Twin Concordance for Autism: a Comparison of Multiple Diagnostic Criteria in a Population-Based Twin Study

Background: Despite assertions that autism is one of the most strongly genetically influenced multi-factorial developmental disabilities, evidence for the strength of genetic influences relies heavily on small English and Scandinavian twin studies ascertained with considerably more narrow diagnostic criteria than used today.
Objectives: To examine contemporary evidence of the genetic contributions to autism and the autism spectrum using a twin study design.

Methods: The study design incorporated a two-pronged approach to participant ascertainment. First, we recruited families from Wisconsin birth records of twins born from 1998-2003, and we screened 3758 two to three year-old twins for autism and the autism spectrum (Kees et al., 2005; Hefter et al., 2008). Second, we undertook statewide case-finding and recruited twins up to 20 years of age who had community diagnoses on the autism spectrum. Primary caregivers of children identified in both prongs were interviewed about each of their twins, using the Social Responsiveness Scale (SRS; Constantino et al., 2003) and the Social Communication Questionnaire (SCQ; Rutter, Bailey, & Lord, 2003), which contains content derived from the Autism Diagnostic Interview–Revised. We then visited the homes of families with at least one twin above cut-off scores on the SRS or SCQ, and we administered the appropriate module of the Autism Diagnostic Observation Schedule (ADOS). We examined extensive medical records, and (blind to zygosity and concordance) we excluded any twin pair in which another medical diagnosis could plausibly render an autism diagnosis nonidiopathic (seizures were not an exclusionary criterion). We analyzed twin concordance for having received a credible community diagnosis and scoring above the thresholds on the ADOS, SCQ, and SRS; uncertain zygosity was determined by genotyping.

Results: MZ pairwise concordance was at least twice DZ pairwise concordance for all measures: community diagnosis = MZ 65% (15/23), DZ 23% (7/31); ADOS-autism cut-off = MZ 53% (10/19), DZ 26% (5/19); ADOS-spectrum cut-off = MZ 77% (17/22), DZ 31% (9/29); SCQ = MZ 57% (12/21), DZ 19% (5/26); SRS-stricter cut-off = MZ 52% (12/23), DZ 11% (3/28); and SRS-broader cut-off = MZ 74% (17/23), DZ 26% (8/31). A combined criterion of ADOS-spectrum cut-off and SCQ led to a MZ pairwise concordance of 50% (10/20) and a DZ pairwise concordance of 15% (3/21), whereas a combined criterion of either ADOS-spectrum cut-off or SCQ led to a MZ pairwise concordance of 62% (13/21) and a DZ pairwise concordance of 17% (4/24). The male:female ratio was relatively consistent across the different diagnostic criteria: community diagnosis = 6.1:1; ADOS-autism cut-off = 5.6:1; ADOS-spectrum cut-off = 4.9:1; SCQ = 4.8:1; SRS-stricter cut-off = 5:1; and SRS-broader cut-off = 4.6:1. We also pursued model-fitting approaches to these data, which will be reported.
Conclusions: These results confirm the classic twin studies in suggesting substantial genetic variance for the autism phenotype. However, unlike some of the classic results, MZ twin concordance is not near 100%, and DZ concordance does not approach zero. Implications for genetic models of autism are discussed.