IVF outcomes of women with discrepancies between age and serum anti-Müllerian hormone levels

This study was a single-center retrospective cohort study that consecutively included women who underwent their first IVF cycles in Center for Reproductive Medicine, Shandong University, from March 2013 to June 2014. Of 9431 women, 7283 women of reproductive age were young (< 35 yr), and the other 2148 were at an advanced age (≥35 yr). In the young group, participants with low AMH levels (the 0-25th percentage, 0.01–1.32 ng/ml) were set as the exposed group (young low-AMH group, n = 1819), while those with medium AMH levels (the 25-75th percentage, 1.32–3.99 ng/ml) and high AMH levels (the 75-100th percentage, 3.99–22.05 ng/ml) were set as the unexposed group (young average-AMH group, n = 3642; young high-AMH group, n = 1822). In the advanced-aged group, women with high (the 75-100th percentage, 2.41–22.05 ng/ml) AMH levels were set as exposed group (older high-AMH group, n = 537), and women with medium (the 25-75th percentage, 0.63–2.41 ng/ml) AMH levels and low (the 0-25th percentage, 0.01–0.62 ng/ml) were set as the unexposed (older average-AMH group, n = 1074; older low-AMH group, n = 537) groups, respectively. Women who had eggs frozen or used donated eggs were excluded.

The serum AMH levels were measured before the first ovarian stimulation cycle. We excluded the factors which that may affect the AMH level, such as pregnancy, smoking and combined oral contraceptives. All blood samples were separated into aliquots within 2 h after collection and frozen at − 80 °C until use. Tests to determine the serum AMH levels were performed in batches using enzyme-linked immunosorbent assay (ELISA) (Ansh Labs, Webster, USA). The intra- and inter-assay coefficients of variation were below 10%.

Different ovarian stimulation protocols were applied in this study and included the natural cycle, gonadotrophin-releasing hormone (GnRH) agonist long protocol [16], and flexible GnRH antagonist protocol [17], which have been described in detail previously. Briefly, for GnRH short protocols, the women received GnRH agonist (Diphereline, Ipsen) 0.1 mg daily starting on day 2 or 3 of the menstrual cycle. Recombinant FSH (Gonal-F, Merk Serono) was started after 2 days. The dosage was adjusted according to the ovarian response. Cycle monitoring was performed by ultrasonography and serum sex steroid hormone measurements. Ovulation triggering was implemented with a dose of 4000 to 8000 IU urinary human chorionic gonadotropin (HCG) when two or more follicles measured 18 mm. Collected oocytes were inseminated through IVF or ICSI. The later was performed when a total sperm count of < 500 000 after gradient centrifugation [18]. Embryos were cultured up to day 3 or day 5, and then, one or two embryos were transferred under ultrasound guidance. Intramuscular progesterone at a daily dose of 80 mg or dydrogesterone tablets (Duphaston) 10 mg two times daily were used by injection or orally for luteal phase support, from the day of oocyte retrieval until 10 weeks after conception.

Surplus embryos were cryopreserved by vitrification. Endometrial preparation was performed through either a natural cycle, an artificial cycle, or ovulation induction cycles. The natural cycle was applied to women with regular menstruation and normal ovulation in the B-ultrasound (US). The natural cycle was monitored by US. The artificial cycle was used for women with oligo/anovulation. Oral estradiol valerate (Progynova, Delpharm Lille) was administered for endometrial preparation. The protocol was the same as that described in our previous study [17]. For women in whom a satisfactory endometrium was not achieved using the above two protocols, we performed ovulation induction using low-dose HMG as an alternative. When the endometrial thickness reached 8 mm, intramuscular progesterone was added at a dose of 80 mg per day. In addition, one or two frozen embryos (day 5 or day 3) were thawed and transferred. The luteal-phase support continued until 10 weeks after conception.

The AFC (antral follicle count) was defined using a 2D technique. Clinical pregnancy was considered as one or more gestational sacs visualized by ultrasound examination [17]. Miscarriage was considered as any spontaneous or therapeutic pregnancy loss during clinical pregnancy [17]. The inclusion of several treatment cycles for one subject the analysis would lead to bias. In our study, only the first treatment cycle was considered. Thus, the starting point for the LBR and CLBR was the first live birth in the first treatment cycle, and the follow-up examinations were ended when all fresh and frozen embryos derived from the first treatment cycle were transferred. Live birth was defined as delivery of at least one viable infant with a gestational age equal to or greater than 28 weeks [17]. The LBR per transfer was calculated by dividing the number of newborns by the transfer times [17]. Cumulative live birth was defined as the occurrence of live birth after the transfer of all fresh and frozen embryos derived from the first stimulation cycle. Each subject was included only once in the analysis.

Since not all the participants had used all of their embryos within the cycle by the time of the data collection, conservative and optimistic CLBRs were calculated as well. The conservative estimates assumed that women who did not return for frozen ET would not have the outcome of a live birth [19]. The 95% confidence interval (CI) was calculated using the standard error from the binomial distribution. The optimistic CLBR assumed that women who did not return for frozen ET would have similar outcomes to those who continued with frozen-thawed ET [19]. The pointwise optimal estimates and 95% CIs were assessed by the Kaplan–Meier estimate.

Categorical data were represented as percentages and frequencies. Continuous data are expressed as the mean ± standard deviation (SD) and were tested for normality using the Kolmogorov-Smirnov test.

A 2 × 3 factorial analysis was conducted to test the effects of age and AMH and their interaction value. P < 0.05 was considered to be statistically significant. Pairwise comparisons were performed when the interaction p value was < 0.05. In women with age and AMH discrepancies, chi-square test or Fisher’s exact test was used for categorical data comparisons, and data with normal distributions were compared between groups using Student’s t-test. P < 0.05 was considered to be statistically significant. We used SPSS 23.0 for statistical analysis (SPSS Inc., USA).

The IVF outcomes of each age-stratified group are listed in Table 2, Additional file 1: Table S1 and Table S2, Additional file 2: Figure S1. In young group, the number of oocytes retrieved (9.15 ± 5.09 vs 13.13 ± 5.49 vs 16.04 ± 6.44, p < 0.01 for all pairwise comparisons), the number of 2 pronuclear zygotes (2PN, 5.32 ± 3.59 vs 7.88 ± 4.21 vs 9.31 ± 4.59, p < 0.01 for all pairwise comparisons), and the frequency of good-quality embryos (GQE, 2.42 ± 2.53 vs 3.50 ± 3.22 vs 4.08 ± 3.78, p < 0.01 compared with the average- and high-AMH groups, respectively) were significantly lower in low-AMH group than in the average-AMH and high-AMH groups. The clinical pregnancy rate (CPR, 58.01% vs 62.90% vs 66.33%, p < 0.01 for all pairwise comparisons), cumulative clinical pregnancy rate (CCPR, 65.37% vs 78.31% vs 83.15%, p < 0.01 for all pairwise comparisons), LBR (48.52% vs 54.65% vs 56.68%, p < 0.01 compared with the average- and high-AMH groups, respectively), and CLBR (56.35% vs 69.99% vs 72.99%, p < 0.01 for compared with the average- and high-AMH groups, respectively) were also decreased. It was shown that age, but not AMH, was the risk factor for miscarriage (Age: OR = 1.05, 95% CI = 1.02–1.08, p < 0.01; AMH: OR = 1.01, 95% CI = 0.97–1.05, p = 0.56, Additional file 1: Table S3).

Among advance-aged groups, the number of oocytes retrieved (12.91 ± 5.73 vs 8.45 ± 4.39 vs 4.53 ± 3.25, p < 0.01 for all pairwise comparisons), the number of 2PN zygotes (7.79 ± 4.39 vs 5.15 ± 3.53 vs 2.70 ± 2.30, p < 0.01 for all pairwise comparisons), the frequency of GQE (3.60 ± 3.32 vs 2.28 ± 2.36 vs 1.31 ± 1.54, p < 0.01 for all pairwise comparisons) were significantly higher in the high-AMH group than in the low-AMH group. Pregnancy outcomes including the CPR (49.44% vs 41.45% vs 33.90%, p < 0.01 for all pairwise comparisons), CCPR (66.11% vs 49.07% vs 30.73%, p < 0.01 for all pairwise comparisons), and CLBR (52.89% vs 39.29% vs 20.11%, p < 0.01 for all pairwise comparisons) were also significantly higher in older women with high AMH levels. The frequency of LBR in the older high-AMH group was significantly higher (37.45% vs 20.34%, p < 0.01) than that in the young low-AMH group, but there was no difference (37.45% vs 32.46%, p = 0.11) compared with the average-AMH group. Moreover, the MR in the high-AMH group was significantly lower (21.50% vs 35.56%, p < 0.01) than that in the low-AMH group, while no difference (21.50% vs 21.68%, p = 0.97) was found compared with their average-AMH counterpart. Age, but not AMH, was correlated with MR in women of advanced age (Age: OR = 1.36, 95% CI = 1.26–1.49, p < 0.01; AMH: OR = 1.00, 95% CI = 0.91–1.10, p = 0.99) (Additional file 1: Table S3).

When the IVF outcomes were compared between the two groups with discrepancies between age and the AMH level (the young low-AMH group and older high-AMH group), it was found that the number of oocytes retrieved (9.15 ± 5.09 vs 12.91 ± 5.73, p < 0.01), the number of 2PN zygotes (5.32 ± 3.59 vs 7.79 ± 4.39, p < 0.01), and the frequency of GQE (2.42 ± 2.53 vs 3.60 ± 3.32, p < 0.01) were significantly lower in the young low-AMH group than in the older high-AMH group. However, the CPR (58.01% vs 49.44%, p < 0.01) and LBR (48.52% vs 37.45%, p < 0.01) was significantly higher in the former. The CCPR (65.37% vs 66.11%, p = 0.75) and CLBR (56.35% vs 52.89%, p = 0.15) were comparable between both groups. No difference was found in MR between the groups (16.37% vs 21.50%, p < 0.01) (Table 3).

The optimistic and conservative CLBRs of women in terms of age and the AMH level are presented in Fig. 1 and Additional file 1: Table S4. In the young women with low AMH levels, the optimistic CLBR increased along with the transfer time from 41.70% (95% CI: 39.37–44.12) for the first ET to 89.28% (95% CI: 81.58–94.76) for the fifth ET. In addition, the conservative CLBR increased from 37.99% (95% CI: 35.75–40.26) for the first ET to 55.91% (95% CI: 53.59–58.21) for the third ET. In the older women with high AMH levels, the optimistic CLBR increased from 36.31% (95% CI: 32.32–40.63) after the first ET to 81.75% (95% CI: 71.19–90.23) after the fifth ET. In addition, the conservative CLBR increased from 34.82% (95% CI: 30.79–39.02) after the first ET to 52.14% (95% CI: 47.82–56.44) after the third ET. After the third ET, the conservative CLBR in both discrepancy groups reached a plateau regardless of whether more ETs were performed. We calculated the ROC curves for prediction of live birth and cumulative live birth in low-AMH young and high-AMH elder groups. However, the predictive value was poor , which indicated that AMH had a significant association with live birth but was a poor independent predictor for live birth (Additional file 3: Figure S2).