Response

Extract

Determining the genetic basis of hereditary prostate cancer has turned out to be an arduous task. Prostate cancer genetic research has been hindered by the late age of onset of the disease, variation in the access to prostate cancer screening, and the inability of investigators to distinguish genetic and nongenetic forms of the disease based on clinical features alone. Hsieh et al. now report a second independent set of families confirming linkage of prostate cancer to HPC1, the putative prostate cancer susceptibility locus on chromosome 1q discovered by Smith et al. ( 1 ) . This confirmation is particularly important given the failure of McIndoe et al. ( 2 ) to demonstrate linkage to chromosome 1q24-25 markers in 49 large, high-density prostate cancer families.

How does one interpret these contradictory findings? Review of the clinical features of the prostate cancer families utilized in the various studies reveals more similarities than differences ( Table 1 ). The average age of onset of prostate cancer in the four published studies is slightly younger than the average age of onset in the general population ( 4 ) . Both studies by Smith et al. and McIndoe et al. targeted large families with many affected men and typed a significant number of unaffected as well as affected family members. Such a strategy is ideal for detecting linkage to a rare dominant disease gene. Our study ( 3 ) and the study by Hsieh et al. identified smaller prostate cancer families and genotyped primarily affected individuals. All four studies employed the same nonparametric statistical tests. In addition, the studies by Smith et al. and McIndoe et al. used parametric statistical tests appropriate for linkage to a rare dominant disease gene. Since the parametric tests rely on debatable model assumptions and parameter estimates, they are, in our opinion, less reliable than the nonparametric tests. As a minor statistical criticism of the study by Hsieh et al., we would have preferred to see nonparametric linkage scores calculated against all markers simultaneously. The lack of information at one marker is often compensated for by the presence of information at a nearby marker.