[Skip to Navigation]
Sign In
Figure.  Changes in Hospital Referral Region–Level Use of Prostate Magnetic Resonance Imaging (MRI), Genomic Testing, and Observation for Newly Diagnosed Prostate Cancer Stratified by Quartile, July 2012-June 2014 to July 2017-June 2019
Changes in Hospital Referral Region–Level Use of Prostate Magnetic Resonance Imaging (MRI), Genomic Testing, and Observation for Newly Diagnosed Prostate Cancer Stratified by Quartile, July 2012-June 2014 to July 2017-June 2019
Table 1.  Characteristics of the Study Cohort at the Patient Level by Study Perioda
Characteristics of the Study Cohort at the Patient Level by Study Perioda
Table 2.  Patient-Level Mixed-Effects Logistic Regression of the Association Between Regional Use of Prostate MRI and Genomic Testing and Observation for Prostate Cancer Among Patients Diagnosed in the Early and Late Study Periods
Patient-Level Mixed-Effects Logistic Regression of the Association Between Regional Use of Prostate MRI and Genomic Testing and Observation for Prostate Cancer Among Patients Diagnosed in the Early and Late Study Periods
Table 3.  Characteristics of HRRs by Quartile of Change in Prostate MRI and Genomic Testing During the Study Period
Characteristics of HRRs by Quartile of Change in Prostate MRI and Genomic Testing During the Study Period
Table 4.  Results of Weighted Multivariable Linear Regression on the Associations of HRR-Level Adoption of Prostate MRI and Genomic Testing and Observation for Prostate Cancer Among 236 HRRs
Results of Weighted Multivariable Linear Regression on the Associations of HRR-Level Adoption of Prostate MRI and Genomic Testing and Observation for Prostate Cancer Among 236 HRRs
1.
Mahal  BA, Butler  S, Franco  I,  et al.  Use of active surveillance or watchful waiting for low-risk prostate cancer and management trends across risk groups in the United States, 2010-2015.   JAMA. 2019;321(7):704-706. doi:10.1001/jama.2018.19941 PubMedGoogle ScholarCrossref
2.
Hoffman  RM, Harlan  LC, Klabunde  CN,  et al.  Racial differences in initial treatment for clinically localized prostate cancer: results from the prostate cancer outcomes study.   J Gen Intern Med. 2003;18(10):845-853. doi:10.1046/j.1525-1497.2003.21105.x PubMedGoogle ScholarCrossref
3.
Leapman  MS, Carroll  PR.  New genetic markers for prostate cancer.   Urol Clin North Am. 2016;43(1):7-15. doi:10.1016/j.ucl.2015.08.002 PubMedGoogle ScholarCrossref
4.
Shore  ND, Kella  N, Moran  B,  et al.  Impact of the cell cycle progression test on physician and patient treatment selection for localized prostate cancer.   J Urol. 2016;195(3):612-618. doi:10.1016/j.juro.2015.09.072 PubMedGoogle ScholarCrossref
5.
Vickers  AJ.  Decision analysis for the evaluation of diagnostic tests, prediction models and molecular markers.   Am Stat. 2008;62(4):314-320. doi:10.1198/000313008X370302 PubMedGoogle ScholarCrossref
6.
Cooperberg  MR, Davicioni  E, Crisan  A, Jenkins  RB, Ghadessi  M, Karnes  RJ.  Combined value of validated clinical and genomic risk stratification tools for predicting prostate cancer mortality in a high-risk prostatectomy cohort.   Eur Urol. 2015;67(2):326-333. doi:10.1016/j.eururo.2014.05.039 PubMedGoogle ScholarCrossref
7.
Eggener  SE, Rumble  RB, Armstrong  AJ,  et al. Molecular biomarkers in localized prostate cancer: ASCO guideline. J Clin Oncol. 2020;38(13):1474-1494. doi:10.1200/JCO.19.02768
8.
Ahmed  HU, El-Shater Bosaily  A, Brown  LC,  et al; PROMIS study group.  Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study.   Lancet. 2017;389(10071):815-822. doi:10.1016/S0140-6736(16)32401-1 PubMedGoogle ScholarCrossref
9.
Kasivisvanathan  V, Rannikko  AS, Borghi  M,  et al; PRECISION Study Group Collaborators.  MRI-targeted or standard biopsy for prostate-cancer diagnosis.   N Engl J Med. 2018;378(19):1767-1777. doi:10.1056/NEJMoa1801993 PubMedGoogle ScholarCrossref
10.
Ahdoot  M, Wilbur  AR, Reese  SE,  et al.  MRI-targeted, systematic, and combined biopsy for prostate cancer diagnosis.   N Engl J Med. 2020;382(10):917-928. doi:10.1056/NEJMoa1910038 PubMedGoogle ScholarCrossref
11.
Leapman  MS, Nguyen  HG, Cooperberg  MR.  Clinical utility of biomarkers in localized prostate cancer.   Curr Oncol Rep. 2016;18(5):30. doi:10.1007/s11912-016-0513-1 PubMedGoogle ScholarCrossref
12.
Lynch  JA, Rothney  MP, Salup  RR,  et al.  Improving risk stratification among veterans diagnosed with prostate cancer: impact of the 17-gene prostate score assay.   Am J Manag Care. 2018;24(1)(suppl):S4-S10.PubMedGoogle Scholar
13.
Canfield  S, Kemeter  MJ, Febbo  PG, Hornberger  J.  Balancing confounding and generalizability using observational, real-world data: 17-gene genomic prostate score assay effect on active surveillance.   Rev Urol. 2018;20(2):69-76.PubMedGoogle Scholar
14.
Horrow  C, Pacyna  JE, Sutton  EJ, Sperry  BP, Breitkopf  CR, Sharp  RR.  Assessing optimism and pessimism about genomic medicine: development of a genomic orientation scale.   Clin Genet. 2019;95(6):704-712. doi:10.1111/cge.13535 PubMedGoogle ScholarCrossref
15.
Giordano  SH, Kuo  YF, Duan  Z, Hortobagyi  GN, Freeman  J, Goodwin  JS.  Limits of observational data in determining outcomes from cancer therapy.   Cancer. 2008;112(11):2456-2466. doi:10.1002/cncr.23452 PubMedGoogle ScholarCrossref
16.
Raffin  E, Onega  T, Bynum  J,  et al. Are there regional tendencies toward controversial screening practices? a study of prostate and breast cancer screening in a Medicare population. Cancer Epidemiol. 2017;50(pt A):68-75. doi:10.1016/j.canep.2017.07.015
17.
Dinan  MA, Mi  X, Reed  SD, Lyman  GH, Curtis  LH.  Association between use of the 21-Gene Recurrence Score Assay and receipt of chemotherapy among Medicare beneficiaries with early-stage breast cancer, 2005-2009.   JAMA Oncol. 2015;1(8):1098-1109. doi:10.1001/jamaoncol.2015.2722 PubMedGoogle ScholarCrossref
18.
Dartmouth Atlas Project. General FAQ. 2019. Accessed October 25, 2020. https://www.dartmouthatlas.org/faq
19.
Schroeck  FR, Kaufman  SR, Jacobs  BL,  et al.  Regional variation in quality of prostate cancer care.   J Urol. 2014;191(4):957-962. doi:10.1016/j.juro.2013.10.066 PubMedGoogle ScholarCrossref
20.
Goodney PR, Dzebisashvili N, Goodman DC, Bronner KK. Variation in the Care of Surgical Conditions. A Dartmouth Atlas of Healthcare Series. Trustees of Dartmouth College; 2014.
21.
Leapman  MS, Wang  R, Ma  S, Gross  CP, Ma  X.  Regional adoption of commercial gene expression testing for prostate cancer.   JAMA Oncol. 2021;7(1):52-58. doi:10.1001/jamaoncol.2020.6086PubMedGoogle Scholar
22.
Tyson  MD, Graves  AJ, O’Neil  B,  et al.  Urologist-level correlation in the use of observation for low- and high-risk prostate cancer.   JAMA Surg. 2017;152(1):27-34. doi:10.1001/jamasurg.2016.2907 PubMedGoogle ScholarCrossref
23.
Kim  SP, Gross  CP, Shah  ND,  et al.  Perceptions of barriers towards active surveillance for low-risk prostate cancer: results from a national survey of radiation oncologists and urologists.   Ann Surg Oncol. 2019;26(2):660-668. doi:10.1245/s10434-018-6863-1 PubMedGoogle ScholarCrossref
24.
Tran  GN, Leapman  MS, Nguyen  HG,  et al.  Magnetic resonance imaging-ultrasound fusion biopsy during prostate cancer active surveillance.   Eur Urol. 2017;72(2):275-281. doi:10.1016/j.eururo.2016.08.023 PubMedGoogle ScholarCrossref
25.
Washington  SL  III, Jeong  CW, Lonergan  PE,  et al.  Regional variation in active surveillance for low-risk prostate cancer in the US.   JAMA Netw Open. 2020;3(12):e2031349. doi:10.1001/jamanetworkopen.2020.31349 PubMedGoogle Scholar
26.
Makarov  DV, Soulos  PR, Gold  HT,  et al.  Regional-level correlations in inappropriate imaging rates for prostate and breast cancers: potential implications for the Choosing Wisely Campaign.   JAMA Oncol. 2015;1(2):185-194. doi:10.1001/jamaoncol.2015.37 PubMedGoogle ScholarCrossref
27.
Shi  Y, Pollack  CE, Soulos  PR,  et al.  Association between degrees of separation in physician networks and surgeons’ use of perioperative breast magnetic resonance imaging.   Med Care. 2019;57(6):460-467. doi:10.1097/MLR.0000000000001123 PubMedGoogle ScholarCrossref
28.
Tannenbaum  SS, Soulos  PR, Herrin  J,  et al.  Surgeon peer network characteristics and adoption of new imaging techniques in breast cancer: a study of perioperative MRI.   Cancer Med. 2018;7(12):5901-5909. doi:10.1002/cam4.1821 PubMedGoogle ScholarCrossref
29.
Presley  CJ, Tang  D, Soulos  PR,  et al.  Association of broad-based genomic sequencing with survival among patients with advanced non-small cell lung cancer in the community oncology setting.   JAMA. 2018;320(5):469-477. doi:10.1001/jama.2018.9824 PubMedGoogle ScholarCrossref
30.
Leapman  M, Presley  CJ, Zhu  W,  et al. Real-world practice patterns and impact of PD-L1 expression testing in patients with advanced non-small cell lung cancer. J Clin Oncol. 2019;37(15 suppl):9059. doi:10.1200/JCO.2019.37.15_suppl.9059
31.
Sanda  MG, Cadeddu  JA, Kirkby  E,  et al.  Clinically localized prostate cancer: AUA/ASTRO/SUO guideline. part II: recommended approaches and details of specific care options.   J Urol. 2018;199(4):990-997. doi:10.1016/j.juro.2018.01.002 PubMedGoogle ScholarCrossref
32.
Mohler  JL, Antonarakis  ES, Armstrong  AJ,  et al.  Prostate cancer, version 2.2019, NCCN Clinical Practice Guidelines in Oncology.   J Natl Compr Canc Netw. 2019;17(5):479-505. doi:10.6004/jnccn.2019.0023 PubMedGoogle ScholarCrossref
33.
Leapman  MS, Wang  R, Park  HS,  et al.  Association between prostate magnetic resonance imaging and observation for low-risk prostate cancer.   Urology. 2019;124:98-106. doi:10.1016/j.urology.2018.07.041 PubMedGoogle ScholarCrossref
34.
Cucchiara  V, Cooperberg  MR, Dall’Era  M,  et al.  Genomic markers in prostate cancer decision making.   Eur Urol. 2018;73(4):572-582. doi:10.1016/j.eururo.2017.10.036 PubMedGoogle ScholarCrossref
35.
Jairath  NK, Dal Pra  A, Vince  R  Jr,  et al.  A systematic review of the evidence for the Decipher Genomic Classifier in Prostate Cancer.   Eur Urol. 2021;79(3):374-383.doi:10.1016/j.eururo.2020.11.021PubMedGoogle Scholar
36.
Löppenberg  B, Friedlander  DF, Krasnova  A,  et al.  Variation in the use of active surveillance for low-risk prostate cancer.   Cancer. 2018;124(1):55-64. doi:10.1002/cncr.30983 PubMedGoogle ScholarCrossref
37.
Sonn  GA, Fan  RE, Ghanouni  P,  et al.  Prostate magnetic resonance imaging interpretation varies substantially across radiologists.   Eur Urol Focus. 2019;5(4):592-599. doi:10.1016/j.euf.2017.11.010 PubMedGoogle ScholarCrossref
Original Investigation
Genetics and Genomics
October 8, 2021

Adoption of New Risk Stratification Technologies Within US Hospital Referral Regions and Association With Prostate Cancer Management

Author Affiliations
  • 1Department of Urology, Yale School of Medicine, New Haven, Connecticut
  • 2Yale Cancer Outcomes, Public Policy, and Effectiveness Research Center, New Haven, Connecticut
  • 3Department of Chronic Disease Epidemiology, Yale School of Public Health, New Haven, Connecticut
  • 4Department of Therapeutic Radiology, Yale School of Medicine, New Haven, Connecticut
  • 5Department of Internal Medicine, Yale School of Medicine, New Haven, Connecticut
JAMA Netw Open. 2021;4(10):e2128646. doi:10.1001/jamanetworkopen.2021.28646
Key Points

Question  Is adoption of prostate magnetic resonance imaging (MRI) and genomic testing associated with increased use of observation rather than active treatment for prostate cancer?

Findings  In this cohort study of 65 530 commercially insured patients with prostate cancer, uptake of prostate MRI and genomic testing was associated with increased use of observation vs active treatment as initial management. Although prostate MRI, genomic testing, and observation increased overall, use was highly varied across hospital referral regions.

Meaning  The findings of this study suggest that adoption of technologies designed to improve decision-making may lead to perceivable reductions in overtreatment of prostate cancer

Abstract

Importance  The clinical decisions that arise from prostate magnetic resonance imaging (MRI) and genomic testing in patients with prostate cancer are not well understood.

Objective  To evaluate the association between regional uptake of prostate MRI and genomic testing and observation vs treatment for prostate cancer.

Design, Setting, and Participants  This retrospective cohort study of commercial insurance claims for prostate MRI and genomic testing included 65 530 patients 40 to 89 years of age newly diagnosed with prostate cancer from July 1, 2012, through June 30, 2019.

Exposures  Patient- and regional-level use of prostate MRI and genomic testing.

Main Outcomes and Measures  Observation vs definitive treatment for prostate cancer. Patient-level analyses examined the association between receipt of testing or residing in a hospital referral region (HRR) that adopted testing and observation. In regional-level analyses, the dependent variable was the change in the proportion of patients observed for prostate cancer at the HRR level between 2 periods: July 1, 2012, to June 30, 2014, and July 1, 2017, to June 20, 2019. The independent study variables included HRR-level changes in the proportion of men undergoing prostate MRI and genomic testing between these periods, and the models were adjusted for contextual factors associated with prostate cancer care and socioeconomic status.

Results  This study identified 65 530 patients, including 27 679 in the early period (mean [SD] age, 58.0 [5.9] years) and 37 851 in the late period (mean [SD] age, 59.0 [5.7] years). Use of prostate MRI increased significantly from 7.2% (95% CI, 6.9%-7.5%) to 16.7% (95% CI, 16.3%-17.1%) from the early to late period. Use of genomic testing increased significantly from 1.3% (95% CI, 1.1%-1.4%) to 12.7% (95% CI, 12.3%-13.0%) from the early to late period. Compared with the lowest, the highest HRR quartiles of prostate MRI and genomic testing uptake were associated with an adjusted 4.1% (SE, 1.1%; P < .001) and 2.5% (SE, 1.1%; P = .03) absolute increase in the proportion of patients receiving observation, respectively.

Conclusions and Relevance  In this cohort study, uptake of prostate MRI and genomic testing was associated with increased use of initial observation vs treatment for prostate cancer. Marked geographic variation supports the need for further patient-level research to optimize the dissemination and outcome of testing.

Introduction

There have been dramatic recent increases in the acceptance of active surveillance for prostate cancer, a period of close monitoring followed by timely treatment if needed.1 Nonetheless, nearly half of eligible patients still receive immediate active treatment, leading to preventable toxic effects and expense.1,2 Magnetic resonance imaging (MRI) of the prostate and several gene expression panels (genomic tests) have been developed and clinically implemented to improve decision-making for localized prostate cancer. Prostate MRI is a staging and diagnostic tool that enhances estimates of cancer grade and stage, whereas genomic tests provide prognostic estimates that are derived from observed associations with disease outcome.3-6 By clarifying the prognosis of screening-detected prostate cancers, it has been assumed that these tools will likely improve the precision of management and increasingly support observational management such as active surveillance.7

Despite rapid diffusion into clinical practice, no randomized clinical trials or observational studies have, to our knowledge, addressed the effects of these tests on the initial management of prostate cancer at the population level. Prostate MRI improves the diagnosis and staging of prostate cancer, and clinical guidelines increasingly endorse MRI to improve risk stratification when considering observational management.8-10 However, little is known about the decisions that arise from real-world use. Retrospective and registry-based studies suggest that genomic tests augment decision-making, steering more patients toward observation, but are limited by their study design and possible external bias from industry sponsorship.11-13 Determining the effectiveness of these technologies is critical given their expense and potential for decisional conflict associated with their results.14 The availability of these technologies also coincides with transformational shifts in the acceptance of observation management strategies, such as active surveillance.10 Thus, the extent to which prostate MRI and genomic testing have directly facilitated the adoption of observation rather than immediate treatment may be difficult to appreciate in analyses conducted at the individual patient level.

We assessed the association between the use of prostate MRI and genomic testing and the initial management of prostate cancer. Conventional methods for assessing causal inference using administrative claims are prone to bias because of unmeasured clinical and pathological factors and the subtleties of preference.15 As a result, even rigorous adjustment at the patient level may fail to counter the effects of unmeasured confounding and reverse causality. One method to overcome these constraints is to perform analyses across groups of patients who reside in different regions, taking advantage of well-known regional variation in health care.16,17 That is, if testing is associated with the decision to observe prostate cancer, we would expect overall increasing use of observation in areas where testing has increased. In this analysis, we use complementary approaches aimed at triangulating effect estimates conducted at the patient and regional levels.

Methods
Study Design and Cohort Selection

We performed a retrospective cohort study, with analysis conducted at the patient and hospital referral region (HRR) levels, using geographic boundaries described in the Dartmouth Atlas of Healthcare.18,19 The primary data source was deidentified claims data from Blue Cross Blue Shield Axis, a federation of 36 individual health insurance organizations and companies that provides care to approximately one-third of all Americans. Using a secure data portal, we accessed deidentified, longitudinally linked claims from Blue Cross Blue Shield Axis. We included patients 40 to 89 years of age who were newly diagnosed with prostate cancer from July 1, 2012, through June 30, 2019. To ensure that we captured newly diagnosed, incident prostate cancers, we required that the diagnosis coincided with a claim for prostate biopsy within 90 days and that patients had been enrolled in Blue Cross Blue Shield Axis for at least 12 months before the date of prostate cancer diagnosis.

The second source was publicly available data from the Dartmouth Atlas Project, a population-based, small area analysis of health care contextual factors, expenditures, services, and outcomes.20 We excluded HRRs that diagnosed fewer than 20 patients in any of the two 24-month period of interest (July 1, 2012, to June 30, 2014, and July 1, 2017, to June 30, 2019). The study schema is presented in eFigure 1 in the Supplement. This study was deemed non–human participant research by the Yale University Institutional Review Board, and the need for informed consent was therefore waived. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline (eAppendix in the Supplement).

Study Variables
Patient-Level Measures

We identified claims for prostate MRI and commercial genomic testing surrounding a patient’s initial diagnosis of prostate cancer. For prostate MRI, we included claims in the 6 months before and after diagnosis. For genomic tests, we used a previously described algorithm based on Current Procedural Terminology codes linked to laboratory-specific National Provider Identifier numbers.21 We focused on tests included in the 2020 National Comprehensive Cancer Network’s Prostate Cancer Clinical Guideline and assessed their use in the 6-month period after diagnosis (eAppendix in the Supplement).

We assessed a patient’s treatment status based on claims for definitive surgery, radiotherapy, androgen deprivation, or focal therapy in the 6-month period after diagnosis based on Current Procedural Terminology codes using previously applied algorithms.22 Patients without definitive treatment within 6 months of their prostate cancer diagnosis were regarded as undergoing observation.

Regional-Level Measures

We assigned patients to HRR boundaries based on billing zip code. We calculated the proportion of patients within each HRR who received genomic testing, prostate MRI, and initial definitive treatment or observation. We determined HRR-level proportions of genomic testing, prostate MRI, and treatment for prostate cancer in the following 2 periods: July 1, 2012, to June 30, 2014 (early), and July 1, 2017, to June 30, 2019 (late). These periods were selected to reflect an early period before widespread availability of these technologies and a later period characterized by wider interest and availability. The intervening July 1, 2014, to June 30, 2017, period was used as a washout. To account for differences in baseline use that existed between regions, we further calculated the absolute change in these measures (genomic testing, MRI, and observation) between the early and late periods and stratified the magnitude of HRR-level change by quartile. We compiled health care contextual measures associated with prostate cancer care that were described at the HRR level in 2014.20

Statistical Analysis

We compared sociodemographic characteristics of HRRs across quartiles of prostate MRI and genomic testing change between the early and late study periods. We assessed the association between change in prostate MRI and genomic testing at the HRR level using frequency tables and Spearman rank correlation coefficient. We used ordinary least squares regression to study the association between the quartile of HRR-level change in the use of prostate MRI and genomic testing and changes in the proportion of patients with prostate cancer undergoing observation. The regression models were weighted by the number of patients in each HRR in the early period. The dependent variable was the change in the proportion of patients observed for prostate cancer at the HRR level between 2 periods: July 1, 2012, to June 30, 2014, and July 1, 2017, to June 30, 2019. The independent study variables included the change in HRR-level use of genomic testing, prostate MRI, and contextual factors of prostate cancer care.

In a secondary analysis conducted at the patient level, we examined the association between prostate MRI and genomic testing and observation for prostate cancer using descriptive statistics. We constructed 2 series of multivariable logistic regression models to estimate the patient-level associations of receiving testing and residing in an HRR that adopted testing to varying degrees. The dependent variable in both models was observation vs treatment for prostate cancer. We first constructed a mixed-effects logistic regression model that incorporated an HRR random intercept to account for clustering of patients within regions with distinct patterns of prostate cancer care. The independent variables were the patient status of prostate MRI and genomic testing. The models were adjusted for age and HRR-level contextual factors (median household income and historical rates of prostate-specific antigen screening). In separate mixed-effects model restricted to the early and late periods, the independent variables included the quartile rank of prostate MRI and genomic testing change between the early and late periods. The model was similarly adjusted for age, diagnosis year, and HRR-level contextual factors. A 2-sided P < .05 was considered to be statistically significant. Statistical analyses were performed in SAS software, version 9.4 (SAS Institute Inc).

Results
Patient-Level Analysis

A total of 65 530 patients were identified, including 27 679 in the early period (mean [SD] age, 58.0 [5.9] years) and 37 851 in the late period (mean [SD] age, 59.0 [5.7] years). Eligible patients with prostate cancer were drawn from 236 HRRs that contributed more than 20 patients in each evaluation period. The proportion of patients who received prostate MRI increased from 7.2% (95% CI, 6.9%-7.5%) to 16.7% (95% CI, 16.3%-17.1%) from the early to late period; the proportion of patients who received genomic testing increased from 1.3% (95% CI, 1.1%-1.4%) in the early period to 12.7% (95% CI, 12.3%-13.0%) in late period (Figure). A total of 64 377 patients (98.2%) who received testing did so with a single modality; 1153 (3.0%; 95% CI, 2.9%-3.2%) received both prostate MRI and genomic testing in the late study period (Table 1).

The proportion of patients whose cases were managed with observation increased from 26.4% (95% CI, 25.9%-26.9%) in the early period to 35.4% (95% CI, 34.9%-35.9%) in the late period (Table 1). Among patients who underwent prostate MRI in the early period, 40.7% (95% CI, 38.6%-42.9%) were observed compared with 25.3% (95% CI 24.8%-25.8%) among those without prostate MRI. In the late period, observation increased to 52.2% (95% CI, 51.0%-53.4%) among patients who received prostate MRI compared with 32.1% (95% CI, 31.5%-32.6%) among patients who had not received an MRI. Similarly, observation was more common among patients who received genomic testing. In the late period, 59.2% (95% CI, 57.8%-60.6%) of patients who received genomic testing were observed compared with 32.0% (95% CI 31.5%-32.5%) of those without genomic testing.

At the patient level, prostate MRI (odds ratio [OR], 2.18; 95% CI, 2.06-2.30) and genomic testing (OR, 3.05; 95% CI, 2.86-3.25) were associated with increased odds of observation. Residence within an HRR with a higher median household income (OR per $10 000 increase, 1.06; 95% CI, 1.03-1.09) was associated with increased odds of observation (eTable 1 in the Supplement). Table 2 gives the results of mixed-effects logistic regression models examining the effects of HRR-level changes in prostate MRI and genomic testing. Residence in a region that subsequently adopted prostate MRI or genomic testing was not associated with increased odds of observation in the early period. Among patients diagnosed in the late period, residence in an HRR with the highest quartile increase in prostate MRI use (OR, 1.21; 95% CI, 1.09-1.34) and greater adoption of genomic testing (third vs first quartile: OR, 1.14; 95% CI, 1.02-1.27) were associated with increased odds of observation (Table 2).

Regional-Level Analysis

Contextual population demographic characteristics from the Dartmouth Atlas differed across levels of prostate MRI and genomic testing uptake. The HRRs with greater prostate MRI uptake had lower proportions of Black patients (mean [SD] prostate MRI change quartile, 12.6% [10.2%] in the highest quartile vs 17.3% [13.6%] in the lowest quartile; P = 0.03 for trend). The HRRs with greater genomic testing uptake had higher levels of college education (mean [SD], 26.9% [8.4%] in the highest quartile vs 25.0% [5.7%] in the lowest quartile; P <.001 for trend) and median income level (mean [SD], $54 286 [$14 248] in the highest quartile vs $51 778 [$9732] in the lowest quartile; P = 0.02 for trend). In addition, contextual prostate cancer care patterns differed across quartiles of HRR-level genomic testing uptake. Regions with greater adoption of genomic testing had historically less use of primary androgen deprivation therapy and less use of prostatectomy (Table 3).

Among the 236 HRRs that contributed more than 20 patients in each evaluation period, substantial regional variation in testing occurred. At the HRR level, use of prostate MRI in the late study period ranged from 0% to 62.7%, and the use of genomic testing ranged from 0% to 58.2%. Similarly, the initial management approach varied substantially, with observation ranging from 15.6% to 58.3% across HRRs. The mean (SD) change over time in HRR-level use was 9.0% (8.0%) for prostate MRI, 11.0% (8.2%) for genomic testing, and 9.0% (9.3%) for observation. Increases in HRR-level use of prostate MRI testing (ρ = 0.23, P < .001) and genomic testing (ρ = 0.18, P = .005) were positively correlated with change in observation (eFigure 2 in the Supplement). In multivariable linear regression, mean (SE) increases in the proportion of observed patients of 4.09% (1.06%) (P < .001) for the highest quartile (vs lowest) of prostate MRI and 2.47% (1.11%) (P = .03) for the highest quartile (vs lowest) of genomic testing uptake were observed. No significant association was seen between the mean (SE) second or third quartile rank of prostate MRI (mean [SE], 1.00 [1.03] in the second quartile [P < .001] vs 1.09 [1.07] in the third quartile [P < .001]) and genomic testing uptake (0.06 [1.05] in the second quartile [P = .96] vs 2.01 [1.14] in the third quartile [P = .08]). Increases in observation were also associated with baseline observation rate (estimate [SE], –0.71 [0.06]; P < .001), contextual median household income (estimate [SE], 1.02 [0.30]; P = .001), and prostate-specific antigen testing rates (estimate [SE], –0.07 [0.03]; P = .04) (Table 4).

Discussion

This cohort study found substantial increases in the use of prostate MRI and prognostic tissue–based genomic testing among commercially insured men with prostate cancer between July 2012 and June 2019. During the same period, the overall proportion of men who were initially observed, rather than treated, for prostate cancer also increased substantively. Despite increases in these practices overall, changes in prostate MRI, genomic testing, and observation varied widely across geographic regions. Examining the association between testing and initial management at the patient and regional levels, we found that uptake of prostate MRI and genomic testing was associated with increases in the use of observation. However, these associations were limited to HRRs with greater uptake of prostate MRI and genomic testing. In the context of an expanding array of new diagnostic and prognostic tools in cancer care, these findings can guide discussions about the clinical decisions associated with their use. Prominent regional variation in the use of risk stratification technologies also emphasizes an opportunity to address emerging gulfs in utilization, particularly as they may relate to overtreatment.

Distinct prostate cancer treatment practices among high-use regions may reflect several coexisting factors. The technologies themselves may exert direct effects as anticipated, leading to changes in management within an HRR that were perceivable only at high levels. Such an explanation would be congruent with analyses conducted at the patient level and prior evidence sources for prostate MRI and genomic testing, drawing the association between the use of refined risk assessment tools and greater use of observation for prostate cancer.23,24 It is also likely that regions with greater uptake of emerging diagnostic staging and risk assessment technologies shared conditions that favored greater increases in observation but did not directly affect the decision for treatment or observation, such as a region’s health care infrastructure and culture.25,26 However, we found that adoption of these technologies did not necessarily occur in HRRs with greater baseline inclinations for observation in the period before more widespread availability of these tools. Future research to better understand the relative contribution of these influences is warranted given the mixed track record of several other diagnostic technologies in oncology, including many that rapidly enter standard clinical care.27-30

Associations between regional expansion of new risk stratification technologies and observation for prostate cancer should be viewed in the context of global shifts in support of active surveillance. During the period in which prostate MRI and genomic testing became clinically integrated, practice guidelines also became explicit in their support for active surveillance as the preferred management strategy for most patients with low-risk prostate cancer, consolidating years of research and advocacy.31,32 Although adjunctive tests, such as prostate MRI and genomic testing, are not required to select patients for active surveillance, evidence supporting their role in refining risk and reducing diagnostic uncertainty has flourished.33 Indeed, we observed significant increases in the overall and HRR-level use of the technologies in line with prior estimates10,34,35 Associations between high HRR-level use and changes in management further imply a role for these technologies within common pathways for observation. However, the strength of the linear associations is weak, and the effect estimates were quite modest. Attenuation of the effect estimates when conducted at the regional vs patient level are also notable and can serve a methodologic caveat when performing effectiveness research using administrative claims. As a result, our results should be interpreted as an indication of alignment between regional increases in observation and tools aimed to facilitate its application, without causal attribution or directionality.

Limitations

This study has limitations. Patient-level analyses are limited by the absence of granular clinical information used in decision-making. Among this undifferentiated sample of men with prostate cancer, use of testing is expected to be confounded by lower-risk status, which may lead to both the exposure (use of testing) and outcome (observation). We suspect that concerns are more prominent for genomic testing, based on firmer indications for favorable-risk disease, but could also affect estimates for prostate MRI. By performing analyses at the regional level, this study is also subject to biases associated with ecologic inference. Therefore, effect estimates must be interpreted at the HRR level and cannot necessarily be extrapolated to individual patients. Similar directions of association in secondary analyses conducted at the patient level are reassuring but not conclusive. By aggregating information within groups of patients, we do not account for individual patient characteristics that might modify the association between the adoption of technologies and observation for prostate cancer. However, patient characteristics explain only a small amount of variation in the use of observational management, which is largely driven by other factors, such as a patient’s geographic region or treating physician and facility.36 As a result, our analysis is responsive to external determinants of initial management strategy, which might be modified by the use of prostate MRI or genomic testing. We also adjusted for historical contextual factors at the HRR level described in the Medicare population, which may not be directly applicable to the younger, privately insured cohort. In addition, because we used anonymized administrative claims, we were unable to incorporate the results of testing or the appropriateness of management decisions that arose from their use.37

Conclusion

Although use increased overall, uptake of prostate MRI and genomic testing surrounding prostate cancer diagnosis varied considerably across regional health care marketplaces measured at the HRR level. Similarly, use of initial observation vs active treatment for prostate cancer increased overall, but the magnitude of change varied significantly by HRR. The highest quartile of uptake of prostate MRI and genomic testing was associated with increases in the use of observation for the initial management of prostate cancer. Less rapid adoption of the technologies was not associated with increases in observation for prostate cancer. These findings suggest alignment between a region’s use of new risk stratification techniques occurring at the extremes and changes in the use of observational management for prostate cancer.

Back to top
Article Information

Accepted for Publication: August 5, 2021.

Published: October 8, 2021. doi:10.1001/jamanetworkopen.2021.28646

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Leapman MS et al. JAMA Network Open.

Corresponding Author: Michael S. Leapman, MD, Department of Urology, Yale University School of Medicine, 310 Cedar S, Brady Memorial Library 238c, New Haven, CT 06520 (michael.leapman@yale.edu).

Author Contributions: Dr Wang 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.

Concept and design: Leapman, Ma, Gross.

Acquisition, analysis, or interpretation of data: Leapman, Wang, Park, Yu, Sprenkle, Dinan, Ma.

Drafting of the manuscript: Leapman.

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

Statistical analysis: Leapman, Wang, Ma.

Obtained funding: Leapman.

Administrative, technical, or material support: Leapman.

Supervision: Gross.

Conflict of Interest Disclosures: Dr Park reported receiving personal fees from Rad Onc Questions LLC and Bristol Myers Squibb outside the submitted work. Dr Yu reported receiving speaking and consulting fees from Boston Scientific Speaking and serving on the advisory board for Galera Pharmaceuticals. Dr Dinan reported receiving grants from the National Cancer Institute outside the submitted work. Dr Ma reported receiving grants from Celgene/Bristol Myers Squibb and personal fees from Bristol Myers Squibb outside the submitted work. Dr Gross reported receiving grants from National Comprehensive Cancer Network (Pfizer/Astra-Zeneca), Johnson & Johnson, and Genentech and travel and speaking fees from Flatiron outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by grant K08CA204732 from the National Institutes of Health/National Cancer Institute (Dr Leapman).

Role of the Funder/Sponsor: The National Institutes of Health/National Cancer Institute had no role in 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.

Additional Contributions: The Blue Cross Blue Shield Association furnished the data for this analysis.

References
1.
Mahal  BA, Butler  S, Franco  I,  et al.  Use of active surveillance or watchful waiting for low-risk prostate cancer and management trends across risk groups in the United States, 2010-2015.   JAMA. 2019;321(7):704-706. doi:10.1001/jama.2018.19941 PubMedGoogle ScholarCrossref
2.
Hoffman  RM, Harlan  LC, Klabunde  CN,  et al.  Racial differences in initial treatment for clinically localized prostate cancer: results from the prostate cancer outcomes study.   J Gen Intern Med. 2003;18(10):845-853. doi:10.1046/j.1525-1497.2003.21105.x PubMedGoogle ScholarCrossref
3.
Leapman  MS, Carroll  PR.  New genetic markers for prostate cancer.   Urol Clin North Am. 2016;43(1):7-15. doi:10.1016/j.ucl.2015.08.002 PubMedGoogle ScholarCrossref
4.
Shore  ND, Kella  N, Moran  B,  et al.  Impact of the cell cycle progression test on physician and patient treatment selection for localized prostate cancer.   J Urol. 2016;195(3):612-618. doi:10.1016/j.juro.2015.09.072 PubMedGoogle ScholarCrossref
5.
Vickers  AJ.  Decision analysis for the evaluation of diagnostic tests, prediction models and molecular markers.   Am Stat. 2008;62(4):314-320. doi:10.1198/000313008X370302 PubMedGoogle ScholarCrossref
6.
Cooperberg  MR, Davicioni  E, Crisan  A, Jenkins  RB, Ghadessi  M, Karnes  RJ.  Combined value of validated clinical and genomic risk stratification tools for predicting prostate cancer mortality in a high-risk prostatectomy cohort.   Eur Urol. 2015;67(2):326-333. doi:10.1016/j.eururo.2014.05.039 PubMedGoogle ScholarCrossref
7.
Eggener  SE, Rumble  RB, Armstrong  AJ,  et al. Molecular biomarkers in localized prostate cancer: ASCO guideline. J Clin Oncol. 2020;38(13):1474-1494. doi:10.1200/JCO.19.02768
8.
Ahmed  HU, El-Shater Bosaily  A, Brown  LC,  et al; PROMIS study group.  Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study.   Lancet. 2017;389(10071):815-822. doi:10.1016/S0140-6736(16)32401-1 PubMedGoogle ScholarCrossref
9.
Kasivisvanathan  V, Rannikko  AS, Borghi  M,  et al; PRECISION Study Group Collaborators.  MRI-targeted or standard biopsy for prostate-cancer diagnosis.   N Engl J Med. 2018;378(19):1767-1777. doi:10.1056/NEJMoa1801993 PubMedGoogle ScholarCrossref
10.
Ahdoot  M, Wilbur  AR, Reese  SE,  et al.  MRI-targeted, systematic, and combined biopsy for prostate cancer diagnosis.   N Engl J Med. 2020;382(10):917-928. doi:10.1056/NEJMoa1910038 PubMedGoogle ScholarCrossref
11.
Leapman  MS, Nguyen  HG, Cooperberg  MR.  Clinical utility of biomarkers in localized prostate cancer.   Curr Oncol Rep. 2016;18(5):30. doi:10.1007/s11912-016-0513-1 PubMedGoogle ScholarCrossref
12.
Lynch  JA, Rothney  MP, Salup  RR,  et al.  Improving risk stratification among veterans diagnosed with prostate cancer: impact of the 17-gene prostate score assay.   Am J Manag Care. 2018;24(1)(suppl):S4-S10.PubMedGoogle Scholar
13.
Canfield  S, Kemeter  MJ, Febbo  PG, Hornberger  J.  Balancing confounding and generalizability using observational, real-world data: 17-gene genomic prostate score assay effect on active surveillance.   Rev Urol. 2018;20(2):69-76.PubMedGoogle Scholar
14.
Horrow  C, Pacyna  JE, Sutton  EJ, Sperry  BP, Breitkopf  CR, Sharp  RR.  Assessing optimism and pessimism about genomic medicine: development of a genomic orientation scale.   Clin Genet. 2019;95(6):704-712. doi:10.1111/cge.13535 PubMedGoogle ScholarCrossref
15.
Giordano  SH, Kuo  YF, Duan  Z, Hortobagyi  GN, Freeman  J, Goodwin  JS.  Limits of observational data in determining outcomes from cancer therapy.   Cancer. 2008;112(11):2456-2466. doi:10.1002/cncr.23452 PubMedGoogle ScholarCrossref
16.
Raffin  E, Onega  T, Bynum  J,  et al. Are there regional tendencies toward controversial screening practices? a study of prostate and breast cancer screening in a Medicare population. Cancer Epidemiol. 2017;50(pt A):68-75. doi:10.1016/j.canep.2017.07.015
17.
Dinan  MA, Mi  X, Reed  SD, Lyman  GH, Curtis  LH.  Association between use of the 21-Gene Recurrence Score Assay and receipt of chemotherapy among Medicare beneficiaries with early-stage breast cancer, 2005-2009.   JAMA Oncol. 2015;1(8):1098-1109. doi:10.1001/jamaoncol.2015.2722 PubMedGoogle ScholarCrossref
18.
Dartmouth Atlas Project. General FAQ. 2019. Accessed October 25, 2020. https://www.dartmouthatlas.org/faq
19.
Schroeck  FR, Kaufman  SR, Jacobs  BL,  et al.  Regional variation in quality of prostate cancer care.   J Urol. 2014;191(4):957-962. doi:10.1016/j.juro.2013.10.066 PubMedGoogle ScholarCrossref
20.
Goodney PR, Dzebisashvili N, Goodman DC, Bronner KK. Variation in the Care of Surgical Conditions. A Dartmouth Atlas of Healthcare Series. Trustees of Dartmouth College; 2014.
21.
Leapman  MS, Wang  R, Ma  S, Gross  CP, Ma  X.  Regional adoption of commercial gene expression testing for prostate cancer.   JAMA Oncol. 2021;7(1):52-58. doi:10.1001/jamaoncol.2020.6086PubMedGoogle Scholar
22.
Tyson  MD, Graves  AJ, O’Neil  B,  et al.  Urologist-level correlation in the use of observation for low- and high-risk prostate cancer.   JAMA Surg. 2017;152(1):27-34. doi:10.1001/jamasurg.2016.2907 PubMedGoogle ScholarCrossref
23.
Kim  SP, Gross  CP, Shah  ND,  et al.  Perceptions of barriers towards active surveillance for low-risk prostate cancer: results from a national survey of radiation oncologists and urologists.   Ann Surg Oncol. 2019;26(2):660-668. doi:10.1245/s10434-018-6863-1 PubMedGoogle ScholarCrossref
24.
Tran  GN, Leapman  MS, Nguyen  HG,  et al.  Magnetic resonance imaging-ultrasound fusion biopsy during prostate cancer active surveillance.   Eur Urol. 2017;72(2):275-281. doi:10.1016/j.eururo.2016.08.023 PubMedGoogle ScholarCrossref
25.
Washington  SL  III, Jeong  CW, Lonergan  PE,  et al.  Regional variation in active surveillance for low-risk prostate cancer in the US.   JAMA Netw Open. 2020;3(12):e2031349. doi:10.1001/jamanetworkopen.2020.31349 PubMedGoogle Scholar
26.
Makarov  DV, Soulos  PR, Gold  HT,  et al.  Regional-level correlations in inappropriate imaging rates for prostate and breast cancers: potential implications for the Choosing Wisely Campaign.   JAMA Oncol. 2015;1(2):185-194. doi:10.1001/jamaoncol.2015.37 PubMedGoogle ScholarCrossref
27.
Shi  Y, Pollack  CE, Soulos  PR,  et al.  Association between degrees of separation in physician networks and surgeons’ use of perioperative breast magnetic resonance imaging.   Med Care. 2019;57(6):460-467. doi:10.1097/MLR.0000000000001123 PubMedGoogle ScholarCrossref
28.
Tannenbaum  SS, Soulos  PR, Herrin  J,  et al.  Surgeon peer network characteristics and adoption of new imaging techniques in breast cancer: a study of perioperative MRI.   Cancer Med. 2018;7(12):5901-5909. doi:10.1002/cam4.1821 PubMedGoogle ScholarCrossref
29.
Presley  CJ, Tang  D, Soulos  PR,  et al.  Association of broad-based genomic sequencing with survival among patients with advanced non-small cell lung cancer in the community oncology setting.   JAMA. 2018;320(5):469-477. doi:10.1001/jama.2018.9824 PubMedGoogle ScholarCrossref
30.
Leapman  M, Presley  CJ, Zhu  W,  et al. Real-world practice patterns and impact of PD-L1 expression testing in patients with advanced non-small cell lung cancer. J Clin Oncol. 2019;37(15 suppl):9059. doi:10.1200/JCO.2019.37.15_suppl.9059
31.
Sanda  MG, Cadeddu  JA, Kirkby  E,  et al.  Clinically localized prostate cancer: AUA/ASTRO/SUO guideline. part II: recommended approaches and details of specific care options.   J Urol. 2018;199(4):990-997. doi:10.1016/j.juro.2018.01.002 PubMedGoogle ScholarCrossref
32.
Mohler  JL, Antonarakis  ES, Armstrong  AJ,  et al.  Prostate cancer, version 2.2019, NCCN Clinical Practice Guidelines in Oncology.   J Natl Compr Canc Netw. 2019;17(5):479-505. doi:10.6004/jnccn.2019.0023 PubMedGoogle ScholarCrossref
33.
Leapman  MS, Wang  R, Park  HS,  et al.  Association between prostate magnetic resonance imaging and observation for low-risk prostate cancer.   Urology. 2019;124:98-106. doi:10.1016/j.urology.2018.07.041 PubMedGoogle ScholarCrossref
34.
Cucchiara  V, Cooperberg  MR, Dall’Era  M,  et al.  Genomic markers in prostate cancer decision making.   Eur Urol. 2018;73(4):572-582. doi:10.1016/j.eururo.2017.10.036 PubMedGoogle ScholarCrossref
35.
Jairath  NK, Dal Pra  A, Vince  R  Jr,  et al.  A systematic review of the evidence for the Decipher Genomic Classifier in Prostate Cancer.   Eur Urol. 2021;79(3):374-383.doi:10.1016/j.eururo.2020.11.021PubMedGoogle Scholar
36.
Löppenberg  B, Friedlander  DF, Krasnova  A,  et al.  Variation in the use of active surveillance for low-risk prostate cancer.   Cancer. 2018;124(1):55-64. doi:10.1002/cncr.30983 PubMedGoogle ScholarCrossref
37.
Sonn  GA, Fan  RE, Ghanouni  P,  et al.  Prostate magnetic resonance imaging interpretation varies substantially across radiologists.   Eur Urol Focus. 2019;5(4):592-599. doi:10.1016/j.euf.2017.11.010 PubMedGoogle ScholarCrossref
×