Survival by Hispanic Ethnicity among Cancer Patients Participating in SWOG Clinical Trials

Abstract

Background

Racial disparities in cancer outcomes have been described. It is still unclear if patients of Hispanic ethnicity have better or worse survival outcomes. In this study we evaluated whether Hispanic participants in SWOG clinical trials had different survival outcomes compared to non-Hispanics.

Methods

Adult patients registered in SWOG phase II/III clinical trials between 1986 and 2012 were analyzed. Studies of similar histology and stage were combined. Within each analysis, Kaplan-Meier survival curves were generated to examine differences in outcome by ethnicity. Multivariate Cox regression was used to estimate the association of ethnicity and survival outcomes, controlling for major disease-specific prognostic factors and demographic variables plus area-level income and education to account for socioeconomic status.

Results

29,338 patients registered to 38 trials were included; 5% of them were Hispanic. Hispanics were more likely to be younger and from lower income and education areas (all p<0.05). No differences in survival were observed across tumor types except in the advanced-stage prostate cancer group, where we observed an association between Hispanic ethnicity and worse overall survival (HR=1.40, p=0.006), progression-free survival (HR=1.36, p=0.007), and cancer specific survival (HR=1.42, p=0.013). After adjusting for multiple comparisons, no differences in outcomes were seen.

Conclusions

Hispanic patients participating in SWOG trials receiving uniform treatment and follow-up had similar survival outcomes compared to non-Hispanic patients, with the single exception of the advanced stage prostate cancer group.

Impact

Our study shows that Hispanic patients receiving uniform treatment and follow-up have similar outcomes compared to non-Hispanics.

BACKGROUND

Over the last few decades a sustained decline in cancer mortality has been evident¹. Declines in cancer death rates have been noted for men and women across all ethnic and racial groups. However, the mortality decline has been consistently greater among whites compared to African Americans (AA) or Hispanics¹ suggesting that therapeutic advances are not benefiting all patients equally^{2, 3}. While differences in tumor biology are likely contributing, differences found within tumor subsets suggest that socio-demographic factors like access to quality care or timely treatment could play a role.

Hispanics represent the largest ethnic minority group in the United States, accounting for 17.4% of the total population⁴. Projections indicate that the size of this population will double by 2050⁵, suggesting the importance of evaluating outcomes for this ethnic group. Hispanics have different behavioral and environmental exposures and a high prevalence of important cancer risk factors like obesity and diabetes. Among Hispanics, cancer is the leading cause of death⁶, and Hispanics are more likely to be diagnosed with advanced stages compared with their non-Hispanic counterparts. Recent reports suggest that the cancer death rates among Hispanics are 30% lower compared with non-Hispanic whites, however these observed discrepancies may be exaggerated due to incomplete death ascertainment and challenges associated with follow-up⁶. Comprehensive analyses using prospective data across a broad range of cancers comparing outcomes among Hispanics and non-Hispanics have not been conducted.

In this study, we evaluated outcomes of cancer patients treated in clinical trials according to Hispanic ethnicity. Studying clinical trials participants allows for the analysis of a homogenous patient population with known baseline prognostic factors, controls for treatment, and follows patients similarly at regular intervals with reliable assessment of outcomes. To the best of our knowledge this is the first study evaluating outcomes according to Hispanic ethnicity in a large cohort of clinical trial participants controlling for demographic and clinically-relevant variables.

METHODS

The data included in this analysis originate from SWOG, an NCI-sponsored clinical trials consortium. Beginning in the late 1980s, SWOG routinely collected self-reported data on ethnicity (Hispanic vs. non-Hispanic). We systematically examined the SWOG trials from 1986–2012. We considered only published phase III or large phase II trials enrolling adult patients with available ethnicity data.

Given that the Hispanic population in the U.S. has a younger age distribution, and cancer occurs primarily in older individuals, the proportion of patients with Hispanic ethnicity in trials is often small. To increase power, we combined trials according to disease site, histology and stage. The primary outcome was overall survival (OS), measured as days from registration to last contact (censored) or death by any cause. We also examined progression-free survival (PFS) and cancer specific survival (CSS) as secondary outcomes. PFS was defined as days from registration to last contact (censored), death by any cause, or evidence of relapse or progressive disease. CSS was measured as days from registration to cancer specific death, death by another cause (censored), or last contact (censored). Approximately 29% of patient deaths were accompanied by detailed cause of death information. For the remaining 71%, any death preceded by documented relapse or progression was considered a cancer specific death. All analyses were limited to the first five years after registration in order to focus on cancer-related and treatment-related survival. Patients with last contact date or death date greater than five years post-registration were censored at five years.

Statistical Analysis

Kaplan-Meier survival curves were generated for each analysis, separately for Hispanic and non-Hispanic patients. Multivariate Cox regression was used to estimate the effect of Hispanic ethnicity on OS, PFS and CSS while controlling for disease-specific prognostic factors and socio-demographic variables. Each analysis accounted for age, race, and sex. To account for differences in socioeconomic status (SES) and educational level, SES was estimated using patient zip code. Zip code was linked to median household income and percent of residents who completed high school via 2000 U.S. census data. If the median household income in a patient’s zip code was higher than the overall 2000 U.S. median household income, the patient was designated as high income. A similar method used percent of residents who completed high school in a given zip code to define the area as having high or low education level. In addition, each analysis accounted for important disease-specific clinical adjustment variables. Only clinical adjustment variables with a known effect on survival that were measured in all studies within an analysis were included. Supplementary Table-1 provides a complete list of the variables included in the multivariable model for each tumor subtype category. Each analysis was stratified by study. Results are expressed as Hazard Ratios (HR) and 95% Confidence Intervals (CI). Follow-up time was truncated at 5 years to emphasize survival related to cancer diagnosis and treatment. In addition to disease-specific analyses, we also conducted an omnibus test of the global trend across the different analyses, based on a one-sample t-test of the analysis-specific (signed) chi-square statistics representing the test of the statistical significance of the association of ethnicity and outcome in the multivariable regression models.

Since males represented a very small fraction of the breast cancer patients, they were removed from the breast cancer analyses. Among patients with Non-Hodgkin Lymphoma (NHL) the International Prognostic Index (IPI) was included as a covariate; since the calculation of this index considers age, NHL analyses did not include age. Finally, because there were relatively few patients in the advanced breast cancer group who were ER and PR-negative, only patients who were ER and/or PR-positive were included. We conducted multiple sensitivity analyses: 1) we compared survival outcomes between Hispanics and non-Hispanic whites (rather than all non-Hispanics); 2) we excluded income and education level; 3) we limited follow-up to the first 10 years after randomization (rather than the first five years only); 4) we examined alternative approaches to identifying cancer-specific deaths; and 5) we examined the potential for different trends in the association between ethnicity and outcomes between patients enrolled prior to vs. after the year 2000. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC), all tests were two-sided and used an exploratory significance level of α=0.05. Observed p-values were also compared to a multiple comparisons Bonferroni-adjusted p-value level. Assuming independent testing for a given endpoint (e.g. overall survival), this level was calculated at p=.0036 (p=.05 divided by 14 analyses per endpoint type). Power calculations encompassing the range of sample sizes used in this analysis are shown in the appendix (Supplementary Table-2).

RESULTS

In total, we obtained patient data from 38 unique trials totaling 37,548 enrollments. 8,210 patients were missing data on Hispanic ethnicity, zip code, or one of the important clinical adjustment variables, leaving 29,338 patients available for inclusion in 14 separate analyses (Table-1).

Table-1.

Clinical trials included in the analysis classified according to disease site and stage

Type of cancer	Trials by study number	Enrollment date range	Major eligibility criteria	Important clinical adjustment variables*	Total sample size (No. Hispanic-% Hispanic)
Adjuvant breast, ER− and PR−	S8897, S9313, S9623, S0012, S0221, S0307	1989–2012	Early stage (I-III), no prior chemotherapy for this breast cancer**	• Number of positive lymph nodes: ≥4 vs. <4 • Tumor size:≤5cm vs. >5cm • Pre-menopausal vs. post-menopausal	4,387 (270–6%)
Adjuvant breast, ER+ and/or PR+	S8814, S8897, S9313, S9623, S0012, S0221, S0307	1989–2012	Early stage (I-III), no prior chemotherapy for this breast cancer**	• Number of positive lymph nodes: ≥4 vs. <4 • Tumor size: ≤5cm vs. >5cm • Pre-menopausal vs. post-menopausal	9,991 (482–5%)
Advanced breast, ER+ and/or PR+	S0226, S0500	2004–2012	Stage IV, no prior chemotherapy for metastatic disease	• Her2- vs. Her2+ • Pre-menopausal vs. post-menopausal	958 (43–4%)
Advanced colorectal	S8905, S9420	1989–1999	Metastatic (disseminated), maximum 1 prior adjuvant chemo/immunotherapy	• Performance status: 0–1 vs. 2	1,038 (81–8%)
Adjuvant gastric	S9008	1991–1998	Stages IB-IV (M0)	• Number of involved lymph nodes: none vs. 1 – 3 vs. ≥4 • T Stage:T1–T2 vs. T3–T4	474 (50–11%)
Adjuvant colorectal	S9304, S9415	1994–2000	Stage M0, no prior systemic or radiation therapy	• Number of nodes involved: N0 vs. N1 vs. N2–N3 • Performance status: 0–1 vs. 2 • Invasion of perirectal fat or adjacent organs: T1–T2 vs. T3–T4	2,521 (87–3%)
Advanced prostate	S8894, S9346	1989–2008	D2 disease	• Performance status: <2 vs. ≥2 • Severity of disease: minimal vs. extensive	2,810 (100–4%)
Advanced prostate, hormone refractory	S9916, S0421	1999–2010	Metastatic prostate cancer that is unresponsive or refractory to hormone therapy. Maximum one prior systemic therapy.	• Extraskeletal metastatic disease: yes vs. no • PSA progression only: yes vs. no • Performance status: < 2 vs. ≥ 2	1,547 (81–5%)
AML	S8600, S9031, S9333, S0106	1986–2009	No previous systemic chemotherapy for acute leukemia	• SWOG performance status: <2 vs. ≥2	979 (45–5%)
Advanced NSCLC	S9308, S9509, S0003	1993–2002	No previous systemic chemotherapy or biologic therapy for NSCLC	• Weight loss: <5% vs. ≥5% • LDH: Normal (≤ULN) vs. abnormal (>ULN) • Stage of disease: IIIB vs. IV	1,137 (31–3%)
Advanced Aggressive NHL	S8516, S9704, S9349, S9125, S9240	1986–2007	Intermediate or high grade histology, no prior chemotherapy or radiation therapy for lymphoma***	• IPI risk: Low vs. low-int vs. high-int vs. high	699 (37–5%)
Advanced Indolent NHL	S0016, S8809, S9800, S9911	1988–2008	Stages IIB-IV, no prior chemotherapy or radiation therapy for lymphoma	• IPI risk: Low vs. low-int vs. high-int vs. high	974 (25–3%)
Multiple Myeloma	S9210, S9321, S0232	1993–2007	No prior chemotherapy for this disease	• Stage: I-II vs. III	1,211 (43–4%)
Advanced GIST	S0033	2000–2001	Metastatic or unresectable	• Performance status: 0–2 vs. 3 • Measurable vs. Non-Measurable disease	612 (28–5%)
Total	38 unique trials	1986–2012			29,338 (1,403–5%)

^*Including age (<65 vs. ≥65), race (African American vs. other), and sex (where appropriate).

^**For S0307, chemotherapy was allowed concurrent with registration.

^***Study S9704 allows a single course of CHOP/CHOP-R chemotherapy.

Patient Characteristics

Patient characteristics and clinical prognostic factors according to cancer type are shown in Table-2. Five percent of the participants were Hispanic. Patient characteristics and clinical prognostic factors according to Hispanic ethnicity are shown in Table-3. Hispanic patients participating in SWOG clinical trials were younger and less likely to report being AA; in addition they lived in lower income and education areas (all p<0.05). There were very few statistically significant differences between Hispanic and non-Hispanic patients with respect to clinical prognostic factors.

Table-2.

Patient characteristics and disease-site specific prognostic factors according to clinical trial subgroup analysis

Type of cancer	Percent of patients who are:						Percent of patients with disease-site specific prognostic factors
	≥65yo	Female	African American	Hispanic	Education≤ U.S. median	Income≤ U.S. median
Adjuvant breast ER− and PR−***	7.5%	100%	12.7%	6.2%	34.2%	50.0%	Tumor > 5 cm: 8.9%	Post-menopausal: 47.5%	≥ 4 nodes: 12.5%**
Adjuvant breast ER+ and/or PR+***	13.2%	100%	7.2%	4.8%	30.5%	46.0%	Tumor > 5 cm: 8.7%	Post-menopausal: 56.0%	≥ 4 nodes: 22.6%**
Advanced breast ER+ and/or PR+	41.9%	100%	9.4%	4.5%	31.8%	53.5%	Post-menopausal: 93.5%	Her2negative: 88.4%**
Advanced colorectal	39.7%	41.2%	13.0%	7.8%	44.3%	60.3%	Performance status 2: 10.4%
Adjuvant gastric	33.5%	27.6%	17.7%	10.5%	50.0%	63.3%	≥ 4 nodes: 42.0%	Stage T3-T4: 67.7%
Adjuvant colorectal	37.9%	40.7%	6.9%	3.5%	35.6%	56.5%	N2–3 nodes: 30.9%	Stage T3-T4: 85.2%	Performance status 2: 2.9%
Advanced prostate	66.7%	0%	20.1%	3.6%	39.9%	55.9%	Performance status ≥ 2: 10.7%	Severity = Extensive: 74.2%
Advanced prostate, hormone refractory	70.7%	0%	13.7%	5.2%	36.3%	51.8%	Extraskeletal metastasis: 52.9%	Progression other than PSA: 81.1%	Performance status ≥ 2: 9.2%
AML	33.7%	44.8%	8.3%	4.6%	39.5%	56.6%	Performance status ≥ 2: 21.9%
Advanced NSCLC	41.4%	33.5%	12.4%	2.7%	45.8%	62.2%	≥ 5% weight loss: 39.3%	Abnormal LDH: 37.6%	Stage IV: 89.0%
Advanced aggressive NHL	15.2%	39.5%	9.2%	5.3%	36.9%	53.8%	IPI risk = High/High-Int.: 40.3%
Advanced indolent NHL	11.9%	42.4%	4.4%	2.6%	32.6%	49.9%	IPI risk = High/High-Int.: 13.0%
Multiple Myeloma	17.3%	40.9%	17.8%	3.6%	37.3%	55.3%	Stage III: 59.1%
Advanced GIST	46.2%	44.8%	11.4%	4.6%	38.2%	48.5%	Performance status 3: 3.3%	Measurable disease: 94.4%
Total	27.5%	40.1%1	10.7%	4.8%	35.1%	51.5%

¹Total percent includes only the non-sex specific cancers

^**Percent of non-missing data

^***S0012 did not collect baseline information on the number of positive lymph nodes. In order to include patients from S0012 in the analyses, a dummy covariate was created to denote missing node involvement information. Similarly, since approximately 11% of the patients in the advanced breast cancer group had missing Her2 status, a dummy covariate was created to denote Her2 status missing information.

Table-3.

Patient characteristics and disease-site specific prognostic factors according to Hispanic ethnicity and clinical trial subgroup analysis (patients with complete data).

Type of cancer	Percent of patients in high risk categories (% of all Hispanics vs. % of all non-Hispanics)
	≥65yo	Female1	African American	Education≤ U.S. median	Income≤ U.S. median	Disease-site specific prognostic factors
Adjuvant breast ER− and PR−	6 vs. 8		3 vs. 13*	68 vs. 32*	68 vs. 49*	Tumor>5 cm: 20 vs. 8*	Post-menopausal: 40 vs. 48*	≥ 4 nodes: 9 vs. 13
Adjuvant breast ER+ and/or PR+	5 vs. 14*		2 vs. 7*	70 vs. 29*	66 vs. 45*	Tumor>5 cm: 13 vs. 8*	Post-menopausal: 51 vs. 56*	≥ 4 nodes: 29 vs. 22*
Advanced breast ER+ and/or PR+	30 vs. 42		5 vs. 10	53 vs. 31*	70 vs. 53*	Post-menopausal: 95 vs. 93	Her2negative: 92 vs. 88
Advanced colorectal	19 vs. 41*	41 vs. 41	2 vs. 14*	83 vs. 41*	78 vs. 59*	Performance status 2: 11 vs. 10
Adjuvant gastric	20 vs. 35*	42 vs. 26*	2 vs. 20*	72 vs. 47*	72 vs. 62	≥4 nodes: 32 vs. 43	Stage T3-T4: 66 vs. 68
Adjuvant colorectal	28 vs. 38	43 vs. 41	0 vs. 7*	74 vs. 34*	76 vs. 56*	N2–3 nodes: 26 vs. 31	Stage T3-T4: 89 vs. 85	Performance status 2: 8 vs. 3*
Advanced prostate	52 vs. 67*		2 vs. 21*	77 vs. 38*	67 vs. 56*	Performance status ≥2: 13 vs. 11	Severity = Extensive: 79 vs. 74
Advanced prostate, hormone refractory	54 vs. 72*		0 vs. 14*	63 vs. 35*	58 vs. 51	Extraskeletal metastasis: 58 vs. 53	Progression other than PSA: 89 vs. 81	Performance status ≥2: 12 vs. 9
AML	16 vs. 35*	47 vs. 45	0 vs. 9*	71 vs. 38*	76 vs. 56*	Performance status ≥2: 11 vs. 22
Advanced NSCLC	39 vs. 42	39 vs. 33	0 vs. 13*	77 vs. 45*	77 vs. 62	≥5% weight loss: 48 vs. 39	Abnormal LDH: 35 vs. 38	Stage IV: 81 vs. 89
Advanced aggressive NHL	8 vs. 16	32 vs. 40	0 vs. 10*	62 vs. 35*	73 vs. 53*	IPI risk=High/High-Int.: 35 vs. 41
Advanced indolent NHL	20 vs. 12	36 vs. 43	4 vs. 4	44 vs. 32	56 vs. 50	IPI risk=High/High-Int.: 20 vs. 13
Multiple Myeloma	9 vs. 18	37 vs. 41	5 vs. 18*	67 vs. 36*	70 vs. 55	Stage III: 65 vs. 59
Advanced GIST	21 vs. 47*	46 vs. 45	0 vs. 12	71 vs. 37*	64 vs. 48	Performance status 3: 4 vs. 3	Measurable disease: 96 vs. 94
Total1	17 vs. 28*	41 vs. 40	2 vs. 11*	70 vs. 33*	68 vs. 51*

¹Percent provided for the non-sex specific cancers only

^*Indicates a significant difference between Hispanic and non-Hispanic groups (p<0.05)

Outcomes by Ethnicity Status

Figure-1 shows Kaplan-Meier OS curves comparing Hispanic and non-Hispanic patients. There were no clear differences according to Hispanic ethnicity in OS; the only exception was among patients with advanced prostate cancer where Hispanics had worse survival. Interestingly, no survival differences were observed among patients with hormone-refractory advanced prostate cancer. Multivariate Cox regression results showed a significant association between Hispanic ethnicity and worse OS in the advanced-stage prostate cancer group (HR=1.40, 95% CI: 1.10–1.78, p=.006). Consistent with the OS results, Hispanic patients in the advanced-stage prostate cancer group also had worse PFS (HR=1.36, 95% CI: 1.09–1.70, p=.007) and CSS (HR=1.42, 95% CI: 1.08–1.87, p=.01). Hispanics had better CSS in early-stage ER and PR-negative breast cancer (HR=0.63, 95% CI: 0.43–0.93, p=.02). There were no other statistically significant differences in survival outcomes. Importantly, none of these results achieved statistical significance under a multiple testing p-value level of p=.0036. Results from the multivariable Cox models are illustrated in Figure-2. There was no evidence of a general trend towards better or worse outcomes for Hispanic patients across the panel of different analyses according to the global tests.

Figure-1. Kaplan-Meier survival curves for Hispanic vs. non-Hispanic patients.

Numbers along the bottom of each graph represent at-risk counts for Hispanic (in blue) and non-Hispanic (in red) patients at the corresponding time in years along the x-axis.

Figure-2. Multivariable analysis of OS, PFS and CSS according to Hispanic ethnicity by tumor site analysis.

Forest plots are shown for each survival outcome by cancer site category and represent Cox regression-based hazard ratios for Hispanic versus non-Hispanic ethnicity. Each horizontal bar represents the 95% confidence interval for the associated hazard ratio (HR; black dot). HRs to the left of the line of equal hazard (HR=1) indicate better survival for Hispanic patients, and HRs to the right of line of equal hazard indicate worse survival for Hispanic patients. Cancer groups are shown in order of the HR from the overall survival Cox regressions.

Sensitivity Analyses

When using non-Hispanic whites as reference category, the results remained grossly unchanged (Supplementary Figure-1). The exclusion of SES variables changed some of the results, highlighting the importance of social factors in cancer outcomes. Similar to the full model, Hispanic patients with advanced stage prostate cancer had worse outcomes. Hispanic patients with early stage ER and/or PR-positive breast cancer had worse PFS than their non-Hispanic counterparts. Hispanics with early stage ER and PR-negative breast cancer had better CSS than non-Hispanics. Otherwise, there were no statistically significant differences between Hispanic and non-Hispanic patients (data not shown). In no disease group was there evidence that the association of Hispanic ethnicity and outcome differed between categories of education or income using interaction tests. Sensitivity analyses evaluating a longer follow-up time of 10 years are shown in Supplementary Figure-2, indicating generally similar results. We examined different approaches to identifying cancer-specific deaths, finding no global pattern of better or worse cancer-specific deaths for Hispanic patients under any of three alternative scenarios (Supplementary Figure-3). Finally, we found no instances of statistically significant associations of Hispanic ethnicity and outcomes in the opposite directions between patients enrolled prior to and after the year 2000, and there was no statistically significant evidence that global patterns in the three outcomes (OS, PFS, and CSS) differed between patients registered prior to vs. after the year 2000 (data not shown).

DISCUSSION

In this large cohort of patients participating in therapeutic SWOG clinical trials we found little evidence that overall, progression-free, or cancer-specific survival differed between Hispanics and other ethnic groups; indeed there were no differences after adjusting for multiple comparisons. These results contrast with recent cancer statistics indicating that stage adjusted mortality rates among Hispanics are different compared to non-Hispanic whites⁶. The mortality rate for males (per 100,000 and age adjusted to the 2000 U.S. standard population) is estimated to be 148 for Hispanics and 211 for non-Hispanic whites; among females mortality rates are 99.4 and 149.2 respectively.

This differential mortality favoring Hispanics has been reported across tumor types and across other chronic diseases and is known as the Hispanic paradox. This term refers to the epidemiological finding that Hispanics in the U.S. tend to have substantially better health outcomes than the average population in spite of what their aggregate socioeconomic indicators would predict^{7, 8, 9, 10}. In the current study, using a homogenous population of clinical trial participants, after controlling for important clinical factors, we demonstrate similar survival outcomes among Hispanics compared to non-Hispanics. Importantly, the patients included in our cohort were treated with similar protocols and were followed prospectively in a uniform fashion. Our findings suggest that current therapies provide similar benefits regardless of Hispanic ethnicity. This finding is important, because cancer among Hispanics accounted for 126,000 new cancer cases in 2015⁶. Moreover, an increasing number of cancer cases among Hispanics is expected in the future since Hispanics in the U.S. are younger on average and cancers are more common in older patients^{5, 11}.

The only group of patients where we observed a detriment in outcomes was the advanced prostate cancer group, although this difference was not statistically significant under multiple testing. Also of interest is the absence of any pattern of statistically significant associations for patients with advanced hormone refractory prostate cancer. Of note, this patient group only included registrations beginning the year 1999, whereas patients with advanced prostate cancer we enrolled from 1989 through 2008. Although we found no general pattern of temporal trends in the association of ethnicity and outcomes, temporal trends could be influential in this disease setting. Differential outcomes according to race/ethnicity have been reported in prostate cancer patients, it has been described that AA patients have mortality rate double that of white men,^{12, 13};¹⁴. In a similar analysis to the one presented here, Albain et al. reported that, after adjusting for important covariates, AA prostate cancer patients despite being treated similarly had a 19% increased risk of dying compared to whites³.

Relatively few studies have reported the mortality rates among Hispanic prostate cancer patients. A recent study using the SEER-Medicare database reported similar all-cause and prostate-specific mortality rates for Hispanics when compared with non-Hispanic whites¹⁵. In the models before adjusting for multiple comparisons, we observed that among patients with advanced prostate cancer, Hispanics had an increased risk of dying. It is possible that comorbidities, obesity or differences in the volume of the disease could have had an effect otcome.^{6, 12, 16,17, 18}. While the implications of worse outcomes among Hispanic patients with prostate cancer could be relevant considering that prostate cancer is the most frequent cancer in Hispanic males (22% of the cases.); we must emphasize that after adjusting for multiple comparisons this difference was no longer of statistical significance.

In our sensitivity analysis excluding SES-related variables, Hispanics with ER and/or PR-positive early stage breast cancer had significantly worse PFS compared to non-Hispanics. While the benefit of adjuvant endocrine therapy among these patients is unquestionable, the therapy is prolonged and costly and this could have a differential effect among Hispanics; known to have overall lower income and educational levels compared to whites. Early discontinuation of endocrine therapy is a frequent problem^{19, 20}, in the largest study to date evaluating adherence in this group of patients, a 32% rate of early discontinuation was observed²¹. Early discontinuation and non-adherence have been associated with worse outcomes and patients of ethnic minorities and disadvantaged income and educational levels are particularly vulnerable^{22, 23}. Unfortunately in our analysis we are not able to take into account adherence to adjuvant endocrine therapies.

Despite the strengths of our unique analysis, we acknowledge that the cohort of patients may not represent the characteristics of the overall Hispanic population. Just 5% of patients overall were Hispanic, which is actually greater than the proportion of Hispanics in the population for the age distribution represented by clinical trial patients²⁹. Nonetheless this low rate presents potential issues with respect to both selection bias and statistical power. It is well known that clinical trial participants tend to be younger, healthier, have fewer comorbidities and belong to higher SES and educational levels compared with non-participants^24,25,26. Different studies have characterized clinical trial participants, with some suggesting that ethnic minorities are underrepresented^{3, 27}. In addition, it is clear that clinical factors such as performance status, number of comorbid conditions and cancer stage also determine clinical trial participation²⁸. We cannot exclude the possibility that Hispanics participating in SWOG clinical trials have biological and socio-demographic characteristics different from the general Hispanic population. However, although these factors likely influence overall estimates of survival outcomes, they are less likely to influence relative differences by ethnicity. Finally, the analysis of cancer-specific survival was limited by the absence of specific cause-of-death information for many patients. Despite these limitations, our study demonstrates for the first time that, when treated homogenously, Hispanics have similar outcomes compared to non-Hispanics. Efforts should continue to promote and engage ethnic minorities to participate in clinical trials. Only by including large and representative samples of patients in clinical trials will we be able to ensure generalizability of the obtained results to the entire population.

In summary, in this study we report for the first time in a large cohort of clinical trial participants outcomes according to Hispanic ethnicity. After adjusting for multiple comparisons, no notable differences in survival outcomes by ethnicity were evident, demonstrating that Hispanic ethnicity was not a predictor of survival among SWOG clinical trial participants.

CONDENSED ABSTRACT.

In this large study evaluating outcomes among cancer patients participating in phase II/III SWOG clinical trials (n=29,338) and receiving uniform treatment and follow-up, we observed similar survival outcomes between Hispanic and non-Hispanic patients.