Divergent understandings in comparative oncology

Butler et al.’s recent exploration of the rates of species diversification and cancer is an important addition to the study of comparative oncology (1). Do lineages that tend to diversify more rapidly get more cancer? The authors hypothesize that a susceptibility to genomic rearrangements might lead to higher diversification and cancer rates. However, because deleterious rearrangements are swiftly eliminated, only lineages with robust, stabilized genomes can diversify (2). Diversification is a signature of genomic stability rather than instability. Therefore, species with more stable genomes might diversify more and have less cancer. Unfortunately, the statistical model, previously published in Butler et al.’s analysis of body size and cancer (3), is deeply flawed. Butler et al. treat tumor counts as additive, therefore, any apparent relationship between diversification and cancer prevalence may simply reflect differences in sample size rather than related to evolution. Their approach estimates a proportion of necropies that will have cancer, which implies that all species have the same rate of cancer. The species differences only come through their differences in the other predictors (body size, change in body size, and diversification rate) which add or subtract a fixed number of cancers, regardless of the number of necropsies observed. With only 87 mammal and 76 bird species in their analysis, Butler et al. may lack the statistical power needed to detect interactions between the number of necropsies and their other predictors (they do not report testing for those interactions). An alternative statistical model allows each species to have a different intrinsic rate of cancer which is then modified by the other predictors like body size or diversification rate. When we reanalyzed their data under these assumptions using methods from Compton et al. (4), we found no evidence of an association between diversification rate and either malignancy or benign tumor prevalence in either mammals or birds (Fig. 1). Nor did we find an association between change in body size and malignancy in either birds or mammals (Fig. 1). While a larger sample size may prove such a relationship in the future, it is unlikely to rescue fundamentally flawed methodology.