Racial and Ethnic Disparities in Interval Colorectal Cancer Incidence: A Population-Based Cohort Study

Introduction

Colorectal Cancer (CRC) is the third most common cancer and second leading cause of cancer-related death in the US.(1) CRC screening is effective in reducing CRC incidence and mortality by detecting pre-cancerous lesions or cancer at more curable stages.(2–7) However, some CRCs develop in screened populations because they were missed at the time of screening or developed between recommended screenings or within surveillance intervals.(8) Interval CRCs, defined as cancers that develop after a negative colonoscopy but before the next recommended test, account for 3–8% of CRCs in the US, though estimates vary by study population.(8–10) Interval CRC risk and its associations with patient demographic and clinical factors as well as physician factors, including quality of colonoscopy metrics, has been examined in a some studies.(9–13) However, patterns of interval CRC risk by race/ethnicity are not well known.(9)

Black-white disparities in interval CRC risk are of particular concern because blacks have the highest CRC incidence and mortality rates of any race or ethnicity in the US, with incidence rates 22–27% higher than whites.(14) Approximately 40% of these disparities in CRC incidence are attributed to lower screening utilization in blacks.(15) The remaining proportion has not been fully explained though likely contributors include differences in socioeconomic status, lack of follow-up after a positive test, and risk factors.(15–18) Another possible reason for this disparity is the difference in the quality of screening tests for CRC.(15) Whereas previous studies have noted poorer quality of mammography in blacks compared with whites;(19, 20) similar detailed evidence pertaining to quality of colonoscopy and interval CRCs by race/ethnicity is not available. Thus, we examined whether interval CRC risk varies by race/ethnicity in Medicare patients 66–75 years of age and whether physician’s polyp detection rate accounts for the potential differences in interval CRC risk between blacks and whites in a population-based cohort study.

Methods

Results

Of the 79,396 Medicare beneficiaries meeting enrollment criteria, 61,433 were included in the analytic cohort, contributing a median of 4.4 and a total of 235,146 person-years of follow-up (Figure 1). The median age at index colonoscopy was 70.0 years and 2,735 interval CRCs were identified.

Most colonoscopies, regardless of race/ethnicity, were performed by physicians whose primary specialty was gastroenterology (Table 1). Receipt of a screening colonoscopy was comparable in blacks (79.5%) and white (80.7%) and polypectomy at index was similar (23.4% in blacks and 24.7% in whites). A higher proportion of blacks (52.8%) received their index colonoscopy from physicians in the lowest two polyp detection rate quartiles compared to whites (46.2%). Relative to patients receiving a colonoscopy from physicians in the highest polyp detection rate quartile, interval CRC risk was higher in patients whose colonoscopy was performed by physicians in first (lowest) (HR=1.95, 95% CI 1.74, 2.20), second (HR=1.53, 95% CI 1.37, 1.71), and third (HR=1.21, 95% CI 1.08, 1.35) (Table 2) quartiles in dose-response manner (Test for trend: χ²= 136.6, p-value <0.001).

The probability of interval CRC by the end of follow-up was 7.1% in blacks, 5.8% in whites, 4.4% in Hispanics, and 3.8% in Asians. (Figure 2) In the Cox regression, interval CRC risk was significantly higher in blacks (HR= 1.32, 95% CI 1.15, 1.51) than in whites after accounting for age, gender and year of colonoscopy (Table 2). Results did not substantially change with further adjustment for characteristics of the zip-code of the patient’s residence (state, poverty level and urban-rural classification), comorbidity and polypectomy at index colonoscopy. Further adjustment for polyp detection rate (Table 2) did not meaningfully affect the black-white differences in interval CRC occurrence (HR=1.31, 95% CI 1.13, 1.51). In the model stratified on individual physicians (Table 3), the point estimate for the black-white difference was farther from the null, but less precise (HR= 1.47, 95% CI 1.13,1.92). In analyses stratified on polyp detection rate quartiles, blacks had higher interval CRC risk relative to whites when colonoscopies were performed by physicians in PDR quartiles three (medium-high) (HR=1.35, 95% CI 1.01,1.82) or four (high) (HR=1.74, 95% CI 1.28–2.37), but not at lower PDR categories. (Table 3)

In analyses stratified on tumor location (Table 4), black patients had higher interval CRC risk for lesions in the distal colon (HR=1.45, 95% CI 1.00, 2.11) and rectum (HR=1.70, 95%CI 1.25, 2.31), but not significantly different for proximal CRCs (HR=1.17, 95% CI 0.96, 1.42). When stratified on stage, compared with whites, blacks had significantly higher risk of interval CRC diagnosed as distant (HR=1.60, 95% CI 1.12, 2.29), but not regional or local disease (Appendix Table 4). Black-white differences were observed for early and late interval CRCs (Appendix Table 5).

Black-white differences were more pronounced among patients who had polypectomy at index colonoscopy (HR=1.41, 95% CI 1.13,1.77) than patients who did not (HR=1.21, 95% CI 0.99, 1.47), but a test for interaction was not significant (p-value for heterogeneity=0.26) (Appendix Table 6). In analyses according to test indication, black-white disparities were similar in patients with screening (HR=1.21, 95% CI 0.97, 1.50) and non-screening colonoscopies (HR=1.21, 95% CI 0.99, 1.48) (Appendix Table 6). Results of analyses restricted to patients receiving their index colonoscopy from a higher-volume gastroenterologist (≥50 or ≥100 colonoscopies) were similar to the main findings (Appendix Tables 7a and 7b) as were analyses that accounted for competing risk of death (Appendix Table 8). Analyses estimating the impact of unmeasured confounding revealed that the confounder-interval CRC and confounder-race/ethnicity associations would each need to exceed 1.50 for the higher risk of interval CRC among blacks to no longer be statistically significant.

Compared to whites, Asians had significantly lower interval CRC risk (HR=0.72, 95% CI 0.57, 0.90) (Table 2). There was no significant difference between interval CRC risk in Hispanics, relative to whites, in the main analyses, or by tumor location or stage.

Discussion

In this population-based study of elderly Medicare enrollees, interval CRC risk was 31% higher in blacks compared to whites, while risk among Asians was lower. Blacks were more likely than whites to have colonoscopies performed by physicians with lower polyp detection rates, a surrogate measure for the quality of colonoscopy. A black-white difference was observed among patients receiving a colonoscopy from physicians with higher polyp detection rates, but not among those who received care from physicians with lower polyp detection rates. However, differences in polyp detection rate of the physician did not explain the observed black-white disparity. Black-white differences in interval CRC risk were more pronounced for distal colon and rectal cancers than for proximal colon cancer.

Missed lesions and polyp detection rate are especially important factors for proximal lesions because the proximal colon is harder to reach endoscopically and it is the most common location of difficult-to-detect sessile polyps.(35–37) A previous study reports that behavioral factors (e.g. smoking, obesity) account for a greater proportion of differences in proximal colon cancer incidence by measures of SES, raising the possibility that lower colonoscopy utilization could account for these differences.(18) Additionally, a SEER-Medicare study reported a protective effect of ever receiving a colonoscopy, relative to not receiving a colonoscopy, that was closer to the null in blacks compared to whites, particularly for distal colon cancers after accounting for physician specialty but not polyp detection rate.(38)

We observed greater black-white differences in distal colon and rectal cancer interval CRC risk than in the proximal colon. This observation, along with greater black-white differences among patients receiving colonoscopies from physicians with higher but not lower polyp detection rates, aligns with previous observations that disparities in outcomes and healthcare utilization often manifest as higher quality or new interventions become available. (39–42) Whether this finding attributable to physician factors, including lower quality of examination, is unknown and warrants exploration.

Blacks and other minorities more frequently received colonoscopies from physicians with lower polyp detection rates and thus lower quality of care. These findings are consistent with previous reports that blacks were more likely to receive healthcare from physicians in lower resource settings and also experienced poorer outcomes.(19) (43)

It is possible that quality factors other than polyp detection rate contribute to this pattern. Cecal intubation rates, withdrawal time and patient quality factors such adequacy of bowel preparation may vary by race, and be correlated with polyp detection rate.(44) Data on incomplete resection of polyps, the second most common reason for interval CRCs,(9) by race/ethnicity are not available in the published literature nor is it captured in SEER-Medicare data. Polyp detection rate is an indirect measure of lesions missed during a colonoscopy, the most commonly cited reason for interval CRCs, though other reasons include incomplete resection of polyps, or rapidly developing or “de-novo” tumors.(8, 9) A study estimated that 37% of interval rectal CRCs were attributable to incomplete polyp detection, compared to 10–16% of proximal tumors, possibly contributing to blacks’ higher risk of interval rectal cancer (45) and interval CRCs diagnosed at an advanced stage compared to whites in our study.

It is important to point out that some interval CRCs (13–24%) are believed to be “de novo” cancers and are thus unavoidable. It is not clear if risk of such lesions differs by race.(9) Previous studies suggest that blacks have more aggressive tumors, because they tend to be younger at diagnosis(46) and have higher prevalence of large polyps.(47) However, overall prevalence estimates for colorectal polyps and adenomas in blacks and whites are similar,(17, 48) findings consistent with similar polypectomy prevalence in our study. There is evidence that sessile serrated polyps are more aggressive and interval CRCs are more likely to exhibit micro-satellite insatiability (MSI) and CpG island methylator phenotype (CIMP).(49) Whether or not these factors account for higher interval CRC risk among black patients is not clear as studies on racial differences in MSI and CIMP are equivocal(50–52) and our study was not designed to directly answer this question. MSI tumors tend to be proximally located, and if this factor was driving racial disparities in interval CRCs, we would have expected especially elevated proximal interval CRC tumor risk among blacks relative to whites. Our data provide little evidence that this may be the case.

Polypectomy at index colonoscopy was similar between blacks and whites, though black-white differences were more apparent among patients with polypectomy than without polypectomy. Depending on the size and number of polyps detected, surveillance colonoscopy is recommended at intervals of up to 10 years of an index procedure in most instances.(22) We were not able to directly measure adherence to recommended follow-up intervals due to a lack of information on histology and polyp size in Medicare claims. Black-white differences were observed within three years following a colonoscopy in our study and two previous SEER-Medicare studies noted that blacks were more likely to undergo a surveillance colonoscopy within three or five years of polyp removal.(53, 54) Taken together, this suggests that differences in recommended surveillance colonoscopy may only moderately contribute to higher risk of interval CRC among blacks, though further study of the utility of race-specific surveillance colonoscopy recommendations is needed.

Lower interval CRC risk among Asians relative to whites observed in the current study is consistent with the previously reported lower overall CRC incidence in this group (55), a pattern commonly attributed to differences in risk factors. Our findings agree with previous studies reporting similar polyp prevalence in Asians and whites.(48, 56) These findings raise the possibility that polyps progress more slowly in Asians, though detailed information on tumor characteristics (e.g.: MSI) in Asians is not available.

In our study, the probability of interval CRCs within 5 years of an index colonoscopy was 5.8% and the majority were proximally located. This observation is consistent with current literature.(9–11) Our findings and those reported elsewhere (11) highlight the importance of attentive examination of the colon and rectum during a colonoscopy to achieve the optimal benefit of this test. A previous SEER-Medicare study covering the period from 1994–2005 noted higher odds of interval CRC relative to screen-detected CRCs for blacks relative to whites.(11) Our findings are consistent with this observation.

There are some limitations of this study. Colonoscopy and polypectomy were identified using billing codes. Compared to an endoscopic database, Medicare data have high sensitivity (>93%) and specificity (98%) for identifying colonoscopy and polypectomy.(57, 58) Research supports the use of administrative data to estimate polyp detection rate (27) as a proxy for adenoma detection rate as the two measures are highly correlated (24–27, 59). The PDR-ADR correlation is stronger in the proximal (r>0.90) than the distal (r=0.58–0.59) colon, (24, 26) likely due to a greater proportion of non-adenomatous polyps in distal colon.(60) However, proximal ADR is more strongly associated with interval CRC risk than distal ADR (10). Test indication was not directly available, though results incorporating algorithm-based indication did not alter our main findings. Further, the minimum number of colonoscopies required to determine polyp and adenoma detection rates is unclear. One study estimated that ≥500 examinations would be needed to determine ADR(61) and the study that served as the basis for establishing the American Gastroenterology Association’s quality metrics included physicians with ≥300 colonoscopies(10), while others have used a threshold of 50 procedures.(25, 27) In the current study, we used a threshold of 500 colonoscopies, represented by 25 colonoscopies in the 5% Medicare sample. Varying the threshold and restricting analyses to higher volume gastroenterologists did not alter our main findings. Polyp detection rates were based on Medicare FFS patients, which may not represent physicians’ total patient population, though relative measures of procedural volume in SEER-Medicare and Medicare were similar and procedural volume in patients <65 and ≥65 years were correlated in previous studies. (62) (63) We did not have information on the involvement of gastroenterology fellows that may increase adenoma detection rates (64) and could vary by institution and patient race. Information on tumor characteristics and polyp histopathology that presumably influence CRC risk(9) were not available in SEER-Medicare data. We also used area-based measures of SES as individual-level measures as were not available.(65, 66) Data on behavioral factors (e.g: smoking, obesity) that clearly increase overall CRC risk were also not available, though the influence of these factors on interval CRCs is unexplored. Analyses estimating the impact of such unmeasured confounders revealed that each of the confounder-interval CRC and confounder-race/ethnicity associations would need to exceed 1.50 to substantially alter our main findings (33), a level not observed in most studies of overall CRC risk.(67) We did not have data on receipt of colonoscopy prior to Medicare enrollment. Receipt of colonoscopy and polypectomy prior to Medicare disrupts the natural history of colorectal carcinogenesis and increases the opportunities to detect and remove polyps, and thus a lower prevalence of polyps. It is possible that lower colonoscopy use prior to Medicare enrollment among blacks, if present, may result in more accumulated lesions and thus the potential for a higher chance of missed lesions in this population, but this requires study. Lastly, our results may not represent disparities occurring in younger populations.

In conclusion, we observed higher risk of interval CRCs in blacks compared to whites in a population-based study of elderly Medicare enrollees. Proximal tumors represented the majority of interval CRCs. Black-white differences were most pronounced for distal colon and rectal cancers and in patients receiving colonoscopies from higher quality physicians, as measured by polyp detection rate. While quality of colonoscopy was associated with interval CRC, it did not account for the racial disparities. Futures studies examining this issue are warranted given the higher overall risk of interval CRC in black populations as well as larger disease burden in this group.

Supplementary Material