Neurodevelopmental problems and extremes in BMI

Study population

The study population was recruited from the ongoing Child and Adolescent Twin Study in Sweden (CATSS), a longitudinal, nationwide study focusing on mental health problems during childhood and adolescence (Anckarsater et al., 2011). CATSS is based on telephone interviews with the parents of all 9-year-old twins in Sweden. (During the first three years of the study, parents of all 12-year-olds were also included (CATSS-9/12)). In the present study, we analyzed the data on twins born between 1992 and 1998 (n = 12,496) who participated in the CATSS-9/12. We excluded 407 children with congenital or early brain damage, known chromosomal syndromes, diabetes, cancer, celiac disease, or other defined medical conditions with potential weight effect, and another 734 (6% of those who participated in the study) because of missing data. The final study population thus consisted of 11,355 children, with close to equal distribution between boys and girls (51.9% and 48.1%, respectively) and 9- and 12-year-olds (47.6% and 52.4%, respectively). Distribution of zygosities in the study was 27% monozygotic (MZ) twins, 35% dizygotic (DZ)–same sex (ss), and 33% DZ–different sex; 5% of the children could not be typed by zygosity. Only twin pairs with complete data on weight, height, and symptoms of NPDs were included in calculating heritability estimates; 1518 MZ and 1900 DZss twin pairs were eligible for inclusion.

Measures

BMI

BMI is calculated as weight in kg/height in m². For children, the use of BMI as a tool for assessing body fat is more complex than for adults. BMI references for children are age- and gender-specific, and the cutoffs for underweight, overweight, and obesity are population-specific. Many countries have their own population references in addition to the well-known international references of the World Health Organization (WHO, 2007). In both clinical practice and research the Karlberg, Luo & Albertsson-Wikland (2001) BMI reference is used as the standard in Sweden. The Karlberg BMI reference was derived from an original study of 3650 healthy singletons born between 1973 and 1975 in the western part of Sweden. Their mean BMI ± 1, 2, and 3 standard deviations (SD) were calculated in addition to previous percentile charts (WHO, 2007) to create a better tool for clinical evaluation (Reilly & Kelly, 2011). Given the geographic origin and the age range of the present study cohort, only data representing the 9-year-old (n = 828 boys; n = 790 girls) and 12-year-old (n = 640 boys; 676 girls) children from the Karlberg study (Karlberg, Luo & Albertsson-Wikland, 2001) were used as BMI references for comparison with the CATSS twins.

In the present study, information about children’s weight and height was reported by their parents. To avoid bias, we constructed a new reference for BMI because (1) we had a much larger database than those used in the WHO and the Karlberg studies, (2) our sample represented twins, not singletons, and (3) our data include newer generations of children. BMI scores were transformed with the help of Box–Cox transformations to ensure normal distribution (for a detailed statistical description, see Appendix S1) and then converted to BMI standard deviation scores. In the present study, low extreme BMI was defined as more than 2SD below the mean, low BMI as more than 1SD but less than 2SD below the mean, normal BMI as within the range of minus one and plus 1SD from the mean, high BMI as more than 1SD but not more than 2SD above the mean, and, finally, high extreme BMI as more than 2SD above the mean (Fig. 1).

Medication

Overall, parents reported 17 different prescription drugs used by children either daily or during the month prior to the interview; only three of these (betamethasone, methylphenidate, and desmopressin) had known effects on weight. In the total study population less than 1% (n = 95) received medication with potential influence on weight, and over half of these (n = 48) did not screen positive for ADHD or ASD. Only one of the 81 children who screened positive for ASD, 20 of the 832 who screened positive for ADHD, and 26 of the 262 children who screened positive for ADHD and ASD combined had medication. Of the 95 children who were on potentially weight-modulating medication, 93% had a normal BMI, 3% had an extremely high BMI, and 4% had an extremely low BMI.

Prevalence of ADHD and ASD diagnoses according to BMI categories

There was a normal distribution of BMI in the comparison group, with 69% of the children having a normal BMI and close to equal distribution of the lower and higher ends of the scale (Fig. 2). The BMI distribution in the ADHD group was similar to the comparison group. However, the ASD group showed almost three times greater prevalence of the high extreme BMI (standardized residual = 2.3) and nearly twice as great a prevalence of low extreme BMI (standardized residual = 1.1), with only about half having a normal BMI (Fig. 2). The ADHD + ASD group showed a roughly two-fold higher prevalence in the high extreme category (standardized residual = 2.0), a significantly increased prevalence in the high category (standardized residual = 3.4), and an increased, but not significantly, prevalence in the low extreme BMI category (Fig. 2) over the comparison group. In general, the NDP groups had higher prevalences than the comparison group in all BMI categories other than normal (X² (12, N = 11,355) = 39.941, P < 0.00).

Association between extremes in BMI and NDPs

Both ADHD and ASD modules were weakly but significantly associated with high and high extreme BMI in the univariable models. In the multivariable model (including both ADHD and ASD as covariates) only the ASD module kept its significant positive association with these dependent variables (Table 2).

When measuring the association of the subdomains in univariable models, problems with attention from the ADHD domain and problems with social interaction and flexibility from the ASD domain increased the risk for high BMI. Only the social interaction subdomain kept its significant positive association with high BMI in a multivariable model. All three ASD subdomains were significantly associated with high extreme BMI in univariable models, while in the multivariable model only the subdomain measuring problems with flexibility retained a significant relation (Table 2). Among the ADHD subdomains, only inattention increased the risk of high extreme BMI in both univariable and multivariable models.

Neither ADHD nor ASD were associated with low BMI, but ASD, in a univariable model, had a weak positive association with low extreme BMI. The flexibility and language subdomains were associated with this risk, but only the flexibility subdomain retained its significant effect in a multivariable model.

Etiology of BMI

Intra-class correlations were measured in gender- and age-specific groups. They were stronger in MZ than in DZ twins (Table 3) in all groups. In the ACE models, the genetic effect was higher in 12-year-olds than in 9-year-olds in both girls and boys, and the effect of common environment was lower.

Twins’ BMI standard versus singletons’

Using the newly constructed Swedish BMI reference (specific to 9- and 12-year-old twins) for this study (Appendix S1), there were no significant differences in the mean BMI of boys or girls compared with Swedish (Karlberg, Luo & Albertsson-Wikland, 2001) or international (WHO, 2007)  BMI references (Fig. 3). However, in our BMI reference, the cutoffs at +1SD, +2SD, and +3SD were higher in all study groups (with the exception of 12–year-old girls (Fig. 3D)), while at −1SD, −2SD, and −3SD, BMI was consequently lower than these BMI values in the Swedish or international BMI references, providing stricter criteria for extremes.

Prevalence of extreme BMIs in children with ADHD and/or ASD

In the present study, children who scored above the diagnostic cutoff for ASD were overrepresented on both extreme ends of BMI. Three times as many children with ASD as children with no ASD or ADHD had high extreme BMIs, and twice as many had low extreme BMIs. BMI in the group of children with ADHD followed a close to normal distribution, and the coexistence of ADHD with ASD seemed less often associated with extreme BMIs than ASD on its own. Extremes in BMI have previously been observed in children with ASD identified as overweight (Chen et al., 2010; Rastam, 2008) and underweight (Hebebrand et al., 1997; Mouridsen, Rich & Isager, 2002). Selective eating has been shown to be more prevalent among children with ASD, but no definite association has been found between selective eating and alterations in BMI among children with ASD (Evans et al., 2012), although restrictive eating in Canadian children 5 to 12 years old was associated with low BMI in one study (Pinhas et al., 2011).

ADHD and ASD domains associated with extremes in BMI

The results of the present study show that the presence of ADHD is associated with a high extreme BMI, while ASD is associated with both the high and the low extremes of BMI.

While there are findings showing that impulsiveness in healthy adult females and males is associated with overeating (Davis et al., 2006; Strimas et al., 2008), an earlier study suggested that inattention is a reason for both low and high extreme weights (Cortese & Vincenzi, 2012; Davis et al., 2006). Our results may support the earlier study, that is, that the ADHD subdomain of inattention rather than the subdomain of impulsiveness/hyperactivity is associated with an increase in BMI. While ADHD scores were associated with a high extreme BMI, ASD scores were associated with both low and high extreme BMIs. Even in the presence of ADHD, the effect of ASD remained a significant risk factor for extremes of BMI in both directions. According to earlier studies, overweight and obesity are common in ASD, though probably not in children before the age of 7 (Emond et al., 2010), but increasingly so from school-age on (Chen et al., 2010). Underweight has also been reported in Danish and German adolescent boys with ASD (Hebebrand et al., 1997; Mouridsen, Rich & Isager, 2008; Sobanski et al., 1999).

Etiology of BMI

As previously shown, the variance in children’s BMI is largely influenced by genetic factors (Dubois et al., 2012). However, in younger children there is also a significant effect of shared environment (most easily translated as the family influence on eating behaviors, reviewed by Birch & Fisher in 1998). The present study and a recent review on the heritability of BMI (Elks et al., 2012) suggest that the shared environmental effects during childhood decrease with age, while genetic predispositions have a larger effect later in life.

Because ASD was overrepresented on both ends of the BMI categories, it was not possible to conduct bivariate analyses of NPDs and BMI. The interpretation of such a model would be difficult and potentially misleading since ASD in the high end of BMI might be a different construct than ASD in the low end. Even with the relatively large number of subjects included in this study, tetrachoric correlations between ASD/ADHD and high and low BMIs could not be performed because too few individuals fulfilled each category to allow any conclusion to be drawn. Therefore, the present study did not analyze the importance of common genetic and environmental factors in the expression of NDPs and extreme BMI.

Twin BMI standard

Because twins are generally smaller than singletons at birth, special twin BMI references are recommended for following them during their first years of life (Estourgie-van Burk et al., 2010; Van Dommelen et al., 2008). The information on BMI in twins older than 3 years old compared with singletons is contradictory. While some studies show that twins’ BMI may still differ from singletons’ at the age of 17 (Pietilainen et al., 1999), others have found no discrepancies in BMI between twins and singletons (Wilson, 1979; Estourgie-van Burk et al., 2010).

Our study confirmed that in pre-pubertal children, the BMI of twin samples could be considered to be representative of the general population of children. However, we found discrepancies between the twins’ and national and international BMI standards for extremes in under- and overweight. This could be explained by the large sample size, which could increase the visibility of those more rare extremes. Another explanation could be that this is an expression of time trends in the BMI development of children. A third explanation would be that although mean BMIs in twins seem to be similar to those of singletons, twins may be overrepresented at the extreme ends of the BMI curve. These findings emphasize the need for a more updated and relevant BMI reference when assessing BMI deviations among twins in research and clinical practice.