Association of nutritional metabolic phenotypes in early pregnancy with preeclampsia: a retrospective study based on cluster analysis

To identify distinct nutritional-metabolic subtypes in early pregnancy using cluster analysis and to evaluate their association with preeclampsia (PE) risk. This study included 23,959 pregnant women. Nine early-pregnancy nutritional and lipid biomarkers (HDL, LDL, total cholesterol, triglycerides, folate, vitamin B12, calcium, ferritin, and vitamin D) were analyzed. K-means clustering was applied to identify latent metabolic patterns. Associations between subtypes and PE and severe PE (sPE) were assessed using logistic regression. A random forest model was used to evaluate the robustness of the clustering structure. Four nutritional-metabolic subtypes were identified: normal control ( n = 6,830), dyslipidemia ( n = 4,712), nutrient-deficient ( n = 5,047), and high-lipid micronutrient-sufficient ( n = 7,370) groups. The nutrient-deficient group showed a significantly higher risk of both PE (OR = 1.43, 95% CI: 1.13–1.81, P = 0.003) and sPE (OR = 1.76, 95% CI: 1.22–2.54, P = 0.002) compared with the normal control group in multivariate analysis. Stratified analyses suggested effect modification by pre-pregnancy BMI and parity, whereas maternal age showed no consistent association across subgroups. The random forest model achieved a classification accuracy of 93.4%, supporting the stability of the clustering approach. Cluster-based subtyping of early pregnancy nutritional and lipid profiles may help identify women at increased risk of preeclampsia and may serve as a potential tool for early risk stratification and individualized preventive strategies.