Presented to the Science Writers' Seminar of the American Cancer Society, San Francisco, California, March 26, 1996.
Characteristics that determine the anatomic site within the colorectum where carcinoma is most likely to occur would be useful in choosing optimal modalities for cancer screening and in the study of etiology. The aim of this study was to identify any association of gender, race, or age with colorectal carcinoma in the proximal or distal colorectum.
A descriptive, population-based epidemiologic study was conducted on colorectal carcinoma (CRC) cases identified through the Illinois Tumor Registry. Among CRC patients diagnosed between 1986 and 1991, age at diagnosis (40-90 years), gender, white or African American race, and subsite of CRC location were identified. Incidence rates were determined with demographic data from the 1990 Illinois census. Logistic regression was used to identify significant subsite specific risk factors for CRC.
During the study period, 38,931 cases of CRC occurred in Illinois among whites and African Americans between the ages of 40 and 90 years. Age, race, and gender were independently significantly associated with both proximal and distal CRC risk (P < 0.0001). The greatest risk for white males was for distal CRC, followed by the risk for African American males of developing CRC in the distal colorectum. African Americans were more likely than whites to develop proximal CRC (odds ratio [OR] = 1.19), whereas whites were more likely to have distal CRC than African Americans (OR = 1.26). Regarding the development of CRC anywhere in the colorectum, male incidence rates adjusted for age and race were greater than rates for females (OR = 1.32 for proximal and 1.68 for distal).
A recent analysis of colorectal carcinoma (CRC) screening strategies by the Office of Technology Assessment of the U.S. Congress has demonstrated that the most cost-effective means of screening average-risk adults for CRC are either fiberoptic flexible sigmoidoscopy or double contrast barium enema.1 Not addressed in that report was the potential for further increasing the effectiveness of each of these strategies by offering barium enema to individuals with predominantly proximal colon carcinoma and sigmoidoscopy to individuals who were most likely to have neoplasia in the distal colorectum. The purpose of this study was therefore to determine if gender, age, or race could be used to differentiate risk for cancer within specific regions of the colorectum.
Data were obtained from the State of Illinois Tumor Registry for the years 1986-1991 for CRC including year of diagnosis, age at diagnosis, anatomic subsite within the colorectum, race, and gender. Data excluded from analysis included carcinoma of the anal canal, carcinoma found within multiple sites of the colorectum, and carcinoma with site unspecified. Also excluded were individuals not of either white or African American race and individuals younger than 40 years or older than 90 years. Population figures in Illinois relating to race, gender, and age were obtained from the 1990 U.S. Census.
Presented in this report are crude case numbers for the 6-year reporting period, the trend in the number of cases reported over the 6 years, and unadjusted incidence rates for carcinoma of the proximal and distal colorectum by age, gender, and race. Age adjustment was made by multiple logistic regression within gender-race groups to model the effect of age, assuming that the mean age in all groups was 60 years. A logistic multiple regression model was then fit using age, race, and gender, but not year of diagnosis, as independent variables. Race and gender were dichotomous variables, and age was adjusted as a continuous variable in 5-year increments from 40 through 90 years. Year of diagnosis was not included in the final model because preliminary regression modeling showed it not to be an independent significant predictor of subsite risk.
When significant interactions were found in the regression between pairs of risk factors (age, gender, and race) for proximal and distal CRC, paired logistic regressions were performed for each group to determine which group was affected the most by each risk factor. The colorectum was divided anatomically for these analyses at the junction of the descending and sigmoid colon; the proximal colon extended from the cecum through the descending colon, and the distal colon included the sigmoid, rectosigmoid, and rectum.
There were 38,931 CRCs included in the analyses for the 6-year period: 19,213 in men and 19,718 in women, 35,058 in whites and 3873 in African Americans, 18,114 cases of carcinoma in the proximal colon and 20,817 in the distal colorectum. A total of 1924 cases fell into one of the 3 exclusion groups and were not included in the analyses. The number of CRC cases declined during the period 1986-1991 (by 1.8% for both genders). Unadjusted incidence rates increased in the proximal colon progressively with age in both races and genders up to age 85 years, after which rate declined in African American males (Fig. 1 (3K)). In the distal colon the incidence findings were similar except for a clearer predominance of males, in particular white males, and a decline in rates for both white males and females after age 85 years (Fig. 2 (3K)). A pattern of racial predominance was seen in age-adjusted incidence rates for the proximal colon, with rates being higher for African Americans of both genders than for whites, and a pattern of gender predominance for the distal colorectum, with risk for males of both races being higher than for females (Fig. 3 (8K)). Each characteristic (gender, race, and age) analyzed independently, controlling for all other factors, was a highly significant contributor to risk of carcinoma in both the proximal and distal colorectum (P < 0.0001), due in part to the large sample size. Age had a greater effect on proximal colon carcinoma risk than distal CRC risk (Table 1). The odds ratios for age were so much lower than for race or gender in Table 1 because age was analyzed as a continuous variable.
Table 1. Odds Ratios for Colorectal Carcinoma in Illinois, 1986-1991, Adjusted for Age, Race, and Gender
Significant interactions were observed between age and both gender and race for proximal and distal CRC, the effect of age being greater among males and whites. The interaction between race and gender was significant for distal CRC; white males were at higher risk than could be explained by the main effect of being white combined with the main effect of being male.
The point of anatomic division of the colorectum used in this report, between the descending and sigmoid colon, was chosen after analysis of SEER (Surveillance, Epidemiology, and End Results program of the National Cancer Institute) age-adjusted CRC incidence data by each of seven subsites, in which a pattern of racial predominance in risk was observed to extend for each subsite from the cecum through the descending colon and a pattern of gender predominance was observed for the sigmoid, rectosigmoid, and rectum above the anal canal (Fig. 4 (9K)).2 Dividing the combined colorectal subsites into proximal and distal between the descending colon and sigmoid colon, including the rectum in distal, clarified the roles gender and race played in subsite risk compared with the roles described in previous publications.3-8 An age-adjusted incidence graph for SEER using this anatomic division is almost identical to what is found in Illinois (Fig. 3 (8K)). This point of anatomic division divides the colorectum by its embryologic anatomy (dividing the midgut from the hindgut) and arterial anatomy (dividing the superior from the inferior mesenteric artery). Clinically, this is the anatomic border that distinguishes the screening effectiveness of fiberoptic sigmoidoscopy, which is limited in length of examination, from that of barium enema, which is limited in its gross anatomic and histopathologic resolution. On the other hand, the points of division of the colorectum used in previous publications, i.e., division of the rectum above the anal canal from the distal sigmoid and division of the colon between the flexures, are problematic. Misclassification of tumors in the region of the rectosigmoid is particularly common because a clear anatomic border between the two sites is lacking.
Several publications have dealt with the issues of age, gender, and race as determinants of anatomic subsite cancer risk in the colorectum. Results have been variable and difficult to interpret. Usually only case numbers were presented, often from single or several institutions, without denominators that allowed incidence rates to be calculated.3-8 One publication did have population-based incidence rates from SEER, with data on gender, race, and subsite, but it was focused on time trend issues and did not analyze or present the data from the perspective described herein.2
It could be hypothesized that prior screening might have had an effect on tumor location. When cancers do develop, populations that are rigorously screened tend to have predominantly right-sided cancer.9 However, recent data from both the Health Interview Survey10 and the Behavioral Risk Factor Survey System11 show that white men are the most frequently screened, which, if prior screening were a major determinant of subsite risk, would result in anatomic patterns opposite to what have been observed in Illinois and SEER. Indeed, the predominance of distal CRC among white males demonstrates just how few people at risk are screened.
These analyses from Illinois have important implications both for screening and in the determination of risk modifiers for CRC. Sigmoidoscopy and barium enema have been found to be the most cost-effective screening strategies for average-risk adults for CRC.1 When choosing which to use, the effectiveness of each might be further improved if sigmoidoscopy were offered as the primary screening strategy to white males, who are more likely to have distal CRC, whereas African American females, who are more likely to have proximal CRC, would be poorly served by sigmoidoscopy and would be better examined by a screening strategy that covered the proximal colon, such as barium enema.
The incidence of CRC declined in African American women relative to all men (Figs. 1 (3K),2 (3K)) at age 65-70 years. This divergence in incidence may be mediated by estrogen. Overweight, which some studies have shown to be greatest in African American women,12, 13 may increase estrogen exposure,14 particularly after menopause. Estrogen has been associated with decreased risk for both proximal and distal CRC.15-17 Conversely, in men, the effect of being overweight18, 19 has been increased risk for CRC. The marked predominance of white males among those at risk for distal CRC is another dominant feature of the incidence graphs (Figs. 2 (3K),3 (8K)). This white male predominance is similar to the pattern of risk that is seen in adenocarcinoma of the distal esophagus and gastric cardia,20 which suggests that risk for cancer at both these locations might be effected by similar exposures.
The decline in CRC case numbers during the period 1986-1991 was small but consistent with what has been observed in SEER,21 and together they represent the first reported decline in CRC incidence in any country. A time trend analysis of suspected risk factors for CRC in light of this declining incidence may yield important clues to CRC causation and facilitate further prevention of disease. Finally, the overall poor survival for African Americans, controlled for stage,22, 23 might at least in part be explained by the findings presented herein, with the predominance of right-sided disease in this racial group and the poorer prognosis associated with proximal lesions.24