LAUSR

BACKGROUND Subsequent primary cancers (SPCs) compete with first cancers and non-cancer events as the primary cause of death among cancer patients. We aimed to assess temporal trends in SPC mortality since 1980 among adult-onset cancer patients in competing risk models. METHODS Patients registered with a first cancer in the population-based Tasmanian Cancer Registry, Australia, between 1980-2009 were followed up to December 2014. Cumulative incidence function (CIF) was used to estimate the cumulative incidence of cause-specific deaths in the presence of competing risks. The hazard ratios of SPC-specific deaths were assessed in two regression models: subdistribution hazard ratios from competing risk models (SHRs) and hazard ratios from Cox models (CHRs). RESULTS Overall, 5339 (9.3%) of 57,288 patients developed SPCs and 2494 died from SPCs during the follow-up. While the cumulative incidence of first cancer deaths at 5, 10, 15 and 20-years gradually decreased over periods of first cancer diagnosis, the cumulative incidence of SPC deaths did not. The SHRs for SPC-specific deaths increased from the reference period 1980-1984 to a peak for first cancers diagnosed in 1995-1999 (SHR = 1.18, 95%CI 1.03-1.35), before a decrease in 2005-2009 (SHR = 0.82, 95%CI 0.70-0.95) in competing risk models. However, this pattern was not consistent in CHRs. For individuals with specific first cancers, those with a first prostate cancer in 1995-1999 ha d the greatest SPC mortality risk (SHR = 2.08, 95%CI 1.29-3.36). CONCLUSION Competing risk models, but not Cox models, demonstrated temporal increases in SPC-specific mortality. Greater detection of non-fatal first prostate cancers appears to have contributed to this trend.