📄 medicine

The Progesterone-to-Oocyte Index Independently Predicts Live Birth in GnRH Antagonist Fresh IVF/ICSI Cycles: Development and Temporal Validation of a Nomogram

This study developed and temporally validated an eight-variable nomogram demonstrating that the progesterone-to-oocyte index (POI) is an independent predictor of live birth in GnRH antagonist fresh IVF/ICSI cycles, offering a valuable tool for risk stratification and clinical decision-making, particularly for poor ovarian responders.

Original authors: Xinxin Li, Yushan Huang, Xiya Zhang, Tian Gan, Weijie Xing, Yue Meng

Published 2026-10-08

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Xinxin Li, Yushan Huang, Xiya Zhang, Tian Gan, Weijie Xing, Yue Meng

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the journey to build a family through assisted reproduction, one of the most critical moments occurs just before an egg is collected from a woman's ovaries. At this stage, doctors measure a hormone called progesterone in the blood. This hormone acts as a signal, telling the lining of the uterus to prepare for a pregnancy. However, if the levels of this hormone rise too early or too high, it can cause the uterine lining to mature too quickly, becoming out of sync with the developing embryo. When this happens, the embryo may struggle to attach, and the chance of a successful pregnancy drops. For years, doctors have debated how to measure this risk. They have looked at the raw amount of progesterone in the blood, but this number changes depending on how many eggs a woman produces. A woman with many eggs might have a high progesterone level simply because she has many active follicles, not because her body is reacting poorly. This has led researchers to search for a better way to measure the hormone's impact, one that accounts for the individual's specific response to treatment.

A team of researchers at the Third Affiliated Hospital of Sun Yat-sen University set out to solve this puzzle by analyzing thousands of past treatment cycles. They focused on a specific type of fertility treatment where a drug is used to prevent the body from releasing eggs too early, allowing doctors to control the timing of the egg retrieval. The researchers gathered data from 2,365 cycles performed between 2015 and 2024. Their goal was to create a tool that could predict, on the very day of egg retrieval, whether a fresh embryo transfer would result in a live birth or if it would be better to freeze the embryos and wait for a later date. To do this, they compared three different ways of measuring progesterone: the raw amount in the blood, a ratio comparing the hormone to the number of follicles seen on an ultrasound, and a new ratio comparing the hormone to the actual number of eggs retrieved.

The study found that the most accurate predictor was the ratio of progesterone to the number of eggs retrieved. The researchers call this the progesterone-to-oocyte index. By using this specific measurement alongside seven other common factors—such as the woman's age, how long she had been trying to conceive, the thickness of her uterine lining, the amount of stimulation medication used, and the number of embryos available—they built a personalized prediction tool. This tool, known as a nomogram, allows a doctor to plug in a patient's specific numbers and receive a calculated probability of having a baby. The analysis showed that this progesterone-to-oocyte ratio was the strongest signal of all. When this ratio was high, the chances of a live birth dropped significantly, even if the woman had many eggs. In fact, the study revealed a surprising twist: while having more eggs usually helps, once the progesterone ratio is taken into account, having a very large number of eggs can sometimes be associated with a lower chance of success if the progesterone burden per egg is too high.

The researchers tested their new tool on two separate groups of patients. First, they used data from 2015 to 2021 to build the model, and then they tested it on data from 2022 to 2024 to see if it would still work with newer patients. The tool performed consistently well in both groups, correctly distinguishing between cycles that would succeed and those that would not. The results were particularly helpful for women who produced fewer eggs, a group where previous methods often failed to give clear guidance. For these patients, the tool was even more accurate, suggesting that the progesterone-to-oocyte ratio is a vital clue for those with lower ovarian response. The study concludes that this index offers a reliable way to make the difficult decision between transferring an embryo immediately or freezing it for later. By identifying when the uterine environment is likely too advanced for a fresh transfer, doctors can choose to pause the cycle, preserving the embryos for a time when the body is more ready, ultimately giving patients a better chance of holding a healthy baby.

Technical Summary: The Progesterone-to-Oocyte Index (POI) as a Predictor of Live Birth in GnRH Antagonist IVF/ICSI Cycles

Problem Statement
In GnRH antagonist fresh IVF/ICSI cycles, the timing of embryo transfer is a critical decision point. While fresh transfers offer benefits such as a shorter time to live birth and lower economic burden, premature elevation of serum progesterone (P) on the trigger day is known to alter endometrial receptivity, potentially compromising implantation. However, the prognostic utility of absolute progesterone levels remains debated, as they are inherently coupled to the number of hormonally active follicles. Previous attempts to normalize this metric, such as the Progesterone-to-Follicle Index (PFI), have shown promise but rely on ultrasound follicle counts, which are subject to inter-observer variability. Furthermore, existing prediction models for live birth often suffer from limited cohort sizes, lack of external validation, or failure to systematically compare different progesterone metrics. There is a need for a robust, individually tailored prediction tool that integrates progesterone dynamics with other clinical parameters to guide the decision between fresh transfer and elective cryopreservation (freeze-all).

Methodology
This retrospective cohort study was conducted at the Third Affiliated Hospital of Sun Yat-sen University, analyzing 2,365 fresh GnRH antagonist IVF/ICSI cycles performed between January 2015 and December 2024. The cohort was chronologically partitioned into a development set (2015–2021, n=1,644) and a temporal validation set (2022–2024, n=721) to mimic prospective deployment and prevent information leakage.

The study employed a rigorous three-stage predictor selection framework:

Univariable Filtering: 17 candidate variables (spanning patient demographics, ovarian reserve, stimulation parameters, and embryological factors) were screened, retaining those with $P < 0.10$ .
Collinearity Management and Progesterone Parameter Selection: To address the interdependence of absolute P, PFI, and the Progesterone-to-Oocyte Index (POI), three parallel multivariable models were fitted. The model with the lowest Akaike Information Criterion (AIC) and highest Area Under the Curve (AUC) was selected.
Dual Criterion Selection: Surviving variables underwent both LASSO logistic regression (10-fold cross-validation) and bidirectional AIC-based stepwise selection. Only variables selected by both methods were retained for the final model.

The primary endpoint was live birth (delivery of one or more live neonates $\ge$ 28 weeks). Model performance was assessed via discrimination (AUC), calibration (Hosmer–Lemeshow test, calibration plots), and clinical net benefit (Decision Curve Analysis). Subgroup analyses were conducted to evaluate model robustness across different patient strata (e.g., poor responders, age groups).

Key Results

Optimal Progesterone Metric: POI (calculated as trigger-day serum P divided by the number of retrieved oocytes) was identified as the superior predictor compared to absolute P and PFI, yielding the lowest AIC (2041.45 vs. 2042.24 for PFI and 2046.44 for absolute P).
Nomogram Construction: The final model incorporated eight independent predictors: female age, infertility duration, endometrial thickness, total Gn dose, oocytes retrieved, POI, available embryos, and embryos transferred.
Predictive Associations:
- POI: Exhibited the strongest inverse association with live birth (Adjusted OR 0.027, 95% CI 0.002–0.305; $P=0.004$ ).
- Oocyte Count Reversal: A notable statistical phenomenon was observed where the coefficient for "oocytes retrieved" reversed from positive in univariable analysis (OR 1.059) to negative in the multivariable model (OR 0.951). The authors attribute this to a suppression effect, where the detrimental impact of total progesterone load (mediated by POI) and embryo quality variables unmask a negative association when adjusted for.
Model Performance:
- Discrimination: The model demonstrated consistent discrimination across cohorts: Apparent AUC 0.663 (training), Optimism-corrected AUC 0.666, and Temporal Validation AUC 0.675.
- Calibration: Satisfactory calibration was confirmed with Hosmer–Lemeshow $P \ge 0.129$ and calibration slopes near unity.
- Subgroup Utility: The model showed enhanced performance in poor ovarian responders, achieving an AUC of 0.791 for patients with an Antral Follicle Count (AFC) < 5 and 0.730 for those with AMH < 1.2 ng/mL.
Clinical Utility: Decision curve analysis indicated a positive net benefit across clinically relevant threshold probabilities, supporting the model's use for risk stratification.

Significance and Claims
The study claims that POI independently predicts live birth in GnRH antagonist fresh IVF/ICSI cycles and offers a mechanistic advantage over absolute progesterone or PFI by normalizing progesterone burden to the functional output of the follicular cohort (retrieved oocytes). The integration of POI into an eight-variable nomogram provides a bedside tool for trigger-day risk stratification.

The authors emphasize that the model's enhanced predictive power in poor responders suggests that absolute progesterone thresholds are least informative in this specific subgroup, where the per-follicle progesterone output is disproportionately high. Mechanistically, the study supports the rationale that elevated progesterone impairs endometrial receptivity rather than embryo ploidy; therefore, when POI is elevated, a "freeze-all" strategy is supported to preserve the cycle's reproductive investment. The temporal validation approach is highlighted as a methodological strength, offering a more realistic estimate of prospective performance than random data splitting. The study concludes that this tool facilitates individualized counseling and evidence-based decisions regarding fresh transfer versus elective cryopreservation.

Technical Summary: The Progesterone-to-Oocyte Index (POI) as a Predictor of Live Birth in GnRH Antagonist IVF/ICSI Cycles

More like this