Antral follicle count, anti-mullerian hormone and inhibin B: predictors of ovarian response in assisted reproductive technology?

Objective The objective of this study was to evaluate the relationship between anti-mullerian hormone (AMH), inhibin B and antral follicle count (AFC) with ovarian response. Design Retrospective study. Setting Fertility unit. Sample AFC was recorded, and a serum sample obtained on day 3 from all patients undergoing in vitro fertilisation (IVF). Patients were given 300 IU/L recombinant follicle stimulating hormone (FSH; Gonal F). The following day blood samples were collected. Methods Serum samples were assayed for FSH, AMH and inhibin B using commercial immunoassay kits and oestradiol using an in house assay. Main outcome measures Response to gonadotrophin stimulation and the number of eggs collected. Results AFC was negatively correlated to age (r=-0.426, P < 0.001). Delta inhibin B (levels of inhibin B on day 4 minus day 3) had the best association to the number of eggs collected (r= 0.533, P < 0.001) followed by basal AMH (r= 0.51, P < 0.001) and AFC (r= 0.505, P < 0.001). The number of eggs fertilised was significantly associated with basal AMH (r= 0.592, P < 0.001) and inhibin B (r= 0.548, P < 0.001). AMH with a cutoff of 0.2 ng/mL had the best sensitivity (87%) and specificity (64%) in predicting poor response. A cumulative score using basal FSH, basal AMH, delta E-2 (levels of oestradiol on day 4 minus day 3), delta inhibin B, AFC and age gives the best predictive statistics to identify poor responders with 87% sensitivity and 80% specificity and a positive likelihood ratio of 4.36. Conclusion Delta inhibin B had the best positive association with the number of eggs collected and basal AMH is the single best predictor of poor response. AFC has a significant association with the number of eggs collected and is predictive of clinical pregnancy. It is evident that a single parameter is of limited value in predicting ovarian response. However, we have demonstrated a cumulative score using all the above markers could be useful in predicting poor response.