Background: A 20-year follow-up study on cancer incidence among people living near nuclear power plants in South Korea ended in 2011 with a finding of significantly, but inconsistently, elevated thyroid cancer risk for females. Reanalysis of the original study was carried out to examine the dose–response relationship further, and to investigate any evidence of detection bias. Methods: In addition to replicating the original Cox proportional hazards models, nested case-control analysis was carried out for all subjects and for four different birth cohorts to examine the effects of excluding participants with pre-existing cancer history at enrollment. The potential for detection bias was investigated using the records of medical utilization and voluntary health checks of comparison groups. Results: The overall risk profile of the total sample was similar to that of the original study. However, in the stratified analysis of four birth cohorts, the cancer risk among people living near nuclear power plants became higher in younger birth cohorts. This was especially true for thyroid cancers of females (hazard ratio (HR) 3.38) and males (HR 1.74), female breast cancers (HR 2.24), and radiation-related cancers (HR 1.59 for males, HR 1.77 for females), but not for radiation-insensitive cancers (HR 0.59 for males, HR 0.98 for females). Based on medical records and health check reports, we found no differences between comparison groups that could have led to detection bias. Conclusions: The overall results suggest elevated risk of radiation-related cancers among residents living near nuclear power plants, controlling for the selective survival effect. This is further supported by the lack of evidence of detection bias and by records of environmental exposure from radiation waste discharge.