Child Health: Is It Really Assisted Reproductive Technology that We Need to Be Concerned About?

The health of children conceived with assisted reproductive technologies (ART) has always been of concern to couples considering infertility treatments and for health practitioners. Worldwide, the number of births conceived by ART continues to climb. 1 – 3 As of 2015, ART contributed to 1.7% of all births in the United States (US). 2 Almost all ART procedures performed in the US are in vitro fertilization (IVF), of which 69% included intracytoplasmic sperm injection (ICSI). 2 Though other types of fertility treatment, including ovulation induction (OI) and intrauterine insemination (IUI), are commonly used as the first line of treatment, less data has been gathered to track their trends over time. Estimates of births conceived with non-ART treatments range from 3–7%. 4 There have been numerous reviews published on the topic of child health after conception by infertility treatment, including one published last year in the current journal 5 and our own focused on cardio-metabolic health. 6 Hence, the current review is not intended to exhaustively review the existing literature in this area. However, in brief, evidence suggests that babies conceived by ART are born earlier and smaller than their peers, which lead to concerns for neurodevelopmental and other long-term health outcomes given the evidence from developmental origins of health and disease research. Nevertheless, in general we 7 , 8 and others have found no major differences in development or growth in early childhood after accounting for plurality. Longer term studies are needed, as well as studies with a substantial group of children conceived by non-ART treatments. Less investigated are early signs of developmental differences such as fetal growth and placental development. 9 Specific techniques and treatment effects such as embryo culture media and oocyte vitrification also continue to be evaluated. The goal of the present review is to discuss the epidemiologic hurdles to studying the health impact of conception by infertility treatment, including how perhaps the reasons for infertility may drive independent health effects long-term. Under the developmental origins of health and disease paradigm, specific conditions, such as maternal obesity, may lead to long-term offspring health risks ranging from obesity to neurodevelopmental delay. These conditions are also related to the risk of infertility. As such, we will also review common risk factors for male and female infertility and discuss potential impacts on developmental programming in order to address how best to investigate whether ART itself possesses risks to child health.

Apart from population and technological considerations, underlying infertility has been a source of confounding giving rise to questions regarding whether associations observed with child health are truly due to treatment effects. Parental health has an impact on child outcomes beginning at delivery. 43 Despite a wealth of research into developmental programming from maternal pregnancy conditions, fewer studies have focused on pre-conceptional health and how infertility risk factors of both partners may play a role. Part of the difficulties lie in measuring participants prior to pregnancy as well as accuracy of participant reported diagnoses of some of the infertility disorders. Below summarizes the evidence for factors which children’s health has been investigated, including parental age, parental obesity, dietary intake, thyroid function, depression, and other female gynecologic conditions. Advanced maternal age (≥35 years) increases risks of pregnancy and neonatal complications. 44 – 46 It can also increase risks of chromosomal abnormalities. Based on increasing rates of preterm delivery and other adverse neonatal outcomes, one anticipates higher risks of adverse long-term outcomes. However, risks of long-term impact on childhood growth and development from advanced maternal age may be modified by postnatal circumstances. Women of older age at childbirth tend to have better socioeconomic status and it is difficult to tease apart those benefits conferred. Indeed, in a study of children from five birth cohorts in low to middle income countries, stunting at 2 years of age (OR 064; 95% CI 0·54–0·77) and failure to complete secondary schooling (OR 0·59; 95% CI 0·48–0·71) were both protected by advanced maternal age as compared to children of mothers of 20–25 years of age. 47 A retrospective cohort study tried to tackle the question of longer term outcomes by using hospital data linking 200,000 deliveries in Israel and their outcomes through 18 years of age. 48 They did not identify clear differences in risk of childhood morbidities (i.e., central nervous system/brain, lymphoma, leukemia) related to maternal age at 35–39 (hazard ratio [HR] 1.06; 95% CI 0.76–1.48) or even at 40–50 years (HR 0.73; 95% CI: 0.36–1.46) when compared to 20–34 years. 48 Yet, other studies have observed high risks of childhood cancers including retinoblastoma and leukemia with increasing maternal age. 49 Further complicating the impact of maternal age in ART is the use of donor or autologous oocytes and the unclear comparison group in those cases. Live-birth rates decline with increasing maternal age in ART but this is particularly true for autologous cycles rather than for donor cycles. 50 Hence, the maternal age effects are suspected to be due to decreasing oocyte quality over time, but aside from short-term outcomes, the impact of maternal age over the life-course may be counterbalanced by positive socioeconomic influences with older age, particularly in terms of childhood growth and development although perhaps less so for risks of cancer. Paternal age may also play a role that sometimes is difficult to isolate from the effects of maternal age as couples tend to have correlated age. For example, the Childhood Leukemia International Consortium identified increased risks of acute lymphoblastic leukemia associated with advanced paternal age with mixed results for advanced maternal age, owing to the collinearity of the age information. 51 Accumulating evidence suggests, however, that paternal age and indeed paternal risk factors in general, may be important to assess. A very comprehensive systematic review of this evidence was published on paternal factors. 52 The authors located the most articles for investigations into paternal age, with much fewer for other aspects (i.e., obesity, smoking) due to fewer studies with paternal information. For short-term outcomes, they found “little or no difference” in rates of preterm, low birth weight or small-for-gestational age due to advanced paternal age. A very slight increase in stillbirths (with low evidence) and in birth defects (with moderate evidence) was also observed. For longer term outcomes, paternal age has been related to risks for autism 53 but evidence of other psychiatric disorders is lacking. Little evidence suggested increased risks for childhood cancer, obesity, or diabetes. 52 About a third of all US adults are obese (BMI≥30), making it a critical concern. Male and female obesity increases risk of infertility. For females, obesity is associated with anovulation, miscarriage and longer time to pregnancy. The impact of adiposity is primarily hypothesized to be through hormonal disruption as adipose tissue is endocrinologically active. For males, increases in DNA fragmentation and epigenetic effects have been observed and are beginning to be further elucidated. Few studies have investigated childhood outcomes related to paternal BMI. 52 In the short-term, paternal height rather than BMI is associated with birthweight. In terms of long-term effects, the Upstate KIDS Study observed that paternal obesity (BMI≥30) was associated with increased odds of failing the personal-social domain of the Ages and Stages Questionnaire (ASQ), rated by parents at 4, 8, 12, 18, 24, 30, and 36 months of age. 54 Although this association was not strong, it was independent of maternal BMI. In fact, maternal obesity was associated with failing the fine motor domain. Severe parental obesity (BMI≥35 in both parents) was additionally associated with failing the problem-solving domain. Findings from Suren and colleagues using data from MoBa, a large birth cohort from Norway, has also demonstrated a stronger, linear association between paternal BMI rather than maternal BMI with risk of developing autism. 55 As the role of epigenetics have come to be demonstrated, however, empirical evidence suggests that the paternal contribution goes beyond traditional heridity. 56 , 57 In 2015, scientists from INSERM demonstrated in a clever experiment that sperm/testis RNA from male mice fed a western diet injected into once-cell embryos of control parents fed standard diets, lead to progeny with higher weight and altered glucose tolerance. 58 They further identified two microRNAs (i.e., miR-19b and miR-29a) to be particularly dysregulated in male mice fed western diets. Specifically, miR-19b micro-injection increased offspring body weight. Apart from miRNA, other small piwi RNA (piRNA) or transfer RNA (tRNA) are also suspected to play a role. 56 Many more studies have investigated the role of maternal obesity on short- and long-term outcomes in child health. Pre-pregnancy BMI is also associated with pregnancy complications such as gestational diabetes and preeclampsia which in turn lead to neonatal complications such as preterm delivery and suboptimal fetal growth (i.e., macrosomia or growth restriction). Apart from the observed associations in Upstate KIDS mentioned above, other studies have also identified maternal obesity as a risk factor for a range of neurodevelopmental outcomes. 59 Moreover, obesity is tied to energy balance, dietary intake, and micronutrient metabolism. For instance, obese individuals are more likely to have low circulating concentrations of 25-hydroxy vitamin D [25(OH)D] since the lipid soluble vitamin is sequestered by adipose tissue. Vitamin D’s potential link with fertility is suggested by evidence that its receptor is prevalent on both male and female reproductive tissues in rodent models. 60 In one study, female rats deficient in vitamin D experienced a 75% reduction in fertility. 61 Investigations in humans have largely focused on pregnancy outcomes of women undergoing ART. As recently reviewed by Pacis et. al., vitamin D deficient women [25(OH)D<20ng/mL] have a lower percent implantation and clinical pregnancy per embryo transfer compared with vitamin D sufficient [25(OH)D≥30ng/mL] women. 62 Maternal vitamin D levels, in turn, have been investigated with regards to child health. Gestational and cord blood 25(OH)D concentrations have been associated with neurodevelopmental outcomes, such as mental or psychomotor development, among infants and toddlers in most, 63 – 68 but not all, 69 studies. In early childhood, low 25(OH)D concentrations from gestational or newborn samples have been related to externalizing symptoms, 70 attention deficit hyperactivity disorder (ADHD)-like symptoms 70 – 72 and autism diagnosis. 73 – 75 Interestingly, no association between gestational or cord blood 25(OH)D and lifetime ADHD diagnosis has been detected. 76 , 77 In middle childhood and early adolescence, gestational 25(OH)D has been inversely associated with language ability 78 and verbal fluency. 79 However, it is unclear if the associations detected between 25(OH)D and markers of infant development translate into other cognitive or behavioral problems later in life since, with one exception, 69 , no relation between low 25(OH)D concentrations and measures of cognitive ability (i.e., IQ or scholastic achievement) 68 , 70 , 77 , 80 or behavior problems 68 , 69 , 77 , 78 , 80 , 81 in middle childhood or adolescence has been found. Nutritional studies remain difficult to conduct. Unlike vitamin D, many other factors such as protein intake or even caloric consumption, are not well captured by biomarkers. Self-report is also limited. Future epidemiologic studies of maternal nutrition may leverage on new technologies such as metabolomics. PCOS is a complex endocrine disorder affecting 5–16% of women. 82 – 84 It is the most common identifiable cause of infertility, and is characterized by menstrual irregularities, hyperandrogenism, polycystic ovaries, and frequently, obesity and insulin resistance. 82 PCOS is a strongly heritable disorder, with studies suggesting that over 70% of PCOS is linked to genetic causes. 85 , 86 Mothers with PCOS are at higher risk of numerous pregnancy complications, including gestational diabetes, pregnancy-induced hypertension, preeclampsia, pre-term birth, and infant mortality, which may all influence offspring health. 87 , 88 Women with PCOS may be hyperandrogenic during pregnancy, and endocrine function during pregnancy may impact fetal development. While these changes do not appear to impact infant growth patterns, accumulating evidence suggests that in utero exposure to a hyperandrogenic environment is associated with neurodevelopmental disorders in offspring. 89 – 94 In the Upstate KIDS Study, maternal PCOS was associated with higher risks of developmental delays in several domains, particularly among girls. 89 This finding has been supported by several large studies in Europe using administrative data, which have found associations between maternal PCOS and/or elevated fetal testosterone and higher risks of autism, ADHD and pervasive developmental disabilities. 90 – 92 , 95 – 99 As with studies of potential health effects of IVF on offspring, studies of PCOS face several methodological challenges. PCOS is a heterogeneous condition with diagnostic criteria that have changed over time, and identification of true cases may be challenging. As PCOS is associated with higher risks of depression and mental health disorders, there has been some concerns that maternal mental health may confound the relationship between PCOS and offspring neurodevelopment, despite a lack of evidence that maternal mental health conditions cause PCOS. A large recent study in the United Kingdom addressed this concern by comparing offspring of women with PCOS to offspring of mothers without PCOS with similar mental health history, finding that PCOS was associated with higher risk ADHD and autism regardless of maternal mental health history. 92 A second concern is that PCOS may be more frequently diagnosed in women pursuing fertility treatment, which in the US is expensive and may reflect a population with higher socioeconomic status (SES). In these cases, lower SES women with PCOS may be underrepresented in some studies if study selection is conditional on pursuit of fertility treatment. This could lead to greater misclassification of PCOS status by SES among a control or non-PCOS reference population compared to a case or PCOS-positive group in research studies. Due to potential associations of SES with PCOS and neurodevelopment, statistical adjustment for SES is an important consideration. However, despite concerns about misclassification of PCOS by SES, the consistency of estimates across different countries and health systems where diagnosis of PCOS may be less related to SES – such as European countries with national healthcare systems – suggests that associations between PCOS and neurodevelopmental disorders are less likely to be wholly attributable to SES or country-specific patterns of diagnosis. The hypothesis that maternal PCOS may increase risk for offspring neurodevelopmement is also supported by animal models, which indicate that a PCOS-like phenotype in mothers can have a host of metabolic and developmental implications for offspring. 100 Studies that show a relationship between fertility treatment and offspring health, particularly neurodevelopment, should consider the role of PCOS and other causes of infertility. Hypothyroidism affects 2–4% of reproductive-age individuals and is most often caused by thyroid autoimmunity. 101 Thyroid hormones including thyroid stimulating hormone (TSH), thyroxine (T4) and triiodothyronine (T3) exert direct and indirect effects on normal ovarian function and may thus be relevant for female infertility. For example, granulosa cells contain T3 binding sites and serum levels of T3 and T4 have been shown to correlate with follicular fluid levels of these hormones. 102 Thyroid hormones may also act indirectly on the hypothalamic-pituitary-ovarian axis via suppression of sex hormone-binding globulin (SHBG) activity and elevation of thyrotropin-releasing hormone (TRH). 103 Specifically, decreased binding activity of SHBG may increase free testosterone and estradiol, resulting in elevated androstenedione and estrone. It is currently unclear whether the effects of thyroid hormones on sex steroid hormone levels and menstrual cycle length are sufficient to influence female fertility, as epidemiologic studies evaluating hypothyroid and infertility are scarce and conflicting. A cross-sectional study of women ages 18–39 reported significantly higher TSH levels among women with unexplained infertility as compared to controls (1.95 versus 1.66 mIU/L; P = 0.0003). 104 Moreover, twice as many women with unexplained infertility had TSH levels ≥ 2.5 mIU/L as compared to controls (P<0.05). These findings are consistent with those of Poppe et al, who observed significantly higher mean TSH levels among 197 infertile women presenting at a fertility clinic as compared to 100 parous women (1.7 +/− 8.8 versus 1.2 +/−0.7 mIU/L; P=0.05). 105 The association was particularly strong for women experiencing ovulatory disorder infertility as compared to control women (1.9 +/− 4.9 versus 1.2 +/−0.7 mIU/L; P=0.003). The presence of thyroid peroxidase antibodies (TPO-Abs) was also associated with 2.5 times the odds (95% CI 1.02–5.12) of infertility as compared to no TPO-Abs. In a more recent prospective study of euthyroid women, preconception subclinical hypothyroidism, as defined by TSH ≥2.5 mIU/L, was not associated with time to pregnancy, pregnancy loss, or live birth among healthy fecund women. 106 The presence of TPO-Abs antibodies was also not associated with any of these outcomes in that population. Hypothyroidism seems to be related to disruption of ovarian function and menstrual cycle disturbance, which may subsequently lead to subfertility and indication for infertility treatment. Collectively, the existing evidence suggests that that overt hypothyroidism is positively associated with infertility, whereas subclinical hypothyroidism is not, though there is much debate about the relevant cut-point for differentiating the two. 107 Furthermore, it is largely unknown what the joint effects of hypothyroidism, infertility and ART treatment may have on the health outcomes of the offspring. Future studies should carefully consider these joint effects in well-designed prospective cohort studies. Thyroid hormones are essential for growth and development of all organs during the fetal period and later. In the brain, they act through nuclear receptors and regulate gene expression influencing neurogenesis, neuronal migration, neuronal and glial cell differentiation, myelination, and synaptogenesis. 108 – 111 During the first weeks of gestation when the fetal thyroid gland is not fully functional, inadequate thyroid hormones in the mother can also be detrimental because the fetus solely relies on maternal transfer of hormones for growth and development. 112 Even though substantial physiological alteration occurs in thyroid function of pregnant women to maintain the need of the mother and fetus, thyroid dysfunction is a common clinical problem during pregnancy. 113 If it remains undetected or inadequately treated, maternal overt hypothyroidism results in cognitive impairments in the offspring, even in the absence of neonatal hypothyroidism. 114 Neuroimaging studies of children born to hypothyroid mothers have shown that brain morphological abnormalities may persist as the child ages despite maternal treatment. 115 , 116 In view of this evidence, guidelines recommend treatment of overt hypothyroidism during pregnancy to prevent adverse outcomes in both mother and child. 117 , 118 In recent years, emerging evidence from epidemiological studies suggests that milder forms of thyroid hormone insufficiency during gestation are also associated with suboptimal cognitive and behavioral outcomes in children. Subclinical hypothyroidism which is defined as high levels of TSH with T4 within the gestational age-specific normal range is associated with a reduction in children’s IQ. 114 Differences in mental development and neuropsychological functioning have been reported in children born to women with subclinical hypothyroidism compared to euthyroid women. 119 , 120 Children born to women with isolated hypothyroxinemia (T4 values below the 2.5 th or 5 th percentile with TSH within the normal range) are at increased risk of suboptimal cognitive function 121 , language delay 122 , ADHD symptoms and other behavioral problems, 123 – 126 autistic symptoms, 127 , 128 motor function and executive function, 124 and schizophrenia. 129 One neuroimaging study reported an association of maternal hypothyroxinemia with smaller grey matter and cortex volumes and larger cerebellar white matter volumes in children at age 6 years. 130 Null associations have been reported as well. 131 Grounded on convincing evidence from observational studies, 121 – 125 , 127 – 130 two randomized clinical trials have examined the benefit of treatment on cognitive function in the children of women who underwent thyroid screening in early pregnancy and had mild thyroid insufficiency. 132 , 133 However, the trials have failed to demonstrate expected benefits of treatment and improvement in children’s IQ. Nonetheless, some speculate that the null findings are attributed to late interventions (after week 12 of gestation) or potential overtreatment. 134 , 135 Considering the lack of strong evidence on the association between milder forms of hypothyroidism and infertility, it is unlikely that subclinical hypothyroidism or hypothyroxinemia explain the potential impact of ART on the offspring’s neurodevelopment. Overt hypothyroidism in women, however, should be considered in studies of ART and child neurodevelopment. Empirical evidence suggests that there is an association between depression and infertility. 136 For instance, a cross-sectional study in Hungary found that infertile women had significantly higher depression scores than their fertile counterparts (14.94 vs. 8.95, p<0.0001). 137 Similarly, a Polish study found higher mean depression scores associated with infertility among women (8.5 vs. 5.2, p<0.0001). 138 Men, although less studied, are not exempt to depressive symptoms associated with infertility. 139 Methodological limitations, including the cross-sectional design, small sample size, lack of appropriate comparison groups, and inadequate adjustment for confounding (particularly with history of depression) present a challenge in determining whether depression is a predictor or a consequence of infertility. 140 Experiencing depressive mood over time may have a negative influence on the regulation of reproductive hormones in male and female, or it may give rise to conditions such as obesity which in turn affects fertility. 141 However, it is less clear whether infertility can be the underlying factor for depressive mood versus a cause of infertility. Given limitations of the literature, more methodologically rigorous studies are needed for a thorough understanding of the etiological direction of the association between depression and infertility. In particular, examining the timing of depression onset, duration, and history during the preconception period may improve the understanding of this relationship. Information on the timing and nature of depression may enable reproductive health researchers to establish a timeline for development of depression with respect to fertility status for couples. Irrespective of the mode of conception, parental psychological health may negatively impact child outcomes. Perinatal depression (depression during pregnancy and postpartum period) particularly has a negative influence on infant and child development. 142 , 143 Various studies have shown that children born to women experiencing perinatal depression are at a higher risk for experiencing adverse behavioral and developmental outcomes including preterm birth 144 , externalizing and internalizing behaviors 145 , 146 , poor growth 147 , and reduced cognitive functioning. 148 Fathers are also at risk for experiencing perinatal depression 149 , though research on fathers’ depression status during this time period is scarce 150 , 151 and its effect on child outcomes is unclear. Several studies suggest fathers’ perinatal depression is associated with children’s adverse emotional status 152 , 153 however a more recent study found that this relationship is mediated by maternal depression and couple conflicts. 154 Depression during the perinatal period may be an indication of a more chronic condition with an onset in the preconception period. However, studies examining the onset, frequency, and duration of depression during preconception period are limited. Furthermore, despite accumulating evidence on the presence of perinatal depression in fathers 155 , 151 , the literature on this topic including predictors of this condition —specifically the timing and the history of depression— and its association with infertility and child outcomes is very limited and remain important to assess. Aside from the aforementioned conditions indicated for infertility treatment, little is known about the long-term health outcomes of children conceived by ART to women with other conditions associated with infertility, such as endometriosis and fibroids. Endometriosis is associated with infertility, and accounts for up to 40% of all infertility causes among couples seeking treatment. 156 Importantly, endometriosis is associated with an inflammatory milieu. 157 Women with endometriosis undergoing ART experience a reduced rate of fertilization, 156 thought to be caused by excessive local production of pro-inflammatory cytokines including interleukins and tumor necrosis factor-alpha (TNF-α). In line with the inflammation hypothesis, women with endometriosis are at increased risk of hypertension and cardiovascular disease, indicating long-term health consequences associated with inflammation beyond just the reproductive years. 158 , 159 Uterine fibroids occur in up to 70% of women and are also associated with infertility. 160 Some have hypothesized that inflammation may also play a role in the pathogenesis of uterine fibroids, and women with fibroids appear to experience higher cardiovascular risk. 161 Because inflammation is associated with adverse pregnancy and child health outcomes, future epidemiologic studies should evaluate long-term child health outcomes among those conceived with ART by mothers with uterine fibroids and endometriosis. Despite the long list of possible related conditions reviewed above, a note must be made that the bias from confounding is dependent on its strength of association with the outcome of interest as well. It is usually best to draw a causal diagram 162 depicting the model one wants to estimate associations for and determine which are true confounders and which may be mediators of associations (e.g., preterm delivery) that need not be adjusted for, or may require more specific types of analyses. Figure 1 shows an example of a conceptual diagram for examining the association between ART and the outcome of neurodevelopment. One can adjust for BMI, underlying infertility and socio-demographics (i.e., age, race, education) to adequately control for the confounders listed. In this case, even if underlying infertility does not directly impact neurodevelopment in some way, due to its association with upstream factors, its adjustment would remain a possible option for controlling for confounding of these other factors (e.g., thyroid function) that might not be well captured in a specific study but are known to lead to neurodevelopmental issues. However, even if there is no information collected on some of the indicators of infertility treatment in a specific study, the impact of confounding by indication for each specific research question can be estimated using bias analyses. 163 Such analyses may then help evaluate whether the missing confounding information would or would not have impacted the associations found in the study. Following the recommended reporting guidelines such as STROBE for observational research 164 , unadjusted estimates should also be published along with models with successive adjustment models to better evaluate the impact of covariate adjustment. In addition, causal evidence for the confounders are required. For the above factors, many of which have been understudied, the true estimates of associations along with biological underpinnings between them and child health remain unclear. Further research is necessary to provide better estimates for bias analyses. In sum, the evidence accumulated on the health of children after conception by ART (and less so for non-ART treatments) suggest that they are largely similar in development and growth as their peers. However, given the heterogeneity and continuing advancements of ART procedures, ongoing research and surveillance that tackles the existing methodological challenges is essential. Even if fertility treatment techniques themselves may not cause marked differences in long-term health, attention should also be given to understanding the impact of the causes of infertility including both male and female factors.