Study selection

As shown in Fig. 1, 650 studies were identified by initial databases and manual searches. A total of 203 studies were removed due to duplicate records, and 365 were excluded after assessing their titles and abstracts. The remaining 82 studies underwent full-text review, and 63 studies were not eligible for inclusion. Finally, 19 studies fulfilled the eligibility criteria and were included in the qualitative analysis, with 10 included in the quantitative analysis (meta-analysis).

Fig. 1

PRISMA flow diagram of the study selection process

Study characteristics

Table 1 summarizes the characteristics of the 19 studies. All 19 studies were cohort studies, and the NOS score ranged from 5 to 9 (median 6). The classifications of the level of serum AMH were not consistent among these studies. Seven studies classified participants into the low-AMH group (the 0-25th percentile), average-AMH group (the 25-75th percentile), and high-AMH group (the 75-100th percentile) [13, 15, 16, 19,20,21,22]; three studies classified participants into the 75–90th percentile group and 90-100th percentile group [14, 23, 24]. The remaining studies did not classify participants into different groups based on AMH levels but analyzed AMH levels as a continuous variable [22, 25,26,27,28,29,30,31,32,33] and thus were not included in the quantitative analysis (meta-analysis). All 19 studies used the Rotterdam criteria for the diagnosis of PCOS.

Table 1 Characteristics of the included studiesPrimary outcomesClinical pregnancy

Seven studies analyzed clinical pregnancy; six stratified their patients into a low-AMH group (the 0-25th percentile), average-AMH group (the 25-75th percentile), and high-AMH group (the 75-100th percentile) [13, 15, 16, 19,20,21]. The overall clinical pregnancy rate was 48.5% (414/853) and 55.0% (473/860) in the high-AMH group and the low-AMH group, respectively. Meta-analysis demonstrated that the odds of a clinical pregnancy were significantly lower if patients were classified in the high-AMH group (OR: 0.77, 95% CI: 0.64–0.93) (Fig. 2). There was a moderate degree of heterogeneity (I2 = 46%). In addition, Du et al. [25] examined clinical pregnancy rate in a cohort of 200 PCOS patients aged 25 to 36 years undergoing IVF-ET. Patients were divided into two different groups based on a cutoff AMH level of 6.99 ng/L. Their results showed that clinical pregnancy rate was significantly lower if AMH was greater than 6.99 ng/L (p = 0.001). The result of the dose–response meta-analysis is shown in Fig. 4a. We observed an inverse linear association between prepregnancy serum AMH level and clinical pregnancy in PCOS patients (p = 0.008). The heterogeneity was not significant (I2 = 44.4%). The OR (95% CI) of clinical pregnancy was 0.996 (0.994, 0.999) per 1% increase in the AMH percentile.

Fig. 2

Meta-analysis results for primary outcomes. A Forest plot for clinical pregnancy rate. B Forest plot for live birth rate

Live birth

Seven studies analyzed live birth; three studies stratified their patients into a low-AMH group (the 0-25th percentile), average-AMH group (the 25-75th percentile), and high-AMH group (the 75-100th percentile) [13, 15, 20]. The incidence of live birth (per treatment cycle) was 49.9% (381/763) in the low-AMH group and 48.9% (315/643) in the high-AMH group. Meta-analysis demonstrated that the odds of a live birth were significantly lower if patients were classified into the high-AMH group than if they were classified into the low-AMH group (OR: 0.71; 95% CI: 0.58–0.87) (Fig. 2). Heterogeneity was not detected (I2 = 0%). The result of the dose–response meta-analysis is shown in Fig. 4b. We observed an inverse linear association between prepregnancy serum AMH level and live birth in PCOS patients (p < 0.001). The heterogeneity was not significant (I2 = 0%). The OR (95% CI) of live birth was 0.995 (0.993, 0.998) per 1% increase in the AMH percentile.

This section contains a summary of findings that cannot be meta-analyzed. Ho et al. examined live birth rate in a cohort of 921 women with PCOS who underwent IVM priming with hCG [27]. While high AMH levels do indicate a high number of oocytes and a high oocyte maturation rate, univariate analysis did not reveal an association between AMH level and live birth after the transfer of the first embryo after IVM (OR: 1.02; 95% CI: 0.98–1.06). Notably, this result may be applicable only to the specific IVM technique used in this study, namely, the transfer of day-2 embryos, which is not a regular practice at many IVF centers. Tabibnejad et al. [26] investigated the relationship between serum AMH levels and ICSI outcomes in 50 PCOS patients with 289 embryos. In this scenario, their findings suggested that AMH was not an accurate predictor of a live birth (AUC = 0.59 [95% CI, 0.42–0.76]). However, among women with tubal factor infertility, AMH had a moderate predictive value for a live birth (AUC = 0.70 [95% CI, 0.55 to 0.85]). Guan et al. [28] analyzed the cumulative live birth rate in 160 PCOS patients of advanced age (≥ 35 years). All patients underwent their first fresh cycles and subsequent frozen cycles within one year. Their results demonstrated that patients with an AMH level above 32.12 pmol/L were likely to have a 72% (HR, 1.72; 95% CI, 1.08–2.73, p = 0.023) and 34% (HR, 1.34; 95% CI, 1.07–1.68, p = 0.010) improvement in cumulative live birth rate compared to those with AMH levels below 7.85 pmol/L and 7.85–32.12 pmol/L, respectively. Acharya et al. [32] divided their patients based on an AMH cutoff level of 12 ng/ml. Their results showed that AMH was negatively associated with live birth (OR, 0.93; 95% CI, 0.90–0.96) up to an AMH level of 12 ng/ml. Beyond 12 ng/ml, the association was attenuated (OR, 1.01; 95% CI, 0.99–1.04).

Ovarian hyperstimulation syndrome

Three studies analyzed the incidence of OHSS. Tal et al. [15] reported a retrospective cohort study in a sample of 184 women with PCOS who underwent their first fresh IVF/ICSI cycles. Women were stratified into 3 groups according to the 0-25th (< 3.32 ng/ml), 25-75th (3.32–8.27 ng/ml), or 75-100th (> 8.27 ng/ml) percentile of serum AMH concentration. The stimulation protocol included either pituitary downregulation via GnRH agonist in a long protocol or a GnRH antagonist to prevent premature ovulation. No difference regarding the OHSS incidence was found among the three groups. When Kamel et al. [33] divided patients into two groups (AMH cutoff value: 4.6 ng/ml) and used GnRH antagonist, they found that the incidence of severe OHSS was significantly higher in patients with AMH > 4.6 ng/ml (p = 0.026). In addition, Muharam et al. [30] tried to determine the cutoff value of AMH to predict hyperresponse in PCOS patients undergoing controlled ovarian stimulation. The AUC of the ROC curve was 0.626 (95% CI; sensitivity: 71.8%; specificity: 52.7%), indicating poor predictive quality.

Secondary outcomesE2 on day of hCG

Four studies analyzed E2 on the day of hCG and stratified their patients into a low-AMH group (the 0-25th percentile), average-AMH group (the 25-75th percentile), and high-AMH group (the 75-100th percentile) [13, 15, 16, 21]. The pooled results found a null association between serum AMH levels and E2 on the day of hCG (SMD: 0.12; 95% CI: -0.98–1.23) (Fig. 3a). Sensitivity analysis demonstrated that the study by Kaya et al. [21] affected the robustness of the meta-analysis. After excluding this study, the difference in E2 on the day of hCG became significant (SMD: 0.63; 95% CI: 0.13–1.14). The result of the dose–response meta-analysis is shown in Fig. 4c. Using a linear model, we did not observe a dose–response association between prepregnancy serum AMH level and E2 on the day of hCG in PCOS patients (p = 0.901).

Fig. 3

Meta-analysis results for secondary outcomes. A Forest plot for E2 on hCG day, B forest plot for number of oocytes retrieved, C forest plot for number of MII oocytes, D forest plot for fertilization rate, E forest plot for implantation rate, and F forest plot for preterm birth

Fig. 4

Dose–response meta-analysis results for A Clinical pregnancy rate, B Live birth rate, C E2 on hCG day, D Number of oocytes retrieved, E Number of MII oocytes, F Fertilization rate, G Implantation rate

Number of oocytes retrieved

Eight studies analyzed the number of oocytes retrieved: four stratified their patients into a low-AMH group (the 0-25th percentile), an average-AMH group (the 25-75th percentile), and a high-AMH group (the 75-100th percentile) [13, 15, 16, 21]. The meta-analysis showed a significantly increased number of oocytes retrieved in women with PCOS who were classified into the high-AMH group compared with the low-AMH group (SMD: 0.90, 95% CI: 0.30–1.51) in a random-effects model. There was a high degree of heterogeneity (I2 = 87%) (Fig. 3b). The result of the dose–response meta-analysis is shown in Fig. 4d. Using a quadratic model, we observed a nonlinear inverted U-shaped association between prepregnancy serum AMH level and the number of oocytes retrieved in PCOS patients (p = 0.002). PCOS patients with an AMH level between the 39th percentile and the 97th percentile had a significantly increased number of oocytes retrieved. The heterogeneity was significant (I2 = 99.1%).

This section contains a summary of findings that cannot be meta-analyzed. Ho et al. [27], using the study design described earlier in this paper, reported that AMH showed a significant positive correlation with the number of oocytes by univariate analysis (coefficient = 0.28; p = 0.001). Similarly, Tabibnejad et al. [26] prospectively evaluated the number of oocytes retrieved in a cohort of 50 PCOS patients undergoing ICSI. They also found a positive correlation (Spearman’s r = 0.45, p = 0.001). However, when using multivariable analysis in a sample of 59 PCOS patients, Chen et al. [29] found a null association (r = -0.059, p = 0.685). Arslanca et al. [14] reported the same outcome in a cohort of 110 PCOS patients who underwent FET. Patients were categorized into the AMH 75-90th percentile group (n = 66) and 90-100th percentile group (n = 44), and no significant differences in terms of the number of oocytes retrieved were noted between the two groups.

Number of MII oocytes

Three studies analyzed the number of MII oocytes. Two stratified their patients into a low-AMH group (the 0-25th percentile), an average-AMH group (the 25-75th percentile), and a high-AMH group (the 75-100th percentile) [13, 21]. Although there was a trend toward a greater number of MII oocytes in the high-AMH group (SMD: 1.85, 95% CI: -1.07–4.78), the difference did not reach significance in a random-effects model. There was a high degree of heterogeneity (I2 = 97%) (Fig. 3c). In addition, Tabibnejad et al. [26] reported a significant positive correlation between AMH concentration and the number of MII oocytes (Spearman r = 0.42, p = 0.002). The result of the dose–response meta-analysis is shown in Fig. 4e. Using a linear model, we did not observe a dose–response association between prepregnancy serum AMH level and the number of MII oocytes in PCOS patients (p = 0.206).

Fertilization

Six studies reported fertilization rate, and four of those studies stratified their patients into a low-AMH group (the 0-25th percentile), average-AMH group (the 25-75th percentile), and high-AMH group (the 75-100th percentile) [13, 15, 16, 21]. The incidence of overall fertilization (per treatment cycle) was 61.9% (5656/9134) in the low-AMH group and 60.0% (6688/11140) in the high-AMH group. The meta-analysis demonstrated significantly decreased odds of fertilization in women with PCOS whose AMH was classified in the high-AMH group in a fixed-effects model (OR 0.92, 95% CI 0.87–0.98) (Fig. 3d). There was a moderate degree of heterogeneity (I2 = 36.1%). In addition, Du et al. [25] and Arabzadeh et al. [31] reported a null association between AMH and the fertilization rate (p > 0.05). The result of the dose–response meta-analysis is shown in Fig. 4f. We observed an inverse linear association between prepregnancy serum AMH level and fertilization in PCOS patients (p = 0.027). The heterogeneity was not significant (I2 = 44%). The OR (95% CI) of fertilization was 0.999 (0.998, 1.000) per 1% increase in the AMH percentile.

Implantation

Five studies analyzed implantation rate, and three of those studies stratified their patients into a low-AMH group (the 0-25th percentile), average-AMH group (the 25-75th percentile), and high-AMH group (the 75-100th percentile) [15, 16, 21]. The overall incidence of implantation was 39.7% (69/174) in the low-AMH group and 31.7% (52/164) in the high-AMH group. The meta-analysis found no association between AMH level and implantation rate (OR: 0.82, 95% CI: 0.28–2.39) (Fig. 3e). There was a high degree of heterogeneity (I2 = 79%). Du et al. [25] reported that the incidence of implantation was 41.1% (37/90) in the low-AMH group (< 6.99 ng/L) and 20.91% (23/110) in the high-AMH group (> 6.99 ng/L). Additionally, Arabzadeh et al. [31] reported that AMH was not associated with implantation rate in women with PCOS (r = -0.299, p = 0.138) but was positively associated with implantation rate in the non-PCOS control group (r = 0.305, p = 0.05). The result of the dose–response meta-analysis is shown in Fig. 4g. We did not observe a dose–response association between prepregnancy serum AMH level and implantation in PCOS patients (p = 0.735).

Cycle cancellation

Two studies analyzed cycle cancellation. Acharya et al. [32] stratified their patients with an AMH cutoff value of 12 ng/ml. In both groups, an increasing AMH level was associated with a higher probability of cycle cancellation (OR: 1.12, 95% CI: 1.10–1.15 and OR: 1.03, 95% CI: 1.01–1.05 in the AMH < 12 ng/ml group and > 12 ng/ml group, respectively). In the analysis of the reasons for cycle cancellation, the authors reported that in the AMH < 12 ng/ml group, each 1-unit increase in AMH level was associated with an 11% increase in the odds of embryo transfer cancellation because of the OHSS risk (OR, 1.11; 95% CI, 1.07–1.16). Xi et al. [16] stratified their 164 participants into 3 groups according to the < 25th (< 4.85 ng/ml), 25th to 75th (4.85–8.82 ng/ml), or > 75th (> 8.82 ng/ml) percentile of serum AMH concentration. Embryo transfers cancelled due to OHSS risk from the low-, middle-, and high-serum AMH groups were 1, 4 and 7 cases, respectively.

Obstetric outcomes

Three studies analyzed miscarriages. Liu et al. [20] examined this outcome in a cohort of 2973 infertile women, including 418 women with PCOS undergoing their first IVF treatments. The incidence of miscarriage was 8.1% (6/74) in the low-AMH group, 19.1% (29/152) in the average-AMH group, and 17.1% (12/70) in the high-AMH group. Although there were more high-quality embryos transferred in the average-AMH group than in the low-AMH group, the difference in miscarriage rate was not significant among these three groups, indicating that AMH was not associated with miscarriage rate among PCOS patients. Du et al. [25] also reported the early miscarriage rate in their two subgroups. The findings suggested that the rate of early miscarriage was significantly lower among the participants in the low-AMH group, with an incidence of early miscarriage of 6.67% (6/90) in the low-AMH group and 19.09% (21/110) in the high-AMH group (p < 0.001). Notably, this study investigated only the early miscarriage rate of patients, while the patient’s late pregnancy process was not studied.

GDM was reported in two studies [14, 25]. In the comparison of the GDM incidence between the serum AMH 75-90th percentile group and the 90-100th percentile group, there was no association between AMH and GDM. However, Du et al. [25] suggested that patients with AMH greater than 6.99 ng/ml had an increased incidence of GDM (p < 0.001). In addition, preeclampsia and PPROM were reported by only one study [14], with no association found.

Neonatal outcomes

Four studies reported preterm birth [14, 22,23,24]. Meta-analysis was performed to compare the preterm birth rate between the 75-90th percentile group and the 90-100th percentile group (AMH) (Fig. 3f), and a null association was found (OR: 1.92, 95% CI: 0.58–6.42). Meanwhile, Du et al. [22] reported that a higher AMH (75-100th percentile) was associated with an increased risk for preterm birth among women with a BMI ≥ 24 kg/m2 (OR: 2.10, 95% CI: 1.01–4.37) but not among women with a BMI < 24 kg/m2 (OR: 0.78, 95% CI: 0.35–1.73) after adjusting for multiple confounding factors, including maternal age, BMI, duration of infertility, and basal antral follicle count. This study also selected “small for gestational age”, “large for gestation age”, “low birth weight”, and “macrosomia” as outcomes of interest. In brief, no significant differences were found in the rates of these outcomes among patients in the different serum AMH groups (adjusted OR ranging from 0.91 to 1.13).