Disparities in Refusal of Locoregional Treatment for Prostate Adenocarcinoma

PURPOSE:

We assessed sociodemographic factors associated with and survival implications of refusal of potentially survival-prolonging locoregional treatment (LT, including radiotherapy and surgery) despite provider recommendation among men with localized prostate adenocarcinoma.

METHODS:

The National Cancer Database (2004-2015) identified men with TxN0M0 prostate cancer who either received or refused LT despite provider recommendation. Multivariable logistic regression defined adjusted odds ratios (AORs) with 95% CI of refusing LT, with sociodemographic and clinical covariates. Models were stratified by low-risk and intermediate- or high-risk (IR or HR) disease, with a separate interaction analysis between race and risk group. Multivariable Cox proportional hazard ratios compared overall survival (OS) among men who received versus refused LT.

RESULTS:

Of 887,839 men (median age 64 years, median follow-up 6.14 years), 2,487 (0.28%) refused LT. Among men with IR or HR disease (n = 651,345), Black and Asian patients were more likely to refuse LT than White patients (0.35% v 0.29% v 0.17%; Black v White AOR, 1.75; 95% CI, 1.52 to 2.01; P < .001; Asian v White AOR, 1.47; 95% CI, 1.05 to 2.06; P = .027, race * risk group interaction P < .001). Later year of diagnosis, community facility type, noninsurance or Medicaid, and older age were also associated with increased odds of LT refusal, overall and when stratifying by risk group. For men with IR or HR disease, LT refusal was associated with worse OS (5-year OS 80.1% v 91.5%, HR, 1.65, P < .001).

CONCLUSION:

LT refusal has increased over time; racial disparities were greater in higher-risk disease. Refusal despite provider recommendation highlights populations that may benefit from efforts to assess and reduce barriers to care.

INTRODUCTION

Disparities in diagnosis, treatment, and survival outcomes among men with prostate adenocarcinoma are well-documented and are associated with sociodemographic and biologic factors.^1–3 For example, race,² age,⁴ and distance from treatment facility⁵ have been shown to influence treatment patterns and outcomes. These differences are mediated by a combination of biologic and societal factors such as access to care.^1–3 One factor may be different rates of refusal of potentially survival-prolonging treatments. Providers may not always recommend treatment to patients because of performance status, comorbidities, or greater perceived risk compared with potential benefit⁶; however, patients may also refuse treatment despite provider recommendations.⁷

Definitive locoregional treatment (LT) is particularly critical for patients with intermediate-risk (IR) and high-risk (HR) prostate cancer, where randomized trials have demonstrated an improvement in survival.^8–11 Patterns of refusal represent a means through which survival disparities may be studied and addressed. Differential patterns of refusal among sociodemographic groups may be symptomatic of disparities at a deeper level. For example, refusal of radiation may be indicative of mistrust toward providers or the logistical inability to attend daily radiation treatments.^12,13

Therefore, examining patterns of treatment refusal can highlight subgroups of patients who may benefit from greater efforts to counsel regarding treatment options and to reduce barriers to accessing care.¹⁴ Although not all patients diagnosed with prostate cancer have sufficient life expectancy to benefit from LT, we focused on patients who had been recommended to have LT, which implies an assessment of life expectancy performed by their physicians.¹⁵ A greater understanding of disparities in treatment decisions may be hypothesis generating and may inform efforts to study biases that affect patients' receipt and providers' recommendation of treatment. Using the National Cancer Database, we aimed to identify sociodemographic factors associated with treatment refusal among men with localized prostate adenocarcinoma and assessed survival implications of treatment refusal.

METHODS

Data Source and Patients

We performed a retrospective cohort analysis of data from 2004 to 2015 using the prostate cancer Participant User File from the National Cancer Database (NCDB).¹⁶ The NCDB, established in 1989, is a nationwide US hospital–based cancer registry sponsored by the American Cancer Society and the American College of Surgeons. The NCDB captures approximately 70% of new cancer diagnoses in the United States. We assessed all men with localized (N0M0) prostate adenocarcinoma with low-risk, IR, and HR disease. Men were categorized as having low-risk (Gleason ≤ 6, prostate-specific antigen [PSA] < 10 ng/mL, and cT1-T2a), IR (Gleason 7, PSA 10-20 ng/mL, or cT2b-T2c), or HR disease (Gleason 8-10, PSA > 20 ng/mL, or cT3-T4) based on AUA/American Society for Radiation Oncology guidelines.^17,18 We excluded patients for whom there were incomplete clinical and sociodemographic data included in this study.

Clinical and Sociodemographic Covariates

The primary dependent variable of interest was refusal of LT despite provider recommendation. Refusal of surgery and radiation is coded in the NCDB as “Surgery of the primary site was not performed; it was recommended by the patient's physician, but this treatment was refused by the patient, the patient's family member, or the patient's guardian. The refusal was noted in patient record” and “Radiation therapy was not administered; it was recommended by the patient's physician, but this treatment was refused by the patient, the patient's family member, or the patient's guardian. The refusal was noted in patient record,” respectively. Previous studies among patients with GI malignancies have used these refusal variables in the NCDB.^19–21 As patients who refused surgery may opt for radiation or vice versa, to best capture disparities in treatment decisions, refusal of LT was defined for patients who refused both surgery and radiation.

The primary independent variables of interest were risk group, age, race (White, Black, Native American/Alaskan, Asian/South Asian/Southeast Asian, and Others), year of diagnosis (stratified by 2004-2009 v 2010-2015), facility type, region within the United States (East, Central, Southwest, and West), rural or urban county, distance from treatment facility (far defined as ≥ 10 miles if in an urban county and ≥ 50 miles if in a rural county), Charlson-Deyo comorbidity coefficient (CDCC) (0, 1, 2, or ≥ 3), insurance status (Medicaid and uninsured v insured), and income. Data on income, a proxy for socioeconomic status, made use of quartiles that were derived from the 2012 American Community Survey for each patient's home ZIP code (Table 1).

TABLE 1.

Baseline Characteristics and Overall Proportion and Adjusted Odds of Refusing Localized Treatment

Statistical Analysis

Multivariable logistic regression defined adjusted odds ratios (AORs) with 95% CI of refusing LT when it was recommended, considering all stages together and adjusting for the independent variables of interest listed above; separate models defined AORs with 95% CIs for AUA/American Society for Radiation Oncology low-risk and IR or HR disease. Separate models with race (Black v White race and Asian v White race) * risk group (IR or HR v low-risk) interaction terms were carried out. Interaction term analysis assesses whether the effect of one variable is different based on the value of another variable, that is, whether the AOR associated with risk group differs significantly based on race.²²

For patients with IR and HR disease who have been shown to experience a survival benefit from treatment,^17,18,23 Cox proportional hazards regression analysis modeled the impact of refusal of recommended surgery on overall survival (OS) while adjusting for the clinical and sociodemographic covariates comprising the independent variables of interest listed above. Analyses were performed using Stata/SE 15.1 (StataCorp, College Station, TX). The hospital institutional review board deemed the study to be exempt given the use of de-identified data.

RESULTS

Baseline Characteristics

Of 887,839 men included in the study, the median age was 64 years; 236,494 (26.6%) had low-risk disease, 448,601 (50.5%) had IR disease, and 202,744 (22.8%) had HR disease. White Americans made up 82.1% (n = 729,282), African Americans 13.6% (n = 120,474), Native Americans 0.2% (n = 1,747), and Asian Americans 1.8% (n = 16,152); 2.3% (n = 20,184) were classified as Others. 31,512 men (3.6%) had no insurance or were on Medicaid (Table 1).

Locoregional Treatment Refusal Disparities

Of 887,839 men included, 2,487 (0.28%) refused LT. In the multivariable analysis, having IR or HR disease was associated with lower rates of treatment refusal (0.20%) compared with having low-risk disease (0.49%; AOR; 0.34; 95% CI, 0.31 to 0.36; P < .001). Black patients were more likely to refuse LT than White patients (0.42% v 0.25%; AOR, 1.46; 95% CI, 1.31 to 1.62; P < .001).

A significant interaction term was observed between Black race (v White) and risk group (IR or HR v low-risk) (P_interaction< .001), but not between Asian race (v White) and risk group (P_interaction = .294). On subgroup analysis, among men with low-risk disease (n = 236,494), Black, Asian, and White patients had similar rates of LT refusal (0.66%, 0.54%, and 0.46%, differences not statistically significant, P > .05 in pairwise comparisons). Among men with IR or HR disease (n = 651,345), Black patients and Asian patients were more likely to refuse LT than White patients (0.35% v 0.29% v 0.17%; Black v White AOR, 1.75; 95% CI, 1.52 to 2.01; P < .001; Asian v White AOR, 1.47; 95% CI, 1.05 to 2.06; P = .027, Fig 1A).

FIG 1.

Proportion of men with low-risk and intermediate-risk or high-risk prostate cancer from the National Cancer Database (2004-2015) who refused localized therapy stratified by race (A, referent: White), by insurance status (B, referent: private insurance), and by socioeconomic status (C, referent: median household income < $38,000 USD). Asterisks represent statistical significance in the different rates of locoregional treatment refusal compared with referent (P < .05). USD, US dollars.

No insurance or Medicaid compared with private insurance was associated with greater odds of LT refusal (0.81% v 0.26%; AOR, 2.82; 95% CI, 2.46 to 3.23; P < .001; Fig 1B). Greater county-wide mean household income (> $63,000 in US dollars [USD]) compared with lower income (< $38,000 USD) was associated with lower odds of LT refusal (0.21% v 0.37%; AOR, 0.74; 95% CI, 0.65 to 0.84; P < .001; Fig 1C). Patients diagnosed from 2010 to 2015 were more likely to refuse LT compared with those diagnosed from 2004 to 2009 (0.36% v 0.21%; AOR, 1.81; 95% CI, 1.67 to 1.97; P < .001). Other factors associated with LT refusal included age over 80 years compared with 50 and under (0.83% v 0.19%; AOR, 7.22; 95% CI, 5.50 to 9.49; P < .001), increased comorbidity with CDCC ≥ 3 compared with CDCC = 0 (0.61% v 0.28%; AOR, 1.69; 95% CI, 1.13 to 2.54; P = .011), and receipt of LT at a Community Cancer Program compared with an Academic or Research Program facility (0.32% v 0.54%; AOR, 0.74; 95% CI, 0.65 to 0.84; P < .001) (Table 1; Fig 2). These disparities generally persisted when stratifying by low-risk and IR or HR disease (Appendix Table A1, online only).

FIG 2.

Forest plot for refusal of localized therapy among men with prostate adenocarcinoma from the National Cancer Database (2004-2015). CDCC, Charlson-Deyo comorbidity coefficient; LT, locoregional treatment; NCI, National Cancer Institute.

Locoregional Treatment Refusal and Survival

For men with IR or HR disease (median follow-up 6.07 years), LT refusal was significantly associated with worse OS (n = 598,910, 5-year OS 80.1% v 91.5%, HR, 1.65, P < .001, Appendix Fig A1, online only). The following factors were also associated with worse OS on multivariable analysis: Black and Native American race, older age, community cancer center facility type, higher CDCC, noninsurance status or Medicaid, and lower socioeconomic status (Appendix Table A2, online only).

DISCUSSION

In this national study of 887,839 men diagnosed with prostate adenocarcinoma from 2004 to 2015, we found that Asian American and African American race was associated with increased odds of LT refusal despite provider recommendation; racial disparities in LT refusal among men with IR or HR disease were greater than among men with low-risk disease (P_interaction < .001). Furthermore, later year of diagnosis, greater comorbidity burden, lack of insurance, treatment at a community cancer center, and lower socioeconomic status were also associated with greater LT refusal, among patients with both low-risk disease and IR or HR disease.²⁴ Refusal of LT was associated with worse OS among men with IR or HR disease.

Disparities in access to care among patients with prostate cancer can come in many forms. Although other studies have shown disparities in the receipt of LT for localized prostate cancer based on race, socioeconomic status, and insurance status,^25–27 these disparities may be due to differences in physician recommendations, rather than patient preference. The current study highlights that although there may also be differences in physician recommendations by sociodemographic factors, there are significant differences in patient adherence with recommended treatment; refusal despite provider recommendation assures that a physician thought that treatment would be worthwhile, likely taking into account patient comorbidity and/or life expectancy.¹⁵ Although we found that LT refusal is a reassuringly rare event, patterns of refusal allow insight into other clinical disparities based on these sociodemographic factors. Although only 0.28% of patients were recorded as having refused LT, patients who refuse LT may represent the tip of the iceberg. More common disparities may be present in subtler ways that we are unable to capture, such as the use of shorter-duration androgen deprivation therapy for HR disease managed with radiation, omission of recommended brachytherapy boost, or omission of adjuvant or salvage RT when recommended. Therefore, the recorded refusals, which represent refusal of any LT in the NCDB, may underestimate refusal of survival-improving therapies. Several of these disparities warrant further discussion.

First, our study found greater racial disparities in rates of LT refusal among men with IR or HR disease compared with men with low-risk disease, with African American and Asian American patients refusing LT approximately twice as often as White patients with IR or HR disease. These findings are concerning, as both radiation and prostatectomy improve survival in unfavorable-risk disease.^15,24,28 Racial differences in the refusal of LT may have several possible explanations. Minority patients may have greater distrust toward the healthcare system or cultural values that influence treatment decisions. Issues of mistrust may be associated with historical abuses by the US healthcare system, particularly for African Americans.^12,13,29 The potential role of mistrust in the refusal of LT and of cultural competency as an avenue to mitigate these disparities merits further exploration.^30,31 Among Asian Americans, culturally associated attitudes such as the Filipino bahala na value, that is, the belief that negative events are beyond a person’s control, may influence patterns of LT refusal^32,33; potentially diverse cultural reasons for LT refusal among various cultural and ethnic subpopulations require further study.^34,35

In addition to race-associated factors, our findings highlight that residence in a poorer county and noninsurance status or Medicaid insurance were also associated with increased odds of LT refusal, suggesting that barriers to care may lead patients to refuse treatment.^36–38 For example, even when treatment itself is covered, some patients—such as those in poorer counties or who qualify for Medicaid—confronted with the other costs of treatment may opt out of treatment for fear of the economic ramifications of cancer care.^39–43 Studies among patients with cancer have demonstrated reduced racial disparities in receipt of treatment among the insured compared with noninsured.⁴⁴ The present findings suggest that even among patients who are insured, there are disparities in the patients who agree to treatment, possibly because of other barriers not addressed by insurance alone. Further policy-level efforts are needed to ensure that patients with cancer are supported as they undergo treatment, and radiation oncologists and surgeons who offer LT for prostate cancer should keep these possible barriers in mind when engaging with patients and making recommendations.^45,46

Receiving care at an academic or research facility was also associated with lower odds of refusing LT compared with receiving care at a community cancer program, which may be explained in part by self-selection by some patients who seek care at high-volume facilities and may be less likely to refuse treatment; barriers to care at high-volume facilities likely contribute as well. Studies of refusal of surgery among patients with GI malignancies similarly demonstrated greater rates of refusal among patients treated at nonacademic centers.^47,48 As patients treated at nonacademic centers may have less comprehensive insurance coverage,⁴⁹ financial barriers may play a role in the decision to refuse treatment. Costs not captured by insurance coverage (eg, parking fees) may play a role.⁵⁰ Later year of diagnosis was also associated with increased odds of LT refusal. The increased popularity of active surveillance strategies for favorable-risk disease⁵¹ may theoretically bias patients toward refusing LT despite unfavorable disease. For example, despite NCCN guidelines endorsing active surveillance mainly for low-risk disease and with caution for favorable IR disease,⁵² recent data suggest that even men with unfavorable IR disease have begun to undergo active surveillance in recent years.⁵³

Finally, the association between LT refusal and worse OS among men with IR or HR disease is likely multifactorial. Unmeasured confounders related to patient comorbidities and performance status likely contribute to the survival decrement associated with LT refusal. However, given that all patients included in our study had treatment recommended by a physician (implying at least 5 years of life expectancy per NCCN guidelines¹⁵) and the differences persisted after adjustment for comorbidity score and age, it remains possible that refusal of LT contributed to their worse survival.

Our study must be viewed in terms of its limitations. First, we could not ascertain the reasons for refusal of treatment, which will be an important area for future qualitative studies. Second, the percentage of patients who refuse LT is small. However, the disparities we identified are likely to manifest in other forms of undertreatment based on patient preference that we were unable to identify. Third, based on the available data, there are likely many other confounders, which we could not adjust for that may partially explain some of the disparities that we identified. Fourth, our findings are limited by the use of discrete race groups that cannot completely capture sociocultural nuances. Therefore, our findings are hypothesis generating and may guide the study of disparities in patient- and provider-level biases, as well as efforts to address and mitigate these disparities.

In conclusion, disparities exist in patterns of LT refusal among men with localized prostate cancer. Sociodemographic and clinical traits including later year of diagnosis, Asian American and African American race, greater comorbidity burden, lack of insurance, treatment at a community cancer center, and lower socioeconomic status were associated with increased odds of LT refusal. Racial disparities were greater among men with IR or HR disease. As LT refusal was associated with worse OS among men with IR or HR disease, assessing and addressing reasons for LT refusal may help mitigate survival disparities in prostate cancer. Furthermore, understanding cultural, sociodemographic, and clinical factors associated with refusal may allow more tailored counseling and improved shared decision making.

Appendix

FIG A1.

Overall survival among men with intermediate-risk or high-risk prostate adenocarcinoma who received local therapy and patients who refused local therapy.

TABLE A1.

Adjusted Odds of Localized Treatment Refusal, Stratified by Risk Group

TABLE A2.

AORs Comparing Survival Among Patients With IR or HR Prostate Cancer Who Refused and Received Localized Treatment

DATA SHARING STATEMENT

The patient de-identified National Cancer Database is available from the American College of Surgeons and is not owned by the authors of this work. Data can be requested via https://www.facs.org/quality-programs/cancer/ncdb/puf.

AUTHOR CONTRIBUTIONS

Conception and design: Edward Christopher Dee, Brandon A. Mahal, Vinayak Muralidhar

Financial support: Paul L. Nguyen

Collection and assembly of data: Edward Christopher Dee, David D. Yang, Vinayak Muralidhar

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Disparities in Refusal of Locoregional Treatment for Prostate Adenocarcinoma

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).