The Development of Intermediate Clinical Endpoints in Cancer of the Prostate (ICECaP)

Abstract

New systemic therapies have prolonged the lives of men with metastatic castration-resistant prostate cancer (mCRPC). Use of these therapies in the adjuvant setting when the disease may be micrometastatic and potentially more sensitive to therapies may decrease mortality from prostate cancer. However, the conduct of adjuvant prostate cancer clinical trials is hampered by taking longer than a decade to reach the meaningful endpoint of overall survival (OS) and the fact that many men never die from prostate cancer, even if they relapse. A validated intermediate clinical endpoint (ICE) in prostate cancer that is a robust surrogate for OS has yet to be defined. This paper details the plans, process, and progress of the international Intermediate Clinical Endpoints in Cancer of the Prostate (ICECaP) working group to pool individual patient data from all available clinical trials of radiation or prostatectomy for localized disease and conduct the requisite analyses to determine whether an ICE can be identified. This paper further details the challenges and the a priori statistical analytical plans and strategies to define an ICE for adjuvant prostate cancer clinical trials. In addition, a brief review of the health economic analyses to model the benefits to patients, society and manufacturers is detailed. If successful, the results from this work will provide a robust surrogate for OS that will expedite the design and conduct of future adjuvant therapy trials using new agents that have proven activity in mCRPC. Moreover, it will also define the health economic benefits to patients and societies.

Despite a high rate of screening resulting in early identification of most of the 230 000 new cases of prostate cancer in the United States in 2014, there were still about 29 000 deaths (1), the second leading cause of male cancer deaths. A closer look reveals that about 5% of the 230 000 cases are de novo metastatic and account for approximately 10 000 deaths (2,3). De novo metastatic prostate cancer patients usually present with symptoms and have a median age of 73 years and a median survival of about three years (4). However, most prostate cancer deaths are in patients who recur after therapy with curative intent for localized disease. An analysis of the Surveillance, Epidemiology, and End Results (SEER) registry data revealed that 8.2% of 548 995 patients with localized prostate cancer between 1998 and 2011 (unpublished data) died of prostate cancer (15 748 [10.0%] of the 157 314 treated with radiation therapy, 10 579 [2.8%] of 371 579 treated with prostatectomy, and 1888 [9.4%] of the 20 102 treated with prostatectomy plus radiation died of prostate cancer). By extrapolation (and some assumption), 8% of 230 000 would account for 17 000 prostate cancer deaths in United States.

A major decrease in the deaths from prostate cancer could be achieved with more effective adjuvant therapy for high-risk clinically localized disease. For example, decreasing the relapses and subsequent death from prostate cancer death by 25% would save approximately 4250 lives in the United States each year. The precedence to support the plausibility of this goal is early identification with prostate-specific antigen (PSA) screening and use of adjuvant androgen deprivation therapy (ADT) (especially for high-grade, high-volume disease) possibly accounts for the 30% decrease death rate from prostate cancer in the United States (5–8). More intense adjuvant therapy is now possible with the emergence of new agents active in the more resistant mCRPC state (9–16).

The current approach of using OS in studies of adjuvant prostate cancer therapies requires long-term follow-up—typically more than 10 years from concept to final report. An examination of 28 randomized trials for which data were available showed the median time from first patient enrolled to database lock for publication of primary outcomes is 11.5 years (interquartile range [IQR] = 9.8–15.8) (Table 1) and does not include time from concept development to study activation. The investment in time and financial resources for a pharmaceutical sponsor has been prohibitive given the patent lifespan of a new drug. Consequently, most adjuvant studies have been conducted by the cooperative groups in Europe, Australia/New Zealand, the United Kingdom, Canada, and the United States.

Table 1.

Trial characteristics of all eligible trials and of trials included in ICECaP as the training set*

Trial details	All potentially eligible trials	Trials participating ICECaP training set
Total No. of trials	102	43
Total No. of patients	61 904	28 905
No. randomized per trial, median (IQR)	395 (201-830)	413 (217-843)
Follow-up, median (IQR), y	7.1 (5.6-10.0)	7.6 (6.0-10.0)
Years from 1st enrollment to database lock for primary outcome analysis, median years, (No. of trails) (IQR)	11.5 (N= 28) (9.8-15.8)	11.1 (N= 19) (10.2-14.2)
Study region, No. (%)	N (%)	N (%)
US	37 (36)	15 (35)
Canada	28 (27)	11 (26)
UK	17 (17)	10 (23)
Europe	37 (36)	15 (35)
ANZ	8 (8)	5 (12)
Trial status, No. (%)
Outcomes Published	64 (63)	35 (81)
Completed Accrual	26 (25)	7 (16)
Terminated	8 (8)	1 (2)
Recruiting	4 (4)
Type of treatment, No. (%)
RT- based
RT+/-ADT/Chemo/Other	30 (29)	14 (33)
RT dose	22 (22)	10 (23)
ADT+/-RT	5 (5)	3 (7)
Adjuvant/Salvage RT	7 (7)	2 (5)
RP- based
RP+/-ADT/Chemo/Other	18 (18)	5 (12)
RP & RT
RP/RT +/-ADT/Chemo/Other	4 (4)	4 (9)
RP vs RT vs Watchful Waiting (No ADT)	10 (10)	3 (7)
No local (WW vs ADT) or Other	6 (6)	2 (4)
Year of first enrollment, No. (%)
1967-1979	4 (4)
1980-1989	13 (13)	9 (21)
1990-1994	13 (13)	6 (14)
1995-1999	29 (28)	19 (44)
2000-2004	16 (16)	4 (9)
2005-2009	13 (13)	1 (2)
Unknown	14 (14)	4 (9)
Year of last enrollment, No. (%)
1967-1979	2 (2)
1980-1989	4 (4)	2 (5)
1990-1994	9 (9)	5 (12)
1995-1999	19 (19)	13 (30)
2000-2004	26 (25)	14 (33)
2005-2009	12 (12)	4 (9)
2010-present	8 (8)
Unknown	22 (22)	5 (12)
OS as the primary endpoint, No. (%)	26 (25)	10 (23)
Endpoints include, No. (%)
OS	88 (86)	42 (98)
PCSM	57 (56)	25 (58)
TTM/MFS	65 (64)	31 (72)
PSA progression	73 (72)	33 (77)

* ADT = androgen deprivation therapy; MFS = metastasis-free survival; OS = overall survival; PCSM = prostate cancer–specific mortality; PSA = prostate-specific antigen; RP = radical prostatectomy; RT = radiation therapy; TTM = time to metastasis; WW = watchful waiting.

This timeframe is due to the lengthy “average” disease course for localized prostate cancer until death from prostate cancer. Moreover, the risk of death from prostate cancer despite PSA surveillance and treatment is extremely variable, ranging from low (<5%) to a high (> 40%) chance of death. Moreover, the risk of prostate cancer death is impacted by the patient’s age and comorbidities. At one extreme is the man age 50 years with high-volume Gleason 9 cancer, a PSA recurrence 18 months after therapy for localized disease followed by metastases three years after PSA rise (17) vs a man age 68 years who had a prostatectomy for Gleason 7 disease and has PSA relapse at age 75 years (18) and a man age 78 years with active coronary artery disease with Gleason 6 who probably does not even require local therapy (19).

Prior Attempts for Finding an Intermediate Clinical Endpoint (ICE) in Prostate Cancer

Potential ICEs as surrogate for OS in the setting of localized prostate cancer treated with radiation therapy have been reported, include time to biochemical failure, PSA-doubling time (PSA-DT), PSA nadir, end of treatment PSA, disease-free survival (DFS), metastasis-free survival (MFS). Examples are the following: An analysis of the TROG 96.01 trial population (20) noted that time to biochemical failure was better than PSA-DT at predicting the trial finding and satisfied all of Prentice’s criteria (21). PSA relapse at cutpoints of less than 1.5 years, less than 2 years, and less than 2.5 years were potential surrogates for prostate cancer–specific survival. PSA nadir and end-of treatment PSA have been investigated in radiation-based trials with or without six months of ADT in the TROG 96.01 and a DFCI trial (22). The authors found that PSA immediately after treatment and lowest PSA concentration (PSA nadir > 0.5ng/mL) predicted the trial finding and satisfied the Prentice’s criteria for surrogacy. A composite endpoint of a PSA of 25ng/mL or greater, initiation of treatment, and disease-free survival in the RTOG92-02 at three years met Prentice’s criteria as surrogate endpoint for 10-year prostate cancer–specific survival (23). In the same dataset, metastasis-free survival at three years but not five years was consistent with Prentice’s criteria as surrogate endpoint for 10-year prostate cancer–specific survival.

Post-treatment PSA doubling time (PSA-DT) of three months or faster met Prentice’s criteria (21) as a potential surrogate endpoint for prostate cancer–specific survival or OS among men treated with radiation therapy or prostatectomy using the CAPSURE and CPDR databases (24). The challenges with PSA-DT are that it can only be applied to those who have a PSA relapse, it requires collecting information on PSA values beyond relapse (which was not done in all studies), and it may be influenced by other treatments, such as ADT. Moreover, this endpoint was not confirmed in the RTOG 92-02 trial population, in which men were randomly assigned to neoadjuvant ADT plus radiation therapy with or without two years of adjuvant ADT (25). In an analysis of the TROG 96.01 trial population, which randomized men to radiation with or without six months of ADT (20), PSA-DT predicted the trial finding and satisfied all Prentice’s criteria at cutpoints of less than 12 months and less than 15 months as potential surrogate endpoint for prostate cancer–specific survival.

Types of Potential ICEs in Prostate Cancer

In this manuscript, we describe a formal approach for the assessment of ICEs that may serve as robust surrogates for OS and hence expedite readouts from future adjuvant prostate cancer trials. Validated ICEs in breast and colorectal adjuvant trials are used for drug development (26-29). Potential prostate cancer ICEs are: 1) time to biochemical failure defined according to local therapy. If patient had a prostatectomy, a PSA of greater than 0.2ng/mL would be deemed a threshold for failure (30), and postradiation to intact prostate, the Phoenix definition of a rise by 2ng/mL or more above the nadir PSA, is considered as biochemical failure (31); 2) metastasis-free survival, with events defined as documented metastatic disease or death from any cause; 3) disease-free survival includes as events the first evidence of recorded clinical recurrence (local/regional progression and/or distant metastases confirmed by imaging or histological evidence) or death from any cause; 4) event-free survival (EFS) includes any documented disease event (same as DFS plus biochemical failure) or death from any cause.

DFS will be the endpoint of primary interest as it can be applied to radiation and prostatectomy trials as well as trials where ADT was deployed. Moreover, it is an endpoint that is reported or can be constructed from the endpoints provided from the individual patient data (IPD) in the ICECaP data repository. It has some potential to be influenced by ADT instituted for preceding biochemical relapse, and the DFS endpoint hence emerges as mCRPC. This has a different prognosis than DFS emerging with no prior ADT—ie, presenting with metastatic disease as first evidence of relapse or having a PSA relapse, but in which ADT is deferred until radiographic progression. An endpoint of secondary interest is MFS, which is a subset of DFS as it does not include local progression events. This endpoint is not uniformly captured in the databases of the clinical trials.

PSA-based endpoints are subject to a number of confounding factors such as recovery of testosterone after adjuvant ADT and institution of ADT for rising PSA after the local therapy or institution of salvage curative therapy (eg, radiation postprostatectomy). As such, although a PSA rise most often precedes DFS endpoints and would be a more expeditious intermediate endpoint, it is subject to many confounders from treatment and the long lead time to metastases and death from prostate cancer in many patients results in patients dying of other diseases. It is possible a decrease in early PSA relapse will be a surrogate of OS but it is also anticipated that for it to be robust ICE, the hypothesized difference in the PSA-based ICE between control and experimental group will have to be substantially larger than that for DFS. This conjecture is based on the expectation that a decrease in DFS (a more substantial endpoint than a biochemical based ICE) will track more closely with decreased death from prostate cancer and resultant increase in OS. In contrast, it is expected a biochemical-based ICE will have more deaths from/competing risks of other causes during the lead time prior to possible death from prostate cancer). Event-free survival (biochemical relapse plus DFS) will therefore be our third endpoint of interest, recognizing very few events will be DFS without preceding PSA rise. The fourth endpoint of interest will be biochemical relapse–free survival.

In this commentary, we will analyze existing data from randomized, controlled clinical trials to identify promising ICEs. At the completion of the analysis of a training set data, one or more ICEs will be taken forward for validation in prospective trials to assess their potential to expedite adjuvant prostate cancer clinical trials. A relevant ICE will be one that is applicable to both radiation and prostatectomy studies and studies with different systemic therapies (ADT, chemotherapy). A viable ICE will be one that is feasible to measure a proposed clinical difference between the experimental group and control with a reasonable sample size. An efficient ICE is one that can be used as early as possible after random assignment to expedite a readout evaluating a new therapeutic strategy for localized disease. We will also assess whether more proximal ICEs are surrogates for later endpoints, such as DFS being an ICE for prostate cancer–specific mortality, which in turn is a surrogate for OS.

Process for Finding a Validated ICE in Prostate Cancer

Trial Selection

The most robust process to assess possible clinical endpoints for surrogacy with OS requires use of individual patient data (IPD) of all randomized, controlled trials. To that end the ICECaP working group is gathering IPD from as many randomized, controlled trials for localized disease as possible, which were conducted in Canada, the United Kingdom, Europe, Australia/New Zealand, and the United States and for which accrual has been closed. Reasons for exclusion are trials with primary endpoints such as safety, toxicity, quality of life, feasibility, dosimetry, and patient decision-making without systematic long-term follow-up. The Systematic Reviews and Meta-Analysis (per PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement) approach has been using PubMed, Clinical Trials registries, Cooperative Groups’ websites, meetings’ websites, review of citations in publications including meta-analysis papers, other sources (eg, personal communications). As of June 2013, 102 trials were identified as potentially eligible. Forty-three of these have data viable for use and agreed to participate (the training set: 28 905 patients); 12 trials with pending publication will participate in the future as the validation set. The selection process and reasons for exclusion are shown in Figure 1.

Figure 1.

Flowchart of section and participation of randomized clinical trials for localized prostate cancer in Intermediate Clinical Endpoints in Cancer of the Prostate.

Trial Characteristics

The characteristics of the 43 trials that participated in ICECaP as well as the 102 potentially eligible trials are listed in Table 1. The 43 trials include data on 28 905 patients with a median of 413 patients per trial (IQR = 217-843, median follow up for primary outcomes is 7.6 years (IQR = 6.0-10.0 years), 29 trials are radiation-based studies, and 13 trials (7171 patients) had prostatectomy as the primary modality. The majority of studies (42 out of 43) reported OS as the outcome; 58%, 72%, and 77% of these studies included prostate cancer–specific mortality (PCSM), time to distant metastasis (TTM) or MFS, and PSA progression as the endpoints, respectively. The trial characteristics of the 43 participating trials are similar to that of the entire 102 potential eligible trials identified (Table 1).

The ICECaP Surrogacy Analysis Plan

Having gathered the data from the randomized controlled trials, a meta-analysis of the IPD will be undertaken. There are a number of surrogacy analyses approaches (reviewed in the Supplementary Material, Statistical Analysis Plan [SAP], available online) using both IPD and trial-level data. The primary analysis of the ICECaP effort will use the correlation approach as described by Burzykowski, Molenberghs, and Buyse (32-34) at both the IPD and trial data levels. Essentially, the correlation approach requires two conditions be fulfilled: 1) the surrogate is prognostic for disease outcome (ie, the surrogate and true endpoints are strongly correlated). This is known as ‘individual level’ surrogacy or ‘outcome’ surrogacy and means that the candidate surrogate endpoint must be associated with outcome in the same fashion as a prognostic marker, without reference to specific interventions; and 2) the effect of intervention on the surrogate is sufficiently correlated with the effect on the true endpoint. This is known as ‘trial level’ surrogacy or ‘effect’ surrogacy, as it must be demonstrated for a group of patients in a clinical trial.

For individual-level surrogacy analysis, the analysis will depend on the type of ICE considered. For time-to-event ICEs, a bivariate copula model will be fitted using IPD for ICEs and OS (32-34). The bivariate copula model (Hougaard’s or Clayton’s model, symmetric in both ICE and OS) will yield a nonparametric association parameter (Kendall’s tau [35]) between the endpoints. Kendall’s tau is a rank correlation for bivariate time-to-event endpoints subject to censoring and measures the dependence between the two endpoints.

For categorical ICEs, the strength of association between the ICE and OS will be assessed through the prognostic effect of the ICE on OS either by examining time to death using a standard Cox regression or assessing OS at specific timepoints (eg, 5-year OS) using logistic regression. The strength of association will be quantified as an odds ratio or a hazard ratio with 95% confidence intervals. The prediction accuracy will be assessed using the receiver operating characteristic methods. Sensitivity, specificity, and positive and negative predictive values for survival will be summarized.

For trial-level surrogacy analysis, the correlation between treatment effects on ICEs and OS will be estimated using an error-in-variables regression model on the estimated treatment effects on ICEs and on OS. Such a model appropriately accounts for estimation errors in the treatment effects, but often convergence issues prevent such a model to be fitted. The correlation coefficient (or its square, R², which is the proportion of variance explained by the regression) will be provided.

An intermediate clinical endpoint will be claimed to be an acceptable surrogate endpoint for OS if there is a strong correlation between endpoints and there is strong correlation between treatment effects on the endpoints. Consistent with other surrogacy assessments (36), we will claim surrogacy when the squared correlation is 0.7 or higher (SupplementaryTable 1, available online). However, the decision to take a potential ICE to validation set will be based on the “relevance” of the proposed ICE) (discussed above) and Health Economic Analysis (discussed below). The surrogate threshold effect (STE) will also be defined as the minimum treatment effect on the surrogate necessary to predict a nonzero effect on OS in a future trial (32). A future trial would then require an upper limit of the confidence interval for the estimated treatment effect on the surrogate to fall below the STE and hence be deemed a reliable surrogate of OS.

To gain greater confidence in the findings, sensitivity analyses will be conducted. These include: 1) exclusion of trials exhibiting extreme treatment benefits on either ICE or OS; 2) exclusion of trials with extremely long or short follow-up; 3) using a trial-specific definition for the ICE (compared with the use of a standardized definition); and 4) adjustment for prognostic factors (ie, trial stratification factors) in the estimation of the treatment effect on the surrogate and OS.

Challenges for Finding an ICE in Prostate Cancer

There are potential challenges in identifying an ICE for prostate cancer, a heterogeneous disease afflicting a heterogeneous group of patients (with many at risk of dying of another illness), managed with heterogeneous treatments, and relapsed disease has a variable natural history. Moreover, therapy may impact the endpoint (eg, ADT as part of primary treatment decreases PSA production and may impact this potential ICE, as some men may not have return of testosterone). Also, a PSA rise postprostatectomy may prompt curative salvage radiation. The impact of comorbidities and non–prostate cancer deaths may also dilute the association of an ICE with the OS-based endpoint.

The primary analysis of surrogacy endpoints will be conducted in the overall population combining different types of therapy. However, to address heterogeneity, subgroup analyses will be conducted by type of therapy because ICEs may be different for different types of treatment. The following treatment types are planned in subgroup analysis: 1) prostatectomy trials vs radiation therapy trials and 2) within radiation therapy trials, analysis with and without ADT. For each study treatment type, IPD from a robust number of trials will be pooled to perform an IPD meta-analysis. When an insufficient number of trials are available for a certain study treatment type, trial design elements such as geographic region may be used as to divide trials into separate units for analysis.

To account for the heterogeneity in cancer prognosis, subgroups with highest risk of prostate cancer death will be prospectively identified (eg, men younger than than age 65 years with high-grade prostate cancer) using risk groups based on established criteria (37-47): age at random assignment (<65 years, 65-74 years, ≥75 years), Gleason score (<7, 7, and >7 or <7 vs ≥7), PSA at diagnosis or at random assignment (≤10, >10 to ≤20 and >20ng/mL), clinical T stage or pathology T stage from prostatectomy (T1, T2, and T3; ≥T2c vs <T2c), risk group (NCCN or D’Amico criteria; low, intermediate, and high), and years of diagnosis (random assignment, grouped by 5-year increments). The main goal of subgroup analysis is to see if an ICE may be solely or particularly more robust for those most at risk of dying of prostate cancer, which will in turn be reflected in improved OS (as the risk of death from a comorbidity did not dilute the impact of decrease risk of prostate cancer death).

Statistical Power for Surrogacy Analysis in ICECaP

From the 43 modeling trials anticipated to be available in 2015, the projected total number of patients is 28 389, with about 8000 deaths from any cause (25%-30% of population) and about 2000 prostate cancer deaths (8%-10% of population). The median number of randomly assigned patients is about 400 per trial (mean = 660; range = 63-3603; IQR = 217-843).

For patient level surrogacy, given the large number of patients included in the meta-analysis approach, the 95% confidence intervals will tightly surround the observed correlation estimates. For the trial-level surrogacy, Table 2 provides the lower bound of the two-sided 95% confidence interval for various correlation coefficients to be detected between the treatment effects on ICEs and OS and for different numbers of trials (units) analyzed. For example, for 50 trials and a correlation coefficient of 0.8 for a potential ICE as a surrogate for OS, the lower bound limit of the two-sided 95% confidence interval is 0.67.

Table 2.

Lower bound of the two-sided 95% confidence interval of the correlation coefficient for different numbers of trials

No. of trials (units)	Correlation coefficients to be detected
	0.2	0.3	0.4	0.5	0.6	0.7	0.8	0.9
20	-0.27	-0.16	-0.05	0.07	0.21	0.37	0.55	0.76
30	-0.17	-0.07	0.05	0.17	0.31	0.45	0.62	0.80
40	-0.12	-0.01	0.10	0.22	0.35	0.50	0.65	0.82
50	-0.08	0.02	0.14	0.26	0.39	0.52	0.67	0.83
60	-0.06	0.05	0.16	0.28	0.41	0.54	0.69	0.84
70	-0.04	0.07	0.18	0.30	0.43	0.56	0.70	0.84

Once a potential surrogate is identified, validation studies will be performed. These will include a cross-validation approach where the model accuracy will be assessed by a leave-one-out cross-validation strategy and a linear model rebuilt on the other n-1 trials weighted by trial size. This model is then applied to the omitted trial to compare the predicted and observed treatment effect on OS. External and independent validation will be done using future clinical trial datasets that will report OS in 2015 and later. The observed and predicted treatment effect on OS (and its 95% CI) will be compared.

The SAP accompanying this manuscript as Supplementary Material (available online) includes a discussion of the statistical methods accounting for heterogeneity.

Health Economic Analysis

Regulatory agencies throughout the world, such as the Food and Drug Administration (FDA), are mandated to ensure safe and effective medical products are approved expeditiously. In 1992, the Prescription Drug User Fee Act (PDUFA) was passed by the United States Congress permitting the FDA to collect fees from sponsors of a drug to fund the new drug approval process with the provision that the FDA meet performance benchmarks with an emphasis on ensuring an expeditious process. A number of alternative approval systems have been developed, as well as an overall recognition of the value of faster access to valuable therapies including using biomarkers and ICEs as surrogates for a given clinical outcome.

The basic tradeoff in using ICE is between the speed and quality of approval decisions. The potential benefits for more rapid approval of a therapy based on a validated ICE include patients getting earlier access to a given therapy. For the developer, the value of the new therapy rises given realization of profits sooner, which also enable further research. Also, the benefit to society is that earlier access to an effective adjuvant therapy will decrease costs for treating fewer patients with metastatic disease and for reducing loss of productivity. However, there are potential “costs” for ICEs if inaccuracy leads to nonapproval of beneficial therapy (eg, from a false-negative study) and approval of nonbeneficial therapy (eg, from a false-positive) and leads to futile therapy with toxicity.

As such, health economist members of the ICECaP Working Group will perform a decision analysis of the quantitative magnitudes involved on both sides of this tradeoff. In particular, they will evaluate the quality of life and economic impact of implementing approval of an adjuvant therapy for prostate cancer based upon an ICE vs using OS. This will be done by modeling high-risk clinically localized prostate cancer from diagnosis to death reflecting current practice and comparing this with an ICE-based approval by the FDA that provides all men with immediately available adjuvant therapy sooner. The primary outcome measures will be life expectancy, quality-adjusted life-year, toxicity and financial cost from a societal perspective (see SAP for expanded discussion). Specifically, the analysis will make use of transition probabilities such as the probability of developing metastasis after biochemical recurrence by using the ICECaP database and literature review. Utilities using preexisting data from published literature will also be assessed and include quality-of-life values on a score of 0 to 1 for particular health states such as no relapse, biochemical relapse treated with ADT, and CRPC.

Given the fact that cost constraints impact delivery and access to care across the globe, potential economic benefits from use of ICE will be assessed. Direct medical costs of treating and monitoring prostate cancer will be assessed using data from Medicare reimbursement schedules. Indirect medical costs (opportunity costs) will evaluate costs of patients’ loss of productivity and time. The model will also assess differences in life expectancy and quality-adjusted life-year expectancy and the costs between expedited (with use of an ICE) and delayed approval (based on OS). In particular, this work will make use of data derived from years of life saved by earlier adjuvant therapy, quality of life improvement because of decrease in incidence of recurrence, costs of treatment of recurrence and metastatic disease, and opportunity cost to society by having used an ICE rather than OS in adjuvant trials. Varying the performances of the hypothetical, adjuvant therapy will be fed into the model to see how it affects the outputs, and then extrapolations will be made to help decide what predictive ability is needed to result in an acceptable cost-effectiveness profile.

Conclusion

Incremental advances in the longevity of men with mCRPC have been made with development of new systemic therapies. Use of these therapies in the adjuvant setting when the disease may be micrometastatic and more sensitive to therapies may decrease deaths from prostate cancer and build upon the prior advances with adjuvant ADT in high-risk disease. With the goal of making this claim a reality, the international ICECaP working group has come together to assess whether an ICE can be used to expedite the design and conduct of future adjuvant therapy trials, using new agents active in mCRPC, as well as deploy health economic analyses to model the benefits to patients and societies around the globe.

Funding

Prostate Cancer Foundation (PCF) Challenge Award; Grants from Industry (awarded to PCF): Millennium-Takeda, Sotio, Janssen, Astellas/Medivation, Sanofi.

Writing Committee: Coordinating Center at Dana Farber Cancer Institute (DFCI): Christopher Sweeney (Chair DFCI), Brandon Bernard (DFCI), Mari Nakabayashi (DFCI), Meredith Regan (DFCI), Wanling Xie (DFCI); Health Economics and Utilities Analyses: Julia Hayes (DFCI), Nancy Keating (Brigham and Women’s Hospital and Harvard Medical School), Suhui Li (George Washington University), Tomas Philipson (University of Chicago); Members Independent of Coordinating Center overseeing Statistical Analysis Plan: Marc Buyse (International Drug Development Institute), Susan Halabi (Duke University), Philip Kantoff (DFCI, Prostate Cancer Foundation [PCF]), A. Oliver Sartor (Tulane University), Howard Soule (PCF); Additional Research: Brandon Mahal (Harvard Medical School) (unpublished analysis of SEER data-base).

Executive Committee: Christopher Sweeney (DFCI)-Chair, Philip Kantoff (DFCI, PCF), Howard Soule (PCF).

ICECaP Working Group Members (in alphabetical order): Ove Andren, John Armstrong, Donald Berry, Michel Bolla, Marc Buyse, Simon Chowdhury, Noel Clarke, Laurence Collette, Matthew Cooperberg, Jim Denham, Mario Eisenberger, James Dignam, Karim Fizazi, Boris Freidlin, Martin Gleave, Muriel Habibian, Susan Halabi, Julia Hayes, Nick James, Jonathan Jarow, Nancy Keating, Philip Kantoff, Gary Kelloff, Laurence Klotz, Suhui Li, Himu Lukka, Brandon Mahal, Malcolm Mason, Andrea Miyahara, Mari Nakabayashi, Wendy Parulekar, Tomas Philipson, Meredith Regan, Howard Sandler, Oliver Sartor, Peter Scardino, Howard Scher, Richard Simon, Jonathan Simons, Eric Small, Howard Soule, Christopher Sweeney, Matthew Sydes, Catherine Tangen, Ian Thompson, Bertrand Tombal, Anders Widmark, Thomas Wiegel, Scott Williams, and Wanling Xie.

The industry funders had no role in the writing of the commentary or the decision to submit it for publication.