{"paper_id":"bbc53f86-fe6b-499e-a6d0-ab422a7db1f0","body_text":"The widespread adoption of freeze-all protocols in IVF cycles has led to a\nprogressive shift in clinical practice towards the preferential use of frozen embryo\ntransfers (FET).\nGlobal data presented at the 2024 ESHRE Annual Meeting demonstrated that, in 2020,\n76.7% of all single blastocyst transfers involved frozen embryos ( Adamson  et al ., 2024 ).\nSimilarly, the 2021 Latin American Registry of Assisted Reproduction reported that\n52.7% of cycles employed a freeze-all approach, with 72.5% of transfers being\nperformed as FET ( Zegers-Hochschild  et\nal ., 2025 ).\nThe optimal protocol for endometrial preparation in FET cycles remains an ongoing\ndebate. The precise physiological synchrony between the euploid embryo and the\nendometrium represents a major focus of current research. The choice of including a\ncorpus luteum (CL) appears to influence maternal health and may have implications\nfor neonatal outcomes throughout pregnancy ( Lawrenz\n et al ., 2020 ), although this remains an area of\nactive discussion ( Pinborg  et al .,\n2023 ;  Ho  et al .,\n2024 ).\nThe primary objective of this study was to evaluate the effectiveness of different\ntypes of endometrial preparation-natural versus artificial cycles-in the context of\nroutine clinical practice, without the restrictive framework of a randomized\ncontrolled trial ( Horwitz  et al .,\n1990 ). As a secondary objective, we sought to compare approaches to\nluteal phase support and examine their association with maternal complications.\n\nRetrospective, observational, real-life cohort study, just single euploid embryo\ntransfers (SET) of thawed day 5 or 6 blastocysts previously frozen using\nvitrification, conducted between January 2022 and May 2024. Embryos resulted\nwith/without the use of donor oocytes.\nPatient data were obtained through review of medical records, following informed\nconsent, at FERTIPRAXIS Clinic - Human Reproduction Center, Rio de Janeiro, Brazil,\nan institution accredited by the Latin American Network of Assisted Reproduction.\nThe study protocol received approval from the Ethics Committee of Maternidade\nEscola, Federal University of Rio de Janeiro (UFRJ), and was registered with\nPlataforma Brasil under protocol number 83384024.3.0000.5275.\nThe choice of endometrial preparation-natural or artificial cycle-was determined by\nthe attending physician. The decision regarding LPS was either shared with the\npatient or made solely by the physician, particularly in cases involving combined\nprogesterone regimens.\nSubmucosal fibroids, endometrial polyps, intramural or serosal fibroids ≥5\ncm, or hydrosalpinx as identified by TVU, as well as an endometrial thickness\n< 7mm, or no confirmation of the presence of a CL in the natural cycle.\nPatients with regular menstrual cycles (21-35 days) underwent a baseline\ntransvaginal ultrasound (TVU) by cycle day 5 to exclude residual follicles,\nfollowed by a control TVU between days 8 and 10 to confirm the presence of a\ndominant follicle (≥14 mm) and endometrial thickening. From the\nidentification of the dominant follicle onward, patients performed urinary\nLH peak detection tests every 12 hours (Clearblue®, SPD Swiss\nPrecision Diagnostics GmbH) notifying their physician upon obtaining a\npositive result, which was considerate day LH + 0 regardless of the time of\nthe test positive. The presence of a CL was confirmed by TVU two days later\n(day LH + 2), as well as an endometrial thickness of ≥7mm, and luteal\nphase support (LPS) started on that very same day. Embryo transfer was\nperformed 5 days later (day LH +7).\nTo avoid transfers on Sundays, LPS could be started on day LH+1, if\nendometrial thickness was at least 7mm, in which case, CL confirmation was\nstill performed on day LH+2, but in that case, ET was performed on day\nLH+6.\nEndometrial preparation commenced on cycle day 2 or 3 with oral estradiol\nvalerate at a dose of 4 mg/day (Primogyna® - Delpharm Lille S.A.S.,\nFrance) or estradiol (Natifa® - Libbs Farmacêutica, Brazil). A\nbaseline transvaginal ultrasound (TVU) was performed, and endometrial\nthickness was monitored between days 8 and 10 of estrogen administration.\nEndometrial thickness ≥7 mm was required and if this threshold was\nnot achieved after 25 days of estrogen use, the cycle was cancelled ( Stormlund  et al .,\n2025 ). Upon reaching the required endometrial thickness, the\nestradiol dose was increased to 6 mg/day, and LPS was initiated (designated\nas day P+0). Embryo transfer was subsequently scheduled on the fifth day of\nprogesterone administration (day P+5).\n\nPreviously frozen day 5 or 6 blastocysts were morphologically graded ( Gardner & Schoolcraft, 1999 ); grades 4, 5,\nor 6, and A/or B, prior to cryopreservation and thawed following standard protocols\n(Ingamed®). Embryos were cultured in CSSNMXC® medium for a minimum of\ntwo hours prior to transfer.\nTransfers were conducted in an ambulatory operating room under positive pressure,\nguided by transabdominal ultrasound using a Wallace® 17G catheter, carefully\nplacing the embryo at the mid-point of the endometrial cavity.\nThree different regimens were utilized, maintained until the 12th week of\ngestation:\na) Micronized vaginal progesterone (MVP) [Junno® -\nFarmoquímica AS, Brazil; or Utrogestan® - Besins, Brazil:\n200 mg every 8 hours.\nb) Dydrogesterone (DYG) [Duphaston® 10 mg, Abbott\nLaboratórios, Brazil; 1 tablet orally every 8 hours.\nc) Combination MVP (200 mg every 12 hours) + DYG (10 mg orally every 8\nhours).\nBiochemical pregnancy was assessed via serum β-hCG testing 12 days\npost-transfer. Clinical pregnancy was defined as the presence of a gestational\nsac with embryonic cardiac activity on TVU, performed 15 days after a positive\nβ-hCG result. Miscarriage was defined as the spontaneous loss of a\nclinical pregnancy before 22 completed weeks of gestation. Live birth rate (LBR)\nwas defined as the number of deliveries resulting in at least one live-born\ninfant per embryo transfer, according to international standards ( Zegers-Hochschild  et al .,\n2017 ).\nThey were performed using R software (version 4.3.2, 2023). Initially,\ndescriptive analyses were conducted for the following variables: age, body mass\nindex (BMI), embryo stage, neonatal birth weight (BW), and the presence of\nendometriosis, stratified by endometrial preparation type (natural\n vs . artificial) and their respective subgroups.\nProgesterone subgroups included: natural-combined, natural-vaginal,\nnatural-oral, artificial-combined, artificial-vaginal, and artificial-oral\ncycles.\nQuantitative variables (age, BMI, and BW) were summarized as means with standard\ndeviations and compared using t-tests and ANOVAS. Categorical variables were\npresented as frequencies and percentages, with comparisons performed using\nChi-square test. A  p -value < 0.05 was considered indicative\nof statistical significance. Additionally, descriptive analyses of birth rates\nwere performed across all groups.\nGeneralized linear models (GLM) were subsequently employed to assess associations\nbetween the study groups and the following outcomes: live birth, biochemical\npregnancy, clinical pregnancy, miscarriage, and neonatal birth weight. To ensure\nthe robustness and reliability of the findings, all models were adjusted for key\npotential confounders: maternal age, BMI, embryo morphology, and the presence of\nendometriosis.\nFor binary outcomes (LBR, pregnancy, clinical pregnancy, and miscarriage),\nresults are presented as odds ratios (ORs) with 95% confidence intervals (CIs).\nFor the continuous outcome of neonatal birth weight, results are expressed as\nestimated coefficients (β) with corresponding 95% CIs.\nAdditionally, outcome weighting was applied across groups to correct for\npotential imbalances in sample sizes, ensuring a more accurate estimation of\ngroup effects within the overall cohort. Weights were calculated based on the\ntotal sample size and the relative number of participants in each group.\n\nA total of 301 SET was included in the analysis. Of these, 202 were performed\nfollowing a natural cycle (NC) protocol and 99 followed an artificial cycle (AC)\nprotocol. Within the NC group, 15 cycles used combined progesterone support, 141\nused vaginal, and 46 used oral progesterone. In the AC group, 26 used combined, 53\nvaginal, and 20 oral progesterone.\nBaseline characteristics of patients in the NC and AC groups are presented in  Table 1 , and subgroup comparisons are shown in\n Table 2 . No statistically significant\ndifferences were observed between groups regarding age, BMI, neonatal birth weight\nand presence of endometriosis. A significant difference was noted in embryo stage\ndistribution between groups ( p =0.016).\nSample Characteristics by Main Groups\nSample Characteristics by Subgroups\nLive birth rates (LBR) by main group and subgroups are summarized in  Table 3 . The overall LBR was higher in the NC\ngroup (45.54%) than in the AC group (34.34%), although this difference did not reach\nstatistical significance (z=1.85;  p =0.06). Among subgroups, the\nhighest LBR was observed in the natural-oral group (58.69%). No statistically\nsignificant differences were identified in subgroup comparisons.\nLive birth rates\nResults of the GLMs for the primary and secondary outcomes by main group are shown in\n Table 4 . After adjustment for age, BMI,\nembryo stage, and presence of endometriosis, clinical pregnancy was significantly\nmore likely in the NC group compared to the AC group (OR=0.37; 95% CI: 0.13-1.02;\n p =0.047). Miscarriage was significantly more frequent in the AC\ngroup (OR=2.96; 95% CI: 1.51-5.96;  p =0.002). No significant\ndifferences were found in birth or neonatal birth weight outcomes.\nResults of Generalized Linear Models (GLM) Analyses (two groups).\nSubgroup analyses are reported in  Table 5 .\nPregnancy was more likely in patients using oral progesterone in a natural cycle\nwhen compared to those using vaginal progesterone in a natural cycle (OR=2.35; 95%\nCI: 1.06-5.21;  p =0.03). In addition, patients in the natural-oral\ngroup had significantly higher pregnancy rates compared to those in the\nartificial-oral group (OR=0.87; 95% CI: 0.22-0.93;  p =0.036).\nMiscarriage was more frequent in the artificial-oral group compared to the\nnatural-oral group (OR=1.31; 95% CI: 1.20-1.95;  p =0.001). Full\ninterand intra-group comparisons are presented in  Supplementary Tables 1 - 3 .\nResults of Generalized Linear Models (GLM) Analyses (six groups)\nClinical complications during pregnancy are presented in  Table 6 . Gestational diabetes occurred significantly more often\nin the AC group (17.6%) compared to the NC group (5.4%) (z=2.156;\n p =0.03). No statistically significant differences were observed\nbetween groups for preeclampsia or HELLP syndrome.\nPregnancy complications in the general groups\n\nSuccessful implantation requires synchrony between a viable embryo and a receptive\nendometrium, within the window of implantation (WOI)-a period that has been studied\nsince the 1970s and is thought to occur between days 19 and 23 of the menstrual\ncycle ( Psychoyos, 1973 ). In this\nretrospective cohort study, we aimed to isolate the impact of endometrial\npreparation protocol by limiting the analysis to euploid SET, thereby controlling\nfor embryonic competence.\nDeferring embryo transfer through frozen-thawed cycles may offer practical\nadvantages, including improved patient autonomy and the ability to align treatment\nwith personal or professional schedules ( Alonso-Mayo\n et al ., 2024 ). AC protocols are often perceived as\nideal in this context due to reduced monitoring and predictability ( Mackens  et al ., 2017 ; 2023).\nAlthough modified natural cycles (mNC) using ovulation induction or hCG trigger are\ncommon, our study focused exclusively on tNC, defined by spontaneous ovulation and\nreliance on the physiological corpus luteum, without pre-ovulatory hormonal support.\nThis approach reflects current interest in more physiological endometrial\npreparation strategies ( Lawrenz  et\nal ., 2020 ).\nMackens  et al . (2023) \ndemonstrated that the addition of just 30 mg of DYG per day to MVP could improve\nreproductive outcomes in women with low serum progesterone levels on the day of FET\nin an HRT cycle. This additional progesterone supplementation was previously\nsupported by evidence from  Gaggiotti-Marre  et\nal . (2019)  as a potential ‘rescue’ for HRT-FET cycles with\nlow serum progesterone on the day prior to FET; however, they had added 25 mg/day of\nsubcutaneous progesterone to MVP. Therefore, we decided to evaluate three options:\n30 mg of DYG alone, 600 mg of MVP alone, or a combination of 400 mg MVP plus 30 mg\nDYG.\nCriticisms of tNC often cite its dependence on precise monitoring and reduced\nscheduling flexibility ( Reljič & Knez,\n2018 ). However, in our setting, patients undergoing tNC required only 3-4\nclinic visits. The protocol was facilitated by urinary LH testing positivity and CL\nconfirmation via ultrasound, which allowed for scheduled initiation of progesterone\nsupport and avoided weekend transfers-consistent with the approach described by\n Gavrić Lovrec  et al .\n(2022) . No serial hormonal monitoring was necessary, in line with earlier\nfindings on WOI timing precision using LH-based scheduling ( Xiao  et al ., 2012 ).\nCancellation rates were comparable between groups: 15.1% in tNC (mostly due to\nanovulation or premature ovulation) and 14.6% in AC (primarily due to undetected\novulation or insufficient endometrial thickness). These rates are notably lower than\nthose reported by  Ho  et al .\n(2024) , who described a 21% cancellation rate in mNC due to premature\novulation or anovulation. Additional cancellations (e.g. COVID-19 positivity or\npatient withdrawal) were equally distributed.\nPrevious evidence comparing tNC and AC has been inconclusive. A 2025 Cochrane review\n( Ghobara  et al ., 2025 )\nfound no significant differences in LBR, although data quality was low and\nmiscarriage outcomes were underreported. In our raw data, embryo morphology differed\nsignificantly between groups, prompting the use of GLM to adjust for age, BMI,\nembryo stage, and endometriosis.\nAfter adjustment, GLMs confirmed significantly higher clinical pregnancy rates in the\ntNC group (OR=0.37; 95% CI: 0.13-1.02;  p =0.047) and a significantly\nhigher miscarriage rate when AC (OR=2.96; 95% CI: 1.51-5.96;\n p =0.002). No significant differences were found in birth rates or\nneonatal birth weight ( Table 4 ). These\nfindings support the hypothesis that the physiological presence of the CL may\nenhance implantation and early pregnancy maintenance ( Lee  et al ., 2022 ). Also, they contradict the\nlack of difference presented by  Ho  et\nal . (2024)  in terms of clinical pregnancies or miscarriage\nrates when comparing AC and tNC.\nSubgroup analysis of six progesterone regimens revealed that oral DYG in tNC was\nassociated with significantly higher pregnancy rates compared to both\nnatural-vaginal (OR=2.35; 95% CI: 1.06-5.21;  p =0.03) and\nartificial-oral protocols (OR=0.87; 95% CI: 0.22-0.93;  p =0.036).\nMiscarriage rates were higher in the artificial-oral subgroup compared to\nnatural-oral (OR=1.31; 95% CI: 1.20-1.95;  p =0.001), with no\nadditional differences between vaginal and combined formulations ( Table 5  and  Supplementary Tables 1 - 3 ).\nConcerns have been raised regarding the absence of the CL in AC protocols,\nparticularly in relation to vascular and endocrine factors such as relaxin, nitric\noxide, and endothelial growth factors ( Lee\n et al ., 2022 ). Although our sample was not powered\nto assess rare obstetric complications, we found no significant difference in\nhypertensive disorders between groups. Notably, gestational diabetes was\nsignificantly more frequent in the AC group (17.6%) compared to tNC (5.4%,\n p =0.03), despite the occurrence of four twin pregnancies (4.6%)\nin the tNC group-one of which resulted in preterm labor following feto-fetal\ntransfusion ablation.\nAlthough BW were slightly higher in the AC group, this difference did not achieve\nsignificance. Two congenital anomalies were reported in the AC cohort: one case of\npostaxial polydactyly in a MVP cycle, and one cardiac malformation following DYG\nexposure, which resulted in neonatal ICU admission and death from iatrogenic\nsepsis.\nOur study is strengthened by its single-center design and uniform embryo quality,\nenabling control of key confounders through multivariable modelling. Limitations\nsuch as serum progesterone levels at FET were not systematically assessed, though\nprevious studies have suggested that inadequate luteal progesterone may compromise\nAC outcomes. Although oocyte donation was more common in AC cycles (17/99\n vs . 8/202), this was controlled for GLM models. Finally,\nC-section was frequent and largely physician directed, as common in Brazilian\nobstetric care.\n\nIn this study true natural cycles were associated with significantly higher clinical\npregnancy rates and lower miscarriage rates compared to artificial cycles. Subgroup\nanalyses suggest a potential advantage for oral progesterone support in natural\ncycles. Our findings reinforce the physiological and clinical benefits of corpus\nluteum-supported endometrial preparation but prospective studies are needed to\noptimize luteal support strategies and assess long-term maternal and neonatal\noutcomes.","source_license":"CC-BY-4.0","license_restricted":false}