P-521 A gene expression risk signature of endometrial failure for prognosis in In Vitro Fertilization (IVF) patients

• Published in: Human reproduction (Oxford) Vol. 38; no. Supplement_1
• Main Authors: Diaz-Gimeno, P, Sebastian-Leon, P, Spath, K, Sanchez-Reyes, J M, Vidal, M C, Pellicer, N, Wells, D, Pellicer, A
• Format: Journal Article
• Language: English
• Published: 22-06-2023
• ISSN: 0268-1161; 1460-2350
• DOI: 10.1093/humrep/dead093.864

Abstract
Study question
It is possible to identify a characteristic pattern of endometrial gene expression indicative of implantation failure, which is independent of implantation window displacements?
Summary answer
An endometrial transcriptomic signature was able to identify patients with a > 3-fold increased risk of implantation/pregnancy failure with 95% accuracy.
What is known already
Implantation failure of endometrial origin is a complex and multifactorial symptom with diverse causes, diagnosed in IVF patients after repeated failures with good quality embryos. A generation of gene expression tools have assumed that Window of implantation (WOI) displacement is the principal cause of this condition, but strategies seeking to counteract this problem by adjusting the day of embryo transfer have not yielded convincing improvements in outcomes. However, it is conceivable that other forms of endometrial disruptions, relevant to embryo implantation, could exist. New endometrial diagnostic strategies are needed to understand, diagnose and potentially treat patients affected with such problems.
Study design, size, duration
A prospective multicenter study between January 2018 and December 2021 recruited 281 Caucasian IVF patients (mean age of 39.4±4.8 years and BMI of 22.9±3.5 kg/m2) undergoing hormone replacement therapy and encompassing 114.5±7.2 h of progesterone administration at time of endometrial biopsy. Following experimental quality controls and clinical follow up, 186 patients who had a good quality embryo transferred in the cycle after endometrial biopsy collection were included for gene discovery and prediction performance.
Participants/materials, setting, methods
The expression of 404 genes selected for their potential to distinguish endometrial timing and/or endometrial disruption was measured. Transcriptomic variation related to progression of the menstrual cycle was removed using transcriptomic endometrial dating (TED) and linear models. Study groups were established according to clinical and gene expression parameters through a semi-supervised artificial intelligence procedure. Gene signature discovery and a cross-validation processes were undertaken to define predictive expression patterns. Reproductive outcomes were compared between prognosis profiles.
Main results and the role of chance
We developed a procedure called Clinically Acute Transcriptomic Stratification (CATranS), combining clinical parameters and deep transcriptomic molecular characterisation to stratify patients according to endometrial prognosis: ‘poor’ (n = 137) or ‘good’ (n = 49). These two transcriptomic profiles were associated with differing reproductive outcomes in the single embryo transfer following biopsy: pregnancy rate (45.1% vs 79.6%, poor vs good prognosis, respectively, p = 3.8E-5); live birth rate (56.4% vs 97.5%, p = 3.0E-06); clinical miscarriage rate (27.9% vs 2.6%, p = 0.0020); biochemical pregnancy rate (20.4% vs 0%, p = 0.0023). Patients with a poor prognosis profile had 3.3-times higher relative risk of a transferred embryo failing to implant or a pregnancy not being sustained to term, compared with good prognosis patients. Initially, a reference dataset was used to build a prototype for diagnosing endometrial failure, revealing that a gene expression signature consisting of 135 genes was the most predictive. Prediction performance was estimated using a 5-fold 100-times cross-validation process, resulting in a median accuracy of 0.92, median sensitivity of 0.96, and median specificity of 0.84. From these 135 predictive genes, 122 were differentially expressed (FDR<0.05) in endometrial poor prognosis, 59 up- and 63 down-regulated, most involved in functional processes such as regulation (17%), metabolism (8.4%), immune response and inflammation (7.8%).
Limitations, reasons for caution
We describe a potential new strategy for evaluating endometrial competence. However, to confirm predictive value, validation using additional samples from patients independent of signature discovery set is required. Further research to identify potential treatments for patients classified as poor prognosis is needed, providing a tailored clinical pathway for these patients.
Wider implications of the findings
The prototype described is a novel concept, potentially leading to development of a new generation of tools for diagnosis of fertility problems related to endometrial factors. The ‘poor’ prognosis profile is not caused by asynchronies in menstrual cycle progression, opening the possibility of finding new treatment pathways for these patients.
Trial registration number
NA