Radiation-induced thyroid neoplasms: evidence for familial susceptibility factors.

To determine if there is a familial component to susceptibility to radiation-induced thyroid neoplasms, we studied 572 individuals who were members of 286 sibpairs who received childhood radiation treatment and for whom follow-up information was obtained. Of these 572 individuals, 240 (42.0%) had thyroid neoplasms (benign and malignant), and 75 (13.1%) had surgically confirmed thyroid cancer. To test the null hypothesis, that neoplasm occurred without regard to family membership, it was necessary to take into account each individual's years at risk and known risk factors. These risk factors, analyzed by the proportional hazards model of Cox, were sex, age at time of radiation treatment, and treatment dose. For each individual, we calculated the cumulative hazard that a neoplasm would occur from that individual's specific risk factors and years at risk. Each individual was also assigned an indicator, D = 1 or 0, according to whether a neoplasm had occurred. Finally, for each individual we computed a residual, D minus the cumulative hazard. In the absence of familial effects, positive and negative residuals would be distributed without regard to family membership, whereas residuals would tend to have concordant signs and magnitudes within families if familial effects were present. Permutational methods, therefore, were used to determine whether the sum among families of the products of residuals within sibpairs was too large, compared to random pairing. For all thyroid neoplasms (both benign and malignant), within-family concordance was significant (P = 0.05, the observed sum among families of the products of residuals was larger than 9468 of 9999 permutations). For thyroid cancer considered alone, the same analysis did not demonstrate familial concordance conclusively, but the results were suggestive (P = 0.18). We conclude that in addition to the previously described risk factors of female sex, younger age at radiation exposure, and higher dose, it is likely that there are independent familial risk factors for developing thyroid neoplasms. Whether these are genetic or environmental factors remains to be determined.

A stratified Wilcoxon-type test for trend.

We propose an extension of Cuzick's non-parametric Wilcoxon-like trend test for situations when the observations can be grouped into strata which are thought to be related to the outcome. We illustrate its use in two examples from cancer epidemiology.