Re: Assessment of Impact of Outmigration on Incidence of Second Primary Neoplasms in Childhood Cancer Survivors Estimated From SEER Data

Abstract

The Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute reports cancer incidence, prevalence, and survival from selected geographic areas of the United States (1). However, estimates of the incidence of secondary cancers among SEER childhood cancer survivors could be invalidated by outmigration because cancer diagnoses for persons residing outside SEER catchment areas are not reportable (2,3). Outmigration bias has not previously been assessed (4). We, therefore, report results from the first study, to our knowledge, of underascertainment in SEER incidence rates for second primary cancers (SPCs) in survivors of childhood or adolescent cancer, a population in which change of residence could be frequent. Data on SPCs in 5-year childhood cancer survivors were obtained from SEER 9 Registries Database Nov 2010 Sub (1973–2008) (1) and the Childhood Cancer Survivor Study (CCSS), an ongoing cohort of survivors identified in 26 North American medical centers between 1970 and 1986 with 85% complete follow-up (5). We included individuals who were diagnosed with an International Childhood Cancer Classification (ICCC) malignancy before age 20 years and who were alive 5 years after the diagnosis of the first cancer (6). SPC in CCSS was the first SEER-reportable primary malignancy subsequent to a childhood cancer. SEER included SPC diagnoses occurring before 2009; CCSS before 2010. Because of differences in the characteristics of the SEER- and CCSS-eligible populations (Table 1), we matched each childhood cancer survivor in SEER to an individual in CCSS based on sex, birth year, age at diagnosis, and first primary cancer diagnosis using the 12 major ICCC categories. SEER survivors without an exact match were excluded. Where multiple matches were possible, one pairing was randomly selected. Table 1. Summary characteristics of 5-year childhood cancer survivor cohorts before and after matching Underascertainment bias was assessed in the matched data set by computing the SPC incidence ratio (IR) as the ratio of the total number of SEER SPC cases to total person-years divided by the ratio of total number of CCSS SPC cases to total person-years. Person-years were calculated as the time from first primary diagnosis to the development of SPC, death, or end of follow-up (December 31, 2008), assuming no loss to follow-up. We also computed the incidence ratio for all-cause mortality, an endpoint that should not be affected by outmigration because mortality data are obtained from the National Death Index. Bootstrap resampling was used to obtain 95% confidence intervals (CIs) for the incidence ratios. SEER reported 7967 5-year childhood cancer survivors whose first cancer was diagnosed before 1987, and CCSS reported 13 771 such individuals. Based on sex, birth year, age at diagnosis, and type of diagnosis, 5604 exact matches were found in the CCSS cohort. There were 422 SPCs among the matched SEER survivors. We found statistically significantly higher incidence of SPCs in SEER survivors than among their CCSS matches (IR = 1.15, 95% CI = 1.06 to 1.27). The incidence ratio was unchanged when Canadians (n = 1017 individuals) were excluded from the analysis (IR = 1.13, 95% CI = 1.04 to 1.24). When we additionally matched on first course radiation treatment and geographic region (San Francisco, Atlanta, Seattle), SPC incidence remained statistically significantly higher in SEER than in CCSS (IR = 1.24, 95% CI = 1.02 to 2.81). We could not determine if this elevated risk was due to differences in childhood cancer treatment between SEER and CCSS survivors, as SEER collects only limited information about the first course of treatment. In contrast to SPC rates, the rates of all-cause mortality were not statistically significantly different between these groups (IR = 0.71, 95% CI = 0.54 to 1.03). In a matched comparison of long-term childhood cancer survivors in SEER and the CCSS cohort, a large sample of survivors treated at North American academic medical centers, we found no evidence of an outmigration bias in the incidence of SPCs in SEER.