INTRODUCTION: While serum luteinizing hormone (LH) on the day of trigger is a conventional predictor of in vitro fertilization (IVF) outcomes, its clinical utility remains inconsistent. This study aimed to evaluate the serum LH/large follicle (LH/LF) ratio as a refined risk-stratification marker for reproductive outcomes in IVF/ICSI cycles. METHODS: A retrospective cohort study was conducted involving 8,329 first fresh IVF/ICSI cycles. Multivariable logistic regression models, adjusted for maternal age, body mass index (BMI), anti-Müllerian hormone (AMH), basal antral follicle count (Basal AFC), endometrial thickness, number of embryos transferred, and embryo developmental stage, evaluated the independent associations of the LH/LF ratio with the probability of clinical pregnancy and live birth. Predictive performance was compared with absolute LH levels using receiver operating characteristic (ROC) curves and the DeLong test. RESULTS: The LH/LF ratio was identified as a significant independent negative predictor of clinical pregnancy (OR = 0.589, 95% CI: 0.540-0.644, P < 0.001) and live birth (OR = 0.599, 95% CI: 0.546-0.656, P < 0.001). For clinical pregnancy, the LH/LF ratio exhibited improved discriminative power compared to absolute LH levels (AUC = 0.717 vs. 0.651; P < 0.001). Stratification analysis revealed that within the elevated absolute LH population, the LH/LF ratio successfully differentiated a distinct subgroup (high LH but low ratio; Group C) that achieved a clinical pregnancy rate of 49.66% and an LBR of 44.14%, which were significantly higher than the 33.39% and 27.73% observed in the dual-high risk cohort (Group B; P < 0.01). CONCLUSIONS: The LH/LF ratio may provide valuable prognostic information for reproductive outcomes compared to absolute LH levels alone, which is consistent with the proposed follicular dilution effect. By integrating systemic hormonal profiles with follicular burden, this ratio provides a potentially useful framework for individualized clinical risk stratification in fresh cycles, pending future prospective external validation.