Etiologic Conclusions from Similar Birth Cohort Effects

In Response: We thank Drs. Stang and Jöckel for their comments (1) on our article (2). In our study, we compared the incidence rates of testicular seminoma and nonseminoma in adult men using data from high-quality population-based registries across Europe, hypothesizing that a consistent observation of homogeneity in generation-specific trends would signify the subtypes share a similar, if currently, enigmatic etiology. We believe that comparisons in the birth cohort dimension serve such a purpose: the exposures that induce testicular germ cancer likely occur within a narrow age range very early in life (3) and thus should manifest themselves as cohort effects (4).We concede that there is considerable random variability in the incidence data in the Czech Republic and in the other populations under study. We restricted the investigation to eight countries for which the incidence in the underlying age-period cells was sufficiently numerous to interpret the trends and chose not to present the rate ratios in the single cells associated with the greatest random variability (e.g., those representing by the earliest and the most recent cohorts).We did not however attempt to statistically test for differences in the rate ratios between histologic subtypes but did counsel in our conclusion that “statements as to the degree of homogeneity of seminoma and nonseminoma trends must be equivocal, given that nonidentifiability precludes the possibility to present and compare unique estimates of the cohort trends.” Although there are compelling reasons to apportion the increasing linear trend to birth cohorts for both subtypes, in view of the lack of uniqueness of the incidence rate ratios, limited additional inference can be gained from a formal test of homogeneity.Our study showed that there was a consistency in the observed cohort-specific incidence rate ratios of seminoma and nonseminoma in seven of the eight European populations. Our conclusions of “largely similar if not identical” etiologies were not drawn in isolation of the biological and epidemiologic evidence for and against such an observation. We certainly did not discount evidence from studies supporting etiologic heterogeneity in the subtypes and liberally cite both descriptive and analytical studies reporting such findings in our article. The analysis of the literature provided by Stang and Jöckel seems selective. They cite Akre et al. (5) as evidence for heterogeneity and ignore that the more recent extension and update of that particular study by Richiardi et al. (6) reported no important heterogeneity. Our conjecture is that a proper, systematic analysis of all the available studies will find no consistent difference in the risk factor patterns for seminoma and nonseminoma.Lastly, the aim of Fig. 2 in our study was to simply show that where rates of seminoma were high so were rates of nonseminoma and vice versa. We commented on the higher seminoma to nonseminoma ratio in Italy and possible explanations for that observation. We also cautioned against overinterpretation of variability by subtype on the period scale. The 5-fold variation in testicular germ cell cancer rates in Europe and the recent leveling off of incidence in several high-risk countries possibly implies the testicular cancer epidemic is in different phases in different countries (7). The recently observed deviations from the uniformly increasing trend in countries in the mature phase (e.g., Denmark and Switzerland) should be seen in the rates of nonseminoma some years ahead of those of seminoma if, as is likely, the former are more aggressive and rapidly growing than the latter.