Measuring correlation in clinical traits among relatives is important to our understanding of the causes of variable expressivity in Mendelian diseases. Random effects models are widely used to estimate intrafamilial correlations, but such models have limitations. We incorporated survival techniques into a random effects model so that it can be used to estimate intrafamilial correlations in continuous variables with right censoring, such as age at onset. We also describe a negative-binomial gamma mixture model to determine intrafamilial correlations of discrete (e.g., count) data. We demonstrate the utility of these methods by analyzing intrafamilial correlations among patients with neurofibromatosis 2 (NF2), an autosomal-dominant disease caused by mutations of the NF2 tumor-suppressor gene. We estimated intrafamilial correlations in age at first symptom of NF2, age at onset of hearing loss, and number of intracranial meningiomas in 390 NF2 nonprobands from 153 unrelated families. A significant intrafamilial correlation was observed for each of the three features: age at onset (0.35; 95% confidence interval (CI) 0.23-0.47), age at onset of hearing loss (0.51; 95% CI, 0.35-0.64), and number of meninginomas (0.29; 95% CI, 0.15-0.43). Significant correlations were also observed for age at first symptom within NF2 families with truncating mutations (0.41; 95% CI, 0.06-0.68) or splice-site mutations (0.29; 95% CI, 0.03-0.51), for age at onset of hearing loss within families with missense mutations (0.67; 95% CI, 0.18-0.89), and for number of meningiomas within families with splice-site mutations (0.39; 95% CI, 0.13-0.66). Our findings are consistent with effects of both allelic and nonallelic familial factors on the clinical variability of NF2.
Copyright 2002 Wiley-Liss, Inc.