This prospective investigation utilized data from the Chinese Miao Mother and Child Cohort(MCCMC), established in September 2022, to examine associations between dietary factors and adverse pregnancy outcomes. Participants were recruited from three hospital sites within the cohort center located in Miao settlement areas of Qiandongnan Miao and Dong Autonomous Prefecture, Guizhou Province, China. Pregnant women were eligible for inclusion if they met the following criteria: (a) age ≥18 years, (b) singleton pregnancy, (c) Gestational age ≤ 12 weeks, (d) Miao ethnicity, and permanent residence in the cohort center region for ≥1 year. Women were excluded if they had a history of pre-pregnancy diabetes or fasting blood glucose ≥7.0 mmol/L, prior diagnosis of GDM during previous pregnancy, severe liver or kidney disease, autoimmune diseases, or long-term use of glucocorticoids or other medications that affect glucose metabolism.

Sample size calculation was based on an observed GDM prevalence of 18.0% in southwest China(29), with a significance level (α) of 0.05, a statistical power of 90%, and an allowable margin of error (d) of 3.0%. The required sample size was therefore estimated to be 630 participants. Initially, 730 participants were recruited; 32 were excluded due to loss to follow-up, and 7 were excluded due to incomplete dietary data. During the analysis phase, we further excluded participants whose reported dietary intake reflected implausible total energy intake (< 2.09 MJ [500 kcal]/day or >20.92 MJ [5,000 kcal]/day) to minimize dietary measurement error.

This study was approved by the Ethics Committee of the Affiliated Hospital of Guizhou Medical University (approval number: 2021 [065-01]), and written informed consent was obtained from all participants prior to enrollment.

This study employed a prospective cohort design to conduct longitudinal follow-up surveys among pregnant women. The baseline assessment, conducted during early pregnancy (< 12 weeks gestation), gathered participant information and recalled dietary and physical activity status from the preceding year. A follow-up assessment was then conducted during mid-pregnancy (13 weeks until GDM outcome) to capture dietary intake and physical activity during early pregnancy.

Dietary intake was evaluated using a validated semi-quantitative FFQ comprising 74 food items. The FFQ was specifically designed to capture both common foods consumed in southwestern China and traditional Miao ethnic cuisine, including sour soup (red and white varieties), cured pork, pickled vegetables, and others. This instrument has demonstrated reliable validity and reproducibility in previous validation studies among Miao pregnant women (30).

Data were gathered via face-to-face interviews conducted by trained investigators. Participants reported their consumption frequency (daily, weekly, monthly, or never) for each food item. When participants provided frequency ranges (e.g., “4–5 times/week”), the median value (4.5 times/week) was recorded.

More accurate portion size estimation was facilitated using standardized food models and food maps as visual references. The models represented a standard 90 kcal portion, and participant intake was recorded as multiples thereof (e.g., 0.5 × , 2 × ). For items lacking physical models, food maps were employed in conjunction with a validated conversion table. This table established weight equivalents for all depicted portions (e.g., 1 plate of shredded potatoes = 100 g) and was developed based on pre-study consultations with local nutritionists and market surveys. Daily food intake (g/day) was calculated by multiplying the daily consumption frequency by the portion weight (g). Nutrient intakes were determined by multiplying daily food intake by the corresponding nutrient density values from the Chinese Food Composition Table (31). Nutrient calculations excluded sour soup, a traditional Miao food, due to limited available data on its detailed nutritional composition, despite studies examining its microbial community diversity (32, 33).

Multiple non-dietary covariates were evaluated as potential confounding factors, including maternal demographics, physical activity, obstetric and gynecological history, and familial diabetes. Pre-pregnancy body mass index (BMI) [kg/m²] was calculated based on weight and height measurements taken on-site. Parity was classified as either nulliparous (0 previous births) or multiparous (≥1 previous births). Family history of diabetes was recorded as positive when participants reported diabetes diagnoses in immediate family members (parents or siblings). The history of polycystic ovary syndrome (PCOS) was documented as yes or no. The International Physical Activity Questionnaire (IPAQ) was used to measure participants' intensity and duration of weekly physical activity during preconception and early pregnancy.

A standardized 2-h 75g oral glucose tolerance test (OGTT) was administered between the 24th and 28th weeks of gestation to screen for GDM. All OGTT results were documented in the electronic medical record systems of the hospital sites within the cohort center. GDM was ascertained if they met at least one of the following criteria: fasting plasma glucose >5.1 mmol/L, 1-h post-load glucose >10.0 mmol/L, or 2-h post-load glucose ≥8.5 mmol/L based on the 2011 criteria established by the Ministry of Health of China (34). For participants who underwent OGTT screening at facilities outside the cohort center, GDM status was ascertained through telephone follow-up interviews. In cases where participants self-reported a GDM diagnosis but complete OGTT values (all three time points) could not be obtained, we recorded the binary outcome of physician-diagnosed GDM without the specific glucose measurements.

We classified 74 foods into 23 food groups based on nutritional similarities. Sour soup was categorized separately due to its distinctive nutrient profiles. Principal component analysis with varimax rotation was used to identify dietary patterns. The Kaiser–Meyer–Olkin test (KMO=0.749 for pre-pregnancy; KMO = 0.722 for early pregnancy) and the Bartlett test of sphericity (all P < 0.001) suggested that the data structure was reasonable. Eventually, we determined the number of patterns using scree plot analysis, eigenvalues (>1), factor interpretability, and explained variance. Food categories contributing to each dietary pattern were determined based on factor loadings ≥0.3. Subsequently, Principal component scores were calculated for each pregnant woman, with the highest value representing the pregnant woman's dietary pattern. For each dietary pattern, the higher the principal component score, the higher the degree of adherence to the pattern. Participants were divided into quartiles (Q1-Q4) according to their dietary pattern scores. Continuous variables are presented as median and interquartile range (Median ± IQR), and comparisons between groups were made using the Wilcoxon rank-sum test. Categorical variables are presented as frequency and percentage [n(%)].

Logistic regression models were used to assess the association between dietary patterns and the risk of GDM, with odds ratios (OR) and 95% confidence intervals (95% CI) calculated. In the multivariate analysis, model 1 was the unadjusted model; model 2 adjusted for other dietary patterns, maternal age, household monthly income per capita, occupation, and family history of PCOS; model 3 further adjusted for pre-pregnancy BMI, physical activity level, and energy intake (quartiles) based on model 2. To assess linear trends, quartile groups for each dietary pattern were treated as continuous variables and included in logistic regression models to calculate P trend values. Additionally, we employed restricted cubic spline analyses to examine potential dose-response relationships between consumption of traditional Miao foods (including sour soup and pickled vegetables) and GDM.

In this prospective cohort study of 683 Miao women, 130 participants (19.03%) developed GDM (Table 1). The mean age of the cohort at enrollment was 30.44 ± 4.76 years. The largest age group was 30–35 years, accounting for 275 (40.26%) of the participants. Most participants (652 [95.46%]) were married. Educational attainment varied considerably: 405 (59.30%) had completed 13-15 years of formal education, while 193 (28.26%) had less than 9 years of schooling. The mean pre-pregnancy BMI was 22.36 ± 3.49. 423 (61.93%) participants were classified as normal weight, 184 (26.94%) as overweight or obese, and 76 (11.13%) as underweight. Regarding occupation, 101 (14.79%) participants were manual workers. More than half of the women (354 [51.83%]) had multiple pregnancies. Medical history included PCOS in 29 (4.25%) participants and a family history of diabetes in 45 (6.59%). Physical activity was predominantly of light intensity during.

Principal component analysis identified two major dietary patterns among Miao women during both preconception and early pregnancy periods (Supplementary Table S1). The first pattern, characterized by high consumption of whole grains, vegetables, fungi/algae, fruits, beans, meat, and eggs, was designated the “prudent” pattern. The second pattern, featuring processed meats, snacks, convenience foods, desserts, and beverages, was termed the “processed” pattern. These patterns explained 11.78% and 7.58% of dietary variance during preconception, and 11.25% and 8.13% during early pregnancy, respectively. Similar clustering was observed in hierarchical analysis (Figure 1).

Women with higher adherence to the prudent pattern consumed significantly greater quantities of vegetables, fungi/algae, beans, meat, and eggs during both periods compared with those following the processed pattern (all P < 0.05; Supplementary Table S2). Conversely, processed pattern adherents had significantly higher intakes of processed meats, pickled vegetables, snacks, convenience food, dairy desserts, and beverages (all P < 0.05). Nutrient analysis revealed that prudent pattern followers had significantly higher intakes of protein, fatty acids, dietary fiber, vitamins, and minerals during both preconception and early pregnancy periods compared with the processed pattern group (all P < 0.05; Table 2).

Table 3 presents the odds ratios of GDM according to quartiles of dietary pattern scores during the preconception year and early pregnancy. For the prudent dietary pattern during preconception, a significant inverse association was observed with GDM risk in the fully adjusted model (Model 3), with the highest quartile showing a 51% reduced risk (OR: 0.49; 95% CI: 0.24, 0.97) compared to the lowest quartile (P for trend = 0.049). Similarly, during the early pregnancy, adherence to the prudent dietary pattern was significantly associated with lower GDM risk in the fully adjusted model, with the highest quartile demonstrating a 56% reduction in risk (OR: 0.44; 95% CI: 0.22, 0.89; P for trend = 0.031). However, we did not observe a significant association between processed dietary prudent and GDM after adjusting for other confounding variables, either in the year before pregnancy or during early pregnancy.

Restricted cubic spline analysis revealed distinct dose-response patterns for traditional Miao foods and GDM risk (Figure 2). Sour soup consumption exhibited protective associations during both study periods, with significant overall associations during preconception (P-overall = 0.044) and early pregnancy (P-overall = 0.011). The association during preconception approached statistical significance for linearity (P = 0.050 for non-linearity), while the relationship in early pregnancy exhibited a linear pattern (P = 0.073 for non-linearity). By contrast, pickled vegetable consumption was not significantly associated with GDM risk during either the preconception (P-overall = 0.337) or early pregnancy (P-overall = 0.493) periods. The trends for both periods were linear (P for non-linearity = 0.141 and 0.511, respectively).