Decrease in incidence of colorectal cancer among individuals 50 years or older following recommendations for population-based screening

Abstract

Background & Aims

The incidence of colorectal cancer (CRC) in the U.S. is increasing among adults younger than age 50 years, but incidence has decreased among older populations after population-based screening was recommended in the late 1980s. Blacks have higher incidence than whites. These patterns have prompted suggestions to lower the screening age for average-risk populations or in blacks. At the same time, there has been controversy over whether reductions in CRC incidence can be attributed to screening. We examined age- and race-related differences in CRC incidence over a 40-year time period.

Methods

We determined the age-standardized incidence of CRC, from 1975 through 2013, using the population-based Surveillance, Epidemiology, and End Results (SEER) program of cancer registries. We calculated incidence for 5-year age categories (20—24 years through 80—84 years and 85 years or older) for different time periods (1975—1979, 1980—1984, 1985—1989, 1990—1994, 1995—1999, 2000—2004, 2005—2009, and 2010—2013), tumor subsite (proximal colon, descending colon, and rectum), and stages at diagnosis (localized, regional, and distant). Analyses were stratified by race (white vs. black).

Results

There were 450,682 incident cases of CRC reported to the SEER registries over the entire period (1975—2013). Overall incidence was 75.5/100,000 white persons and 83.6/100,000 black persons. CRC incidence peaked during 1980 through 1989 and began to decrease in 1990. In whites and blacks, the decreases in incidence between the time periods of 1980—1984 and 2010—2013 were limited to the screening-age population (ages 50 years or older). Between these time periods, there was a 40% decrease in incidence among whites compared with a 26% decrease in incidence among blacks. Decreases in incidence were greater for cancers of the distal colon and rectum, and reductions in these cancers were greater among whites than blacks. CRC incidence among persons younger than 50 years decreased slightly between 1975—1979 and 1990. However, among persons 20—49 years old, CRC incidence decreased from 8.3/100,000 persons in 1990—1994 to 11.4/100,000 persons in 2010—2013; incidence rates in younger adults were similar for whites and blacks.

Conclusions

Based on an analysis of the SEER cancer registries from 1975 through 2013, CRC incidence decreased only among individuals 50 years or older between the time periods of 1980—1984 and 2010—2013. Incidence increased modestly among individuals 20—49 years old between the time periods of 1990—1994 and 2010—2013; the decision of whether to recommend screening for younger populations requires a formal analysis of risks and benefits. Our observed trends provide compelling evidence that screening has had an important role in reducing CRC incidence.

Background

Colorectal cancer (CRC) incidence and mortality in the U.S. have changed strikingly in recent decades.¹ Overall CRC incidence declined by more than 30% from 1975 (59.5 per 100,000) to 2013 (37.9 per 100,000),² with particularly steep declines among those over the age of 65. Mortality rates have similarly declined during the same period.¹

The declines in CRC incidence have not occurred equally in all populations. In sharp contrast to the decline in CRC incidence and mortality among older individuals, incidence is actually rising in adults younger than age 50.^3–11 These findings have led to recommendations to extend screening among average-risk individuals to ages below 50 years.^12–14 Because blacks have higher CRC incidence than whites, and the age-related acceleration in incidence starts at a younger age, calls have also been made to lower the screening age for blacks.^15–18

At the same time, there has been controversy regarding the role of screening in the reductions in CRC incidence. Unlike screening for other cancers (e.g., breast, prostate), CRC screening reduces incidence via excision of premalignant lesions. The use of CRC screening has become increasingly common in the U.S. since it was first formally recommended in the late 1980s (largely driven by the use of colonoscopy),^19–22 but the extent to which it explains declines in CRC incidence has been debated. Some have argued that screening accounts for much of the improvement in incidence and mortality, as observed in clinical trials of fecal occult blood testing (FOBT) and sigmoidoscopy.^23–25 However, others have suggested that screening has had only a modest impact, and changes in the prevalence of risk factors may be more important.^{26, 27}

The trends in CRC incidence raise two related questions: what is the role of screening in explaining the declining rates? And which populations should be screened? To clarify these questions and better understand the clinical and public health implications of CRC incidence trends, we examined age- and race-related differences in incidence over a 40-year time period.

Methods

Incidence of invasive CRC was derived from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program during 1973–2013. SEER routinely collects data on patient demographics, primary tumor site, tumor morphology, and stage for all cancers diagnosed in defined geographic regions. The SEER 9 registries cover approximately 10% of the U.S. population and include Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Puget Sound, and Utah. Age-adjusted incidence (using the 2000 U.S. standard population) was obtained using SEER*Stat version 8.2.1 as incidence rates per 100,000 persons.

To account for differences in incidence by age, we estimated incidence rates across 5-year age categories (20–24 to 80–84 and 85+) by time period, tumor subsite, and stage at diagnosis. We divided the study period into eight approximate 5-year time periods (1975–79, 1980–84, 1985–89, 1990–94, 1995–99, 2000–04, 2005–09, 2010–13). Tumor subsites included proximal colon (cecum, ascending colon, hepatic flexure, transverse colon), descending colon (splenic flexure, descending colon, sigmoid colon), and rectum (rectosigmoid junction, rectum) according to the International Classification of Disease for Oncology, 3^rd edition (ICD-O-3). Stage at diagnosis was based on SEER summary staging. Localized disease was defined as being limited to the large bowel, regional as limited to nearby lymph nodes or other organs, and distant disease as systemic metastasis.

To illustrate trends, we plotted incidence rates by age, contrasting time periods, tumor subsites, and stages at diagnosis. All analyses were stratified by race (white vs. black). Corresponding 95% confidence intervals were calculated as modified gamma intervals.¹⁸

This study was approved by the Institutional Review Board at the University of North Carolina at Chapel Hill (#15-1957).

Results

There were 450,682 incident cases of CRC diagnosed over the entire period 1975–2013. Overall incidence was 75.5 per 100,000 and 83.6 per 100,000 among whites and blacks, respectively. Detailed incidence rates by age, time period, tumor subsite, and stage at diagnosis are available in Supplementary Table 1 (whites) and Supplementary Table 2 (blacks).

Incidence peaked during 1980–89, with a subsequent decline beginning around 1990. In both whites (Figure 1A) and blacks (Figure 1C), the declines in incidence between 1980–84 and 2010–13 were limited to the screening-age population. Incidence curves diverged sharply from previous years, but only at ages 50 years and higher. However, whites experienced a greater decline: a 40% decrease (from 90.4 to 54.0 per 100,000) versus only 26% (from 90.7 to 67.2 per 100,000) among blacks. For persons aged 65–69 years, CRC incidence in whites decreased by 51% (from 228.7 to 113.2 per 100,000) from 1980–84 to 2010–13 compared to a 19% decrease (from 214.2 to 173.2 per 100,000) for blacks.

Figure 1. Age-adjusted (2000 U.S. standard population) incidence of colorectal cancer by age group and time period (1980–84 vs. 2010–13a) for whites and blacks, SEER 9, 1975–2013.

^aWe chose to compare incidence rates from the most recent time period (2010–13) to 1980–84 in the figure because incidence peaked in 1980–84 and subsequently declined through to 2013

Shaded regions in figure denote 95% confidence intervals

In younger populations (<50 years), there were hints of decreases in incidence between 1975–79 and about 1990, differing by 5-year age group and time period. Subsequently, in marked contrast to populations over age 50, there were continued increases through to 2010–13. In the whole population aged 20–49 years, rates increased from 8.3 per 100,000 in 1990–94 to 11.4 per 100,000 in 2010–13. This difference of 3.1 per 100,000 was a 37% increase, similar in whites (Figure 1B) and blacks (Figure 1D).

Declines in CRC incidence by tumor subsite were also limited to screening-age populations. At ages over 50, the incidence of distal colon and rectal cancers decreased substantially from 1980–84 to 2010–13 for both whites and blacks (Figure 2). Declines in the incidence of proximal colon cancer were smaller, and for blacks, incidence differed little over time. Although blacks consistently had higher overall (and age-specific) CRC rates than whites, they had lower rectal cancer rates overall (19.8 per 100,000 in blacks vs. 21.5 per 100,000 in whites) and in older age groups (Supplementary Tables 3 and 4)

Figure 2. Age-adjusted (2000 U.S. standard population) incidence of proximal colon, descending colon, and rectal cancer by age group and time period (1980–84 vs. 2010–13) for whites and blacks, SEER 9, 1975–2013.

Proximal colon includes the cecum, ascending colon, hepatic flexure, and transverse colon; descending colon includes the splenic flexure, descending colon, and sigmoid colon; rectum includes the rectosigmoid junction and rectum.

Shaded regions in figure denote 95% confidence intervals

Among whites, incidence of local, regional, and distant disease each decreased from 1980–84 to 2010–13 (Supplementary Figure 1), only at ages 50 and older. There was very little decrease in incidence of local disease among blacks (and limited to the oldest ages). For the screening-age population, the absolute difference in incidence of distant disease between whites and blacks was generally about 15 per 100,000. The higher incidence of metastatic disease among blacks accounted for essentially all of the difference in overall CRC rates between whites and blacks (Supplementary Tables 5 and 6).

Discussion

Trends in incidence across age, race, subsite, and stage provide compelling evidence of the impact of screening on CRC incidence in the U.S. Specifically, the dramatic decline in CRC incidence from its peak in the 1980s to 2010–13 occurred exclusively among individuals older than age 50: the recommended age for the initiation of CRC screening in average-risk populations. Moreover, incidence rates have fallen most prominently in the distal colorectum, where screening most reduces cancer incidence.^28–31

The differences in CRC incidence by race offer additional support for the impact of screening. As noted by others,^32–37 blacks have experienced smaller declines in incidence over time (particularly for localized disease), as would be expected in a population that historically has had lower screening rates than whites.^{19–22, 38, 39} That blacks presented more frequently with metastatic disease, a difference that accounted for much of the differences in the overall rates with whites, may also be related to lower use of screening. In addition, there is evidence that blacks have been screened less effectively than whites: more frequently with FOBT and less frequently with colonoscopy.¹⁹ The effectiveness of FOBT may be reduced when patients do not adhere to a regular schedule.^40–43 Evidence is limited concerning differences in adherence by race, but the prevalence of repeat FOBT remains low, between 14 and 54%.^44–47 FOBT or sigmoidoscopy screening alone may also miss proximal cancers, which are more common in blacks.^{15, 48–52} Our findings support efforts to target screening programs in the underserved and the need for tailored interventions that address minority populations.⁵³

There were no continued declines in CRC incidence among younger adults, a population that is not regularly screened. The relative increases in incidence summarized above, combined with the benefits of screening observed in older populations, have led some to promote changing screening guidelines to begin average-risk screening at a younger age, such as age 40,^{13, 14} or in blacks.^15–18 Indeed, the relative increases in incidence rates in younger populations are dramatic, and the temptation is to believe the clinical implications are equally important. These relative increases, however, are based on an absolute increase of only a few additional cases per 100,000 persons over a 20-year time period. For example, at age 45–49, the relative increase in incidence from 1990 to 2013 is nearly 30%, but the absolute difference over the same time period is a modest 6.4 cases per 100,000. The differences are even smaller for younger ages. Consequently, it is not clear whether it is appropriate to lower the recommended age at screening. To ensure a net benefit, such a recommendation should assess all benefits and harms in absolute terms, not solely on relative changes in one outcome, CRC incidence.⁵⁴ These considerations also pertain to the question of whether blacks should be recommended to start CRC screening at an earlier age.

There are factors other than screening that could account for some of the reductions in CRC incidence. The large declines in certain CRC risk factors (e.g., smoking,⁵⁵ red and processed meat consumption⁵⁶) have likely contributed to lower incidence rates. Yet obesity, a clear risk factor for CRC,⁵⁷ has been increasing in prevalence⁵⁸ while CRC incidence continues to decline in adults of screening-age. These temporal trends in risk factors are not biologically consistent with the divergent incidence trends in younger and older populations. It is unlikely that risk factor trends would parallel the declines in incidence among older populations but be inversely related to incidence rates in younger (age <50) adults. Changes in the prevalence of risk factors across the U.S. population may play only a minor role in CRC incidence patterns.

Concerns have also been raised regarding the timing of the decline in CRC incidence relative to recommendations for population-based screening.²⁷ In the early 1980s, scientific evidence began supporting the effectiveness of population-based screening with FOBT or sigmoidoscopy, based on findings from feasibility studies in the U.S.⁵⁹ and European trials.⁶⁰ Screening targets were subsequently included in the National Cancer Institute’s 1987 working guidelines on the early detection of cancer.⁶¹ Uptake of CRC screening has been slow compared to other cancer screenings (e.g., breast, cervical). However, national surveys⁶² as early as 1982 show nearly half (44%) of men and women had ever received FOBT. Our finding that the incidence of distal colon and rectal cancer started to decline around 1985–89 is consistent with this early adoption of stool-based screening tests. The decline in the incidence of proximal cancer did not start until the early 2000s (and in 2005–09 for blacks; see Supplementary Tables 3 and 4), which may correspond to more recent trends favoring colonoscopy as the preferred screening modality.²⁹

Others²⁷ describing declines in CRC mortality (vs. incidence) over a similar time period have commented that screening does not fully explain recent trends. We have discussed the effects of CRC “screening,” but it is important to keep in mind the distinction between early cancer detection and the removal of premalignant lesions through polypectomy. An integral component of the CRC screening program in the U.S., polypectomy affects incidence to a greater extent than it does mortality. CRC mortality, on the other hand, has a broader set of influences than incidence, such as treatment advances or earlier evaluation of symptomatic cancers. We focused exclusively on patterns of CRC incidence to provide a clear picture of the impact of screening.

There are some limitations of cancer incidence data, which may not be uniformly accurate for all populations (including racial/ethnic minorities) and geographic regions. We did not have the ability to examine incidence rates in populations other than white and black race. Although CRC incidence is generally lower in Asian and Hispanic populations, some evidence shows rates are increasing among them.^{63, 64} We also had incomplete data on stage and anatomic subsite. In addition, cancer registries not provide information about the prevalence of risk factors or screening in the population, limiting the ability to study incidence patterns in conjunction with secular trends.

In summary, patterns of CRC incidence point to screening as an important underlying explanation for the large declines in incidence. It is unlikely that the declining incidence rates are due to changes in the prevalence of risk factors over time. Reductions in incidence begin around age 50, are more prominent in the distal colorectum, and are less marked in blacks. These trends suggest screening has played a dominant role in declines in CRC incidence. As incidence in older adults has declined, the proportion of cancers diagnosed under age 50 has increased, apparently alarmingly. The absolute increase in incidence before age 50 is modest, and the question of screening in younger populations requires a formal analysis of risks and benefits. However, it is clear that achieving the public health goal of 80% screening adherence in age-eligible populations^{65, 66} could avert over 200,000 new cancer cases and deaths in the next 20 years.⁶⁷ Continued work to improve the uptake of CRC screening in older and minority populations may result in further improvements in CRC outcomes.

Abbreviations

CRC

colorectal cancer

FOBT

fecal occult blood testing

SEER

Surveillance, Epidemiology, and End Results

ICD-O-3

International Classification of Disease for Oncology, 3^rd edition

SEER

Surveillance, Epidemiology, and End Results