|Home | About | Journals | Submit | Contact Us | Français|
Does IVF increase the risk of autism spectrum disorders (ASDs)?
No association between IVF and ASDs or any of its subtypes was found in this sample.
Certain prenatal factors may increase the risk of ASDs. Studies on the association between IVF and ASDs have shown inconsistent results. IVF is known to increase the risk of perinatal problems but many of them are related to multiple pregnancies.
This case–control study included 4164 autistic cases and 16 582 matched controls born in Finland in 1991–2005. The cases were diagnosed with ASDs by the year 2007. The maximum age at diagnosis was 16 years.
Four controls were matched to each case. For singletons the matching criteria were date of birth, place of birth, sex and residency in Finland. For twins the birth order within a twin pair was included as well. In the whole sample, there were 63 cases (1.51%) and 229 controls (1.38%) born after IVF.
No significant association was found between IVF and ASDs (adjusted odds ratio (OR): 0.9, 95% confidence interval (CI): 0.7–1.3) or its subtypes childhood autism (OR: 0.8, 95% CI: 0.4–1.5), Asperger's syndrome (OR: 0.9, 95% CI: 0.5–1.6) or other pervasive developmental disorder (OR: 1.0, 95% CI: 0.6–1.6). When only singletons were included, there was an association between IVF and Asperger's syndrome in an unadjusted analysis (OR: 2.0, 95% CI: 1.1–3.5) but this was not significant when adjusted for mother's socioeconomic status or parity. When the analyses were conducted separately for boys and girls, there was a significant association between IVF and Asperger's syndrome for boys in an unadjusted analysis (OR: 2.1, 95% CI: 1.2–3.7) but this was not significant in the final adjusted model.
Information both on IVF and on ASDs was based on registers and it is possible that there is some misclassification. No information on different subtypes of IVF or other assisted reproduction techniques was available. Statistical power may have been insufficient.
This study showed no increased risk of ASDs in children born after IVF but studies with larger sample sizes and information on different subtypes of IVF are needed to confirm the finding.
The study was supported by Autism Speaks, NIMH 1K02-MH65422 and NIEHS 1R01ES019004. There are no competing interests.
Assisted reproduction techniques (ARTs) are under continuous development and the number of babies born after IVF is increasing. In Finland, >3% of babies are currently born following IVF and its subtypes, such as ICSI and frozen embryo transfers (FETs) after IVF or ICSI (Gissler and Heino, 2010).
The course of IVF pregnancies and the health of children born after IVF are of major public health interest. IVF pregnancies are associated with an increased risk of perinatal problems, such as preterm delivery, perinatal mortality, low birthweight, being small for gestational age and admission to neonatal intensive care (Helmerhorst et al., 2004; Jackson et al., 2004), as well as congenital anomalies (Rimm et al., 2004; Hansen et al., 2005; Wilkins-Haug, 2008). Many of these problems are probably associated with multiple pregnancies (Sutcliffe and Ludwig, 2007; Basatemur and Sutcliffe, 2008) but even singletons have been reported to have an increased risk of certain health complications. It is not clear, however, whether the increased risks are better explained by treatment per se or rather by parental factors leading to infertility (Ludwig and Diedrich, 2002; Reddy et al., 2007; Sutcliffe and Ludwig, 2007).
Because IVF children are often followed only during the perinatal or infant periods, there is less information about their well-being in later life (Ludwig et al., 2006; Middelburg et al., 2008; Wagenaar et al., 2008). In several studies conducted to date, early neurodevelopmental health (Sutcliffe and Ludwig, 2007; Wagenaar et al., 2008), later cognitive function and school performance (Middelburg et al., 2008; Wagenaar et al., 2008), physical development (Ludwig et al., 2006; Basatemur and Sutcliffe, 2008) and psychosocial well-being (Ludwig et al., 2006; Basatemur and Sutcliffe, 2008; Wagenaar et al., 2008) were similar to spontaneously conceived children. Some studies with sufficient follow-up time, however, have shown an increased risk of cerebral palsy and other neurological problems (Ludwig et al., 2006; Hvidtjorn et al., 2009) but others suggest that the effects of IVF are mediated by risk factors such as preterm birth (Basatemur and Sutcliffe, 2008; Middelburg et al., 2008).
Several perinatal factors have been suggested to increase the risk of autism spectrum disorders (ASDs), including maternal prenatal medication use, prematurity, growth restriction, intrapartum hypoxia and bleeding (Hultman et al., 2002; Glasson et al., 2004; Kolevzon et al., 2007; Gardener et al., 2009; Haglund and Källén, 2011; Lampi et al., 2012). Relatively few studies on the association between IVF and ASDs have been conducted, with contradictory results that may be secondary, in part to methodological limitations (Hvidtjorn et al., 2009). Most are cohort studies based on Scandinavian registers. The largest of these, a Swedish study, had a sample size close to 1.5 million (Ericson et al., 2002). No significant association was found between IVF and ASDs but the outcome was grouped together with other developmental disturbances. Two large Danish register-based cohort studies (Lidegaard et al., 2005; Hvidtjorn et al., 2011) did not show a significant association between IVF and ASDs with the exception that in one of these studies (Hvidtjorn et al., 2011) a relationship was observed for the subgroup in which the women had been exposed to FSH and received IVF treatment. These two studies did not report different ASD types separately. In addition, these studies had relatively short follow-up times. In one of them (Lidegaard et al., 2005), the follow-up time was a little >4 years on average, which was probably too short to identify all cases, as disorders such as Asperger's syndrome are often diagnosed later. Studies conducted in the UK, in which children with Asperger's syndrome have been diagnosed in the 1990s and 2000s have suggested that the median age at diagnosis is almost 10 years (Williams et al., 2008) or the mean age around 7 years depending on the year of diagnosis (Latif and Williams, 2007). The only cohort study in which a significant positive association between IVF and ASDs has been found [odds ratio (OR): 1.68, 95% confidence interval (CI): 1.11–2.53] was based on a Finnish cohort (Klemetti et al., 2006) but the definition of the outcome was even less precise than in the previously mentioned studies. In addition to ASDs, practically all childhood developmental and psychiatric disorders were included (Klemetti et al., 2006). In a rather small Danish case–control study (Maimburg and Vaeth, 2007) the outcome was restricted to childhood autism but the exposure included hormone therapy both with and without IVF. In the Maimburg and Vaeth (2007) study, a significantly lower risk (OR: 0.37, 95% CI: 0.14–0.98) of childhood autism was found in children born following ART compared with children who were conceived naturally.
Our aim was to examine the association between IVF and risk of different subtypes of ASDs. In this study we were able to overcome some of the main limitations of previous studies. First, we employed a large sample size. Secondly, we made use of comprehensive register data which provided a representative sample of cases and controls, reducing the potential for selection bias. Thirdly, we analyzed the findings separately by ASD subtype.
The study is part of the Finnish Prenatal Study of Autism, which is a nested case–control study based on a national birth cohort, and aims to identify pregnancy, infancy and childhoood risk factors for ASDs. An overview of the study has been presented previously (Lampi et al., 2011). The study was authorized by the Ministry of Social Affairs and Health of Finland with approvals from the National Institute for Health and Welfare (THL), the Ethics Committee of the Intermunicipal Hospital District of Southwest Finland, and the Institutional Review Board of the New York State Psychiatric Institute. No informed consent was required for this study. To assess the association between IVF and ASDs, we conducted a linkage between two national registers for 4164 autistic cases and their 16 582 matched controls born in 1991–2005. There were 111 cases and 436 controls, which were twins. There were originally five cases (0.1% of all cases) born from triplet pregnancies but they were excluded from this study because it would have been difficult to find matched controls. Of all the children born in Finland during the same period 0.09% were born from pregnancies with more than two fetuses (National Institute for Health and Welfare, 2012).
The data on IVF were collected from the Finnish Medical Birth Register (FMBR), which is a national register maintained by THL. Information was collected on maternal background, pregnancy, delivery and early outcomes of the newborn for all births in Finland. The register includes all mother's and child's personal identification numbers, which can be linked to one another. The FMBR commenced in 1987, and data on fertilization treatment has been collected since October 1990. In 1990–1995, IVF was denoted by the item ‘assisted reproduction,’ including IVF with or without ICSI or FET. In 1996, two items, ‘in vitro fertilization’ and ‘other assisted reproduction’ were introduced. A validation study (Gissler et al., 2004), however, showed that the latter category also included mostly IVF children, and that the two categories could not be separated. Hence, in the current study the two items were combined. In 2004, the terminology was revised again, and the treatments, which include IVF, could be clearly differentiated from the others. By merging the items from all three time periods, we obtained one variable referring to IVF treatments. Different subtypes of IVF, however, could not be specified.
Children born in 1991–2005 and diagnosed with ASDs by the year 2007 were identified from the Finnish Hospital Discharge Register (FHDR), another nationwide register maintained by THL. The FHDR includes the personal identification numbers and covers the diagnoses and the days of admission and discharge in all public and private inpatient care units in Finland. The diagnoses are based on the International Classification of Diseases and Related Health Problems (ICD). All diagnostic codes indicating an ASD were collected according to the ICD-9 in 1991–1995 and ICD-10 in 1996–2007. The most recent registry diagnosis was used. The diagnostic categories included in this study were, first, the whole group of ASDs (F84) and, secondly, three of its subcategories: childhood autism (F84.0), Asperger's syndrome (F84.5) and other pervasive developmental disorder (PDD) and PDD, not otherwise specified (PDD NOS) (F84.8 and F84.9). A validation study (Lampi et al., 2010) has shown that the validity of childhood autism diagnosis in the FHDR is very good: when a re-assessment using Autism Diagnostic Interview-Revised was made, 96% of the cases with registry diagnoses of childhood autism met the diagnostic criteria.
Four controls selected from the FMBR were matched to each case. For singletons the matching criteria were date of birth (± 30 days), place of birth, sex and residency in Finland. For twins the matching criteria were date of birth (±6 months but for eight controls the range ±11 months had to be used), region of birth, sex and birth order within a twin pair. No data for whether twins were mono- or dizygotic were available for analysis because this information is not included in the FMBR. The exclusion criteria for controls were ASDs or severe/profound mental retardation according to the FHDR, and eight controls were excluded for this reason. In addition, for 66 controls the personal identification number of the mother or the child was incorrect or incomplete. Altogether 74 controls were removed from the database, which led to 16 582 controls.
An additional analysis was conducted in which stratification by intellectual disability was performed. For this purpose, information on diagnoses indicating intellectual disability (F70-79 in ICD-10 and 317–319 in ICD-9) was collected from the FMBR.
Prior to adjustment, bivariate analyses had been conducted to test the significance of association between covariates and IVF among controls as well as between covariates and ASDs (Table I). Maternal age, mother's socioeconomic status (SES), gestational age and parity were all significantly associated with both the exposure and outcome. The data on these covariates were collected from the FMBR. All variables were defined categorically. When studying maternal age at birth, age 20–34 years was used as a reference. For mother's SES, upper white collar workers were used as a reference and the comparison was made with lower white collar workers, blue collar workers and others, which includes entrepreneurs and people outside the labor force, such as students, housewives and unemployed people. The measure of SES follows the national classifications on occupations and socioeconomic groups (Statistics Finland, 1987; Statistics Finland, 1989), and it was primarily based on occupational status, even though educational level was also considered for white collar workers. For gestational age, the category 38–41 weeks was used as a reference. For parity, the reference category was no previous births.
Associations between IVF and different ASDs were quantified by calculating ORs with 95% CIs using conditional logistic regression analysis. In addition to crude ORs, we calculated ORs adjusted first for maternal age at birth, mother's SES, gestational age and parity separately, and then for all variables simultaneously. The analyses were conducted separately for singletons and for the whole sample, which included both singleton and twin births. Because of the secular changes in the frequency of IVF treatment, the prevalence of ASDs and the developments in IVF techniques, we assessed whether the association between IVF and ASDs was modified by birth year. Data about the children born in 1991–1995, 1996–2000 and 2001–2005 were analyzed separately and the results of the three time periods were compared with each other. Data were also analyzed separately for children with and without intellectual disability, and separately for boys and girls. The sample size of this study was sufficient to detect a 1.5- to 2.1-fold difference in IVF between cases and controls for total ASDs, childhood autism, Asperger's syndrome and PDD when using 80% power. A two-sided test with type I error rate (alpha) of 0.05 and matched case control study design were applied for this calculation (Dupont, 1988). Statistical analysis was performed using SAS statistical software (SAS Institute Inc. SAS Version 9.3. Cary, NC, USA). The power calculations were performed with the R (2.15.1) packages epicalc (http://cran.r-project.org/web/packages/epicalc/epicalc.pdf) and epiR (http://cran.r-project.org/web/packages/epiR/epiR.pdf).
The numbers of cases and controls in the diagnostic categories of childhood autism, Asperger's syndrome and PDD/PDD NOS are shown in Table II. The total number of cases and controls in these three categories combined is less than in the whole group of ASDs (diagnostic code F84) because rare diagnoses, such as Rett's syndrome, were not studied separately.
Unadjusted and adjusted ORs for the relationship between IVF and each of the ASDs categories are shown in Table III. In the whole sample, 1.51% of children had been born after IVF. When the ASD categories were combined, no significant association with IVF was found in unadjusted or adjusted analyses. In the unadjusted analyses for singletons, among the three diagnostic subtypes, a significantly increased risk for only Asperger's syndrome was found among IVF children (OR: 2.0, 95% CI: 1.1–3.5). In the whole sample including twins there was no significant association between IVF and any of the ASD subtypes. When the analyses were conducted separately for boys and girls, no new significant associations were found except for Asperger's syndrome. In the whole sample, the association between IVF and Asperger's syndrome, which was close to significant when both genders were combined, became significant when only boys were included (OR: 2.1, 95% CI: 1.2–3.7). In addition, among the singletons the association between IVF and Asperger's syndrome was even stronger when only boys were included (OR: 2.4, 95% CI: 1.3–4.5). No significant associations were found for girls but the analyses were limited by the small number. Even among singletons, the association between IVF and Asperger's syndrome became non-significant when controlled for mother's SES (OR: 1.7, 95% CI: 0.9–3.1) or parity (OR: 1.6, 95% CI: 0.9–2.9). When adjusted for all covariates, the OR was further decreased to 1.1 (95% CI: 0.6–2.1). This association was 1.4 (95% CI: 0.7–2.9) for boys but could not be calculated for girls, because only two girls with Asperger's syndrome were born after IVF and an adjusted model did not converge owing to missing covariates. The data for the whole ASD group were also analyzed in three equal time periods of births but no statistically significant differences in the proportion of IVF children among the cases and controls were found during any of these periods (1991–1995: OR: 0.8, 95% CI: 0.5–1.5; 1996–2000: OR: 1.4, 95% CI: 0.95–2.1; 2001–2005: OR: 0.6, 95% CI: 0.2–1.7). No differences in the association between IVF and ASDs were found when the analysis was conducted separately for those with and without intellectual disability: the OR remained at OR: 1.1 (95% CI: 0.8–1.5) for both groups.
This study showed no increased risk for ASDs, analyzed as a single category and by individual subtypes, in children born after IVF. This is in line with most previous studies. The risk for Asperger's syndrome was increased in the unadjusted analyses of singleton births but was attenuated and not statistically significant following adjustment for mother's SES or parity. A possible explanation for the confounding by SES is that women with high SES obtain IVF treatment more often, and their children may be more likely to be diagnosed if they have Asperger's syndrome. Differences in access to ART by region and SES have been reported for Finland (Klemetti et al., 2007). These differences, however, are relatively small, because IVF treatments are also given in the public sector and social sickness insurance covers a part of the treatment costs for most women using private services (Klemetti et al., 2007). It has been shown in the UK that ASD cases are more likely to become identified in clinical practice if parents have a long education (Baird et al., 2006), and this may apply particularly to Asperger's syndrome because the impairment is usually less severe and thus non-identification may be more common compared with other types of ASDs. No increased risk with IVF was shown for childhood autism or PDD/PDD NOS.
The study had several strengths. This is the first study of IVF and ASDs in which subcategories of ASDs were analyzed separately and the results compared with one other. A second strength is that the number of ASD cases was greater than any of the previous studies on IVF and ASDs. Thirdly, the universal public health coverage and the use of a national register including information on both inpatients and outpatients improved case ascertainment and generalizability of the findings. The last of these methodological advantages, together with the selection of controls from a national register with complete coverage of the population, reduced the potential for selection bias. Furthermore, the prospective assessment of exposure minimized the possibility of recall bias.
However, some limitations concerning both exposure and outcome data should be considered. First, owing to the combination of the two items that refer to ART, the number of IVF children may have been overestimated in the years 1996–2003. However, Gissler et al. (2004) estimated that the proportion of children in the ‘other assisted reproduction’ group who were incorrectly classified as IVF children is no higher than 11%. It has also been reported that some IVF information may be missing in the FMBR, especially in the early 1990s (Gissler et al., 2004). The possible bias is still likely to be small because these two types of misclassification are expected to be non-differential between cases and controls. A second, related limitation is that no information on specific types of IVF treatment or on other forms of ART was available. Thirdly, misclassification of diagnoses of Asperger's syndrome and PDD/PDD NOS may have occurred because the register diagnoses of these disorders were not directly verified by a structured research interview, unlike childhood autism. Finally, the negative findings may suggest that there is no association between IVF and ASDs but these may also be explained by the lack of statistical power. The association which was closest to significance was that between IVF and Asperger's syndrome. We had 6733 cases and controls in that group but we would have needed 24 000 to be able to detect an OR of 1.5. There might also be some factors, which were not included in this study, which masked a true positive association. For example, maternal ASDs may decrease the likelihood of getting IVF treatment and thus the proportion of children born to mothers with autistic features may be higher among children who were conceived naturally. Insufficient statistical power may complicate the interpretation of previous studies as well. The studies on IVF and ASDs (Lidegaard et al., 2005; Hvidtjorn et al., 2011) and those in which ASDs together with other developmental problems has been included (Ericson et al., 2002; Klemetti et al., 2006) have shown moderate risk estimates, which have been mostly non-significant (with the exception of the study by Klemetti et al.) but in the direction of a positive association. A similar pattern was shown in our unadjusted analysis but it did not remain when adjusted for all covariates. A previous study on ART and childhood autism showed a non-significant association in the direct of a negative association (Maimburg and Vaeth, 2007). This is in line with our results.
In conclusion, the findings suggest that IVF is not likely to increase the risk of ASDs. However, studies with larger sample sizes are needed to confirm the finding. In addition, it will be important to study not only different types of ASDs but also different types of fertilization treatments separately.
V.L., A.S.B., M.G. and A.S. participated in designing the study. M.G., M.R. and A.S. contributed to the data collection. All authors participated in designing the statistical analyses and M.R. and A.S. conducted them. V.L. drafted the first version of the manuscript and all other authors contributed critically to its revision. The final version was approved by all authors.
The study was supported by Autism Speaks, NIMH 1K02-MH65422 and NIEHS 1R01ES019004. The funding sources have no role in designing or conducting the study, in collecting, managing, analyzing or interpreting the data or in preparing, reviewing or approving the manuscript.