This study examined the incidence rates and risk factors for colorectal cancer in 9 geographic divisions in the United States.
This study examined the incidence rates and risk factors for colorectal cancer in 9 geographic divisions in the United States.
The colorectal cancer cases were diagnosed between 1998 and 2001 in 39 states and the District of Columbia (grouped into 9 geographic divisions in the United States). The association between colorectal cancer and geographic division was analyzed using the Poisson regression model controlling for demographics and ecologic measures of education, behavioral factors and colorectal cancer screening data extracted from the Behavioral Risk Factor Surveillance System.
The age-adjusted incidence rates of colorectal cancer were highest in the Middle Atlantic division, followed by New England division, East and West North Central divisions, East South Central and South Atlantic divisions, West South Central and Pacific divisions, with the lowest rate observed in the Mountain division. Old age, male gender, black race, less than a twelfth-grade education, smoking, and no physical activity were significantly associated with higher incidence rates of colorectal cancer, whereas having sigmoidoscopy/colonoscopy in the past 5 years, fecal occult blood test in the past year, and obesity were associated with lower incidence rates of colorectal cancer. The relative ranking of incidence rates of colorectal cancer across divisions changed after adjusting for these factors.
Significant geographic variation in colorectal cancer exists in the United States. Risk factors, including demographics, education, behavior, and screening use, can only partially explain the differences across geographic divisions. Cancer 2006. © 2006 American Cancer Society.
Epidemiologic, genetic, and experimental studies have suggested that colorectal cancer results from complex interactions between inherited susceptibility and environmental factors.1–10 Previous research has shown that there are variations in colorectal cancer incidence rates with respect to age, gender, race, and access to health care.11–15 Colorectal cancer incidence rates increase with advancing age, with the highest age-specific incidence rates per 100,000 persons observed in males and females aged 70 years and older. Males generally have a higher incidence rate than females, and blacks experience a higher incidence rate than their counterpart white population. Previous studies indicate that a higher risk of colorectal cancer may be associated with obesity, excessive dietary fat consumption, high meat and calcium intake, use ofpostmenopausal female hormone supplements, lack of physical activity, alcohol consumption, overuse of certain vitamins, and cigarette smoking, in addition to older age, male gender, and black race.16–18 These factors, however, may not completely explain the gender or racial variation associated with colorectal cancer.
Age-adjusted incidence rates of colorectal cancer also vary geographically.4, 19–21 Generally, incidence rates of colorectal cancer are higher in the East than in the West. Data from the United States Cancer Statistics of 2000 showed an age-adjusted rate of 48.8 cases of colorectal cancer per 100,000 males in New Mexico, whereas the incidence rate in Rhode Island was nearly 75% higher at 79.4 cases per 100,000 males. Unfortunately, previously published studies on geographic variation are limited, as they cover only small areas and/or single years.
The study describes geographic differences in the incidence of colorectal cancer, using data collected over a period of 4 years by the Centers for Disease Control and Prevention National Program of Cancer Registries (NPCR) and the National Cancer Institute's Surveillance, Epidemiology, and End Results registries (SEER).19, 22 This study also examines whether other potential factors might explain the geographic variation of colorectal cancer, including age, gender, race, and population prevalence of risk factors characterized by the Behavioral Risk Factor Surveillance System (BRFSS).23
This study included 538,237 invasive colorectal cancers with the International Classification of Diseases for Oncology third Edition (ICD-O-3) topography codes (C18.0-C18.9, C19.9, C20.9, and C26.0) that were first diagnosed during the period between 1998 and 2001.24 Only data that were obtained from the state central registries participating in the NPCR and the SEER Program were included in the analysis.25 Data from the SEER Atlanta registry (which is not a statewide registry for Georgia) were therefore not included in the analysis. For data from state central registries to be included in this analysis, statewide registries must also have met data quality standards for all invasive cancer sites combined.19 Registries in 39 states and the District of Columbia (representing 87% of the U.S. population) met these criteria. In addition to the District of Columbia cancer registry, the 39 state cancer registries include Alabama, Arizona, Alaska, California, Colorado, Connecticut, Florida, Hawaii, Idaho, Illinois, Indiana, Iowa, Kansas, Kentucky, Louisiana, Massachusetts, Maine, Michigan, Missouri, Minnesota, Montana, North Carolina, Nebraska, New Jersey, New Mexico, New York, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Texas, Utah, Vermont, Washington, Wisconsin, West Virginia, and Wyoming. Data also include further classification of the cancers using SEER summary stage categories: localized (confined to colon or rectum), regional (extension to adjacent tissues or regional lymph nodes), distant (metastasis to other areas of the body), and unknown stage.26 Because complete summary stage data for California were not available, the staging data from the Pacific division were not included in the analysis involving stage data.
Demographic data for this study were also obtained from NPCR and SEER registries. Education information that is age-, gender-, and state-specific was obtained from the United States Census Bureau American Community Survey Summary.27 Specifically, the proportion of persons aged 18 years and older with less than a twelfth-grade of education within each 5-year age group and for both genders was extracted for all states and included in the analysis. Finally, data concerning the prevalence of risk factors, health care access, and use of colorectal cancer screening were obtained from the BRFSS survey.23 These BRFSS variables are state-level data and include the following:
Geographic divisions within the United States were defined by modifying the United States Census Bureau's divisional grouping system28 to include only the 39 states and the District of Columbia registries that met the cancer incidence data quality standard. These divisions are listed below:
New England: Connecticut, Maine, Massachusetts, Rhode Island, and Vermont
Middle Atlantic: New Jersey, New York, and Pennsylvania
South Atlantic: District of Columbia, Florida, North Carolina, South Carolina, and West Virginia
East South Central: Alabama and Kentucky
West South Central: Louisiana, Oklahoma, and Texas
East North Central: Illinois, Indiana, Michigan, Ohio, and Wisconsin
West North Central: Iowa, Kansas, Minnesota, Missouri, and Nebraska
Mountain: Arizona, Colorado, Idaho, Montana, New Mexico, Utah, and Wyoming
Pacific: Alaska, California, Hawaii, Oregon, and Washington
The percentage of the population covered by each division is shown in Table 1. Age-adjusted incidence rates, the distribution of stage of disease at diagnosis, and the BRFSS survey results were calculated by geographic division. All incidence rates were age-adjusted to the 2000 U.S standard population, using 19 age categories. Race was categorized as white, black, and other, because of the small number of colorectal cancers in some divisions for Asian/Pacific Islanders, American Indians, and Alaska natives. The age-adjusted rates and corresponding 95% confidence intervals were calculated using SEER*Stat software.22 The prevalence estimates and the corresponding standard errors of the BRFSS variables were generated with sampling weights using SUDAAN version 8.0 software to reflect underlying population differences among states.29 Details concerning the sampling weights calculation are described elsewhere.23
|Modified division*||Population size||Population coverage, %||Number of cases†||Rate per 100,000‡||95% CI|
|East South Central||33,876,820||54.4||19,036||55.5||54.69||56.27|
|West South Central||114,649,599||91.5||50,579||51.3||50.81||51.71|
|East North Central||180,372,917||100.0||104,087||57.9||57.54||58.24|
|West North Central||71,180,592||95.5||42,902||57.7||57.10||58.20|
Poisson multivariate regression with incidence rates as the dependent variable was used to examine differences in the effect of the geographic division on the incidence of colorectal cancer rates while adjusting for demographics and ecologic BRFSS variables, including behavioral risk factors and colorectal cancer screening data.30 The rate ratio and 95% confidence intervals derived from the Poisson regression modeling using the SAS GENMOD procedure (SAS Institute, Inc., Gary, NC) were calculated for all independent factors and used to describe the difference in the colorectal cancer incidence rates between any 2 comparison groups/divisions. The potential interaction between race and division was not tested because of small numbers in some race and division combination groups. The data included in the Poisson regression were grouped by 5-year age group, gender, race, and state, although state-specific BRFSS data were applied uniformly to all age-, gender-, and race-combinations within a state. The BRFSS variables were analyzed as continuous variables. To avoid multicollinearity between the variables ‘ever had FOBT’ and ‘had FOBT in past year,’ as well as between ‘ever had sigmoidoscopy/colonoscopy’ and ‘had sigmoidoscopy/colonoscopy in the past 5 years,’ 2 separate Poisson regression models were analyzed. One model included variables for ‘ever had FOBT’ and ‘ever had sigmoidoscopy/colonoscopy,’ and the other model included ‘had FOBT in past year’ and ‘had sigmoidoscopy/colonoscopy in the past 5 years.’ Because BRFSS questions on use of colorectal cancer screening focused on persons aged 50 years and older, the Poisson regression was performed for all ages in the dataset as well as for persons aged 50 years and older only.
A total of 538,237 invasive colorectal cancers were included in this study. The age-adjusted incidence rates of colorectal cancer were highest in the Middle Atlantic division, followed by the New England division, East and West North Central divisions, East South Central and South Atlantic divisions, West South Central and Pacific divisions, with the lowest rate observed in the Mountain division (Table 1).
The age-adjusted incidence rates of colorectal cancer diagnosed in males and females in each of the 9 U.S. divisions are shown in Figure 1. Males had significantly higher incidence rates of colorectal cancer compared with females across all divisions (P < 0.0001). Age-adjusted incidence rates in males ranged from 54.8 (95% CI, 53.9–55.7) cases per 100,000 persons in the Mountain division to 74.9 (95% CI, 74.2–75.5) cases per 100,000 persons in the Middle Atlantic division. The overall rates by division for females ranged from 39.6 (95% CI, 39.0–40.3) cases per 100,000 persons in the Mountain division to 53.6 (95% CI, 53.1–54.0) cases per 100,000 persons in the Middle Atlantic. As with males, lower rates among females were noted in the Mountain and Pacific divisions, and higher rates were observed in the Middle Atlantic and New England divisions.
Figure 2 presents colorectal cancer incidence rates by division among whites, blacks, and other races. In general, blacks had significantly higher incidence rates of colorectal cancer (P < 0.0001) compared with whites, but the higher incidence rates in blacks than in whites were not observed across all divisions. The highest rate for blacks, 67.5 (95% CI, 64.4–70.7) cases per 100,000 persons, occurred in the West North Central division, whereas a high of 62.5 (95% CI, 62.0–62.9) cases per 100,000 was reported for whites in the Middle Atlantic division. Both black and white incidence rates were lowest in the Mountain division, with 53.6 (95% CI, 48.5–58.6) and 46.5 (95% CI, 45.9–47.1) cases per 100,000 persons, respectively. The age-adjusted incidence rates for other races combined (including Asian/Pacific Islanders, American Indians, and Alaska Natives) were the lowest in all 3 racial groups in all divisions. As expected, the corresponding 95% confidence intervals for this combined race category were wide as a result of the small sample size. Significant geographic variations in incidence rates of colorectal cancer were also observed for persons of the same race.
The distribution of stages of colorectal cancers atdiagnosis was significantly different across geographic divisions (Fig. 3; P < 0.0001). The distribution of stage at diagnosis was similar among the New England, East South Central, West North Central, and Mountain divisions with higher percentages of local stage disease and lower percentages of unstaged disease than the other divisions. In all divisions, the percentage of distant stage disease was about 16%.
Table 2 presents descriptive data from the BRFSS survey results across the 9 U.S. divisions. The New England and Mountain divisions had the lowest prevalence of obesity (18% and 19%, respectively), and the East South and West South Central divisions had the highest prevalence (25% and 24%, respectively). The percentage of residents with no physical activity was lowest in both Mountain and Pacific divisions (10%), and the East and West South Central divisions had the highest percentages with no physical activity (25% and 24%, respectively). Finally, the prevalence of current smokers was similar throughout the country, ranging from 27% in the East South Central division to 21% in both the Mountain and Pacific divisions. There was wide variation in the proportion of those aged 50 years or older who ever had a FOBT, ranging from 50% in the New England division to 37% in the East South Central division. Of those aged 50 years or older who resided in the New England division, 51% had ever had a sigmoidoscopy/colonoscopy, whereas only 40% of those aged 50 years or older living in the West South Central division had ever had a sigmoidoscopy/colonoscopy.
|Modified division†||Obesity (BMI ≥ 30)||No physical activity||Smoker||Ever had FOBT‡||FOBT in past year‡||Ever had sigmoidoscopy colonoscopy‡||Sigmoidoscopoy/colonscopy in past 5 Yr‡|
|New England||18 ± 1.02||13 ± 0.50||22 ± 1.97||50 ± 3.69||31 ± 3.12||51 ± 2.62||43 ± 3.09|
|Middle Atlantic||21 ± 1.28||16 ± 2.36||23 ± 1.69||44 ± 1.36||24 ± 0.51||47 ± 0.69||39 ± 1.88|
|South Atlantic||22 ± 2.49||16 ± 1.55||25 ± 2.99||47 ± 6.07||27 ± 4.91||47 ± 6.03||39 ± 6.77|
|East South Central||25 ± 0.09||23 ± 10.93||27 ± 5.00||39 ± 3.25||14 ± 10.21||42 ± 3.80||35 ± 2.85|
|West South Central||24 ± 1.01||21 ± 5.83||25 ± 3.19||37 ± 1.61||20 ± 2.27||40 ± 2.25||32 ± 1.94|
|East North Central||23 ± 1.63||13 ± 3.31||26 ± 1.94||46 ± 4.33||23 ± 1.68||49 ± 6.98||39 ± 6.37|
|West North Central||22 ± 1.35||16 ± 6.74||23 ± 2.07||49 ± 4.09||24 ± 1.14||47 ± 8.99||38 ± 8.6|
|Mountain||19 ± 1.82||10 ± 2.16||21 ± 3.50||43 ± 6.25||20 ± 4.72||46 ± 2.37||35 ± 1.68|
|Pacific||22 ± 1.80||10 ± 3.03||21 ± 3.27||48 ± 9.66||25 ± 6.92||50 ± 3.41||40 ± 3.03|
The adjusted rate ratios for old age (data not shown), black race, less than a twelfth-grade education, smoker, and no physical activity were all greater than 1, indicating that these factors were associated with higher incidence rates of colorectal cancer, whereas female gender, having sigmoidoscopy/colonoscopy in the past 5 years and FOBT in the past year, and obesity were associated with lower incidence rates of colorectal cancer (Table 3). The Poisson regression analyses confirmed that the geographic variation in colorectal cancer incidence across divisions persisted, even after controlling for age, gender, race, and ecologic measures of education and behavioral factors. However, the relative ranking of incidence rates of colorectal cancer across divisions changed after adjusting for independent factors, as shown by the magnitude of the rate ratio. For example, the colorectal cancer incidence rate in the New England division was 1.07 times higher than that in the Pacific division (P < 0.0001). However, the incidence rate in the New England division was not significantly different from that observed in the Middle Atlantic division.
|Variable||Rate ratio†||95% CI||P|
|East South Central||0.91c||0.88–0.95||<0.0001|
|West South Central||0.95c||0.93–0.97||<0.0001|
|East North Central||0.97||0.96–0.99||0.0038|
|West North Central||1.00b||0.98–1.01||NS|
|Pacific||1.00 (reference group)||—||—|
|Male||1.00 (reference group)||—||—|
|White||1.00 (reference group)||—||—|
|12th–Grade education or less‖||1.02||1.01–1.03||0.0043|
|No physical activity||1.06**||1.04–1.07||<0.0001|
|FOBT in past year||0.97**||0.95–0.98||0.0045|
|Sigmoidoscopy/colonoscopy in past 5 yr||0.97**||0.96–0.98||<0.0001|
The geographic distribution of the incidence rates of colorectal cancer remained similar when ‘having had a sigmoidoscopy/colonoscopy in past 5 years’ and ‘having had FOBT in the past year’ were replaced by ‘ever had FOBT’ and ‘ever had sigmoidoscopy/colonoscopy’ (results not shown). Similarly, the geographic distribution across divisions remained unchanged when only persons aged 50 years and older were included in the Poisson regression analysis (results not shown).
Significant geographic variation in the incidence of colorectal cancer exists in the United States. Our study also showed that old age, male gender, black race, less than a twelfth-grade education, and no physical activity were all associated with higher incidence rates of colorectal cancer. Yet, significant differences in colorectal cancer incidence rates continued to be present across the 9 U.S. divisions after controlling for these factors. The Northeast (namely the Middle Atlantic and the New England divisions) had the highest age-adjusted rates, and the Mountain division had the lowest age-adjusted rates of colorectal cancer for the diagnosis years during 1998 and 2001.
Because incidence rates were age-adjusted, age is unlikely to be a factor that can be used to explain the geographic variation noted across the divisions. Males typically had higher age-adjusted incidence rates of colorectal cancer than did females across all divisions.1, 2, 13–15 Blacks, as has been shown in many other studies, experienced higher rates of colorectal cancer compared with their counterparts in the white population.11–15 The racial difference, however, does not explain why the Middle Atlantic and New England divisions have the highest colorectal cancer burden among the 9 divisions. When only whites are considered, the New England and Middle Atlantic divisions continue to have the highest incidence rates of colorectal cancer. The New England division has a relatively small black population, yet had a colorectal cancer incidence rate similar to that in the Middle Atlantic division. Furthermore, the black population in the Middle Atlantic division did not have the highest burden of colorectal cancer compared with black populations in other divisions. In fact, the West North Central (Nebraska, Kansas, Missouri, Iowa, and Minnesota) and the East North Central (Wisconsin, Michigan, Ohio, Indiana, and Illinois) divisions had the highest incidence of colorectal cancers among blacks. It is noteworthy that the divisions in which whites had the highest incidence rates were not consistent with those divisions in which blacks were shown to have the highest rates of colorectal cancer. Because of the small numbers of other races (e.g., Asians and American Indians) in some divisions, it is difficult to derive the inference on geographic variation in this subpopulation across the U.S. divisions. Because race was adjusted for in the analysis, it is unlikely that black race alone is the determining factor in the geographic differences in colorectal cancer incidences across the United States.
This study has also observed some differences in various assessments of behavioral factors across the 9 divisions. The Mountain division had the lowest incidence of colorectal cancer and a very low percentage of the population that was obese, engaged in no physical activity, and smoked. Conversely, the Middle Atlantic division, the geographic area that was observed to have the highest incidence rate of colorectal cancer, had a higher percentage of the population that was obese, engaged in no physical activity, and smoked. Although these correlations and data from previous research1, 9, 31–33 suggest that obesity, lack of physical activity, and smoking increase the risk for colorectal cancer, the results for the New England division, the East and West South Central divisions call this hypothesis into question. The New England division, which was ranked the second highest in the incidence of colorectal cancer, had the lowest prevalence of obesity in the nation and a relatively low prevalence of no physical activity. The East and West South Central divisions had the highest percentage of the population that was obese, engaged in no physical activity, and smoked, yet they had the second and third lowest colorectal cancer burden (after the Mountain division). Our observation that obesity was associated with lower incidence rates of colorectal cancers is inconsistent with the literature.1, 19, 31–33 This finding may be an artifact due to ecologic measurement of this factor. The question of whether the observed low use of colorectal cancer screening in both the East and West South Central divisions was associated with their low incidence rates warrants continued surveillance. The low use of screening could suggest that the unusually low burden of colorectal cancer in this division may be dueto under-diagnosis of the disease, particularly because the population in this division has the highest proportion of blacks. A recent study of more than 5500 adults aged 50 and older14 showed that blacks and Hispanics were less likely than whites to have been screened, after controlling for age, gender, access to care (income and insurance), family cancer history, smoking, and obesity. It will be important to monitor whether the incidence of colorectal cancer rises as screening for colorectal cancers increases inthe division. Also, the question of whether a relatively high utilization of colorectal cancer screening observed in the New England division can explain the observed high incidence rate of colorectal cancer remains to be tested and is described below.
Although screening and early diagnosis of colorectal cancers in the New England division seem to contribute to the elevated age-adjusted incidence rates of colorectal cancer in this division,3 the similarity in the distributions of stages at diagnosis in this division with other divisions weakens conclusions concerning the role of screening and early diagnosis. Use of screening would initially increase colorectal cancer rates by identifying subclinical cases, but, in the long run, colorectal cancer rates should be reduced because of the detection of premaligant disease that would have been treated after identification.2 In fact, this study showed that use of FOBT in the past year and the use of sigmoidoscopy/colonscopy in the past 5 years were associated with lower incidence rates. Because colorectal cancer screening data were not available for all states until the 2001 BRFSS survey, it is unlikely to sort out the dynamic impact resulting from screening use in various communities. Furthermore, the BRFSS questions related to the use of colorectal cancer screening were only applied to respondents aged 50 years and older. However, in separate analyses, where we restricted the analysis to include only individuals aged 50 years and older, the relative geographic variation across divisions remained unchanged.
Although this study covers a large proportion of the U.S. population, it also has some limitations. For some divisions only a few states were included in the study. For example, Delaware, Georgia, Maryland, and Virginia, which are part of the South Atlantic division, were not included because they did not meet the inclusion eligibility criteria. Our study results may not be generalizable to those states that were not included in the study, but the results are generalizable to the 87% of the U.S. population that was included. Another drawback of our study is that more precise measures of socio-economic status (SES), such as income and/or occupation, were not available. Although educational status is often used as one dimension of SES, it does not always capture the whole spectrum of SES, including lifestyle factors, exposure to toxins, or the interaction between race/ethnicity and social class. In addition, ethnicity (Hispanic and non-Hispanic) was not considered in this study because of small numbers in some regions and the inconsistency in the definition of Hispanic ethnicity used. This factor may be important in explaining divisional differences in colorectal cancer burden.
For years, the colon cancer mortality rate in the Northeast was known to be elevated and was largely attributable to higher socioeconomic levels among urban populations (possibly related to dietary and nutritional factors).1 This study confirms that the Northeast inherently has the highest incidence of colorectal cancers in the United States despite adjustment for various factors, including age, gender, race, behavioral factors, and access to early screening and diagnosis. Unfortunately, differences in dietary habits (including fat and fiber consumption) across divisions were not included in this analysis because of incomplete BRFSS data for 30 states. As research continues to suggest associations between dietary fat and fiber consumption and colorectal cancer risk, data on fat consumption in population are needed.
All data included in this study were collected using uniform items and codes, thereby ensuring a reasonable degree of comparability at every level. Nevertheless, incidence rates may be partially influenced by discrepancies in timeliness, completeness, and accuracy among registries or from one reporting period to the next. Furthermore, because state registries continually update previously submitted data, national totals are typically in a constant state of flux. Finally, a fundamental concern in this study is the use of BRFSS state-level ecologic data for lifestyle factors and screening use; these observations may not occur at the individual level.
In conclusion, this study has greater coverage of the U.S. population than did previous studies that examined the geographic variation of colorectal cancer incidence in the United States. Including cases diagnosed over a period of 4 years further stabilized the estimates of colorectal cancer incidence rates. Our study also provides stronger evidence to confirm that geographic variation in colorectal cancer incidence exists across the United States, and it underscores the need for continued surveillance and research of population-level data for the purpose of explaining geographic differences in colorectal cancer.
We appreciate the in-kind support from all the contributors to this monograph and also are grateful for the contributions of Jessica King for the preparation of analytic files and to Faruque Ahmed for leading the colorectal cancer monograph project.