If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Personalizing Localized Prostate Cancer: Validation of a Combined Clinical Cell-cycle Risk (CCR) Score Threshold for Prognosticating Benefit From Multimodality Therapy
Address for correspondence: Jonathan D. Tward, MD, PhD, Huntsman Cancer Institute, University of Utah, 1950 Circle of Hope Dr, Salt Lake City, UT 84112
A CCR score prognosticated metastasis better than risk group.
•
A score threshold determines who can be safely treated with a single modality.
•
The score can determine the absolute benefit of multimodality therapy.
Abstract
Introduction
The combined clinical cell-cycle risk (CCR) score is a validated model that combines the cell-cycle progression (CCP) score with the University of California San Francisco Cancer of the Prostate Risk Assessment (CAPRA) score. This score determines the risk of progressive disease for men with prostate cancer. Here, we further validate the prognostic ability of the CCR score and evaluate its ability to help determine which patients may safely forgo multimodality therapy.
Patients and Methods
We evaluated the CCR and a CCR-based multimodality threshold (2.112) in a retrospective, multi-institutional cohort of men with National Comprehensive Cancer Network intermediate- or high-risk localized disease (N = 718). These men received single or multimodality therapy (androgen deprivation with radiation [RT], or surgery with adjuvant RT or hormones).
Results
CCR score prognosticated metastasis for single-modality therapy, as a continuous variable (hazard ratio, 3.97; 95% confidence interval [CI], 2.61-6.06) and when dichotomized at the threshold (hazard ratio, 15.90; 95% CI, 5.43-46.52). The 10-year Kaplan-Meier risk for those receiving single-modality (RT or surgical) therapy with CCR scores below and above the threshold for single-modality treatment was 4.3% (95% CI, 1.0%-17.1%) and 20.4% (95% CI, 13.2%-30.7%), respectively. Using the threshold, 27% of men with newly diagnosed high-risk and 73% with unfavorable intermediate-risk disease could avoid multimodality therapy.
Conclusions
Patients with CCR scores below the multimodality threshold (2.112) may safely forgo multimodality therapy. The CCR score can be used as a decision aid to counsel men whether or not single-modality therapy would be sufficient for their intermediate- or high-risk prostate cancer.
The National Comprehensive Cancer Network (NCCN) recommends that men with unfavorable intermediate- or high-risk prostate cancer who choose radiation (RT) combine it with androgen deprivation therapy (ADT).
Clinically localized prostate cancer: ASCO Clinical Practice Guideline Endorsement of an American Urological Association/American Society for Radiation Oncology/Society of Urologic Oncology Guideline.
Although multimodality therapies can reduce the risk of recurrence and metastasis and increase overall survival, they increase the risk of morbidity and, frequently, mortality.
As such, accurate methods are needed to identify men with a higher risk of disease progression who certainly benefit from multimodality therapies to those with lower risk who may safely avoid it.
The Prolaris gene expression classifier test combines the University of California San Francisco Cancer of the Prostate Risk Assessment (CAPRA)
The University of California, San Francisco Cancer of the Prostate Risk Assessment score: a straightforward and reliable preoperative predictor of disease recurrence after radical prostatectomy.
score with a cell-cycle progression (CCP) score derived from a tumor RNA expression profile. This combined clinical cell-cycle risk (CCR) score provides significantly improved prognostic information over clinicopathologic factors and is a better predictor of progression than CAPRA or CCP scores alone.
Comparison of the prognostic utility of the cell cycle progression score for predicting clinical outcomes in African American and non-African American men with localized prostate cancer.
This study aimed to determine whether a “multimodality threshold” based on the CCR score could identify patients choosing RT who could safely avoid ADT, determine the likelihood of needing postoperative RT for those choosing surgery, and if the CCR model could be used to calculate a personalized absolute risk reduction of metastasis for single- versus multimodality therapies.
Materials and Methods
CCP Testing
CCP testing was performed at Myriad Genetics, Inc, as described previously.
Transatlantic Prostate Group. Prognostic value of an RNA expression signature derived from cell cycle proliferation genes in patients with prostate cancer: a retrospective study.
Development of the CCR Score Multimodality Threshold
The CCR score multimodality threshold was developed using a large commercial dataset comprising individuals with NCCN unfavorable intermediate- and high-risk prostate cancer and known CCR scores (N = 15,669). All individuals provided informed consent for clinical testing, and data were de-identified and aggregated for analysis. Among 15,669 individuals, 4615 (29.5%) had prostate cancer classified as NCCN high-risk. The multimodality threshold score was pre-specified at a CCR score of 2.112 so that the proportion of individuals with scores above the threshold would not exceed 29.5%, mirroring the ratio of individuals with NCCN high-risk disease.
Study Cohort
The study cohort was derived from datasets from 5 tertiary referral centers (Durham Veterans Affairs Medical Center, Durham, NC [treated 1994-2005]
Comparison of the prognostic utility of the cell cycle progression score for predicting clinical outcomes in African American and non-African American men with localized prostate cancer.
; see Supplemental Figure 1 [in the online version]). Detailed characteristics of all but the Huntsman Cancer Institute dataset have been published previously.
Comparison of the prognostic utility of the cell cycle progression score for predicting clinical outcomes in African American and non-African American men with localized prostate cancer.
Treatment modality was based on physician-patient preference within the usual context of routine clinical practices at each center. The CCP score was used to inform treatment choice only for the Huntsman Cancer Institute dataset; all others were retrospective datasets with CCP testing performed after treatment was delivered. In a sensitivity analysis excluding patients from the Huntsman Cancer Institute, no significant results changed substantially (see Supplemental Table 1 in the online version). We obtained institutional review board approval with a waiver of informed consent to use de-identified data for each dataset.
A dataset query identified men with NCCN intermediate- or high-risk prostate cancer, known outcomes, and CCR scores. De-identified records were combined into a limited-use, pooled cohort (N = 718). Men were classified as having single-modality therapy if they received RT alone (external beam RT [EBRT], brachytherapy, or both), or surgery without planned adjuvant RT or ADT. Men who had surgery and then progressed to receive salvage RT or ADT remained in the single-modality group based on intention-to-treat. Men who received combination ADT and RT, or who after surgery received adjuvant RT, ADT, or the combination, were classified in the multimodality therapy group. Patient ADT duration in this study was not controlled or collected.
Statistical Analysis
This study followed the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) criteria for evaluating prognostic biomarkers.
The pre-specified primary endpoint, time to metastasis, was defined as time to the first occurrence of any nonregional nodal, bone, or visceral progression after initial therapy. For the primary endpoint, patients lost to follow-up for any reason before 10 years were censored at the time of the last follow-up; patients with metastasis or last follow-up after 10 years were censored at 10 years. The secondary endpoint, time to progression, was the earliest time to biochemical recurrence (BCR), metastasis, or salvage RT use. For patients undergoing RT, BCR was defined using Phoenix criteria,
Defining biochemical failure following radiotherapy with or without hormonal therapy in men with clinically localized prostate cancer: recommendations of the RTOG-ASTRO Phoenix Consensus Conference.
and for surgical patients as prostate-specific antigen ≥ 0.2 ng/mL. Eleven patients missing dates of salvage RT were censored at the time of the last follow-up. Those patients lost to follow-up for any reason before 5 years were censored at the last follow-up; patients with progression events or last follow-up after 5 years were censored at 5 years.
The ability of the NCCN risk group, CAPRA score, CCP score, CCR score, and multimodality therapy status to predict metastasis was evaluated using univariable and multivariable Cox proportional-hazards regression analyses.
The multimodality threshold was validated in patients receiving single-modality therapy. The ability to predict metastasis-free survival was evaluated using Harrell’s concordance statistic (C-index).
Cumulative incidence of metastasis over time was evaluated using the Kaplan-Meier method, with confidence intervals (CIs) based on the log-log transformation.
We performed classification and regression tree (CART) analysis to evaluate prognostic covariates that might alter the benefit of multimodality therapy above or below the threshold (see Supplemental Methods in the online version).
The absolute risk reduction of metastasis for RT + ADT versus RT alone at any CCR score was derived from a bivariate Cox proportional hazards model fit on patients receiving primary RT using treatment modality and CCR score as a continuous variable.
Results
Baseline characteristics for the full study cohort (N = 718) are shown in Table 1. There were 232 patients treated initially with RT and 486 treated with surgery. There were 41 (8.4%) patients treated with multimodality therapy in the surgery cohort and 116 (50.0%) in the RT cohort.
Table 1Characteristics of the Study Population for the Full Study Cohort and Separated by Treatment Type
RT (N = 232)
Surgery (N = 486)
All Patients (N = 718)
N (%) or Median (IQR)
N (%) or Median (IQR)
N (%) or Median (IQR)
CCR score
2.11 (1.29-2.91)
1.55 (1.05-2.12)
1.65 (1.11-2.34)
CCP score
0.4 (−0.3 to 1)
0.1 (−0.4 to 0.6)
0.1 (−0.4 to 0.8)
NCCN risk group
Favorable intermediate
38 (16.4)
133 (27.4)
171 (23.8)
Unfavorable intermediate
105 (45.3)
249 (51.2)
354 (49.3)
High
89 (38.4)
104 (21.4)
193 (26.9)
CAPRA score
5 (3-6)
4 (3-5)
4 (3-5)
Low (0-2)
22 (9.5)
72 (14.8)
94 (13.1)
Intermediate (3-5)
123 (53)
348 (71.6)
471 (65.6)
High (6-10)
87 (37.5)
66 (13.6)
153 (21.3)
Clinical T-stage
T1
122 (52.6)
303 (62.3)
425 (59.2)
T2
90 (38.8)
177 (36.4)
267 (37.2)
T3
20 (8.6)
6 (1.2)
26 (3.6)
Gleason score
<7
26 (11.2)
79 (16.3)
105 (14.6)
7
151 (65.1)
326 (67.1)
477 (66.4)
>7
55 (23.7)
81 (16.7)
136 (18.9)
PSA, ng/mL
8.2 (5.7-14.5)
6.8 (4.9-11.3)
7.3 (5.2-11.8)
Treatment modality
Multimodal
116 (50)
41 (8.4)
157 (21.9)
Single mode
116 (50)
445 (91.6)
561 (78.1)
Ancestry
African American
74 (31.9)
109 (22.4)
183 (25.5)
Caucasian
157 (67.7)
373 (76.7)
530 (73.8)
Other
1 (0.4)
4 (0.8)
5 (0.7)
Abbreviations: CAPRA = University of California San Francisco Cancer of the Prostate Risk Assessment; CCP = cell-cycle progression; CCR = clinical cell-cycle risk; IQR = interquartile range; NCCN = National Comprehensive Cancer Network; PSA = prostate-specific antigen; RT = radiation therapy.
The median follow-up time was 5.1 years; there were 37 events within 10 years. As a continuous variable, the CCR score was a significant prognostic indicator, with each 1-unit increase in CCR score corresponding to a nearly 4-fold increased risk of metastasis (hazard ratio [HR], 3.75; 95% CI, 2.71-5.20; P = 1.6×10−16). When stratified by surgery and RT, the CCR score remained significant, with the HR essentially unchanged (HR, 3.58; 95% CI, 2.52-5.08; P = 7.0×10−14). The univariate C-indices for the ability of the CCR, CAPRA, or CCP scores, or the NCCN risk group to predict metastases in the full cohort were 0.87, 0.80, 0.79, and 0.76, respectively.
We evaluated the performance of the CCR score separately in the surgery and RT cohorts using bivariate prognostic models after adjusting for CAPRA score, NCCN risk group, or treatment modality (Table 2). After adjusting for CAPRA score, CCR score remained a significant prognostic indicator in the surgical (HR, 4.08; 95% CI, 1.90-8.78; P = 5.7×10−4) and RT (HR, 4.30; 95% CI, 2.23-8.30; P = 6.2×10−5) cohorts, indicating that the molecular component of the CCR score adds important prognostic information to clinical variables regardless of treatment. We obtained similar results after adjusting for NCCN risk groups.
Table 2Performance of CCR and Bivariable Cox Regressions in Predicting Metastasis for Radiation and Surgery Cohorts
NCCN models in RT patients run with Firth penalty.
CCR score
4.03 (2.20-7.37)
2.8 × 10−7
2.91 (1.50-5.64)
1.2 × 10−3
NCCN risk group
Favorable intermediate
reference
.23
reference
.85
Unfavorable intermediate
0.81 (0.03-18.95)
1.70 (0.19-14.91)
High
0.56 (0.02-14.94)
1.94 (0.17-21.59)
Univariate and bivariate Cox regression analyses are shown.
Abbreviations: CAPRA = University of California San Francisco Cancer of the Prostate Risk Assessment; CCP = cell cycle progression; CCR = clinical cell-cycle risk; CI = confidence interval; HR = hazard ratio; NCCN = National Comprehensive Cancer Network; RT = radiation therapy.
a NCCN models in RT patients run with Firth penalty.
In those receiving single-modality therapy specifically, the CCR score prognosticated metastasis for patients receiving RT (HR, 5.52; 95% CI, 2.22-13.68; P = 3.4×10−5), those receiving surgery (HR, 3.67; 95% CI, 2.22-6.09; P = 5.8×10−7), or the surgery and RT cohorts combined (HR, 3.97; 95% CI, 2.61-6.06; P = 1.6×10−10). The treatment type effect was not significant in a bivariate model accounting for CCR score (P = .10), and neither was the interaction between CCR score and treatment type (P = .65). Therefore, the risk of metastasis for any individual selecting a single modality therapy can be prognosticated for any CCR score (see Supplemental Figure 2 in the online version).
Validation of the CCR Multimodality Risk Threshold
At the predefined threshold (CCR score = 2.112), the 10-year risk of metastasis was 8.1% (95% CI, 4.0%-15.8%). When dichotomized at the threshold, men treated with single-modality therapy in the above-threshold group had a nearly 16-fold predicted risk of developing metastasis than those in the below-threshold group (HR, 15.90; 95% CI, 5.43-46.52; P = 8.1×10−10]. The cumulative incidence of metastasis at 10 years in the below- and above-threshold groups for those receiving single-modality therapy was 4.3% (95% CI, 1.0%-17.1%) and 20.4% (95% CI, 13.2%-30.7%), respectively (Table 3). For men who elected surgery specifically, the cumulative incidence of progression in the below- and above-threshold groups was 23.2% and 52.2%, respectively (Table 3 and Figure 1B). This implies that men above the threshold are at least twice as likely to be recommended salvage RT.
Table 3Point Estimates for Metastasis and Progression
Risk of Metastasis
Risk of Disease Progression
Below Threshold
Above Threshold
Below Threshold
Above Threshold
n
5-year Risk, %
10-year Risk, %
n
5-year Risk, %
10-year Risk, %
n
5-year Risk, %
n
5-year Risk, %
Full cohort
479
0.6
5.2
239
13.0
17.8
479
21.9
239
35.7
Single modality
RT
92
2.2
–
24
22.7
–
92
15.6
24
39.7
Surgery
339
0.4
4.1
106
12.9
17.1
339
23.2
106
52.2
Combined
431
0.7
4.3
130
14.8
20.4
431
22.0
130
50.0
Multimodality
RT
25
0.0
–
91
13.1
–
25
7.1
91
17.9
Surgery
23
0.0
8.3
18
0.0
0.0
23
35.1
18
22.6
Combined
48
0.0
7.7
109
10.6
14.1
48
21.8
109
18.6
Threshold of CCR score = 2.112.
Dashes indicate that there were no more evaluable patients at risk at that time point.
Figure 1Kaplan-Meier 10-Year Risk Estimates for Metastasis (A) or Progression (B) in Men Who had Radical Prostatectomy (Surgery), Stratified by Clinical Cell-Cycle Risk Scores Above or Below the Multimodality Threshold, and by Single-Modality or Multimodality Therapy
Kaplan-Meier survival estimates for surgery (Figure 1) and RT (Figure 2) patients illustrate the prognostic value of CCR scores below and above the threshold, stratified by single- or multimodality therapy for metastasis (Figures 1A and 2A) and disease progression (Figures 1B and 2B). In all cases, men with CCR scores above the threshold who received single-modality therapy had the worst outcomes. The cumulative incidence at 5 and 10 years for metastasis and at 5 years for progression are further summarized in Table 3.
Figure 2Kaplan-Meier 10-Year Risk Estimates for Metastasis (A) or Progression (B) in Men Who had Radiation Therapy, Stratified by Clinical Cell-Cycle Risk Scores Above or Below the Multimodality Threshold, and by Single-Modality or Multimodality Therapy
CART analysis of the below-threshold group revealed that CCR score, multimodality therapy, NCCN risk group, and treatment type did not identify subgroups at increased risk of metastases (see Supplemental Figure 3 in the online version). In the above-threshold group, multimodality therapy reduced the predicted risk of metastasis (relative HR, 0.38 for multimodality vs. 1.03 for single-modality; P = .04), but only up to CCR score 4.1 (relative HR, 7.72; P < .01). This suggests that men with an extremely high risk of metastasis may need further treatment intensification. Treatment type and NCCN risk group did not significantly alter the risk of metastasis of the above-threshold group.
If the CCR threshold was used to determine who might safely avoid initial multimodality therapy, then 27% of men with NCCN high-risk disease and 73% of those with NCCN unfavorable intermediate-risk disease might be spared. Only 2% of men with NCCN favorable intermediate-risk disease were in the above-threshold group.
Clinical Utility of the Multimodality Threshold in RT Patients
The risk of metastasis for discrete CCR scores can determine an individual’s risk of developing metastasis within 10 years after therapy. We focused these analyses on those who selected RT, where the decision to use multimodality therapy is based on biopsy tissue and clinical details exclusively. When accounting for CCR score, multimodality therapy reduced the risk of metastasis in the RT-treated cohort (HR, 0.25; 95% CI, 0.09-0.69; P = .013) (Figure 3A). This absolute risk reduction was used to determine an NNT for RT + ADT relative to RT alone to prevent metastasis, for any CCR score. CCR scores from 0 to 4 correspond to an NNT range of approximately 500 to 3 (Figure 3B). The reduction in Kaplan-Meier estimated risk for metastasis based on adding ADT to RT for the population of men below the threshold was 2.2%, whereas for those above the threshold it was 20.3%. If all the above-threshold men were treated with multimodality therapy, the NNT to prevent 1 metastasis would be 5, and for those below the threshold it would be 46. By comparison, if all men with NCCN high-risk disease were treated with multimodality therapy, the NNT would be 10. Consistent with the CART analysis, men with CCR scores above approximately 4.5 appear to derive less benefit from added ADT, indicating that they may require additional treatment intensification.
Figure 3A, Risk Prediction for Metastasis at 10 years in RT Cohort, Stratified by Single or Multimodality Therapy and by CCR Score. B, Absolute Risk Reduction for Adding ADT to RT and Number Needed to Treat as a Function of CCR Score
These tests report an individual’s risk of metastasis based on studies that combine multimodality and single-modality treatments. This is the first analysis to show that a gene expression classifier can prognosticate the benefit of using single- or multimodality therapy, based on the biopsy, before the initial treatment decision.
Prognostic biomarkers provide information that is independent of the treatment rendered.
We have demonstrated that CCR is a highly accurate and precise biomarker for prognosticating metastasis regardless of a man’s treatment path. It is therefore unnecessary, and perhaps unethical, to perform another randomized trial to test if intermediate- or high-risk men with low CCR scores benefit from adding ADT to RT. The relative benefit of adding ADT to RT has been proven several times over by large, well-powered randomized trials designed to show even small differences in outcomes.
We do, however, take the position that the trial-proven benefit of adding ADT to RT, which is based on a relative risk reduction, becomes marginal when translated into an absolute risk reduction below the CCR threshold. For men with CCR scores ≤ 2.112, single-modality therapy will likely be judged as sufficient by most patients, as these men only had a 4% risk overall of developing metastasis at 10 years, and the absolute reduction of metastatic risk for adding ADT to RT was 2.2%. At this low absolute as opposed to relative benefit, 90% or more of expert radiation oncologists who specialize in treating genitourinary cancers would omit its use for themselves.
For men with CCR scores > 2.112, multimodality therapy translates into a more clinically meaningful absolute risk reduction, which can be discretely estimated at the individual level.
The CCR risk model was a better prognosticator of metastasis than were CAPRA score, NCCN risk group, or CCP score models when used independently. CCR-based risk stratification also reduced the NNT 2-fold compared with using the NCCN high-risk definition to identify multimodality therapy candidates. For CAPRA and NCCN risk models alike, the additional information captured by the molecular CCP score significantly improved prognostic accuracy as measured by Cox proportional hazards analysis. In a meta-analysis of randomized trials, metastasis-free survival proved a robust surrogate marker of overall survival among men treated for localized prostate cancer.
Therefore, it serves as one of the most important outcomes men and their advisors should consider when evaluating treatments.
Prospectively-randomized trials with EBRT have established that the addition of ADT can reduce the risk of BCR, metastasis, and all-cause mortality; however, these clinically-significant benefits were restricted primarily to those with NCCN intermediate- or high-risk disease.
Long-term results with immediate androgen suppression and external irradiation in patients with locally advanced prostate cancer (an EORTC study): a phase III randomised trial.
5191: Timing of radiotherapy (RT) after radical prostatectomy (RP): first results from the RADICALS RT randomised controlled trial (RCT) [NCT00541047].
Eastern Cooperative Oncology Group study EST 3886 Immediate versus deferred androgen deprivation treatment in patients with node-positive prostate cancer after radical prostatectomy and pelvic lymphadenectomy.
have used previously described clinical trials to support ADT + RT recommendations in appropriately risk-stratified men or for considering adjuvant or early salvage RT after surgery when pathologic circumstances suggest potential benefit. The magnitude of benefit for multimodality therapy is seldom evident in the guideline sections to which practitioners and patients routinely refer, limiting shared decision-making.
Multimodality therapy comes with high costs, including increased medical and financial toxicity, a decrease in short-term quality of life, and the risk of more significant side effects than single-modality therapies. How patients weigh clinical endpoints is highly individual, not predictable, and might be offset by concerns about other costs and care goals.
A multimodality threshold score can inform physician-patient discussions when determining an appropriate initial treatment choice. Men with CCR scores below the threshold might be spared therapies unlikely to improve oncologic outcomes substantially but might bring treatment-related morbidities. Conversely, men with CCR scores above the threshold more clearly benefit from upfront multimodality therapy when choosing RT, potentially improving oncologic outcomes. For those choosing surgery, above-threshold CCR scores can help identify those most likely to benefit from adjuvant or early salvage RT following initial surgical intervention. Therefore, a high CCR score can be a useful tool for physicians and patients weighing the pros and cons of choosing surgery plus a high probability of receiving postoperative RT, versus RT combined with ADT upfront.
Limitations of the current study include the fact that most of the patients were identified retrospectively. As such, some specific details regarding treatments (RT, ADT, surgical) are unknown. Study strengths include the ethnic diversity, the multi-institutional and international setting, and sizeable statistical power. Randomized trials of RT ± ADT in various risk-group populations typically show HRs in the range of 0.50 to 0.65, whereas we have shown a more dramatic effect of ADT + RT (HR, 0.25) when accounting for CCR score.
Because this study was not randomized, this finding should be considered carefully, but it supports the idea that the CCR model can improve the selection of patients who benefit from omitting or adding ADT to RT because the true magnitude of the effect of adding ADT is likely smaller than reported here.
Conclusion
Men with either NCCN unfavorable intermediate-risk or high-risk prostate cancer still may elect to forgo multimodal therapy once they understand the personalized absolute-risk reduction and NNT to prevent metastasis. Likewise, men with NCCN favorable intermediate-risk disease, given the same personalized risk information, may elect multimodality therapy. With the CCR score multimodality threshold as a decision aid, patients and providers can evaluate the personalized benefits of different treatments and identify if it may be appropriate to forgo multimodal therapy and avoid unnecessary morbidities. This should engender higher quality decisions that balance patient autonomy with clinician expertise in patient-centered models of shared decision-making.
The CCR score improves prognostic information for men with prostate cancer compared with other risk stratification systems using clinical information alone. Men with unfavorable intermediate- or high-risk prostate cancer are recommended multimodality therapies if pursuing RT, although the magnitude of benefit over single-modality therapy may be clinically insignificant in some persons.
•
We show that a genome expression classifier combined with clinical information can develop a CCR score that accurately prognosticates metastasis in patients receiving single- or multimodality therapy better than existing risk classifiers like those of the NCCN or the CAPRA score.
•
We evaluated a predetermined score threshold under which patients had excellent outcomes regardless of treatment with single- or multimodal therapy. Thus, the score threshold can be used as a decision aid to counsel men whether or not single-modality therapy would be sufficient for their intermediate- or high-risk prostate cancer. The CCR score can also be used by patients to understand the absolute risk reduction of metastasis using combined over single-modality therapies at any risk score. Therefore, men may make an informed choice about the therapeutic ratio of using single- or multimodality therapy regardless of the threshold.
Disclosure
J. Tward has served on an advisory board and consulted for Myriad Genetics, Inc and Decipher Biosciences; has received research funding from Bayer for work outside of the present manuscript. S. Freedland received support from Myriad Genetics, Inc during the conduct of the study. Myriad Genetics, Inc employed L. Lenz, D. Flake, T. Cohen, and S. Stone during the study. The remaining authors have stated that they have no conflicts of interest.
CRediT authorship contribution statement
Jonathan D. Tward: Conceptualization, Methodology, Validation, Formal analysis, Data curation, Resources, Writing - original draft, Writing - review & editing, Supervision, Funding acquisition, Project administration. Thorsten Schlomm: Resources, Writing - review & editing, Supervision. Stephen Bardot: Resources, Writing - review & editing, Supervision. Daniel J. Canter: Resources, Writing - review & editing, Supervision. Troy Scroggins: Resources, Writing - review & editing, Supervision. Stephen J. Freedland: Resources, Writing - review & editing, Supervision. Lauren Lenz: Methodology, Validation, Formal analysis, Data curation, Writing - review & editing. Darl D. Flake: Methodology, Validation, Formal analysis, Data curation, Writing - review & editing. Todd Cohen: Resources, Writing - review & editing, Supervision, Funding acquisition, Project administration. Michael K. Brawer: Resources, Writing - review & editing, Supervision, Funding acquisition, Project administration. Steven Stone: Resources, Writing - review & editing, Supervision, Funding acquisition, Project administration. Jay Bishoff: Resources, Writing - review & editing, Supervision.
Acknowledgments
The authors wish to thank Jennifer Logan, PhD, and Stephanie Meek, PhD, for editorial assistance with the manuscript. This work was supported by Myriad Genetics, Inc; the Huntsman Cancer Foundation; and the National Cancer Institute of the National Institutes of Health [award number P30CA042014]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Supplemental Methods
No proprietary statistical code was used in this study. All analyses used R software version 3.5.0 or higher (R Core Team, Vienna Austria), or STATA (StataCorp, LLC, 2017. Stata Statistical Software: Release 15. College Station, TX).
Classification and Regression Tree (CART) Analysis
CART analysis was performed to identify men in the below-threshold group who might benefit from multimodality therapy and, conversely, those in the above-threshold group who might not, based on other prognostic covariates. Benefit was defined as no metastasis. Martingale residuals of a Cox model were used to calculate approximate χ2 values for all possible cutpoints on CART covariates, including clinical cell-cycle risk (CCR) score as a continuous variable, single- or multimodality therapy, radiation (RT) or surgery as initial therapy, and National Comprehensive Cancer Network (NCCN) risk group. When the use of single- or multimodality therapy created a subgroup split, a threshold of P ≤ .05 with a minimum of 10 patients and 3 failures per subgroup was deemed sufficient for further analysis.
Derivation of Absolute Risk Reduction and Number Needed to Treat
To predict the absolute risk reduction of metastasis for RT + androgen deprivation therapy versus RT alone at any CCR score, a bivariate Cox proportional hazards model was first fit on patients receiving primary RT using treatment modality and CCR score as a continuous variable. The predicted risk of metastasis at 10 years with confidence intervals was then calculated for all CCR scores between 0 and 5, for multimodality versus single-modality therapy. The absolute risk reduction for single- versus multimodality therapy at any CCR score was the difference between these risk estimates. The number needed to treat with single- versus multimodality therapy to prevent metastasis at any CCR score was computed using the reciprocal of the absolute risk reduction. At the multimodality threshold or by NCCN risk group, the absolute risk reduction was estimated using Kaplan-Meier methods.
Abbreviations: CAPRA = University of California San Francisco Cancer of the Prostate Risk Assessment; CCP = cell cycle progression; HCI = Huntsman Cancer Institute; NCCN = National Comprehensive Cancer Network.
Supplemental Figure 3Classification and Regression Tree Analysis for Metastasis Risk. Relative Hazard Ratios Compared With the Baseline Population Hazard for the Below-Threshold and Above-Threshold Cohorts. Covariates Used in the Model Were CCR Score, Use of Radiation or Radical Prostatectomy, NCCN Risk Group, and Single-Modality or Multimodality Therapy. The Green Box Indicates Risk Below the Baseline Hazard, whereas the Red Box Indicates Risk Above the Baseline Hazard
Supplemental Table 1Sensitivity Analysis With and Without the Huntsman Cancer Institute Cohort
Analysis
Subgroup
Full Cohort (N = 718)
Cohort Without HCI Patients (N = 580)
N
HR (95% CI)
P Value
N
HR (95% CI)
P Value
Prognostic ability of CCR score
N/A
718
3.75 (2.71-5.20)
1.6 × 10−16
580
3.70 (2.60-5.25)
2.7 × 10−14
Prognostic ability of CCR score after accounting for CAPRA
Patients who received surgery
486
4.08 (1.90-8.78)
5.7 × 10−4
407
4.16 (1.78-9.72)
1.6 × 10−3
Patients who received RT
232
4.30 (2.23-8.30)
6.2 × 10−5
173
3.91 (1.95-7.85)
4.5 × 10−4
Prognostic ability of multimodal threshold
Patients treated with single-mode therapy
561
15.90 (5.43-46.52)
8.1 × 10−10
443
12.46 (4.19-37.06)
9.7 × 10−8
Impact of multimodal therapy after accounting for CCR score
Patients who received RT
232
0.25 (0.09-0.69)
.013
173
0.24 (0.08-0.68)
.012
Abbreviations: CAPRA = University of California San Francisco Cancer of the Prostate Risk Assessment; CCR = clinical cell-cycle risk; CI = confidence interval; HCI = Huntsman Cancer Institute; HR = hazard ratio; RT = radiation.
Clinically localized prostate cancer: ASCO Clinical Practice Guideline Endorsement of an American Urological Association/American Society for Radiation Oncology/Society of Urologic Oncology Guideline.
The University of California, San Francisco Cancer of the Prostate Risk Assessment score: a straightforward and reliable preoperative predictor of disease recurrence after radical prostatectomy.
Comparison of the prognostic utility of the cell cycle progression score for predicting clinical outcomes in African American and non-African American men with localized prostate cancer.
Transatlantic Prostate Group. Prognostic value of an RNA expression signature derived from cell cycle proliferation genes in patients with prostate cancer: a retrospective study.
Defining biochemical failure following radiotherapy with or without hormonal therapy in men with clinically localized prostate cancer: recommendations of the RTOG-ASTRO Phoenix Consensus Conference.
Long-term results with immediate androgen suppression and external irradiation in patients with locally advanced prostate cancer (an EORTC study): a phase III randomised trial.
5191: Timing of radiotherapy (RT) after radical prostatectomy (RP): first results from the RADICALS RT randomised controlled trial (RCT) [NCT00541047].
Immediate versus deferred androgen deprivation treatment in patients with node-positive prostate cancer after radical prostatectomy and pelvic lymphadenectomy.
00013-6&id=gr1.jpg){kind=link}
00013-6&id=gr2.jpg){kind=link}
00013-6&id=gr3.jpg){kind=link}
00013-6&id=fx1.jpg){kind=link}
00013-6&id=fx2.jpg){kind=link}
00013-6&id=fx3.jpg){kind=link}