Response to DeSantis and Jemal
Year: 2018
Type: article
Abstract: We thank DeSantis and Jemal for bringing additional attention to the important issue of racial and ethnic disparities in breast cancer incidence in their correspondence. We reviewed their analysis of female invasive breast cancers occurring in all ages in the NAACCR and SEER 13 registries databases. On two points, we agree completely: first, estimates for the whole United States may vary depending on the collection of registries analyzed; and second, aggregated data may mask important heterogeneity at the local or regional level. After reviewing their concerns, we stand by our results (1), including our projections that breast cancer incidence through 2030 is expected to slowly increase among non-Hispanic Whites (NHW) and slowly decrease among non-Hispanic Blacks (NHB). Our age-period-cohort forecasts are not simply an extrapolation from the overall age-standardized rate curve; rather, they account for age interactions over time, which are a consequence of birth cohort effects that impact each individual age group. For the SEER 13 data analyzed in our report (1), we show in Figure 1 that the cohort-specific breast cancer relative risk among NHW has been statistically significantly increasing in successive cohorts born since 1956 by 0.9% per birth year (95% confidence interval 0.6 to 1.2%/birth year), but is not statistically significantly changing in NHB. A two-sided Wald test was used to determine statistical significance. These birth cohort trends are incorporated into our forecasts and are largely responsible for the projected future increases in invasive breast cancer incidence in NHW and the slight decreases in NHB women. Estimated cohort rate ratios with JoinPoint analysis for primary invasive female breast cancer incidence among non-Hispanic White women (stars) and non-Hispanic Black women (diamonds), using single-year SEER 13 data for ages 30 to 84 and calendar years 1992 to 2014. Age-period-cohort forecasting models estimate cohort-specific age-specific incidence for observed and forecast years (2015–2030) using estimated age-specific incidence for the 1957 reference birth cohort (data not shown) multiplied by the observed (1908–1984) and future (1985–2000) birth cohort relative risks as shown. The JoinPoint model for non-Hispanic White women has four statistically significant segments, and the model for non-Hispanic Black women has three. Shaded bands show point-wise 95% confidence intervals (CI) for the cohort rate-ratios for the observed and future cohorts. For more details on projections, see Best et al. 2018 [4]. DeSantis and Jemal took exception to our exclusion of incidence rates for females younger than age 30 years and older than age 84 years. We did so because only 0.5% of cases occurred in girls or women younger than age 30 years, and SEER does not provide age-specific population counts for persons ages 85 years and older. Thus, we began our analysis at age 30 years, not at age 20 years or age 0 years, consistent with prior analyses (2). DeSantis and Jemal conclude when comparing their analysis of SEER 13 vs NAACCR incidence trends that SEER 13 overestimates national incidence rates. We concur that differences are to be expected given heterogeneity across registries. In addition, there are other potential contributors to heterogeneity between two different sets of analyses. For example, given our etiological emphasis, we opted to restrict our SEER 13 analysis to the first primary breast cancer for each woman rather than all primary breast cancers, which would lead to slightly different results. As we noted in our brief report (1), a complete evaluation of racial disparities in first primary invasive breast cancer will need to account not only for ethnicity but also for hormone receptor status, screening effects (via age stratification) (2) and temporospatial heterogeneity (3); this will be the subject of future investigations. This research was supported by the Intramural Research Program and the Cancer Prevention Fellowship of the National Cancer Institute at the National Institutes of Health. Affiliations of authors: Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD (BCDL, PSR, WFA, GLG), Division of Cancer Prevention (BCDL) The funders had no role in the writing of this manuscript, data analysis, or decision to submit it for publication. The authors have no disclosures to declare.
DOI: 10.1093/jnci/djy163
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: green
Funder: National Institutes of Health
OpenAlex ID: W2897809501