Résumé
Comparative oncology is a rapidly expanding field that seeks to explain variation in cancer risk across species by examining trends between tumour prevalence and key risk factors such as body mass, longevity, life history traits, and mutation rates. These trends are then used to address fundamental questions in the field, including the discovery of potential novel anti-cancer therapies, improvements to species conservation efforts, and understanding how cancer has influenced the evolution of multicellularity. They thus must be robust. This study demonstrates that when estimated on captive species those trends are heavily influenced by their scientific and public popularity, and that accounting for this bias can substantially alter their direction and magnitude. Hence, we reanalysed published captive vertebrate datasets examining the associations between neoplasia, malignancy, and lethal tumour prevalences with body mass, longevity, life history traits, and germinal cells mutation rates. When we included proxies of species popularity in our analyses, the previously reported weak effect of body mass on tumour and malignancy prevalences disappeared entirely. Similarly, the previously reported positive relationship between germline mutation rate and cancer mortality was eliminated after controlling for popularity bias. For life history traits, the effect of clutch size on cancer neoplasia and malignancy prevalences in birds doubled in magnitude, and while the negative trend between gestation length and tumour prevalence in mammals was not greatly affected, our analyses revealed that baseline tumour prevalences were underestimated for popular animals. Finally, the previously observed association between hemochorial placentation and cancer mortality in mammals was eliminated when confounding variables were included. Collectively, these results demonstrate that current comparative analyses based on tumour prevalence in captive animals are heavily influenced by species’ scientific and public popularities. Future studies utilising such datasets should incorporate measures of species popularity as confounding variables to ensure more robust conclusions and misleading research directions.