Identifying cycle-specific predictive factors for clinical pregnancy after recurrent implantation failure: a retrospective cohort study

Abstract

Objective

To investigate whether the clinical predictors of achieving a clinical pregnancy differ between fresh and frozen embryo transfer cycles in patients with recurrent implantation failure (RIF) undergoing their first ART treatment after diagnosis.

Methods

This retrospective study reviewed 24,070 ART cycles performed at Peking University Third Hospital from 2021 to 2023 among patients with at least two documented embryo transfer cycles. Patients diagnosed with RIF were identified, and the first transfer cycle following diagnosis was assessed. After applying strict inclusion and exclusion criteria, 377 fresh and 922 frozen cycles were included. Clinical, hormonal, endometrial, and procedural characteristics were collected. Clinical pregnancy was defined as the primary outcome. Predictive factors were evaluated using univariate and multivariate logistic regression models.

Results

Independent predictors of clinical pregnancy included younger age (OR = 0.939, 95% CI: 0.912–0.967, p < 0.001), higher AMH levels (OR = 1.045, 95% CI: 1.002–1.089, p = 0.039), elevated androgen (OR = 1.031, 95% CI: 1.004–1.059, p = 0.026), lower FSH (OR = 0.955, 95% CI: 0.913–0.998, p = 0.040), and greater endometrial thickness (OR = 1.057, 95% CI: 1.013–1.103, p = 0.011). The predictive factors differed between cycle types. In fresh cycles, the number of transferable embryos and the use of long or ultra-long stimulation protocols were additional significant predictors. In frozen cycles, only age, AMH, and endometrial thickness remained significant.

Conclusion

Predictors of clinical pregnancy in RIF patients differ between fresh and frozen embryo transfer cycles. Younger maternal age, favorable ovarian reserve (reflected by higher AMH and lower FSH), elevated androgen, and greater endometrial thickness are overall associated with higher pregnancy likelihood. Embryo availability and stimulation protocol further influence outcomes in fresh cycles, whereas frozen cycles exhibit a more limited predictor profile focused on patient ovarian reserve and endometrial conditions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12958-026-01525-0.

Introduction

Infertility is defined as the failure to achieve a clinical pregnancy after 12 months or more of regular, unprotected sexual intercourse [1].It has become a increasingly prominent global health concern, affecting approximately 17.8% of couples of reproductive ages [2, 3]. Assisted reproductive technology (ART) including in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI) has enabled many couples to conceive and deliver healthy children [4]. However, due to the considerable heterogeneity in the causes of infertility, many patients still struggle to achieve live births despite multiple treatment attempts.

Recurrent implantation failure (RIF) defined as the failure to achieve a clinical pregnancy after the transfer of at least four good-quality embryos in a minimum of three fresh or frozen cycles in women under the age of 40 [5]. Implantation failure may result from diverse contributing factors; therefore, a comprehensive assessment of ovarian reserve [6, 7], sperm quality [8, 9], uterine cavity environment [10–12], tubal condition [13], and other related parameters is essential [14]. Impaired endometrial receptivity is estimated to account for approximately two-thirds of implantation failure cases among patients with RIF [15]. As a multifactorial and heterogeneous condition, RIF often requires tailored therapeutic strategies to optimize treatment outcomes.

Given the inherent uncertainty of ART outcomes in RIF patients, recent research has increasingly focused on identifying predictive factors to facilitate individualized clinical decision-making. A recent study involving over 5,000 ART patients demonstrated that low anti-Müllerian hormone (AMH) levels, insulin resistance, and autoimmune abnormalities substantially increase the risk of RIF, and proposed a predictive model with high accuracy (AUC = 0.900) [16]. Additionally, another study developed and validated a clinically practical nomogram integrating six independent predictors, including age, body mass index (BMI), endometrial thickness, embryo transfer type, prior biochemical pregnancy, and duration of RIF, to estimate live birth probability in first subsequent cycles, showing good discrimination (AUC = 0.727 in training and 0.771 in validation cohort) and calibration [17]. Building upon these efforts, we sought to identify key factors associated with achieving clinical pregnancy in the first embryo transfer cycle following RIF diagnosis, with the goal of improving individualized counseling and supporting more couples of reproductive age.

Materials and methods

Patients and cycles selection

In this retrospective study, we included all ART cycles performed at the Reproductive Center of Peking University Third Hospital between January 2021 and December 2023 among patients with more than two documented embryo transfer cycles, yielding a total of 24,070 cycles from approximately 4,359 individuals. This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the Ethics Committee of the Peking University Third Hospital (Approval No.M2022471) and written informed consent was obtained from all participants prior to enrollment. After applying strict inclusion and exclusion criteria (Fig. 1), 1299 cycles were eligible for the final analysis, including 377 fresh embryo transfer cycles and 922 frozen embryo transfer(FET) cycles. The inclusion criteria were: age under 40 years; having undergone at least two consecutive cycles of fresh or freeze-thaw embryo transfer; ongoing ART treatment; and selection of the first treatment cycle following the confirmed diagnosis of RIF. The exclusion criteria were: severe internal or surgical conditions requiring long-term medication (such as diabetes, hypertension, or autoimmune diseases); hereditary disorders requiring preimplantation genetic testing (PGT); missing baseline or follow-up data; receipt of special treatments after RIF diagnosis, including intrauterine PRP, hCG, or G-CSF infusion, ERA testing, microbiota assessment, and surgical procedures for newly developed conditions such as hydrosalpinx; and receipt of additional treatments at external institutions with untraceable regimens or protocols.

Fig. 1.

Flowchart of patient selection

Data collection and outcome measures

We collected comprehensive baseline characteristics from female participants, including age, BMI, infertility type, duration of infertility, and underlying infertility factors. Infertility factors were categorized into six categories: tubal (e.g., adhesion, hydrosalpinx), ovarian (e.g., diminished reserve), intimal (e.g., endometriosis, chocolate cysts), musculature (e.g., fibroids, adenomyosis), unexplained, and male factors (e.g., oligospermia). Notably, female patients could present with one or multiple contributing factors. Comprehensive hormonal profiles were assessed, including follicle-stimulating hormone (FSH, mIU/mL), estradiol (E2, pmol/L), progesterone (P, ng/mL), prolactin (PRL, ng/mL), luteinizing hormone (LH, mIU/mL), testosterone (T, ng/mL), androstenedione (A, ng/mL), and anti-Müllerian hormone (AMH, ng/mL). Additionally, endometrial thickness and morphology scores, cycle type, number of previous pregnancies and births, type of fertilization, number of embryos transferred, and embryo type were recorded for all cycles. What’s more, for fresh cycles, further parameters were evaluated, including gonadotropin (Gn) duration and dosage, number of oocytes harvested, 2-pronuclei (2PN) zygotes, high-quality embryos, transferable embryos, and ovarian stimulation protocols. For FET cycles, endometrial preparation protocols were also documented. The primary outcome was clinical pregnancy, defined as the presence of an intrauterine gestational sac confirmed by transvaginal ultrasound.

ART procedures

Ovarian stimulation protocols included the long, prolonged, or short agonist protocols, or the antagonist protocol [18]. Follicular development was monitored by transvaginal ultrasonography and serum hormone measurements. When at least two dominant follicles reached a diameter of 18 mm, final oocyte maturation was triggered using either 250 µg recombinant hCG, 0.2 mg GnRH agonist (triptorelin), or a dual trigger. Oocyte retrieval was performed 36–38 h after the trigger, followed fertilization via IVF or ICSI. Pronuclear (PN) evaluation was conducted 16–18h after fertilization. Embryos were cultured in GM medium until day 3, when their developmental status and quality were assessed. High-quality embryos were selected for fresh transfer, while others were either cryopreserved immediately or cultured to the blastocyst stage for vitrification and subsequent frozen-thawed transfer. For FET cycles, endometrial preparation protocols included hormone replacement therapy, natural cycle, or ovulation induction cycle [19].

Statistical analysis

All statistical analyses were performed using SPSS software version 26.0 (IBM Corp., Armonk, NY, USA). Continuous variables with a normal distribution were expressed as mean ± standard deviation ( ± s) and compared between groups using the independent-samples t-test or appropriate non-parametric tests. Categorical variables were presented as frequencies and percentages [n (%)], with comparisons performed using the chi-square test (χ²) for nominal variables and Fisher’s exact test for ordinal variables. Variables with p < 0.05 in univariate analysis were included in the multivariate binary logistic regression to identify independent predictors. Multicollinearity among selected predictors was assessed, and in cases of high correlation, the variable with the strongest association with RIF was retained. A two-sided p value < 0.05 was considered statistically significant. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to estimate the strength of association between each variable and the occurrence of RIF.

Results

Baseline characteristics of the study population

To iderntify predictive factors for treatment success in the first subsequent cycle of RIF patients, a total of 1,299 cycles were analyzed and categorized by pregnancy outcome: clinical pregnancy failure (n = 639) and success (n = 660). Baseline characteristics of the two groups were compared (Table 1). Patients achieving clinical pregnancy were significantly younger than those who did not(34.96 ± 3.91 vs. 36.54 ± 4.62, p < 0.001). The distribution of infertility type differed between the groups, with a higher proportion of primary infertility in the clinical pregnancy group (64.24% vs. 58.84%, p = 0.045), while the infertility duration was slightly shorter(4.98 ± 3.31 vs. 4.99 ± 3.59 years, p = 0.026). Significant differences in baseline hormone levels were also observed, with the clinical pregnancy group showing lower FSH and higher P, A, AMH, and LH levels. Furthermore, the number of previous live births was associated with pregnancy outcome, as the proportion of nulliparous women was higher in the clinical pregnancy group (95.30% vs. 91.55%, p = 0.022). No significant differences were observed in other baseline characteristics between the two groups.

Table 1.

Comparison of baseline characteristics between RIF patients who achieved clinical pregnancy and those who did not in the cycle following diagnosis

Characteristics	Failed clinical pregnancy (N = 639)	Successful clinical pregnancy (N = 660)	p
Maternal age, year	36.54 ± 4.62	34.96 ± 3.91	< 0.001
Maternal BMI, kg/m2	22.76 ± 3.47	22.54 ± 3.47	0.668
Type of Infertility, No. (%)			0.045
Primary	376 (58.84%)	424 (64.24%)
Secondary	263 (41.16%)	236 (35.76%)
Infertile duration, year	4.99 ± 3.59	4.98 ± 3.31	0.026
Causes of infertility			0.274
Tubal factor	233(36.46%)	266(40.30%)
Ovulatory factor	243(38.03%)	211(31.97%)
Intimal factor	51(7.98%)	59(8.94%)
Musculature factor	50(7.82%)	62(9.39%)
Unknown factor	67(10.49%)	76(11.52%)
Male factor	451(70.58%)	456(69.09%)
FSH, mIU/mL	6.44 ± 3.25	5.89 ± 2.43	< 0.001
E2, pmol/L	157.33 ± 86.53	166.55 ± 94.40	0.713
P, ng/mL	1.61 ± 3.58	2.00 ± 6.39	0.033
PRL, ng/mL	12.93 ± 36.51	12.85 ± 41.29	0.766
LH, mIU/mL	3.45 ± 3.05	3.73 ± 3.58	0.048
T, ng/mL	0.99 ± 4.75	0.77 ± 3.46	0.097
A, ng/mL	4.21 ± 4.16	5.16 ± 4.87	< 0.001
AMH, ng/mL	2.60 ± 2.62	3.39 ± 3.54	< 0.001
Intimal thickness, mm	8.80 ± 2.57	9.21 ± 2.72	0.689
Intimal morphology			0.577
A	562(87.95%)	587(88.94%)
A-B&B	77(12.05%)	73(11.06%)
Type of cycle			0.423
Fresh	192(30.05%)	185(28.03%)
Frozen-thawed	447(69.95%)	475(71.97%)
Number of previous pregnancies			0.082
0	394(61.66%)	438(66.36%)
1	129(20.19%)	135(20.45%)
2	68(10.64%)	59(8.94%)
3	34(5.32%)	15(2.27%)
4	8(1.25%)	8(1.21%)
5	6(0.94%)	5(0.76%)
Number of previous productions			0.022
0	585(91.55%)	629(95.30%)
1	47(7.36%)	28(4.24%)
2	7(1.10%)	3(0.45%)
Type of fertilization			0.826
IVF	393(61.50%)	402(60.91%)
ICSI	246(38.50%)	258(39.09%)
Number of embryos transferred			0.089
1	366(57.28%)	347(52.58%)
2	273(42.72%)	313(47.42%)
Type of embryo transferred			0.510
D3(Cleavage embryo)	355(55.56%)	360(54.55%)
D5(Blastocyst)	271(42.41%)	280(42.42%)
D3&D5(Cleavage embryo&Blastocyst)	13(2.03%)	20(3.03%)

BMI Body mass index;, FSH Follicle-stimulating hormone, LH Luteinizing hormone, E2 Estradiol, P Progesterone, PRL‌ Prolactin, LH‌ Luteinizing Hormone, T Testosterone, A‌ Androgen, AMH Anti-Mullerian hormone, IVF In vitro fertilization, ICSI Intracytoplasmic sperm injection

Logistic regression analysis of clinical pregnancy

Based on the preceding univariate analysis, several variables—including female age, infertility type and duration, FSH, P, LH, A, AMH, and the number of previous live births—were identified as potentially associated with clinical pregnancy outcomes. Endometrial thickness was additionally included in the multivariate model due to its established clinical relevance for embryo implantation (Table 2). Multivariate logistic regression analysis indicated that increasing age (OR = 0.939, 95% CI: 0.912–0.967, p < 0.001) and higher FSH levels (OR = 0.955, 95% CI: 0.913–0.998, p = 0.040) were negatively associated with clinical pregnancy. In contrast, elevated levels of A (OR = 1.031, 95% CI: 1.004–1.059, p = 0.026), higher AMH levels (OR = 1.045, 95% CI: 1.002–1.089, p = 0.039), and greater endometrial thickness (OR = 1.057, 95% CI: 1.013–1.103, p = 0.011) were positively associated with the likelihood of clinical pregnancy. These results are largely consistent with our previous clinical experience.

Table 2.

Multivariate analysis of factors influencing pregnancy outcomes in the first assisted reproductive cycle after diagnosis of RIF patients

Factors	OR	95%CI	p
Maternal age, year	0.939	0.912–0.967	< 0.001
Type of infertility	1.036	0.808–1.328	0.780
Infertile duration, year	1.010	0.977–1.044	0.565
FSH, mIU/mL	0.955	0.913–0.998	0.040
P, ng/mL	1.010	0.986–1.034	0.429
LH	1.017	0.978–1.057	0.389
A, ng/mL	1.031	1.004–1.059	0.026
AMH, ng/mL	1.045	1.002–1.089	0.039
Intimal thickness, mm	1.057	1.013–1.103	0.011
Number of productions	0.689	0.448–1.061	0.091

OR Odds ratio, CI Confidence interval, FSH follicle-stimulating hormone, P Progesterone, A‌ Androgen, AMH Anti-Mullerian hormone

Predictors of clinical pregnancy after RIF in fresh cycles

Our cohort was further stratified by cycle type into 377 fresh cycles and 922 frozen cycles. In the fresh cycle subgroup, consistent with the overall cohort, significant differences were observed between the clinical pregnancy and non-pregnancy groups in terms of maternal age, infertility duration, A levels, and AMH levels (Supplementary Table 1). Notably, patients achieving clinical pregnancy demonstrated superior embryological outcomes, including a greater number of 2PN zygotes, high-quality embryos and transferable embryos. Additionally, ovarian stimulation protocols differed significantly between groups (p = 0.007), with higher proportions of long protocol (21.08% vs. 16.67%) and ultra-long protocol (18.92% vs. 9.90%) use observed in the clinical pregnancy group. Multivariate logistic regression analysis in the fresh cycle subgroup identified several significant predictors of clinical pregnancy (Table 3). Increasing female age was negatively associated with pregnancy outcomes (OR = 0.922, 95% CI: 0.874–0.973, p = 0.003), corresponding to a 7.8% decrease in pregnancy likelihood per additional year. A levels were positively associated with clinical pregnancy (OR = 1.059, 95% CI: 1.000–1.121, p = 0.049), while the number of transferable embryos emerged as the strongest predictor (OR = 1.485, 95% CI: 1.173–1.879, p = 0.001). Ovarian stimulation protocols were also significantly related to outcomes (OR = 0.823, 95% CI: 0.694–0.976, p = 0.025), highlighting the importance of protocol selection.

Table 3.

Multivariate analysis of factors influencing pregnancy outcomes in the fresh cycle subgroup

Factors	OR	95%CI	p
Maternal age, year	0.922	0.874–0.973	0.003
A, ng/mL	1.059	1.000–1.121.000.121	0.049
AMH, ng/mL	1.021	0.915–1.139	0.708
2PN	1.029	0.909–1.166	0.648
Numbers of quality embryos	0.962	0.807–1.147	0.666
Number of transferable embryos	1.485	1.173–1.879	0.001
Ovarian hyperstimulation protocols	0.823	0.694–0.976	0.025
Intimal thickness, mm	1.036	0.963–1.116	0.344

OR Odds ratio, CI Confidence interval, A‌ Androgen, AMH Anti-Mullerian hormone, 2PN Two-pronuclear

Predictors of clinical pregnancy after RIF in FET cycles

Here we present a comparative analysis of 922 FET cycles (Supplementary Table 2). Women who achieved clinical pregnancy were significantly younger (34.87 ± 3.86 vs. 36.04 ± 4.43 years, p = 0.014) and demonstrated more favorable endocrine profiles, including higher circulating P (2.25 ± 7.50 vs. 1.70 ± 4.21 ng/mL, p = 0.026), A (5.35 ± 4.95 vs. 4.37 ± 4.48 ng/mL, p = 0.008) and AMH levels (3.78 ± 3.70 vs. 2.95 ± 2.89 ng/mL, p < 0.001), as well as lower FSH levels (6.06 ± 2.21 vs. 6.46 ± 3.33 mIU/mL, p < 0.001). Multivariate analysis of FET cycles revealed distinct predictors(Table 4). Advanced maternal age remained a consistent negative predictor(OR = 0.939, 95% CI: 0.924–0.954, p < 0.001), whereas endometrial thickness emerged as a novel positive predictor (OR = 1.058, 95% CI: 1.036–1.080, p = 0.010). AMH maintained predictive value (OR = 1.056, 95% CI: 1.009–1.104, p = 0.018), though its effect differed from that observed in fresh cycles.

Table 4.

Multivariate analysis of factors influencing pregnancy outcomes in the FET cycle subgroup

Factors	OR	95%CI	p
Age of Female, year	0.939	0.924–0.954	< 0.001
FSH, mIU/mL	0.973	0.925–1.024	0.293
P, ng/mL	1.013	0.988–1.039	0.309
A, ng/mL	1.030	0.999–1.061	0.056
AMH, ng/mL	1.056	1.009–1.104	0.018
Intimal thickness, mm	1.058	1.036–1.080	0.010

OR Odds ratio, CI Confidence interval, FSH Follicle-stimulating hormone, P Progesterone, A‌ Androgen, AMH Anti-Mullerian hormone

Discussion

In this retrospective study involving 1,299 embryo transfer cycles following a confirmed diagnosis of RIF, we identified several independent predictors of clinical pregnancy outcomes in the immediate subsequent treatment cycle. Across the overall cohort, younger maternal age, higher A and AMH levels, as well as increased endometrial thickness were positively associated with the likelihood of achieving a clinical pregnancy, whereas elevated FSH levels were negatively correlated with pregnancy outcomes. In fresh cycles, younger age and higher A levels remained significant predictors, and a greater number of transferable embryos and the choice of stimulation protocol further contributed to improved pregnancy outcomes. In contrast, in FET cycles, endometrial thickness and AMH emerged as key positive predictors, while advanced maternal age consistently reduced the probability of clinical pregnancy.

Baseline levels of FSH and AMH have been widely recognized as reliable predictors of IVF success, as they are strongly associated with oocyte quality and overall female fertility [20]. Serum AMH levels have been extensively validated as predictors of ovarian reserve [21], embryo quality [22, 23], and reproductive outcomes [24, 25]. A 2025 study using a random forest model identified AMH as the strongest predictor of RIF, with CE, intrauterine adhesions, high BMI, and FSH also contributing to elevated risk [26]. In our study, lower FSH and higher AMH levels were consistently observed in the clinical pregnancy group across cycle types, and multivariate analysis further confirmed their independent association with reproductive outcomes.

In our cohort, we also observed a significant correlation between androgen levels and successful clinical pregnancy in the subsequent cycle of RIF patients, which sparked our interest. Androgens have been shown to play a direct and essential role in regulating female reproductive function, with appropriate levels contributing to follicular development [27]. Emerging evidence from randomized controlled trials indicates that testosterone supplementation may improve outcomes in women with poor ovarian response [28, 29] or diminished ovarian reserve [30] undergoing ART. However, given the interconversion between androstenedione and testosterone [31], the precise mechanisms underlying their roles in reproductive physiology remain to be clarified.

Consistent with previous findings, female age [32, 33] and endometrial thickness were also important determinants of clinical pregnancy outcomes in RIF patients. It is well established that endometrial thickness is closely associated with ART outcomes. A large-scale 2018 study involving 25,767 IVF cycles demonstrated that endometrial thickness was significantly related to both miscarriage and live birth rates [34]. What’s more, a retrospective study in 2022 focusing on FET cycles further revealed a nonlinear association between endometrial thickness and pregnancy outcomes, showing a positive correlation with clinical pregnancy rates when the thickness was below 9.5 mm [35]. Consistent with prior studies, while univariate analysis revealed no significant intergroup difference in endometrial thickness, the clinical pregnancy group exhibited a trend toward a thicker endometrium.

Subgroup analysis clearly indicated that factors influencing clinical pregnancy outcomes differ between patients undergoing fresh embryo transfer and those undergoing FET. Multivariate analysis revealed that maternal age was a consistent negative predictor of clinical pregnancy across the total population as well as in both fresh and FET subgroups. This finding underscores the well-established detrimental effect of advancing age on reproductive potential. In line with previous studies [36, 37], endometrial thickness emerged as a significant positive predictor only in the FET subgroup (OR = 1.058, 95% CI: 1.036–1.080, p = 0.010), whereas it showed no predictive value in the fresh cycle group, highlighting the need for careful hormonal modulation of endometrial receptivity in FET cycles to improve implantation outcomes. Notably, the number of transferable embryos was a significant predictor of clinical pregnancy in the fresh cycle subgroup (OR = 1.485, 95% CI: 1.173–1.879, p = 0.001), suggesting that ovarian responsiveness may be associated with reproductive outcomes [38]. While A, FSH, and AMH levels were significant predictors in the overall population, their predictive value varied across fresh and frozen cycles, indicating the need for further exploration of cycle-specific endocrine influences.

Our study has several notable strengths. First, the inclusion of a relatively large cohort of patients with RIF (n = 1299) enhances statistical power to identify key factors influencing clinical pregnancy outcomes in subsequent treatment cycles. Second, to the best of our knowledge, this is the first study to systematically compare cycle-specific prognostic factors between fresh and frozen embryo transfer groups, it reveals cycle-stratified predictors with direct implications for clinical decision-making. Third, the incorporation of comprehensive clinical and laboratory parameters—including AMH, A levels, and endometrial thickness—strengthens the robustness and clinical utility of the predictive model. However, several limitations of this study should be acknowledged. As a retrospective analysis, this study is subject to inherent limitations, including potential selection bias and residual confounding, which may affect the interpretation of the observed associations. What’s more, we were unable to comprehensively trace and quantify specific treatments or interventions received by patients between the diagnosis of RIF and the subsequent embryo transfer cycle, such as medication use, hysteroscopic evaluation, or other adjunctive procedures. Finally, the classification of ovarian stimulation protocols and endometrial preparation regimens was necessarily simplified, which may have introduced heterogeneity within each category.

Taking together, this study highlights both common and cycle-specific predictive factors for clinical pregnancy outcomes in RIF patients. These findings provide an evidence-based foundation for personalized prognostic counseling and the development of tailored treatment strategies following RIF diagnosis.

Author contributions

Ruiqi Wang: Data curation, Statistical analysis, Writing – original draft, Investigation. Congcong Ma: Data curation, Statistical analysis, Writing – original draft. Zhonghong Zeng: Formal analysis, Software, Writing – original draft. Shilin Fang & Nan Zhang: Data curation, Resources. Yang Yu & Ping Zhou: Writing – review & editing. Rong Li: Conceptualization, Methodology, Project administration, Supervision, Writing – review & editing.

Funding

This study was supported by National Key Research and Development Program (2022YFC2702500), the Key Projects of Yunnan Province Science and Technology Department (202302AA310044), the Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine (ZYYCXTD-C-202404) and the Clinical Cohort Construction Program of Peking University Third Hospital (No. BYSYDL2024001).

Data availability

The datasets analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Ethics Committee of the Peking University Third Hospital (Approval No.M2022471).

Consent for publication

Informed consents for publication were obtained from all participants.

Competing interests

The authors declare no competing interests.