Abstract Purpose To assess change in incidence of retinoblastoma in Northern Europe and to compare commonly used methods for calculating its incidence against birth cohort analysis. Design Retrospective cohort study. Participants Individual and pooled data of 291 Swedish and 174 Finnish children diagnosed with retinoblastoma between 1958 and 1998. Main outcome measures Incidence per 1 million children younger than 5 years of age (37 812 035 person- years at risk) and per 100 000 live births (7 152 265 live-born children at risk). Methods Data were from Swedish and Finnish Cancer Registries and corresponding national referral centers for retinoblastoma. Incidence was calculated both by standard analysis per children younger than 5 years of age and per live births, and by birth cohort analysis. Curves were smoothed with robust, locally weighted regression. Linear regression was used to fit pooled data. Results The number of new retinoblastoma cases per year ranged from 0 to 13 (1–13 per birth cohort) in Sweden and from 0 to 10 in Finland (1–9 per birth cohort). The mean incidence was 11.8 (95% confidence interval [CI], 10.5–13.1) and 11.2 (95% CI, 9.4–13.0) per 1 million children younger than 5 years of age in Sweden and Finland, respectively, and 6.7 (95% CI, 5.9–7.5) and 6.2 (95% CI, 5.3–7.2) per 100 000 live births, respectively. Analysis based on year of diagnosis suggested moderate increase in incidence since 1990, but by birth cohort analysis, incidence rates were stable for both countries. The pooled incidence by birth cohort was 6.0 (95% CI, 5.4–6.6) per 100 000 live births, corresponding to 1 in 16 642 (95% CI, 15 105–18 528) live births. Conclusions The data suggest that the incidence of retinoblastoma is stable in Northern Europe. Analysis based on birth cohort is recommended for future epidemiologic studies, because it minimizes the effect of variable age at diagnosis of this developmental cancer and results in less variable incidence rates than standard analysis based on year of diagnosis.