Disparities among Black and Hispanic colorectal cancer patients: Findings from the California Cancer Registry

Abstract Background Colorectal cancer (CRC) is the third most common cancer in California and second among Hispanic/Latinx (H/L) males. Data from the California Cancer Registry were utilized to investigate the differential impact on CRC outcomes from demographic and clinical characteristics among non‐Hispanic white (NHW), non‐Hispanic Black (NHB), U.S. born (USB), and non‐U.S. born (NUSB) H/L patients diagnosed during 1995–2020. Methods We identified 248,238 NHW, 28,433 NHB, and 62,747 H/L cases (32,402 NUSB and 30,345 USB). Disparities across groups were evaluated through case frequencies, odds ratios (OR) from logistic regression, and hazard ratios (HR) from Cox regression models. All statistical tests were two‐sided. Results NHB patients showed a higher proportion of colon tumors (75.8%) than NHW (71.5%), whereas both NUSB (65.9%) and USB (66.9%) H/L cases had less (p < 0.001). In multivariate models, NUSB H/L cases were 15% more likely than NHW to have rectal cancer. Compared to NHW, NHB cases had the greatest proportion of Stage IV diagnoses (26.0%) and were more likely to die of CRC (multivariate HR = 1.12; 95% CI = 1.10–1.15). Instead, NUSB H/L patients were less likely to die of CRC (multivariate HR = 0.87; 95% CI = 0.85–0.89) whereas USB H/L did not differ from NHW. Conclusions NHB and H/L cases have more adverse characteristics at diagnosis compared to NHW cases, with NHB cases being more likely to die from CRC. However, NUSB H/Ls cases showed better survival than NHW and US born H/L patients. These findings highlight the importance of considering nativity among H/L populations to understand cancer disparities.


| INTRODUCTION
Colorectal cancer (CRC) is the third most common cancer in the USA, and the second cause of cancer death representing 8% of all new cancer cases. 1 Non-Hispanic Black (NHB) individuals have higher incidence of CRC compared to non-Hispanic White (NHW) individuals and greater mortality. 2Among NHB populations CRC accounts for 14% and 11% of the cancer burden, and 10% and 9% of cancer mortality, among males and females, respectively. 2ncidence and mortality rates of CRC in the United States are lower among Hispanic individuals, also referred as Latino/a/x individuals (henceforth referred to as H/L individuals) than NHW individuals.However, cancer accounts for a higher proportion of the cancer burden among H/L populations than NHW, with 11% and 9% of deaths among men and women, respectively. 3Among US H/L individuals, CRC is the second and third most common cancer and most common cause of cancer death among males and females, respectively. 3Previously, we and others reported disparities in CRC incidence by region of origin, with H/L of Mexican origin having the lowest incidence. 4,5][8] In California, CRC is the fourth most common cancer and second cause of cancer death, with the highest incidence among NHB individuals, who represent ~5% of the California population.H/L individuals account for 39.3% of the California population and represent 25.9% of the total US H/L population. 9About 35% of H/L in California were born outside the mainland USA, with 82.4% coming from Mexico, 9.4% from Central America, 2.4% from South America, 0.7% from Cuba, 1.5% from Puerto Rico, and 3.6% from other origins. 9n this observational descriptive study, our main objective was to report key socio-demographic, clinical characteristics, and survival patterns of CRC among H/Ls and NHB individuals living in California between 1995 and 2020, and compare them to those observed among NHW individuals, considering differences within H/L patients by nativity, given previous reports that suggest region of origin can be a source of heterogeneity in the H/L population.

| Case identification
We used data from the California Cancer Registry (CCR) including primary CRC cases who were at least 18 years old and diagnosed between 1995 and 2020.Cases were identified by National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results Program (SEER) site codes (21041-21052).H/L status was defined by the NAACCR Hispanic Identification Algorithm (NHIA). 10We assigned nativity as U.S. born (USB) and non-U.S.born (NUSB) based on birthplace information.Individuals born in Puerto Rico were included among the NUSB H/L category, even though those born after 1941 are US born, given previously reported disparities between island and mainland born Puerto Rican individuals.For individuals with missing birthplace (42.37%), we used an algorithm previously reported to impute nativity, 11 and had 3.18% of individuals with missing values.Combining nativity and race we created four mutually exclusive groups: NHW, NHB, USB H/L, and NUSB H/L.Cause of death were identified based on the International Classification of Diseases, 9th Revision and 10th Revision (ICD-9 and ICD-10).

| Study variables
We considered the following sociodemographic variables: age at diagnosis (<62, 62-76, >76, based on tertiles of study cases), insurance status (not insured, managed care, Medicaid, Medicare, other), marital status (single, married, separated/divorced/widowed, unknown), religion (Roman catholic, None/agnostic/atheist, Christian Conclusions: NHB and H/L cases have more adverse characteristics at diagnosis compared to NHW cases, with NHB cases being more likely to die from CRC.
However, NUSB H/Ls cases showed better survival than NHW and US born H/L patients.These findings highlight the importance of considering nativity among H/L populations to understand cancer disparities.

K E Y W O R D S
African American, colorectal cancer, disparities, Hispanic, Latino, registry not catholic, Jewish, other western, Eastern, unknown), nativity (non-U.S. born, U.S. born), and neighborhood Socioeconomic Status (nSES) with quintiles as cut-off point, and as a three-level ordinal variable, which was constructed based on CCR's previously published areabased methodologies, 12,13 grouping variables as: (1) Low: includes lower and lower-middle nSES; (2) Middle nSES; (3) High: includes upper-middle and highest nSES.
We considered the following clinical characteristics: cancer stage based on AJCC staging system (Stage I-IV), as well as the SEER summary stage at the time of diagnosis (localized, regional, distant), tumor size (<2, 2-5, and >5 cm), presence of lymph nodes involvement (N0 and N1), presence of metastasis (M0 and M1), and tumor location (colon vs. rectum, ICD-O-3 codes for colon are C180-189 and C260, and ICD-O-3 codes for rectum are C199 and C209).The tumor location for the colon was further stratified into left-sided (C185-187), right-sided (C180-184), and not otherwise specified (NOS) (C188-189 and C260).We considered the following treatment variables: surgery, chemotherapy, radiation, immunotherapy, hormone therapy, transplant, and other therapy.We also created a composite overall treatment status variable (no treatment, had treatment, treatment recommended or not given or refused), and we considered time from diagnosis to first line of treatment.

| Statistical analyses
Comparison of case frequencies for all variables across racial/ethnic groups and comparisons within H/Ls by nativity, were done using Chi-squared tests.Odds ratios (OR) and 95% confidence intervals (CIs) from logistic regression models were used to identify determinants of CRC tumor location, receipt of treatment, and time from diagnosis to treatment, considering key confounders as described in the appropriate tables.Hazard ratios (HR) and associated 95% CIs were estimated using multivariable Cox proportional hazards regression to evaluate survival differences between NHB and H/L relative to NHW.We tested the proportional hazard assumptions graphically by plotting log follow-up time against log survival ("log-log" plot) and the survival curve derived from the Cox regression model (that imposes a proportional hazard assumption) against the Kaplan-Meier survival curve (without such assumption).Confounding analyses were conducted to determine covariate inclusion in multivariable models, a threshold of >10% in the ORs or HRs was a priori determined and implemented.Cross-product interaction terms were fitted to evaluate potential effect modification by year of diagnosis and religion when appropriate, and multicollinearity was assessed with variance inflation factor (VIF) computation with values of VIF <10 at individual covariatestrata deemed acceptable.All analyses were done using STATA software, version 15.1 (StataCorp LLC, Texas).All tests of statistical significance are two-sided.A pvalue <0.05 was deemed statistically significant.

| Sociodemographic and clinical characteristics of CRC cases in California
Statistically significant differences were observed for year and age at diagnosis, nSES, insurance status, marital status, sex, and religion between NHB, USB H/L, and NUSB H/L cases, when compared to NHW cases (p-values <0.001), and between USB H/L and NUSB H/L cases (pvalues <0.001) (Table 1).We observed a decrease in the proportion of NHWs diagnosed from 2005 onwards, with an increase in the percentage of NHB, USB H/Ls and NUSB H/Ls in more recent years.Moreover, a greater proportion of diagnoses before age 62 among USB H/L (45.6%) and NUSB H/L patients (42.1%) compared to NHB (37.4%) and NHW (26.6%) cases was found.Compared to NHW patients, USB H/L, and NUSB H/L patients had more than double the proportion of diagnoses younger than 50 years old (7.7% NHW vs. 18.2%USB H/Ls and 15.1% NUSB H/ Ls, 11% NHB) and were on average, of lower nSES, uninsured, married, and Roman Catholic.Across all racial/ ethnic categories there was higher proportion of diagnoses among females than males, with the highest proportion among USB H/L patients (55.4%), followed by NUSB H/L patients (53.2%).
Disparities in clinical characteristics were found (Table 2).NHB cases tended to have a higher proportion of colon tumors than NHW patients (74.6% in NHB vs. 71.5% in NHW), with 43.3% of those being right-sided among NHB cases.In contrast, USB H/L patients showed a higher proportion of rectal tumors when compared to NHW patients (33.1% vs. 28.5%,respectively) which increased among NUSB H/L cases (34.1%) when compared to USB H/L patients.A greater proportion of more adverse tumor characteristics were observed among NHB and USB H/L patients when compared to NHW patients.For example,

| Determinants of tumor localization, treatment receipt, and time to treatment reception
When contrasting patients with rectum versus colon tumor localization, sex, age, and marital status were statistically significant determinants using multivariable analyses (Table 3).Compared to NHW cases, NHB patients were 29% less likely to have cancer in the rectum (OR = 0.71, 95% CI = 0.70-0.74,p < 0.001), whereas USB H/L cases were 6% (OR = 1.06, 95% CI = 1.03-1.08,p < 0.001) and NUSB H/L cases were 15% (OR = 1.15, 95% CI = 1.12-1.18,p < 0.001) more likely to have a tumor in the rectum than in the colon (Table 3).Being older, and being female were associated with reduced risk of rectal tumor localization (Table 3).Married cases were slightly less likely to have rectal tumors than single cases.
When determinants of receipt of first line of treatment within 3 months after CRC diagnosis were evaluated, we observed that NHB (OR = 1.14, 95% CI = 1.09-1.19),USB H/L (OR = 1.14, 95% CI = 1.09-1.19) and NUSB H/L (OR = 1.27, 95% CI = 1.22-1.32)patients were all more likely to receive treatment after 3 months since diagnosis compared to NHW patients, with the largest disparity for NUSB H/L cases, after adjustment for age at diagnosis, sex, marital status, nSES, insurance, AJCC stage and tumor location (Table 3).Cases with high nSES were more likely to receive treatment within 3 months compared to those with low nSES.Patients on Medicaid were more likely to be diagnosed after 3 months (OR = 1.25; 95% CI = 1.22 or with stage II-IV were also more likely to receive treatment after 3 months of diagnosis.Compared to single cases, married CRC cases were less likely to have treatment after 3 months from date of diagnosis (Table 3).

| Survival analysis
Results from multivariate analyses (Table 4) showed that when compared to NHW, NHB patients were 12% more likely to die of CRC (HR = 1.12; 95% CI = 1.10-1.15,p < 0.001), whereas NUSB H/L patients were less likely to die (HR = 0.87; 95% CI = 0.85-0.89,p < 0.001).No difference was reported between USB H/L and NHW cases (p = 0.68).Characteristics associated with higher risk of CRC death included being male, older age, low nSES, not receiving treatment, uninsured status, being single, rectal localization, and higher stage.When laterality was evaluated, CRC patients with left-sided tumors were 7% less likely to die of CRC when compared to patients
Stage-specific survival analyses suggested that compared to NHW individuals, the increased risk of death among NHB individuals was larger for Stage I and Stage II, and the reduced risk of CRC death among NUSB H/L individuals was larger among Stage IV patients (Table S3).Moreover, whereas analyses adjusting for stage showed that USB H/Ls individuals did not differ in risk of death compared to NHW individuals (Table 4), stratified analyses by stage at diagnosis showed that USB H/L patients had 11% higher risk of dying of CRC compared to NHW patients for stage II CRC (Table S3).

| DISCUSSION
We report disparities regarding sociodemographic, clinical characteristics, and survival for CRC patients of different racial/ethnic groups in California, and among H/L patients when considering nativity.Key disparities included a greater proportion of younger diagnoses among USB H/L patients and rectal tumors among all H/L, when compared to NHW cases.In addition, we observed that NHB patients had more adverse characteristics at diagnosis, such as the greatest proportion of Stage IV cancer, and presence of metastases, followed by NUSB H/L patients.Importantly, compared to NHW, NHB and H/L patients were less likely to receive treatment and were more likely to receive it after 3 months from diagnosis.Consistent with these disparities, NHB patients were more likely to die of CRC compared to NHW cases.In contrast, despite being diagnosed with many adverse characteristics, USB H/L cases did not differ from NHW in their likelihood of dying of CRC from NHW, and NUSB H/L cases had lower likelihood of dying.
5][16] A study using CCR data suggests the survival disparity between NHB and NHW years has disappeared starting in 2009; however, we saw no statistically significant interactions between year of diagnosis and race/ethnicity. 17Our finding of comparable risk of death for USB H/L and NHW, but reduced risk of death for NUSB H/L is supported by previous studies done in California, 11,17 both adjusting for similar covariates as we did in our study.In addition, previous SEER studies also reported no differences in survival or reduced risk of death for H/L patients compared to NWH, although neither study considered nativity, [14][15][16] and one did not consider clinical variables. 16In addition, a study in Nevada also reported that H/L patients had improved survival compared to NHW, when considering various demographic factors, albeit not nativity, as well as stage and tumor characteristics. 18tudies focusing on mortality have either reported no differences in mortality for H/L patients when compared to NHW, without consideration of SES or nativity, 19 or reduced mortality rates for H/L compared to NHW in California and Texas. 20Interestingly, in the latter study, when disaggregating by nativity, US born H/L males had higher mortality rates than NHW whereas NUSB H/L males and females had reduced mortality in both states.Female USB H/L patients in California also had reduced mortality compared to NHW, whereas those in Texas did not differ from NHW. 20 Finally, a SEER-based study also reported on lower mortality rates for H/L patients and reported that this mortality gap is narrowing, with NHW showing more than double the decline in mortality over the years 2000-2011 compared to H/L patients. 21n contrast to the above findings, 10% higher CRCrelated risk of dying was reported for H/L patients in a study with SEER data between 1988 and 2000, for H/L patients of Mexican origin and H/L patients of unknown origin, but not among those with origin in Puerto Rican, Cuban or South/Central America. 22This study considered treatment received but not SES or nativity.Another study in Texas also reported that H/L patients had ~16% increased risk of dying of CRC compared to NHW, 23 however, this study did not consider treatment received or nativity.][26] However, these studies did not consider differences by stage, nativity, SES, or treatment received, 24,25 or were focused only on metastatic cases and did not adjust for SES or nativity. 26In our study, we provide evidence that nSES is a confounder in the association between race/ethnicity, receipt of treatment, time to treatment, and CRC mortality.Furthermore, higher nSES was inversely correlated with increased likelihood of no treatment reception, prolonged time to diagnosis, and CRC-specific mortality in a linear fashion (all p-trend >0.005) (Tables 3 and 4).
Overall, the observed improved CRC-specific survival among NUSB H/L patients compared to USB H/L and NHW patients is counter-intuitive given the adverse clinical characteristics displayed in greater proportion among NUSB H/L patients.Whether this advantage is due to socio-demographic characteristics, such as increased social support among NUSB H/L patients living in ethnic enclaves, or better health determinants (e.g., diet, lifestyle factors) as previously proposed, 27 or biological differences given potential different genetic admixture patterns among NUSB H/L compared to USB H/L patients, is unclear.Finally, we cannot discard the possibility that some NUSB H/L individuals may have returned to their countries of origin, thus not being captured by vital statistics records in the US.8][29] More studies are needed to understand the key determinants of this potential survival advantage.What is learned for the NUSB H/L population could potentially inform recommendations to improve survival among NHW, NHB, and USB H/L cancer patients.
Strengths of our study include the use of large data from the CCR, with large representation of H/L patients, and consideration of nativity which allowed us to uncover possible disparities within H/L individuals.The main weaknesses of our study are the missing values of cancer stage based on Stage SEER for approximate 15% of cases, non-specified country of origin on 31.9% of NUSB H/Ls; which precluded analyses based on this variable, and missing nativity for 42.37% of study participants.However, sensitivity analysis showed that the distribution of sociodemographic and clinical variables was not affected by the imputation algorithm except in the dichotomous treatment reception variable, which showcased no statistically significant difference between NUSB H/Ls when compared to NHW and USB H/Ls (p-value = 0.783 & 0.076, respectively) (Table S4).Furthermore, when multivariable models were fitted with non-imputed nativity participants, all associations remained virtually unchanged when compared to main findings (Table S5).
In summary, our study reports disparities in CRC survival as well as sociodemographic and clinical characteristics among H/L and NHB patients compared with NHW in California.These findings highlight the importance of considering nativity in the evaluation of cancer patterns among H/L populations and the need to conduct follow-up studies to understand the possible determinants captured by nativity that may explain the observed differences in survival.

3
Multivariable logistic regression models for determinants of tumor localization, receipt of treatment and time between diagnosis and treatment in California CRC patients (1995-2020).
Sociodemographic characteristics of NHW, NHB, and H/L CRC patients in California (1995-2020).werelesslikely to have tumors <2 cm, with a smaller proportion of Stage I diagnosis, localized lesions, and receipt of treatment, including surgery (all p-values <0.001).No statistically significant difference were reported for lymph node involvement and metastasis (p-values = 0.556 & 0.073, respectively) (Table2).
T A B L E 1 a greater proportion of larger tumors (5+ cm), more advanced AJCC stage, distant infiltration, metastasis, and prolonged time to treatment reception (3-6 months) were observed at diagnosis (all p-values <0.001).When compared to USB H/L, NUSB H/L patients -1.40) compared to those without insurance.Patients diagnosed with cancer in the rectum Clinical characteristics of NHW, NHB, and H/L CRC patients in California (1995-2020).
T A B L E 1 (Continued) T A B L E 2 Multivariate CRC-specific survival analyses for CRC cases in California (1995-2020).
*p-trend; five-strata nSES variable (with quintiles as cut-off point) was fitted as a continuous variable in independent model.