Do IVF culture conditions have an impact on neonatal outcomes? A systematic review and meta-analysis

Abstract

Purpose

Are embryo culture conditions, including type of incubator, oxygen tension, and culture media, associated with obstetric or neonatal complications following in vitro fertilization (IVF)?

Methods

A systematic search of MEDLINE, EMBASE, and Cochrane Library was performed from January 01, 2008, until October 31, 2021. The studies reporting quantitative data on at least one of the primary outcomes (birthweight and preterm birth) for the exposure group and the control group were included. For oxygen tension, independent meta-analysis was performed using Review Manager, comparing hypoxia/normoxia. For culture media, a network meta-analysis was carried out using R software, allowing the inclusion of articles comparing two or more culture media.

Results

After reviewing 182 records, 39 full-text articles were assessed for eligibility. A total of 28 studies were kept for review. Meta-analysis about the impact of incubator type on perinatal outcomes could not be carried out because of a limited number of studies. For oxygen tension, three studies were included. The pairwise meta-analysis comparing hypoxia/normoxia did not show any statistical difference for birthweight and gestational age at birth. For culture media, 18 studies were included. The network meta-analysis failed to reveal any significant impact of different culture media on birthweight or preterm birth.

Conclusion

No difference was observed for neonatal outcomes according to the embryo culture conditions evaluated in this review. Further research is needed about the safety of IVF culture conditions as far as future children’s health is concerned.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10815-024-03020-0.

Introduction

Assisted reproductive technology (ART) is one of the most dynamic and fast-growing fields in medicine with the goal of fulfilling the dream of millions of couples of giving birth to a healthy baby. Since inception, the results of ART have hovered between 0 and 100%. These results could be explained in part by the limitations of clinical treatments and/or laboratory procedures, but they could also stem from the inherent limitations of human reproduction. Although we might tend to believe that ART can outperform natural conception, by using controlled ovarian stimulation and various embryo selection tools, it remains however limited by the poor efficacy of human reproduction.

The advent of high-performance in vitro fertilization (IVF) culture media and high-yield embryo cryopreservation through vitrification has led to critical improvement of singleton conceptions and also to a significant reduction of obstetric and perinatal complications. However, as the number of children conceived following ART increases, obstetric and perinatal outcomes as well as long-term health are key issues in the field of reproductive medicine. Although the majority of children are healthy, some authors reported that perinatal outcomes following a single pregnancy obtained following ART appear to be more unfavorable than for natural conception [1]. Particularly, fresh embryo transfer (fresh ET) has been associated with an increased risk of low birthweight (LBW), small for gestational age (SGA), and preterm birth (PTB) [2–4] when compared to spontaneously conceived singletons. Recent evidence suggests that the consequences of these poor neonatal outcomes for the short- and long-term health of children might be noteworthy, with some studies suggesting increased risks of cardiovascular diseases, mental health disorders, and social difficulties in the case of LBW [5, 6]. Moreover, weight differences could also extend into the postnatal and child growth period because of a tendency to show a postnatal catch-up growth [7], reported as a risk factor for cardiometabolic diseases in adulthood [8, 9].

Various causes for those poorer outcomes have been proposed, such as patient characteristics; subfertility by itself [10–12]; controlled ovarian stimulation regimens [4], but also embryo culture media [13, 14]; and possibly, more globally, embryo culture conditions.

Indeed, the IVF process involves in vitro culture, during which embryos are exposed to an artificial environment, influenced by the type of incubators, oxygen levels, and culture systems and media. Historically, traditional box-style incubators with atmospheric oxygen tension (20%) were widely used. A shift was progressively observed toward low oxygen tension (5%) in the 2010s, a more physiological condition for embryo development, minimizing oxidative damage, further leading to improved livebirth rates [15]. Then, bench-top incubators, and sometimes time-lapse systems, were introduced, in order to provide a stable environment by limiting disturbances [16]. Culture media have also strongly evolved, from simple culture media based on blood serum during the early years of IVF, to commercially produced complex media containing a variety of different substances, such as amino acids, human albumin, antibiotics, and growth factors [17]. Media currently used can be divided into sequential, using different compositions throughout the embryo development, reflecting the changes in metabolic requirements, and single-step media, using only one single medium containing the necessary components for the embryos to choose what they need during the whole culture. Although it has been proposed that single-step culture, by decreasing handling and environmental fluctuations, could reduce stress undergone by embryos [18], controversies remain about the real efficacy in terms of pregnancy and live birth rates of these different approaches for culture conditions [19]. Moreover, this distinction between sequential/single-step is not enough to highlight the precise differences in composition between the culture media, each of whose components should be considered from the point of view of consequences for embryo development and for potential epigenetic effects. Particularly, glucose, lactate, pyruvate, and amino acid concentrations may have an influence that is also linked to oxygen tension due to fine redox regulation [20], as well as methionine and folate concentrations that are crucial for supporting DNA methylation changes during preimplantation development [21, 22].

Considering embryo culture conditions is even more crucial that the number of cycles using blastocyst transfer rises [23], resulting in more and more embryos being cultured for a long period. Indeed, it is noteworthy that, in a recent systematic review, fresh blastocyst transfer was associated with higher risks of LGA and very PTB when compared to fresh cleavage-stage embryo transfer [24], underlining the need to look at the impact of culture conditions on obstetric and perinatal outcomes from every angle. Despite growing literature, knowledge concerning the effects of embryo culture conditions during IVF on the health of children remains to be explored.

With this in mind, we aimed to perform a systematic review and meta-analysis reporting obstetric or perinatal complications according to incubator type, oxygen levels, and culture media.

Methods

Search strategy and selection criteria

The search strategy, selection criteria, data extraction, quality assessment, and statistical analyses described below were defined a priori in version-controlled documents. The conduct and reporting of this systematic review and meta-analysis were guided by PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines [25, 26] and prospectively registered (PROSPERO CRD42021285883).

All studies reporting obstetric or perinatal outcomes for at least one of the three culture conditions (incubator type, oxygen levels, culture systems, or media) after the use of conventional IVF or intracytoplasmic sperm injection (ICSI) were included in the initial screening. The main outcomes were mean of birthweight, low birthweight (defined as the rate of birthweight less than 2500 g), high birthweight (defined as the rate of birthweight more than 4500 g), and preterm birth (PTB) (defined as the live birth or stillbirth which occurs after 22 weeks but before 37 weeks of amenorrhea). Studies were included if they report values of at least one of the main outcomes for at least one of the three different culture conditions. The secondary outcomes were maternal and placental disorders during pregnancy (antepartum and postpartum haemorrhage, gestational diabetes, hypertensive disorders of pregnancy, digestive or urinary complications during pregnancy, cholestasis, placenta praevia, accreta or abruption), miscarriage and stillbirth (late miscarriage, intrauterine fetal death), sex ratio, fetal malformations, cesarean delivery, gestational age at birth, large for gestational age (LGA) (fetus > 90th percentile), small for gestational age (SGA) (fetus < 10th percentile), and intrauterine growth restriction. Exclusion criteria were other assisted reproductive techniques such as intrauterine insemination or in vitro maturation and cycles with preimplantation genetic testing.

The electronic databases PubMed, Embase, and the Cochrane Library were searched for publications from 1 January 2008–31 October 2021. The search strategy was limited to articles published after 2008 in order to favor the inclusion of articles evaluating culture conditions that are as close as possible to today’s laboratory, including low oxygen tension, bench-top incubators, and current culture media. It was also limited to articles published in English or French involving human subjects. The search strategies for electronic databases are described in detail in Supplemental Data. The literature search strategies were performed by an information specialist in association with the referral Inter-University Library of Medicine of Paris Descartes, Paris 5, France.

Study selection and data extraction

Two independent reviewers (NS and CS), blind to authors, institutions, journal titles, and study results, performed an initial screening of the title and abstract of all articles. Based on the pre-established inclusion criteria, full texts of potentially relevant articles were retrieved and assessed for inclusion by two reviewers independently (NS and CS). Methodological validity was also assessed prior to inclusion in the review. Any disagreement or uncertainty was resolved by discussion among reviewers to reach a consensus. Two independent reviewers carried out study selection, bias assessment, and data extraction (NS and PP or CS and NA). Data were extracted from included articles using a data extraction form designed by the authors.

The following study details were collected to characterize the included studies: country, study design, study type, inclusion and exclusion criteria, and study period and for each culture group: number of women, maternal age, body mass index (BMI), IVF/ICSI method, rank, stimulation protocol, number of cycles, number of transfers, number of oocytes, and type of transferred embryos, as well as main and secondary outcomes, as described above, according of the 3 studied culture conditions.

Risk of bias and study quality assessment

The methodological quality of the selected studies was assessed by two independent reviewers using the Cochrane Handbook methods and by the adapted Newcastle-Ottawa Quality Assessment Scale for cohort studies [27]. This system evaluates studies based on three criteria: participant selection, comparability, and ascertainment of outcomes. Risks of bias were assessed according to the type of study, using ROBINS-I [28] for cohort studies or the RoB-2 tool [29] for RCT. ROBINS-I covers seven domains: confounding, selection of participants, classification of intervention, deviations from intervention, missing data, measurement of outcome, and selection of reported results. Each risk of bias criteria was judged as “low”/“moderate”/“serious”/“critical”/“no information.” RoB2 covers five domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Each risk of bias criteria was judged as “low,” “high,” or some concerns’ risk.

Data analysis

All qualified studies with quantitative data for singleton on at least one of the outcomes were included in the meta-analysis. No replacement of missing data was done. Each outcome was analyzed independently.

By using continuous or dichotomous data, the summary statistics were the standard difference in means (SMD) or the risk ratio (RR; risk exposure/risk control) alternatively, with their 95% 2-sided confidence interval (CI). Using a random-effects model, we applied the inverse variance method for SMD and the Mantel-Haenszel method for RR. Pooled effect sizes were deemed statistically significant at p < 0.05. In addition to the estimation of between-study variance (τ²), the Q chi-squared test was used to test the heterogeneity between the studies. The inconsistency across studies was quantified using the I² statistic and interpreted following the Cochrane Collaboration guide [30].

For incubator type and oxygen tension, meta-analyses were performed using Review Manager 5.4.1 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). For culture media, published studies evaluated various culture media and not systematically compared outcomes according to single-step and sequential culture systems. A “conventional” meta-analysis was therefore not relevant and the netmeta version 2.1.0 of R-4.1.3 software was used to perform frequentist (a Graph Theoretical Model) network meta-analyses (NMAs).

The following steps were applied to all NMAs. In the first step, the network structure was determined. The network geometry was visualized using a network graph in order to show which culture media have been compared directly in the qualified articles and which can only be informed indirectly. Then, pairwise meta-analyses of all directly compared culture media were carried out so that the statistical heterogeneity of studies within each comparison can be evaluated. Common heterogeneity was assumed to act as only a few articles per direct comparison were expected. We conducted a statistical evaluation of consistency using the design-by-treatment interaction model (globally) and separate direct from indirect evidence test (locally). The most commonly reported medium in the included studies was Vitrolife G-series, and only the comparisons with this medium were displayed. We also fitted the design by treatment interaction model to evaluate the presence of inconsistency in the entire network [31]. Collected studies appeared to be sufficiently similar with respect to the distribution of effect modifiers (selection criteria, female age), and random-effects NMAs were conducted to synthesize all evidence for each outcome. Measure of culture media effect is displayed with their 95% CI in league tables.

The quality of evidence for each outcome was judged using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group methodology [32].

Sensitivity analysis

Pre-specified subgroup analyses included in the statistical analysis plan were performed: type of transferred embryo (fresh/frozen) and type of embryo stage (D2/3, D5/6).

Results

Studies description

The literature search yielded a total of 444 articles. After removing duplicates, 182 abstracts were reviewed, 143 records were excluded (out of topic n = 114, main outcome missing n = 19, review n = 4, other n = 6), and 39 full-text articles were assessed for eligibility. Among them, a total of 28 studies were retained for the review, including 20 that were included in the meta-analysis (Fig. 1). Indeed, 5 studies could not be used in the meta-analysis because of the corresponding duplicate data or missing main outcome. Moreover, 3 studies reporting neonatal outcomes according to the incubator type were excluded, as no meta-analysis could be performed due to different study groups between the studies [33–35].

Fig. 1.

Flow chart of study selection for systematic review and meta-analysis

Three studies assessed the influence of oxygen concentration during embryo culture on neonatal outcomes (Table 1): one randomized clinical trial (RCT) [36] and two retrospective cohort studies [37, 38]. The study sample sizes ranged from 170 to 15,552 singleton live births, for a total of 16,166 singleton neonates. One study included only results following fresh embryo transfers [37], and the two others presented results following fresh and frozen embryo transfers [36, 38].

Table 1.

Characteristics of studies included in the meta-analysis

Author, year Country	Study design	Participants	Female age	Culture conditions			Type of embryo transfer	Stage of embryo transfer	Total number of live births	Outcomes
				O2 concentration						PTB	GA	BW	LBW	HBW	LGA	SGA
Castillo et al. (2020) UK	Retrospective, cohort	All IVF/ICSI cycles (HFEA register) Jan 2011–Dec 2013	NA	5–6%	20%		Fresh	Mixed	15,552 (11,957 for 5–6%, 3,595 for 20%)		NS	NS
Rendón Abad et al. (2020) Spain	Prospective, RCT	Oocyte donation cycles Nov 2009–Apr 2012	25–45 (recipient)a	6%	20%		Fresh	D3	26(147 for 6%, 115 for 20%)	NS	NS	NS	NS	NS	NS	NS
							Frozen-thawed	D3	182 (93 for 6%, 89 for 20%)	NS	NS	NS	NS	NS	NS	NS
Van Montfoort et al. (2020) The Netherlands	Retrospective, cohort	Jan 2012–Dec 2013	34.0 ± 4.3b	5%	20%		Fresh	D2	124 (35 for 5%, 89 for 20%)		NS	NS
							Frozen-thawed	D2	46 (14 for 5%, 32 for 20%)		NS	NS

Author, year Country	Study design	Participants	Female age	Culture conditions			Type of embryo transfer	Stage of embryo transfer	Total number of live births*	Outcomes
				Culture media						PTB	GA	BW	LBW	HBW	LGA	SGA
Castillo et al. (2020) UK	Retrospective, cohort	All IVF/ICSI cycles (HFEA register) Jan 2011–Dec 2013	NA	Global (CSC, Global, Vitrolife)	Sequential (Cook, Medicult, Quinn’s, Vitrolife)		Fresh	Mixed	16,930 singletons		NS	NS
Stimpfel et al. (2020) Slovenia	Retrospective, cohort	IVF/ICSI cycles with ≥ 3 fertilized oocytes, sibling oocyte approach Sep 2015–Apr 2016	33.9 ± 4.5b	Global (Vitrolife)	Sequential (Cook)		Both	D5	46 singletons (25 G-TL, 21 Cook)			NS
Cimadomo et al. (2018) Italy	Prospective, cohort	IVF/ICSI cycles with ≥ 1 mature oocyte Sep 2013–Sep 2015	38.5 ± 3.6b	Global (CSC)	Sequential (Quinn’s)		Both	D5/6	388 neonates (378 singletons and 10 twins)		NS	NS
Bouillon et al. (2016) France	Prospective, RCT	Singletons live births following IVF/ICSI 2008 (6 months)	NA	Global (Global)	Global (SSM)		Fresh	D2/3	71 singletons (40 Global, 31 SSM)	NS	NS	NS	NS		NS	NS

Gu et al. (2016) China

Retrospective, cohort

Singleton live births following IVF/ICSI with D2/3 transfer Jan 2009–Dec 2012

31.3 ± 4.1b

Sequential (Vitrolife)

Sequential (Quinn’s)

Fresh

D3

1,755 singletons (419 G-series, 1336 Quinn’s)

NS

NS

NS

NS

NS$

NS

NS

Frozen-thawed

615 singletons (202 G-series, 413 Quinn’s)

Kleijkers et al. (2016) The Netherlands

Prospective, RCT

First IVF/ICSI cycles Jul 2010–May 2012

33.8 ± 4.4b

Global (HTF)

Sequential (Vitrolife)

Fresh

Mixed

300 singletons (137 HTF, 163 G-series)

NS

NS

0.005

0.02

NS

NS

NS

De Vos et al. (2015) Belgium

Retrospective, cohort

Singleton live births following fresh embryo transfer, female age < 40 Apr 2004–Dec 2009

32.1 ± 4.2b

Sequential (Medicult)

Sequential (Vitrolife)

Fresh

Mixed

2,098 singletons (1,388 Medicult, 710 G-series)

NS

NS

NS

NS

NS

Yin et al. (2015) China

Retrospective, cohort

Singleton live births following D3 fresh embryo transfer Jan 2010–Dec 2012

30.4 ± 0.4b

Sequential (Quinn’s)

Global (SSM or CSC)

Fresh

D3

678 singletons (459 Quinn’s, 138 SSM, 81 CSC)

NS

NS

NS

NS

Lemmen et al. (2014) Denmark

Retrosepctive, cohort

Singleton live births following fresh embryo transfer Jan 2004–Dec 2009

32.6 ± 3.9b

Sequential (Cook)

Sequential (Medicult)

Both

D2/3

1,124 singletons (977 Cook, 147 Medicult)

NS

NS

NS

NS

Wunder et al. (2014) Switzerland

Retrospective, cohort

Healthy singleton live births following fresh or frozen IVF/ICSI Jan 2000–Dec 2004

33.9 ± 4.1b

Sequential (Vitrolife)

Sequential (Cook)

Both

D2/3

525 singletons (352 G-series, 173 Cook)

NS

NS

NS

NS

NS

NS

Zhu et al. (2014) China

Retrospective, cohort

IVF/ICSI cycles with fresh embryo transfer, female age < 41 Jan 2009–Jun 2012

31.9 ± 0.7b

Global (Global)

Sequential (Vitrolife or Quinn’s)

Fresh

D5

96 singletons (31 Global, 53 G-series, 12 Quinn’s)

NS

NS

Carrasco et al. (2013) Spain

Retrospective, cohort

IVF/ICSI Oct 2006–Dec 2010

36.3 ± 4.4b

Sequential (CooK)

Sequential (Medicult)

Sequential (Vitrolife)

Fresh

D2

523 singletons (154 Cook K-SICM, 172 ISM1, 197 G-series)

NS

NS

NS

NS

Eskild et al. (2013) Norway

Retrospective, cohort

Singleton live births following IVF/ICSI Jan 1999–Dec 2011

NA

Sequential (Medicult)

Sequential (Vitrolife)

Fresh

D2/3

901 singletons (402 ISM1, 449 G-series)

NS

0.02

Hassani et al. (2013) Iran

Prospective, RCT

ICSI cycles with ≥ 2 follicles Feb 2009–Jun 2009

32.1 ± 5.7b

Sequential (ISM1)

Sequential (Vitrolife)

Fresh

D2/3

164 singletons (86 ISM1, 78 G-series)

0.001

Lin et al. (2013) China

Retrospective, cohort

Singleton and twin live births following IVF/ICSI, female age < 40 Jan 2008–Dec 2010

31.2 ± 3.9b

Global (Global)

Sequential (Vitrolife or Quinn’s)

Fresh

D3

1,201 singletons (460 Global, 596 G-series, 145 Quinn’s)

NS

NS

NS

NS

NS

Eaton et al. (2012) USA

Retrospective, cohort

Singleton and twin live births following fresh D3 SET and DET, respectively Jan 1999–Dec 2008

32.7 ± 3.5b

Global (Global)

Sequential (Vitrolife)

Fresh

D3

155 singletons (53 Global, 102 G-series)

NS

NS

Nelissen et al. (2012) The Netherlands

Prospective, cohort

IVF/ICSI, female age < 41 Jul 2003–Dec 2006

32.5 ± 3.8b

Sequential (Vitrolife)

Sequential (K-SICM Cook)

Fresh

D2/3

292 singletons (168 G-series, 126 Cook)

NS

NS

0.006

0.006

NS

NS

NS

Frozen-thawed

67 singletons (22 G-series, 45 Cook)

NS

NS

NS

NS

NS

NS

NS

Vergouw et al. (2012) The Netherlands

Retrospective, cohort

Ongoing pregnancies following fresh or frozen single embryo transfer Jan 2008–Dec 2010

34.1 ± 4.0b

Global (HTF)

Sequential (Quinn’s)

Fresh

D3

358 singletons (99 HTF, 259 Quinn’s)

NS

NS

NS

NS

NS

Frozen-thawed

118 singletons (32 HTF, 86 Quinn’s)

NS

NS

NS

NS

< 0.05

BW birthweight, D2/3 day 2/3 (cleavage stage), D5/6 day 5/6 (blastocyst stage), HBW high birthweight, HFEA Human Fertilisation and Embryology Authority, ICSI intra cytoplasmic sperm injection, IVF in vitro fertilization, GA gestational age, LBW low birthweight, LGA large for gestational age, NA not available, PTB preterm birth, RCT randomized controlled trial, SGA small for gestational age

*Singleton only, except Cimadomo et al.

^$HBW > 4000 g (4500 g for other studies)

^aRange

^bMean ± SD

A total of 18 studies assessing the influence of culture media on neonatal outcomes were included in the NMA (Table 1): three RCTs [39–41], two prospective cohort studies [18, 42], and 13 retrospective cohort studies [37, 43–54]. Five studies compared one single-step versus one sequential culture media [18, 41, 45, 50, 51], four compared one or several single-step formulations versus one or several sequential formulations [37, 49, 53, 54], eight compared two or more different sequential culture media [40, 42–44, 46–48, 52], and one compared two different single-step formulations [39]. The study sample sizes ranged from 46 to 16,930 live births, for a total of 27,898 singleton neonates. Overall, 11 studies only included results following fresh embryo transfer, three displayed separate results for fresh and frozen embryo transfer, and four mixed results.

The overall quality of evidence was judged low for birthweight according to oxygen tension or culture media and low for low birthweight and preterm birth risk according to culture media. The majority of included studies showed low to moderate bias (Supplementary Table 1).

Type of incubator

No meta-analysis could be performed, due to different study groups between the studies. In a retrospective study of 378 live births, no difference for obstetric outcomes was observed when a time-lapse incubator was used rather than a standard incubator [33]. Concerning the perinatal outcomes, the main series published evaluated the perinatal outcomes in more than 40,000 IVF cycles leading to nearly 11,000 live births [37], showing no difference in birthweight and gestational age according to the use of benchtop or standard incubators. In one study on 593 live births following fresh embryo transfer, the mean birthweight was higher in the time-lapse group when compared to the control group (either benchtop or standard incubator) [35]. Two other small series comparing live births following embryo culture in time-lapse systems or standard incubators failed to find any significant difference [33, 34].

Oxygen tension

No studies evaluating obstetric outcomes according to oxygen level could be included. For perinatal outcomes, a total of 12,246 singletons born from embryos cultured under 5% oxygen tension (hypoxia) were compared to 3920 singletons born from embryos cultured under 20% oxygen tension (normoxia). Mean birthweight did not significantly differ between the 2 groups (SMD 45 g, 95% CI [− 24–114], I² = 0%) (Fig. 2). The exclusion of live birth following frozen embryo transfer did not change the result (data not shown). No significant difference in gestational age at birth was shown according to the oxygen tension (0.08 week, 95% CI [− 0.17–0.32], I² = 0%) (Supplementary Figure 1). Risks of low birthweight and of prematurity according to the oxygen tension during embryo culture were only reported in Rendon Abad et al. without any significant difference [36]. In this study, no significant difference for preterm births, cesarean sections, or congenital malformations was found either [36].

Fig. 2.

Influence of oxygen tension: birthweight forest plot

Culture media

Independent NMAs were performed on each outcome in order to assess the effect of culture media on obstetric and perinatal outcomes.

Concerning obstetric outcomes, the risks for pregnancy-related complications such as hypertension and diabetes were similar across 3 or 2 different media ([54] and [52], respectively) as well as the risk for miscarriage or stillbirth [39, 40, 43, 50].

Concerning perinatal outcomes, Fig. 3 represents the network of eligible comparisons for birthweight. The network is well connected. In total, 18 articles including 9 culture media and 27,867 livebirths were entered in the NMA. The NMA shows a high heterogeneity/inconsistency (τ²= 15,590, I² = 94.9%, 95% CI [93.3–96.1%]). No statistical difference was shown between culture media, with SMD ranging from − 193 to + 105 g (Table 2). No statistical difference was shown between indirect and direct comparisons (e.g., comparisons versus G-Series® (Vitrolife®) (Supplementary Figure 2). We deemed that the transitivity assumption was held, and there was no suggestion of global or local inconsistency in the network. Sensitivity analyses could be performed only in the subgroup of fresh embryo transfer (12 articles, 2375 livebirths) and found no significant difference between culture media (Supplementary Table 2).

Fig. 3.

Network graph of eligible comparisons for birthweight. The thickness of the lines is proportional to the number of studies evaluating each direct comparison. The size of the nodes is proportional to the number of livebirths reported for each culture medium

Table 2.

League table of pairwise comparisons for birthweight

Pairwise comparisons of the culture medium in the row versus the culture medium in the column. For each pairwise comparison, a number from row minus column is presented. Culture media are reported in alphabetical order. Data are SMD of birthweight (grams) with a 95% confidence interval. Network estimates from NMA in the lower triangle (grey boxes) and the direct estimates from pairwise comparisons in the upper triangle (white boxes). For the random-effects model, the direct estimates are based on the common between-study variance τ² from the NMA

Figure 4 represents the network of eligible comparisons for low birthweight. Overall, 8 articles including 7 culture media and 6859 livebirths were entered in the NMA. The NMA shows a very low heterogeneity (τ² = 0.21). No statistical difference was shown between culture media, and none of the RR was significant (Table 3). We deemed that the transitivity assumption was held, and there was no suggestion of global or local inconsistency in the network. No sensitivity analysis could be performed for low birthweight.

Fig. 4.

Network graph of eligible comparisons for low birthweight. The thickness of the lines is proportional to the number of studies evaluating each direct comparison. The size of the nodes is proportional to the number of livebirths reported for each culture medium

Table 3.

League table of pairwise comparisons for low birthweight

Pairwise comparisons of the culture medium in the row versus the culture medium in the column. For each pairwise comparison, a number from row minus column is presented. Culture media are reported in alphabetical order. Data are risk ratio (RR) of low birthweight with a 2-sided 95% confidence interval. Network estimates from NMA in the lower triangle (grey boxes) and the direct estimates from pairwise comparisons in the upper triangle (white boxes). For the random-effects model, the direct estimates are based on the common between-study variance τ² from the NMA

Figure 5 shows the network of eligible comparisons for preterm delivery. In total, 13 articles including 8 culture media and 8209 livebirths were entered in the NMA. The NMA shows a very low heterogeneity/inconsistency (τ² = 0). Only one RR significantly differs from 1.0 (Table 4). It concerned the direct estimate for comparison between G-Series® (Vitrolife®) and HTF® (RR 3.92, 95% CI [1.15–13.37]). However, the indirect estimate of this comparison was not statistically significant (RR = 2.10, 95% CI [0.86–5.13]). One indirect estimate for the comparison between Global® and HTF® was borderline (RR = 2.56, 95% CI [1.00–6.58]). Finally, no statistical difference was shown for comparisons of direct and indirect estimates (e.g., comparisons versus G-Series® (Vitrolife®), Supplementary Figure 3). We deemed that the transitivity assumption was held, and there was no suggestion of global or local inconsistency in the network. Sensitivity analyses could be performed only in the subgroup of embryos transferred at the D2/D3 stage and show no statistical difference between culture media (data not shown).

Fig. 5.

Network graph of eligible comparisons for preterm birth. The thickness of the lines is proportional to the number of studies evaluating each direct comparison. The size of the nodes is proportional to the number of livebirths reported for each culture medium

Table 4.

League table of pairwise comparisons for preterm delivery

Pairwise comparisons of the culture medium in the row versus the culture medium in the column. For each pairwise comparison, a number from row minus column is presented. Culture media are reported in alphabetical order. Data are risk ratio (RR) of preterm delivery with a 2-sided 95% confidence interval. Network estimates from NMA in the lower triangle (grey boxes) and the direct estimates from pairwise comparisons in the upper triangle (white boxes). For the random-effects model, the direct estimates are based on the common between-study variance τ² from the NMA

Discussion

A systematic review on the impact of embryo culture conditions on both obstetric and neonatal outcomes was performed. No significant difference in birthweight according to culture under hypoxia or normoxia conditions was noted. The network meta-analysis did not reveal any difference in birthweight or risk of prematurity across various culture media, suggesting no superiority or inferiority of one culture medium over another in terms of perinatal outcomes.

Child health following IVF is a constant concern for reproductive specialists. Although the absolute risks remain relatively low, some unfavorable perinatal outcomes, such as SGA following fresh ET or LGA following frozen embryo transfer, have been underlined by numerous studies and may have serious short and long-term consequences on children’s health [5, 6].

Among the pathophysiological hypothesis, culture conditions could influence gene expression profile, especially through epigenetic mechanisms [55, 56], while embryos are undergoing major genomic and epigenomic changes since major methylation dynamics take place during the preimplantation stages of embryo development. Several studies have then shown that the expression of genes can be altered depending on the culture conditions used and that the establishment and maintenance of DNA methylation are impaired during IVF [57]. Those modifications could lead to long-term epigenetic changes with developmental and growth disorders which could influence further clinical outcomes, in line with the developmental origins of health and disease (DOHaD) hypothesis [14]. Indeed, as imprinted genes play a role in placental development, genetic imprinting disorders during the pre-implantation period may have an impact on placental functions and hence on the occurrence of adverse obstetrical and/or neonatal outcomes [58].

In this review, obstetric or perinatal complications according to incubator type, oxygen levels, and culture media were examined. The available literature mainly reports results about birthweight and/or prematurity risks.

Concerning incubators, insufficient data was published to perform a meta-analysis, in particular regarding time-lapse incubators. If this technology theoretically allows for optimal embryo selection with minimal disruption to culture conditions, there is no good evidence for any advantage or disadvantage in terms of pregnancy, live birth, or miscarriage rates [59]. Although time-lapse incubators contribute to reduce the changes in culture conditions, investigation about their safety remains limited, and further investigations should be conducted in order to conclude on their potential capacity to improve perinatal outcomes.

For oxygen tension, our meta-analysis did not reveal any significant difference in birthweight or gestational age at birth between cultures under atmospheric oxygen tension or low oxygen tension. Several studies on human embryos have reported the superiority of low oxygen tension in terms of embryo quality and blastulation rates [60, 61], and, although the evidence was of low quality, the most recent meta-analysis revealed a small but significant increase in live birth rates when low oxygen tension is used during embryo culture [15]. However, although supposedly more physiological, investigating whether such a change in embryo culture conditions could affect obstetric or perinatal outcomes remains crucial. Although only three studies were available in the literature [36–38], our results show that oxygen tension during embryo culture does not seem to affect birthweight and gestational age at birth, with low heterogeneity. One study also evaluated other perinatal outcomes, without any significant difference for preterm births, cesarean sections, or congenital malformations [36]. Taken altogether, these results suggest that embryo culture using incubators offering low oxygen tension should be preferred, as recommended by the latest guidelines published by the European Society of Human Reproduction and Embryology (ESHRE) [62].

While many studies have evaluated the efficacy of different culture media, no consensus has yet been reached on whether certain culture systems (sequential vs. one-step) or certain media themselves would offer some advantages over others in terms of live birth rates. Studies showed that conflicting results and comparisons were made difficult by the numerous commercially available culture media [63, 64]. Safety studies evaluating obstetric and neonatal outcomes following the use of different culture media are receiving increasing attention. Among the published studies, some of them did report significant differences in terms of mean birthweight depending on the medium used for embryo culture [40–42, 46]. Although the precise mechanisms explaining these findings remain to be clarified, the proposed hypothesis relies on the influence of the culture medium on levels of embryo gene expression [55], as well as on epigenetic regulation during early embryo development [65]. However, most of the published studies did not find any significant difference in perinatal outcomes of different culture media [18, 37, 44, 47, 48]. Here too, studies evaluated different culture media, limiting the comparability across studies and underlining the interest of a network meta-analysis. Finally, our network meta-analysis failed to show any significant difference for birthweight, risk of low birthweight, or preterm birth according to the type of culture media. These findings are in line with the comparable DNA methylation signature that was observed in a recent study evaluating genome-wide analysis of IVF neonates following embryo culture in two different media [66]. At last, no difference in birthweight was observed in the sensitivity analysis which was performed on fresh embryo transfers only. Indeed, cryopreservation could introduce a major variable since multiple studies have shown that frozen embryo transfers are associated with improved neonatal outcomes such as reduced risks of PTB, LBW, and SGA compared with fresh embryo transfer [4] but with higher risks of macrosomia and LGA [67].

Our study presents some limitations. Unfortunately, we were unable to perform any meta-analysis about obstetric outcomes, due to limited published evidence. Available neonatal outcomes were mostly birthweight and risk of prematurity or gestational age at birth, and we could not conclude about other neonatal outcomes. For oxygen tension, one single study represented 60% of the results, limiting the interpretability of our results. Moreover, outcome truncation could introduce a bias that may lead to misinterpretation of the results since higher live birth rates are expected with low oxygen tension when compared to atmospheric oxygen tension. Furthermore, as the type of incubator often dictates oxygen tension and even culture medium, correlations between those three culture conditions should be considered, especially as embryo development could be affected differently by the culture media formulation according to oxygen tension [20]. Due to the lack of RCT in the literature, most of the included studies were retrospective cohort studies, relying on observational data with intrinsic limits since culture conditions are confounded with clinical practices and time. Besides, only unadjusted estimates have been considered, and other potential factors that could have been associated with perinatal outcomes, such as parity, ethnicity, or socioeconomic status, were hardly never mentioned in the included studies, making it impossible to perform a meta-regression. However, since our NMA did not show any difference, it is unlikely that any adjustment for potential confounding factors could reveal a significant difference between culture media. Finally, although we were able to evaluate oxygen tension and culture media, 2 key variables, there may be other lab parameters that could affect embryonic development and fetal growth.

Nevertheless, our review presents several strengths. So far, previous reviews were based on comparing the results of individual studies mostly examining the differences between 2 (sometimes 3) culture media. They could not conclude on the question of the association with perinatal outcomes, since there is a very large variety of culture media, which makes it impossible to perform any pairwise meta-analysis. We were able to go further relying on a network meta-analysis, which has the advantage of combining direct and indirect evidence from published studies. The included studies provided a dense and well-connected network for analyses, and procedures to investigate the possibility of inconsistency were performed in order to limit the risks of misleading conclusions. Moreover, the relatively small deltas between the groups confirm that this lack of difference should not correspond to a lack of power.

In the context of the ongoing debate about the putative superiority of one medium over another in terms of LBR [19, 37], potential consequences on children’s health could be considered in the choice of culture media in an IVF lab. Our findings indicate that there is currently no formal evidence to recommend any culture medium over another in terms of perinatal outcomes and suggest that we could be quite reassured regarding the potential consequences of IVF culture conditions on children’s health. Intuitively, it is indeed likely that the DOHaD effects of these few days in the laboratory are minor for human embryos when compared to the potential impacts during the preconception and post-implantation stages.

Nevertheless, the available literature and the quality of evidence appear as currently insufficient to formally conclude about the potential consequences of culture conditions on obstetric and perinatal outcomes. In a recent retrospective study analyzing singleton deliveries from a large validated database, no statistically significant differences were found between single-step and sequential media systems in the odds of placental abnormalities, gestational diabetes, and hypertension. On the other hand, single-step culture was associated with an increased risk of LGA, but not with prematurity, SGA, or low birthweight [68]. In another recent retrospective cohort study, a sequential Vitrolife medium was associated with a higher risk of placenta praevia, while the sequential Cook medium was associated with a higher risk of macrosomia [69]. According to the authors, these recent results support the hypothesis of the influence of culture media on obstetric and perinatal outcomes of IVF children. However, given their limitations, further research is needed to evaluate these culture condition–induced effects on human embryos. Ideally, randomized prospective studies should be proposed, controlling for all other laboratory variables, as the choice for the type of incubator, oxygen tension, and culture medium is often intimately intricated. Moreover, even reassuring results concerning perinatal outcomes would be insufficient to exclude any effect of culture conditions on the health, growth, and development of children conceived following IVF. Culture medium composition has also been linked to differences in postnatal weight [70, 71] and childhood developmental profile [39], advocating not only for further studies but also for transparency and disclosure from the industry about precise formulations.

In conclusion, no substantial difference was observed for birthweight or risk of prematurity according to oxygen tension and culture medium used during the IVF process. These results underline that we currently do not have any formal evidence that could lead to a recommendation to use one or another culture condition when dealing with perinatal outcomes. Further research evaluating the effects of culture conditions both on success rates and on long-term health issues should be carried out in order to increase data about the safety of our culture conditions regarding obstetric and neonatal outcomes.

Author contribution

CS: collection of data, statistical analysis, interpretation of data, and writing of the manuscript. NAY: collection of data, interpretation of data, and critical revisions of the manuscript. PP: collection of data, interpretation of data, and critical revisions of the manuscript. CS: interpretation of data and critical revisions of the manuscript. MG: study design, analysis and interpretation of data, and critical revisions of the manuscript. NS: study design, collection of data, analysis and interpretation of data, and writing of the manuscript.

Funding

This work was sponsored by an unrestricted grant from Gedeon Richter France. The funder had no role in the study design, screening, quality assessment, data extraction.

Declarations

Competing interests

The authors declare no competing interests.