Cerebral palsy in children born after assisted reproductive technology:a meta-analysis

Fang-Fang Wang 1,2 · Tao Yu · Xiao-Lu Chen · Rong Luo · De-Zhi Mu

Abstract

Keywords Assisted reproductive technology · Cerebral palsy · Meta-analysis

Introduction

Cerebral palsy (CP) is a group of permanent motor disorders that are often accompanied with a wide range of comorbidities [ 1]. Prevalent in about 1.5-3.0 per 1000 live births around the world [ 2], CP is one of the most common causes of motor disability in children and is associated with a substantial burden on families and society [ 3]. The etiology of CP remains uncertain. The etiology is considered to be multifactorial and heterogeneous, and events of the prenatal, perinatal, and postnatal periods alone or in combination may be implicated [ 2, 4]. Many risk factors have been established, including preterm delivery, multiple pregnancy,low birth weight, male sex, intrauterine infection, congenital anomalies, genetic variation, maternal diseases and a maternal age of 40 years or older [ 5- 7]. In recent years, growing evidence has shown that assisted reproductive technology(ART) treatment is associated with an increased risk of adverse perinatal sequelae, such as preterm delivery, multiple birth, low birth weight, birth defects, and intrauterine growth restriction [ 8- 10], which may account for the development of CP.

ART is the technology used to establish pregnancy in infertility treatments, including in vitro fertilization(IVF), intracytoplasmic sperm injection (ICSI), artificial insemination, and ovarian stimulation (OI), among others [ 11]. The estimated proportion of infants born after ART ranges from 1.5 to 6% in the developed world[ 8, 12- 14], and about 1 million babies are born each year after ART worldwide [ 15, 16]. With its acceptability by more and more people, the effect of ART on the risk of CP has attracted close attention of researchers, and several studies have assessed the risk of CP in ART-conceived children. However, the results have been inconsistent. Some studies have reported that ART children were at higher risk for the occurrence of CP compared with those born after naturally conception[ 14, 17- 26], while others have found no risk [ 27- 30],or even a potential protective effect [ 31]. Two metaanalyses have been made while attempting to evaluate the long-term neurodevelopmental outcome of offspring conceived following ART [ 32, 33]. However, the evidence was limited because these meta-analyses either were conducted more than 10 years ago, and thus did not include recent studies, or were based on relatively small sample sizes. Therefore, studies that have comprehensively pooled evidence on the association between CP risk and the use of ART remain sparse.

Using recently accumulating evidence, our goal was to estimate the risk of CP in ART children compared with naturally conceived children by conducting a systematic review and a meta-analysis of published studies. Considering the elevated risk of multiple births and preterm deliveries in children born after ART, we also reported the findings of singletons, multiples, and preterm births separately whenever possible.

Methods

Literature search

The present study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)[ 34] and was registered with the International Platform of Registered Systematic Review and Meta-Analysis Protocols (INPLASY202130060). PubMed, Embase, the Cochrane Library, and Web of Science were searched independently by two authors (WFF and YT) from inception to November 22, 2020 for purposes of identifying relevant articles that evaluated the relationship between CP and ART. The following medical subject headings(MeSH) and keywords were used to search for the first theme: “assisted reproductive technology”, “assisted reproductive technique”, “in vitro fertilization”, “IVF”,“intracytoplasmic sperm injection”, “ICSI”, “artificial insemination”, “ovarian stimulation”, and “infertility treatment”, and “OR” was used to connect relevant search terms. For the second theme, MeSH and key words such as “cerebral palsy”, or “little disease” were used. The first theme and the second theme were combined using“AND”. We restricted our search to human studies written in English. We also hand searched the reference lists of the eligible articles to obtain potentially relevant studies.We did not attempt to locate unpublished studies.

Study selection

Two authors (WFF and YT) separately screened duplicates, the titles and abstracts, and full texts of the studies.Studies were included if they met the following inclusion criteria: (1) studies investigated the relationship between CP and ART; (2) the study population contained all live born children, singletons or multiple births; (3) the exposure of interest was ART; (4) the control group was naturally conceived children; (5) the outcome of interest was CP; (6) the study design was cohort or case-control study; and (7) the effect estimates [odds ratios (ORs)or risk ratios (RRs)] could be extracted or calculated.Reviews, case reports, conference abstracts, comments,animal studies or studies with insufficient information were excluded. If more than one paper involved the same population, then the most comprehensive and recent study was selected.

Data extraction and quality assessment

Two authors (WFF, YT) independently extracted data to ensure the reliability of the results, using a standard data extraction form. Any disagreements were resolved through discussion with a third author (CXL). The following information was extracted from each included study:first author’s name, year of publication, study region,study period, study design, source of the study population, ascertainment of ART and CP, ART type, sample size, confounding factors and effect measures (ORs,RRs) with corresponding 95% confidence interval (CI).If effect estimates were not provided in an article, we calculated ORs and 95% CIs from the raw data. Where both adjusted estimates and crude estimates were available from a particular study, the adjusted estimates were extracted and used in data syntheses.

The methodological quality of all the included studies was examined using the Newcastle-Ottawa Scale (NOS)by two authors (WFF, YT) independently. Any potential disagreements were resolved through discussion with a third investigator (CXL). The NOS is designed for nonrandomized studies and consists of an eight-item process for assessing the selection of the study population(maximum 4 scores), comparability (maximum 2 scores),and the evaluation of exposure and outcome (maximum 3 scores). Quality was assigned as high with 7-9 scores,moderate with 4-6 scores and low with 0-3 scores.

Statistical analysis

We calculated pooled effect estimates for all studies combined as well as singletons, multiples, and preterm births separately whenever possible. Since the absolute risk of CP was low, the OR was considered to be equivalent to the RR.Therefore, we used the OR to estimate the risk of CP in ART children. The ORs and their 95% CIs were inputted into the software and then were logarithmically transformed. The corresponding standard error (SE) was calculated using the equation from Cochrane handbook: SElogOR(log(upper 95%CI) - log(lower 95% CI))/3.92 [ 35]. The logOR and SElogORwere combined using an inverse-variance approach. A fixed effects model was adopted to combine data when there was no obvious heterogeneity among studies; otherwise, a random-effects model was used [ 36]. The heterogeneity was assessed using theI2 inconsistency test and Cochran’sQtest [ 37].I2 ＜ 50% andPvalue ＞ 0.10 for theQtest indicated no obvious heterogeneity. To examine the sources of potential heterogeneity, we conducted subgroup analyses on study quality, study location, study design, CP sources and whether they were adjusted for confounding factors. To assess whether a single study significantly affected pooled estimates, sensitivity analysis was performed by removing one study at a time. Funnel plot together with the Begg’s test were used to assess potential publication bias [ 38]. All statistical analyses were conducted using Stata 14 (StataCorp LP, College Station, TX). The significance level was set asP＜ 0.05, unless otherwise specified.

Fig. 1 Flow chart of study selection

Results

Literature selection

We identified 430 potential records (47 from PubMed, 224 from Embase, 114 from Web of Science, and 45 from the Cochrane Database). Of these records, 107 were excluded because they were duplicates, and 296 were further excluded following screening of titles and abstracts. The remaining 27 records were eligible for full-text screening. Of these remaining records, five were excluded because they did not report CP as an outcome, three were excluded because they had no or inappropriate control groups, and 10 were excluded owing to overlapping data [ 22- 26, 28, 30, 39- 41].Finally, nine articles were included in our meta-analysis [ 14,17- 21, 27, 29, 42]. A flow diagram of the literature selection process in accordance with PRISMA is shown in Fig. 1.

Study characteristics

Our meta-analysis included seven retrospective cohort studies and two case-control studies involving 89,214 ART children and 4,160,745 natural conceived children.The included studies were published between 2006 and 2018. A total of 75,772 (85%) ART children were born in Europe, 9118 (10%) were born in Oceania and 4324 (5%)were born in Asia. The characteristics of included studies are listed in Table 1. There were eight studies using birth registry data, and one study using birth data from a single medical center. Seven studies selected all living births(involving both singletons and multiples) for analysis, and two were restricted to singletons. In terms of the type of ART, seven studies reported data on CP concerning IVF or IVF/ICSI together, three studies specifically focused on OI, and in one study the type of ART included also gamete intrafallopian transfer. Cases of CP were identified through CP registries in five studies and through hospital discharge registers in four studies. In most studies, the diagnosis of CP was confirmed not earlier than at 2 years of age. Two studies provided only crude ORs or raw data, whereas others provided adjusted ORs that controlled for conventional risk factors for CP, such as year of birth, sex, maternal age, parity, socioeconomic status, and maternal smoking during pregnancy. Six studies were considered of higher methodological quality, and these studies contributed 85%of ART children in the present meta-analysis.

Data analysis

The pooled OR based on all nine studies combined was 2.17 (95% CI 1.72-2.74) indicating a 117% increased risk of CP in ART children. This pooled estimate included studies of singletons only and of studies that have grouped all births together. The individual point estimates for these studies varied from 0.83 to 4.90. There was substantial heterogeneity among the studies (I2 = 65.3%,P= 0.003)(Fig. 2).

The heterogeneity among studies was no longer significant when we restricted the analysis to studies where the CP risk had been discussed separately in ART singletons or ART multiples. The pooled OR for 63,602 ART singletons compared with 3,864,213 naturally conceived singletons was 1.36 (95% CI 1.16-1.59). A low heterogeneity was found(I2 = 20.8%,P= 0.265) (Fig. 3). The pooled OR comparing CP in 22,941 ART multiples and 92,533 naturally conceived multiples was 1.05 (95% CI 0.86-1.29). No evidence of heterogeneity was found (I2 = 0.0%,P= 0.948) (Fig. 4).Some studies also discussed the effects of preterm birth on the association between CP risk and ART. Hvidtjørn et al. [ 18] found the observed effect changed from 1.72(95% CI 1.39-2.12) to 0.96 (95% CI 0.77-1.19) after adjustment for gestational age. Two studies [ 19, 21]reported the estimates in a stratum of children carried to preterm, and the pooled OR for CP risk in preterm children born after ART was 1.53 (95% CI 0.66-3.56) (Fig. 5).For singletons, one study evaluated the risk of CP in ART stratified by gestational age and found the risk was only increased in the under 32-week group (OR = 2.7, 95% CI 1.0-6.9) [ 14].

Sensitivity analysis and publication bias

Given the low number of included studies in our metaanalysis, we carried out sensitivity analyses by removing one study a time. The pooled ORs for all analyses (i.e.,all studies combined, singletons only, multiples only) did not vary materially when omitting any single study. Therefore, no individual study in this meta-analysis significantly affected our pooled OR estimates, and our results were stable and reliable (Supplementary Fig. 1). The funnel plots for all studies combined, singletons, and multiples were fairly symmetrical (Supplementary Fig. 1). According to the Begg’s test, there was no evidence of obvious publication bias among the included studies for all studies combined (P= 0.466), for singletons only (P= 0.902), or for multiples only (P= 0.452).

Subgroup analysis

A summary of the main and subgroup meta-analyses within the different plurality groups stratified by study quality,study location, study design, CP sources, and adjustment for confounders is provided in Table 2. The pooled estimates based on high quality studies showed significantly increased risk of CP in ART children and ART singletons and were comparable to that of overall results. For all studies combined, significantly stronger associations between CP risk and ART were found in European studies, in cohort studies with adjustment for confounders, and in which CP was ascertained through CP registries. For singletons, significantly stronger associations between CP risk and ART were found in Oceanian studies, in cohort studies with adjustment for confounders and in which CP was ascertained through CP registries. The summary risk estimates for CP in ART multiples were generally similar and comparable to that of the overall result when stratified by different factors.

Discussion

The present study is a comprehensive systematic review and meta-analysis of published studies to address the issues of whether an increased risk of CP exists in ART children compared with those conceived naturally and whether the risk differs when singletons, multiples and preterm births are examined separately. Our findings involving 89,214 ART children and 4,160,745 naturally conceived children suggest a 117% increased CP risk in ART children. Our findings also suggest that this risk decreases when singletons are examined separately (36%). However, a statistical association between ART and CP is not observed either in ART multiples or in ART preterm births. The findings are further confirmed when the pooled estimates were restricted to high methodological quality studies. This is an important observation because it indicates areas of subsequent investigation.

There are two published meta-analyses that compared CP in ART and naturally conceived children [ 32, 33]. They both reported a significantly higher risk of CP in ART children, as we did. The meta-analysis of Hvidtjørn et al. [ 32]included only three cohort studies in the Scandinavian countries. Care must be taken when generalizing these results to areas where socioeconomics, ART practices, demographics,and health care systems differ. The meta-analysis of Djuwantono et al. [ 33] was interested in assessing a range of neurodevelopmental outcomes after ART compared with naturally conceived children. Although the studies included pooled estimates of CP data, the authors’ search for CP data was not thorough and their results included only five studies involving 22,139 ART children. In addition, the authors did not evaluate singletons and multiples separately. In our meta-analysis, we added studies from Oceania and Asian populations containing available data that could provide sufficient statistical power to generate precise effect estimates.We pooled data from 89,214 ART children, and 85% of these children came from the six studies which we considered of high quality for assessing CP risk. We also performed stratified analyses for singletons, multiples, and preterm births separately.

Table 1 Characteristics of the included studies

The advent of ART has led to an increase in the rates of multiple birth and preterm delivery worldwide during the past 4 decades [ 43- 46]. Both of these two variables are also well-known risk factors for CP development. Several studies have evaluated the role of multiplicity and prematurity in the association between CP and ART. In a Swedish cohort, the investigators reported that the overall risk of CP in IVF children was increased almost fourfold compared with matched controls. Using logistic regression analyses and stratifying for plurality, low gestational age showed independent effects on increasing the risk for CP in all IVF children and IVF singletons [ 23]. Subsequently, a Danish study showed the importance of prematurity as an independent risk factor for CP in IVF children [ 25]. The authors found that nearly 59% of CP cases among children born after IVF may be attributed to preterm birth, whereas this fraction was only 19% in naturally conceived children. Because twin births accounted for 78% of the preterm IVF children in the study,the authors indicated the importance of prevention of the high frequency of multiple pregnancies in IVF children.Klemetti et al. [ 17] reported a threefold increased CP risk in IVF children when both multiples and singletons were included. Similarly, a higher proportion of preterm births in the IVF group was found (23.6% in the IVF group vs.5.5% in natural conceptions), and this was largely attributed to higher rates of multiple births. The authors concluded that reducing the number of embryos transferred would improve the health of IVF children. In a 2010 Danish cohort study, Hvidtjørn et al. [ 18] reported an increased risk of CP among 33,139 ART children (RR = 1.72, 95% CI 1.39-2.12)compared with naturally conceived children. The authors found more babies came from multiple pregnancy (29.1%vs. 2.4%) and preterm delivery (19.5% vs. 5.6%) after ART than natural conception. After adjustment for multiplicity or gestational age, the observed effect decreased with CI’s including 1. The author concluded that the increased risk for CP in ART children was largely associated with multiplicity and prematurity in these children [ 18]. In our meta-analysis stratification for plurality or prematurity of significance for CP also resulted in reductions or loss of significant risk estimates. These results highlight the impact of multiple pregnancy and prematurity on the overall rates of CP in ART children. This is of great importance from a public health perspective because the number of multiple pregnancies and premature births in ART can clearly be reduced by use of single embryo transfer (SET).

Although previous researches have yielded conflicting results regarding risk of CP in singletons born after ART[ 14, 18- 20, 23, 24], our study confirmed that ART singletons experienced a significantly increased risk for CP.Because only one study evaluated the risk of CP in ART singletons stratified by gestational age, it was impossible for us to examine pooled estimates in singletons born preterm.The etiology behind the increased risk of CP in ART singletons remains unclear and warrants further research. One possible explanation is the vanishing embryo phenomenon(number of transferred embryos was larger than the number of infants born) in ART pregnancy [ 39, 47, 48]. Vanishing embryo rates have been reported to vary between 5 and 30%following ART [ 49, 50], and surviving singletons following a vanishing embryo are at higher risk of adverse outcomes,such as preterm birth, low birth weight, and intrauterine growth restriction [ 51, 52]. As the use of SET increases,any impact of vanishing embryo may diminish. Indeed, two recent studies reported no cases of CP were found in singletons born after SET [ 14, 18].

Another potential explanation for the increased CP risk in ART singletons is the subfertility itself, including the relatively advanced age of subfertile couples and the underlying cause of their subfertility. There is some evidence of moderate quality that naturally conceived singletons from subfertile couples have increased risks of preterm birth and low birth weight when compared with singletons from fertile couples [ 53]. Subfertile women have had a higher risk for adverse birth outcomes even without infertility treatments [ 54]. One Australian study found the risk of CP was increased among children of subfertile couples even when they were conceived without the aid of ART [ 27]. A 2005 Swedish study also reported that the relation between CP and IVF disappeared after correcting for the years of unwanted childlessness [ 28]. However, Zhu et al. failed to find a significant association between the risk of CP and untreated subfertility measured by time to pregnancy [ 26].The impact of subfertility on the overall risk of CP still needs further study.

Third, the procedures of ART may increase the risk of CP,such as ovarian stimulation, in vitro culture of the embryo,the freezing and thawing of embryos, possible damage during the process of ICSI and altered hormonal environment.Using regression analysis, Strömberg et al. [ 23] found IVF independently contributed to risk of development of CP.Some studies reported a significantly higher risk of preterm birth and low birth weight in IVF/ICSI versus naturally conceived singletons from a subfertile population [ 55, 56].A meta-analysis of preterm birth in ART versus non-ART siblings including two Nordic studies showed a significantly higher risk for preterm in the ART sibling [ 53]. These results indicate that factors related to ART treatment per se also may play a part.

Fig. 2 Forest plot of cerebral palsy risk after assisted reproductive technology in all studies. ES effect size, CI confidence interval

Fig. 3 Forest plot of cerebral palsy risk after assisted reproductive technology in singletons. ES effect size, CI confidence interval

In addition, there appear to be epigenetic alternations,such as DNA methylation and gene expression changes,in ART children related both to multiple manipulations required during the ART procedures and to the poor quality of gametes [ 57- 60]. A meta-analysis by Lazaraviciute et al. [ 61] reported an increased risk of imprinting disorders in ART children with a pooled OR of 3.67(95% CI 1.37-9.74). Genetics have been reported to play a significant role in causation of CP and up to one third of CP cases likely have genetic causes [ 6, 62]. A recent study by Mohandas et al. [ 63] showed consistent regional differences in DNA methylation in CP children. However,this study did not analyze children conceived after ART.Considering the very limited data about genetic studies in ART children, whether or not there is a genetic mechanism in the pathogenesis of CP after ART remains unclear.

Fig. 4 Forest plot of cerebral palsy risk after assisted reproductive technology in multiples. ES effect size, CI confidence interval

Fig. 5 Forest plots of cerebral palsy risk after assisted reproductive technology in prematurities. ES effect size, CI confidence interval

There was substantial heterogeneity when all studies were combined to assess the risk of CP and ART, which was not surprising given the differences in sample source,sample size, geographic region, characteristics of populations, ART types, assessment of CP, quality score, and adjustment for confounding factors. The heterogeneity was still observed in some subgroup analyses, which indicated that the sources of heterogeneity among studies may be complex. However, the association between CP risk and ART did not differ significantly after the sensitivity analyses indicating that our results were stable and reliable.

There are several limitations in the present study. First,the small number of included studies limited the generalizability of our findings, as well as the meta-regression analysis to identify sources of heterogeneity. Second,there were a lack of data for subgroup analysis based on type of ART or for data analysis that adjusted for known confounders, such as small gestational age, low birth weight and male gender. In addition, the extent to which the observed CP risk was mediated through multiple gestations cannot be identified. Third, the biological mechanism underlying the relationship between CP and ART cannot be specified in the present study. Fourth, the inherent limitation of included studies is a concern. Studies of CP risk and ART often choose children born to normal women with natural conceptions rather than infants born to subfertile women with natural conceptions as a control group,which may not be appropriate because subfertility per se is associated with adverse birth outcomes, such as preterm birth and low birth weight. It is difficult to eliminate the effects of factors related to the underlying subfertility.Further studies with singletons from subfertile couples as control groups or studies with sib-ship designs may add significantly to our knowledge in this field.In summary, the present study suggests that the risk of CP is increased in children born after ART, and the increased CP risk for ART is largely due to the high rates of multiplicity and preterm birth in ART children. A more widespread use of SET may warrant consideration to improve the long-term health of ART children. Due to the limited number and the substantial heterogeneity of included studies, our findings should be viewed with caution, and further well-designed studies are needed to understand the strength of association and explain the underlying mechanism.

Table 2 Summary of subgroup analysis of the meta-analysis results

FundingThis work was funded by the Sichuan Provincial Department of Science and Technology Regional Innovation Cooperation Project(2020YFQ0021).

Compliance with ethical standards

Ethical approvalThis article does not contain any studies with human participants or animals performed by any of the authors.

Conflict of interestNo financial or non-financial benefits have been received or will be received from any party related directly or indirectly to the subject of this article. The authors have no conflict of interest to declare.

Data availabilityThe datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Supplementary InformationThe online version contains supplementary material available at https:// doi. org/ 10. 1007/ s12519- 021- 00442-z.

Author contributionsWFF and YT contributed equally to this manuscript. WFF and YT contributed to study concept and design, acquisition, analysis and interpretation of data, drafting the manuscript. CXL contributed to analysis and interpretation of data, drafting the manuscript. LR contributed to study concept and design, critical revision of the manuscript, acquisition of funding, and supervision. LR had full access to all the data in the study and takes responsibility for the integrity of the data. MDZ contributed to analysis and interpretation of data, and critical revision of the manuscript. All the authors read and approved the final version of the manuscript.