Previous studies had primarily confirmed the correlation between the abnormal pre-pregnancy body mass index (BMI) of pregnant women and their adverse pregnancy outcomes, but rare studies had focused on the stratification of the women with pre-pregnancy normal BMI and its impact on the preterm birth of the women. Methods: 418,397 women with pre-pregnancy BMIs of 18.5-23.9kg/m2 were randomly divided into the training cohort (313,799 women) and the validation cohort (104,598 women) at a 3:1 ratio. The levels of the pre-pregnancy blood glucose, liver enzymes and thyroid-stimulating hormones of the women were clustered using K-Means. Logistic regression analysis and sensitivity analysis were used to identify the correlation between the BMI of the women in different subgroups and their preterm birth. Results: Four distinct feature groups were identified in the training cohort, and which showed the consistent presence in the validation cohort, including the elevated liver enzyme group (ELE), the optimal blood glucose supply group (OBGS) and the elevated thyroid-stimulating hormone group (ETSH), and the low metabolic status group (LMS). Compared to that of the LMS group, only the OBGS group was associated with a reduced risk of the premature birth ［aOR (95% CI) of the training cohort=0.88 (0.84-0.92); aOR (95% CI) of the validation cohort=0.87 (0.80-0.95)］. Conclusion: Four replicable new subgroups exist in the women with the normal pre-pregnancy BMI. There is a lower rate of the premature births among the women with normal pre-pregnancy BMI when their pre-pregnancy blood glucose supply is optimal.