Intellectual Disability in Children Conceived Using Assisted Reproductive Technology

This retrospective population-based cohort study included LBs from 1994 to 2002 in Western Australia. ART was defined as children conceived by using standard IVF or ICSI. Data on all children with ID were obtained from the population-based Intellectual Disability Exploring Answers (IDEA) database.¹ Cases in this register are ascertained from the state-governed Disability Services Commission and the public education system (for children with ID who receive education support). Criteria for inclusion are a full IQ score of <70 and deficits in adaptive functioning with evidence of developmental delay before 18 years of age. For Disability Services Commission–ascertained cases, information is available about the main cause of ID assigned by physicians and recorded in the IDEA database using the American Association on Intellectual and Developmental Disabilities (formerly the American Association on Mental Retardation) classification system.^17,18 Diagnoses of ID to December 2010 were included, giving at least 8 years of follow-up for all children in the cohort. IDEA database records were linked to 4 other data sources: (1) the Midwives Notification System, a statutory collection of antenatal and perinatal data on all Western Australian births from 20 weeks’ gestation (or ≥400 g); (2) the Reproductive Technology Register, a statutory collection of treatment data from Western Australian fertility clinics; (3) the Western Australian Register of Developmental Anomalies (WARDA), in which data on BDs diagnosed up to 6 years of age (major or minor classification and the exclusion list are available on the King Edward Memorial Hospital Web site¹⁹) and diagnoses of cerebral palsy²⁰ are collected; and (4) the Registry of Deaths (linkage with this data set was updated in March 2016). Data were linked by the Western Australian Health Department Data Linkage Branch using probabilistic matching and deidentified for provision to researchers. Ethical approval was obtained from the Western Australian Department of Health Human Research Ethics Committee (2011/80).

Figure 1 describes the children excluded from our cohort. These comprised (1) Aboriginal births because they represented <0.5% of our ART cohort (versus 6% of non-ART births) and are at increased risk of ID²; (2) births conceived by using gamete intrafallopian transfer (n = 178; 1 with ID), partial zona dissection, and subzonal insemination (n = 15; no ID), which are too few in number to allow for meaningful comparisons; (3) children who died before the age of 7 years without a diagnosis of ID (N = 1056) because some diagnoses may not be made until a child attends school (the characteristics of children who were excluded for this reason are presented in Supplemental Table 6); and (4) children with a known postneonatal cause for ID (eg, postnatal injury [N = 31]).

ID was categorized into mild or moderate (IQ 40–69) or severe (IQ <40).¹ Known causes of ID recorded in the IDEA database or in linked WARDA records were reviewed by 3 authors (M.L., H.L., and G.B.), who were blinded to conception status, and presented in a hierarchical fashion similar to that of Yeargin-Allsopp et al,²¹ with each child classified according to 1 main cause of ID. In children with no identified sufficient cause, we describe the presence of pregnancy and/or birth complications and coexisting conditions. ART was categorized as “any or none,” with subcategories of IVF, ICSI, and fresh and frozen-thawed embryo transfer. Analyses were stratified by plurality and prematurity when numbers permitted. The following variables were considered potential confounders: child’s sex, parity (3 groups), maternal and paternal age at birth (7 groups each), maternal socioeconomic status (SES) (area-based social disadvantage index in centiles²²), maternal smoking during pregnancy, private health insurance (PHI) at birth, year of birth (1994–1996, 1997–1999, and 2000–2002), delivery method (vaginal, elective cesarean delivery, and emergency cesarean delivery), and marital status (married and/or cohabiting versus other).

Univariate statistics (cross-tabulations with χ² tests and means and/or t tests) were used to identify variables associated with both ID and ART. Risk ratios (RRs) and 95% confidence intervals (CIs) were estimated by using Poisson regression with robust SEs. Regression models were adjusted for covariates with no missing data (child’s sex, maternal age, parity, year of birth, delivery mode) and then marital status and PHI. In final models, we used generalized estimating equations (specifying an exchangeable correlation structure) to account for the potential correlation in results between births from the same mother. Marital status was missing for 0.1% of births, and PHI was missing for ∼6%, but this did not differ by ID status. The following variables met the criteria for confounding but were not included: paternal age (highly correlated with maternal age), maternal smoking (only available from 1997 and made only a 3% change to the model), and maternal SES (more missing data than for PHI and less representative of the socioeconomic gradient between ART and non-ART births, being an area-based rather than individual measure). Supplementary analyses were used to investigate the impact of missing data and of additionally adjusting for maternal SES.

RRs for any ID, mild or moderate ID, and severe ID were estimated for all children. Analyses of any ID were also conducted for singletons and twins separately and by preterm or term status. Data were too sparse to allow for an analysis of higher-order multiples (HOMs) separately. ART subtypes were compared by using both non-ART and standard IVF as comparison groups.

A 2-sided P < .05 was considered statistically significant. Data were analyzed with SPSS version 24 (IBM SPSS Statistics, IBM Corporation, Armonk, NY).

There were 225 728 LBs in Western Australia between 1994 and 2002; of which 3118 (1.4%) were conceived with ART. After exclusions (Fig 1), 2876 ART-conceived and 207 751 non–ART-conceived children were available for analysis. Although all children had at least 8 years of follow-up, those in the ART group were younger on average than those in the non-ART group (median age 10.6 years versus 12.8 years), reflecting the increasing use of ART over time. All ART procedures involved cleavage stage transfers; 182 (6.3%) were single embryo transfers (SETs).

Mothers of ART-conceived children were older, less likely to have smoked, more often primiparous, and more likely to have PHI and be of higher SES (Table 1). ART-conceived children were more likely to be twins or HOMs (33.6% ART versus 2.5% non-ART), be born preterm, have lower birth weight, have been delivered by cesarean, and to have a major BD (Table 1).

There were 3551 children diagnosed with ID (16.9 per 1000 LBs), 60 of whom were ART conceived (20.9 in 1000), comprising 37 singletons and 23 multiples. The proportion of severe ID was higher in ART-conceived (12.9%) than in non–ART-conceived children (5.9%).

ART-conceived children were at an increased risk of any ID, mild or moderate ID, and severe ID in adjusted analyses (Table 2). Risks remained when analyses were restricted to singletons and were highest (although not statistically significant) for those born very preterm. A similar pattern of higher risk was observed for ART-conceived twins of <32 weeks’ gestation. Median tests revealed no significant difference in gestational age of preterm (<32 and 32–36 weeks) ART-conceived and non–ART-conceived singletons or twins (data not shown), suggesting that the increased risk of ID among ART births in these strata was not attributable to lower gestational age. Missing data had minimal impact on the RRs, as did additional adjustment for maternal SES (Supplemental Tables 7 and 8).

When comparing ART subtypes to non-ART, ICSI-conceived children had the highest risk of ID (RR 2.54; 95% CI 1.69–3.83); both fresh and frozen-thawed embryo transfer were associated with a 60% elevation in risk (Table 2). Surgically retrieved sperm was used in 190 (20%) ICSI cycles resulting in an LB. A greater proportion of children were diagnosed with ID after ICSI with ejaculated (3.48%) versus surgically retrieved sperm (1.58%). The risk of ID was also elevated for ICSI-conceived children versus those who were conceived using IVF (RR 1.98; 95% CI 1.12–3.48; Table 3). When stratified by sex (Supplemental Table 9), the highest RRs were seen for ICSI-conceived girls (regardless of whether the comparison group was non–ART-conceived or IVF-conceived girls).

Known causes of ID among ART-conceived and non–ART-conceived children are shown in Table 4. Although the numbers in some categories are small, children born after ICSI were more likely to have a known genetic condition than IVF-conceived or non–ART-conceived children (28% vs 13% and 12%, respectively). Down syndrome was the most frequent genetic cause of ID for both ART-conceived and non–ART-conceived children, accounting for 10% of ART and 5% of non-ART cases. Conditions that are sometimes caused by imprinting errors (Angelman, Prader-Willi, and Russell-Silver syndromes) were diagnosed in 5% of ART versus 0.3% of non-ART cases (data not shown). No cause of ID was identified for 72% of ART-conceived and 82% of non–ART-conceived children. Among the children with no identified cause for ID, several pregnancy complications, low birth weight, being a twin or HOM, and having cerebral palsy were more likely in ART-conceived than in non-ART-conceived children (Table 5).

With our findings, we suggest that children conceived using ART have a small increased risk of ID even when analyses are restricted to singleton births. The risk of ID is doubled among very preterm births, after ICSI, and for severe ID. ID was also more common in children born after ICSI versus standard IVF.

The cumulative prevalence of ID in all children in Western Australia remained stable over our study period.² Comparisons with other studies are complicated by differences in study methods, case definitions, and follow-up periods. However, our findings are consistent with Knoester et al,²³ who found that 5- to 8-year-old children born after ICSI (compared with non-ART) had a lower IQ, and Zhu et al,²⁴ who found that singletons born after infertility treatment had twice the risk of severe developmental delay at age 18 months compared with singletons born to fertile couples. The birth cohorts in 2 population-based studies covered time periods similar to ours: Bay et al²⁵ reported no increase in “mental retardation” in IVF-conceived and ICSI-conceived children in Denmark, whereas Sandin et al²⁶ reported an 18% increase in risk for ART-conceived versus non–ART-conceived children in Sweden and a 51% increase for ICSI versus IVF. The underlying population prevalence of ID in these studies was ∼6 per 1000 births (versus 17 per 1000 in our study), which was similar to Western Australian estimates before the inclusion of education data (7.6 in 1000).¹ The Western Australian register may therefore be capturing milder cases of ID than the Danish psychiatric and Swedish patient registers, including those diagnosed only when a child attends school. A 2011 meta-analysis of ID prevalence in population-based studies of children and adolescents reported a pooled prevalence of 18.3 in 1000,²⁷ which was similar to our study.

ID is a heterogeneous condition with many possible causes. We therefore stratified by known mediating factors (prematurity and plurality) that are more likely to occur after ART. The observed increased risk of ID after ART in very preterm births warrants further investigation in larger cohorts. We recently observed a higher risk of cerebral palsy in preterm (<32 weeks) ART versus non-ART singletons²⁸ and postulated that the loss of a comultiple (estimated to occur in 10% of ART singleton births resulting from multiple embryo transfer) might play a role in the greater risk of adverse outcomes in this group.^29,–31 

ICSI use was restricted to couples with severe male-factor subfertility in our cohort, but we cannot determine whether increased risks of ID relate to the procedure itself or to the underlying subfertility. Contrary to Sandin et al,²⁶ we found no evidence that ICSI with surgical sperm extraction was associated with a greater risk of ID than ICSI with ejaculated sperm. ICSI-conceived children were more likely to have a known genetic cause of ID. We showed previously that ICSI singletons had a twofold increased risk of chromosomal anomalies compared with non-ART births.³² Genetic abnormalities occur more frequently in men who are infertile, so ICSI (which bypasses natural selection barriers) may allow for the transmission of chromosomal anomalies to the offspring.¹² Advanced paternal age is also associated with higher risk of genetic mutations in sperm, several (autosomal-dominant) monogenic disorders, autism, and subtle deficits in neurocognitive outcomes.^33,–36 Thus, it is noteworthy that ICSI fathers had the highest median age (37 years) and that ICSI fathers of children with ID of a known genetic cause had a median age of 40 years (data not shown). Syndromes related to aberrant imprinting have been associated with ART.¹² Although 5% of ART-conceived children with ID were diagnosed with conditions that are sometimes caused by imprinting disturbance, we had no information to determine whether imprinting defects occurred in these children.

ICSI is now used to treat a broader group of patients,³⁷ which will allow for comparisons of ID risk in the offspring of couples with a range of causes for their subfertility. Comparisons of ID risk when ICSI is used for specific indications (eg, reduced semen parameters, preimplantation genetic screening, and failed IVF) would be particularly informative; however, these data are not usually collected in ART registers.

The prevalence of ID is known to be higher in male patients,³⁸ but in our study, it was similarly elevated in both girls and boys who were conceived using ICSI (31 per 1000 births). This finding needs confirmation in larger studies.

Strengths of this study include population-based ID data ascertained from both disability service and education sources.³⁹ The minimum 8-year follow-up makes it likely that we captured most ID cases in our cohort. Because ID may be diagnosed only at school commencement and ART-conceived children who died before age 7 years had higher rates of prematurity, low birth weight, and were more often multiples (all of which may be associated with a greater risk of ID^13,–15; Supplemental Table 6), we excluded children who died before 7 years of age without a diagnosis of ID. Although ART-conceived children have greater contact with the health system⁴⁰ (which may increase the likelihood of ID being diagnosed), such ascertainment bias is unlikely to affect diagnoses of severe ID (which were increased in our study) or differentially affect IVF and ICSI births. We were able to describe the known causes of ID using information from both the IDEA database and WARDA. This included 20% of ART-conceived and 12% of non–ART-conceived children with known genetic disorders. In contrast, Sandin et al²⁶ report that only 1.7% of ART-conceived and 2.3% of non–ART-conceived children with ID in their study had a known genetic disorder.

Study limitations included the relatively few ID cases in ART-conceived children, leading to imprecise estimates in some subgroup analyses. We made no adjustment for multiple comparisons and note that many RRs had lower confidence limits close to unity. Information about maternal obesity (potentially associated with both ID⁴¹ and ART⁴²) or parental psychiatric history was not available (although adjustment for psychiatric history did not alter risk estimates in the large Scandanavian studies^25,26). We were unable to identify children who were conceived using ovulation induction alone or intrauterine insemination, and they are therefore included in the non-ART comparison group. If these treatments are also associated with increased risks of ID, our risk estimates may be biased toward the null; however, there is limited and conflicting information in the literature in this regard.¹⁶ Missing data for 2 covariates did not appear to materially affect our RR estimates.

Our study included children born from 1994 to 2002, when multiple embryo transfer was common practice in Western Australia. Therefore, our observations that ART-conceived children with no identified cause for ID were much more likely than non–ART-conceived children to be twins or HOMs and have very low birth weight are not surprising. There have been major shifts in clinical practice in Australia since that time; SET is now used in 86% of cycles, resulting in important reductions in the prevalence of multiple births and pregnancies affected by vanishing twins and improvements in perinatal health.⁵ Preimplantation genetic testing has also increased and was used in 8.6% of cycles in 2015.⁵ However, in many other countries, the shift toward SET has occurred far more slowly. Risk of ID after ART may be particularly high in regions such as North America, Latin America, and the Middle East, where multiple embryo transfer is widely used and ICSI use rates are high.^37,43,44 

To prevent ID and reduce the burden it places on families and communities, it is important to identify even small modifiable risk factors so that they can be ameliorated. For ID associated with ART, such actions may include reducing preterm births via treatment strategies that favor SET and restricting ICSI treatment to couples with severe male-factor subfertility. These couples may opt to use preimplantation genetic testing to maximize the transfer of chromosomally normal embryos.

ART

assisted reproductive technology

BD

birth defect

CI

confidence interval

HOM

higher-order multiple

ICSI

intracytoplasmic sperm injection

ID

intellectual disability

IDEA

Intellectual Disability Exploring Answers

IVF

in vitro fertilization

LB

live birth

PHI

private health insurance

RR

risk ratio

SES

socioeconomic status

SET

single embryo transfer

WARDA

Western Australian Register of Developmental Anomalies

We thank Prof Elizabeth Milne for her helpful comments on our article draft. We also thank the Western Australian Data Linkage Branch and Department of Health data custodians for the linkage of and access to core health data sets, the Western Australian Reproductive Technology Council for permission to use data from the Reproductive Technology Register, and the Register and Fertility Centre staff for their assistance with data queries.