Objective To investigate the predictive value of pre-transfer serum fibroblast growth factor 21 （FGF-21）， anti-Müllerian hormone （AMH）， and neuropilin-1 （NRP-1） for early pregnancy loss at 6 weeks following in vitro fertilization-embryo transfer （IVF-ET）， and to establish an early predictive model based on serum biochemical markers. Methods This prospective study consecutively enrolled 322 women who achieved clinical pregnancy after IVF-ET at our center between September 2022 and September 2024. Participants were randomly divided into a modeling cohort （n = 225） and a validation cohort （n = 97） at a 7：3 ratio. According to ultrasound findings at 6 weeks of gestation， patients in the modeling cohort were classified into an early pregnancy loss group （n = 59） and an ongoing pregnancy group （n = 166）. Baseline clinical characteristics and pre-transfer serum levels of FGF-21， AMH， and NRP-1 were collected. Multivariate logistic regression was applied to identify independent risk factors for early pregnancy loss and to construct a predictive model. Model discrimination， calibration， and stability were evaluated using receiver operating characteristic （ROC） curves， the Hosmer-Lemeshow goodness-of-fit test， and bootstrap resampling in both cohorts. Results Univariate analysis revealed that the FSH/LH ratio， antral follicle count， and number of retrieved oocytes were significantly associated with early pregnancy loss （P < 0.001）. Compared with the ongoing pregnancy group， women with early loss showed significantly elevated pre-transfer serum FGF-21 levels， whereas AMH and NRP-1 levels were markedly reduced （P < 0.001）. Multivariate logistic regression demonstrated that an FSH/LH ratio < 1.8 （OR = 1.629， P = 0.002） and higher FGF-21 levels （OR = 1.338， P = 0.002） were independent risk factors， while higher AMH （OR = 0.741， P = 0.010） and NRP-1 levels （OR = 0.874， P = 0.007） were protective. Stratified analysis indicated that among patients with FSH/LH ≥ 1.8， FGF-21 levels were significantly higher and AMH and NRP-1 levels significantly lower （all P < 0.001）. Interaction analysis further suggested that the FSH/LH ratio significantly modified the associations between these biomarkers and pregnancy loss risk （P for interaction < 0.05）. Specifically， in the higher FSH/LH subgroup， the risk effect of FGF-21 was amplified， while the protective effects of AMH and NRP-1 were more pronounced. The combined predictive model achieved C-indices of 0.869 （95%CI： 0.826 ~ 0.926） in the modeling cohort and 0.835 （95%CI： 0.811 ~ 0.907） in the validation cohort. Its AUC for predicting early pregnancy loss was 0.934 in the modeling cohort and 0.909 in the validation cohort， both significantly outperforming individual markers （AUCs： FGF-21 = 0.867， AMH = 0.881， NRP-1 = 0.853； Z = 2.024， 1.831； P < 0.001）. Decision curve analysis showed that the model provided consistent net clinical benefit across threshold probabilities of 0.1 ~ 0.4， underscoring its clinical utility. Conclusions Elevated pre-transfer serum FGF-21 and reduced AMH and NRP-1 levels are strongly associated with early pregnancy loss at 6 weeks after IVF-ET. The predictive model developed in this study demonstrates robust accuracy and stability， offering substantial clinical application value for early risk stratification.