Development of a predictive model for luteal phase oocyte retrieval in poor responders undergoing natural cycle IVF

The aim of this study is the development of a prediction model indicating successful application of Oocyte Retrieval performed during the Luteal Phase (LuPOR) in poor responders, as defined by the retrieval of at least one MII oocyte. Recruitment included 1688 poor responders diagnosed as per Bologna Criteria, undergoing natural cycle ICSI between 2012 and 2020. Oocyte collections were performed during the follicular phase and during the luteal phase similarly. Antral Follicle Count (AFC), Estradiol (E₂) levels evaluated on both trigger days prior to Follicular Phase Oocyte Retrieval (FoPOR) and LuPOR, and the number of small follicles 8-12 mm that were not aspirated during FoPOR were identified as predictive factors indicative of an efficient LuPOR practice with an Area Under the Curve (AUC) of 0.86, 0.86, 0.89 as well as 0.82 respectively. The combination of the above-mentioned characteristics into a prediction model provided an AUC of 0.88, specificity and a sensitivity of 0.73 and 0.94 respectively and an accuracy of 0.89. The model provided a positive predictive value (PPV) of 93.5% and a negative predictive value (NPV) of 46.8%. The clinical conclusion of the present study aims to be of added value to the clinician, by providing a prediction model defining the POR population benefiting from LuPOR. The high PPV of this model may renders this tool helpful for the practitioner that considers LuPOR.