Racial Disparities in Recurrence and Overall Survival in Patients With Locoregional Colorectal Cancer

Rebecca A. Snyder, MD, MPH; Chung-Yuan Hu, MPH, PhD; Syed Nabeel Zafar, MBBS, MPH; Amanda Francescatti, MS; George J. Chang, MD, MS


J Natl Cancer Inst. 2021;113(6):770-777. 

In This Article


Study Population

Patients older than 18 years with American Joint Committee on Cancer (AJCC) stage I, II, or III adenocarcinoma of the colon or rectum treated with definitive surgical resection in 2006–2007 were identified from the National Cancer Database (NCDB). A random sample of up to 10 patients with CRC from each Commission on Cancer (CoC) facility was selected for detailed primary data collection regarding surveillance testing and recurrence as part of the commission's special study as previously described.[3] For facilities with fewer than 10 eligible patients, data for all eligible patients were abstracted. These data were then merged with corresponding NCDB records. The study period (2006–2007) was selected to allow for a minimum of 5 years of follow-up.

Patient race was determined from predefined NCDB data based on assignment by a CoC registrar according to fixed categories.[4] Only patients of White or Black race were included for the primary study analysis. Patients with unknown tumor location, income, education, distance traveled, or race were excluded. Patients who did not undergo any surveillance testing following treatment of the primary cancer were also excluded because they could not be assessed for the primary outcome of recurrence.

The study analysis was considered exempt by The University of Texas MD Anderson Cancer Center Institutional Review Board.


CRC clinical and pathologic stage was defined based on the AJCC 6th edition staging manual.[5] High-risk stage II colon cancer was defined by the presence of 1 or more of the following features: T4 tumor (based on AJCC T-stage), fewer than 12 lymph nodes (LN) examined, positive margin, high tumor grade (grades 3–4), perineural invasion, or lymphovascular invasion.

Assessment of receipt of treatment of primary colon cancer was based on LN retrieval and chemotherapy. Guideline concordant care (GCC) was defined as 1) removal of at least 12 LN and all nodes negative (AJCC stage I or low-risk stage II), 2) removal of fewer than 12 LN but receipt of chemotherapy (high-risk stage II), or 3) removal of at least 12 LN and 1 or more nodes positive and receipt of chemotherapy (stage III). Failure to receive GCC was defined as 1) removal of fewer than 12 LN and no receipt of chemotherapy (high-risk stage II), or 2) removal of any number of LN with at least 1 or more nodes positive and no receipt of chemotherapy (stage III). Assessment of GCC among patients with rectal cancer was defined as receipt of neoadjuvant or adjuvant radiation in patients with either clinical stage II or III or pathologic stage II or III disease.

Surveillance testing included computed tomography, magnetic resonance imaging, positron emission tomography, or carcinoembryonic antigen tests as previously described.[3]

Outcome Measures

The primary outcomes of interest were disease recurrence (locoregional or distant) and OS. Local recurrence was defined as recurrence at the site of primary tumor, anastomosis, or adjacent structure. Regional recurrence was considered recurrence at the regional LN associated with the primary tumor site, and distant recurrence was defined as recurrence outside the local or regional sites. Recurrence could be confirmed either clinically or pathologically. If synchronous locoregional and distant recurrence were diagnosed, both were recorded.

A prespecified sensitivity analysis was performed to assess recurrence rates and OS only among patients who received GCC, defined above.

Statistical Analysis

Baseline clinical and demographic characteristics were compared across racial groups using the χ2 test for categorical variables, t test for means, and Kruskal-Wallis test for medians of continuous variables. Patients were censored at the time of death, loss to follow-up, or end of the surveillance study period (5 years). Cumulative recurrence rates were determined using the Kaplan-Meier method and tested using log-rank test. Multivariable Cox proportional hazards regression model was performed to yield statistical inferences. The proportional hazards assumption was verified graphically using the log-log plot, where the -ln{−ln(survival)} curves of the covariate vs ln(analysis time) were checked. Inverse probability of treatment weighting using propensity score was performed to confirm the findings. Covariates adjusted in Cox and logistic regression for propensity score estimation were based on the previously published recurrence risk model and included age at diagnosis, sex, histology, pathologic tumor stage, index of metastasis (number of positive LN divided by total number of LN examined), lymphovascular invasion, perineural invasion, surgical margin status, chemotherapy, and tumor site, as well as Charlson-Deyo comorbidity index.[6] Two variables had greater than 5% missing values: lymphovascular invasion and perineural invasion. Multiple imputation by chained equations was used to substitute predicted values for missing values with 20 imputed values.

All tests were 2-sided, with an alpha of .05. All analyses were performed using SAS software (version 9.1.3; SAS Institute Inc) for data processing and Stata MP (version 13.1; StataCorp) for statistical analyses. Study findings are reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology statement guidelines.[7]