around 15 g per meal (about 21 percent), with a consistent pull towards under-counting; 63 percent of meals were under-estimated (Brazeau 2013), and the larger the error, the wider the glucose swings.2 The honest target is within 10 g if you can, with 20 g as the boundary you do not want to cross; then refine against what the CGM actually shows. Structured education is the evidence base under all of this: the DAFNE programme improved HbA1c by about 1.0 percentage point in 169 adults (DAFNE 2002).3
| Counting accuracy | Population finding | Source |
|---|---|---|
| Within 10 g | No meaningful glucose difference at most meals | Smart 2009 (paediatric) |
| Around 20 g out | Meaningfully higher hypo risk if dosed for the larger meal | Smart 2009 / 2012 (paediatric) |
| Adult real-world | About 15 g (21 percent) error per meal, 63 percent under-counted | Brazeau 2013 (adult) |
The meals that change the curve: fat and protein
Carbohydrate is not the whole story. Fat and protein add their own, later glucose rise, from around 2 hours out to 5 hours. A systematic review of 31 studies (Bell 2015) found fat, protein and glycaemic index all meaningfully change post-meal glucose in type 1 diabetes.4 In a closed-loop study a high-fat dinner needed about 42 percent more insulin than a low-fat one at identical carbohydrate (Wolpert 2013).4 A four-fat dose-finding trial showed the effect grows with the fat: about 20 to 40 g of added fat needed only roughly 6 percent more insulin, while 60 g needed about 21 percent, and the fat lowered glucose in the first 2 hours before raising it in the late window (Bell 2020).4 A high-fat, high-protein meal is the one that moves the curve most: on average about 65 percent more insulin (range 17 to 124 percent), delivered as a split over roughly two and a half hours (Bell 2016).4 The population-level response is to spread the dose so more insulin lands late, an effort that is a care-team conversation, not a fixed recipe.
Illustrative curve shapes, not measured patient data. The point is the late tail: a high-fat, high-protein meal keeps glucose up after an ordinary meal has already come down. Glucose axis in mmol/L (mg/dL); direction is reliable, the size is personal.
Carbohydrate quality: glycaemic index and load
Two meals can carry the same grams of carbohydrate and still behave differently, because how fast the carbohydrate arrives matters. That is glycaemic index (GI), a coarse ranking of how quickly a food raises glucose; glycaemic load (GL) combines that speed with the portion size. The international GI tables now list over 4,000 foods, a useful coarse tool for the quality dimension (Atkinson 2021).5 The honest caveat is that GI is a population-mean for a healthy reference group, not a personal value: a study of commercial breads found that loaves labelled “wholemeal” or “wholegrain” did not reliably produce a lower glucose response, so the label does not predict the curve (Calter 2026).6 The practical reading: lower-GI, higher-fibre choices tend to give a gentler, more predictable rise, which is one lever among several; the real T1D response is GI plus quantity plus fat plus protein plus pre-bolus timing plus the individual.
For a typical mixed meal with a rapid-acting insulin, taking the dose about 15 to 20 minutes before eating gives the insulin a head start on the glucose and lowers the post-meal peak (Cobry 2010, Slattery 2018, ISPAD Grade A).7 Shorter for very high-GI meals or faster insulins; longer for slow meals; flexible or after the meal for very young children whose intake is unpredictable. This is a population-average principle to explore with your care team, not a fixed instruction.
“65 percent more insulin” sounds precise until you read the range beside it: 17 to 124 percent. The honest headline is the direction, not the decimal. Whenever a single number is offered for something as variable as a meal, ask how wide the spread around it really is.The Goldacre lens, evidence-grade discipline, one of the four GNL appraisal lenses.
Why one ratio handles most meals, and where it breaks
Why one ratio handles most meals
The insulin-to-carbohydrate ratio is more capable than it looks. Pemberton’s clinical synthesis frames it as covering roughly 130 percent of the carbohydrate-only insulin demand for a typical mixed meal,8 because the ratio you settle on with your care team absorbs the average protein and fat load implicitly. That is why an ordinary meal does not need fat-and-protein arithmetic on top: it is already baked in. The mechanistic reason sits underneath: carbohydrate alone explains only about 49 to 57 percent of the post-meal insulin response; the rest is fat and protein (Bell 2014).4 The ratio is doing quiet work you do not see.
The macronutrient envelope, and the two meals that leave it
The ratio holds while a meal stays inside the usual envelope, broadly carbohydrate 40 to 55 percent, protein 10 to 20 percent, fat 20 to 40 percent of energy (ISPAD 2024 macronutrient frame, Grade C).9 Two kinds of meal step outside it. A near-zero-fat, carbohydrate-heavy meal can run higher than the ratio predicts, because there is less fat to slow the carbohydrate. A high-fat, high-protein meal exceeds the envelope the other way and needs the spread-bolus layer. GNL teaches two compatible population-average routes for those outliers, and labels both as working rules, not prescriptions:
| Meal type | Population-average pattern | Evidence anchor |
|---|---|---|
| Ordinary mixed meal | Standard ratio, dosed about 15 to 20 minutes ahead | ISPAD Ch10 A |
| High-fat, high-protein | More insulin, spread over roughly 2 to 3 hours; the size is learned with the team | Bell 2016 / 2020 A |
| Fat-protein unit (FPU) method | 100 kcal of fat and protein counted as one unit, dosed at the carb ratio over an extended bolus | Pankowska 2012 A |
D The percentage uplift and the split duration are population-average shapes drawn from the trials above, not a personalised dose. Every figure assumes the user’s own ratio and correction factor, set with the diabetes care team. GNL does not tell anyone how many units to take; if you do not know your ratio, do not act on a figure here, speak to your team. The paediatric carbohydrate-dose floor in GNL calculators caps body weight at 60 kg.
Does an automated system make counting optional?
Partly, and honestly. An automated insulin delivery (AID) system compensates when a meal is under-bolused, but it does not fully replace an accurate meal announcement. In adolescents on one hybrid closed-loop system, precise carbohydrate counting reached about 80 percent time in range against about 74 percent for a simplified small / regular / large approach (a roughly 7 percentage-point gap, Petrovski 2023), and the system delivered nearly twice the auto-correction insulin trying to close it.10 The simplified approach still met international targets for most participants, so it remains a meaningful option when counting is burdensome. The residual gap shows up as more time above range after meals, not as more lows. Counting well and bolusing on time make the algorithm’s job easier, not redundant.
It is the rare, large swing that does the lasting damage, not the average Tuesday dinner. A meal you under-count you fix at the next reading; the move that pays off most is the dull one, a repeatable default plate that holds under load, so your bandwidth is free for the days that genuinely need it.The Taleb lens, robustness to outliers, one of the four GNL appraisal lenses.
A teaching rule is only as honest as the label on it. The 130 percent coverage figure and the spread-bolus uplift are clinical syntheses on a Grade A and B trial base, not a per-person law; that is a clean, declared lens, not the whole view. Name what would strengthen it, and never sell the rule as the territory.The Hayes lens, technical and methodological rigour, one of the four GNL appraisal lenses.
The whole guide, summarised
Glucose never lies; it just records what the plate, the count and the timing did, in their own time. Make the food predictable, count close enough, and learn the late curve with your team.
This page is the taster. The full journey on Nutrition and type 1 diabetes, three modules and their 30 questions, with your progress saved, lives in Learn with Grace. Glucose never lies; come and learn to read it.
References
Evidence grades A (strongest) to D (editorial or working analysis).
- Smart CE, et al. In children using intensive insulin therapy, a 10 g variation in carbohydrate did not significantly affect postprandial glycaemia. Diabet Med. 2009;26(3):279-285; and the 20 g failure boundary in the companion paediatric work. A graded B per GNL methodology (small paediatric crossover).
- Brazeau AS, et al. Carbohydrate counting accuracy and blood glucose variability in adults with type 1 diabetes. Diabetes Res Clin Pract. 2013;99(1):19-23 (50 adults, 448 meals, 15.4 g / 20.9 percent mean error, 63 percent under-counted). A graded B per GNL methodology (cross-sectional observational).
- DAFNE Study Group. Training in flexible, intensive insulin management to enable dietary freedom in people with type 1 diabetes: dose adjustment for normal eating (DAFNE) randomised controlled trial. BMJ. 2002;325(7367):746 (169 adults, about 1.0 percentage-point HbA1c improvement). A
- Bell KJ, et al. Impact of fat, protein and glycaemic index on postprandial glucose control in type 1 diabetes, systematic review of 31 studies. Diabetes Care. 2015;38(6):1008-1015; Wolpert HA, et al. Dietary fat acutely increases glucose and insulin requirements. Diabetes Care. 2013;36(4):810-816 (about 42 percent more insulin for a high-fat dinner); Bell KJ, et al. Optimised mealtime insulin dosing for fat and protein in type 1 diabetes. Diabetes Care. 2016;39(9):1631-1634 (about 65 percent more insulin, range 17 to 124 percent); Bell KJ, et al. Amount and type of dietary fat, postprandial glycaemia, and insulin requirements in type 1 diabetes. Diabetes Care. 2020;43(1):59-66 (20 to 40 g fat about +6 percent, 60 g about +21 percent); Bell KJ. PhD thesis, University of Sydney, 2014 (carbohydrate explains about 49 to 57 percent of the insulin response). A
- Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J. International tables of glycaemic index and glycaemic load values 2021: a systematic review. Am J Clin Nutr. 2021;114:1625-1632 (over 4,000 entries; the reference for any quoted GI value). A
- Calter, et al. Glycaemic properties of commercially available breads. 2026 (breads labelled “wholemeal” or “wholegrain” did not consistently produce a lower glycaemic response). A graded B per GNL methodology. CLAIM-PENDING-MA: full citation (journal, volume, pages) to be verified at Medical Affairs sign-off.
- Cobry E, et al. Timing of meal insulin boluses to achieve optimal postprandial glycaemic control. Diabetes Technol Ther. 2010;12(3):173-177 (20-minute pre-bolus lowest excursion); Slattery D, Amiel SA, Choudhary P. Optimal prandial timing of bolus insulin in diabetes management: a review. Diabet Med. 2018 (15 to 20 minutes pre-meal); ISPAD 2024 Chapter 10 (pre-bolus Grade A). A
- Pemberton J. The Ultimate Guide to Mealtime Insulin Dosing in Type 1 Diabetes, 2021 (the about-130 percent ICR coverage rule; the within-10 g adult accuracy threshold; the spread-bolus working rule). D GNL clinical synthesis on a Grade A/B base.
- Annan SF, Higgins LA, Jelleryd E, et al. ISPAD Clinical Practice Consensus Guidelines 2024, Chapter 10: Nutritional management in children and adolescents with diabetes. Horm Res Paediatr / Pediatr Diabetes. 2024 (macronutrient frame: carbohydrate 40 to 50 percent, fat under 35 percent, protein 15 to 25 percent of energy, Grade C; insulin-to-carbohydrate matching Grade A; additive fat and protein effect Grade B). A
- Petrovski G, et al. Simplified meal announcement versus precise carbohydrate counting on the MiniMed 780G system in adolescents with type 1 diabetes, 12-week randomised study. 2023 (about 80 percent time in range with precise counting versus about 74 percent simplified). A graded B per GNL methodology. CLAIM-PENDING-MA: full citation (journal, volume, pages) to be verified at Medical Affairs sign-off.
One page, three voices: Jude, Grace, John. Population-average, not personalised.
Keep learning: Mealtime insulin · HbA1c and Time in Range · Alcohol and type 1 diabetes · Exercise