Certification of mobile apps for health care--reply.

In Reply We share Dr Constantino’s interest in determining if there are differences in the etiology of ASD between males and females. We did not compare the heritability of ASD between males and females in the study for 2 reasons. First, even a large population-based study such as ours has limited power to detect statistically significant sex differences in heritability. Second, we addressed the question of sex effects in our recurrence risk analysis. We did not observe important or statistically significant differences in recurrence risk in siblings and cousins for the different sibling and cousin sex combinations. The relative risk for males or females was not dependent on the sex of the proband. Furthermore, an epidemiological study from Denmark also did not find support for sex-related differences in recurrence risk.1 We look forward to seeing this topic addressed further in future populationbased studies. Several factors may affect concordance rates. Some earlier twin studies used interviews, rating scales, or both to score different autistic traits and applied preselected cutoffs to define autism.2,3 The pairwise concordance using such an approach may differ from pairwise concordance based on clinical diagnosis as reflected in our study. The source population, methods of ascertainment, handling of time trends, and participation rates may also affect observed concordance rates. The models we fitted are complex and, like any statistical model, do come with some underlying assumptions. Furthermore, the models can only give crude estimates of the sources and nature of the disease, and therefore we were cautious not to overinterpret our estimates or speculate what the exact sources underlying our estimates might be. Yet one of the potential important results of an analysis such as ours is to facilitate and stimulate a discussion about the specific causes of the disorder. One way to address current concerns about behavioral genetic models of heritability is to complement them with modeling methods using molecular data to estimate heritable and nonheritable factors. We agree with Constantino that de novo mutations would inflate the nonshared environment component term and this may be important.