- Split View
-
Views
-
Cite
Cite
Zheng Chang, Paul Lichtenstein, Brian M D’Onofrio, Catarina Almqvist, Ralf Kuja-Halkola, Arvid Sjölander, Henrik Larsson, Maternal age at childbirth and risk for ADHD in offspring: a population-based cohort study, International Journal of Epidemiology, Volume 43, Issue 6, December 2014, Pages 1815–1824, https://doi.org/10.1093/ije/dyu204
- Share Icon Share
Abstract
Background: Women who give birth at younger ages (e.g. teenage mothers) are more likely to have children who exhibit behaviour problems, such as attention-deficit/hyperactivity disorder (ADHD). However, it is not clear whether young maternal age is causally associated with poor offspring outcomes or confounded by familial factors.
Methods: The association between early maternal age at childbirth and offspring ADHD was studied using data from Swedish national registers. The sample included all children born in Sweden between 1988 and 2003 (N = 1 495 543), including 30 674 children with ADHD. We used sibling- and cousin-comparisons to control for unmeasured genetic and environmental confounding. Further, we used a children-of-siblings model to quantify the genetic and environmental contribution to the association between maternal age and offspring ADHD.
Results: Maternal age at first birth (MAFB) was associated with offspring ADHD. Teenage childbirth (<20 years) was associated with 78% increased risk of ADHD. The association attenuated in cousin-comparison, suggesting unmeasured familial confounding. The children-of-siblings model indicated that the association between MAFB and ADHD was mainly explained by genetic confounding.
Conclusions: All children born to mothers who bore their first child early in their reproductive lives were at increased risk of ADHD. The association was mainly explained by genetic factors transmitted from mothers to their offspring that contribute to both age at childbirth and ADHD in offspring. Our results highlight the importance of using family-based designs to understand how early life circumstances affect child development.
Women who give birth at younger ages (e.g. teenage mothers) are more likely to have children who exhibit behaviour problems such as ADHD. However, the nature of this association is not clear.
All offspring born to mothers who began childbearing early are at increased risk of ADHD.
The association between early maternal age and offspring ADHD is mainly explained by genetic confounding.
Introduction
Attention-deficit/hyperactivity disorder (ADHD) is the most common neurobehavioural disorder of childhood. This highly prevalent disorder affects about 5% of children worldwide,1 with a substantial degree of persistence over time.2 In addition to the high heritability of ADHD,3 accumulating epidemiological studies highlight the importance of environmental factors in the aetiology of ADHD.4,5 Teenage childbearing is recognized as an important public health concern worldwide and is associated with a wide range of negative developmental outcomes in offspring.6 Although a few studies have shown that young maternal age is associated with an increased risk for ADHD,5,7,8 the nature of this association remains unclear for two major reasons.8
First, not all studies that investigate the effect of maternal age on offspring ADHD make the distinction between maternal age at each birth (MAEB) and maternal age at first birth (MAFB): that is, whether the effect of young maternal age is specific to each child or shared by siblings. Some studies found that MAFB is a better index than MAEB for behavioural outcomes in offspring of women who have more than one child.9,10 If this is the case, all children born to mothers who began childbearing early should have a similar risk of ADHD.
Second, most studies have not rigorously examined the role of plausible confounding factors (including familial factors) which could provide alternative explanations for the observed association.11,12 Young maternal age might be causally associated with offspring ADHD. Early childbearing may interfere with the developmental trajectory of teenage mothers, introducing social and economic stressors that constrain their ability to parent effectively.13,14 This increases the likelihood of a less supportive or disadvantaged home environment, factors which are themselves associated with ADHD.15,16 Alternatively, young maternal age might not be causally associated with offspring ADHD; rather, it could be a marker for environmental or genetic confounding.6,14 Previous research suggests that teenage mothers tend to come from disadvantaged backgrounds (e.g. low-education and high-poverty communities),6 and these environmental risk factors are also independently associated with ADHD.13,17 In addition, both age at childbearing and ADHD are influenced by genetics,3,12 and genetic predisposition for putative environmental risk factors (i.e. young maternal age) could therefore be passed down from parents to their offspring, and account for offspring’s risk for ADHD.
Because experimental studies are not feasible for studying exposures such as maternal age, family-based designs (e.g. sibling-comparison, cousin-comparison) represent an important alternative for causal inference in observational studies.18,19 The current study explored the association between early maternal age at childbirth and offspring ADHD in a large population-based cohort of children, with family-based designs. Specifically, we first examined whether MAEB or MAFB is a better index for the risk of offspring ADHD. Second, we explored the mechanisms that underlie the association between early maternal age at childbirth and offspring ADHD. We did this by comparing siblings within nuclear families and cousins within extended families to control for familial confounding. We also used a children-of-siblings model to quantify the contribution of genetic and environmental influences on the observed association.
Methods
Subjects
We used data through linkage of longitudinal, population-based registers in Sweden using unique personal identification numbers.20 The Medical Birth Register (MBR) includes information on newborn children and their mothers for more than 99% of births in Sweden.21 The National Patient Register (NPR) contains diagnoses according to the World Health Organization’s (WHO’s) International Classification of Diseases (ICD-10), from inpatient care and outpatient visits since 2001.22 The Prescribed Drug Register (PDR) includes data on dispensed pharmaceuticals. Information regarding drug identity according to the Anatomical Therapeutic Chemical (ATC) classification system, quantity and dosage of the prescribed drug and date of prescription has been registered since July 2005.23 The Multi-Generation Register (MGR) was used to link all individuals to their parents,24 thus allowing us to identify different types of family structure. The Swedish Twin Register was used to identify monozygotic (MZ) and dizygotic (DZ) twins.25 The Migration Register and Cause of Death Register were also used to account for migration and death.
We studied a cohort of children born in Sweden between 1988 and 2003, consisting of 1 640 694 newborn children identified from the MBR. We excluded those who: had severe congenital malformations (n = 59 390); were from multiple births (n = 43 681); were stillborn (n = 5162); died (n = 5460); emigrated before age 3 years or year 2001 (n = 31 398); or received an ADHD diagnosis before age 3 years (n = 60). The final sample included 1 495 543 eligible individuals (91.2% of the targeted population), who were offspring of 896 389 mothers (Figure 1). We identified 988 625 full siblings from 443 555 mothers among the full sample, and we also identified children of mothers who had sisters, half-sisters or same-sex twins (Figure 1).
Measures
MAEB and MAFB
Maternal age at each birth (MAEB), as a continuous measure, was obtained from the MBR. By linking to the MGR, we also obtained maternal age at first birth (MAFB) for each mother. We dichotomized maternal age into teenage and adult childbirth, in line with previous research.26 The binary measure of MAEB/MAFB was defined as teenage childbirth if the mother’s age at each/first childbirth was below 20 years and adult childbirth if the mother’s age was at least 20 years. The proportion of teenage childbirth in this sample was 2.3% of all births, and 5.0% of all first births, which is consistent with recent demographic studies in Sweden.27 The continuous measures of MAEB/MAFB were re-centred at 26 years, the mean maternal age at first birth.
Offspring ADHD
Children with ADHD diagnoses in the NPR (ICD-10 code: F90) or treated with ADHD medication [methylphenidate (ATC code: N06BA04); atomoxetin (N06BA09); amphetamine (N06BA01); dexamphetamine (N06BA02)] according to the PDR were identified as ADHD cases, as this combined approach has been found to be a valid indicator of ADHD.28 In total, we identified 30 674 children with ADHD from these two registers (23 548 from the NPR and 23 912 from the PDR); the date of diagnosis was defined as the date of first record in any registers.
Statistical analyses
We explored the association between MAEB and offspring ADHD using four models with progressive adjustment for potential confounders. Cox regression was used to deal with the right-censored outcome measure. The first model examined the population-wide association between MAEB and offspring ADHD with adjustment for children’s sex, birth order and birth year in categories. In addition, the second and third model further adjusted for paternal age at childbirth and MAFB. The fourth model examined the same association using a sibling-comparison, which controls for unmeasured genetic and environmental factors that are shared by siblings.29 This was done by fitting a fixed-effect model30 (stratified Cox regression) to the subsample of full siblings. If MAEB causally increased the risk of ADHD, then children born to young mothers should have a higher risk of ADHD than their later-born siblings, and the increased risk should remain in all models. Otherwise, if the increased risk disappears when controlling for familial factors (Models 3 and 4), it suggests that all children born to mothers who began childbearing early have a similar risk of ADHD, and MAFB is a better index for this risk.
The mechanisms underlying the association between MAFB and ADHD in offspring were subsequently explored using two models. The first model examined the population-wide association between MAFB and offspring ADHD. The second model examined the association using cousin-comparison, which controls for unmeasured genetic and environmental factors that are shared within extended families.9,31 This was done by stratified Cox regression in the subsample of full cousins. If MAFB is causally associated with offspring ADHD, then the population-wide association between them should remain in the cousin-comparison. To investigate the modifying effect of birth order, we conducted analyses on firstborn cousins and non-firstborn cousins separately. All siblings/cousins were used in the sibling-/cousin-comparison described above, no matter if they were discordant on exposure or not. The discordant sibling/cousin pairs contributed information in adjusted stratified Cox regression, and siblings/cousins discordant for any other covariate or length of follow-up are also informative for the within-family estimates.
To quantify the magnitude of the different processes (e.g. causal, genetic and environmental confounding) explaining the observed association between MAFB and offspring ADHD, we used a children-of-siblings (CoS) model by fitting the structural equation model illustrated in Figure 2. The variances of MAFB and offspring ADHD were decomposed into four sources of components: additive genetic (A), extended-family environmental (T), nuclear-family environmental (N) and unique environmental (E) influences. We also estimated the genetic correlation (rA), the extended-family environmental correlation (rT) and the nuclear-family environmental correlation (rN) between MAFB and ADHD. These parameters can be used to quantify the contributions from genetic influence, extended-family environmental influence and nuclear-family environmental influence on the phenotypic correlation between MAFB and offspring ADHD, which support the interpretations of genetic confounding, environmental confounding and causal effect, respectively.
We randomly selected up to two sister mothers in each extended family (and included all twin sisters). We included six types of extended families: mothers with sisters, mothers with maternal half-sister, mothers with paternal half-sister, mothers with MZ or DZ twin sister, and those without sisters (Figure 1). Then we randomly drew up to two of each mothers’ offspring, thus constructing up to two nuclear families within each extended family. Within each nuclear family, offspring could be single, two full siblings or two maternal half-siblings.
To maximize power, we used the continuous measure of MAFB. A liability threshold model was applied to the binary variable (ADHD), assuming that the ordered categories reflect an imprecise measurement of an underlying normal distribution of liability.32 The structural equation modelling package OpenMx33 in R was used to perform full information maximum likelihood model-fitting with raw data. In the online supplement (available as Supplementary data at IJE online), we provided a more detailed description of the CoS model.
Results
Sample demographic characteristics are presented in Table 1. These characteristics were symmetrically distributed in the entire cohort, siblings and cousins, except for birth order and year of birth. MAEB was moderately correlated with paternal age (r = 0.67) and MAFB (r = 0.61).
Variables . | Entire cohort . | Full siblings . | Full cousins . |
---|---|---|---|
Number of individuals | 1 495 543 | 988 625 | 383 511 |
Sex (%) | |||
Male | 51.2 | 51.3 | 51.2 |
Female | 48.8 | 48.7 | 48.8 |
Birth order (%) | |||
1st | 41.9 | 36.8 | 40.7 |
2nd | 36.2 | 41.6 | 37.3 |
3rd | 15.2 | 14.9 | 15.8 |
>=4th | 6.7 | 6.7 | 6.2 |
Maternal age at birth (mean y ± SD) | 29.3 ± 5.1 | 29.0 ± 4.7 | 29.1 ± 4.8 |
Maternal age at first birth (mean y ± SD) | 26.0 ± 4.7 | 26.0 ± 4.3 | 26.0 ± 4.5 |
Paternal age at birth (%) | |||
<20 y | 0.6 | 0.4 | 0.5 |
20–24 y | 9.3 | 9.1 | 9.7 |
25–29 y | 28.9 | 31.3 | 31.5 |
30–34 y | 32.3 | 33.5 | 33.5 |
>=35 y | 28.3 | 25.5 | 24.6 |
Missing | 0.6 | 0.2 | 0.2 |
Calendar year of birth (%) | |||
1988 | 6.7 | 5.1 | 6.7 |
1989 | 7.0 | 5.5 | 7.2 |
1990 | 7.4 | 6.9 | 7.8 |
1991 | 7.4 | 7.8 | 8.0 |
1992 | 7.3 | 8.2 | 7.9 |
1993 | 7.0 | 8.1 | 7.4 |
1994 | 6.7 | 7.8 | 7.1 |
1995 | 6.2 | 7.2 | 6.4 |
1996 | 5.8 | 6.7 | 5.9 |
1997 | 5.4 | 6.3 | 5.5 |
1998 | 5.2 | 6.0 | 5.2 |
1999 | 5.2 | 5.9 | 5.1 |
2000 | 5.4 | 5.7 | 5.1 |
2001 | 5.5 | 4.9 | 5.0 |
2002 | 5.8 | 4.0 | 4.9 |
2003 | 6.0 | 3.9 | 4.8 |
Variables . | Entire cohort . | Full siblings . | Full cousins . |
---|---|---|---|
Number of individuals | 1 495 543 | 988 625 | 383 511 |
Sex (%) | |||
Male | 51.2 | 51.3 | 51.2 |
Female | 48.8 | 48.7 | 48.8 |
Birth order (%) | |||
1st | 41.9 | 36.8 | 40.7 |
2nd | 36.2 | 41.6 | 37.3 |
3rd | 15.2 | 14.9 | 15.8 |
>=4th | 6.7 | 6.7 | 6.2 |
Maternal age at birth (mean y ± SD) | 29.3 ± 5.1 | 29.0 ± 4.7 | 29.1 ± 4.8 |
Maternal age at first birth (mean y ± SD) | 26.0 ± 4.7 | 26.0 ± 4.3 | 26.0 ± 4.5 |
Paternal age at birth (%) | |||
<20 y | 0.6 | 0.4 | 0.5 |
20–24 y | 9.3 | 9.1 | 9.7 |
25–29 y | 28.9 | 31.3 | 31.5 |
30–34 y | 32.3 | 33.5 | 33.5 |
>=35 y | 28.3 | 25.5 | 24.6 |
Missing | 0.6 | 0.2 | 0.2 |
Calendar year of birth (%) | |||
1988 | 6.7 | 5.1 | 6.7 |
1989 | 7.0 | 5.5 | 7.2 |
1990 | 7.4 | 6.9 | 7.8 |
1991 | 7.4 | 7.8 | 8.0 |
1992 | 7.3 | 8.2 | 7.9 |
1993 | 7.0 | 8.1 | 7.4 |
1994 | 6.7 | 7.8 | 7.1 |
1995 | 6.2 | 7.2 | 6.4 |
1996 | 5.8 | 6.7 | 5.9 |
1997 | 5.4 | 6.3 | 5.5 |
1998 | 5.2 | 6.0 | 5.2 |
1999 | 5.2 | 5.9 | 5.1 |
2000 | 5.4 | 5.7 | 5.1 |
2001 | 5.5 | 4.9 | 5.0 |
2002 | 5.8 | 4.0 | 4.9 |
2003 | 6.0 | 3.9 | 4.8 |
Variables . | Entire cohort . | Full siblings . | Full cousins . |
---|---|---|---|
Number of individuals | 1 495 543 | 988 625 | 383 511 |
Sex (%) | |||
Male | 51.2 | 51.3 | 51.2 |
Female | 48.8 | 48.7 | 48.8 |
Birth order (%) | |||
1st | 41.9 | 36.8 | 40.7 |
2nd | 36.2 | 41.6 | 37.3 |
3rd | 15.2 | 14.9 | 15.8 |
>=4th | 6.7 | 6.7 | 6.2 |
Maternal age at birth (mean y ± SD) | 29.3 ± 5.1 | 29.0 ± 4.7 | 29.1 ± 4.8 |
Maternal age at first birth (mean y ± SD) | 26.0 ± 4.7 | 26.0 ± 4.3 | 26.0 ± 4.5 |
Paternal age at birth (%) | |||
<20 y | 0.6 | 0.4 | 0.5 |
20–24 y | 9.3 | 9.1 | 9.7 |
25–29 y | 28.9 | 31.3 | 31.5 |
30–34 y | 32.3 | 33.5 | 33.5 |
>=35 y | 28.3 | 25.5 | 24.6 |
Missing | 0.6 | 0.2 | 0.2 |
Calendar year of birth (%) | |||
1988 | 6.7 | 5.1 | 6.7 |
1989 | 7.0 | 5.5 | 7.2 |
1990 | 7.4 | 6.9 | 7.8 |
1991 | 7.4 | 7.8 | 8.0 |
1992 | 7.3 | 8.2 | 7.9 |
1993 | 7.0 | 8.1 | 7.4 |
1994 | 6.7 | 7.8 | 7.1 |
1995 | 6.2 | 7.2 | 6.4 |
1996 | 5.8 | 6.7 | 5.9 |
1997 | 5.4 | 6.3 | 5.5 |
1998 | 5.2 | 6.0 | 5.2 |
1999 | 5.2 | 5.9 | 5.1 |
2000 | 5.4 | 5.7 | 5.1 |
2001 | 5.5 | 4.9 | 5.0 |
2002 | 5.8 | 4.0 | 4.9 |
2003 | 6.0 | 3.9 | 4.8 |
Variables . | Entire cohort . | Full siblings . | Full cousins . |
---|---|---|---|
Number of individuals | 1 495 543 | 988 625 | 383 511 |
Sex (%) | |||
Male | 51.2 | 51.3 | 51.2 |
Female | 48.8 | 48.7 | 48.8 |
Birth order (%) | |||
1st | 41.9 | 36.8 | 40.7 |
2nd | 36.2 | 41.6 | 37.3 |
3rd | 15.2 | 14.9 | 15.8 |
>=4th | 6.7 | 6.7 | 6.2 |
Maternal age at birth (mean y ± SD) | 29.3 ± 5.1 | 29.0 ± 4.7 | 29.1 ± 4.8 |
Maternal age at first birth (mean y ± SD) | 26.0 ± 4.7 | 26.0 ± 4.3 | 26.0 ± 4.5 |
Paternal age at birth (%) | |||
<20 y | 0.6 | 0.4 | 0.5 |
20–24 y | 9.3 | 9.1 | 9.7 |
25–29 y | 28.9 | 31.3 | 31.5 |
30–34 y | 32.3 | 33.5 | 33.5 |
>=35 y | 28.3 | 25.5 | 24.6 |
Missing | 0.6 | 0.2 | 0.2 |
Calendar year of birth (%) | |||
1988 | 6.7 | 5.1 | 6.7 |
1989 | 7.0 | 5.5 | 7.2 |
1990 | 7.4 | 6.9 | 7.8 |
1991 | 7.4 | 7.8 | 8.0 |
1992 | 7.3 | 8.2 | 7.9 |
1993 | 7.0 | 8.1 | 7.4 |
1994 | 6.7 | 7.8 | 7.1 |
1995 | 6.2 | 7.2 | 6.4 |
1996 | 5.8 | 6.7 | 5.9 |
1997 | 5.4 | 6.3 | 5.5 |
1998 | 5.2 | 6.0 | 5.2 |
1999 | 5.2 | 5.9 | 5.1 |
2000 | 5.4 | 5.7 | 5.1 |
2001 | 5.5 | 4.9 | 5.0 |
2002 | 5.8 | 4.0 | 4.9 |
2003 | 6.0 | 3.9 | 4.8 |
MAEB and offspring ADHD
Teenage childbirth was associated with substantially increased risk of offspring ADHD [Model 1, hazard ratio (HR) = 2.24, Table 2]. The association was attenuated but remained substantial after adjustment for paternal age at childbirth (Model 2, HR = 1.57). However, the increased risk disappeared when adjusting for MAFB (Model 3, HR = 0.90) or adjusting for unmeasured genetic and environmental factors using sibling-comparison (Model 4, HR = 0.80). A similar pattern of result was observed when using the continuous measure of MAEB. One-year earlier MAEB increased the risk of ADHD by 6% (HR = 1.06), but the increased risk disappeared when adjusting for MAFB or comparing siblings. Compared with a quadratic model, we found a linear model provided a good approximation of the association (Supplementary Figure 1, available as Supplementary data at IJE online). We also found similar results from analyses using the full sample, and only using the siblings (Supplementary Table 1, available as Supplementary data at IJE online). Taken together, the results indicated that offspring born when their mothers were at young age do not have a higher risk of ADHD than their later-born siblings.
MAEB . | Model 1a . | Model 2b . | Model 3c . | Model 4d . |
---|---|---|---|---|
Binarye | 2.24 (2.12–2.36) | 1.57 (1.48–1.67) | 0.90 (0.84–0.96) | 0.81 (0.71–0.94) |
Continuous | 1.06 (1.05–1.06) | 1.05 (1.04–1.05) | 0.99 (0.99–1.00) | 0.98 (0.97–0.99) |
MAEB . | Model 1a . | Model 2b . | Model 3c . | Model 4d . |
---|---|---|---|---|
Binarye | 2.24 (2.12–2.36) | 1.57 (1.48–1.67) | 0.90 (0.84–0.96) | 0.81 (0.71–0.94) |
Continuous | 1.06 (1.05–1.06) | 1.05 (1.04–1.05) | 0.99 (0.99–1.00) | 0.98 (0.97–0.99) |
aPopulation-wide association, adjusted for offspring’s sex, birth order and birth year in categories.
bIn addition to Model 1, adjusted for paternal age at childbirth in categories.
cIn addition to Model 2, adjusted for MAFB.
dSibling-comparison, adjusted for unmeasured genetic and environmental factors shared by siblings and measured covariates.
eMAEB < 20 y.
MAEB . | Model 1a . | Model 2b . | Model 3c . | Model 4d . |
---|---|---|---|---|
Binarye | 2.24 (2.12–2.36) | 1.57 (1.48–1.67) | 0.90 (0.84–0.96) | 0.81 (0.71–0.94) |
Continuous | 1.06 (1.05–1.06) | 1.05 (1.04–1.05) | 0.99 (0.99–1.00) | 0.98 (0.97–0.99) |
MAEB . | Model 1a . | Model 2b . | Model 3c . | Model 4d . |
---|---|---|---|---|
Binarye | 2.24 (2.12–2.36) | 1.57 (1.48–1.67) | 0.90 (0.84–0.96) | 0.81 (0.71–0.94) |
Continuous | 1.06 (1.05–1.06) | 1.05 (1.04–1.05) | 0.99 (0.99–1.00) | 0.98 (0.97–0.99) |
aPopulation-wide association, adjusted for offspring’s sex, birth order and birth year in categories.
bIn addition to Model 1, adjusted for paternal age at childbirth in categories.
cIn addition to Model 2, adjusted for MAFB.
dSibling-comparison, adjusted for unmeasured genetic and environmental factors shared by siblings and measured covariates.
eMAEB < 20 y.
MAFB and offspring ADHD
Considering MAFB, teenage childbirth was associated with 78% increased risk of offspring ADHD at population level (HR = 1.78, Table 3). Using cousin-comparison, the effect of teenage childbirth was somewhat attenuated (HR = 1.33), but remained significantly associated with offspring ADHD. We saw the same pattern of results using the continuous measure of MAFB. One-year earlier MAFB increased the risk of offspring ADHD by 7% (HR = 1.07), and the association was attenuated in cousin-comparison (HR = 1.03). The association in firstborn cousins [HR = 1.03, 95% confidence interval (CI): 1.00–1.05] was similar to those in non-firstborn cousins (HR = 1.03, 95% CI: 1.01–1.05), suggesting a limited role of birth order in modifying the observed association. In the Supplementary Table 2 (available as Supplementary data at IJE online), we also report the association between MAFB and offspring ADHD in other relationships. Taken together, the results suggest that MAFB is strongly associated with offspring ADHD, and at least part of the association is confounded by unmeasured familial factors.
MAFB . | Population-widea . | Cousin-comparisonb . |
---|---|---|
Binaryc | 1.78 (1.72–1.84) | 1.33 (1.18–1.50) |
Continuous | 1.07 (1.06–1.07) | 1.03 (1.02–1.04) |
MAFB . | Population-widea . | Cousin-comparisonb . |
---|---|---|
Binaryc | 1.78 (1.72–1.84) | 1.33 (1.18–1.50) |
Continuous | 1.07 (1.06–1.07) | 1.03 (1.02–1.04) |
aPopulation-wide association, adjusted for offspring’s sex, birth order in categories, birth year in categories and paternal age at childbirth in categories.
bCousin-comparison, adjusted for unmeasured genetic and environmental factors shared by cousins and measured covariates.
cMAEB < 20 y.
MAFB . | Population-widea . | Cousin-comparisonb . |
---|---|---|
Binaryc | 1.78 (1.72–1.84) | 1.33 (1.18–1.50) |
Continuous | 1.07 (1.06–1.07) | 1.03 (1.02–1.04) |
MAFB . | Population-widea . | Cousin-comparisonb . |
---|---|---|
Binaryc | 1.78 (1.72–1.84) | 1.33 (1.18–1.50) |
Continuous | 1.07 (1.06–1.07) | 1.03 (1.02–1.04) |
aPopulation-wide association, adjusted for offspring’s sex, birth order in categories, birth year in categories and paternal age at childbirth in categories.
bCousin-comparison, adjusted for unmeasured genetic and environmental factors shared by cousins and measured covariates.
cMAEB < 20 y.
The children-of-siblings model
Table 4 shows the genetic and environmental influence on MAFB, offspring ADHD and the association between them. MAFB showed a moderate heritability, with genetic influence explaining 49% of the variance. ADHD showed a high heritability, with genetic influence explaining 73% of the variance. There was a significant phenotypic correlation between MAFB and ADHD (r = 0.12). The phenotypic correlation between MAFB and offspring ADHD was mainly explained by the genetic correlation (rA = 0.40) between MAFB and ADHD. This result confirmed and extended the findings from the cousin-comparison in suggesting that the association between MAFB and offspring ADHD was not causal and was, instead, explained largely or entirely through genetic factors that influence the liability to both MAFB and ADHD.
. | Genetic and environmental influence on MAFB and ADHD . | |||
---|---|---|---|---|
A% . | T% . | N% . | E% . | |
MAFB | 49 (45–54) | 6.7 (4.5–8.8) | 44 (42–46) | |
ADHD | 73 (63–85) | 5.9 (2.6–9.5) | 2.1 (1.5–3.4) | 19 (14–26) |
Genetic and environmental influence onthe association between MAFB and ADHD | ||||
rA | rT | rN | Phenotypic correlation | |
Cross-phenotype correlationa | 0.40 (0.24–0.60) | 0.17 (−0.42–0.78) | −0.13 (−0.99–0.99) | |
Contribution to phenotypic correlationb | 0.12 | 0.01 | −0.01 | 0.12 |
. | Genetic and environmental influence on MAFB and ADHD . | |||
---|---|---|---|---|
A% . | T% . | N% . | E% . | |
MAFB | 49 (45–54) | 6.7 (4.5–8.8) | 44 (42–46) | |
ADHD | 73 (63–85) | 5.9 (2.6–9.5) | 2.1 (1.5–3.4) | 19 (14–26) |
Genetic and environmental influence onthe association between MAFB and ADHD | ||||
rA | rT | rN | Phenotypic correlation | |
Cross-phenotype correlationa | 0.40 (0.24–0.60) | 0.17 (−0.42–0.78) | −0.13 (−0.99–0.99) | |
Contribution to phenotypic correlationb | 0.12 | 0.01 | −0.01 | 0.12 |
A, additive genetic; T, extended-family environmental; N, nuclear-family environmental; E, unique environmental influence.
arA, rT, rN, cross-phenotype correlation of each variance component (A, T, N).
bPhenotypic correlation explained by additive genetic correlation is calculated as 0.5*√AMAFB*√AADHD*rA, by extended-family environmental correlation √TMAFB*√TADHD*rT and by nuclear-family environmental correlation √NMAFB*√NADHD*rN. The sum of the three components is the phenotypic correlation.
. | Genetic and environmental influence on MAFB and ADHD . | |||
---|---|---|---|---|
A% . | T% . | N% . | E% . | |
MAFB | 49 (45–54) | 6.7 (4.5–8.8) | 44 (42–46) | |
ADHD | 73 (63–85) | 5.9 (2.6–9.5) | 2.1 (1.5–3.4) | 19 (14–26) |
Genetic and environmental influence onthe association between MAFB and ADHD | ||||
rA | rT | rN | Phenotypic correlation | |
Cross-phenotype correlationa | 0.40 (0.24–0.60) | 0.17 (−0.42–0.78) | −0.13 (−0.99–0.99) | |
Contribution to phenotypic correlationb | 0.12 | 0.01 | −0.01 | 0.12 |
. | Genetic and environmental influence on MAFB and ADHD . | |||
---|---|---|---|---|
A% . | T% . | N% . | E% . | |
MAFB | 49 (45–54) | 6.7 (4.5–8.8) | 44 (42–46) | |
ADHD | 73 (63–85) | 5.9 (2.6–9.5) | 2.1 (1.5–3.4) | 19 (14–26) |
Genetic and environmental influence onthe association between MAFB and ADHD | ||||
rA | rT | rN | Phenotypic correlation | |
Cross-phenotype correlationa | 0.40 (0.24–0.60) | 0.17 (−0.42–0.78) | −0.13 (−0.99–0.99) | |
Contribution to phenotypic correlationb | 0.12 | 0.01 | −0.01 | 0.12 |
A, additive genetic; T, extended-family environmental; N, nuclear-family environmental; E, unique environmental influence.
arA, rT, rN, cross-phenotype correlation of each variance component (A, T, N).
bPhenotypic correlation explained by additive genetic correlation is calculated as 0.5*√AMAFB*√AADHD*rA, by extended-family environmental correlation √TMAFB*√TADHD*rT and by nuclear-family environmental correlation √NMAFB*√NADHD*rN. The sum of the three components is the phenotypic correlation.
Discussion
The current study examined the association between early maternal age at childbirth and offspring ADHD, in a large population-based cohort with family-based designs. Our results support two main conclusions. First, maternal age at first birth predicted offspring ADHD. Thus, all offspring born to mothers who began childbearing early were at increased risk of ADHD. Second, the association between MAFB and offspring ADHD was mainly explained by genetic confounding: that is, genetic factors transmitted from mothers to children contribute to both age at childbirth in mothers and ADHD in offspring.
In line with previous research,5,7,8 we found that MAEB was associated with an increased risk of offspring ADHD. However, the increased risk disappeared when controlling for MAFB or conducting sibling-comparison, suggesting that the population-wide association between maternal age and offspring ADHD was due to risk factors at a family level. Similar findings have been observed for other behavioural outcomes.9,10 Thus, all offspring born to mothers who began childbearing early were at increased risk of ADHD, which may have important implications for the planning of family-based prevention efforts. If anything, the within-family analyses were indicative of a protective effect of MAEB. There are at least two alternative explanations to this finding. Advancing maternal age might increase the risk of neurodevelopmental disorders, as suggested by a meta-analysis on maternal age and autism.34 Alternatively, because of the correlation between maternal age and paternal age, the protective effect could also be explained by residual confounding of advancing paternal age, which has been reported as a risk factor for psychiatric disorders.8,35 Further research is required to explore whether advancing maternal age or paternal age increases the risk of ADHD.
Family-based designs (e.g. sibling- and cousin-comparisons) have been used to test causal hypotheses about putative environmental risk factors,9,10,36 but no previous study with such designs has investigated the association between young maternal age and offspring ADHD. Compared with population-wide estimates, the association between MAFB and offspring ADHD substantially decreased in the cousin comparisons, suggesting that the association was largely explained by unmeasured familial factors. Similar results were found in a study on maternal age and cognitive test score.10 In contrast, other studies have found support for a causal mechanism underlying the association between MAFB and offspring disruptive behaviors and criminal convictions (including one Swedish study).9,36 These different results might suggest that MAFB is one aetiological factor that differentiates ADHD from other externalizing disorders. This interpretation is consistent with a meta-analysis showing that shared environmental influences accounted for 10–19% of the variance in externalizing disorders, but in contrast had a very limited impact on ADHD.37
To further understand the mechanism of the familial influences on the association between MAFB and offspring ADHD, we used a CoS model which was designed to explore the effect of a risk factor at the family level. We found that both MAFB and ADHD were heritable, and the association between them was mainly explained by shared genetic influences, which suggests that genetic factors influencing MAFB were passed down from parents to their offspring and accounted for most of the risk of ADHD. These results are consistent with findings from a recent twin study showing that the association between ADHD and adolescent sexual risk behaviour is due to genetic factors.38
Determination of causal connections between parental characteristics and child outcomes is one of the key questions in the field of developmental psychopathology.39,40 The sibling- and cousin-comparisons provided insight into whether familial factors account for the association between parental characteristics and child outcomes, and had the important advantage of conceptual clarity.18 However, they cannot quantify the magnitude of different processes explaining the observed association. The CoS model used in this study provided a powerful quantitative approach to disentangle and quantify the underlying processes explaining the observed association, which had several advantages. First, the model used multiple relative groups, thus increasing analytical power and improving generalizability. Second, more than one child from each nuclear family can be included in the model. Therefore, it was possible to study family-level risk factors that influenced all children in the same family. Third, the model can quantify the degree to which the association between parental characteristics and child outcomes is consistent with a causal influence, or instead is due to confounding factors (environmental or genetic).
The results in this study should also be considered in the context of its limitations. First, although recent validation checks of ADHD diagnoses in Swedish registers indicated low numbers of false-positives, our case identification strategy could not avoid false-negatives, especially for older children in this cohort. The potential misclassification should be non-differential for population-based comparison, and would lead to null findings; hence, our findings are probably conservative estimates of the actual effect of maternal age on ADHD. Second, teenage childbirth is relatively rare in Sweden compared with the USA and other countries.26 Therefore, generalizations from these results to other countries should be made with caution. Third, although the cousin-comparison partially controls for some aspects of the environment shared within extended families, the fact that there is no measured information on the extent to which the cousins share an environment should be considered as a limitation. Fourth, the CoS model assumed equal environment among all types of relative groups (except for mothers with paternal half-sisters). This means that MZ twins and their family would spend similar amounts of time together as the families of DZ twins and full sisters.39 Fifth, the model also assumed no assortative mating. In case of assortative mating, the genetic confounding might also be explained by the father’s genetic risk. As such, additional research is needed to explore this question using other study designs.
To conclude, the current study found that MAFB predicted offspring ADHD, and all offspring born to mothers who began childbearing early were at increased risk of ADHD. The association between young maternal age and offspring ADHD was mainly explained by genetic confounding. Teenage childbearing is internationally recognized as a public health issue with adverse consequences for both young mothers and their children.26 Although young maternal age itself does not cause offspring ADHD, it is an important risk marker for all children born to young mothers. Our results suggested that public policy initiatives should aim not only to promote later childbearing in the population, but also identify individual at-risk mothers and their children who may need support. Our findings are also likely to contribute to the understanding of the aetiology of ADHD. What is usually considered as an environmental risk factor (teenage childbearing) is likely a marker of genetic predisposition. Our study highlights the importance of using family-based designs when trying to understand how early life circumstances affect child development, and underscores the need for additional research on this topic to ensure that prevention efforts and public policy are evidence based.
Funding
This study was supported in part by the Swedish Research Council (2010-3184; 2011-2492), the Swedish Initiative for Research on Microdata in the Social And Medical Sciences (SIMSAM, 340-2013-5867), the Swedish Council for Working Life and Social Research (2006-1625) and the National Institute of Child Health and Human Development (HD061817).
Conflict of interest: Dr H. Larsson has served as a speaker for Eli Lilly & Co., and his association with Eli Lilly was not related to this publication in any way. All other authors declared no conflicts of interest.
References