Background
The percentage of people without health insurance has been estimated to be 15% of the US population, affecting approximately 48 million individuals in 2012.1 A recent study focusing on the 10 most deadly cancers in the SEER database showed that lack of insurance is associated with advanced-stage disease at diagnosis, less use of cancer-directed surgery and/or radiation therapy, and worse survival.2 These results confirmed 2 previous retrospective studies on the US National Cancer Data Base (NCDB), wherein adequate insurance was associated with the reception of proper cancer screening and timely access to medical care.3,4
In this context, only few studies evaluated the impact of insurance status on prostate cancer characteristics,5,6 and none of these examined the interplay between insurance status, cancer characteristics, and treatment selection. This is of utmost importance, because tumor characteristics represent established predictors of cancer prognosis and play a major role in treatment selection.7,8 Specifically, active surveillance might be considered a valid option in low-risk patients,9 whereas local treatment of the tumor is the preferred option in patients with intermediate- and high-risk prostate cancer,10,11 even in the presence of advanced disease.12–16
For this reason, we aimed to examine the impact of insurance status on tumor risk at diagnosis, and on the selection of initial treatment, after accounting for baseline tumor characteristics. Our hypothesis stated that uninsured status operates as a barrier for appropriate treatment, and specifically that the magnitude of such barrier effect is more relevant in patients with favorable characteristics.
Materials and Methods
Patient Population
We relied on the 2010–2011 SEER database, which records demographic information, tumor site, histology, stage, grade, and treatments performed. The population residing within the areas served by SEER cancer registries is comparable to the general US population, because the catchments for the 18 SEER registries comprise approximately 28% of the US population.17 Collection of information on insurance status within SEER began in 2007. We focused on patients diagnosed in 2010 and 2011, because SEER provided detailed information on clinical tumor characteristics at diagnosis (eg, biopsy Gleason score, number of total biopsy cores taken, and number of positive biopsy cores) starting in 2010.
Individuals with histologically confirmed prostatic adenocarcinoma (International Classification of Diseases for Oncology [ICD-O] code 61.9, histologic code 8140) were identified (n=116,398). Patients aged 65 years or older (n=65,623) were excluded, because this is the age at which most patients become eligible for Medicare.2,17 Patients with metastatic disease at diagnosis (n=3,160) were not included in this study, because systemic therapies (not available in SEER) are generally the primary treatment for these patients. Moreover, excluded were individuals with unknown prostate-specific antigen (PSA) levels (n=5,345), clinical stage (n=204), biopsy Gleason score (n=739), percentage of positive cores (n=17,551), race (n=500), marital status (n=1,796), or insurance status (n=1,087). These selection criteria yielded 20,393 evaluable patients.
Variables Definition
Demographic information included age at diagnosis, race (white vs African American vs Hispanic vs other), marital status (married vs not married), annual family income (provided at the county level in SEER), and insurance-related information. Specifically, insurance status was identified using the “Primary Payer at Diagnosis” codes and was categorized as insured (insured or insured/no specifics), Medicaid coverage (any Medicaid), or uninsured. The SEER definition for insured includes those with private insurance (managed care, health maintenance organization, or preferred provider organization), Medicare, and coverage from the military or Veterans Affairs at the time of initial diagnosis and/or treatment.2,17
Clinical tumor characteristics included PSA level, clinical stage, biopsy Gleason score, number of cores taken at biopsy, and number and percentage of positive biopsy cores. The PSA value for each patient corresponded to the highest PSA value recorded before diagnosis. Clinical stage (CS) was defined using variable “CS Clinical Extension” and was categorized as T1 versus T2a versus T2b–c versus T3 or greater. Biopsy Gleason score was identified using variables “Site-Specific Factor 7–8” and was categorized as 6 or less, 7, and 8 or greater. Information regarding number of cores taken and number of positive cores were identified through variables “Site-Specific Factor 12–13.” Risk groups were defined according to the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Prostate Cancer12: low risk (PSA level <10 ng/mL, clinical stage ≤T2a, and biopsy Gleason score ≤6); high risk (PSA level >20 ng/mL, and/or clinical stage ≥T3, and/or biopsy Gleason score ≥8); or intermediate risk (all the remaining) (to view the most recent version of these guidelines, visit NCCN.org).
Treatment alternatives were identified using variables “Site-Specific Surgery,” “Radiation Therapy,” and “Radiation Sequence with Surgery,”15 and consisted of surgery, external-beam radiation therapy, brachytherapy, or nonlocal treatment of the primary tumor.
Outcomes Definition
The aim of our study was to assess the impact of insurance status on tumor characteristics at diagnosis and treatment selection. Specifically, the 2 evaluated outcomes were: (1) presenting with low-risk prostate cancer at diagnosis, defined as a PSA level less than 10 ng/mL, clinical stage T2a or less, and biopsy Gleason score of 6 or less; and (2) local treatment of the prostate, defined as receiving radical prostatectomy, external-beam radiation therapy, or brachytherapy as initial treatment.
Statistics
Statistical analyses consisted of 3 main steps. First, multivariable logistic regression analysis was used to test the relationship between insurance status (insured vs Medicaid-covered vs uninsured) and presenting with low-risk prostate cancer at diagnosis. Covariates consisted of patient age, race (white vs African American vs Hispanic vs other), marital status (married vs not married), and annual family income.
Second, we examined the impact of insurance status on receiving local treatment of the prostate. We wished to assess whether rates of local treatment were different by baseline risk by testing an interaction with insurance status and recurrence risk. The latter was defined using 2 established pretreatment predictive models: the Stephenson nomogram7 and the UCSF Cancer of the Prostate Risk Assessment (CAPRA) score.8 The interaction term between insurance status and recurrence risk was determined as follows: [Recurrence Risk] [Insurance Status], wherein the risk of recurrence was calculated using both the Stephenson nomogram coefficients and the CAPRA score, whereas insurance status was categorized as “insured vs Medicaid-covered vs uninsured.” Locally weighted scatterplot smoothing (LOWESS) method was used to graphically explore the interactions among insurance status, use of local treatment, and baseline risk of cancer recurrence.
Third, multivariable logistic regression analysis was used to test the relationship between insurance status (insured vs Medicaid-covered vs uninsured) and receipt of local treatment of the prostate. Covariates consisted of patient age, race (white vs African American vs Hispanic vs other), marital status (married vs not married), annual family income, PSA level, clinical stage (T1 vs T2 vs ≥T3), biopsy Gleason score (≤6 vs 7 vs ≥8), and percentage of positive biopsy cores.
All statistical analyses were performed using STA-TA v.12 (Stata Corp, College Station, TX). All tests were 2-sided, with a significance level set at 0.05.
Results
Descriptive characteristics of patients are shown in Table 1. Overall, 18,993 patients (93%) were insured, 849 (4.3%) had Medicaid coverage, and 551 (2.7%) were uninsured.
In the first step of our analyses, we evaluated the association between insurance status and presenting with low-risk disease at diagnosis, which was observed in 5,799 (31%), 185 (22%) and 106 (19%) of insured, Medicaid-covered, and uninsured patients, respectively. Findings from the multivariable logistic regression analysis indicated that Medicaid-coverage (odds ratio [OR], 0.67; 95% CI, 0.57, 0.80; P<.0001) and uninsured status (OR, 0.57; 95% CI, 0.46, 0.71; P<.0001) were significantly associated with a lower likelihood of presenting with low-risk disease at diagnosis, after adjusting for patient age, race, marital status, and annual family income (Table 2).
In the second step of our analyses, we explored the variation of local treatment use according to insurance status, and the baseline risk of cancer recurrence, which was estimated using both the Stephenson nomogram and CAPRA score. In both cases, interaction tests for the hypothesis that treatment disparities among insurance groups vary according to baseline recurrence-risk were statistically significant (insured vs Medicaid-covered, P<0.0001; insured vs uninsured, P<.0001). Figure 1 stratifies patients according to insurance status and shows the proportion of patients receiving local treatment of the prostate plotted against the baseline 5-year cancer recurrence risk, as determined by the Stephenson nomogram (Figure 1A) and the CAPRA score (Figure 1B). Local treatment was used more frequently in insured patients compared with Medicaid-covered and uninsured patients, regardless of baseline cancer recurrence risk. Moreover, treatment disparities significantly increased with the increment in baseline risk of recurrence. Specifically, differences between insured and uninsured patients increased from 10% when the risk was low, to more than 20% when the baseline cancer recurrence risk exceeded 40%, or the CAPRA score was 8 or greater. Medicaid-covered patients showed intermediate rates of local treatment between insured and uninsured patients.
In the third step of our analyses, we tested the relationship between insurance status and local treatment of the prostate. Findings from the multivariable logistic regression analysis indicated that Medicaid-coverage (OR, 0.72; 95% CI, 0.60, 0.86; P=.0003) and insured status (OR, 0.45; 95% CI, 0.37, 0.55; P<.0001) were significantly associated with a lower likelihood of receiving local treatment of the prostate, after adjusting for patient age, race, marital status,
Characteristics of 20,393 Patients With Prostate Cancer Diagnosed Between 2010 and 2011 Within the SEER Database, Stratified According to Insurance Status
Discussion
Our hypothesis stated that uninsured status operates as a barrier for appropriate treatment, and specifically that the magnitude of such a barrier effect is more relevant in patients with favorable characteristics.
Our results rejected our hypothesis, and showed that treatment disparities become more pronounced as the baseline cancer recurrence risk increases.
Multivariable Logistic Regression Analysis Predicting Low-Risk Disease at Diagnosis in 20,393 Patients With Prostate Cancer Diagnosed Between 2010 and 2011 in the SEER Database
First, uninsured patients were more likely to present with intermediate- or high-risk prostate cancer at diagnosis. Moreover, the proportion of patients affected by a Gleason score of 8 or greater at biopsy was significantly lower in the insured group (10%) compared with the Medicaid-covered (18%) and uninsured group (17%). This result could be partially related to access to care, because a delayed diagnosis may be associated with a finding of more advanced and more aggressive disease.18 These findings corroborate a recent NCDB-based study showing that Medicaid-covered and uninsured patients had greater PSA levels at diagnosis and a higher odds of high Gleason score and advanced clinical stage.5 However, a major limitation of such a study is related to the
Proportion of patients receiving local treatment of the prostate plotted against the 5-year cancer recurrence risk as determined by the Stephenson nomogram (A) and the CAPRA score (B). The shaded area represents the distribution of the overall population.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 13, 11; 10.6004/jnccn.2015.0164
Proportion of patients receiving local treatment of the prostate plotted against the 5-year cancer recurrence risk as determined by the Stephenson nomogram (A) and the CAPRA score (B). The shaded area represents the distribution of the overall population.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 13, 11; 10.6004/jnccn.2015.0164
Proportion of patients receiving local treatment of the prostate plotted against the 5-year cancer recurrence risk as determined by the Stephenson nomogram (A) and the CAPRA score (B). The shaded area represents the distribution of the overall population.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 13, 11; 10.6004/jnccn.2015.0164
Multivariable Logistic Regression Analysis Predicting Local Treatment of the Primary Tumor in Patients Between 2010 and 2011 in the SEER Database
Second, Medicaid-covered and uninsured patients had a lower probability of receiving local treatment of the prostate. This finding is in line with a previous SEER-based study, in which insured patients were 2-fold more likely to receive local treatment of the prostate compared with uninsured patients.6 However, we extended our analyses by evaluating the variation of treatment selection according to baseline risk of cancer recurrence. This step is of extreme importance, because tumor risk can significantly influence initial treatment selection. We used 2 previously published and externally validated predictive models to assess the patients' risk of recurrence.7,8 We found that the higher the risk of recurrence, the more pronounced the treatment disparities were among insurance groups. Specifically, compared with insured patients, the risk of not receiving local treatment doubled in uninsured patients when the baseline cancer recurrence risk increased from 10% to 40% according to the Stephenson nomogram, or the CAPRA score increased from 2 to 8. These findings are of great concern and imply that uninsured patients receive local treatment less, when they need it more. Such disparities may importantly compromise cancer outcomes in these individuals, because local treatment of the prostate has been shown to significantly improve cancer control outcomes in high-risk patients.14,25
Third, our study focused on patients younger than 65 years and diagnosed with prostate cancer in the contemporary era. Given the ongoing debates about health care coverage, our findings high-lighted the crucial importance of insurance status for receiving appropriate cancer care, especially for patients with longer life expectancy. Although the Affordable Care Act substantially expands access to health insurance and health care, the latter remains inaccessible for many US consumers. Among the most serious barriers to accessing coverage under the current law is the so-called “family glitch.” Internal Revenue Service regulations for implementing the Affordable Care Act's premium tax credit program deny credits to the families of employees if the employee is offered affordable employer-sponsored coverage even though family coverage is unaffordable. This interpretation of the statute deprives access to premium tax credits to millions of US consumers who have no other affordable insurance options. This barrier could easily be removed, even if by regulation alone.26
Finally, racial disparities persisted even in this contemporary series of patients with prostate cancer. Indeed, findings from the multivariable analysis indicated that African American and Hispanic patients were less likely to receive local treatment of the prostate compared with white patients, after adjusting for demographic and tumor characteristics. These findings were in line with a recent population-based study focused on lung, prostate, breast, and colorectal cancers in which racial disparities did decrease over time.27 Further studies are needed to evaluate whether treatment disparities between racial and sociodemographic groups will persist in the future as Americans move toward equivalent health care access.28
To the best of our knowledge, this is the first study evaluating treatment disparities among insurance groups according to precise assessment of baseline cancer features. The current study relied on the SEER database, which is the only comprehensive population-based data set in the United States, covering approximately 28% of the US population. Therefore, our findings may be more generalizable than those of previous reports.
Despite its strengths, our study is not devoid of limitations. First, we were unable to further discern nonlocal treatment of the tumor, because this information is not available in the SEER database. Specifically, some patients classified as receiving nonlocal treatment of the tumor may have undergone active surveillance or watchful waiting, whereas others may have received systemic therapies. However, systemic therapies, such as androgen deprivation therapy, are not indicated as initial treatment for nonmetastatic prostate cancer.12,13 Second, the SEER database does not contain information on comorbidities that could importantly affect the clinical decision-making. However, a recent SEER-based study has highlighted the absence of a significant association between insurance status and non–prostate cancer–specific mortality, suggesting no difference in baseline comorbidities.6 Therefore, differences in comorbidity profiles among insurance groups are unlikely to explain the observed treatment disparities.
Third, we were unable to address the impact of insurance status on cancer-specific and overall survival, because the current study included only patients diagnosed in the most recent years (2010–2011) with no sufficient follow-up.
Fourth, the exclusion of individuals older than 65 years to avoid the confounding effect due to Medicare eligibility is an additional limitation. However, patients with longer life expectancy might maximally experience barriers to adequate treatment. Moreover, the insured or Medicaid status in the SEER database is captured at a given time point. This fact represents an additional limitation of the study, because prior studies have shown the importance of accounting for history of Medicaid enrollment.29
Finally, a large number of cases were excluded because of missing values, mainly for PSA level (n=5,345) and percentage of positive cores (n=17,551). These exclusion criteria could bias the results of this population-based study. For this reason, we examined the distribution of 2 clinical variables (clinical stage and biopsy Gleason score) before and after excluding patients with unknown PSA level and percentage of positive cores. In the final population (n=20,393), we observed that 70%, 27%, and 3.0% had T1, T2, and T3 or greater disease, respectively. Interestingly, in the initial population (n=47,615) we observed a very similar distribution of clinical stage, as 67%, 30%, and 3.0% of patients had T1, T2, and T3 or greater disease, respectively. Considering biopsy Gleason score, in the final population (n=20,393) 52%, 37%, and 11% had Gleason score of 6 or less, 7, and 8 or greater, respectively. At the same time, in the initial population (n=47,615) the distribution was 54%, 36%, and 10%, respectively. Additionally, we evaluated the distribution of insured, Medicaid-covered, and uninsured patients in both the initial and final population. We observed a similar distribution: 93%, 4.3%, and 2.7%, versus 93%, 4.1%, and 2.9%, respectively. In conclusion, we excluded a large number of cases because of missing values. However, our findings were unlikely to be influenced by these selection criteria.
Conclusions
Insurance status independently impacts prostate cancer features at diagnosis and treatment selection. Specifically, uninsured patients are less likely to present with low-risk disease, and are less likely to receive local treatment of the prostate. Treatment disparities become more pronounced as the baseline cancer recurrence risk increases. Future expansion of insurance programs might lead to earlier cancer detection and improved prostate cancer management, especially in patients with high-risk disease.
The authors have disclosed that they have no financial interests, arrangements, affiliations, or commercial interests with the manufacturers of any products discussed in this article or their competitors. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
References
- 1↑
DeNavas-Walt C, Proctor BD, Smith JC. Income, poverty, and health insurance coverage in the United States: 2012. Current population reports. Available at: https://www.census.gov/prod/2013pubs/p60-245.pdf. Accessed October 12, 2015.
- 2↑
Walker GV, Grant SR, Guadagnolo BA et al.. Disparities in stage at diagnosis, treatment, and survival in nonelderly adult patients with cancer according to insurance status. J Clin Oncol 2014;32:3118–3125.
- 3↑
Ward EM, Fedewa SA, Cokkinides V, Virgo K. The association of insurance and stage at diagnosis among patients aged 55 to 74 years in the national cancer database. Cancer J 2010;16:614–621.
- 4↑
Halpern MT, Ward EM, Pavluck AL et al.. Association of insurance status and ethnicity with cancer stage at diagnosis for 12 cancer sites: a retrospective analysis. Lancet Oncol 2008;9:222–231.
- 5↑
Fedewa SA, Etzioni R, Flanders WD et al.. Association of insurance and race/ethnicity with disease severity among men diagnosed with prostate cancer, National Cancer Database 2004-2006. Cancer Epidemiol Biomarkers Prev 2010;19:2437–2444.
- 6↑
Mahal BA, Aizer AA, Ziehr DR et al.. The association between insurance status and prostate cancer outcomes: implications for the Affordable Care Act. Prostate Cancer Prostatic Dis 2014;17:273–279.
- 7↑
Stephenson AJ, Scardino PT, Eastham JA et al.. Preoperative nomogram predicting the 10-year probability of prostate cancer recurrence after radical prostatectomy. J Natl Cancer Inst 2006;98:715–717.
- 8↑
Cooperberg MR, Pasta DJ, Elkin EP et al.. The University of California, San Francisco Cancer of the Prostate Risk Assessment score: a straightforward and reliable preoperative predictor of disease recurrence after radical prostatectomy. J Urol 2005;173:1938–1942.
- 9↑
Klotz L, Zhang L, Lam A et al.. Clinical results of long-term follow-up of a large, active surveillance cohort with localized prostate cancer. J Clin Oncol 2009;28:126–131.
- 10↑
Bill-Axelson A, Holmberg L, Garmo H et al.. Radical prostatectomy or watchful waiting in early prostate cancer. N Engl J Med 2014;370:932–942.
- 11↑
Vickers A, Bennette C, Steineck G et al.. Individualized estimation of the benefit of radical prostatectomy from the Scandinavian Prostate Cancer Group randomized trial. Eur Urol 2012;62:204–209.
- 13↑
Heidenreich A, Bastian PJ, Bellmunt J et al.. EAU guidelines on prostate cancer. Part II: treatment of advanced, relapsing, and castration-resistant prostate cancer. Eur Urol 2014;65:467–479.
- 14↑
Wilt TJ, Brawer MK, Jones KM et al.. Radical prostatectomy versus observation for localized prostate cancer. N Engl J Med 2012;367:203–213.
- 15↑
Culp SH, Schellhammer PF, Williams MB. Might men diagnosed with metastatic prostate cancer benefit from definitive treatment of the primary tumor? A SEER-based study. Eur Urol 2014;65:1058–1066.
- 16↑
Fossati N, Trinh QD, Sammon J et al.. Identifying optimal candidates for local treatment of the primary tumor among patients diagnosed with metastatic prostate cancer: a SEER-based study. Eur Urol 2015;67:3–6.
- 17↑
Aizer AA, Falit B, Mendu ML et al.. Cancer-specific outcomes among young adults without health insurance. J Clin Oncol 2014;32:2025–2030.
- 18↑
O'Kelly F, Thomas A, Murray D et al.. Can delayed time to referral to a tertiary level urologist with an abnormal PSA level affect subsequent Gleason grade in the opportunistically screened population? Prostate 2013;73:1263–1269.
- 19↑
Dall'era MA, Hosang N, Konety B et al.. Sociodemographic predictors of prostate cancer risk category at diagnosis: unique patterns of significant and insignificant disease. J Urol 2009;181:1622–1627; discussion 1627.
- 20↑
Swan J, Breen N, Coates RJ et al.. Progress in cancer screening practices in the United States: results from the 2000 National Health Interview Survey. Cancer 2003;97:1528–1540.
- 21↑
Moyer VAU.S. Preventive Services Task Force. Screening for prostate cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med 2012;157:120–134.
- 22↑
Carlsson S, Vickers AJ, Roobol M et al.. Prostate cancer screening: facts, statistics, and interpretation in response to the US Preventive Services Task Force Review. J Clin Oncol 2012;30:2581–2584.
- 23↑
Wolf AM, Wender RC, Etzioni RB et al.. American Cancer Society guideline for the early detection of prostate cancer: update 2010. CA Cancer J Clin 2010;60:70–98.
- 24↑
Carter HB, Albertsen PC, Barry MJ et al.. Early detection of prostate cancer: AUA Guideline. J Urol 2013;190:419–426.
- 25↑
Bastian PJ, Boorjian SA, Bossi A et al.. High-risk prostate cancer: from definition to contemporary management. Eur Urol 2012;61:1096–1106.
- 27↑
Aizer AA, Wilhite TJ, Chen MH et al.. Lack of reduction in racial disparities in cancer-specific mortality over a 20-year period. Cancer 2014;120:1532–1539.
- 28↑
Ellimoottil C, Miller DC. Anticipating the effect of the Patient Protection and Affordable Care Act for patients with urologic cancer. Urol Oncol 2014;32:55–58.
- 29↑
Koroukian SM, Bakaki PM, Raghavan D. Survival disparities by Medicaid status: an analysis of 8 cancers. Cancer 2012;118:4271–4279.
