In 78 pregnant women in southwest China, each step towards a diet of sweet drinks, processed meat and convenience food went with more than double the odds of gestational diabetes

Gestational diabetes, high blood sugar that first shows up during pregnancy, affects roughly one pregnancy in seven worldwide by the most widely used criteria. It is usually found with a glucose drink test between 24 and 28 weeks, and it raises the chances of a large baby, a caesarean and pre-eclampsia, and of type 2 diabetes for the mother years later. Diet is one of the few risk factors a woman can change, so researchers keep asking which ways of eating matter most. A study published in Frontiers in Nutrition on 28 October looked at that question in a group rarely studied: Miao women in Qiandongnan, a mountainous prefecture of Guizhou province in southwest China.

The team, from Guizhou Medical University and Shenzhen University, recruited Miao women having routine antenatal care at three hospitals in Kaili city between March 2023 and January 2025. Each woman filled in a food questionnaire built and tested for Miao pregnant women, and gave a fasting blood sample on the same morning, before anyone knew her glucose test result. Thirty-nine women were then diagnosed with gestational diabetes, and each was paired with a woman of similar age, stage of pregnancy and pre-pregnancy weight whose test was normal. About a third of the 78 were assessed at 10 to 12 weeks; the rest at 24 to 28 weeks, on the day before their glucose test.

Rather than score single foods, the researchers used two ways of eating that an earlier analysis of 705 pregnant women from the same cohort had found. One, which they call the processed pattern, was marked by drinks (sweetened beverages loaded most heavily), alcohol, sweet dairy desserts, processed meat, refined grains, convenience foods, red meat and snacks, and by less milk and fewer eggs. The other, a balanced pattern, leaned on mushrooms and seaweed, fish, whole grains, root and leafy vegetables, eggs, dairy, fruit and poultry. Every woman got a score on each pattern.

The processed score was the one that tracked the diagnosis. After allowing for total calories and for how many times a woman had been pregnant and given birth, each one-step rise in the score (one standard deviation, roughly the gap between a typical woman and one noticeably more reliant on those foods) went with 2.33 times the odds of gestational diabetes. With both patterns in the model it was 2.55 times. The balanced pattern pointed towards less risk in some models but never gave a clear result. One step was shakier than the rest: allowing for calories alone pulled the processed result just short of statistical significance (2.06 times, with a range that included no effect).

The second half of the study looked for a trace of this in the blood. Out of 1,658 proteins detected, 344 passed quality checks, and two got special attention: apolipoprotein M, which travels mostly on HDL ('good' cholesterol) particles, and serum amyloid A4, a protein made by the liver. Both were lower in the women with gestational diabetes, and women with higher processed-diet scores had lower apolipoprotein M (the link with amyloid A4 was borderline). When the two proteins were added to the model, the diet score itself no longer stood out, which fits the idea that the proteins sit on the path between eating and blood sugar, though a study like this cannot show that.

The authors are frank about the limits. Seventy-eight women is a small group for a study that tests hundreds of proteins, and only four of the 34 proteins flagged as different survived a correction for multiple testing. Diet was recorded once per woman and by questionnaire, physical activity wasn't recorded at all, and income and schooling were not adjusted for. A protein-based score that sorted women very well in this sample was built and tested on the same people, after a data-cleaning step that already knew who had diabetes, so the team calls it a proof of concept, not a test anyone should use. What the study adds is a careful look at one community, and a result that lines up with larger studies elsewhere: diets heavy in sweet drinks, processed meat and ready-made food keep turning up alongside diabetes in pregnancy.