Our results, based on a large and representative cohort, show consistently increased mortality (i.e., decreased survival) among Blacks, compared to Whites, that is partially but not completely explained by demographic, disease, treatment, and SES characteristics. This racial/ethnic disparity was evident for mortality due to colorectal cancer as an underlying cause of death, as well as all-cause mortality, although the magnitude of the disparity was larger for colorectal cancer mortality. However, within stage III colon and stages II/III rectal cancer, for which guideline treatments are established, survival differences between Blacks and Whites were considerably smaller and completely explained by SES. We also found decreased survival among Hispanic males that was completely attributable to their being more likely to be diagnosed with advanced stage cancer, and increased survival among Japanese that was due in part to their being more likely to be diagnosed with early stage disease and their place of residence. Our study extends prior population-based racial/ethnic studies of colorectal cancer survival by including additional explanatory factors, including information on comorbidities, radiation, chemotherapy, and area-based SES. In an analysis based on national SEER data, Chien et al. found persistently decreased stage-adjusted survival among certain racial/ethnic groups and postulated that the residual differences were attributable to SES and/or comorbidities [5
]. Using SEER data linked to Medicare claims, we were able to evaluate the impact of these factors, plus additional treatment information, on racial/ethnic differences in survival, and found that although SES had some impact on Black-White differences in survival, comorbidities and treatment did not affect the racial/ethnic differences in survival, at least among the Medicare-eligible population. In fact, the presence of comorbidities had minimal impact on colorectal cancer-specific survival in general, and only among non-Hispanic Whites. As expected, comorbidities were more strongly associated with all-cause survival across the racial/ethnic groups. However, we cannot discount the possibility that more detailed measures of comorbidity (e.g., specific conditions or specific combinations of conditions) or a more comprehensive data collection method (e.g., through medical record review [71
]) may have more significant impacts on survival. Our results confirm, in a national Medicare population, those from the NCI Black/White Cancer Survival Study conducted in the mid-1980's [11
], that the Black-White differential in colorectal survival is partly attributable to stage and SES, but some proportion of the difference remains unexplained. We further found that this pattern was true for all-cause survival as well. A recent review and meta-analysis of 10 studies on colon cancer survival by Du et al. showed that the adjusted hazard ratio for Blacks compared to Whites was 1.14 (1.00–1.29) for all-cause mortality and 1.13 (1.01–1.28) for colon cancer-specific mortality [72
]. Among those studies that adjusted simultaneously for age, comorbidity, SES, and treatment, the estimates were 1.16 (1.03–1.32) and 1.30 (1.01–1.67), respectively [72
]. These estimates are similar in magnitude to those found in our study, 1.14 (1.09–1.20) and 1.21 (1.14–1.29), respectively, with our results showing slightly tighter confidence intervals.
Thus far, it appears that the only studies showing statistically equal stage-adjusted survival between Blacks and Whites are those done in a VA setting [4
], suggesting that there may be attributes of the VA system (equal access to screening and treatment, uniform treatment) or its population (similar SES, equal utilization) that are unique from the general population. An analysis comparing lung and colon cancer outcomes between Black and White VA patients who received uniform evaluation and treatment showed that cancer outcomes were similar despite lower SES (individual-level) among Blacks [74
]. Shavers and Brown synthesized several studies examining racial/ethnic disparities in colorectal cancer treatment and found that population-based studies tended to find significant racial/ethnic differences in treatment, whereas studies showing no racial/ethnic differences tended to be non-population-based, from a single clinic or locality, or of VA populations [75
]. A recent study of a cohort of insured patients found that Black-White differences in survival were explained primarily by stage and receipt of surgery [76
]. In a conceptual framework illustrating potential barriers to the receipt of optimal cancer treatment, Shavers and Brown advocated the importance of "structural barriers", which include presence and type of insurance coverage, institutional and geographic factors [75
]. In fact, a number of recent, population-based studies have demonstrated a modest but independent effect of surgeon and hospital characteristics [8
] and of specific types of health care coverage [10
], but, to our knowledge, these factors have not been evaluated for their impact on racial/ethnic differentials in survival following colorectal cancer.
The increased survival of Japanese compared to other race/ethnic groups and other Asian subgroups has been noted in other studies [5
]. We found that about half of the survival difference between Japanese and Whites was due to earlier stage at diagnosis among Japanese, the other half due to the greater proportion of Japanese living in Hawaii and the increased survival in Hawaii seen in some racial/ethnic groups. Choe et al found, in a survival analysis of national SEER data (with imputed nativity data) that US-born Asians experienced more favorable survival than foreign-born Asians [17
]; as only 28% of Japanese (in California) are foreign-born [78
] and they tend to be the most acculturated of other Asian subgroups, future research might focus on specific factors associated with being US-born Asian, or Japanese in particular, that confer a survival advantage. In a previous paper, we noted decreased stage-adjusted survival among Filipino males [19
]; evidently, the poor survival was limited to both earlier years of diagnosis and younger Filipino men, thus, this pattern was not seen in this analysis.
Our race/ethnicity-specific survival models also revealed differential effects of some of the covariates on colorectal-specific and all-cause survival within each racial/ethnic group. Surprisingly, the area-based SES variables did not appear to be associated with colorectal cancer nor all-cause survival in most racial/ethnic groups. Only the measures poverty and education demonstrated marginally significant effects on survival among Whites, and income among Filipinos. The differential effect of marital status is also interesting: whereas single patients tended to have decreased survival than those who were married, this was not true for Blacks nor Filipinos. The quality of marital status information in the cancer registry is uncertain, so these results should be cautiously interpreted, but the beneficial effects of being married on cancer survival has been previously explored and is hypothesized to be due to greater spousal social support [79
]. We also observed significantly decreased all-cause survival among males compared to females in nearly all racial/ethnic groups. As this gender difference was not seen for colorectal-specific survival among most racial/ethnic groups, this finding suggests that, relative to females, male colorectal patients experienced decreased survival due to causes of death other than colorectal cancer.
There were also differences in the effects of tumor characteristics and treatment. For example, histologic grade, reflecting tumor aggressiveness, did not appear to impact survival among Japanese and Filipinos. Although having any extent of surgical resection was uniformly associated with increased survival across all groups, a beneficial effect of radiation was only seen among Blacks. Chemotherapy was associated with increased survival in all groups, but not in Asians. These racial/ethnic differences in treatment impact on survival may reflect differential distributions of stage and colorectal cancer subsite among race/ethnic groups [63
], or of a differential selection by physicians of healthier patients to receive these treatments [71
]. In addition, because our measures of surgery, radiation, and chemotherapy are crude, the differences may also reflect differences in the quality of treatment. Further study examining survival patterns within specific stage and subsite, and incorporating more specific treatment data would be helpful. Overall, the results from these race/ethnicity-specific analyses support the value of evaluating each group on its own, as this type of analysis may reveal patterns that can result in more targeted and meaningful prevention efforts and strategies for improving survival for each group.
Stage remains the single most independently predictive factor on survival following diagnosis of colorectal cancer across and within racial/ethnic groups. For persons aged 50 or older, the American Cancer Society recommends one of five screening schedules, depending on the modality (fecal occult blood test (FOBT) yearly, flexible sigmoidoscopy or barium enema every 5 years, or colonoscopy every 10 years) [81
]; however, data based on the California Health Interview Survey showed that only 53% of respondents reported having had an FOBT in the past year or sigmoidoscopy/colonoscopy in the past 5 years [82
]. Moreover, Latinos reported lower screening for colorectal cancer (37%) compared to 56.4% among Whites [82
]. Research on promoters and barriers to cancer screening consistently show that, among Asians and Latinos, screening is lower among recent migrants with limited English proficiency [82
]. There is ongoing research to examine strategies to increase screening in different cultures [88
]. Since early detection could potentially reduce some of these differences in survival, particularly among Hispanic men, it is clearly important to learn about barriers to screening among specific communities and to elucidate the most effective strategies for promoting screening in these communities.
Although our study provided a comprehensive evaluation of multiple factors in relation to survival across and within race/ethnic groups, our results may be limited by the quality of some of the SEER and Medicare data. For example, research of race/ethnic misclassification in medical admissions and registry data shows that although there is high consistency between self-report and hospital data for Whites and Blacks, consistency is lower for other groups, in particular Hispanics, American Indians, and certain Asian subgroups [25
]. Our study may also be compromised by completeness of surgery, radiation, and, in particular, chemotherapy data. Comparisons of surgery and radiation recorded in SEER with other sources (Medicare admissions, physician survey) show that registry data are, for the most part, complete and do not vary greatly with regards to race/ethnicity [60
]. Warren et al showed that physician claims data capture about 70% of the chemotherapy administered among colorectal cancer patients [62
]; however, the prevalence of receipt of chemotherapy and of guideline treatment are lower than those seen in other studies [64
] raising concerns about some under-ascertainment of chemotherapy in our data. Thus, we cannot discount the idea that our results, particularly the Black-White differences in survival, may be affected by completeness in our treatment data, particularly if it is differential across racial/ethnic groups. Our analyses could benefit from the addition of potentially useful information on insurance, hospital and physicians characteristics to investigate institutional factors that create unequal access to medical care [8
]. Additionally, although information on colorectal cancer screening is theoretically available from Medicare claims data, insurance coverage for these screening modalities went into effect only recently [95
] and the completeness may be further compromised by how providers bill for these procedures. Furthermore, we included patients diagnosed with colorectal cancer during the years 1992–1997 to allow for observation of a reasonable amount of time between diagnosis and follow-up; however, racial/ethnic survival patterns, particularly between Blacks and Whites, may have changed during this time. Data from the SEER registries showed that while 5-year survival improved among White males and females and Black females between 1992–1997 and 1997–2002, it has worsened slightly among Black males. Thus the survival disparity between Black and White males has increased [14
], making the implications of our analysis even more noteworthy.