Minor physical anomalies in neurodevelopmental disorders: a twin study

Participants were part of the Roots of Autism and ADHD Twin Study in Sweden, described elsewhere in detail [14], from August 2011 to November 2013. They were recruited through the Child and Adolescent Twin Study in Sweden via advertisements in journals of national interest organization, referrals from clinical units (e.g., child psychiatry, habilitation centers), and the Swedish patient registry. A total of 116 twins, representing 51 MZ pairs and seven DZ pairs underwent diagnostic, behavioral, and MPA assessments. The study was approved by the Regional Swedish Ethical Review Board.

Participants were either TD or diagnosed as having a NDD or for some, more than one NDD (indicating co-morbidity of diagnoses), using a consensus process with several experienced clinicians according to DSM-5 criteria corroborated by information from standardized instruments: Autism Diagnostic Observation Schedule (ADOS-2); Autism Diagnostic Interview-Revised (ADI-R); Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS); and Diagnostic Interview for ADHD in Adults (DIVA 2.0) (see Additional file 1: Table S1 for instrument details). IQ testing was performed with the following measures: Wechsler Adult Intelligence Scale-IV (WAIS-IV); Wechsler Intelligence Scale for Children-IV (WISC-IV); Leiter International Performance Scale-Revised; and the Peabody Picture Vocabulary Test (PPVT-4). Autistic traits were measured with the Social Responsiveness Scale-2 (SRS-2). As recommended for research settings, the SRS-2 total raw score was used, with increasing scores (0–195) indicating more severe traits. The twin pairs were categorized as either NDD-concordant (i.e., both twins meeting criteria for NDD diagnosis), NDD-discordant (i.e., only one twin in pair meeting criteria for NDD diagnosis), or TD (i.e., neither twin meeting criteria for NDD diagnosis). Structural magnetic resonance imaging (sMRI) was conducted to identify any gross brain alterations. Blood and saliva were collected to confirm zygosity. Individuals reported to have known genetic syndromes (e.g., Down) were excluded from the study to prevent confounding variables, similar to previous research [4, 7].

A checklist (Additional file 2: Table S2) containing a total of 179 anomalies for males (24 body regions) and 171 for females (23 body regions) was used in the study. The checklist was developed and utilized by two experienced clinical geneticists (authors BMA, AN) based on the London Dysmorphology Database, Elements of Morphology [15], and long-standing clinical expertise. Standardized terms for MPAs are based on the Human Malformation Terminology from the Elements of Morphology [15]. The checklist includes MPAs, as well as anomalies that are commonly assessed for in physical exams (e.g., overweight, vision impairment, etc.). Vision impairments included myopia, hyperopia, colorblindness, astigmatism, etc. and 35% of participants with impairments had corrective lenses. Common phenotypic variants from the original checklist were removed if they were present in more than 4% of the European population [4, 16]. The MPA assessments took approximately 1 h to complete for each pair. The assessments were completed by either one or both clinical geneticists, with the majority completed by both. Any disagreements related to the presence of MPAs were resolved on the spot. Participants received a score of “1” for the presence of every MPA on the checklist, as well as a score of “1” for any clinically relevant findings on the sMRI (as read by a radiologist; 93% had data). Length and weight were measured by a research nurse and corrections were made to the overweight and underweight items on the checklist for each participant based on their actual body mass index (BMI) for individuals 20 years of age or older and BMI-for-age for individuals less than 20 years old. Additionally, a physician trained in dysmorphology assessed photographs of 28 twin pairs from the study to establish interrater agreement on all items mutually visible on both the in-person MPA assessment and through photographs (90 items).

Only pairs with full MPA assessments were included in the analyses (n = 58 pairs) to allow for statistical procedures making comparison within twin pairs. Two pairs of twins were removed a priori due to one twin with TD and another twin with ASD not consenting to the MPA assessment. Medians and interquartile ranges are reported for MPA scores by diagnosis and concordance type and means and standard deviations are used for demographic data related to the participants. To adjust for the clustering in pairs, a conditional regression model was fitted using generalized estimating equation (GEE) analyses [17] for assessments of the association between the number of MPAs and (i) all NDD diagnoses (including ASD and ADHD), (ii) ASD only, and (iii) ADHD only. The number of MPAs was the independent variable and the diagnostic category or traits were the dependent variables. Two models were fitted: first, a linear regression model for estimates of associations in the cohort with clustered standard errors accounting for the twin correlation, and next, a conditional regression model for estimates of association within-pairs after adjusting for factors shared within twins like genetics and environment. The cohort linear model assesses for associations between variables, but accounts for the dependence of participants as they are twins and have more in common with each other than other samples. The within-pairs conditional regression model estimates the association among variables of interest using the differences in outcome values (in this case, the MPA scores) in monozygotic pairs, to account for their genetic relatedness and other shared factors. Previous publications recommend controlling for gender and IQ in analyses of MPAs [4, 9], but only IQ was controlled for in the analyses as there was no association in this sample between gender and MPAs. Additional analyses explored the cross-trait association between the MPA score in one twin with the (i) total raw SRS-2 score and (ii) IQ for the other twin to explore the genetic correlation of the measures. Results for the cohort and within-pairs GEE analyses described above are reported as odds ratios (OR) and/or β estimates with 95% confidence intervals (CI). Statistics were calculated with SPSS version 24 and R version 3.3.2.

Of the 116 participants in the study, 53% (n = 61) had a NDD diagnosis and 47% (n = 55) were TD. The percentage of specific NDD diagnoses in the sample were the following: 24% (n = 28) had a diagnosis of ASD, 32% (n = 37) ADHD, 8% (n = 9) ID, and 26% (n = 30) with other NDDs (e.g., motor, communication, or specific learning disorders). Fifty-seven percent (n = 66; 62% with NDDs) were male, and 43% (n = 50; 40% with NDDs) were female. Ages at examination ranged from 9 to 23 years (mean = 14.05, standard deviation = 3.40) (see Table 1). Details related to participants’ parents can be found in Additional file 3: Table S3.

The interrater agreement between the two sets of raters ranged from 70.4 to 100% across the MPAs. The MPAs with the lowest agreement were downslanted palpebral fissure (70.4%), long eyelashes (70.4%), and straight eyebrows (74.1%). There were numerous MPAs with 100% agreement, including ectrodactyly of the hands and feet, broad thumbs, prominent heels, wide mouths, and hirsutism.

Twin pairs concordant for NDDs had a median of six MPAs, followed by a median of five MPAs for NDD-discordant co-twins and affected twins, respectively (Fig. 1a). In comparison, pairs concordant for TD had a median of three MPAs. These trends were similar for MZ twins (Fig. 1b). The most common MPAs in participants with NDDs were overweight (33%), arachnodactyly/long toes (31%), hypermobility (26%), straight eyebrows (25%), and vision impairment (21%; 46% of these with corrective lenses). In comparison, microtia (24%) and arachnodactyly/long toes (22%) were the only MPAs present in greater than 20% of participants with TD.

No significant association was found between a diagnosis of any NDD and the number of MPAs in the cohort (OR = 1.09, 95% CI = .94–1.27, p = .256). Since there was an association between IQ and MPAs in the cohort (β = − .95, SE = .32, 95% CI = − 1.59 to − .32, p = .003), in that for every MPA, the IQ score decreased by about one point, we controlled for IQ in the model testing the association between any NDD diagnosis and MPA, and the results did not change (OR = 1.05, 95% CI = .89–1.24, p = .592). Conversely, a cross-trait analysis comparing the MPA score from one twin to the IQ in the co-twin showed a strong association (β = − .88, 95% CI = − 1.52 to − .24, SE = .327 p = .007; Additional file 4: Table S4). The within-pairs association between NDD diagnosis and MPAs was neither significant for all participants (OR = 1.10, 95% CI = .46–2.65, p = .832), nor MZ twins only (OR = 1.11, 95% CI = .44–2.82, p = .824) (Additional file 5: Table S5).

Twins pairs concordant for ASD had descriptively the highest median (Md) number of MPAs (Md = 9), followed by a median of six-and-a-half and six MPAs for ASD-discordant co-twins and affected twins, respectively (Fig. 1a). These trends were similar for MZ twins (Fig. 1b). The most common MPAs in participants with ASD included overweight (39%), hypermobility (36%), pes planus (29%), straight eyebrows (29%), vision impairment (25%; 29% of these with corrective lenses), arachnodactyly/long toes (25%), long eyelashes (21%), and microtia (21%).

There was an association between a diagnosis of ASD and MPAs (OR = 1.29, 95% CI = 1.00–1.66, p = .047). The association remained when controlling for ADHD (OR = 1.29, 95% CI = 1.02–1.63, p = .032), other NDD diagnoses (OR = 1.33, 95% CI = 1.05–1.67, p = .016), but not IQ (OR = 1.26, 95% CI = .98–1.61, p = .073). The within-pairs association for ASD was neither significant for all participants (OR = 1.28, 95% CI = .60–2.75, p = .529), nor MZ twins only (OR = 1.42, 95% CI = .63–3.19, p = .398) (Additional file 5: Table S5).

Due to the association between the number of MPAs and a clinical diagnosis of ASD, further analyses were conducted to examine if the association held when the number of MPAs was analyzed in relation to the severity of autistic traits. Using the linear regression model, there was an association between MPA scores and SRS-2 scores in the entire sample (β = 3.02, SE = .98, 95% CI = 1.09–4.94, p = .002), indicating that for every MPA, there was an approximately three-point increase in SRS-2 scores. The association remained when controlling for IQ (β = 2.28, SE = .97, 95% CI = .37–4.18, p = .019), but neither in the within-pairs analysis (β = 1.49, SE = 1.80, 95% CI = − 2.04 to 5.02, p = .409), nor MZ only twins (β = 2.11, SE = 1.85, 95% CI = ⁻1.51–5.73, p = .254) (Table 2 and Additional file 5: Table S5). A cross-trait analysis comparing the MPA score from one twin in each pair to the total SRS-2 raw score showed a strong association (β = 3.34, 95% CI = 1.39–5.30, SE = 1.00, p < .001; Additional file 4: Table S4).

We also examined the presence of MPAs in all seven MZ ASD-discordant twin pairs in the sample. Of note, the overall number of MPAs was similar within pairs or even higher at times for the unaffected co-twin, but two out of seven ASD affected twins had scoliosis, which was absent in their co-twins (Additional file 6: Table S6).

Twin pairs concordant for ADHD had a median of six MPAs each, followed by a median of four and three MPAs for affected twins and co-twins in ADHD-discordant pairs, respectively (Fig. 1a). These trends were similar for MZ twins (Fig. 1b). The most common MPAs in participants with ADHD were overweight (32%), hypermobility (30%), vision impairment (24%; 29% of these with corrective lenses), and straight eyebrows (22%). However, no association was found between ADHD and MPAs in the cohort (OR = 1.05, 95% CI = .93–1.19, p = .435), in the within-pairs analysis for all participants (OR = .53, 95% CI = .24–1.19, p = .123), or when looking at MZ twins only (OR = .36, 95% CI = .09–1.44, p = .151) (Additional file 4: Table S4). We examined the presence of MPAs in the three ADHD-discordant MZ twin pairs in our sample. No major differences in the number or types of MPAs were noted within these twin pairs (Additional file 6: Table S6).

The number of MPAs within MZ twin pairs was highly correlated [Spearman correlation (r_s) = .88, p < .001]. In contrast, no correlation was seen in the DZ pairs (r_s = − .19, p = .676) (Fig. 2 and Additional file 7: Figure S1). MZ pairs had higher numbers of identical MPAs within pairs (Md = 4) compared to DZ pairs (Md = 1; z = − 2.764, p = .006). Additionally, we found that within MZ pairs, there was a smaller median difference in the specific MPAs present within pairs (Md = 2) compared to DZ pairs (Md = 4, z = − 1.066, p = .287).