Population-Based Assessment of Determining Treatments for Prostate Cancer

Importance  Many men with indolent prostate cancer often opt for radical prostatectomy or radiotherapy treatment for their disease. These men may experience considerable detriments of quality of life owing to sexual, urinary, and/or rectal toxic effects associated with these treatments. Without a better understanding of the mutable agents and predictors of treatment types, diffusion of expectant management among these men will be slow.

Objective  To determine population-based predictors for treatment and use of watchful waiting or active surveillance for indolent prostate cancer.

Design, Setting, and Participants  We used Surveillance, Epidemiology, and End Results (SEER)–Medicare linked data. A total of 37 621 men in the general community diagnosed as having prostate cancer from 2004 to 2007 were followed until December 31, 2009.

Exposures  Watchful waiting or active surveillance, radiation therapy, or radical prostatectomy.

Main Outcomes and Measures  We used mixed-effects logistic regression analysis to determine the factors associated with aggressive treatment and use of watchful waiting or active surveillance for men with prostate cancer.

Results  The most common treatment type is radiation therapy (57.9% [95% CI, 57.4%-58.4%]), followed by radical prostatectomy (19.1% [95% CI, 18.7%-19.5%]) and watchful waiting or active surveillance (9.6% [95% CI, 9.3%-9.9%]). Moreover, patients and providers significantly integrate age (odds ratio [OR], 0.32 [95% CI, 0.29–0.35]) and comorbidities (OR, 0.62 [95% CI, 0.56–0.68]) when determining radical prostatectomy, while regional variation (OR, 0.57 [95% CI, 0.47–0.68]) and referral patterns (OR, 44.46 [95% CI, 41.04–48.17]) influence the use of radiation therapy. Patient demographics and tumor characteristics significantly account for 40% of patients undergoing prostatectomy, 12% choosing watchful waiting or active surveillance, and only 3% undergoing radiotherapy.

Conclusions and Relevance  There is increased use of radiotherapy among patients with indolent prostate cancer with limited to no correlation with tumor biology. Active surveillance was underused, and a significant proportion of the variance was unexplained. Further research into qualitatively describing the contributing factors that drive decision-making recommendations for prostate cancer patients is needed.

Prostate cancer remains the most commonly diagnosed solid organ tumor among US men, with an estimated 233 000 new cases and 29 480 deaths in 2014.¹ With recent stage migration,² the natural history of prostate cancer has shifted toward a more indolent course in most newly diagnosed cases.³ Early detection of prostate cancer cases via prostate-specific antigen (PSA) screening has produced a downward stage migration at presentation.⁴ Earlier diagnosis, along with therapeutic advances, has facilitated the increased use of aggressive local treatment, particularly radical prostatectomy, and radiation therapy.⁵Adverse effects associated with these treatments are substantial from both a clinical and an economic standpoint.^6,7

With more indolent cancers being diagnosed, active surveillance protocols have been established, with promising oncologic results, and are associated with the greatest quality-adjusted life expectancy.⁸ Moreover, the benefit of local treatment, such as surgery vs watchful waiting with respect to prostate cancer-specific mortality, seems to be limited to men younger than 65 years and in those with intermediate-risk prostate cancer.⁹ Prior studies have suggested patient, clinician, and socioeconomic factors contributing to overtreatment of low-risk disease.^10-13 Given these factors, patients must often consider the recommendations of their physician, the aggressiveness of their cancer, and whether active surveillance is preferred over definitive treatment and the pursuant morbidity (mainly urinary, bowel, and sexual dysfunction) and health care costs.^14,15 The fact that patients continue to pursue treatment for indolent disease underscores the importance of a physicians’ recommendation in this complex decision-making process.¹⁴ Barriers to active surveillance use have included concerns about long-term disease outcomes, the perception that most men will ultimately undergo treatment, and concerns about the quality of life of men who elect active surveillance.^8,16 In addition, recent studies have suggested that increased self-referral and non–tumor biology–related factors might contribute substantially to physicians’ treatment recommendations.^13,14 The purpose of our population-based study was to identify determinants for use of watchful waiting or active surveillance (WW-AS) among men with indolent prostate cancer, as well as various treatments in a contemporary cohort of elderly Americans.

Box Section Ref ID

We used linked Surveillance, Epidemiology, and End Results (SEER)–Medicare data from the National Cancer Institute, which contains claims records on individuals 65 years or older. We restricted our analysis to patients with prostate cancer diagnosed from 2004 through 2007 and followed until December 31, 2009. Because complete PSA level, clinical stage, and biopsy grade information was not made available in SEER until 2004, we excluded patients prior to this date to derive our cohort. SEER data are summarized in the Patient Entitlement and Diagnosis Summary File (PEDSF) (http://appliedresearch.cancer.gov/seermedicare/aboutdata/PEDSF.pdf ); the database contains information on patient demographics, tumor characteristics, and follow-up information. Race/ethnicity was reported in the study and defined by SEER to capture demographics and observe any disparities. The PEDSF was linked with 100% of Medicare claims from inpatient, outpatient, and national claims history files, and was restricted to patients who had Medicare Fee-for-Service coverage, and for whom Medicare Parts A and B claims data were available 12 months prior to and 24 months after diagnosis of prostate cancer.¹⁷ Using the Medicare Provider Analysis and Review, outpatient, and carrier files, we were able to identify and subsequently categorize treatment type into WW-AS, cryotherapy, radiation therapy, radical prostatectomy, and androgen-deprivation therapy (ADT).

We identified 45 408 men from SEER-Medicare linked data diagnosed as having prostate cancer between 2004 and 2007 with follow-up of Medicare services through 2009. We excluded men for the following reasons: the diagnosis was obtained from a death certificate or autopsy; this was not the first and only malignant neoplasm; prostate cancer was not pathologically confirmed; the patient was enrolled in Medicare for end-stage renal disease or disability; the date of diagnosis in SEER differed from that in Medicare by more than 3 months; the patient was younger than 65 years at diagnosis; the month of diagnosis was invalid; the patient had concurrent health maintenance organization coverage and/or was not enrolled in Medicare Part A and B throughout the study period; information was lacking from 1 year prior to and 2 years after diagnosis; the initial diagnostic biopsy for prostate cancer was lacking; the Gleason grade, PSA level, and clinical stage were unknown; and socioeconomic and comorbidity data were unknown. The final cohort consisted of 37 621 men.

From the PEDSF, we determined patient age, race/ethnicity, marital status, Gleason grade, PSA level, clinical stage, SEER region, and area of residence. Using Gleason grade, PSA level, and clinical stage, we were able to categorize tumor risk strata according to D’Amico risk—low, intermediate, or high.¹⁸ We imputed patient socioeconomic status by using 2000 US Census data to derive quartiles of zip code–level median household income and percentage of residents who were high-school graduates.¹⁹ We used the modification by Warren et al¹⁷ of the Charlson Comorbidity Index to quantify severity of preexisting comorbidities. For each patient, we noted the clinician and institution where the initial prostate cancer was diagnosed, using the Unique Physician Identifier Number (UPIN) and the corresponding institution. We limited our cohort to those who had survived at least 2 years to determine varying treatments during the initial 2-year period after diagnosis.

The aim of our study was to examine the use of various treatment options and identify patient- and clinician-level variables to drive treatment type. Treatment options were categorized into WW-AS surveillance (no definitive treatment within 2 years of diagnosis), cryotherapy, radiation therapy (brachytherapy, intense modulated radiation therapy, and external beam radiation therapy), radical prostatectomy, and ADT. To examine the factors that drive the decision to pursue popular treatment options, we limited our multilevel analysis to those who underwent WW-AS, radiation therapy, or radical prostatectomy. Patients who received combination treatments were classified according to primary treatment modality received. Because the long-term benefits of varying treatment options are attributable to tumor risk and the overall health of the patient, we quantified the proportion of the variance that was attributable to patient demographics, tumor characteristics, region, and year, as well as consultation with other specialists.

We first compared patient demographics and tumor characteristics with treatments—WW-AS, cryotherapy, radiation therapy, radical prostatectomy, and ADT—using χ² analyses and Fisher exact test. Since receipt of health care services may be clustered on the treating physician, we generated mixed-effects logistic regression models to account for both fixed and random effects associated with treatment type. Each model included patient age, race/ethnicity, marital status, Charlson comorbidity score, education, household income, region, area of residence, clinical stage, Gleason grade, PSA level, D’Amico tumor risk, and consultation with subspecialists as fixed terms, while each UPIN was appended to the random effects part of the mixed-effects model. Estimates in the multivariate mixed-effects model are reported as odds ratios (ORs) with corresponding 95% confidence intervals.

Partitioning of variance was conducted using the following equation: (σ²_F)/(σ²_F + τ²₀ + σ²_R), where σ²_F is defined as the variance of the fixed term (covariate or group of covariates) derived from latent-variable approach, τ²₀ is defined as the intercept (level 2) variance, and σ²_R is defined as the level 1 residual variance (π²/3 in our logistic model).²⁰ Groups of patient- and clinician-level variables were included as fixed effects for each treatment type. We stratified these groups as the following: (1) sociodemographic (age, race/ethnicity, marital status, socioeconomic status, comorbidities), (2) tumor risk (clinical stage, PSA level, and Gleason grade), (3) region and year (SEER region, area of residence, and year of diagnosis), and (4) consultation with other specialists (medical oncology and radiation oncology). Clinician-attributable residual intraclass correlation coefficient (ICC)—representing unexplained clinician-level variance—was estimated from the full model of each outcome measure. Unexplained surgeon factors were derived from the intraclass correlation coefficient of the unconditional or null model of each treatment. Unexplained patient factors were derived from the residual variance. Assessment of goodness of fit was determined by a likelihood-ratio test that compares the mixed-effects logistic model with standard (marginal) logistic regression. Because the likelihood-ratio test comparing the model to ordinary logistic regression is significant for all 3 treatment types (WW-AS, prostatectomy, and radiotherapy), we chose to use the mixed-effects logistic regression model.

All analyses were performed using SAS statistical software (version 9.2; SAS Inc) and STATA software (version 11.1; StataCorp LP). All statistical tests were 2-tailed, and the probability of a type I error was set at P < .05. The institutional review board at the University of California, Los Angeles, exempted our study protocol.

The plurality of the cohort was 70 to 74 years of age, white, married, without any comorbid conditions, and diagnosed in a metropolitan environment in the western United States (Table 1). The tumors of most men were stage T1, with PSA levels of 4.1 to 9.9 ng/mL, Gleason grade 6 or lower, and D’Amico intermediate risk disease.

We performed a bivariate analysis comparing varying treatment options according to the tumor biological characteristics as shown in Table 2. While the effect size in differences in outcomes varied, statistical significance was maintained across all metrics. Greater frequency of undergoing radiotherapy was significantly maintained regardless of associated tumor risk categories when compared with other treatment options. The most common treatment type is radiation therapy (57.9% [95% CI, 57.4%-58.4%]), followed by radical prostatectomy (19.1% [95% CI, 18.7%-19.5%]), ADT (10.8% [95% CI, 10.4%-11.1%]), WW-AS (9.6% [95% CI, 9.3%-9.9%]), and cryotherapy (2.6% [95% CI, 2.4%-2.8%]). Treatment with radiation therapy was the most common treatment (48%–66%) irrespective of stage, PSA level, Gleason grade, or D’Amico tumor risk. Radical prostatectomy was significantly influenced by PSA level: from 24% for those with PSA level of 4.0 ng/mL or lower to 9% for those with values of at least 20 ng/mL. Use of WW-AS was guided by clinical stage, Gleason grade, and D’Amico tumor risk strata. Androgen-deprivation therapy was significantly influenced by clinical stage, PSA level, Gleason grade, and D’Amico tumor risk.

We then examined the association between patient demographics, tumor characteristics, and regional factors with treatment type (Table 3). The use of WW-AS increased with advanced age from age 70 to 74 years (OR, 1.82 [95% CI, 1.61-2.06]) to age 80 years or older (OR, 5.12 [95% CI, 4.48-5.86]). Consultation with a medical oncologist resulted in increased use of WW-AS (OR, 1.83 [95% CI, 1.51-2.22]). Asian descent was associated with least likely use of WW-AS (OR, 0.77 [95% CI, 0.59-0.99]) as well as married men (OR, 0.69 [95% CI, 0.62-0.77]). As expected, lower use of WW-AS was associated with tumor characteristics (PSA level, 4.1-19.9 ng/mL; Gleason grade >6; D’Amico high-risk disease). Men referred to a radiation oncologist were less likely to be offered WW-AS (OR, 0.19 [95% CI, 0.17-0.21]). Increased use of radiation therapy was found with advancing age, more significant comorbidities, and tumor characteristics (PSA level, 4.1-19.9 ng/mL) and most likely used when referred to a radiation oncologist (OR, 44.5 [95% CI, 41.0-41.2]).

To determine the source in variation of treatment type, we quantified the relative contribution of patient demographics, tumor characteristics, and referral patterns in our cohort (Figure). Only a small proportion of variation of treatment type is attributed to tumor characteristic, region, and year. The variation in the decision to pursue radical prostatectomy is significantly affected by patient demographics (40%), referral to other specialists (24%), and unexplained surgeon factors (23%). The variation in the decision to pursue radiotherapy is significantly affected by referral to other specialists (44%), as well as unexplained surgeon (20%) and patient factors (30%). The variation in the decision to pursue WW-AS is significantly affected by unexplained patient factors (58%) with less variation attributed to other consultants (14%), unexplained surgeon factors (12%), and patient demographics (12%).

Our study has several important findings. First, irrespective of prostate cancer risk stratification, most low-risk patients are being treated with radiotherapy. This is striking, given that our study was limited to men 65 years or older, who are at greater risk for death owing to competing risks. Furthermore, it has been previously demonstrated that aside from age, there is overuse of treatments in men with low-risk disease and clinically significant comorbidities.²¹ Our finding that increased use of radiotherapy among patients with indolent disease portends to a collaborative need for increased dissemination of prostate cancer treatment guidelines among our radiation oncology colleagues. In fact, the National Institutes of Health Consensus Conference emphasized the importance of further research into determinants associated with the offer and acceptance of active surveillance.²² Potential reasons for decreased acceptance of active surveillance by radiation oncologists may not necessarily include lack of knowledge but rather increased clinician anxiety owing to potentially underdiagnosed cancer from prostate biopsy pathologic findings, the perception that most men will ultimately undergo treatment, and concerns regarding the quality of life of men who elect active surveillance.^8,23,24 Recent data from Sweden suggest the feasibility of dissociating prostate cancer diagnosis and treatment as an modality to increase acceptance of active surveillance; however, there remains a critical need to discern ways to increase adherence and acceptance of active surveillance in low-risk prostate cancer.¹³ While we were unable to identify tumor biological factors as determinants for patients undergoing radiotherapy, prior studies have suggested that self-referral patterns may lead to increased use and costs of medical care.^25,26 Furthermore, consultation with radiation oncologists and regional variation have a significant impact on use of radiotherapy treatment options.²⁷ The magnitude of use of radiotherapy treatment options is significantly increased with integration of urology and radiation oncology practices into prostate cancer center groups.²⁷ In addition, companies that sell turnkey intensity-modulated radiation treatment (IMRT) programs to urology practices market the potential for increased IMRT revenue to replace lost earnings from ADT, for which reimbursement decreased sharply as part of the 2003 Medicare Modernization Act.^27,28 Further research into ongoing patient and clinician factors determining treatment decisions are needed to limit costs and overtreatment.

Second, we found that greater frequency of radical prostatectomy was predominantly owing to patient- and tumor-related factors. These findings corroborate prior findings—that surgery is less likely to be offered to patients with increased comorbidities or in the setting of high-risk disease.²⁹ However, many of these patients are offered radiotherapy, which may not be most appropriate, especially in the setting of low-risk disease and clinically significant comorbidities.^29,30 Jacobs et al³¹ recently identified overall treatment rates for low-risk disease to remain relatively stable; however, use of advanced treatment technologies, including IMRT and robotic surgery, have increased. In addition, rates of other forms of radiotherapy and open surgery decreased, suggesting that this newer technology has been rapidly adopted prior to demonstrating superiority to prior treatments.³¹ Further comparative effectiveness research separating the diagnosis and treatment of prostate cancer is needed to limit overtreatment of low-risk disease.^13,21

Finally, although greater frequency of WW-AS was inversely associated with prostate cancer risk, the differences were not as strikingly different. Unexplained patient and clinician factors accounted for greater than 70% of the variance in use of WW-AS. Prior data have shown that overtreatment of low-risk disease and undertreatment of high-risk disease are not explained by “measurable factors.”²⁹ While age, marital status, and education level have been associated with selection of active surveillance, the current study suggests that this accounts for only 11% of the explanation.¹³ Further comparative effectiveness research is needed to help guide evidence-based decision-making pathways.

While our findings are policy relevant, they must be interpreted in the context of the study design. First, SEER-Medicare is limited to men 65 years or older, and our results may not be generalizable to younger men diagnosed as having prostate cancer. Second, we were unable to identify patient-based determinants for treatments and further research into factors which drive decision-making processes that are needed. Third, while we attempted to identify predictors for undergoing WW-AS, there are undetermined patient and clinician factors that need to be discerned. Fourth, SEER-Medicare lacks the percentage of prostate cores involved with cancer and/or any specific information regarding the greatest percentage of core involved that has been previously demonstrated to increase upgrading on final pathologic findings at radical prostatectomy.^23,24 Therefore, it is unclear whether some of the radiotherapy offered to men with low-risk disease may have been driven by these other important clinical factors, bringing into question the degree to which radiotherapy was overused and WW-AS underused. Moreover, it is known that as men advance in age, high-grade disease becomes more prevalent and more likely to be missed on prostate needle biopsy owing to increasing biopsy sampling error as the prostate gland enlarges as a result of benign prostatic hyperplasia.³² Therefore, in this advanced-age cohort, underestimating the risk group may be more likely, and the potential for underdiagnosing clinically significant prostate cancer should be considered. Fifth, neither SEER nor Medicare explicitly identify those men who are being treated with WW-AS. We imputed this treatment category based on the lack of any treatment over a 2-year period. However, those who have no Medicare-paid treatment claims may not be undergoing WW-AS. It is also entirely possible that these patients refused or were not offered treatment, were lost to follow-up, or were treated but care was paid for by another insurer. While the definition of WW-AS may not be always correct, our utilization of a 2-year window is considerably more exclusive than other studies that may have limited to a shorter treatment window. Finally, the intensity of screening has a considerable impact on the likelihood of finding cancers that are low risk and suitable for active surveillance. At present, we cannot be certain of how intensely patients were screened and whether patients diagnosed were at higher risk. It has been previously demonstrated that when compared with younger patients (<75 years), older patients are sometimes less likely to be screened and actually at greater risk of aggressive disease and death when diagnosed.³³

There remains an increased use of treatments in men diagnosed as having prostate cancer and underuse of active surveillance in men with low-risk disease. There is an increased use of radiotherapy among all risk groups and in particular patients with indolent disease with limited correlation according to tumor biological characteristics and patient health. Further research into identifying determinants that drive decision-making recommendations for patients diagnosed with low-risk prostate cancer are needed. These findings must be balanced when considering health care reform initiatives to improve quality of care.

Corresponding Author: Jim C. Hu, MD, MPH, Department of Urology, David Geffen School of Medicine at UCLA, 924 Westwood Blvd, Ste 1000, Los Angeles, CA 90024 (jimhumd@gmail.com).

Accepted for Publication: December 3, 2014.

Published Online: February 19, 2015. doi:10.1001/jamaoncol.2014.192.

Author Contributions: Dr Hu had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: All authors.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Chamie.

Obtained funding: Hu.

Administrative, technical, or material support: Williams, Hu.

Study supervision: Williams, Hu.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work is supported by a Department of Defense Prostate Cancer Physician Training Award (W81XWH-08-1-0283) presented to Dr Hu and the NIH Loan Repayment Program (L30 CA154326) presented to Dr Chamie.

Role of the Funder/Sponsor: The design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication has been approved by the Department of Defense.

Disclaimer: The interpretation and reporting of these data are the sole responsibility of the authors.

Additional Contributions: We acknowledge the efforts of the Applied Research Program, NCI; the Office of Research, Development and Information, CMS; Information Management Services Inc; and the SEER Program tumor registries in the creation of the SEER-Medicare database. This study used the linked SEER-Medicare database.