The validity of analyses testing the etiology of comorbidity between two disorders: comparisons of disorder prevalences in families.

摘要

Klein and Riso proposed several alternative models explaining the causes of comorbidity between two disorders (KR models). For each comorbidity model, they also presented a set of predictions for comparing the prevalence of disorder A-only, disorder B-only, and disorder AB (i.e., both disorders) among the relatives of probands with A-only, B-only, AB and controls (i.e., the KR predictions). Neale and Kendler provided the quantitative expectations for these prevalences (i.e., the NK models) and suggested biometric model fitting as an alternative way of testing comorbidity models. Neale and Kendler also suggested that the KR predictions have limited use because variations in the model parameters may lead to different predictions. We tested the KR predictions on two sets of data simulated under the assumptions of the KR/NK models. The results predicted by Klein and Riso and the results derived from the simulated datasets matched in most cases, but there were several notable discrepancies between the two sets of results. First, these discrepancies may be due to variations in the model parameters although the KR predictions are valid tests of the model for some model parameter sets. Second, several KR predictions may not be valid because they do not consider the necessary conditions for the diagnosis of A-only and B-only or alternative routes to comorbidity that are not hypothesized in their comorbidity model, including the fact that some comorbid cases will result by chance in all comorbidity models.