The association between ADHD and physical health: a co-twin control study

Attention-deficit/hyperactivity disorder (ADHD) has been associated with increased risk for physical comorbidity. This study used a twin cohort to investigate the association between physical diseases and phenotypic variations of ADHD. A twin cohort enriched for ADHD and other neurodevelopmental conditions were analysed. The Attention Problems subscale of the Child Behavior Checklist/Adult Behavior Checklist (CBCL/ABCL-AP) was used to measure the participants’ severity of ADHD symptoms. Physical health issues were obtained with a validated questionnaire and were tested in relation to ADHD symptom severity in a co-twin control model. Neurological problems were significantly associated with a diagnosis of ADHD. A conditional model for the analysis of within-twin pair effects revealed an inverse association between digestive problems and the severity of ADHD symptoms, after adjusting for co-existing autism spectrum disorder and ADHD medications. Our findings suggest that individuals with ADHD are susceptible to neurological problems, why a thorough neurological check-up is indicated in clinical practice for this population. In addition, health conditions of digestive system could be considered as a non-shared environmental factor for behavioral phenotypes in ADHD. It supports the possible role of gut-brain axis in the underpinnings of ADHD symptoms, at least for a subgroup of individuals with certain genetic predisposition.

psychodiagnostic assessment administered by experienced clinicians in the RATSS team 30 . Clinical diagnoses were based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) 1 , endorsed by results from standardized diagnostic instruments, including the Kiddie Schedule for Affective Disorders and Schizophrenia-Present and Lifetime Version (K-SADS-PL) 34 , and the Diagnostic Interview for ADHD in Adults (DIVA 2.0) 35 , the Autism Diagnostic Interview-Revised (ADI-R) 36 , and the Autism Diagnostic Observation Schedule Second Edition (ADOS-2) 37 . Full-scale IQ was estimated with the Wechsler Intelligence Scales for Children or Adults, Fourth Editions 38,39 . Dimensional ADHD symptoms were quantified with the CBCL/ ABCL-AP, consisting of 10/17 items, assessing behaviour symptoms related to attention problems within the past 6 months. The CBCL (6-to 18-year-olds version) 40 was completed by the parents, and the ABCL (for ages   41 was completed by the participant's next of kin (mostly parents or spouse). Both the CBCL and ABCL are broadband screeners and part of the Achenbach System of Empirically Based Assessment (ASEBA), a family of tools evaluating problem behaviours and competencies from a wide range of perspectives, including internalizing and externalizing symptoms, as well as social problems and somatic complaints. Items on CBCL/ABCL are rated on a 3-point Likert scale (ranging from 0 to 2), with higher scores indicating more severe symptoms.
Medical history and present physical comorbidity. A parent-/self-report questionnaire validated against medical registry data and designed to obtain information on medical history and present physical health issues is used in RATSS 42 . The questionnaire consists of one open question "Has the child (Have you) been seriously ill during his/her (your) childhood?" and 33 closed questions inquiring about whether the participant had ever had a specific physical health diagnosis or problem. The full list of physical conditions and the distribution of these physical health issues in our sample are summarized in Table 2, except for cardiovascular diseases, for which too few cases were reported to be included in the analysis (four in ADHD, seven in non-ADHD). The comorbid physical problems were categorized into groups based on different physical systems/etiologies (history of infectious diseases, neurological problems, gastrointestinal problems, and immune dysregulation). Different conditions in the same group were summed up to generate a predictive estimate (as a variable indicating the frequency of problems in each group) for clinical ADHD diagnosis or the severity of ADHD symptoms. The reasons to use this variable were: (1) our sample might not have enough power to detect the difference between ADHD and non-ADHD for each physical illness (2) we sought to examine the prognostic power of different levels of physical morbidity severity within each physical system on ADHD both categorically and dimensionally. Among these conditions, two of the items "migraine" and "headache" were combined into one condition "headache" and coded as one physical health issue since these two conditions were clinically highly correlated.
Statistical analysis. All statistical analyses were performed with IBM SPSS software version 25 (SPSS Inc., Chicago, IL, USA), the drgee package 43 and BayesFactor package 44 in R version 3.5.1. Student's t test was used to compare the frequency of physical comorbidity between the individuals with ADHD and without ADHD in the whole sample. To examine if the amount of physical problems differed within twins discordant for ADHD diagnosis or ADHD symptoms, Wilcoxon sign-rank test was used in consideration of the sample size of those twin pairs. Owing to the overall exploratory nature of the research approach, Bayes factors (alternative hypothesis [H 1 ]: null hypothesis [H 0 ]) were also calculated for these comparisons. For the co-occurring physical health conditions which were identified with significantly higher frequency in ADHD in the whole sample, the association with ADHD phenotypes (both categorical and dimensional) was tested. Conditional multivariate logistic and   Other GI problems, n (%) 4 (3.9) www.nature.com/scientificreports/ linear regression analysis with twin pairs clustered was used to explore the adjusted associations between physical health and ADHD diagnosis as well as dimensional ADHD symptoms, after adjustment for potential confounding variables. To determine the within pair effect of physical comorbidity on the variation of ADHD symptoms among quantitatively differing twins, we used conditional generalized estimating equations (CGEE), a multiply adjusted (conditional) linear regression model, to eliminate the influence of pair-consistent confounders 43 . Child (< 18 years) and adult (≥ 18 years) participants were analysed separately in addition to the whole sample which included participants of all ages, except for the adult discordant twins due to limited sample sizes. All tests were two-tailed and p-values of 0.05 or less were considered statistically significant. A Bonferroni correction was applied for multiple comparisons in all the analyses (p < 0.0125 in Tables 2 and 4; p < 0.05 in Table 3).

Results
Comparisons of physical problems between ADHD and non-ADHD. In the total twin sample of all ages, participants with ADHD diagnosis had significantly more neurological problems (t = 2.55, p = 0.012) compared to those without ADHD ( Table 2 [a]). However, when dividing the sample into children group and adult group, the differences of the frequency of neurological problems between ADHD and non-ADHD became nonsignificant in both groups. Still, there was a trend that children participants with ADHD had more neurological health issues than controls (t = 2.28, p = 0.024, larger than 0.0125, the adjusted value for statistical significance here). For infectious diseases, gastrointestinal problems, and immunological problems, there was no difference found between participants diagnosed with ADHD and non-ADHD twins. Comparisons within pairs of MZ and DZ twins discordant for ADHD diagnosis and dimensional ADHD symptoms revealed no difference of coexisting physical problems between ADHD twins and co-twins (  Within-pair effect of physical comorbidity on ADHD symptoms for dimensionally discordant twin pairs. In the conditional logistic model, within-twin pair increases in GI problems were associated with decreases in CBCL-AP subscale scores in MZ children twins quantitatively differing for ADHD after adjusting for ADHD medication (β = − 2.72, p = 0.001, Table 4), but not for all-age group of MZ twins and not for DZ  www.nature.com/scientificreports/ twins. There was no significant within-pair association with ADHD symptoms found for infectious diseases, neurological problems, and immunological problems in all age groups of MZ and DZ twins.

Discussion
This is the first study to investigate the association between co-existing physical problems and ADHD using a well characterized twin sample enriched for NDDs. In addition, we examined the role of physical health issues on the severity of ADHD symptoms in twins quantitatively differing for ADHD. Our results revealed that neurological problems among children were associated with the diagnosis of ADHD. However, for MZ twins with differing dimensional ADHD symptoms, GI problems showed protective within pair effects, even after adjusting for ADHD medications. Our findings support that ADHD is a neurodevelopmental condition, in which the complex underpinning of altered brain development might not only affect multiple domains of cognitive and other behavioural function, but also increase the susceptibility of neurological health issues 45−47 . Moreover, childhood GI tract health could be considered as a non-shared environmental factor associated with severity of ADHD symptoms, which might exert influence via the interplay with a specific genetic background. More research is warranted to disentangle the mechanisms contributing to the overlap between ADHD and neurological complications, as well as the role of gut-brain axis on the phenotypic variation of ADHD 20 .
We examined the differences in frequency of physical problems between ADHD and non-ADHD individuals within our sample and the results varied for different physical systems. In line with the previous literature 22, 48 , individuals with ADHD showed higher rates of neurological health issues, despite the fact that the result disappeared when data was analysed separately in children and adults. We did not observe differences between ADHD and non-ADHD in terms of infection history, co-existing GI problems, and comorbid immunological diseases, in contrast to the results of prior epidemiological studies and also meta-analyses which synthesized data of asthma, atopic diseases, and allergic diseases in ADHD 10,11,22,49−51 . Still, these discrepancies may predominantly reflect the relatively small effect sizes of the differences between ADHD and the general population in general, as well as the heterogeneity of ADHD individuals from aetiology to clinical profiles.
The reported odds ratios of neurological, immunological, infectious and digestive problems in ADHD compared to non-ADHD are probably low and less than 1.5 22,50 . Therefore, it is likely that large samples are required to detect such modest differences of physical comorbidity between individuals with and without ADHD. Moreover, although the prevalence of physical comorbidity in ADHD has been reported to be increased, there is still information lacking regarding whether and how much of these co-occurring physical problems account for the variation of ADHD symptoms 52 . Also, these physical comorbidities are only present in a minority of the ADHD population, rather than accompanying ADHD symptoms consistently. In view of the heterogeneity in aetiology and possible biological pathways of ADHD 14 , the mechanisms underlying the overlap of ADHD and physical health issues are likely to be various across individuals. Thus, the results of prevalence studies investigating physical health in ADHD could be more consistent in population-based samples and varying more among clinical samples or samples that do not exemplify the full spectrum of traits in the target population 10,11,49 . Finally, our sample included both children and adults. However, ADHD is a developmental condition with changing symptomatology over time, which might reflect the dynamic influences of genetics, surroundings, psychosocial factors, and the maturation progress of brain function 25,53−55 . Individuals with ADHD whose symptoms continue to meet the diagnostic criteria when they move into adulthood, only comprise about 50% of the childhood ADHD population 56 . Research revealed that adult ADHD, or the non-remitters, may have distinct risk gene variants 57 . In addition, the psychiatric comorbidity pattern of ADHD also changes throughout the lifespan, such as the increasing rates of personality disorders and bipolar disorders in adulthood 58 . Hence, adult ADHD might be a specific subgroup in terms of genetic basis and developmental trajectory, and could also exhibit different profiles of comorbid physical diseases 59 . For instance, the prevalence of obesity in adult ADHD was found to be more than twice as high as the one for childhood ADHD (28.2% vs 10.3%) 12 . More studies on physical comorbidity in adult ADHD are needed to further enhance our understanding of whether the association between ADHD and physical health issues persists into adulthood.
Our results showed that neurological problems are associated with ADHD diagnosis in childhood. Although alterations in attentional capacity can be secondary to frequent seizure attacks, headache episodes, and antiepileptic medication 60−64 , the found association might suggest that children with ADHD are more vulnerable to neurological health issues subsequently, as prospective associations found in previous studies indicate 7,46,65 . Mechanisms contributing to the higher comorbidity rate between ADHD and specific neurological problems have been proposed. From a biological perspective, predisposing genes 66,67 , disturbances of the norepinephrine and dopamine systems 47,68 , and altered brain functional networks 69 have been postulated to be involved in the co-occurrence of ADHD, epilepsy, and migraine. The association between ADHD and headaches could be mediated by other psychiatric problems, such as sleep disorders and affective disorders 47,70 . Moreover, headache has been commonly reported as one of more frequent side effects of ADHD medications 71 .
Given the possible disabling consequences of neurological complications, it is imperative for practitioners working with ADHD to be attentive and to provide adequate management for the comorbid nervous system conditions in this population. Likewise, ADHD should also be screened for in children with neurological problems for early identification and intervention 18 . Regarding the association between neurological conditions and the severity of ADHD symptoms, we did not find a significant association. Since the individuals with ADHD in our sample were not drug-naïve, possible the treatment and side effects of ADHD medications could not be ruled out as a moderator influencing the outcomes. Further investigations with medication naïve subjects would be helpful to explore the impact of neurological health issues on ADHD symptoms. www.nature.com/scientificreports/ Based on our analysis of within pair effects of physical comorbidity, GI system health could be considered a non-shared environmental factor to the severity of ADHD symptoms among monozygotic twin children. Among the digestive problems presented in the twins with less severe ADHD symptoms, lactose intolerance accounted for the majority of those (7/12, 58.3%). The others were irritable bowel syndrome, abdominal pain, intestinal polyps, diarrhea, and gastroenteritis. We speculate that those children with GI tract health issues were more likely to have diet adjustment to avoid food possibly inducing GI symptoms 72,73 . Their caregivers and relatives might also pay more attention to their diet preparation, including nutritional balance and elimination of food with artificial additives 74 . The diet alteration may change the gut microbiota 75 , which had been proposed to link with the potential pathophysiology of ADHD symptoms through vagus nerve, neuro-metabolites, and neuroinflammation pathways 20 . In addition, the association between possible dietary change and improvement of ADHD symptoms might reflect the effectiveness of diet intervention in previous double-blind placebo-controlled trials in children with ADHD 76 . However, our results did not demonstrate similar effects of existence of digestive problems among dizygotic twins, whose genetic makeup differs. This may indicate that the contribution of GI tract health to the variation of ADHD symptoms could be synergistic under certain circumstances, such as a genetic predisposition towards ADHD. This is consistent with the findings from trials of diet treatment, in that only a subgroup of children with ADHD responded to the administration of dietary change 76 . Therefore, it is recommended for future research to focus on the predictors of recognizing those children who would benefit from diet intervention, as well as the mechanisms underlying the association of diet, digestive problems, and ADHD symptoms. Still, other alternative explanations for the inverse association between GI problems and ADHD symptoms are needed to be considered. For instance, children's gastrointestinal illness might increase parents' tolerance for their children's behavioral problems.
There are several limitations to this study that need to be addressed. First, although mainly selected from a population-based study 30,31 , some twins participating in RATSS were also recruited from via other sources and not sampled randomly, why the overall limited representativeness of our study sample and the generalizability of the results must be kept in mind regarding the frequency of physical problems. Moreover, our findings from a twin sample should be interpreted with caution when extending to singleton samples. Twin pregnancy has increased risk for perinatal morbidity 77 , which is associated with neurodevelopmental conditions and neurological complications 3,78 . Second, the sample size of adult ADHD in our study was limited, making results for association between physical comorbidity and ADHD among children and adults harder to compare. In addition, we did not have enough pairs of twins discordant for ADHD diagnosis to explore the within pair effect of physical problems on the clinical phenotypes of ADHD. Third, the CBCL/ABCL-AP is not a symptom scale derived from DSM-5, making analysis for categorical ADHD and dimensional ADHD symptom less comparable. Despite the validity of CBCL/ABCL-AP for identifying ADHD 79 , instruments designed for quantifying ADHD symptoms such as the Conners Rating Scale-Revised (CRS-R) 80 , the Swanson, Nolan, and Pelham Questionnaire (SNAP-IV) 81 , and adult ADHD self-report scale (ASRS) 82 should also be considered in future studies to measure the core symptoms of ADHD more directly. Fourth, information on physical comorbidity and medical history of infectious diseases was reported by either parents of children with ADHD or adult participants with ADHD via questionnaires. Thus a risk of reporting and recall bias which might lead to an overestimation of the found associations cannot be ruled out. Fifth, the contribution of each physical problem to the association with ADHD symptoms may not be equal, which means the effect size of each problem in the association analysis could be varying. Therefore, the results of our study might be limited with the unweighted approach. Still, the grouping of physical problems might not be accurate either with regard to underlying mechanisms. Sixth, we were unable to confirm whether twins with more digestive problems compared to their co-twin were administered diet adjustments by their parents or not. Further investigations are needed to clarify the association between dietary change and the variation of ADHD symptoms among twins dimensionally differing for ADHD. Finally, our participants with ADHD were not free from treatment. Hence, our results could also be limited in view of the effects of ADHD medication or non-pharmacological interventions.
In conclusion, controlling for the contribution of complex genetics and other common confounders, our findings suggest that health conditions of digestive system are associated with ADHD symptom presentation among twins, and thus form a non-shared environmental factor for behavioural phenotypes in ADHD. Hence, our results support that the gut-brain axis might play some role in the underpinnings of ADHD symptoms, at least for a subgroup of individuals with certain genetic predisposition. In non-responders and those intolerant to ADHD medications, it could be of clinically valuable to identify those individuals with ADHD who might benefit from diet treatment, which has been believed to alter the gut microbiota composition 75 . In addition, we found that neurological problems are associated with ADHD diagnosis among children, in line with previous prevalence studies. Since the emergence of neurological conditions could be either prior to or subsequent to ADHD, it is recommended for clinicians to be aware of the higher rate of comorbidity, and to provide adequate assessment and intervention to improve both physical and psychosocial outcomes of children with ADHD and neurological health issues. www.nature.com/scientificreports/