Sporadic vs familial classification given etiologic heterogeneity: I. Sensitivity, specificity, and positive and negative predictive value

Environmental factors are etiologically important in many non-Mendelian familial disorders in man. Because such disorders often occur as "sporadic" cases, (ie, an affected individual with no affected relatives), it is tempting to assume that such cases represent an "environmental" form of the disorder. This paper presents an evaluation of the sensitivity, specificity, and positive and negative predictive power (PPV and NPV) of this "sporadic vs familial classification." The model assumes etiologic heterogeneity with a subpopulation of cases due to a "major" environmental event acting independent of genotype and the remaining cases resulting from a generalized single major locus (SML). Sibship size is modeled by a truncated negative binomial distribution. For rare disorders, this classification has high sensitivity and NPV but low specificity and PPV. As the disorder becomes more common, sensitivity and NPV fall while specificity and PPV rise. The power of the method increases substantially with increasing sibship size up to four or five, but further increases in power are minimal. MZ twins add considerable power to the method but aunts and uncles add little if anything. Both a correlational (phi) and an agreement-based (kappa) statistic indicate that, under most realistic circumstances, the relationship between etiology and family history is modest.