Are There Gender Differences in Colorectal Cancer Test Use Prevalence and Correlates?

Among both males and females, colorectal cancer is the third most frequent type of cancer diagnosed and accounts for ∼10% of cancer deaths in the United States (1). The early detection of colorectal cancer and/or removal of precancerous polyps may contribute to decreased incidence of and mortality from colorectal cancer (2-4). However, only about 40% of colorectal cancers are diagnosed at an early stage when the 5-year survival rate exceeds 90% (5, 6). Furthermore, despite guidelines from authoritative groups (1, 7), screening rates are low (8, 9). For example, the 2000 National Health Interview Survey showed that 16.8% of males and 17.5% of females had a fecal occult blood test (FOBT) in the past year, and 37.4% of males and 31.1% of females had endoscopy within the past 10 years (10). For either test within recommended time intervals, the prevalence was 44.5% for males and 41% for females (10). These data are consistent with other studies that report somewhat higher endoscopy or any colorectal cancer test use among males and higher FOBT use among females (11-14). Although the differences in test use by gender may seem small, they are consistently statistically significant and may be important at a population level, especially given the low rates of colorectal cancer test use for either gender. The 2002 to 2003 Health Information National Trends Survey (HINTS) was the first national survey to use a standardized set of self-reported measures for colorectal cancer test use that was developed and tested using cognitive interviewing methods (15). This set of measures also included a question about repeat test use: a question not previously included in national surveys.

Correlates of colorectal cancer test use assessed in national surveys have been limited to demographics, access to health care, health status, and health behaviors, and only one study examined whether these correlates differ by gender (12, 16-19). Furthermore, the scant existing data examining gender differences are from regional studies that show inconsistent results for variables such as family history of cancer (11, 12, 18) and health behaviors (12, 18). Additionally, no prior studies have examined gender-specific correlates by colorectal cancer test type (e.g., FOBT and endoscopy). The HINTS was designed and conducted by the National Cancer Institute to collect data on knowledge, beliefs, communication about cancer, and cancer-related behaviors, such as screening. Thus, the HINTS expanded upon the types of correlates previously collected in national surveys by including psychosocial variables.

We conducted exploratory, hypothesis-generating analyses to examine patterns in the data and offer suggestions for future research. Our rationale for an exploratory approach is that although many of the selected correlates have been previously examined for their associations with colorectal cancer test use, published results by gender are inconsistent; therefore, there is little basis for hypothesis testing. Furthermore, the HINTS was not designed to answer the research questions posed here. Because the sample size was not sufficient to yield precise estimates, we looked for consistency in the direction and magnitude of odds ratios (OR) across gender and colorectal cancer test type. If gender differences exist in the factors influencing test use, interventions may benefit from targeting messages by gender. We addressed the following questions: (a) Are prevalence rates for lifetime, recent, and repeat FOBT and endoscopy similar for males and females? (b) Are the demographic, health status, access to health care, and health behavior correlates of FOBT and endoscopy use previously reported in the literature similar for males and females? (c) Are the patterns of these correlates similar to findings from other national surveys? (d) Are psychosocial variables in the HINTS, including knowledge, cancer-related beliefs, and cancer communication, associated with FOBT and endoscopy use, and are the associations similar for males and females?

The HINTS is a cross-sectional telephone survey that was conducted in the United States between October 2002 and April 2003 using random digit dialing. One adult (≥18 years) per household was eligible to participate. Telephone exchanges were geographically stratified to oversample from exchanges estimated to have ≥15% African American and Hispanic residents. Final data were weighed to be nationally representative. Telephone interviewers were able to reach 19,509 households. The final response rate for an initial screening interview was 55%, calculated according to the guide of Standard Definitions published by the American Association for Public Opinion Research (20). The final response rate for the full HINTS interview was 62.8%. Full details of the sampling plan are reported elsewhere (21).

Of the 6,369 telephone surveys completed by adults ages ≥18 years, 2,734 were ages ≥50 years. Of those, 2,686 had no personal history of colon or rectal cancer and comprised the sample used in this report; 63% were female and 74% were non-Hispanic White. The average age was 64.4 years old (SD = 10.4 years) and ranged from 50 to 95 years.

Only those questions from the HINTS relevant to the present study are described below. The complete survey instrument can be obtained online at http://cancercontrol.cancer.gov/hints/instrument.jsp.

To assess colorectal cancer test use, individuals were first read a description of one of three test types (home FOBT, sigmoidoscopy, and colonoscopy) and asked whether they had ever heard of the test (no/yes). Respondents who had heard of the test were then asked whether they had ever had the test (lifetime use: no/yes), when the test was completed, and when the next-to-last test was completed. Persons who reported experience with both sigmoidoscopy and colonoscopy were asked to report when their most recent endoscopic test was completed. The survey did not distinguish which endoscopy test was the most recent and so responses reflected either procedure. Individuals were considered currently adherent to guidelines for colorectal cancer testing if they reported having a home FOBT within the past year or endoscopy within the past 10 years. Use of double-contrast barium enema was not assessed in the HINTS.

Reference groups for the multivariable analyses are listed first in the description of the variables.

Demographic variables included age in years (≥65, 50-64 years), race/ethnicity (non-Hispanic White, Hispanic, or Black/other/missing), marital status (unmarried, married), education level completed (less than high school, more than or equivalent to high school graduate), and income (<US$25,000, ≥US$25,000).

Access to health care variables included the number of physician visits in past year (0, ≥1), having a regular physician (no/yes), and insurance status (none/any).

Health status variables included general health status (excellent/good versus fair/poor), body mass index (normal weight, <25; overweight, 25-29; obese, ≥30), and both personal and family history of cancer (no/yes). Family history of cancer was assessed with the following question: “Have any of your brothers, sisters, parents, children, or other close family members ever had cancer?” (no/yes).

Health behaviors included recent mammography (past 2 years) or prostate-specific antigen test use (past year: no/yes), daily fruit and vegetable intake (<5, ≥5), engaging in regular physical activity long enough to work up a sweat at least once a week (no/yes), and smoking status (never, <100 cigarettes in their lifetime; former, ≥100 cigarettes in lifetime but none currently; current).

Knowledge items included knowing (correct, incorrect) the recommended screening interval for FOBT (12 months) and endoscopy (10 years), the age at which most people should start using each colorectal cancer test (50 years old), and the age category of highest risk for colorectal cancer (≥60 years). The items were examined individually.

Beliefs about cancer risk included multiple single-item questions: absolute perceived risk of developing colon cancer in the future (low/moderate, high, don't know), comparative perceived risk of developing colon cancer “compared with the average (man/woman) your age” (equally/less likely, more likely, don't know), whether participants believed “There's not much people can do to lower their chances of getting cancer” (agree, disagree, no opinion), and for which cancer types (if any) participants believed that exercise could lower risk (none, colon cancer, other cancers, don't know).

Beliefs about colorectal cancer test use were assessed with four questions. Respondents were asked whether they disagreed, agreed with, or had no opinion about the following statements: “Getting checked for colon cancer is too expensive.” “Arranging to get checked for colon cancer would be easy.” “Getting checked regularly for colon cancer increases the chances of finding cancer when it's easy to treat.” and “You are afraid of finding colon cancer if you were checked.”

Beliefs about cancer in general included which cancer was perceived to cause the most deaths (other, colon, don't know) and level of agreement with “There are so many different recommendations about preventing cancer, it's hard to know which ones to follow” and “It seems like almost everything causes cancer” (disagree, agree, no opinion).

Negative Affect. Cancer worry was assessed with the following single item: “How often do you worry about getting colon cancer?” (sometimes/rarely/never versus often/all the time). The Kessler Psychological Distress Scale (22-24) was used to assess nonspecific psychologic distress in the past 30 days. Sum scores were created from the six-item scale for respondents with complete data. For respondents with more than half of the items complete, their mean score for the completed items replaced missing values to allow for a sum of scores for all items. Imputation was needed for only 36 respondents. Item responses ranged from 0 (none of the time) to 4 (all of the time). Scale scores could range from 0 to 24 and were dichotomized (0-12 and 13-24) based on the literature (24).

Cancer Communication. We created a sum score from five questions that assessed the degree to which participants paid attention to any health or medical topics via television, radio, newspapers, magazines, and the Internet. Likewise, we created a sum score from five items that assessed participants' trust in the same five media sources for cancer information. Responses to the five items were summed and divided by the number of items answered by the respondent and dichotomized (a lot/some versus a little/not at all).

Single items assessed participants' trust in health care providers and family or friends for cancer information. Responses were dichotomized for health care providers (a lot versus some/a little/not at all) and family or friends (a lot/some versus a little/not at all) based on the distribution of responses.

Single items assessed whether someone other than a health care provider had ever looked for cancer information for the respondent (no/yes), and whether the respondent had ever looked for information about cancer from any source (no/yes).

Information seeking self-efficacy was assessed with the following item: “Overall, how confident are you that you could get advice or information about cancer if you needed it?” (very confident versus somewhat/slightly/not at all confident).

We chose to stratify analyses by gender instead of testing interactions with gender to explore differences in the patterns of associations. Descriptive analyses were conducted using SAS and Survey Data Analysis (SUDAAN) to report both observed sample sizes and weighed percentages.

Prevalence of Colorectal Cancer Test Use. We examined colorectal cancer test use prevalence rates by gender for lifetime, recent, and repeat use of home FOBT, sigmoidoscopy, and colonoscopy alone and in combination, as well as whether any of the three tests had been completed. We used 95% confidence intervals (95% CI) to compare the rates for males and females. Prevalence estimates were age adjusted. The denominator for repeat colorectal cancer test use was restricted to respondents who would be eligible to complete two screening tests (i.e., two FOBTs within 2 years or two endoscopy tests within 20 years). We conducted analyses using SUDAAN software and a replicate weight jackknife estimate of variance to account for the sampling design and to calculate appropriate population estimates; therefore, the results reflect weighed and design-adjusted data.

Additionally, reasons for not completing a colorectal cancer test have been previously reported but have not been examined for gender differences. Identifying gender-specific and test-specific barriers to colorectal cancer test use may be important for future interventions designed to increase colorectal cancer test use. To explore whether reasons for not being tested differed by gender, we conducted descriptive analyses with a subset of respondents who had seen a health care provider in the past year but had not had a colorectal cancer test within recommended time intervals. Respondents could give multiple reasons for not having had a FOBT or endoscopy. We categorized responses into patient- or system-level factors following Klabunde et al. (25).

Correlates of Colorectal Cancer Test Use. We used multivariable logistic regression analysis with SUDAAN to examine correlates of colorectal cancer test use by gender and test type (FOBT and endoscopy), thereby creating four regression models (one for each combination). Our two dependent variables were home FOBT use in the past year and recent endoscopy (i.e., sigmoidoscopy or colonoscopy in the past 10 years). The reference group for both variables was no colorectal cancer test of any type within recommended intervals. We used ORs and 95% CIs to summarize the results. CIs for males and females that do not overlap may suggest gender differences.

Analysis Strategy. To facilitate comparison with other national surveys (refs. 10, 16; research questions 2 and 3), we used a two-step analysis procedure for all four regression models. In step 1, we examined variables measuring demographics, access to health care, health status, and health behaviors that have been previously examined with nationally representative samples and explored whether their associations with colorectal cancer test use differed by gender. Correlates that were statistically significant (P < 0.05) in at least one of the four models in step 1 were retained in all step 2 regression models. In step 2, we added variables from the HINTS that have not been previously examined in national surveys (i.e., knowledge, beliefs, and cancer communication). Family history of cancer has been consistently associated with colorectal cancer test use in the literature and was retained in step 2, although it was not statistically significant in step 1. Because results did not change when this variable was removed from analyses, we chose to retain it to enable comparisons with other studies.

Effect estimates with cell sizes ≤ 5 were not presented. “Don't know” or missing responses were included when the percent of missing data was large and potentially meaningful (e.g., income and perceived risk) or when “don't know” was similar in meaning to a valid “no opinion” response choice (e.g., cancer beliefs).

In general, the distributions of the independent variables were similar for males and females; exceptions are bolded in Table 1.

After hearing a description of each test, 70.5% of 999 males and 79.7% of 1,687 females reported that they had heard of the home FOBT, and 79.2% of males and 87.6% of females had heard of sigmoidoscopy or colonoscopy.

Prevalence of colorectal cancer test use was low, and there were few gender differences (Table 2). Females reported slightly higher lifetime (ever) and current (past year) use of FOBT only, whereas more males reported repeat endoscopy use.

To further explore the prevalence data, we conducted an analysis comparing the 989 respondents (34.4% of males and 34.7% of females) who reported more than one type of colorectal cancer test with the 767 respondents (25.9% of males and 30.8% of females) who reported only one test type. Experience with more than one colorectal cancer test type was associated with greater adherence to any test type within recommended time intervals for both males (OR, 11.12; 95% CI, 5.46-22.68) and females (OR, 8.72; 95% CI, 5.62-13.52).

Because of stratification and the number of correlates examined, the precision of the OR estimates was reduced, and in the majority of cases, the CIs included 1.0 (Table 3). All of the CIs across gender overlapped, which may suggest no statistically significant gender differences. Nevertheless, several variables showed gender-specific patterns of association with colorectal cancer test use and are discussed below.

Of the demographic variables examined, men younger than 65 years were significantly less likely to have had a colorectal cancer test compared with older men (data available on request). Although the patterns of association were similar for women, the CIs included 1.0. Hispanics and Blacks were less likely to have had a colorectal cancer test compared with non-Hispanic Whites, but this effect was only statistically significant for endoscopy among males. Being married, having more education, and a higher income were associated with greater colorectal cancer test use for both genders, but with the exception of education and endoscopy use among females, the CIs included 1.0. All three measures of access to health care were associated with increased use of colorectal cancer tests for both genders. The strongest association was observed for the number of physician visits in the past year, which was statistically significant for endoscopy for both genders. For health status indicators, there was no consistent pattern between self-reported general health status or body mass index and use of colorectal cancer tests by gender or test type. Although not statistically significant, personal and family history of cancer (other than colorectal cancer) were associated with greater test use except for FOBT among males. Of the health behaviors examined, recent screening for breast or prostate cancer showed a significant association with increased use of colorectal cancer tests across gender and test type. Among males only, fruit and vegetable consumption was significantly associated with increased use of FOBT, whereas weekly exercise was significantly associated with increased endoscopy use. Smoking status showed no consistent pattern by gender or test type.

Step 1 variables that remained significant correlates of colorectal cancer test use in step 2 were age, number of physician visits, and gender-specific prior cancer screening (Table 3). Due to sample size limitations, only one measure of access, number of physician visits, was included in step 2. Only correlates of test use that were statistically significant (P < 0.05) in at least one of the four models are presented in Table 3, but all estimates are available upon request.

Of the HINTS psychosocial variables added to step 2 analyses, only one of the three questions assessing knowledge (knowing the correct colorectal cancer test-specific screening interval) was significantly associated with increased colorectal cancer test use for both males and females (Table 3).

Of the four beliefs about cancer risk, only comparative risk was statistically significantly associated with increased test use and only for endoscopy use among females. However, the pattern of ORs suggested a positive association with FOBT for females as well. In contrast, the ORs were in the opposite direction for males (Table 3).

Of the four beliefs about colorectal cancer test use examined, three showed gender differences in the patterns of association with colorectal cancer test use. Of particular interest, thinking that colorectal cancer tests are too expensive seemed to increase use among males but decrease use among females; however, the CIs included 1.0 for most estimates. “Don't know or no opinion” responses to the question about cost were consistently associated with less colorectal cancer test use for both genders (Table 3). The belief that colorectal cancer testing leads to early detection was associated with increased colorectal cancer test use for both genders; however, the CIs included 1.0 except for FOBT among females. Fear of finding cancer through colorectal cancer testing was significantly associated with increased endoscopy use among females but not males. Although the CIs for FOBT and fear included 1.0, the ORs were elevated for both men and women.

The ORs for the four beliefs about cancer in general and the two measures of negative affect showed less consistent patterns of association by gender or test type. Of these six variables, only one belief about cancer in general was significantly associated with colorectal cancer test use. Responding “don't know” to the statement “everything causes cancer” was associated with increased FOBT use among males (Table 3).

Of the seven cancer communication variables, two showed gender-specific patterns of association with colorectal cancer test use (Table 3). Attention to health information in the media was significantly associated with increased colorectal cancer test use for males but not for females. Greater trust in media information sources was significantly associated with less endoscopy use among males, and the pattern was similar for FOBT use.

Respondents who had seen a health care provider in the past year but had not had any colorectal cancer test within recommended time intervals were asked why they had not been tested. No significant gender differences were found in the reasons for not being tested. The most common reasons included patient-related lack of awareness, such as no reason and not knowing there was a need for testing, as well as system-related reasons, such as not having a physician recommendation or test order (Table 4).

Similar to prevalence estimates from other national surveys, we found that females were more likely to report lifetime and current FOBT use compared with males (11-14). In contrast to previous reports of higher rates of recent use among males (11-14), we found comparable endoscopy use among males and females. However, rates of repeat colorectal cancer test use, a measure not previously assessed in other national surveys, showed that males were more likely to report repeat endoscopy. Regardless of test type, colorectal cancer test use remains below desired levels for both males and females and has increased only modestly since 1987 (9, 26, 27).

Correlates of colorectal cancer test use identified in step 1 of our analysis were similar to those found with National Health Interview Survey and Behavioral Risk Factor Surveillance System (Table 5; refs. 10, 16). Visits to a physician or having an annual medical check-up has been consistently associated with increased test use for both males and females (11, 12, 16-18). Frequency of contact with a physician, especially if for another cancer screening test, is likely to increase the opportunity to discuss or receive recommendations for colorectal cancer screening (10, 28). Older age is also consistently associated with colorectal cancer test use for both genders. To increase the uptake of colorectal cancer test use, future interventions should target younger adults eligible for screening (50-64 years old) and those without a recent visit to a healthcare provider.

Unlike previous findings reported in the literature, family history of cancer and smoking status were not significant correlates of colorectal cancer test use in this study. Family history of colorectal cancer has been found to increase colorectal cancer test use (10, 18, 29), but observed associations have not been consistent by gender (11, 12, 17, 18). Because the HINTS measured family history of any cancer, our results may have differed if respondents had been asked specifically about family history of colorectal cancer. Although smoking was negatively associated with colorectal cancer test use in previous work (10, 13, 18, 30-32), our study showed no consistent pattern.

The gender differences reported for correlates of colorectal cancer test use vary across studies. Similar to previous findings (18), in step 1 analyses, we found that physical activity was positively associated with endoscopy among males. Unlike previous studies, body mass index was not a positive correlate for women (18), and personal history of any cancer was not a positive correlate for men (11). Lastly, education was not associated with test use in our multivariable analyses, but it has been positively associated with test use for both males and females (16) and for males only (12) in other studies. More research examining gender differences in correlates and predictors of colorectal cancer test use is needed to identify factors that can be used in future interventions to identify subgroups who consistently underuse colorectal cancer tests.

Previous research examining knowledge of colorectal cancer risk factors and test use found no associations (33-35); however, we found a consistent positive association between test use and knowledge of the correct colorectal cancer test-specific interval but not for age to begin testing or age of highest colorectal cancer risk. These findings suggest that knowing the age of increased risk for colorectal cancer and the recommended age to initiate screening should be given more emphasis by health care providers and in public health campaigns.

Although not statistically significant across all models, other patterns of association suggested gender differences in the correlates of colorectal cancer test use that could be explored in future studies. Believing that one's risk for colorectal cancer was higher than peers of the same age and gender was positively associated with test use among females but was negatively associated among males. In previous studies, a positive association between perceived risk and colorectal cancer test use was reported among males only (11), and a positive association between perceived risk and stage of change for test use was reported among women only (17). Because perceived risk is frequently an intermediate focus of health promotion interventions, further examination of gender differences using different measures of perceived risk is needed. Although fear of finding cancer has been reported as a barrier to colorectal cancer test use (36), we found a significant positive association with endoscopy use among females. Further cognitive testing with this item may identify how individuals interpret and respond to such hypothesized barriers to screening behavior. Believing that colorectal cancer tests are too expensive may be a barrier to test use among females but not among males. However, for both males and females, not knowing whether colorectal cancer tests were too expensive was associated with less test use. Therefore, questions about cost and insurance coverage may be important but modifiable barriers to test use. A substantial number of participants responded “don't know or no opinion” to survey items assessing cancer-related beliefs, which may reflect a lack of awareness regarding colorectal cancer and test options (37-40).

Cancer communication variables, including measures of social influence and exposure to various sources of health information, were a unique focus of the HINTS. Our finding that attention to health topics in the media was positively associated with test use among men suggests that media campaigns may be a means for improving colorectal cancer screening rates, especially among men. However, our results also showed that greater trust in media sources for health information was associated with less endoscopy use among men, and the patterns of association were similar for the other models. The influence of the media in delivering health information on colorectal cancer screening deserves more attention, particularly with regard to gender differences. Such differences, if confirmed, may support the use of gender-specific intervention strategies.

The reasons for not being screened were similar to previous studies, and our results suggest that they do not differ by gender. Consistent with data from the 2000 National Health Interview Survey (10, 25, 41), the primary reasons for not having a colorectal cancer test included lack of awareness or no reason and no physician recommendation. Previous research suggests that compared with patient self-reports, physicians overestimate the influence of perceived pain and embarrassment as patient-related reasons for not being screened (25). Future research should explore the reasons for this apparent incongruence in beliefs of patients and health care providers.

The HINTS was a cross-sectional survey, and no causal interpretations of the observed associations can be made. All data were based on self-report; however, other studies have shown that self-reported colorectal testing is quite accurate (42, 43). Because our analyses were exploratory and designed to generate hypotheses for future studies, subsequent studies will need to confirm the associations reported here. However, our ORs for step 1 variables were of similar magnitude to those from other national surveys, and differences in statistical significance may be due to the HINTS smaller sample size. Similar to the Behavioral Risk Factor Surveillance System (16), the HINTS did not distinguish colorectal cancer test completion for screening versus diagnosis or between sigmoidoscopy and colonoscopy. Like Seeff et al. (10), sigmoidoscopy compliance was probably overestimated given our use of a 10-year testing interval. None of these limitations, however, are likely to have differentially affected the comparisons by gender.

Similar to previous reports using national data (10, 16, 26), more recent data from the HINTS shows that colorectal cancer test use falls below rates for other recommended cancer screening tests. Colorectal cancer is the first cancer for which we have a recommended screening test for both males and females and are therefore able to examine gender differences in both prevalence and correlates or predictors of test use. Our findings suggest avenues for the future exploration of gender differences. If confirmed in future studies, gender-specific intervention strategies or messages may improve colorectal cancer screening rates.