Bayesian estimates of the incidence of rare cancers in Europe. (June 2018)

• Record Type: Journal Article
• Authors: Botta, Laura; Capocaccia, Riccardo; Trama, Annalisa; Herrmann, Christian; Salmerón, Diego; De Angelis, Roberta; Mallone, Sandra; Bidoli, Ettore; Marcos-Gragera, Rafael; Dudek-Godeau, Dorota; Gatta, Gemma; Cleries, Ramon
• Is Part Of: Cancer epidemiology. Volume 54(2018:Jun.)
• Page Start: 95
• Page End: 100
• Publication Date: 2018-06
• Subjects: European countries -- Incidence -- Rare cancer -- Population-based cancer registries -- Bayesian analysis; Cancer -- Epidemiology -- Periodicals; Cancer -- Prevention -- Periodicals; Cancer -- Diagnosis -- Periodicals; Carcinogenesis -- Periodicals; 616.994005
• DOI: 10.1016/j.canep.2018.04.003
• Languages: English
• ISSNs: 1877-7821

Abstract

Highlights: The classical approach may lead to unreliable estimates when dealing with rare cancers. We use a Bayesian approach to give incidence for every rare cancer and country. We suggest using this approach for cancers with <150 observed cases in a country. The provided estimator expected year needed to observe one case is very informative. Abstract: Background: The RARECAREnet project has updated the estimates of the burden of the 198 rare cancers in each European country. Suspecting that scant data could affect the reliability of statistical analysis, we employed a Bayesian approach to estimate the incidence of these cancers. Methods: We analyzed about 2, 000, 000 rare cancers diagnosed in 2000–2007 provided by 83 population-based cancer registries from 27 European countries. We considered European incidence rates (IRs), calculated over all the data available in RARECAREnet, as a valid a priori to merge with country-specific observed data. Therefore we provided (1) Bayesian estimates of IRs and the yearly numbers of cases of rare cancers in each country; (2) the expected time (T) in years needed to observe one new case; and (3) practical criteria to decide when to use the Bayesian approach. Results: Bayesian and classical estimates did not differ much; substantial differences (>10%) ranged from 77 rare cancers in Iceland to 14 in England. The smaller the population the larger the number of rare cancers needing a Bayesian approach. Bayesian estimates were useful for cancers with fewer than 150 observed cases in a country during the study period; this occurred mostly when the population of the country is small. Conclusion: For the first time the Bayesian estimates of IRs and the yearly expected numbers of cases for each rare cancer in each individual European country were calculated. Moreover, the indicator T is useful to convey incidence estimates for exceptionally rare cancers and in small countries; it far exceeds the professional lifespan of a medical doctor.