📄 medicine

Corpus Luteum Absence Is an Independent Risk Factor for Early Pregnancy Loss After ART: A Risk Stratification Model with Temporal Validation

This study demonstrates that corpus luteum absence is an independent risk factor for early pregnancy loss following assisted reproductive technology, enabling the development of a validated seven-variable risk stratification model that significantly outperforms age alone in predicting pregnancy outcomes.

Original authors: Huiyi Xie¹, Yifang Wu¹, Yunni Cai¹, Jin'gu Li

Published 2026-09-30

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Huiyi Xie¹, Yifang Wu¹, Yunni Cai¹, Jin'gu Li

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

When a pregnancy begins, the body relies on a tiny, temporary organ called the corpus luteum to keep it going. This structure forms in the ovary after an egg is released and acts as a vital support system, pumping out hormones that prepare the uterus to nurture the developing embryo. In natural conception, this organ is always present. However, in modern fertility treatments, doctors sometimes use medications to prepare the uterus without triggering the release of an egg. When this happens, the corpus luteum never forms, and the body must rely entirely on external hormones to sustain the pregnancy. For years, scientists have wondered if this missing organ makes a difference in whether a pregnancy survives its earliest, most fragile weeks. The question is not just academic; it touches on how millions of families approach the complex journey of having a baby through medical assistance.

A team of researchers at Quanzhou Women and Children's Hospital in China set out to answer this question by looking at a massive collection of real-world data. They examined nearly 6,000 cycles where women had achieved a confirmed pregnancy through assisted reproductive technology between 2012 and 2026. The researchers sorted these pregnancies into two main groups: those where the woman's body had produced its own corpus luteum, and those where it had not. The first group included women who had fresh embryo transfers or frozen transfers in a natural cycle, both of which involve ovulation. The second group consisted of women who underwent frozen embryo transfers using hormone replacement therapy, a method that suppresses ovulation and leaves the corpus luteum absent. By comparing the outcomes of these two groups, the study aimed to see if the presence or absence of this specific biological structure influenced the risk of early pregnancy loss.

The results revealed a clear and consistent pattern. Pregnancies that occurred when the corpus luteum was present had the lowest rates of loss, regardless of whether the embryo was fresh or frozen. In contrast, pregnancies where the corpus luteum was missing faced a significantly higher risk of ending in miscarriage. The difference was measurable and distinct: the group with the missing organ experienced a loss rate of roughly 19.4 percent, while the group with the organ present saw a rate of about 14.8 percent. This gap held true even after the researchers accounted for other known factors that affect pregnancy success, such as the age of the mother, her body mass index, and her history of previous miscarriages. The data suggested that the absence of the corpus luteum is an independent risk factor, meaning it carries its own weight in determining the outcome, separate from other variables.

To make these findings useful for doctors and patients, the researchers built a prediction model that combined the status of the corpus luteum with six other routine clinical details. This seven-variable tool was tested against a fresh set of data from later years to ensure it would work for future patients. The model proved to be significantly better at identifying risk than looking at a patient's age alone. It successfully sorted patients into four distinct risk categories, ranging from a low-risk group with a loss rate of just 3.2 percent to a very high-risk group where the loss rate climbed to 39.1 percent. This wide spread demonstrates that the model can effectively distinguish between those who are likely to have a successful pregnancy and those who face a much steeper challenge.

However, the study also highlighted a limitation in how the data was gathered. Because the group of women who had frozen embryos in a natural cycle was quite small, the researchers could not completely separate the effect of the missing corpus luteum from the effect of freezing and thawing the embryo. While the evidence strongly points to the missing organ as a key culprit, the precise contribution of the freezing process itself remains slightly blurred. Despite this, the study provides the strongest quantitative evidence to date that preserving the body's natural hormone-producing structures is crucial. It suggests that when medically possible, fertility protocols that allow the corpus luteum to form should be prioritized, as they offer a better chance of keeping a pregnancy alive through its critical early stages.

Technical Summary: Corpus Luteum Absence as an Independent Risk Factor for Early Pregnancy Loss After ART

Problem Statement
Assisted Reproductive Technology (ART) cycles utilize various endometrial preparation protocols, notably Natural Cycle (NC), Hormone Replacement Therapy (HRT), and GnRH agonist downregulation plus HRT (GnRHa + HRT). A critical biological distinction exists between these protocols: NC and fresh embryo transfers (ET) involve an endogenous corpus luteum (CL), whereas HRT and GnRHa + HRT protocols lack a CL due to the absence of ovulation. While the "corpus luteum hypothesis" suggests that CL-derived vasoactive factors (e.g., relaxin, VEGF) are essential for placental vascular remodeling and pregnancy maintenance, empirical evidence validating CL status as an independent predictor in a multivariable risk model has been lacking. Furthermore, existing prediction models for pregnancy loss often rely on internal cross-validation, lack temporal validation, and rarely incorporate endometrial preparation protocols or CL status as predictors.

Methodology
This single-center retrospective cohort study analyzed 5,865 ART cycles with definitive outcomes (live birth or spontaneous abortion) from Quanzhou Women's and Children's Hospital between 2012 and 2026.

Cohort Stratification: Cycles were categorized by CL status:
- CL-Present: Fresh IVF/ICSI and Frozen Embryo Transfer (FET) in Natural Cycles (NC).
- CL-Absent: FET using HRT or GnRHa + HRT protocols.
Data Split: The dataset was temporally split to ensure rigorous validation. The training set comprised data from 2012–2022 ( $n=3,582$ ), and an independent temporal validation set comprised data from 2022–2026 ( $n=2,283$ ).
Statistical Modeling: A 7-variable logistic regression model (Model A) was developed using routinely available clinical variables: age, Anti-Müllerian Hormone (AMH), Body Mass Index (BMI), infertility duration, CL absence (binary), multiple gestational sacs, and history of prior spontaneous abortion.
Handling Missing Data: Missing AMH data were handled using Multiple Imputation by Chained Equations (MICE, $m=20$ ). The missingness exhibited a marked temporal gradient, ranging from near-complete absence in 2012–2016 (99.6%) to 21.5% in 2023–2026, reflecting the gradual adoption of AMH testing.
Validation Metrics: Model performance was assessed via Area Under the Curve (AUC), DeLong tests for comparison against age-only models, Continuous Net Reclassification Improvement (NRI), Integrated Discrimination Improvement (IDI), calibration curves, Brier scores, and Decision Curve Analysis (DCA).

Key Contributions

Whole-ART-Cycle Validation: This study provides the first large-scale, whole-ART-cycle validation of the CL hypothesis, systematically comparing pregnancy loss risks across fresh transfers, FET-NC, and CL-absent FET protocols.
Independent Predictor Integration: It elevates CL status from a qualitative grouping variable to an independently quantifiable predictor within a multivariable risk stratification model.
Temporal Validation: Unlike many prior studies relying on internal cross-validation, this work employs an independent temporal validation set (2022–2026) to assess true generalizability and account for temporal distributional drift.
Risk Stratification Tool: The development of a parsimonious 7-variable model that effectively stratifies patients into distinct risk tiers, offering a clinical decision aid that outperforms age alone.

Results

Loss Rate Gradient: A clear gradient in pregnancy loss rates was observed, consistent with the CL hypothesis.
- CL-Present Groups: Fresh ET (14.9%) and FET-NC (13.3%) showed the lowest and comparable loss rates ( $P=0.689$ ).
- CL-Absent Groups: FET-GnRHa + HRT (18.9%) and FET-HRT (19.8%) showed significantly higher loss rates ( $P < 0.001$ for four-group trend).
- Pooled Comparison: The pooled CL-present group (14.8%) had a 4.6 percentage point lower loss rate than the CL-absent group (19.4%) ( $P < 0.001$ ).
Multivariable Analysis: After adjustment, CL absence remained an independent risk factor for pregnancy loss (adjusted Odds Ratio [aOR] = 1.128; 95% CI 1.029–1.236; $P=0.010$ ).
Model Performance:
- The 7-variable model achieved an AUC of 0.6464 in the independent temporal validation set.
- This significantly outperformed an age-only model (AUC = 0.6011; $\Delta$ AUC = 0.0453; DeLong $P < 0.001$ ).
- The model successfully stratified patients into four risk tiers with observed loss rates ranging from 3.2% (low risk) to 39.1% (very high risk), representing a 12-fold gradient.
Embryo Morphology: Contrary to expectations, embryo morphology did not provide significant independent predictive value for loss risk once clinical pregnancy was established, suggesting a "ceiling effect" where maternal factors dominate post-implantation outcomes.
Cryopreservation Effect: While FET-NC (CL-present + cryopreservation) had low loss rates, the small sample size ( $n=143$ ) prevented the statistical isolation of the cryopreservation effect from the CL effect.

Significance and Claims
The paper claims that CL absence is a robust, independent risk factor for early pregnancy loss across ART cycles. The primary significance lies in the development of a risk stratification model that integrates CL status with routine clinical factors, demonstrating that this integration significantly improves discrimination over age alone.

The authors position the model not as a precise individual probability predictor (due to suboptimal individual-level calibration) but as a risk stratification tool to identify patient tiers. This tool supports clinical decision-making, particularly in selecting endometrial preparation protocols: prioritizing Natural Cycle protocols for ovulatory patients to preserve endogenous CL function, and considering GnRHa + HRT over pure HRT when NC is not feasible.

The study acknowledges limitations, including the single-center retrospective design, potential confounding by indication, and the inability to fully separate cryopreservation effects due to the small FET-NC sample size. The authors emphasize that while the model offers a clear risk gradient and outperforms age-based assessment, future multicenter prospective studies are required to confirm causality and further refine model performance.

Technical Summary: Corpus Luteum Absence as an Independent Risk Factor for Early Pregnancy Loss After ART

More like this