Dealing with emigration in cohort studies: follow-up of mortality and cancer incidence among Norwegians born between 1967 and 1976.

Emigration causes loss to follow-up. The study aim was to assess the influence of the choice of handling migration in population-based cohort studies on estimated mortality and cancer incidence in the population of origin. All persons born in Norway between 1967 and 1976 and who were not registered dead before 1992 (N = 614,176) were followed up in national registries regarding migration movements, death, and incident cancer between 1992 and 2004. A total of 40,366 (6.6%) of the study population had between 1 and 13 migration movements and 5,354 deaths and 4,447 first cancer cases were recorded during follow-up. Four different follow-up scenarios concerning migration were analysed: considering only person-time before emigration; considering person-time as national residents both before emigration and after repatriation; disregarding whether emigration took place or not; and excluding all who emigrated during follow-up. Mortality and cancer incidence rates were compared in Poisson regression models. Mortality and cancer incidence were only marginally influenced by choice of follow-up scenario. Mortality was higher after repatriation, in particular during the first year of follow-up (rate ratio 2.03; 95% confidence interval 1.02-4.03). This excess had little influence on total population rates. Cancer incidence was not affected by repatriation status. Mortality rates after repatriation were probably elevated because persons who expected to die shortly were more prone to return to their native country ("salmon bias"). The analytical choice concerning follow-up has little influence on outcome occurrences in populations with rather low migration rates. However, the best solution is apparently to censor out persons at the date of emigration in order to avoid salmon bias.