ABSTRACT
RESEARCH QUESTION: Can an optimal LH threshold algorithm accurately predict timing of ovulation for natural cycle-intrauterine insemination (NC-IUI)? DESIGN: A retrospective cohort study (2018-2022) including 2467 natural cycles. Ovulation timing for these cycles was determined using a previously developed AI model. Two LH thresholds, low and high, were determined in the LH algorithm. Being below the low threshold meant that ovulation is likely to occur in ≥ 4 days, suggesting another daily blood test. Between the two thresholds meant that ovulation was likely in 2-3 days, suggesting IUI the next day. Above the high threshold meant that ovulation will likely occur tomorrow, suggesting performing IUI on the same day. RESULTS: The optimal LH model with a high threshold of 40 mIU/ml and a low threshold of 11 mIU/ml succeeded in correctly predicting timing for IUI (day - 1, - 2 relative to ovulation) in 75.4% (95%CI 75.3-75.4). In 23.1% (95%CI 23.0-23.2), the algorithm predicted "error," suggesting performing insemination when in fact it would have been performed on a non-optimal day (0 or - 3). A previously described 3-hormone-based (LH, estradiol, progesterone) AI model performed significantly better in all parameters (93.6% success rate, 4.3 "error" rate). CONCLUSIONS: An LH threshold model, representing common practice, evaluating all possible high and low LH threshold combinations, was successful in accurately scheduling timing for IUI in only 75% of cases. Integrating all three hormones as performed in the AI model may have an advantage in accurately predicting the optimal time for IUI, over the use of LH only.