Guidelines from the National Comprehensive Cancer Network (NCCN) and American Urological Association (AUA) encourage active surveillance (AS), also referred to as watchful waiting, expectant management, or deferred treatment, as a preferred management option for men with low-risk prostate cancer (PCa).1,2 The guidelines reference peer-reviewed evidence on rates of mortality and metastatic disease, incidence of procedure-related adverse events, and patients’ quality of life from randomized trials, combined with a preponderance of supportive evidence from prospective, observational studies.
Several studies reveal trends and variations in AS use among men seen in real-world settings. In an analysis using the National Cancer Database, Löppenberg and associates3 assessed 115,208 men diagnosed with clinically low-risk PCa between 2010 and 2014 and found that 14,180 (12.3%) were initially managed by AS. Use of AS varied between 0% and 100%, with more than 58% of the variation associated with patient-level factors, such as age, race, stage, prostate-specific antigen (PSA) level, and comorbidities. Approximately 10% of the variation in AS use was associated with nonmedical factors, such as treatment facility type or facility volume. In another study, using the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) national registry, Cooperberg and Carroll4 reported that surveillance use for clinically low-risk disease (Cancer of the Prostate Risk Assessment [CAPRA] score range, 0-2) was <15% from 1990 through 2009, and increased in 2010 through 2013 to 40%. The Michigan Urological Surgery Improvement Collaborative (MUSIC), a consortium of 43 academic and community urology practices, reported AS use among 2643 men diagnosed with low-risk PCa between 2012 and 2016.5 Investigators reported a median AS use of 57% (range 30%-73%), an increase relative to prior studies that they attributed to implementing focused quality improvement initiatives.
Attempts to further increase guideline adherence have included healthcare system integration (eg, accountable care organizations), payment reforms, clinical pathways, clinical report cards, and staff training on shared-medical decision making.6-10 For example, Ehdaie and associates11 enlisted five physicians for a 1-hour training session on counselling patients on the implications of AS. They enrolled 1003 low-risk patients (761 before training, 242 after) and found a modest, nonsignificant increase in AS use (9.1%; 95% CI, −0.4 to 19.4%).
Individualized, objective risk stratification using molecular biomarkers and multiparametric prostate magnetic resonance imaging (MRI) have been assessed as potential solutions to further increase guideline-adherent AS use in men with clinically low-risk PCa.12,13 The Oncotype DX® Genomic Prostate ScoreTM (GPS; Genomic Health, Redwood City, CA) is a 17-gene molecular assay performed on formalin-fixed, paraffin-embedded diagnostic prostate biopsy samples. The expression of 12 cancer-related genes from 4 molecular pathways is normalized to the expression of 5 reference genes. A GPS result is indexed to the patient’s NCCN risk category to produce an estimated likelihood for high-grade disease and/or non–organ-confined disease. The GPS assay was clinically validated as an independent predictor of adverse pathology, biochemical recurrence after prostatectomy, metastases, and prostate-specific death in men with clinically low-risk disease.14-17 Three prior studies have demonstrated that GPS testing is associated with an absolute 21% to 29% difference in AS use compared with baseline (non-GPS testing).18-20 GPS testing is discussed in the NCCN guidelines as an option for improved risk stratification for men with early stage PCa.2 Prostate MRI and genomic testing with the GPS assay have shown weak correlation, suggesting that the two modalities represent independent predictors.13 Guidelines conclude MRI may be useful for integration into clinical staging, especially among men with Gleason score ≥ 7, but its role in low-risk PCa remains investigational.2,21
Our primary aim was to estimate AS use in a large US payer system in men with low-risk PCa who received GPS testing, MRI imaging, or no testing by either GPS testing or MRI. We also assessed the effect of GPS testing and/or MRI on AS use, changes by year of diagnosis, and level of AS use between 6- and 12-month follow-up.
A retrospective cohort study was conducted using the OptumTM Research Database (ORD; Eden Prairie, MN), which includes electronically stored medical records and administrative claims data linked to enrollment information and laboratory data from a large US health insurer offering both commercial and Medicare Advantage health plans. The ORD includes more than 34 million persons per year who are geographically diverse across the United States, with the greatest proportion in the Midwest and South US Census Bureau regions. The age and sex distribution of the enrollees is similar to that reported by the US Census Bureau for both the commercially insured and the Medicare managed care populations. The insurer provides coverage for physician, hospital, and prescription drug services. The use of the ORD as a data source for real-world, comparative effectiveness research has been described previously.22,23 De-identified records stripped of identifiable protected health information was extracted or accessed during the study in compliance with the US Health Insurance Portability and Accountability Act.
De-identified records were extracted for patients enrolled in the health plan from January 2013 to June 2016 who had ≥ 1 record of PCa using the International Classification of Diseases (ICD)-9 185 and ICD-10 C61 diagnostic codes. Inclusion criteria included age ≥ 18 years, AUA low-risk PCa (stage T1-T2a, PSA ≤10 ng/mL, Gleason score = 6), clinical activity for at least 12 months before and 6 months after diagnosis, and at least 1 PSA measurement within 12 months before or after diagnosis. PCa diagnosis date was defined as the patient’s earliest observed PCa diagnosis (no record of a diagnosis 12 months prior or during the observational window). Because the GPS assay became commercially available in June 2013, we restricted our analysis to records of patients enrolled in the database from June 2013 to June 2016. Records from GHI were submitted to the ORD for linkage to assure identification of the patients who had GPS testing. Patients with American Medical Association’s Current Procedural Terminology® (CPT) codes for genomic tests that were not the GPS assay were excluded because the small number of patients (n = 40) precluded meaningful comparisons with other groups.
Baseline patient characteristics included age at diagnosis, year of diagnosis, census geographic region, insurance status, and number of comorbidities. Genomic testing was recorded using the CPT codes 84999, 81479, and/or 81599. Other tests were based on the test’s name and measurement. For example, “test name = prostate-specific antigen, measurement type = Gleason.” PCa-related procedures were recorded for patients in the cohort at 6 and 12 months of follow-up, and included radical prostatectomy, radiation therapy, brachytherapy, cryotherapy, or hormone therapy (Table 1). A PCa-related procedure was reported if the patient had only one procedure during the 6- or 12-month follow-up period. A patient with more than one procedure over the interval was designated as having had multiple procedures. A patient with no recorded PCa-related procedures was designated as having undergone AS.
Summary statistics for all variables were reported from the ORD; number of patients and frequencies were reported for categoric variables, and means and standard deviations were reported for continuous variables. Subsequent data management and analyses were performed using STATA® 15/IC 15.1 for Windows (StataCorp LLC, College Station, TX).
A Pearson χ2 test was performed (2-sided, α = 0.05) to assess differences, unadjusted by covariates, in frequency of AS use by test utilization. Multivariate logistic regression of AS use was performed to control further for variations in baseline covariates; postestimation predictive margins are reported using confidence intervals computed by delta method.24,25 The primary analysis was to assess the difference in AS use by test utilization adjusted for baseline covariates. Secondary analyses included assessing differences stratified by the year patients were enrolled in the database and between the 6- and 12-month follow-up periods. We also assessed the distribution of procedures—prostatectomy, radiation therapy, and other (cryotherapy, brachytherapy, hormone therapy, or multiple therapies)—by GPS and/or MRI testing.
Screening identified 290,163 patients with one or more electronic health record (EHR) PCa records between June 2013 and June 2016 (Figure 1). Applying inclusion/exclusion criteria and linking records between ORD and GHI resulted in an analysis population of 8920 patients. Table 2 shows the distributions of covariates, by cohorts of patients who had 6 or 12 months of follow-up; 81% of men followed to 6 months had follow-up recorded up to 12 months. Among men with 6 months of follow-up, 375 had GPS testing (300 had only GPS testing and 75 had GPS testing plus MRI); 1099 patients had MRI only.
The frequency of AS use significantly varied among different combinations of GPS and MRI testing from a low of 42% for MRI only at 12 months of follow-up to 89% for GPS testing only (Figure 2; Pearson χ2 = 245, P <.001). Among patients followed for 6 months, AS utilization was 31.2% higher (95% CI, 22.6% to 39.7%; P <.001) in patients undergoing GPS testing only versus patients who did not receive GPS testing or MRI (Table 3, Figure 3). AS use was 5.1% (95% CI, −9.9% to 20.0%; P=NS) higher among patients undergoing MRI than among those who had neither GPS nor MRI. Combined GPS and MRI testing also correlated with significantly higher AS use, which was similar to the effect for patients with GPS testing only. AS use declined by 3.2% (95% CI, −4.7% to −1.7%, P <.001) between the 6- and 12-month follow-up periods.
Increasing age was significantly associated with AS use (Table 2; 7.8% increase per year; 95% CI, 3.4% to 12.3%; P <.001). No significant differences in AS use were detected based on region, insurance status, or baseline symptoms/diagnosis. Other than in 2013, AS use was stable between 2014 and 2016 in each of the groups identified (Figure 4). Prostatectomy was the most common procedure (22%) among men who did not receive GPS testing or MRI (Figure 5), compared with 5% use of radiation therapy and 4% for other procedures (cryotherapy, brachytherapy, hormone therapy, or multiple procedures). Prostatectomy was >70% less common with GPS testing relative to no GPS testing or MRI. In contrast, radiation therapy was more than twice as common for patients with MRI only compared with no GPS testing, no MRI, and all other testing groups (P <.001).
The use of AS was significantly higher among patients undergoing GPS testing compared with MRI or with no GPS or MRI testing. The difference in AS use with GPS testing versus no GPS or MRI testing was stable regardless of adjustment by cohort year or differences in mean covariate levels, or whether AS use was assessed at 6 or 12 months. The level of AS use declined between 6 and 12 months by 3.2%. MRI was the most common test ordered (13%) and was associated with lower AS use and higher use of radiation therapy, compared with GPS testing or with no GPS or MRI testing.
Our findings are consistent with previous studies of the effect of GPS testing on AS use (Table 4). Dall’Era and colleagues20 showed that men with low-risk disease (n = 124) who underwent GPS testing had a 24% higher rate of AS use compared with untested men with similar risk based on clinicopathologic factors alone. Albala and colleagues19 compared 180 men (100 identified retrospectively without testing and 80 followed prospectively who underwent testing) and found a 21% absolute increase in AS utilization for NCCN very low- and low-risk patients. In a prospective study of physician management change (recommendation pretesting versus actual treatment after testing), Eure and coworkers18 found that AS use was 22% higher than it would have been without the additional information from GPS testing. The baseline rate of AS in these three studies varied between 38% and 43%.
Our study is one of the largest investigations into AS utilization in an unselected (real-world) population and included a study design that permitted assessment and control for baseline characteristics and changes in patterns of care over time. We found a slightly greater association of GPS testing with AS use compared with previous studies (by an absolute difference of 6%-9%). We also found that the effect of GPS testing on AS use was stable over time, and that AS rates among those without GPS or MRI testing remained relatively unchanged from 2014 to 2016. Because of our contemporaneous study design (patients who underwent GPS testing versus those who did not receive testing in the same year and over multiple-year cohorts), the absolute effects seen herein are not likely a result of time bias.
We found that 17% of patients with low-risk PCa underwent some evaluation with GPS testing or MRI within 6 months of diagnosis. It was not possible based on the CPT code and other EHRs to ascertain why the genomic test or MRI were ordered. Citing retrospective cohort data, guidelines have suggested that MRI may be useful to detect extracapsular extensions to aid in decision making for nerve-sparing surgery.2 Although MRI has been considered an adjunct in risk assessment, the lower use of AS with MRI is consistent with the possibility that MRI is being used more for treatment staging rather than as a decision aid for AS use. The observation that approximately one in six patients underwent GPS testing and/or MRI may reveal an unmet need by physicians (and/or their patients) for more information beyond stage, Gleason score, and PSA value to aid in treatment decisions.
The use of AS varied widely with GPS and/or MRI testing and with age; in contrast, AS use did not vary based on geographic region, insurance status, or recorded information on presence of urologic symptoms or respiratory/chest diagnoses. AS use was approximately 43% for patients who underwent MRI, 60% for patients with no GPS or MRI testing, and 89% among patients with GPS testing only. Although published rates of AS use during the 2000s were low (<10%), recent studies have shown notably higher rates (40%-50%) with considerable variation between and among practices. The higher rates often are attributed to confidence in AS for patients with low-risk disease due to publication of the Prostate Intervention Versus Observation Trial (PIVOT) and the Prostate Testing for Cancer and Treatment (ProtecT) trial.26,27 It is notable that the AS rates with no GPS or MRI testing observed in our study are consistent with rates recently reported by the MUSIC groups.5 Despite the publication of large clinical trials, the level of AS use in general remains below full adherence with preferred management recommendations published in guidelines.
Several limitations are inherent to a study design with an EHR and claims database. First, some patients followed to 6 months had no follow-up data at 12 months, which may be due to patients seeking care elsewhere, an insurance change, or a small risk of mortality between 6 and 12 months. Second, because no unique CPT code for AS exists, we defined it as no definitive therapy during the observation period. This definition likely captured the majority of AS patients, but may have permitted some misclassification (some selected patients may not have truly undergone guideline-based AS). Finally, physicians who ordered the GPS test may have had different, unmeasurable, perspectives toward adoption of emerging medical technologies compared with those who did not order the test and thus may not be representative of the field at large.
Clinicians, guideline panels, and insurers want answers to relevant questions when considering policies for the adoption of novel technologies such as genomic testing. For example, how does the intervention affect clinical management, especially among diverse patient populations seen in different clinical settings? Questions about real-world adoption and effects are well answered by collecting data in a broad range of settings. These results add to those from three such studies18-20 of the Oncotype DX GPS assay and demonstrate consistent positive impact of the GPS test on AS use among different physicians and practices, geographic regions, and patients with different insurance coverage, over the >4 years that GPS has been commercially available.
AS is a guideline-“preferred” intervention for low-risk PCa due to the avoidance of side effects of definitive therapy, detrimental effects on quality of life, and unnecessary treatment of small, indolent cancer.2 Combined with the results of other studies from multiple settings,18-20 GPS testing is consistently associated with higher rates of AS compliance compared with MRI or no GPS/MRI testing at all.
The authors wish to acknowledge Bethann Hromatka, Michele Lee, Ruixiao Lu, and Donna Polizio for their copy editing, support, and overall review of the manuscript.
Funding for this research provided by Genomic Health, Inc. (Redwood City, CA).