A Randomised Phase 2 Trial of Intensive Induction Chemotherapy (CBOP/BEP) and Standard BEP in Poor-prognosis Germ Cell Tumours (MRC TE23, CRUK 05/014, ISRCTN 53643604)

Background: Standard chemotherapy for poor-prognosis metastatic nonseminoma has remained bleomycin, etoposide, and cisplatin (BEP) for many years; more effective regimens are required. Objective: To explore whether response rates with a new intensive chemotherapy regimen, CBOP/BEP (carboplatin, bleomycin, vincristine, cisplatin/BEP), versus those in concurrent patients treated with standard BEP justify a phase 3 trial. Design, setting, and participants: We conducted a phase 2 open-label randomised trial in patients with germ cell tumours of any extracranial primary site and one or more International Germ Cell Cancer Collaborative Group poor-prognosis features. Patients were randomised between 2005 and 2009 at 16 UK centres. Intervention: BEP (bleomycin 30 000 IU) was composed of four cycles over 12 wk. CBOP/ BEP was composed of 2 CBOP, 2 BO, and 3 BEP (bleomycin 15 000 IU). Outcome measurements and statistical analysis: Primary end point was favourable response rate (FRR) comprising complete response or partial response and normal markers. Success required the lower two-sided 90% confidence limit to exclude FRRs <60%; 44 patients on CBOP/BEP gives 90% power to achieve this if the true FRR is 80%. Equal numbers were randomised to BEP to benchmark contemporary response rates. Results and limitations: A total of 89 patients were randomised (43 CBOP/BEP, 46 BEP); 40 and 41, respectively, completed treatment. CBOP/BEP toxicity, largely haematologic, was high (96% vs 63% on BEP had Common Terminology Criteria for Adverse Events v.3 grade 3). FRRs were 74% (90% confidence interval [CI], 61–85) with CBOP/BEP, 61% with BEP (90% CI, 48–73). After a median of 58-mo follow-up, 1-yr progression-free survival (PFS) was 65% and 43%, respectively (hazard ratio: 0.59; 95% CI, 0.33–1.06); 2-yr overall survival (OS) was 67% and 61%. Overall, 3 of 14 CBOP/BEP and 2 of 18 BEP deaths were attributed to toxicity, one after an overdose of bleomycin during CBOP/BEP. The trial was not powered to compare PFS. Conclusions: The primary outcome was met, the CI for CBOP/BEP excluding FRRs <61%, but CBOP/BEP was more se an international pha


Introduction
The management of metastatic germ cell tumours (GCTs) with platinum-based chemotherapy represents a major success story. However, a poor prognostic group can be defined that achieved cure rates <50% in an international pooled analysis [1].
CBOP/BEP results from previous studies [18,19] suggested high activity with increased toxicity. However, case selection or improved management over time hinders historical comparisons. This randomised phase 2 trial (ISRCTN 53643604) evaluated CBOP/BEP and BEP at the same time.

2.
Patients and methods

Patients
Applicable regulatory and ethics approvals and written informed consent were obtained. Eligible patients were ! 16

Treatment
The control arm comprised four 3-weekly cycles of Indiana-style BEP, and the CBOP/BEP arm comprised six cycles over 15 wk (see Fig. 1  Patient summary: In this study we tested a new, more intensive way to deliver a combination of drugs often used to treat men with testicular cancer. We found that response rates were higher but that the CBOP/BEP regimen caused more short-term toxicity. Because most patients are diagnosed when their cancer is less advanced, it took twice as long to complete the trial as expected. Although we plan to carry out a larger trial, we will need international collaboration.

Statistical considerations
A single-stage Fleming design was used, assuming an FRR for CBOP/BEP !80% would warrant further study and a rate <60% that historical data suggest with BEP [3] would not. With 44 patients randomised to CBOP/ BEP, the trial had 90% power to exclude response rates 60% with a = 5% (one-sided) when the true response rate was !80%. Thus ''success'' for the primary analysis required the one-sided 95% confidence limit (equivalently, the lower limit of the 90% two-sided CI) to exclude rates <60%. An equal number were randomised to BEP to benchmark the FRR; the trial was not powered to compare arms definitively with respect to efficacy. Continuation to a phase 3 trial powered for PFS required success as defined earlier in the primary analysis, that the BEP FRR was within the anticipated range, and that trial recruitment rate was adequate.
[ ( F i g . _ 1 ) T D $ F I G ] Two sensitivity analyses were prespecified: (1) The NED after surgery group was included as favourable responders; (2) the late orchidectomy partial response group was excluded from the favourable response category.
FRRs were also assessed according to receipt of pre-protocol stabilising chemotherapy (planned subgroup analysis) and according to histologic diagnosis (exploratory analysis). A per protocol population was defined a priori to include eligible patients receiving one cycle or more of protocol chemotherapy; however, all patients met these criteria. Preplanned timeto-event analyses included Kaplan-Meier curves for PFS and OS with treatment hazard ratios (HRs) derived from Cox regression models.

Results
A total of 89 patients were randomised from 16 UK centres, 46 to BEP and 43 to CBOP/BEP (Fig. 2). Baseline characteristics were well balanced ( Distinguishing between symptoms of severe disease, toxicity, and intercurrent illness or infection made it difficult to classify cause of death in several cases. Two on-treatment deaths in the CBOP/BEP arm (one in a patient who had received a bleomycin overdose) and one BEP death 3 mo after treatment were thought likely to be a result of lung damage associated with bleomycin. Bleomycin toxicity was also a possible contributory factor in three further deaths (one BEP, two CBOP/BEP) from infective respiratory conditions (postoperative adult respiratory distress syndrome [ARDS] [one], pneumonia [two]) that occurred 2-3 mo after completion of protocol chemotherapy.
[ ( F i g . _ 2 ) T D $ F I G ] A further two non-GCT deaths on each arm occurred during treatment. In the BEP arm these were due to haemorrhage of brain metastases when heparinised (a complication of neutropenic sepsis management) and a combination of severe disease, toxicity, and pneumonia after the first week of BEP. In the CBOP/BEP arm they were due to ARDS and septicaemia (a complication of neutropenic sepsis that developed in the first week of treatment) and to multiorgan failure associated with sepsis in a patient who was not neutropenic.
The policy change mandating G-CSF prophylaxis led to 98% (previously 40%) of BEP patients and 76% (previously 47%) of CBOP/BEP patients receiving G-CSF in the required cycles. However it did not have a substantial impact on the incidence or grade of neutropenia or febrile neutropenia (29% in CBOP/BEP before and after the protocol amendment), although an impact on duration of symptoms (not assessable) cannot be excluded.

Efficacy
Response was evaluated in all 89 patients (  (Table 3) were broadly consistent.
The FRR was 58.3% (39.7-75.4%) in patients having preprotocol stabilising chemotherapy and 70.8% (60.1-79.9%) in those that did not, and it was 64.2% (52.0-75.1%) in those with a histologic diagnosis compared with 72.2% (57.4-75.1%) in those without. The difference between arms appeared more marked in both those receiving pre-protocol chemotherapy and those without a histologic diagnosis (Table 3).
There were 14 non-GCT deaths (6 BEP, 8 CBOP/BEP) in total. In addition to the 10 deaths that occurred during and/ or were considered related to protocol treatment, described previously, 4 later deaths occurred that were not due to GCT: two (one in each arm) due to toxicity from second-line treatment; one thought likely due to recurrent teratoma,   but a second nonhaematologic malignancy could not be ruled out (CBOP/BEP); and one due to primary lung cancer (BEP).

Discussion
The primary efficacy end point was met, with BEP FRRs approximately as anticipated (61%); the FRR with CBOP/BEP was 74%, with the 90% CI excluding rates <60%. This was supported by encouraging PFS and OS data, particularly in patients with aggressive disease needing stabilising chemotherapy; however, acute toxicity with CBOP/BEP was high. Accrual was slower than anticipated; hence the criteria for proceeding immediately to phase 3 were not met, and it is clear that an adequately powered phase 3 trial would require international collaboration. The two previous nonrandomised CBOP/BEP studies also suggested high activity with increased toxicity. The first [18], in 54 IGCCCG poor-prognosis patients, showed 3-yr relapse-free survival of 83.2% (95% CI, 68.8-91.3) and 3-yr survival of 91.5% (95% CI, 78.6-96.8). In European Organisation for Research and Treatment of Cancer (EORTC) 30948 [19], 29 of 66 eligible patients had poor prognosis. After a median 40-mo follow-up, 1-yr PFS in the poorprognosis group was 81.8% (95% CI, 72.5-91.1). Two-year survival was 84.5% (95% CI, 75.6-93.3). In both studies, the major toxicities were haematologic. There were three treatment-related deaths in the first study but none in the second.
A systematic review (including searches of Medline, American Society of Clinical Oncology annual meeting abstracts [2007][2008][2009][2010][2011][2012][2013], and reference lists from related reviews [20][21][22]) identified 12 randomised trials of novel treatments versus BEP in intermediate-or poor-prognosis patients (Table 4). Even the largest of these was only powered to detect an absolute PFS benefit of 15% [3]; several failed to recruit the targeted sample size. With the exception of the Genito-Urinary Group of the French Federation of Cancer Centres (GETUG) 13 trial [23], which randomised a subset of poor-risk patients showing inadequate marker decline after 1 Â BEP to continue BEP or switch to a dose-dense regimen, none show clear superiority over BEP. Results of high-dose chemotherapy trials [6,7,24,25] are notably divergent and were most favourable in EORTC 30974 that closed early due to poor accrual. CBOP/BEP compares well, demonstrating the greatest estimated relative PFS benefit over BEP. Both the BEP results and relative benefit are remarkably similar to those attained in EORTC 30974 [7] despite what appears to be a higher treatment-related death rate in the present study. Overall grade 3/4 toxicity rates are not presented for EORTC 30974, but grade 4 neutropenia rates were higher arm for arm than in TE23 (47% vs 24% for BEP; 82% high dose vs 65% CBOP/BEP). Rates of grade 4 leukopenia (11%, 20%) and thrombocytopenia (3%, 12%) across the BEP arms of EORTC trials in intermediate [26] and poor-prognosis disease [7], respectively, are suggestive of increased toxicity associated with more advanced disease even on standard therapy. Nevertheless, these more intensive schedules do carry the risk of increased toxicity, so the risk-benefit ratio may be debatable. Any such debate does need to consider the impact of salvage treatment on cumulative toxicity burden in less intensively treated patients who relapse.
Restricting more intensive treatment to those most in need is desirable. Use of dynamic markers to identify patients who are not responding sufficiently well to [ ( F i g . _ 3 ) T D $ F I G ]  standard therapy, as in GETUG 13, is a promising strategy, although the optimal time point to assess rate of marker decline and the method of intensification is still debatable. A further potential strategy for which there are limited data at present [27][28][29] is dose density, giving BEP every 2 rather than every 3 wk, with G-CSF support.

Conclusions
Improved treatments for poor-prognosis disease that increase cure rate by first-line therapy are needed. At present, there is no accepted alternative to BEP, and to challenge its status as standard therapy requires an adequately powered phase 3 trial. This in turn will need international agreement. Because CBOP/BEP met its preset activity goals and attained PFS rates equivalent or better than other approaches to poorrisk disease, it merits consideration in the development of any such trial. Author contributions: Robert A. Huddart 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.
Critical revision of the manuscript for important intellectual content: Pollock, White, Shamash, Cullen.
Administrative, technical, or material support: Pollock.