TL;DR
This randomized controlled trial investigated how genomic testing (Oncotype DX GPS) influences urologists' treatment preferences for favorable-risk prostate cancer. The study found that providing GPS results significantly increased urologists' likelihood of recommending active treatment (prostatectomy/radiation) over active surveillance, particularly in a predominantly Black patient cohort.
What Problem Does This Paper Address?
Active Surveillance (AS) has become a standard approach for managing favorable-risk prostate cancer (PCa), aiming to avoid the side effects of immediate treatment. However, adoption rates vary, especially among Black men, who historically have lower AS rates. This disparity raises questions about potential biases, differences in disease aggressiveness, and the influence of various factors on treatment decisions. While genomic tests like Oncotype DX GPS are designed to refine risk stratification and potentially increase AS uptake, their impact on urologists' treatment preferences, especially in diverse populations and safety-net settings, remains underexplored. This paper addresses the gap in understanding how objective genomic data influences the physician's recommendation, which heavily weighs on patient choice.
Methods
The study employed a randomized controlled trial (RCT) design, the ENACT trial, involving 200 men with favorable-risk PCa across three Chicago hospitals. Participants were randomized into two arms: a control group receiving standard counseling and a GPS arm receiving standard counseling plus the Oncotype DX GPS assay results. Ten urologists participated, with their treatment preferences recorded at two visits: before and after the patient received counseling and, for the GPS arm, the genomic test results.
The primary outcome was the urologists' treatment preference (active surveillance/watchful waiting vs. active treatment like prostatectomy/radiation) at Visit 2. Statistical analysis involved intention-to-treat principles, using chi-square tests for baseline comparisons and multivariable binary logistic regression to identify factors independently associated with urologists' treatment preference. The models were built to account for initial preferences, patient demographics, and clinical factors.
Datasets
The study utilized data from the ENACT trial, which enrolled 200 men diagnosed with favorable-risk prostate cancer between 2016 and 2019. The dataset included clinical information (PSA, Gleason grade, NCCN risk level), patient demographics (age, race/ethnicity, education, health literacy), and urologist treatment preferences. The Oncotype DX Genomic Prostate Score (GPS) was the key genomic data generated for the intervention arm. No new benchmarks were introduced; the study focused on evaluating the impact of an existing diagnostic tool.
Key Insights and Ideas
- Genomic Data Shifts Physician Preference: The most striking insight is that providing objective genomic risk information (GPS) significantly nudged urologists away from recommending active surveillance and towards recommending active treatment. This suggests that while genomic tests are intended to help patients choose AS, they can also increase physician confidence in recommending more aggressive therapy when negative prognostic markers are present.
- Safety-Net Setting Matters: The study's focus on a safety-net setting with a predominantly Black patient population is crucial. It provides evidence on how genomic testing impacts treatment decisions in a context where AS disparities are a significant concern, and where factors like compliance and access to advanced imaging might influence AS recommendations.
- Initial Preference is a Strong Predictor, but GPS Can Override: While a urologist's initial preference was a very strong predictor of their final decision, the GPS assay still demonstrated a significant influence, capable of shifting preferences.
- Health Literacy as a Counterbalance: Interestingly, higher patient health literacy was associated with a decreased likelihood of urologists preferring active treatment, suggesting that informed patients might still steer towards AS, even with genomic data suggesting otherwise.
Key Results
- Doubled Preference for Active Treatment: In the GPS arm, urologists' preference for active treatment (prostatectomy/radiation) almost doubled at Visit 2, increasing from 15.4% to 29.3%, compared to a smaller increase in the control arm (11.5% to 14.1%). This difference was statistically significant (p=0.01).
- GPS as a Significant Predictor: In multivariable analysis, randomization to the GPS arm was associated with a 3.25 times higher odds of urologists preferring prostatectomy/radiation (95% CI: 1.25–9.38, p=0.02).
- Favorable Intermediate Risk and GPS Interaction: Patients with favorable intermediate NCCN risk levels were more likely to be recommended active treatment, and this effect was amplified by GPS testing.
- Health Literacy Impact: Higher patient health literacy was associated with a reduced likelihood of urologists preferring active treatment (OR: 0.78, 95% CI: 0.62–0.99, p=0.044).
Limitations and Future Work
The authors acknowledge several limitations:
- Small Sample Size of Urologists: Only ten urologists participated, which limits the generalizability of findings related to physician behavior.
- Multicenter but Homogeneous Setting: While multicentered, all sites were safety-net hospitals, potentially limiting generalizability to other healthcare settings.
- Patient Population Characteristics: The cohort had a significant proportion of very low-risk PCa patients (40%), which might limit the generalizability of findings to higher-risk populations.
- GPS Report Changes: The GPS report format evolved during the trial, which could have introduced variability.
- Unmeasured Factors: The authors suggest that unmeasured factors like perceived patient compliance or trust might influence provider decisions.
Future work could involve larger studies with more urologists, exploring these unmeasured factors, and assessing the long-term impact of GPS on both physician and patient treatment decisions, particularly in diverse populations. Further investigation into how genomic data interacts with patient-specific factors like trust and compliance is also warranted.
Summary updated