Same calories, a quarter of the carbs: a granola breakfast took a third off the after-breakfast glucose peak in women with gestational diabetes. Twelve of the 32 who started are missing from the result

Ask a woman with gestational diabetes which finger-prick reading she dreads and there is a good chance she will say the one after breakfast. Insulin resistance tends to be at its worst in the morning, so the same slice of toast that behaves itself at lunch can push glucose over the line at 8 am. Diet is the first treatment anyone is offered, and yet the advice on what that breakfast should actually be rests on thin evidence.

A trial published online in The American Journal of Clinical Nutrition on 10 October 2026 tested one specific answer. It is called GLIDING, it was run at Royal Prince Alfred Hospital in Sydney with colleagues in Colorado and Boston, and its design is simple enough to explain in a sentence: each woman ate one breakfast for seven days, went back to her usual diet for seven days, then ate the other breakfast for seven days, in an order decided at random.

The two breakfasts were matched for energy, 245 and 247 kcal. What differed was the carbohydrate: 10.9 g in the low glycaemic load breakfast, 43.4 g in the high one. That is a quarter of the carbohydrate for the same calories. The trial's registry entry says what was in the bowls. The low one was a nut and seed granola with almond milk. The high one, described there as a traditional breakfast, was cereal, cow's milk and fruit.

Each woman wore a continuous glucose monitor throughout, and it was blinded, which means she could not see her own readings. That is important. A woman who watches her glucose climb after cereal may walk it off or eat less at lunch, and then you are no longer measuring the breakfast.

The result at one hour after breakfast was not subtle. On the cereal breakfast, glucose averaged 7.5 mmol/L (134 mg/dL). On the granola breakfast it averaged 5.0 mmol/L (89 mg/dL). The confidence intervals do not come close to touching: 7.2 to 7.8 against 4.7 to 5.2. By our arithmetic that is 2.5 mmol/L lower, a third off the peak.

The second result is about the whole day. The researchers used a measure of how much glucose swings around, the coefficient of variation, which is the spread of the readings divided by their average. Over 24 hours it was 20% on the cereal weeks and 17% on the granola weeks. So one changed meal left a mark on the full day's trace, although a three-point difference is far smaller than the gap after breakfast itself.

Nothing else they checked moved. Fasting blood fats, ketones, how full the women felt and how active they were showed no detectable difference between the two weeks. Ketones are the thing people worry about when carbohydrate is cut in pregnancy, so it is reassuring that a low-carbohydrate breakfast inside an otherwise ordinary diet did not raise them here. With this few women, though, 'no detectable difference' is weak evidence that there is none.

Now the numbers behind the numbers. Thirty-two women were randomised. The analysis has 20: those who finished both weeks and had at least 48 hours of sensor data for each breakfast. That leaves out 12, or three in every eight. The summary does not say why they dropped out or whether more of them left during one breakfast than the other. If the women who found the granola hard to live with are the ones who went missing, the trial tells us what happens when the breakfast is eaten, and less about what happens when it is prescribed.

There is also a mismatch with the plan. The registry entry, filed in April 2023 before the first woman joined, names two main outcomes: glucose after breakfast, and glycaemic variability measured as time in range, the share of the day spent inside the target band. The published summary reports the first of those and the coefficient of variation. It does not give time in range at all. We could only read the summary, not the full paper, so the figure may well be in there. But it is the number a clinic would ask for first, and it is not in the place most readers will look.

The closing line of the summary says targeting breakfast may 'reduce morning bolus insulin requirements'. The registry lists current insulin or metformin treatment as a reason to be excluded from the trial. So, as far as we can tell from the registry and the summary, nobody in the trial was taking mealtime insulin when she joined and no insulin doses are reported. The claim about insulin is a reasonable guess from the glucose curve. It is not something this trial measured.

And the biggest gap is the one every short diet trial in pregnancy has. Seven days of sensor data say nothing about the baby: not birth weight, not the baby's own blood sugar after delivery, not how the birth went. Lower glucose after breakfast is a means to those ends. The registry calls this a pilot trial, and that is the right way to read it.

What it does show is useful all the same. In women with gestational diabetes, the size of the morning glucose rise tracked the carbohydrate in the bowl very closely, at equal calories, when nobody could see the readings. Anyone managing gestational diabetes should take their meal plan from their own diabetes team, because how much carbohydrate a pregnancy needs across the day is a clinical question. But the idea that breakfast is the meal where a swap buys the most now has a clean, small experiment behind it.