It is often of interest to know whether there is increased occurrence of a trait in a pedigree or other structured set of epidemiological data. In answering such questions most current methods use aggregate measures, such as relative risk, that may not relate the outcome for each individual to that individual's risk. In this paper we present a simple method, and its computational algorithm, to overcome this limitation. This new method also permits one to identify high-risk families or subsets of a collection of data, which is not always possible using other approaches.
In a study of cancer risk among relatives of retinoblastoma patients, by applying this new method it was found that 11 of 33 families each obtained through a unilateral retinoblastoma patient are at statistically high risk of cancer at all sites combined, while there are 15 of 47 such families obtained through a bilaterally affected proband. These results are unlikely to have occured by chance, indicating an overall excess risk in the ancestors of these retinoblastoma cases.
The proposed test procedure does not specify the cause of elevated risk; however, a method is proposed that provides some indication regarding possible causal mechanisms under some circumstances.