Assisted Reproductive Techniques, ADHD, and School Performance

All residents of Sweden are assigned a unique identification number, which enables tracking over time in national registers.⁶  The study is based on a linkage of multiple registers with the Medical Birth Register (MBR), which was established in 1973 and includes information from antenatal care, maternal characteristics, maternal health before and during pregnancy, and delivery details.⁷  The National Patient Register (NPR) includes diagnoses from all hospital admissions for psychiatric and somatic care since 1973 and 1987, respectively, and from outpatient visits to specialist care since 2001.⁸  The codes in both the MBR and NPR are based on the International Classification of Diseases (Revisions 8, 9, and 10)⁹  and Swedish classification of surgical and medical care measures. The Prescribed Drug Register was established in July 2005 and records all pharmacy-dispensed medications using the Anatomical Therapeutic Chemical Classification System codes.¹⁰  Registers on migration and causes of death provided information about potential censoring before the end of follow-up on December 31, 2018. The study falls under the permission obtained from the regional ethics review board to link these registers (reference number: 2013-1849-31/2).

Since 1982, the standardized interview at enrollment to antenatal care has been used to survey potential trouble conceiving and to ask affected women to report their time to pregnancy in years and, since 1996, also whether it was achieved with ART. Between 1982 and 2006, all IVF clinics in Sweden provided treatment details for all births achieved with ART to the MBR. From 2007 onward, they instead routinely record all treatment cycles into a nationwide quality register for ART (also included in our linkage). We considered children to have been conceived with ART when any of the following criteria were met: (1) matched IVF clinic record (MBR or quality register for ART); (2) clinical diagnostic or procedure codes recorded in the NPR within 1 month of estimated conception date; or (3) maternal self-report. We further identified whether the fertilization was achieved with ICSI or standard IVF and whether transferred embryos were fresh or frozen-thawed. We considered births to have occurred to couples with known infertility when either of the following criteria were met: (1) self-report of infertility, defined as having tried unsuccessfully to achieve pregnancy for ≥1 year,¹¹  (2) diagnosis of infertility any time before the birth for primipara and any time between previous and current birth for multipara, or (3) any use of fertility assistance as defined above.

In Sweden, ADHD is diagnosed and treated in specialized care.¹²  Identification of cases based on records of diagnosis in the NPR and/or dispensation of ADHD medication in the Prescribed Drug Register (details in Supplemental Table 3) has been previously used^3,13  and validated.¹³  We considered the first record, regardless of type, the incident date of diagnosis. School performance was assessed at the completion of 9 years compulsory school (lower secondary), when final subject grades were recorded to the National School Register.¹⁴  Overall performance was assessed through the final average grade (0–20) across 16 subjects and eligibility for upper secondary school (passing grade in at least 12 subjects, including mathematics, English, and Swedish).

To optimize ascertainment of the outcomes of interest, we identified 2 distinct birth cohorts for follow-up (Fig 1). With outpatient specialist care records available from 2001, ADHD was evaluated in children born between 1996 and 2012 (N = 1 655 873) to allow diagnostic coverage from at least age 5. Children were followed from birth until incidence of ADHD, emigration, death, or the end of 2018, whichever came first. Lower secondary school performance was available up to 2017, and, although the vast majority of children in Sweden graduate in the year they turn 16, we assessed children born 1986 to 2001 (N = 1 555 573) to allow an extra year to complete the ninth grade.

We used subject-matter knowledge to identify all known or suspect determinants of ART use that could also influence offspring neurodevelopment and school performance and thereby give rise to noncausal associations (confounding). In Sweden, ART is available through both public and private care, and eligible women aged <40 years can receive up to 3 treatment cycles free of cost. Couples’ infertility is the primary indication, with a small addition of same-sex couples since 2005 and, more recently, single women using donor sperm. Covariates were selected for their ability to block the influence of confounding, and identified factors included parental demographic factors (age, country of origin, cohabitation status, and highest level of education) and mental health history (mood disorders and nonaffective psychosis), and maternal characteristics at the start of pregnancy (parity, tobacco smoking, BMI, and county of residence) and history of health conditions related to infertility (Supplemental Table 3). Because of poorer ascertainment of somatic conditions in the 80s and 90s, we did not consider maternal history of infertility-related conditions in the evaluation of school performance.

For each study cohort, we first compared background characteristics and outcome distributions across 3 mutually exclusive groups consisting of all children born after use of ART (1), all other children born to couples with known infertility (2), and children born to couples with no known infertility (3).

The cumulative incidence of ADHD was obtained with Kaplan-Meier estimation, whereas school performance was summarized through group means of final average grades and proportions eligible for upper secondary school. ADHD was modeled by using Cox proportional hazard regression with attained age as the underlying time scale and by using strata to allow the baseline hazard to vary with year of birth. In all models, we tested the proportional hazards assumption by plotting Schoenfeld residuals and found no evidence that the ART parameter violated the assumption. For school performance, the mean final grade average was modeled in linear regression and the proportion eligible for upper secondary school by using logistic regression. To account for correlated outcomes between siblings and twins, we used robust SEs for Cox regression and generalized estimating equations for linear and logistic regression.¹⁵  All analyses were conducted by using SAS version 9.4 (SAS Institute, Inc, Cary, NC) and Stata 15.1 (Stata Corp, College Station, TX).

With our modeling strategy, we aimed to evaluate the influence of confounding through sequentially increased degree of confounder adjustment. First, we followed the approach of most previous studies, comparing children conceived with ART with all other children (the majority of whom did not have parents with infertility). The baseline model (model 1) was simply stratified on birth year, after which we adjusted for systematic differences in parental demographic factors and maternal prepregnancy status (model 2). We were further able to adjust for parental history of psychiatric conditions, and, in the evaluation of ADHD, also maternal history of health conditions related to infertility (model 3). The analytic approach was then extended to ensure the comparison only concerned those eligible for ART intervention, by restriction to children of couples with known infertility (following a similar modeling strategy: model 1’ to model 3’). In a final step, we further restricted the comparison to only concern children conceived with ART, by directly contrasting different procedures, first ICSI to standard IVF, and then frozen to fresh embryo transfer. Because these additional techniques were introduced in the early 1990s, the school performance cohort was limited to the subsample born 1992–2001 for this evaluation.

Among the characteristics a priori identified as relevant confounders, the most substantial missingness occurred for early-pregnancy BMI (Table 1). Opting to not include BMI as a covariate, we conducted our main analyses in complete case samples that represented 89% to 90% of the targeted cohorts (Fig 1).

In sensitivity analysis, we performed multiple imputation on the covariates with >5% missing (cohabitation status, maternal smoking, and BMI). We then replicated all main analyses in the imputed data and took the opportunity to assess models with and without the inclusion of BMI. We also repeated all analyses using more specific definitions of ADHD by requiring (1) 2 diagnostic records separated by at least 1 month; and (2) at least 1 diagnostic or dispensation record after the age of 5.

For each of the 2 study cohorts, Table 1 reveals parental and pregnancy characteristics in relation to infertility status and potential use of ART. Both infertility and use of ART increased across the study period; in 1986–2001, 7% of births were to couples with known infertility and 13% of these were achieved with ART, compared with 11% and 26%, respectively, in 1996–2012. In both cohorts, couples with infertility were more likely older and married or cohabitating, compared with couples with no known infertility. Among infertile couples, those that conceived with ART had, on average, higher age and education, and the women were less likely to smoke. As expected, ART use was associated with much higher occurrence of multiple gestations.

The estimated cumulative incidence of ADHD by age 15 was 6.2% (95% confidence interval [CI]: 5.9% to 6.5%) in children conceived with ART, 7.3% (95% CI: 7.2% to 7.5%) in children of couples with infertility that did not use ART, and 7.1% (95% CI: 7.0% to 7.1%) in children born to couples with no known infertility (Fig 2). Comparing children conceived with ART with all other children (irrespective of parental infertility), the former were at lower risk of ADHD (hazard ratio [HR] 0.83; 95% CI: 0.80 to 0.87; model 1, Fig 3, left panel). Adjustment for differences in parental characteristics turned the association to a slightly elevated risk of ADHD with ART (adjusted hazard ratio [aHR] 1.07; 95% CI: 1.02 to 1.12; model 2), which was somewhat attenuated by further adjustment for parental health history (aHR 1.05; 95% CI: 1.00 to 1.09; model 3). Furthermore, when the comparison was restricted to children born to couples with infertility, ART use was associated with a lower risk of ADHD (aHR 0.80; 95% CI: 0.77 to 0.84; model 1, Fig 3, right panel), but adjustments for parental characteristics and health history attenuated the association toward the null (aHR 0.95; 95% CI: 0.90 to 1.00; model 3). Finally, the direct contrast of ART procedures revealed no difference between use of ICSI and standard IVF and a tendency toward lower risk with frozen compared with fresh embryo transfer (not statistically significant; Table 2).

Findings for school performance followed similar patterns (Fig 3). Children conceived with ART initially appeared to have a slight advantage with respect to final grade average and eligibility for upper secondary school. Again, adjustment for parental characteristics reversed the associations into a slight disadvantage compared with all other children, whereas the adjusted comparison with children born to couples with infertility revealed no differences. When contrasting specific ART procedures in the subsample born 1994–2001, the children conceived with ICSI appeared at a slight advantage compared with standard IVF, whereas no difference in school performance was noted between frozen and fresh embryo transfer (Table 2).

Sensitivity analysis to explore the robustness of findings with respect to ADHD definition, model specification, and missing information in some covariates produced minor fluctuations in point estimates of no consequence for the overall interpretation (Supplemental Tables 4 and 5).

In this large follow-up of nationwide birth cohorts, we observed lower risk of ADHD and slightly better overall school performance in children conceived with ART compared with all other children. Differences in parental characteristics appeared to completely explain and even slightly reverse the associations. When the comparison was restricted to children of couples with known infertility, no differences were seen. Additionally, procedures like ICSI or frozen embryo transfer had no adverse influence on children’s risk of ADHD or school performance compared with standard IVF and fresh embryo transfer, respectively.

Our findings confirm the mostly reassuring reports on neurodevelopment in children conceived with ART to date. For ADHD specifically, previous findings have been somewhat conflicting. A few early reports of normal behavior and attention in a small Dutch evaluation of children born after use of ART^16,17  was followed by an alarmingly high prevalence of self-reported ADHD in a small cross-sectional evaluation of young adults conceived with IVF in the United States.¹⁸  Subsequent large Scandinavian register-based studies instead revealed that, in comparison with the general population, children born after ART use were at lower risk of ADHD, although this was explained by differences in maternal characteristics.^19,20  Because the adjustment for such factors even led to a slight reversal of the association in the Swedish study, it was taken to indicate a weak positive association between IVF and (drug-treated) ADHD.¹⁹  Findings from previous studies of school performance follow a similar pattern. When the previously mentioned Dutch study also evaluated several measures of school performance between ages 8 and 18, no differences were seen in relation to children born to couples with infertility that did not use ART.²¹  In contrast, a US study of similar size and age groups revealed overall higher test scores when children conceived with IVF were compared with a random sample of spontaneously conceived children.²²  More recently, authors of 2 register-based studies in Denmark²³  and Sweden²⁴  made a similar observation with respect to school performance in grade 9, finding a small advantage for singletons conceived with ART compared with the general population. Notably, after also seeing associations reversed once some parental characteristics were taken into account, the researchers in both studies found children conceived with ART at a slight disadvantage overall.^23,24 

When applying a similar comparison and adjustment approach, we reproduced the pattern of findings in previous register-based studies. Compared with the general population, children conceived by ART were at an advantage with respect to both ADHD and school performance, but this was reversed to a slight disadvantage once differences in parental characteristics were taken into account. In contrast with previous large register studies we proceeded to restrict the comparison to children of couples with infertility and, once the parental characteristics that still differed between the groups were accounted for, we saw no association between ART and ADHD or school performance. Previous research has revealed children born to couples with infertility at higher risk of ADHD (HR 1.33; 95% CI: 1.26 to 1.42),⁵  and, although a direct causal effect cannot be excluded, this seems most likely due to underlying parental risk factors. As such, infertility acts as an important confounder in the evaluation of ART safety (blocking the influence of common causes of ART use and adverse outcomes in the offspring). To compare the outcome of children whose parents had trouble conceiving further provides a more meaningful clinical contrast by ensuring everyone had the primary indication for ART.

Finally, the direct contrast of ART procedures balances the comparison groups even further by only considering those that underwent treatment. Reassuringly, we did not find any support that the increasingly popular techniques of ICSI or freezing of embryos would put children at any disadvantage with respect to risk of ADHD or school performance compared with standard IVF or fresh embryo transfer.

In addition to thorough consideration of relevant cofounders and emphasis on analytic strategies, our study has several strengths from relying on nationwide prospectively gathered data in registers with high validity. Because the study was based on national birth cohorts, no sampling was made. We were able to combine several sources of ART information, with maternal self-reports complementing IVF clinic reports and outpatient records (which would not include treatments received abroad). The identification of ADHD from diagnostic and dispensation records in Swedish registers has been previously validated,¹³  and our findings were additionally robust to alternative, stricter outcome definitions.

Our study also had some potential limitations. Although the degree of missing information was minor, it concerned several of the relevant covariates in both cohorts. Under the assumption of missing at random, we opted to perform complete case analysis, but we also performed multiple imputation as a sensitivity analysis, providing consistent results. The current study findings reflect the situation in Sweden across a specific time and may not generalize to other populations or time periods. Calendar year plays an essential role in both the implementation and evolution of ART. Aside from the increasing reliance on ICSI (also without male factor infertility), this includes the single-embryo transfer recommendation gradually implemented in Sweden 2001–2003. The consequential drop in the rate of multiple gestation after ART has led to a reduction in the risk of preterm birth and low birth weight.^25,26  Because of these common complications, several previous studies evaluate ART in singletons only,^16,17,21,24  or twins and singletons separately.²³  Considering multiple gestation part of the consequence of the intervention (and thus in the causal pathway), we opted to evaluate all children conceived with ART together. Our overall reassuring findings and their similarity with those of previous large studies in singletons indicate that twinning per se has no adverse influence on the overall neurodevelopment of children conceived with ART.

In this study, we found no indication for concern about children’s risk of ADHD or school performance after use of ART. Our findings thus provide additional reassurance concerning offspring neurodevelopment after use of ART while also stressing the importance of considering the role of underlying infertility in studies of ART safety.

ADHD

attention-deficit/hyperactivity disorder

aHR

adjusted hazard ratio

ART

assisted reproductive technique

CI

confidence interval

HR

hazard ratio

ICSI

intracytoplasmic sperm injection

IVF

in vitro fertilization

MBR

Medical Birth Register

NPR

National Patient Register