Age-specific Trends in Colorectal Cancer Incidence for Women and Men, 1935–2017

Abstract

Although colorectal cancer (CRC) incidence has decreased over time in adults aged ≥50 years, largely due to advancements and uptake in screening, incidence has dramatically increased over the past 4 decades in adults aged < 50 years in the United States.1Ansa B.E. et al.J Clin Med. 2018; 7: 22Crossref Scopus (64) Google Scholar Previous studies have only evaluated age-specific trends in CRC incidence from the mid-1970s or later, providing only a few years of data before the documented increase in early-onset CRC. This makes it difficult to establish the temporal relationship between potential risk factors and CRC incidence, given the induction time of most cancers. For example, it is not known if early-onset CRC incidence was increasing before the rise in obesity rates in the United States, which also began to increase across age groups in the mid-1970s.2Flegal K.M. et al.JAMA. 2002; 288: 1723-1727Crossref PubMed Scopus (5253) Google Scholar To further inform hypotheses about the etiology of CRC, we evaluated age- and sex-specific trends in CRC incidence rates from 1935 to 2017 using the oldest population-based cancer registry in the United States, which includes almost 4 more decades of data compared with national cancer registries. We used data from the Connecticut Tumor Registry (CTR), a population-based registry of all cancers diagnosed in Connecticut residents since 1935. Cases were ascertained by active reporting from hospitals and by death certificates. Before 1945, about 25% of cases were confirmed by death certificates only compared with <3% after 1967.3Heston J.F. et al.Forty-five years of cancer incidence in Connecticut: 1935–79. US Department of Health and Human Services, Public Health Service, National Institutes of Health, National Cancer Institute, 1986Google Scholar CRC was defined using the Surveillance Epidemiology and End Results (SEER) Site Recode definitions of colon cancer excluding the appendix (21041, 21043–21049), rectosigmoid junction cancer (21051), and rectal cancer (21052) (see https://seer.cancer.gov/siterecode/). We used Joinpoint software (version 4.6.0.0; National Cancer Institute) to evaluate the annual percent change (APC) in CRC incidence rates from 1935 to 2017 (for details see Supplementary Methods). A positive APC indicates the annual rate at which incidence is increasing over a specified time interval, whereas a negative APC indicates the annual rate at which incidence is decreasing. We evaluated CRC incidence trends stratified by sex and the age groups 25–49 years (25–39 years was also considered), 50–69 years, and 70–84 years. We also evaluated trends by tumor subtype, including distal colon, proximal colon, and rectal. We compared data between the CTR and the SEER 9 Registries (1975–2017), which includes the CTR.4SEER Program.https://seer.cancer.gov/registries/terms.htmlGoogle Scholar All rates were age-standardized against the 1970 US Standard Population. We used deidentified data aggregated at the registry level, precluding this study from institutional review board review. As shown in Figure 1, CRC incidence statistically significantly increased in 25–49-year-old women in the CTR at an annual rate (ie, APC) of 1.75% (95% confidence interval, 1.18–2.32) from 1985 to 2017; this was preceded by over 3 decades (1953–1985) of declining incidence. In 25–49-year-old men CRC incidence was stable from 1935 to 1974, decreasing from 1974 to 1981 (although the APC for this segment was not statistically significantly different from zero) and statistically significantly increasing from 1981 to 2017 (APC, 1.99%; 95% confidence interval, 1.55–2.43). Similar trends were found when we restricted the age group to 25–39-year-olds (data not shown). When we stratified by subtype, only the incidence of rectal cancer, but not distal or proximal colon cancer, significantly increased in recent decades in 25–49-year-old women and men (Supplementary Figure 1). In adults aged ≥50 years CRC incidence rates statistically significantly decreased across most time points from 1985 to 2017, including for distal colon, proximal colon, and rectal cancer. Age-specific incidence trends in the CTR paralleled trends in the SEER 9 Registries from 1975 to 2015, the period with overlapping data from both sources (Supplementary Figure 2). Using historical data from the oldest population-based cancer registry in the United States, we found that CRC incidence did not begin to increase for adults aged <50 years until 1981–1985. These data support the finding that the increase in CRC incidence in younger adults is attributed to risk factors, such as obesity,5O'Sullivan D.E. et al.Clin Gastroenterol Hepatol. 2021; PubMed Google Scholar that have been changing in recent birth cohorts. Before 1985 early-onset CRC incidence had been significantly decreasing for over 3 decades in women. This differs from data found in a previous study of early-onset breast cancer incidence, which has been steadily increasing since at least 1935.6Lima S.M. et al.JAMA Open Netw. 2020; 3e200929Crossref PubMed Scopus (9) Google Scholar It is not known what caused the decrease in early-onset CRC incidence in women before the mid-1980s, but possible explanations might include improvements in lifestyle factors (eg, dietary habits, reduced smoking, physical activity) and a reduction in environmental exposures (eg, Helicobacter pylori infection7Zuo Y. et al.Medicine. 2020; 99e21832Crossref PubMed Scopus (1) Google Scholar). Although advancements and uptake in screening have contributed to the decrease in CRC incidence in older adults over time, they are not likely to explain trends in adults aged <50 years who are below routine screening age. This study was limited to cancer registry data from the CTR, which may not be representative of other areas of the United States, particularly with regards to its racial composition.8Gibson C. et al.Historical census statistics on population totals by race, 1790 to 1990, and by Hispanic origin, 1790 to 1990, for the United States, regions, divisions, and states. US Census Bureau, Washington, DC2002Google Scholar However, we found that CRC incidence trends from the CTR closely aligned with incidence trends in the SEER 9 Registries. This is important given that no other cancer registry can provide historical data to evaluate CRC incidence trends before the 1970s. We note that annual incidence rates, particularly before 1968, might have included prevalent cases because of the inclusion of death certificate–only data,3Heston J.F. et al.Forty-five years of cancer incidence in Connecticut: 1935–79. US Department of Health and Human Services, Public Health Service, National Institutes of Health, National Cancer Institute, 1986Google Scholar and we cannot rule out the possibility that early trends were influenced by changes in data quality over time in the CTR. However, the steady increase in early-onset CRC occurred when >97% of cases were confirmed as incident cases through hospital records. This study was also limited in that we did not evaluate stage-specific incidence trends because of substantial missing data in early years. We also did not have data on potential risk factors, such as state-level obesity rates, going back to 1935. In conclusion, this study offers strong support that early-onset CRC did not begin to significantly increase in women or men until the 1980s. Future studies should thus consider cohort-specific trends in CRC incidence and exposures that have changed in recent birth cohorts. This includes obesity-related risk factors, given that our findings support that early-onset CRC incidence was not increasing for several decades before the rise in obesity in the United States. The marked declines in CRC incidence rates in older adults over time demonstrate that CRC incidence trends can be reversed through concerted public health efforts. Rebecca Kehm, PhD, MPH (Conceptualization: Equal; Formal analysis: Lead; Methodology: Lead; Writing – original draft: Lead). Sarah M. Lima, MPH (Data curation: Supporting; Writing – review & editing: Equal). Katrina Swett, MS (Data curation: Lead; Writing – review & editing: Equal). Lloyd Mueller, PhD (Writing – review & editing: Equal). Wan Yang, PhD (Writing – review & editing: Equal). Lou Gonsalves, PhD (Writing – review & editing: Equal). Mary Beth Terry, PhD (Writing – original draft: Supporting). We estimated the APC in CRC incidence rates using segment log-linear regression performed with Joinpoint software (version 4.6.0.0; National Cancer Institute); multiple permutation testing was used to identify inflection points in trends. Models allowed for 0 to 4 joinpoints and a minimum of 5 observations between adjacent inflection points in our primary analysis. Year of diagnosis was specified as the independent variable predicting log-transformed age-adjusted annual incidence rates stratified by sex and age group.

Topics & Concepts