Predictive value of androgens and multivariate model for poor ovarian response

Area of Science:

• Reproductive Endocrinology
• Infertility Research
• Ovarian Physiology

Background:

• Predicting poor ovarian response (POR) remains challenging, with existing models lacking sufficient sensitivity and specificity.
• Androgens are implicated in folliculogenesis, suggesting their potential role in predicting ovarian response.
• The Bologna criteria define POR, but effective predictive models are still needed.

Purpose of the Study:

• To evaluate dehydroepiandrosterone sulphate (DHEAS) and basal testosterone as predictors of POR.
• To develop and validate a multivariate model for predicting POR and clinical pregnancy.
• To compare the predictive efficacy of the multivariate model against antral follicle count (AFC).

Main Methods:

• Retrospective analysis of 79 poor responders and 128 normal responders undergoing ovarian stimulation.
• Measurement of serum FSH, LH, oestradiol, DHEAS, and testosterone on day 3 of the menstrual cycle.
• Development of a multivariate predictive model using age, AFC, FSH, FSH/LH, and testosterone.

Main Results:

• DHEAS levels did not differ significantly between poor and normal responders or pregnant and non-pregnant women.
• Basal testosterone showed limited predictive ability for POR as a single predictor.
• The multivariate model demonstrated high predictive accuracy for POR (ROC AUC = 0.976) and pregnancy (ROC AUC = 0.716), outperforming AFC.

Conclusions:

• Basal testosterone, unlike DHEAS, has some predictive value for POR, but is insufficient alone.
• A multivariate model incorporating age, AFC, FSH, FSH/LH, and testosterone reliably predicts POR and clinical pregnancy.
• This novel multivariate model offers a promising tool for identifying patients at risk of POR and optimizing fertility treatment protocols.