A total of 10037 patients (≤ 35 years old) from January 2018 to March 2023 were analyzed for ovarian reserve, including precycle parameters (AMH and AFC) and postcycle parameters (oocyte yield and embryo yield). According to ROC analyses and the POSEIDON criteria, 4484 patients were included and further divided into three groups (1153 patients in the DOR group, 682 patients in the high-risk DOR group, and 2649 patients in the NOR group). The FET outcomes were followed up to December 31, 2023. The outcomes of 5841 ET cycles were analyzed. PSM was conducted to further confirm the differences in ovarian response and pregnancy outcomes between the high-risk DOR group and NOR group. A profile summary of the study is shown in Fig. 1.

Fig. 1

Flow chart of patient selection, grouping and comparisons

Patients grouped according to ROC curve analyses and POSEIDON criteria

In the present study, the ROC curve (Fig. 2A) of the number of oocytes retrieved for predicting one viable embryo (one top-quality embryo yield or one viable blastocyst) indicated that the threshold of the number of oocytes retrieved was 6.5 (AUC = 0.867, standard error = 0.013, 95% confidence interval: 0.842–0.891, p < 0.001, sensitivity: 0.875, specificity: 0.733, Youden’s index: 0.608). Additionally, to obtain 6.5 oocytes, ROC curves of the AMH (Fig. 2B) and AFC (Fig. 2C) prediction models suggested that the cutoff of AMH was 2.53 ng/mL (AUC = 0.847, standard error = 0.006, 95% CI: 0.829–0.853, p < 0.001, sensitivity: 0.751, specificity: 0.797, Youden’s index: 0.548) and the cutoff of AFC was 10.5 (AUC = 0.836, standard error = 0.006, 95% CI: 0.822–0.847, p < 0.001, sensitivity: 0.661, specificity: 0.876, Youden’s index: 0.537). According to the AMH and AFC prediction models in the present study and the POSEIDON criteria, patients were further categorized into three groups: (1) the DOR group (AMH values ≤ 1.2 ng/mL and/or AFC ≤ 5; n = 1153); (2) the high-risk DOR group (1.2 < AMH values < 2.5 and 5 ≤ AFC ≤ 10; n = 682); and (3) the NOR group (2.5 ≤ AMH values ≤ 5.5 and/or 11 ≤ AFC ≤ 20; n = 2249).

Fig. 2

Receiver operating characteristic (ROC) curve analyses of the young females younger than 35 years. (A) ROC analysis of the number of oocytes retrieved for predicting one viable embryo obtained (one top-quality embryo or one viable blastocyst); area under the ROC curve (AUC) 0.867 (95% CI: 0.842–0.891), sensitivity 0.875, specificity 0.733, Youden’s index 0.608, p < 0.001; (B) ROC curve of the number of oocytes retrieved for predicting the cutoff of AMH; AUC 0.847 (95%CI: 0.829–0.853), sensitivity 0.751, specificity 0.797, Youden’s index 0.548, p < 0.001; (C) ROC curve of the number of oocytes retrieved for predicting the cutoff of AFC; AUC 0.836 (95%CI: 0.822–0.847), sensitivity 0.661, specificity 0.876, Youden’s index 0.537, p < 0.001

Patient characteristics

The baseline characteristics of the patients among the three groups are shown in Table 1. The patients in the DOR group and in the high-risk DOR group were older than those in the NOR group (all p < 0.001). The patients in the high-risk DOR group had a longer duration of infertility than those did in the NOR group (p < 0.001). There were significant differences in the percentages of primary infertility between the DOR group and NOR group (50.1% vs. 44.1%, p < 0.014). As expected, there were significant differences on the number of AFC and AMH values among the three groups compared to each other (DOR group < high-risk DOR group < NOR group, all p < 0.001), and patients in the DOR group had the minimum mean number of AFC (6.47) and lowest mean value of AMH (0.952 ng/mL).

Table 1 Characteristics of patients undergoing IVF/ICSI among the three groupsFollicle development, and oocyte performance in oocyte aspiration cycles

The oocyte aspiration cycle characteristics of the patients among the three groups are shown in Table 2. There were significant differences in the number of retrieved oocytes (5.33 ± 3.790 vs. 8.28 ± 3.890 vs. 14.38 ± 5.765), MII oocytes (4.85 ± 3.297 vs.7.23 ± 3.559 vs. 12.47 ± 5.359), top-quality D3 embryos (1.96 ± 1.731 vs. 2.66 ± 2.036 vs. 4.43 ± 3.009), viable blastocysts in blastocyst culture cycles (1.27 ± 1.562 vs. 1.65 ± 1.870 vs. 2.98 ± 2.773), and viable blastocysts in oocyte aspiration cycles (0.64 ± 1.277 vs. 1.20 ± 1.757 vs. 2.74 ± 2.775) among the three groups (DOR group < high-risk DOR group < NOR group, p < 0.001). Compared with the high-risk DOR group and NOR group, the DOR had the highest percentages of cycles with no retrieved oocytes (3.38%, p < 0.001) and no MII oocytes (2.33%, p < 0.001), but it had the lowest percentages of cycles with viable embryos retrieved (84.39%, p < 0.001).

Table 2 Oocyte aspiration cycle characteristics of the patients among the DOR group, high-risk DOR group and NOR group

To evaluate the relative predictive value of group allocation for the number of cycles with viable embryos obtained and cycles with blastocyst culture, logistic regression analyses were performed for the respective populations of each comparison (Fig. 3). High-risk DOR was associated with greater odds of the number of cycles with viable embryos obtained (OR 4.321, 95%CI: 2.867–6.513, p < 0.001) and blastocyst culture cycles (OR 2.611, 95%CI: 2.131–3.199, p < 0.001) than the DOR group, and with lower odds of the number of cycles with viable embryos obtained (OR 0.277, 95%CI: 0.165–0.464, p < 0.001) and cycles with blastocyst culture (OR 0.236, 95%CI: 0.190–0.293, p < 0.001) than the NOR group.

Fig. 3

Univariate logistic regression analyses of the associations between group allocation and the outcomes of IVF/ICSI cycles among the three groups

Outcomes of fresh embryo transfer and frozen-thawed embryo transfer cycles

The outcomes of fresh ET and FET are presented in Table 3. Compared with those in the NOR group, the live birth rates per total ET, including fresh ET and FET, were significantly lower in the DOR group (39.7% vs. 47.8%, p < 0.001) and high-risk DOR group (43.2% vs. 47.8%, p = 0.019). There were significant differences in the percentages of CLBRs per oocyte aspiration cycle among the three groups (43.0% vs. 59.0% vs. 78.2%, p < 0.001), and the DOR group had the lowest percentage of CLBRs per oocyte aspiration cycle. To investigate the effects of embryo development stage and the effects of whether embryos were frozen, the outcomes of D3 embryos and blastocysts which were frozen-thawed or not, were analyzed separately. Compared with those in the NOR group, the live birth rates per frozen D3 ET were significantly lower in the DOR group (35.9% vs. 47.8%, p < 0.001) and high-risk DOR group (39.2% vs. 47.8%, p = 0.004), compared to that of NOR group.

Table 3 Pregnancy outcomes of embryo transfer cycles among the three groups

Considering that the number of fresh ET cycles was limited, we focused on the effects of embryo development stage (D3 embryo or D5/6 embryo) on IVF/ICSI outcomes. In D3 ET cycles, significant differences were detected in HCG-positive rates per ET (53.8% vs. 64.4% vs. 64.4%), clinical pregnancy rates per ET (45.6% vs. 56.3% vs. 56.3%), ongoing pregnancy rates per ET (39.5% vs. 43.1% vs. 50.2%), and live birth rates per ET (38.3% vs. 42.1% vs. 48.8%) among the three groups compared with each other (DOR group < high-risk DOR group < NOR group, p < 0.05). Interestingly, the significant difference in live birth rates disappeared in the D3 ET cycles of the positive HCG population. The pregnancy loss rates per ET or per positive HCG did not differ significantly among the three groups in the D3 ET cycles (p > 0.05). There were no significant differences in any of the pregnancy outcomes in the blastocyst ET cycles (p > 0.05).

Logistic regression analyses were further performed to evaluate the relative predictive value of group allocation for live birth chance in total ET cycles, cumulative live birth in oocyte aspiration cycles, the blastocyst ET number/D3 ET number ratio, and pregnancy outcomes of D3 and D5/6 ET (Fig. 3). Compared with the NOR group, the high-risk DOR group was associated with lower odds of live birth (OR 0.831, 95%CI: 0.714–0.968, p < 0.001). Compared with the NOR group, the high-risk DOR group was associated with lower odds of the blastocyst ET/D3 ET ratio (OR 0.450, 95%CI: 0.383–0.530, p < 0.05); Compared with the DOR group, the high-risk DOR group was associated with higher odds of the blastocyst ET number/D3 ET number ratio (OR 1.367, 95%CI: 1.114–1.678, p < 0.05). In oocyte aspiration cycles, compared with the NOR group, the high-risk DOR group was associated with lower odds of cumulative live birth cases (OR 0.401, 95%CI: 0.332–0.486, p < 0.001), whereas the high-risk DOR group was associated with higher odds of cumulative live birth cases compared with the DOR group (OR 1.911, 95%CI: 1.558–2.344, p < 0.001). Compared with the NOR group, high-risk DOR was associated with lower odds of HCG positive cases (OR 0.726, 95%CI: 0.604–0.874, p < 0.001), ongoing pregnancy cases (OR 0.750, 95%CI: 0.624–0.901, p = 0.002) and live birth cases (OR 0.764, 95%CI: 0.635–0.918, p = 0.005) in D3 ET cycles. In D5 ET cycles, group allocation was not associated with pregnancy outcomes in the respective population of each comparison.

Comparison between high-risk DOR and NOR after propensity score matching

As shown in Table 4, the basic parameters, including age, infertility duration, BMI, and infertility type, were not significantly different after PSM (p > 0.05). Table 4 showed that the high-risk DOR group had a significantly lower number of retrieved oocytes (8.23 ± 3.971 vs. 14.54 ± 5.822, p < 0.001), a lower number of MII oocytes (7.16 ± 3.633 vs. 12.67 ± 5.390, p < 0.001) and a lower number of top-quality D3 embryos (2.60 ± 2.03 vs. 4.26 ± 3.02, p < 0.001) than did the NOR group. A lower number of cycles with blastocyst culture was observed in the high-risk DOR group than in the NOR group (p < 0.001). In the cycles with blastocyst culture, the number of blastocysts was significantly lower in the high-risk DOR group than in the NOR group (1.55 ± 1.850 vs. 2.18 ± 2.477, p < 0.001). The number of viable embryos (3.47 ± 2.02 vs. 5.43 ± 2.61, p < 0.001), including D3 embryos (2.36 ± 1.26 vs. 3.52 ± 1.57, p < 0.001) and blastocysts (1.11 ± 1.72 vs. 1.91 ± 2.44, p < 0.001), that could be frozen and transferred in the high-risk DOR group was lower than that in the NOR group. The percentages of cycles with no MII oocyte obtained (1.3% vs. 0.2%, p = 0.038) and no embryos obtained (4.6% vs. 1.0%, p < 0.001) were significantly greater in the high-risk DOR group than in the NOR group.

Table 4 Baseline characteristics of patients and oocyte aspiration cycle characteristics between the high-risk DOR group and NOR group after propensity score matching

Logistic regression analyses further revealed (Fig. 4) that the high-risk DOR group was associated with lower odds of the number of cycles with viable embryos obtained (OR 0.206, 95%CI: 0.085–0.502, p < 0.001) and with lower odds of the number of cycles with blastocysts retrieved (OR 0.550, 95%CI: 0.438–0.692, p < 0.001) than the NOR group was. The difference in the normal fertilization rates between the groups was not significant, although the methods of insemination differed between the groups.

Fig. 4

Univariate logistic regression analyses of the associations between group allocation and the outcomes of IVF/ICSI cycles in the high-risk DOR group and NOR group

The pregnancy outcomes of fresh ET and FET between the high-risk DOR group and NOR group after PSM are presented in Table 5. Compared with the NOR group, the high-risk DOR group presented a lower number of mean ET cycles(1.285 ± 0.677 vs. 1.519 ± 0.854, p < 0.002), lower live birth rates per total number of ET cycles (45.9% vs. 51.1%, p = 0.035), and lower CLBRs per number of oocyte aspiration cycles (59.0% vs. 77.6%, p < 0.001). Owing to the lower number of blastocysts obtained in the high-risk DOR group, the percentages of D3 ET cycles (74.3% vs. 69.4, p = 0.027) differed between the two groups. In the D3 ET cycles, compared with the NOR group, the high-risk DOR group had a significantly lower number of transferred embryos (1.88 ± 0.334 vs. 1.95 ± 0.219, p < 0.001), lower HCG-positive rates per ET (57.5% vs. 65.1%, p = 0.007), lower clinical pregnancy rates per ET (50.5% vs. 58.2%, p = 0.008), lower ongoing pregnancy rates per ET (44.8% vs. 52.7%, p = 0.007), and lower live birth rates per ET (43.9% vs. 51.3%, p = 0.011). No significant difference was found in the pregnancy loss rates between the two groups in the D3 ET cycles. In the blastocyst ET cycles, the number of transferred embryos was significantly lower in the high-risk DOR group than in the NOR group (1.10 ± 0.295 vs. 1.28 ± 0.451, p < 0.001). The HCG-positive rates per ET, clinical pregnancy rates per ET, ongoing pregnancy rates per ET, live birth rates per ET and pregnancy loss rates per ET did not differ between the two groups in terms of the number of blastocyst ET cycles. Interestingly, in the HCG positive cycles, the pregnancy outcomes, including HCG-positive rates per ET, clinical pregnancy rates per ET, ongoing pregnancy rates per ET and live birth rates per ET, did not differ between the two groups in the D3 ET cycles and blastocyst ET cycles.

Table 5 Pregnancy outcomes of embryo transfer cycles between the high-risk DOR group and NOR group after propensity score matching

Logistic regression analyses were performed between the high-risk DOR group and NOR group after PSM (Fig. 4). Compared with NOR group, the high-risk DOR group was associated with lower odds of HCG-positive cases (OR 0.725, 95%CI: 0.574–0.913, p = 0.007), clinical pregnancy cases (OR 0.734, 95%CI: 0.585–0.920, p = 0.008), ongoing pregnancy cases (OR 0.729, 95% CI: 0.581–0.913, p = 0.007), and live birth cases (OR 0.745, 95%CI: 0.594–0.933, p = 0.011) in D3 ET cycles. In D5 ET cycles, group allocation was not associated with pregnancy outcomes between the high-risk DOR group and NOR group. Compared with the NOR group, the high-risk DOR was associated with lower odds of live birth cases in all ET cycles (OR 0.813, 95%CI: 0.671–0.985, p = 0.035). Compared with the NOR group, the high-risk DOR had lower odds of cumulative live births in oocyte aspiration cycles (OR 0.416, 95%CI: 0.323–0.534, p < 0.001).