Obstetric and neonatal outcomes after natural versus artificial cycle frozen embryo transfer and the role of luteal phase support: a systematic review and meta-analysis.

It has been more than 30 years since the first successful frozen embryo transfer (FET) ( Trounson and Mohr, 1983 ; Zeilmaker et al. , 1984 ). Nowadays, FET is increasingly applied throughout the world and in 2015 FET accounted for about 40% of all IVF cycles in Europe ( De Geyter et al. , 2018 ; ESHRE, 2018 ; Pereira et al. , 2019 ; European IVF-Monitoring Consortium (EIM) for the European Society of Human Reproduction and Embryology (ESHRE) et al. , 2021 ). The rise of FET is mainly due to improvement in laboratory techniques and use of the freeze-all strategy ( Zaat et al. , 2021a ). In FET one needs to synchronize the endometrium with the developmental stage of the embryo to facilitate implantation. The two most common ways to prepare the endometrium and to time the FET are artificial cycle FET (AC-FET) and natural cycle FET (NC-FET). In AC-FET exogenous estrogen is administered to develop the endometrium, subsequently progesterone is used to prepare the endometrium and time FET after optional downregulation using GnRH agonist. In NC-FET, there is a natural build-up of the endometrium while using detection of the LH surge to time the embryo transfer. In spontaneous ovulatory cycles, the resulting corpus luteum (CL) is believed to effectively supply all that is necessary for embryo implantation. When hCG is used for ovulation triggering, its long half-life renders it also as a form of luteal phase support (LPS) ( Casper and Yanushpolsky, 2016 ). A recent meta-analysis concluded—based on low-quality evidence with high heterogeneity in treatment protocols—that progesterone administration for LPS may be beneficial following NC-FET in terms of clinical pregnancy and live birth rates ( Mizrachi et al. , 2021 ). A Cochrane review that compared different types of endometrium preparation for FET found comparable effectiveness in terms of pregnancy chance ( Glujovsky et al. , 2020 ). Safety issues were not considered in the Cochrane review and no specific analysis was performed on NC-FET with or without LPS. Whether or not women should receive LPS following NC-FET is controversial but, nonetheless, NC-FET with LPS is often used in clinical practice ( Weissman, 2020 ; Mizrachi et al. , 2021 ). The latest observational data assessing obstetrical and neonatal outcomes after FET suggests higher risk of early pregnancy loss, rates of hypertensive disorders of pregnancy (HDP), gestational diabetes (GDM), placental pathology, postpartum hemorrhage (PPH), higher birthweights, more babies being born as large for gestational age (LGA), and macrosomia in AC-FET cycles compared with NC-FET cycles ( Hatoum et al. , 2018 ; Saito et al. , 2017 ; Ginstrom Ernstad et al. , 2019 ; Saito et al. , 2019 ; Wang et al. , 2019 ; von Versen-Hoynck et al. , 2019a , b ; Wang et al. , 2020a , b ; Moreno-Sepulveda et al. , 2021 ; Rosalik et al. , 2021 ; Severino and Povoa, 2021 ; Zaat et al. , 2021a ; Busnelli et al. , 2022 ; Vinsonneau et al. , 2022 ). In these studies, no analysis was performed on the effectiveness of the use of LPS in NC-FET compared to NC-FET without using LPS. No meta-analyses have been performed on early pregnancy loss in AC-FET compared to NC-FET and in previous meta-analysis, there has not been a stratified analysis on women with or without polycystic ovary syndrome (PCOS). Given the increasing use of FET, it is important to evaluate the safety and effectiveness of all its specific elements. Knowledge on how to prepare the endometrium for FET should not only focus on of optimal pregnancy rates but also on the safest outcome for mother and baby. With our systematic review, we aim to determine whether NC-FET with or without LPS decreases the risk of adverse obstetric and neonatal outcomes compared with AC-FET.

This systematic review and meta-analysis was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The protocol was registered at the International Prospective Register of Systematic Reviews [PROSPERO CRD42020163086]. A literature search was performed using the following databases: Pubmed, CINAHL, and EMBASE from inception to 10 October 2022. In addition, references of selected articles were examined to identify other relevant studies. Supplementary Data File S1 shows the complete search strategy. Following the search, we have used Covidence systematic review software where duplicates were removed (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia. Available at www.covidence.org ). Two investigators (T.R.Z. and P.K.) independently reviewed titles, abstracts and full text articles, and selected the studies. Any disagreements were resolved by discussion with a third author (M.v.W.) until consensus was reached. Two authors (T.R.Z. and P.K.) independently assessed the quality of the included studies and the quality of the evidence (T.R.Z. and E.K.). Disagreement regarding type and quality of the study was resolved after discussion with the third author (M.v.W.). Three criteria for inclusion were determined, which all had to be fulfilled. First, the study design was a retrospective cohort study, post hoc analysis or follow-up studies of randomized controlled trials (RCTs), because it is not possible to accomplish conditions of prevalence in study designs other than cohort studies. Second, the studies should consist of a study population of women who conceived after NC-FET (including true NC-FET, defined as home-based monitoring of ovulation to time FET and modified NC-FET, defined as ultrasound monitoring and hCG trigger for ovulation to time FET) and a control group of women who conceived after AC-FET. Finally, data had to be available about the obstetric and/or neonatal outcomes. Exclusion criteria were other study designs, studies without control group or studies that did not contain data about obstetric and/or neonatal outcomes. In our research protocol we stated the exclusion criterion: studies including anovulatory women and women with polycystic ovary syndrome (PCOS). However, in the final review and meta-analysis, this exclusion criterion was rejected because of the small number of studies only including ovulatory women. We performed a subgroup analysis on studies that excluded women with PCOS. We chose birthweight as the main outcome and report this as: absolute birthweight (g); LGA (birthweight > 90th percentile); macrosomia (as defined by the authors of the included studies); low birthweight (LBW; birthweight < 2500 g); and small for gestational age (SGA; birthweight < p10). Additional outcomes included: Obstetric outcomes Obstetric outcomes included: early pregnancy loss (defined as a miscarriage before 20 weeks of gestation expressed per woman with a registered pregnancy, although usually only first trimester pregnancy loss could be extracted); GDM (as defined by the authors of included studies); HDP (including pregnancy-induced hypertension, pre-eclampsia (PE) and hemolysis, elevated liver enzymes, and low platelets in the blood (HELLP syndrome); all as defined by the authors of included studies); PPH (as defined by the authors of included studies); placenta previa; preterm birth (PTB: delivery <37 weeks of gestation); and very preterm birth (very PTB; delivery <32 weeks of gestation). Neonatal outcomes Neonatal outcomes were congenital malformations and neonatal mortality. To ensure that pooling was valid, we assessed the similarity between the eligible studies in their design and clinical characteristics using the I 2 statistic. An I 2 >50% was labeled as marked heterogeneity ( Higgins et al. , 2003 ). The ROBINS-I bias tool, for assessing the quality of studies in meta-analyses was used to evaluate the included studies ( Sterne et al. , 2016 ). The quality of the included studies was evaluated according to the following variables: confounding, selection of participants, classification of intervention, deviations from intended interventions, missing data, measurement of outcomes, and selection of reported results. The included studies were graded as low quality, high quality, or unknown. A risk of bias summary was created in Review Manager software (version 5.4; The Cochrane Collaboration, 2020). The quality of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach ( Atkins et al. , 2004 ). Quality of evidence was downgraded by one level for serious concerns and by two levels for very serious concerns for risk of bias, inconsistency, indirectness, imprecision, and publication bias. The assessment of bias and grading of evidence was performed independently by two authors. Any disagreements were resolved by discussion or consultation with a third author. Studies that met the inclusion criteria were selected for analysis. Extracted data from included studies were pooled using StataCorp (2015. Stata Statistical Software: Release 14. College Station, TX, USA: StataCorp LP). We calculated pooled odds ratios (ORs), pooled risk differences (RDs), pooled adjusted ORs, and prevalence estimates with 95% CI using a random effect model, while heterogeneity was assessed by the I 2 . We performed stratified analyses on LPS use. As sensitivity analysis, we stratified studies on completely excluding women with PCOS and studies that also included women with PCOS.

The conducted search identified 2436 studies, then 890 duplicates were removed and 1546 studies were screened based on the abstract. In total, 1509 studies were excluded based on abstract. The remaining 37 studies were considered eligible by at least one of the reviewers. Subsequently, we excluded seven studies by screening of full text. Thirty studies met the inclusion criteria ( Nakashima et al. , 2013 ; Lathi et al. , 2015 ; Guan et al. , 2016 ; Cerrillo et al. , 2017 ; Saito et al. , 2017 ; Ernstad et al. , 2019 ; Jing et al. , 2019 ; Saito et al. , 2019 ; von Versen-Hoynck et al. , 2019a , b ; Bu et al. , 2020 ; Levi Setti et al. , 2020 ; Lin et al. , 2020 ; Makhijani et al. , 2020 ; Pan et al. , 2020 ; Zong et al. , 2020 ; Wang et al. , 2020a , b ; Aslih et al. , 2021 ; Asserhoj et al. , 2021 ; Hu et al. , 2021 ; Li et al. , 2021 ; Waschkies et al. , 2021 ; Zaat et al. , 2021b ; Dallagiovanna et al. , 2022 ; Gu et al. , 2022 ; Li et al. , 2022 ; Roelens et al. , 2022 ; Xu et al. , 2022 ; Yang et al. , 2022 ; Zhou et al. , 2022 ). Detailed information about the selection of studies for inclusion is shown in the PRISMA flow diagram in Fig. 1 Characteristics of the included studies are reported in Table 1 . Prisma flow diagram of the systematic search. Characteristics of included studies. Ovulatory FET including tNC-FET, mNC-FET and sC-FET: (286); AC-FET: (348) RRBC MC tNC-FET: (168); mNC-FET: (611); AC-FET: (357) RC SC mNC-FET: (2224); AC-FET: (4299) RC SC NC-FET: (2469); AC-FET: (5998) PC SC tNC-FET: (50); mNC-FET: (68); AC-FET: (93) RC SC NC-FET (495); AC-FET (97) RRBC MC tNC-FET: (6297); AC-FET: (1446) RC MC NC-FET: (499) AC-FET: (900) RC SC mNC-FET: (427); AC-FET: (794) RC SC tNC-FET: (3790) AC-FET: (2561) RC SC NC-FET: (8425); AC-FET: (2611) RC SC mNC-FET: (519); AC-FET: (106) RC SC tNC-FET: (567) mNC-FET: (1749) AC-FET: (585) RC SC tNC-FET: (1921) AC-FET: (1583) RC SC mNC-FET: (314) AC-FET: (1726) RC MC mNC-FET: (513); AC-FET: (287) RC SC tNC-FET (384); AC-FET: (391) Cleavage stage 20.5% Blastocyst stage 79.5% Not stated per group tNC-FET: (875); NC-FET: (6244) AC-FET: (6115) RC MC tNC-FET: (683); AC-FET: (225) RC SC NC-FET: (325) AC-FET: (211) tNC-FET: (6287); AC-FET: (10 235) mNC-FET: (29 760); AC-FET: (75 474) PC SC By administration of the Study Coordinator. Included women with anovulation and/or PCOS (NC-FET 9.7%; AC-FET 22.3%) mNC-FET: (127) AC-FET: (94) RC MC NC-FET including mNC-FET (22) and sC-FET: (46); AC-FET: (114) RC SC All women with a singleton live birth after FET, age <43 years, vitrified embryo(s) derived from the first IVF/ICSI cycles. Included women with irregular cycle (NC-FET 2.5%; AC-FET 25.7%; P < 0.001) NC-FET: (959) AC-FET: (2029) RC SC tNC-FET: (1783) AC-FET: (550) RC MC mNC-FET: (57); AC-FET: (41) RC SC NC-FET including tNC-FET and mNC-FET: (10 211); AC-FET: (4162) RC SC mNC-FET: (1225) AC-FET: (2136) RC SC mNC-FET: (4727); AC-FET: (1642) The studies of Wang (2020a , b ), Gu (2022) , and Zong (2020) only included women with live birth after 28 weeks of gestation. N/A: not available; RRBC: retrospective register-based cohort; RC: retrospective cohort; PC: prospective cohort; MC: multiple center; SC: single center; FET: frozen embryo transfer; NC-FET: natural cycle FET (not specified whether true or modified); tNC-FET: true natural cycle FET; mNC-FET: modified natural cycle FET; sC: stimulated cycle FET; AC: artificial cycle FET; PGT: preimplantation genetic testing; PCOS: polycystic ovary syndrome. Data from 30 studies (NC-FET n = 56 445; AC-FET n = 57 231) were included in the meta-analysis ( Table 1 ). Five were retrospective register-based cohort studies ( Nakashima et al. , 2013 ; Saito et al. , 2017 ; Ernstad et al. , 2019 ; Saito et al. , 2019 ; Asserhoj et al. , 2021 ). Twenty-three were retrospective cohort studies ( Lathi et al. , 2015 ; Guan et al. , 2016 ; Jing et al. , 2019 ; Bu et al. , 2020 ; Levi Setti et al. , 2020 ; Lin et al. , 2020 ; Makhijani et al. , 2020 ; Pan et al. , 2020 ; Zong et al. , 2020 ; Wang et al. , 2020a , b ; Aslih et al. , 2021 ; Hu et al. , 2021 ; Li et al. , 2021 ; Waschkies et al. , 2021 ; Zaat et al. , 2021b ; Dallagiovanna et al. , 2022 ; Gu et al. , 2022 ; Li et al. , 2022 ; Roelens et al. , 2022 ; Xu et al. , 2022 ; Yang et al. , 2022 ; Zhou et al. , 2022 ). Two were prospective cohort studies ( Cerrillo et al. , 2017 ; von Versen-Hoynck et al. , 2019a , b ). Twenty studies were single-center studies ( Lathi et al. , 2015 ; Guan et al. , 2016 ; Cerrillo et al. , 2017 ; Jing et al. , 2019 ; von Versen-Hoynck et al. , 2019a , b ; Bu et al. , 2020 ; Levi Setti et al. , 2020 ; Makhijani et al. , 2020 ; Zong et al. , 2020 ; Wang et al. , 2020a , b ; Aslih et al. , 2021 ; Hu et al. , 2021 ; Li et al. , 2021 ; Dallagiovanna et al. , 2022 ; Li et al. , 2022 ; Roelens et al. , 2022 ; Xu et al. , 2022 ; Yang et al. , 2022 ; Zhou et al. , 2022 ). The other 10 studies were multicenter studies ( Nakashima et al. , 2013 ; Saito et al. , 2017 ; Ernstad et al. , 2019 ; Saito et al. , 2019 ; Lin et al. , 2020 ; Pan et al. , 2020 ; Asserhoj et al. , 2021 ; Waschkies et al. , 2021 ; Zaat et al. , 2021b ; Gu et al. , 2022 ). Across all studies, data were extracted from national registers or hospitals records ( Table 1 ). Twenty studies used LPS in NC-FET ( Lathi et al. , 2015 ; Cerrillo et al. , 2017 ; Jing et al. , 2019 ; Saito et al. , 2019 ; von Versen-Hoynck et al. , 2019a , b ; Levi Setti et al. , 2020 ; Lin et al. , 2020 ; Makhijani et al. , 2020 ; Pan et al. , 2020 ; Zong et al. , 2020 ; Wang et al. , 2020a , b ; Aslih et al. , 2021 ; Hu et al. , 2021 ; Waschkies et al. , 2021 ; Gu et al. , 2022 ; Li et al. , 2022 ; Xu et al. , 2022 ; Yang et al. , 2022 ; Zhou et al. , 2022 ), six studies did not use LPS ( Ernstad et al. , 2019 ; Asserhoj et al. , 2021 ; Li et al. , 2021 ; Zaat et al. , 2021b ; Dallagiovanna et al. , 2022 ; Roelens et al. , 2022 ), and for four studies this was unclear ( Nakashima et al. , 2013 ; Guan et al. , 2016 ; Saito et al. , 2017 ; Bu et al. , 2020 ). The protocols used for LPS in NC-FET varied widely between studies, as presented in Table 2 . The use of LPS in case of pregnancy ranged from 5 to 12 weeks of gestation ( Table 2 ). Luteal phase support during NC-FET in the included studies. Dydrogesterone 10 mg (Duphastone ® Abbott, Biologicals); or Micronized progesterone 100 mg (Endometrin ® , Ferring); or MVP gel 90 mg (Crinone ® 8%, Merck Serono) Oral Vaginal Vaginal Dydrogesterone, dose not reported (not reported) or; Dydrogesterone, dose not reported (not reported) combined with progesterone, dose not reported (not reported) Oral Oral combined with vaginal mNC-FET: 4 days after hCG injection tNC-FET: 3 days after LH surge Micronized progesterone 200 mg (Prometrium, Rottapharm S.p.a., or Progeffik, Effik Italia S.p.a) or; MVP gel 90 mg (Crinone ® 8%, Merck Serono) Vaginal Vaginal mNC-FET: 2 days after hCG injection tNC-FET: on day of FET Progesterone 90 mg (Crinone, Merck Serono, UK) daily or; Dydrogesterone 30 mg (Abbott Biologicals B.V., the Netherlands) Vaginal Oral Progesterone 80 mg (not reported) Progesterone gel 90 mg (not reported) Intramuscular Vaginal FET: frozen embryo transfer; tNC-FET: true natural cycle FET; mNC-FET: modified natural cycle FET; E2: estradiol; P4 progesterone. Not all studies demonstrated all pre-defined outcomes, and included outcomes per study are presented in Table 3 . Outcomes reported per included study. LGA: large for gestational age; SGA: small for gestational age; LBW: low birthweight; EPL: early pregnancy loss; GDM: gestational diabetes mellitus; HDP: hypertensive disorders of pregnancy; PE: pre-eclampsia; PPH: postpartum hemorrhage; PTB: preterm birth; cong: congenital. Birthweight was lower following NC-FET versus AC-FET (MD 26.35 g; 95% CI 11.61–41.08, I 2  = 63%). NC-FET compared to AC-FET resulted in a lower risk of LGA (OR 0.88, 95% 0.83–0.94, I 2  = 54%; RD −0.016, 95% CI −0.024 to −0.008), macrosomia (OR 0.81; 95% CI 0.71–0.93, I 2  = 68%; RD −0.007, 95% CI −0.012 to −0.002), and LBW (OR 0.81, 95% CI 0.77–0.85, I 2  = 41%; RD −0.012, 95% CI −0.018 to −0.005) ( Table 4 , Fig. 2 , Supplementary Figs S1, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, S13, S14, S15, S16, S17, S18, S19, S20, S21, S22, S23, S24, S25, S26, S27 , and S28 ). Prevalence estimates for these main outcomes following NC-FET and AC-FET are depicted in Supplementary Table S1 . Difference in birthweight for NC-FET versus AC-FET, stratified by luteal phase support use. NC: natural cycle; AC: artificial cycle; LPS: luteal phase support. The summary of findings for NC-FET versus AC-FET with the grading of the evidence. LGA: large for gestational age; SGA: small for gestational age; LBW: low birthweight; GDM: gestational diabetes mellitus; HDP: hypertensive disorders of pregnancy; PE: pre-eclampsia; PPH: postpartum hemorrhage; PTB: preterm birth; LPS: luteal phase support; NC: natural cycle; AC: artificial cycle; FET: frozen embryo transfer; NA: not available. Stratified analyses on LPS use in NC-FET suggested that compared to AC-FET, NC-FET with LPS decreased PTB risk, while NC-FET without LPS did not (OR 0.75, 95% CI 0.70–0.81). Pooled adjusted OR resulted in similar estimates ( Table 4 ). NC-FET compared to AC-FET resulted in a lower risk of early pregnancy loss (OR 0.73; 95% CI 0.61–0.86, I 2  = 71%; RD −0.04, 95% CI −0.06 to −0.03). NC-FET compared to AC-FET resulted in a lower risk of PTB (OR 0.80; 95% CI 0.75–0.85, I 2  = 20%; RD −0.015, 95% CI −0.020 to −0.010) and very PTB (OR 0.66, 95% CI 0.53–0.84, I 2  = 0%; RD −0.004, 95% CI −0.007 to −0.001), HDP (OR 0.60, 95% CI 0.50–0.65, I 2  = 61%; RD −0.022, 95% CI −0.031 to −0.020), PE (OR 0.50; 95% CI 0.42–0.60, I 2  = 44%; RD −0.036, 95% CI −0.053 to −0.019), placenta previa (OR 0.84, 95% CI 0.73–0.97, I 2  = 0%; RD −0.002, 95% CI −0.004 to 0.001), and PPH (OR 0.43; 95% CI 0.38–0.48, I 2  = 53%; RD −0.052, 95% CI −0.963 to −0.009). Pooled adjusted OR resulted in similar estimates. The risk of GDM did not differ between NC-FET and AC-FET (OR 1.01, 95% CI 0.85–1.19, I 2  = 80%; RD 0.000, 95% −0.010 to 0.010) ( Table 4 , Fig. 3 , Supplementary Figs S1, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, and S12 and S15, S16, S17, S18, S19, S20, S21, S22, S23, S24, S25, and S26 ). S ummary of all averaged secondary outcomes for NC-FET versus AC-FET, expressed as odds ratio with 95% CI. NC: natural cycle; AC: artificial cycle; OR: odds ratio. Prevalence estimates for these outcomes following NC-FET and AC-FET are depicted in Supplementary Table S1 . The risk of congenital malformations (OR 0.86; 95% CI 0.66–1.11, I 2  = 3%; RD −0.000, 95% CI −0.010 to 0.010) and neonatal mortality (OR 0.80; 95% CI 0.56–1.13, I 2  = 10%; RD −0.000, 95% CI −0.010 to 0.0010) did not differ between NC-FET compared to AC-FET ( Table 4 , Supplementary Table S1 , Supplementary Figs S12 and S13 and S27 and S28 ). We performed subgroup analyses on studies that excluded or included women with PCOS ( Table 1 ), for all outcomes. No statistically significant differences were observed, and the direction of effect was the same, but there were some differences in size of the effect. Compared to studies including women with PCOS, we found that in studies excluding women with PCOS there may be an increased risk of LGA (OR 0.84, 95% CI 0.76–0.92 for studies excluding PCOS versus OR 0.92, 95% CI 0.84–1.01 for studies included women with PCOS) and early pregnancy loss (OR 0.65, 95% CI 0.52–0.81 for studies excluding PCOS versus OR 0.88, 95% CI 0.63–1.23 for studies including women with PCOS) after AC-FET. In studies excluding women with PCOS, there may be a lower risk of macrosomia (OR 0.91, 95% CI 0.79–1.05 for studies excluding PCOS versus OR 0.76, 95% CI 0.61–0.95 for studies including women with PCOS) after AC-FET, compared to studies including women with PCOS ( Supplementary Figs S1, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, S13, and S14 ). Stratified analyses on LPS use in NC-FET suggested that, compared to AC-FET, NC-FET with LPS decreased PTB risk, while NC-FET without LPS did not (OR 0.75, 95% CI 0.70–0.81 with LPS versus OR 0.96, 95% CI 0.82–1.11 without LPS). LPS use did not modify the other outcomes ( Table 4 , Fig. 4 , Supplementary Figs S15, S16, S17, S18, S19, S20, S21, S22, S23, S24, S25, S26, S27, and S28 ). Preterm birth for NC-FET versus AC-FET, stratified by luteal phase support use. NC: natural cycle; AC: artificial cycle; LPS: luteal phase support. In total, three studies reported on birthweight of babies born after true NC-FET versus modified NC-FET. Birthweight did not differ between true NC-FET when compared with modified NC-FET (MD 44.90 g; 95% CI −186.8–96.9), I 2  = 70%). The significance of the use of LPS in this groups remains to be studied ( Fig. 5 ). Owing to a lack of data, the comparison between true NC-FET versus modified NC-FET could not be pooled for other outcomes. Subgroup analysis on birthweight for true NC-FET versus modified NC-FET. LPS: luteal phase support; NC: natural cycle; FET: frozen embryo transfer; mNC: modified natural cycle; tNC: true natural cycle. Two of the included studies in the meta-analysis were ranked as having a high risk of bias on the domains of confounding bias, selection bias and reporting bias. One study was ranked as low risk of bias. The other 27 included studies were ranked as moderate risk of bias ( Supplementary Fig. S29 ). The GRADE tool was used for grading the quality of evidence. The quality of evidence ranged from very low to moderate ( Table 4 ).

This systematic review and meta-analysis has shown that singletons born from NC-FET might have a lower birthweight and a lower risk of early pregnancy loss, LGA, macrosomia, SGA, LBW, HDP, PE, PPH, and PTB compared to singletons born from AC-FET, based on low to moderate quality of evidence. In combination with comparable effectiveness of the two approaches, the interpretation is that NC-FET is the preferred treatment in women undergoing FET when the risks of obstetrical complications and potential neonatal complications are considered. Based on the analysis of current evidence on the effectiveness and safety outcomes reported in this review, NC-FET should be the preferred treatment in women with ovulatory cycles undergoing FET. The difference between NC-FET and AC-FET may be partly related to the use of LPS in NC-FET for the outcome of PTB. This finding warrants further research on the efficacy of LPS before applying LPS in all NC-FET cycles since no head-to-head studies are currently available.

This systematic review and meta-analysis shows an increase in normal range birthweight and a decrease in LGA, macrosomia, LBW, early pregnancy loss, PTB, very PTB, HDP, PE, placenta previa, and PPH in NC-FET compared to AC-FET. Therefore, the risk of adverse obstetric and neonatal outcomes is lower in NC-FET compared with AC-FET. The use of LPS in NC-FET decreases PTB risk when NC-FET with or without LPS is compared to AC-FET. The quality of evidence was very low to moderate mainly because this is a review based on observational studies and because of the substantial inter-study heterogeneity obtained, which was assumed to be caused by the variation between study populations. We estimate that for each adverse outcome the use of NC-FET may prevent 4 to 22 cases per 1000 women with a singleton live birth. The large sample size of 113 676 live births is a major strength of this study. This is a comprehensive and updated systematic review, which includes analyses of pregnancies following NC-FET and AC-FET. As LPS might have an impact on obstetric and perinatal outcomes, we provided separated analyses of pregnancies resulting from NC-FET with or without LPS. The present systematic review and meta-analysis was carried out in accordance with the PRISMA statement, ensuring high methodological quality. Moreover, the risk of bias of the included studies was assessed using the ROBINS-I tool. The validity of our results is notably improved owing to these factors. The majority of published articles in this review comprised observational studies. There is a great variety in the included studies in terms of study populations, timeline of the study, development stage of the embryos transferred with FET, freezing protocols, the use of pre-implantation genetic testing, and numbers of embryos transferred. Protocols for LPS can hardly be compared between studies because of the major variety in medication used, starting day of LPS and continuation of LPS in case of gestation ( Table 2 ). Adjustment for relevant confounders was not possible in our main analysis owing to lack of individual patient data. We did perform an adjusted analysis by pooling adjusted ORs of the included studies, resulting in no differences in outcomes ( Table 4 ). It should be noted however that confounders, such as vanishing twins, could not be analyzed and may have influenced the outcomes. In addition, unpublished data as full-text articles and in languages other than English were excluded from the meta-analysis. Clear definitions for some of the outcomes were not reported in all publications. Definitions of GDM, HDP, PE, PPH congenital anomalies, and perinatal mortality were inconsistent across the included studies. In total, 17 of the included studies also included women with irregular cycles, anovulation and/or PCOS ( Table 1 ). In our research protocol, we stated to exclude women with PCOS/anovulation and revised this during the execution of the study. This deviation deserves attention because including these women may be associated with a higher risk of perinatal complications, such as HDP, PE, GDM, and PTB, possibly distorting the outcomes of the analyses ( Palomba and La Sala, 2016 ). Therefore, we performed a subgroup analysis on studies that excluded women with PCOS. Although this did not affect our general findings, the results suggest that in studies excluding women with PCOS the RD between NC-FET and AC-FET was larger for LGA and early pregnancy loss and smaller for macrosomia, compared to studies that also included women with PCOS. For the other outcomes, including HDP, PE, GDM, and PTB, no differences were found. Furthermore, we aimed to include only singleton deliveries in our meta-analysis. Unfortunately, in five studies ( Saito et al. , 2019 ; Levi Setti et al. , 2020 ; Pan et al. , 2020 ; Li et al. , 2022 ; Roelens et al. , 2022 ), data were not presented separately for singleton and multiple deliveries. We contacted the authors of these studies for separate data on singleton births. One author responded but was not able to provide data in the short term. The other authors did not respond to our request. The number of multiple deliveries in these studies was comparable between study groups and therefore not likely to have influenced the results of our meta-analysis. The comparison of our findings with those of three recently published meta-analyses ( Moreno-Sepulveda et al. , 2021 ; Rosalik et al. , 2021 ; Busnelli et al. , 2022 ) exploring the same topic are reported in Table 5 . Busnelli et al. (2022) included 19 studies in the meta-analysis, Moreno-Sepulveda et al. (2021) included 13 studies and Rosalik et al. (2021) included 15 studies, where we included 30 studies. The findings of a decreased risk in LGA and macrosomia in NC-FET compared to AC-FET were found in all four studies ( Table 5 ). The decrease in birthweight after NC-FET compared to AC-FET was also found in the study of Rosalik et al. (2021) but not reported in the other two studies ( Table 5 ). Our findings concerning a decrease in the risk of HDP, PE, PPH and PTB after NC-FET compared to AC-FET were also reported by two studies ( Moreno-Sepulveda et al. , 2021 ; Busnelli et al. , 2022 ) ( Table 5 ). We found a higher risk of early pregnancy loss after AC-FET compared to NC-FET, and none of the other meta-analyses reported on this outcome ( Table 5 ). We found no difference in SGA between groups and this was also reported by the studies of Busnelli et al. (2022) and Moreno-Sepulveda et al. (2021) ( Table 5 ). Our finding of a decrease in LBW after NC-FET compared to AC-FET was in line with the result of Moreno-Sepulveda et al. ; however, Busnelli et al. , did not find a difference in LBW between groups ( Table 5 ). The decrease in risks of placenta previa and very PTB after NC-FET compared to AC-FET was also reported by Busnelli et al. (2022) ( Table 5 ). No difference was found in congenital malformations between babies born following NC-FET and those following AC-FET in any meta-analysis ( Table 5 ). No difference in neonatal mortality was found based on our results and the results of Moreno-Sepulveda et al. (2021) ( Table 5 ). Comparison of results with those of three previous meta-analyses. =: no difference; ↑: increased risk; ↓: reduced risk. LGA: large for gestational age; SGA: small for gestational age; LBW: low birthweight; GDM: gestational diabetes mellitus; HDP: hypertensive disorders of pregnancy; PE: pre-eclampsia; PPH: postpartum hemorrhage; PTB: preterm birth; EPS: early pregnancy loss. None of the other meta-analysis reported on the use of LPS in NC-FET. Discrepancies among different meta-analyses could be explained by the different criteria used to assess study eligibility and by the varying amount of data covering different periods of time. For a while now it has been known that babies born from FET have a higher mean birthweight and are more likely to be LGA compared with babies born from fresh embryo transfer ( Maheshwari et al. , 2018 ). The biological explanation of this phenomenon is still unknown. It has been hypothesized that epigenetic disturbances during the early embryonic stages, occurring as a result of the freezing and warming procedures, might affect the development of fetal and placental tissues. This may result in asynchrony between the embryo and endometrium and cause disturbance in fetal growth, resulting in an increase in birthweight after FET compared to fresh embryo transfer ( Pinborg et al. , 2016 ). Historically, FET cycles have been scheduled using AC-FET and therefore this type of endometrial preparation has already previously been suggested as a possible confounder for increased birthweight, LGA and macrosomia after FET ( Ginstrom Ernstad et al. , 2019 ). Based on our meta-analysis and the meta-analysis of Rosalik et al. (2021) , we now can conclude that birthweight and the risk of LGA and macrosomia are indeed increased in AC-FET compared to NC-FET. The exogenous oestrogen and progesterone that is used in AC-FET may affect the endometrium and subsequent placental development. Furthermore, in early pregnancy, progesteron has been described to induce decidualization of endometrial stromal cells, regulate extravillous trophobast (EVT) invasion, and vascular remodeling ( Beltrame et al. , 2018 ). However, aberrant progesterone and oestradiol levels in early pregnancy after AC-FET may lead to abnormal invasion of the EVT, impaired spiral artery remodeling and dysfunction of the trophoblast cells ( Schatz et al. , 2016 ; Labarrere et al. , 2017 ; Beltrame et al. , 2018 ). This non-physiological increase in steroids during AC-FET in early pregnancy has been linked to PE, abnormal placentation, stillbirth, fetal growth restriction (FGR), and many cases of PTB ( Schatz et al. , 2016 ; Labarrere et al. , 2017 ; Beltrame et al. , 2018 ). It is hypothesized that HDP is an etiologically heterogeneous disorder that occurs in at least two subsets, one involving placental dysfunction and FGR, and another with normal or enhanced placental function ( Rasmussen and Irgens, 2003 ; Pinborg et al. , 2014 ). Notably, our results showed that SGA was not different between NC-FET and AC-FET but LBW was increased during AC-FET compared to NC-FET. We hypothesize that AC-FET may more often result in the latter subset of HDP, both clinical and subclinical, causing enhanced placental function owing to very early alterations in implantation and placental development. This hypothesis may explain the increase in birthweight, LGA and macrosomia after AC-FET. However, the majority of the included studies did use progesterone supplementation for LPS in NC-FET, which makes it difficult to tease out the role of progesterone, leaving mainly exogenous estrogen and/or the lack of CL as possible explanations. This needs to be further explored to determine the biological plausibility underlying these adverse outcomes. As reported in our results and in the other recent meta-analysis, the risk of HDP, PE, and abnormal placentation (placenta previa) is decreased in NC-FET compared to AC-FET. The hypothesised rationale of the biological plausibility is discussed in the previous paragraph. Another possible, or perhaps combined, explanation is the absence of the CL during AC-FET. During AC-FET, estrogen substitution causes suppression of a dominant follicle and therefore no ovulation and CL will appear. Already in 1991, a cohort study showed that relaxin, a vasoactive hormone produced by the CL, was not detected in serum of anovulatory women who conceived with oocyte donation in an artificially prepared endometrium throughout pregnancy ( Johnson et al. , 1991 ). In 2019, a group from Stanford-university/University-of-Florida performed a similar study up to 12 weeks pregnancy in ovulatory women who conceived with AC-FET and could not detect relaxin ( von Versen-Hoynck et al. , 2019a , b ). Blood pressure, endothelial function, and the number of circulating endothelial progenitor cells were also affected. The findings of this study support that conception following an artificially prepared endometrium (hence, the lack of CL) has a negative influence on maternal vascular health in early pregnancy compared with pregnancies conceived following NC-FET or natural conception ( von Versen-Hoynck et al. , 2019a , b ). In 2020, a prospective study of two periconception cohorts showed that, during the first trimester, pregnancies conceived in the absence of a CL are characterized by lower circulating renin and prorenin concentrations compared with those conceived naturally ( Wiegel et al. , 2020 ). The absence of vasoactive factors produced by the CL in AC-FET led to deficient circulatory adaptations during early gestation and probably led to increased risks of abnormal placentation and HDP ( Conrad, 2011 ; Conrad and Baker, 2013 ; Conrad et al. , 2019 ; 2019; von Versen-Hoynck et al. , 2019a , b ; Conrad et al. , 2020 ; Singh et al. , 2020 ; Pereira et al. , 2021 ). Abnormal placenta invasion in its turn is associated with an increased risk of PTB ( Morgan, 2016 ) and may also be the reason for an increased risk of PPH ( Busnelli et al. , 2022 ). Based on our exploratory stratified analysis, the use of LPS in NC-FET decreases the risk of PTB and when NC-FET, with or without LPS, is compared to AC-FET (very low quality of evidence). The other outcomes were not influenced by the use of LPS. We need to take into account that the decrease in risk of PTB may be a spurious finding, given the limited number of included studies and the observational nature of these studies. Also, it should be noted that the protocols used for LPS in our meta-analysis vary widely and are barely comparable, which leads to a high level of clinical heterogeneity. Half of the studies on LPS did not report the duration of LPS in case of pregnancy. It was not clear from the included studies if PTB had a spontaneous onset or was induced because of pregnancy complications. To determine a biological rationale for our findings on LPS we searched the literature. For prevention of spontaneous PTB in high-risk women, daily vaginal progesterone or weekly 17-hydroxyprogesterone caproate is recommended from 16 weeks of gestation to 34 weeks of gestation by the guidelines of the National Institute for health and Care Excellence, the American College for Obstetricians and Gynecologists and the International Federation of Gynaecology and Obstetrics ( (NICE) NIfHaCE, 2022 ; American College of Obstetricians and Gynecologists’ Committee on Practice Bulletins—Obstetrics, 2021 ; Shennan et al. , 2021 ). The use of progesterone supplementation limited to the first trimester as a preventative measure for spontaneous PTB has never been studied to our knowledge. Two RCTs included women with recurrent miscarriage and compared the use of vaginal micronized progesterone supplementation to placebo vaginal capsules until 12 and 16–17 completed weeks of gestation, respectively. Progesterone lowered the risk of another miscarriage. The risk of PTB was not reported in this trial but no difference in live birth before 34 weeks of gestation was found ( Coomarasamy et al. , 2016 , 2020 ). A recently published systematic review assessed the long-term effect of prenatal progesterone treatment on child development, behavior, and health. All included studies compared progesterone to placebo in second and/or third trimester for the prevention of PTB. The authors concluded that based on the latest evidence there is no effect of prenatal progesterone on child development. Outcomes after first trimester progesterone alone remain unclear ( Simons et al. , 2021 ). The use of LPS following NC-FET may be beneficial in terms of pregnancy rates, based on a low level of evidence ( Mizrachi et al. , 2021 ). In clinical practice, NC-FET with LPS is often applied ( Weissman, 2020 ; Mizrachi et al. , 2021 ). Currently, a multicenter RCT is comparing the efficacy of NC-FET with or without LPS in China (trial registration number: ChiCTR2200057498). In terms of safety, LPS after NC-FET has not been assessed thus far. The explanation for the decrease in risk of PTB after NC-FET with LPS remains unclear. Perhaps the use of LPS somehow reduces any alterations in adaption of the cardiovascular system and placental invasion during early pregnancy. Future research on the efficacy and safety of LPS use is of great importance as it remains a major gap in knowledge in the field of ART. Owing to these uncertainties, we cannot make any recommendation about the use of LPS. Future research should focus on the efficacy of LPS in NC-FET and follow-up studies need to investigate any safety issues concerning its use. The efficacy of triggering ovulation compared to monitoring natural ovulation has been investigated in the Antarctica-2 RCT, which is currently in its follow-up phase ( Zaat et al. , 2021c ). When hCG is used for ovulation triggering, its long half-life renders it also as a form of LPS ( Casper and Yanushpolsky, 2016 ). This form of LPS has not been proven to be beneficial following NC-FET in terms of clinical pregnancy and live birth rates, based on low quality of evidence ( Mizrachi et al. , 2021 ). In terms of safety, we performed a sub-analysis on true NC-FET (natural ovulation) compared to modified NC-FET (triggering ovulation) and did not find any differences in birthweight between groups. For other outcomes this sub-analysis could not be performed owing to lack of data. This question remains to be studied in future research. The association between the endometrium preparation method for FET and obstetric and neonatal complications merits further attention and awareness in clinical practice in order to optimize the health of both mothers and children after FET. It has been more than 30 years since the first baby was conceived after FET. From the start, FET was performed in an artificially prepared endometrium because its first application was in fresh oocyte donation cycles. A solution for women without oocytes was to apply an ‘artificial cycle’ to grow the endometrium, in which the natural hormones, as produced in ovulatory women by the CL, are partly substituted with estrogen pills and progesterone vaginal capsules. AC-FET has been a very reliable, effective and predictable protocol for the fertility laboratory and therefore is still popular for FET in ovulatory women. However, the time has come to re-evaluate the use of AC-FET. Many studies, including our meta-analysis, report on increased risks for mother and child in AC-FET compared to NC-FET. The advantage of AC-FET nowadays is the easy alignment of the time point of thawing and transferring embryos with organizational necessities of the IVF laboratory, the treating doctors and the patient, which does not outweigh the disadvantages in terms of adverse obstetric and neonatal outcomes ( Zaat et al. , 2022 ). These data on safety outcomes suggest that NC-FET is preferred over AC-FET in ovulatory women. NC-FET in combination with LPS might be considered, as LPS use may be beneficial in terms of pregnancy rates, did not result in worse safety outcomes and might result in a lower PTB risk. Our recommendations do not concern anovulatory women. Now that so many studies on safety are available, an individual patient data meta-analysis that includes the original databases of these studies would be welcome. Such an analysis allows to study whether there are differences in safety profile in specific subgroups while adjusting for confounders. This could also provide us with compound adverse outcomes in terms of pregnancy complications. Concerning implications for future research, the development and use of an extended core outcome set for obstetric and neonatal outcomes in fertility care is needed ( Duffy et al. , 2017 , 2018 , 2021 ). In our meta-analysis, we captured several outcomes, such as early pregnancy loss, HDP, and GDM, based on individual study definition, which introduces heterogeneity. A standardized set of outcomes across studies would facilitate evidence synthesis in meta-analyses and systematic reviews. Furthermore, data on the actual growth during pregnancy of babies born after FET should be collected and analyzed. Previous research demonstrates that crown-rump-length (CRL) is increased in babies after FET and leads to higher birthweight ( Zaat et al. , 2021b ). It would be of great interest to investigate the association between CRL and actual birthweight in a large study in order to look at the risk of FGR after FET. Even babies born with a normal range birthweight could suffer from FGR, which can lead to increased risk of perinatal mortality and morbidity ( De Reu et al. , 2010 ; Audette and Kingdom, 2018 ). FGR is commonly defined as a condition in which the fetus does not reach its intrinsic growth potential ( Marijnen et al. , 2022 ). Placental insufficiency, resulting from a variety of placental lesions, is the common underlying pathophysiologic mechanism ( Burton and Jauniaux, 2018 ).

Click here for additional data file.