Introduction

Frozen embryo transfer (FET), which is fueled by significant advancements in vitrification techniques, coupled with its expanding indications, has emerged as an integral component of assisted reproductive technology (ART) [1,2]. The “freeze-all” strategy can reduce the risk of ovarian hyperstimulation syndrome, enhance endometrial receptivity, and allow for preimplantation genetic testing, all of which have contributed to the rise of FET [3–6]. However, the successful of FET depends on the precise synchronization between embryo developmental stage and endometrial receptivity, which depends on choosing the optimal endometrial preparation protocol [7].

The predominant methods for endometrial preparation in current practice include the natural cycle (NC), modified natural cycle (mNC), and artificial cycle with hormone replacement therapy (HRT). Both the NC and the mNC protocols rely on endogenous hormone production and corpus luteum formation. The mNC approach is considered more flexible due to the active intervention that to promotes the maturation of the dominant follicle. However, the NC and the mNC protocols may lead to a relatively high cycle cancellation rate, up to 20%, reported in some studies, due to the absence of the dominant follicle [8]. It is worth emphasizing that the HRT protocol provides better control over endometrial development by administering estradiol and progesterone externally, it may increase the risks of obstetric and perinatal complications due to the lack of a corpus luteum [9–11].

Despite numerous randomized controlled trials (RCT) comparing the pregnancy outcomes and maternal-fetal safety of these endometrial preparation methods, the superiority of one approach over another remains elusive, mainly due to the heterogeneity in study designs, specifically in monitoring protocols, use of ovulation triggers, and luteal phase support (LPS). A recent systematic review and meta-analysis assessing obstetric and neonatal outcomes following NC and HRT endometrial preparation, encompassing 30 studies with over 110,000 patients, revealed that the NC group receiving LPS had a lower risk of preterm birth compared to the HRT group while NC without LPS did not. However, the interpretation of these findings could not lead to a definite conclusion due to the substantial heterogeneity across studies and the low to moderate quality of evidence [12].

In HRT cycles, the sole reliance on vaginal micronized progesterone may result in suboptimal absorption and lower serum progesterone concentrations in a subset of patients, with detrimental effects on pregnancy outcomes following FET [13,14]. To circumvent this drawback, it has been advocated to monitor serum progesterone prior to embryo transfer. Rescue protocols involving additional progesterone administration in patients with low levels has been shown to restore pregnancy rates similar to those of patients with adequate progesterone levels [15,16]. Concurrently, some evidence came from mNC cycles that clinical pregnancy rates could be improved by a supplementation with exogenous progesterone [17,18].

Given the potential role of intensive LPS in optimizing the endometrial milieu and pregnancy outcomes, the question arises whether the clinical effectiveness of HRT and mNC protocols would still differ significantly if both incorporated vaginal micronized progesterone and oral dydrogesterone for LPS. To address this issue, we conducted a retrospective cohort study to compare the pregnancy outcomes of HRT and mNC FET cycles in ovulatory women who received intensive LPS.

Materials and methods

Results

We collected a total of 2365 autologous FET cycles, comprising 1892 HRT and 473 mNC cycles, among which 1419 HRT and 473 mNC cycles were included in the final analysis after the propensity score matching. The patient demographic and cycle characteristics before and after propensity score matching are presented in Table 1. Prior to matching, there were significant differences between the two groups in terms of BMI, and the history of cesarean section and endometrial thickness. After matching, the baseline covariates were well balanced, as evidenced by the lack of remarkable differences between the matched groups and the standardized mean differences falling below 0.1 for all variables (Table 1 and S1-S2 Figures).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Patient and cycle characteristics before and after propensity score matching.

https://doi.org/10.1371/journal.pone.0327470.t001

Table 2 summarizes the pregnancy outcomes of the propensity-matched cohorts. The live birth rate was similar between the mNC and HRT groups (34.7% and 34.8%, respectively; adjusted RR 1.02, 95% CI 0.80–1.29, p = 1.00). Also, there were no significant differences in the secondary outcomes including positive hCG (54.3% vs 51.3%, p = 0.276), clinical pregnancy (42.9% vs 42.0%, p = 0.767), ongoing pregnancy (35.5% vs 35.7%, p = 1.000), implantation (35.2% vs 35.6%, p = 0.859), biochemical pregnancy (11.4% vs 9.3%, p = 0.212), miscarriage (7.8% vs 7.1%, P = 0.683), and multiple pregnancy (4.0% vs 2.7%, p = 0.218) rates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Pregnancy outcomes of the propensity score-matched cohorts.

https://doi.org/10.1371/journal.pone.0327470.t002

The incidence of major congenital anomalies was low and similar in both arms, with 1 case (<1%) in the mNC group and 2 cases (<1%) in the HRT group. Reported birth defects included a solitary testicle in the mNC group and 2 cardiac malformations (1 pulmonary artery anomaly and 1 case of mitral valve insufficiency) in the HRT group. The frequencies of preterm birth were also low and comparable between the two protocols for both singleton and twin pregnancies (Table 2). Mean birth weights were in the normal range and did not differ significantly for singletons (3159 gram vs 3171 gram) or twins (2458 gram vs 2432 gram) in the mNC and HRT groups, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Crude and adjusted relative risks for pregnancy outcomes.

https://doi.org/10.1371/journal.pone.0327470.t003

There was no significant difference in pregnancy outcomes between the mNC and HRT groups, both before and after matching. In the unadjusted analysis, the mNC protocol had a relative risk of 1.00 (95% CI 0.86–1.15) for live birth, 1.02 (95% CI 0.91–1.15) for clinical pregnancy, 1.00 (95% CI 0.87–1.15) for ongoing pregnancy, and 1.02 (95% CI 0.90–1.16) for implantation relative to the HRT protocol. After matching, the corresponding adjusted relative risks remained virtually unchanged at 1.02 (95% CI 0.80–1.29) for live birth, 1.07 (95% CI 0.85–1.35) for clinical pregnancy, 1.02 (95% CI 0.80–1.29) for ongoing pregnancy, and 1.04 (95% CI 0.93–1.17) for implantation (Table 3).

Discussion

In this large retrospective cohort study using propensity score matching, we found that mNC and HRT protocols with intensive LPS resulted in equivalent pregnancy and live birth outcomes. Among well-matched patient cohorts, the live birth rate was 34.7% in the mNC group and 34.8% in the HRT group. The rates of pregnancy, clinical pregnancy, ongoing pregnancy, implantation, pregnancy loss, multiple pregnancy, gestational age at delivery, birthweight, preterm birth rates were also similar between the two protocols. Our findings suggest that with sufficient luteal phase support, the choice of endometrial preparation for frozen embryo transfer in ovulatory women can be individualized based on patient preference and convenience without compromising outcomes.

These findings are consistent with a recent RCT by Ho et al. (2024) in 1,428 Vietnamese women, which reported similar live birth rates after one FET cycle in the mNC (33%) and HRT (34%) groups, with the notable difference being that the mNC group did not receive progesterone for LPS, and the HRT group used micronized progesterone alone [8]. This contrasts with our study, which employed a combination of vaginal micronized progesterone and oral dydrogesterone in both groups. Our strategy was based on recent evidence regarding optimizing LPS. In HRT cycles, Labarta et al. (2021) demonstrated that low serum progesterone levels on the day of embryo transfer were associated with decreased clinical pregnancy rates when only vaginal progesterone was used [14]. Furthermore, the MIDRONE study also showed that luteal phase support with oral dydrogesterone added to vaginal progesterone led to a higher live birth rate and lower miscarriage rate compared with vaginal progesterone alone [23]. In mNC cycles, the necessity and optimal regimen for LPS remain debated due to the presence of an endogenous corpus luteum. While several small randomized trials and meta-analyses suggest improved clinical pregnancy rates with vaginal progesterone compared to no supplementation [17,18], the rationale for intensive support as used in our study warrants discussion. Existing evidence indicates significant variability in endogenous progesterone levels during the natural luteal phase; for instance, Saupstad et al. (2024) reported a wide range (4.9–91.8 nmol/L) on transfer day in mNC cycles, which did not correlate with clinical pregnancy rates [24]. Currently, there is no established optimal progesterone threshold in mNC, making it difficult to identify patients who might benefit from supplementation. Therefore, providing uniform, intensive LPS to all mNC patients, as done in our study and reflecting common practice in some regions, serves two purposes: 1) it ensures luteal sufficiency even for those with potentially lower endogenous production, and 2) it standardizes the luteal environment between the mNC and HRT groups, thereby isolating the comparison to the endometrial preparation method itself rather than confounding it with differing progesterone levels. Therefore, while Ho et al. demonstrated comparable outcomes between mNC without LPS and HRT with vaginal progesterone alone, our study addresses a different but equally relevant clinical scenario by comparing these protocols under conditions where both receive standardized, intensive luteal support, potentially mitigating concerns about insufficient endogenous or exogenous progesterone in either arm under less intensive regimens.

Our results stand in contrast to two recent retrospective cohort studies that also employed propensity score matching to compare mNC and HRT protocols. Sun et al. (2024) reported significantly higher live birth (48.1% vs 41.2%) and clinical pregnancy (58.3% vs 51.1%) rates in the mNC group compared to the HRT group [25]. Similarly, Wang et al. (2023) found a higher probability of healthy live birth (35.8% vs 30.6%), and lower risks of pregnancy loss and hypertensive disorders of pregnancy with the mNC protocol [26]. This discrepancy may be explained by the lower progesterone doses used for LPS in the HRT groups of those studies. Whereas 800 mg/day vaginal progesterone plus 30 mg/day dydrogesterone were prescribed for both groups in our protocol, the studies by Wang et al. and Sun et al. used lower doses of vaginal progesterone (200 mg/day) with either oral dydrogesterone (20 mg twice daily) [26] or intramuscular progesterone (60 mg) to transform the endometrium, followed by 10 mg/day of oral dydrogesterone the next day and further increase to 30 mg/day after three days [25].

Importantly, our finding that high-dose progesterone supplementation in the mNC group yielded outcomes comparable to the HRT group challenges concerns about potential deleterious effects of supraphysiologic progesterone levels. Our results align with those of Saupstad et al. (2024) and others, suggesting that within a wide range, elevated progesterone levels in mNC cycles (whether endogenous or supplemented) do not negatively impact clinical pregnancy rates [24]. The similar rates of pregnancy, biochemical pregnancy loss, and miscarriage observed between the mNC and HRT groups in our study further support the notion that early exposure to intensive progesterone support, as provided, does not adversely impact embryo implantation or early development. This provides reassurance regarding the safety of adopting intensive LPS strategies in mNC protocols if deemed clinically appropriate. While our study was not designed to determine the absolute necessity of LPS in mNC or identify subgroups who might better benefit, it demonstrates that an intensive regimen is not detrimental compared to the same regimen in HRT cycles.

Moreover, the low and comparable incidence of congenital malformations in both groups provides reassuring evidence for the safety of oral dydrogesterone use in early pregnancy, corroborating the findings of a recent meta-analysis that reported no increased risk of birth defects with this progestin compared to population-based estimates [27]. These findings lend support to the concept of potentially using intensive LPS (like the combination of vaginal micronized progesterone and oral dydrogesterone) in both HRT and potentially mNC cycles without routine blood monitoring. This approach, consistent with suggestions by Lawrenz et al. (2024) for HRT cycles, could simplify clinical workflow and reduce patient burden [19].

The major strengths of our study include the large sample size, the use of the propensity score matching to minimize selection bias and confounding, and the homogeneity of the luteal support protocol. With over 2300 FET cycles included in the analysis, our study is nowadays the largest set up to compare mNC and HRT protocols in the setting of intensive LPS. The propensity score matching created well-balanced cohorts with standardized mean differences <0.1 for all baseline covariates, allowing valid comparisons.

However, the study has some limitations. First, due to the retrospective design, we could not fully control for all potential unmeasured confounders, although we applied strict exclusion criteria and performed multivariable analyses to mitigate their impact. Second, as previously mentioned, the choice of the endometrial preparation protocol was not randomized but resulted from the discretion of the treating physician and patient preferences, so confounding bias, despite PSM, could not be entirely controlled. Third, this study was conducted at a single academic fertility center, which may limit the generalizability of the findings to other populations or settings with different patient demographics or clinical practices. While the internal validity is strengthened by the large sample size and rigorous methodology, the external validity of our study can be limited compared to multi-center studies involving more diverse populations. Finally, a significant limitation relates to the collection of data on obstetric outcomes, particularly pregnancy complications like preeclampsia and gestational diabetes. For patients who delivered at outside facilities, these data were obtained via self-report during telephone interviews. We recognize that this method is prone to recall bias and potential inaccuracies compared to data extracted directly from medical records, especially for complex diagnoses requiring specific clinical criteria. Given these concerns about data reliability for such complications, we decided not to include these variables in the main comparative analysis to maintain the integrity of our primary findings. While data on core outcomes like live birth, miscarriage, preterm birth, and birth weight were considered more reliably reported, the absence of robust data on critical complications like preeclampsia is a notable omission, particularly given the ongoing debate about differential risks between HRT (lacking a corpus luteum) and mNC cycles. Although the expected incidence of these complications might be low in our relatively young, healthy population, potentially limiting statistical power even with complete data, future research ideally utilizing direct medical record abstraction is necessary to comprehensively evaluate the impact of these protocols on the full spectrum of obstetric outcomes.

Conclusion

Our results suggest that in patients receiving intensive LPS with a combination of vaginal micronized progesterone and oral dydrogesterone, the clinical effectiveness of FET with HRT and mNC protocols is equivalent. The choice of endometrial preparation can be based on individual patient characteristics and preferences.

Supporting information

Acknowledgments

The authors wish to acknowledge the invaluable support of all staff members at the participating assisted reproduction center.