Early environmental predictors for attention-deficit hyperactivity disorder (ADHD), autism spectrum disorder (ASD) and their co-occurrence: The prospective ABIS-Study

ADHD and ASD are highly heritable and show a high co-occurrence and persistence into adulthood. This study aimed to identify pre and perinatal risk factors, and early psychosocial exposures related to later diagnosis of ADHD, ASD, and their co-occurrence. 16,365 children born 1997–1999 and their families, involved in the prospective population-based ABIS study (All Babies in Southeast Sweden), were included in this sub-study. Pre and perinatal factors and early environmental psychosocial exposures were collected from parental-questionnaires at birth and 1-year follow-up. Diagnoses from birth up to 23 years of age were obtained from the Swedish National Diagnosis Register in 2020. The cumulative incidence of ADHD, ASD, and their co-occurrence in the ABIS-cohort Study were 4.6%, 1.7%, and 1.1%, respectively. Being male was associated with an increased risk for ADHD, ASD, and their co-occurrence (aOR 1.30, 1.56, and 1.91, respectively), while higher household income reduced it (aOR 0.82, 0.73, and 0.64). Serious life events during pregnancy (aOR 1.40) and maternal smoking (aOR 1.51) increased the risk of ADHD, while older maternal age (aOR 0.96), higher parental education (aOR 0.72 maternal and aOR 0.74 paternal) and longer exclusive breastfeeding (aOR 0.72) reduced it. Non-Swedish paternal nationality (aOR 0.40) and higher maternal education (aOR 0.74) were associated with a lower risk of ASD, while a family history of autoimmune diseases increased the risk of the co-occurrence of both disorders (aOR 1.62). Obtained results suggest that the etiology of ADHD, ASD, and their co-occurrence is independently associated with environmental psychosocial predictors. The co-occurrence seems to overlap the etiology of ADHD, in which psychosocial determinants have a larger role, however, it is also independently influenced by a family history of autoimmune diseases.


Predictors
Data was collected from the questionnaires used in the ABIS-Study at birth and 1-year follow-up.

Pre and perinatal
Maternal disease during pregnancy was a composite measure including hypothyroidism and hyperthyroidism, B12 deficiency, systemic lupus erythematosus, adrenal insufficiency, type 1 and type 2 diabetes, gestational diabetes, celiac disease, rheumatism, and inflammatory bowel disease.Autoimmune disease heredity was dichotomized as no autoimmune heredity or at least one first-degree relative with any of the autoimmune diseases previously mentioned 22 .Infections during pregnancy, and iron and vitamins/minerals supplements use during pregnancy, were dichotomic variables.Parity was categorized in "first parity" or "previous parity", twins (single fetus or multiple fetuses), way of delivery (vaginal, cesarean section, or other complications), gestational age, and birth weight, were also included.

Psychosocial
Child's sex, maternal and paternal age at child's birth, maternal and paternal ethnicity (born in Sweden or abroad), civil status at child's birth (married/living with a partner or single parent), support during pregnancy, and security for mother and child were also included.Serious life events during pregnancy and at 1 year of age 23 , maternal smoking, and tobacco exposure at 1 year of age, were also included.Exclusive breastfeeding duration was categorized as "less than 4 months", "5-8 months", and "9 months or more", according to WHO recommendations 24 .Maternal and paternal education was classified according to the International Standard Classification of Education (ISCED) and graded in three levels: low = ISCED level I-II, medium = ISCED level III-IV, and high = ISCED level V-VII 25 .The disposable household income for the year 2000 was obtained from the Swedish Income and Tax register and was categorized into three levels based on percentage distribution, low income (bottom quintile), medium income (second to fourth quintile), and high income (top quintile) 26 .

Statistical methods
All statistical analyses were performed in SPSS software version 28.0 (IBM SPSS Inc., Chicago, IL, USA).Dichotomous predictors were presented as frequencies and percentages, and differences were assessed using Chi-squared test, while quantitative predictors were described using mean and standard deviation, and differences between groups were assessed using t-test.A P-value ≤ 0.05 was considered statistically significant, and multiple comparisons between the groups with any assessed outcome were adjusted using Bonferroni correction (Table 1).Missing values were imputed via multiple imputation in which all predictor and outcome variables were employed in the imputation process which used 5 imputation samples 27 .All reported univariate and multiple multinomial logistic regression analyses were fitted to the data after multiple imputation.Identification of statistically independent discriminators used a backward elimination algorithm in which all univariately statistically significant discriminators (unadjusted model-Table 2) were entered into a single full model.Individual discriminators were then removed from the model in a stepwise fashion if their individual P-value was > 0.05 until all remaining discriminators were statistically significant (Multivariable model of independent predictors-Table 2).Automatic model selection algorithms are known to be prone to over-optimism 28 and hence we regard the final model selected as a candidate model which is in need of independent replication.Effect sizes were reported as odds ratios (OR) or adjusted odds ratios (aOR) with 95% confidence intervals (95% CI) and 2-tailed P-values.We conducted a response-bias analysis.

Statistically independent predictors of ASD
Being male, lower household income, and lower maternal education together with paternal Swedish nationality, were the risk factors that remain significant in the multiple multinomial logistic regression analyses (Table 2).

ADHD and ASD co-occurrence predictors
The co-occurrence group was characterized by a higher prevalence of autoimmune diseases in the family (P < 0.001), and a higher proportion of preterm babies (P = 0.027).Regarding psychosocial factors, this group was predominantly associated with low parental education level (P < 0.001), single parental status (P < 0.001), low household income (P < 0.001), lack of support during pregnancy (P = 0.002), and lack of security for mother and child (P < 0.001).This group also reported maternal smoking (P < 0.001) and serious life events at 1 year of age (P = 0.002) to a greater extent than the control group (Table 1).

Statistically independent predictors of the co-occurrence
Being male, lower household income, together with a family history of autoimmune disorders remain significant in the multiple multinomial logistic regression analyses (Table 2).

Discussion
This is the first-ever prospective study considering early environmental psychosocial exposures as potential etiological factors not only for ADHD and ASD, but also for their co-occurrence.
Being male was a strong and independent predictor for later development of ADHD, ASD and their cooccurrence.Previous studies have found a male-to-female ratio of 3:1 for both ADHD 29 and ASD 30 while we found a ratio of 2:1.Genetic and endocrine causes have been proposed in previous studies, however, it is also possible that, due to cultural or social expectations, females may report their ADHD symptoms less frequently or as less disabling than males, thus not meeting the diagnostic criteria 31 .
Lower household income level was also independently associated with an increased risk of ADHD, ASD and their co-occurrence.Previous studies conducted in ABIS cohort and in other high-income countries, have shown marked socioeconomic inequality associated with several child health outcomes (T1D, cardiovascular risk factor and infectious diseases) 26,32,33 .On this line, previous studies found that children in families with low SES were twice as likely to have ADHD than children in high SES families 34  www.nature.com/scientificreports/as a background factor, could also influence both their own SES and ADHD development in their offspring 34 .
According to previous studies 35 , we found that younger mothers increased the risk for ADHD in their offspring, that together with a lower education and income level could evidence socio-economic and emotional difficulties in child-rearing 36 .SES was also related to breastfeeding, as it was shown that mothers with low education levels wean earlier 37 .
In our study, longer breastfeeding was associated with lower rates of ADHD, which is in line with previous studies 15 .Breastfeeding could have a protective effect since it facilitates the intimate contact between mother and child.Furthermore, breastfeeding provides long-chain polyunsaturated fatty acids (omega 3 fatty acids, PUFA) that could affect the infant's microbiota regulating the gut-brain-axis, important for brain development 38 .Some studies showed decreased risk for ASD also, which we could not confirm 15 .We found a positive association between smoking (at pregnancy and 1 year of age) and ADHD.Smoking per se could be a biological risk factor for ADHD, but could also indicate the exposition to a vulnerable psycho-social environment.Previous studies also reported this association 9 .Some reports have also shown that maternal stress during pregnancy is a risk factor highly associated with ASD and ADHD 8 .We found an association between serious life events during pregnancy and ADHD risk, but not ASD.Many studies have shown that prenatal exposure to maternal stress was associated with abnormalities in neurodevelopment, neurocognitive function, and cerebral processing, which lead to changes in both the hypothalamic-pituitary-adrenal axis (HPA) and the autonomic nervous system 8,39  Table 1.Distribution of potential predictors of ADHD, ASD, and their co-occurrence among the groups with any outcome and the one with none.ADHD attention-deficit/hyperactivity disorder, ASD autism spectrum disorder, AD autoimmune disease.Values are presented as mean (SD) or as absolute numbers (percentages).# Continuous variables categorized only for descriptive purposes.All P-values were calculated for each group with any outcome against the one without any assessed outcome from Chi-squared test and t-test.The multiple testing correction threshold was P < 0.016.The last column indicates the statistically significant differences found between the groups with any outcome (a: ADHD, b: ASD, c: co-occurrence), and its magnitude.Significant values are in bold.
The underlying genetic and biological component of ASD seems to differ from the etiology of ADHD.The role of infections and the immune system in the etiology of ASD has been widely debated 40 .We consistently found that infections during pregnancy increased the risk of ASD.Accumulating evidence suggests that the immune system and abnormal immune function, including inflammation, cytokine dysregulation, and anti-brain autoantibodies, influence the trajectory of ASD 10 .Intriguingly, we found that the children of mothers with autoimmune diseases have a higher risk for ASD.Maternal hypertension, anemia, overweight, diabetes, and several other medical conditions have been also associated with an increased risk of ASD in offspring 41,42 .A meta-analysis revelated that delivery complications and cesarean section are risk factors for ASD 43 , but when differentiating between ASD and the co-occurrence group, we found that cesarean delivery seems to be a protective factor for ASD.This might be explained by avoidance of brain damage and/or hypoxia of the new-born which might otherwise have happened in a vaginal delivery.A recent systematic review, evidenced, after adjustment for familial confounding, that perinatal hypoxia and respiratory stress were consistently associated with ASD 11 .Regarding the role of genetics, a study on a large cohort of Swedish children showed that the heritability of ASD is approximately 50% and the risk of ASD increases tenfold if a sibling has the disease and twofold if a cousin is diagnosed with ASD 44 .Regarding the ADHD/ASD co-occurrence seems to overlap the etiological pattern of ADHD, while the possible biological etiology of ASD tends to be obscured.There is evidence suggesting that autistic traits are common among children with ADHD, being ADHD a better predictor of the co-occurrence, rather than ASD 45 .A family history of autoimmune diseases was found to be a strong and independent factor for the co-occurrence (not each disorder separately), which may suggest that both disorders could share similar genetic or environmental background.Common immune-mediated diseases such as asthma and eczema have repeatedly been linked to ADHD [46][47][48] , similarly, ADHD has been related to autoimmune diseases, such as celiac disease, ulcerative colitis, psoriasis, ankylosing spondylitis and T1D 47,49 .It was suggested that the co-occurrence is associated with greater impairment, increased severity of psychosocial problems, and may be less responsive to standard treatments for either disorder 20 .

Strengths and limitations
Although our results are based on a large prospective birth cohort from the general population with a follow-up for more than 20 years and the strength of merging doctor-set diagnoses of ADHD and ASD via the National Diagnosis Register, our study has some limitations.The family history of ADHD and ASD diagnoses was not possible to obtain.The lack of explicit criteria for ADHD and ASD co-occurrence could lead to variability in diagnosis and may have contributed to under-recognition or misdiagnosis of this comorbidity.Most data are based on questionnaires answered by the parents, usually the mothers, therefore there is a risk of recall bias.We have validated several data against information registered in the journals of Well Baby Clinics with good agreement, and there is no reason to believe that such registers have a systematic misdistribution that would influence our results.There are dropouts from birth to one year of age, but the remaining ABIS population is still representative of the general population.Thus, in summary, we believe that our results are robust and representative of the Swedish population.

Conclusion
Our study includes pre and perinatal, and early environmental psychosocial exposures as etiological factors of ADHD, ASD, and their co-occurrence in a long-term prospective follow-up of a general population-based birth cohort.Observed associations suggest a genetic and biological component underlying ASD, and a larger role of environmental psychosocial factors in ADHD etiology.The co-occurrence seems to overlap the etiology of ADHD but is also influenced by a family history of autoimmune diseases.This study shed light on the factors that may confer risk for the expression and/or diagnosis of ADHD, ASD and their co-occurrence and despite these factors are not necessarily causal, may constitute important incentives for preventive measures in child health.

Figure 1 .
Figure 1.Study population flow-chart.Definition of groups based on the cumulative incidence rates for ADHD, ASD, and their co-occurrence from birth until the age of 23 years.ADHD attention-deficit/hyperactivity disorder, ASD autism spectrum disorder.

Table 2 .
Multinomial logistic regression analysis, effect sizes represented as odds ratios.Groups with any outcome compared to the one with none.ADHD attention-deficit/hyperactivity disorder, ASD autism spectrum disorder, AD autoimmune disease, OR odds ratio, aOR adjusted odds ratio.a Unadjusted independent variables for ADHD (n = 755), ASD (n = 272), the co-occurrence of both disorders (n = 188), and 15.150 subjects without any assessed outcome.b Multivariable model of independent predictors.# Vaginal delivery acts as the reference category.*P < 0.05, **P < 0.01 and ***P < 0.001.