Optimal patient selection for yttrium-90 glass plus chemotherapy in the treatment of colorectal liver metastases: additional quality of life, efficacy, and safety analyses from the EPOCH study (2024)

Search for other works by this author on:

Article history

Abstract

Introduction

Yttrium-90 (Y-90) glass microspheres (TheraSphere ) is a brachytherapy which can deliver a tumoricidal radiation dose to liver tumors when administered into the hepatic artery and deposited into tumor-feeding arteries. This procedure is commonly referred to as transarterial radioembolization (TARE). The previously published EPOCH study (evaluating TARE in patients with metastatic colorectal carcinoma of the liver who have progressed on first-line chemotherapy) was a 1:1 randomized, open-label, global, multicenter phase III trial comparing outcomes using glass microsphere transarterial radioembolization combined with chemotherapy (TARE/Chemo arm) vs chemotherapy alone (Chemo arm) as second-line therapy for advanced colorectal liver cancer metastases.¹ The study was positive based on both primary endpoints being met. These included progression-free survival (PFS; hazard ratio (HR): 0.69, 95% CI: 0.54-0.88, 1-sided P = .0013) and hepatic PFS (hPFS; HR: 0.59, 95% CI: 0.46-0.77, 1-sided P < .0001), supporting the efficacy of combination therapy. Secondary efficacy endpoints included overall response rate (ORR), which was superior for TARE/Chemo and disease control rate and overall survival (OS), both of which were similar between the 2 study arms.¹ New information reported herein includes efficacy, safety, and time to deterioration in quality-of-life (QoL) data in 2 newly defined patient subgroups as well as additional outcome and time to deterioration in QoL data in the primary population.

KRAS mutation status, Eastern Cooperative Oncology Group (ECOG) status, and carcinoembryonic antigen (CEA) levels are known prognostic factors for metastatic colorectal cancer (mCRC), but only KRAS was a randomization stratification factor in EPOCH. Due to this limitation, there was an imbalance between the 2 study arms with a greater proportion of patients with better prognostic factors in Chemo arm, namely ECOG 0 (62.4% Chemo and 55.3% TARE/Chemo) and baseline CEA < 35 ng/mL (46.9% Chemo and 42.3% TARE/Chemo). This imbalance could affect outcomes. Analyses using 34 covariates demonstrated that these 3 prognostic factors can impact QoL outcomes (measured as time to deterioration in QoL) and likely contributed to the lack of difference in time to deterioration in QoL seen between treatments in the primary population. Given the importance of QoL when selecting treatment, additional analyses were undertaken to identify patient subgroups demonstrating a QoL benefit with TARE/Chemo, taking the 3 prognostic factors into account and ensuring equitable distribution of patients with KRAS mutant vs KRAS wt, ECOG 0 vs ECOG 1, baseline CEA < 35 ng/mL vs CEA ≥ 35 ng/mL between treatments. Efficacy outcomes in these subgroups, namely PFS, hPFS, time to subsequent therapy, and OS, were also determined. Whether changes in baseline CEA can predict a tumor response in these subgroups and was also explored.

Additional efficacy analyses not previously reported using the primary population are herein reported and include time to deterioration of QoL, duration of response duration of disease control, depth of response, time to progression (TTP), hepatic TTP (hTTP), and time to subsequent therapy.

Combination therapy often increases treatment-related adverse events (TEAEs) compared to single therapy. Given that a higher frequency of grade ≥ 3 TEAEs was previously reported in the combination arm, further analyses were undertaken to explore the safety profile of TARE/Chemo treatment in the primary population in greater detail and reported alongside safety in the 2 subgroups.

Cumulatively, the knowledge gained in the exploratory subanalyses reported herein may help inform optimal patient selection and future clinical trial design for TARE treatment for colorectal liver cancer metastases.

Methods

Study design and participants

EPOCH randomized 428 patients to receive unilobar or same-day bilobar administration of TARE plus standard of care (SOC) Chemo or Chemo (1:1) as second-line colorectal liver cancer metastases therapy.¹ The study design, patient selection criteria, and methodology used in the primary analyses have been previously reported.^1,2 The study was performed under an investigational device exemption granted by the FDA and with protocol and consent form approved by presiding ethics committees. Briefly, consenting patients aged ≥ 18 years with unresectable, RESIST 1.1 measurable unilobar/bilobar colorectal liver cancer metastases able to receive second-line irinotecan or oxaliplatin-based chemotherapy with or without targeted therapies, were eligible if their baseline ECOG status was 0 or 1, baseline bilirubin ≤ 1.2 upper limit normal and baseline albumin ≥ 3.0 g/dL. Additionally, patients with prior liver arterial/radiotherapy, clinically evident ascites, unresolved first-line therapy toxicities, confirmed extrahepatic metastases, or contraindication to any procedures or therapies within the protocol, were excluded.

Identification of subgroups

Subgroups were identified from the primary EPOCH study population using the preplanned multivariate Cox proportional hazard model analysis for time to deterioration in QoL (using the 34 covariates previously used in the primary population subgroup covariate analyses). This analysis identified only 4 patient prognostic factors for time to deterioration in QoL with a 1-sided P < .025 namely, KRAS status, ECOG status, extrahepatic metastases, and baseline CEA level. Of these, KRAS wt, ECOG 0, and baseline CEA < 35 ng/mL demonstrated a trend for improved time to deterioration in QoL in the TARE/Chemo arm (ie, 1-sided P < .05). Using combinations of these 3 prognostic factors, 2 subgroups were identified which had improved time to deterioration in QoL in the combination arm vs Chemo alone. Subgroup A excluded patients with both ECOG 1 and CEA ≥ 35 ng/mL and included 303 patients (71% of the primary study population) which corresponds with 67% of the primary population treated with TARE/Chemo. Subgroup B further excluded patients with KRAS-m status and included 168 patients (39% of the primary population) which corresponds with 36% of the primary population treated with TARE/Chemo.

Outcomes

Time to deterioration in QoL, assessed using the Functional Assessment of Cancer Therapy-Colorectal (FACT-c), was defined as time from randomization to a decrease from baseline in the FACT-c total score of ≥ 7 points or to death from any cause, whichever occurred first. Duration of overall response calculated from the time of first overall response (complete response or partial response) until date of disease progression or death due to any cause, whichever occurred first. Duration of disease control was similarly determined but with the addition of stable disease. Depth of response was calculated using the percent change from baseline to nadir in the sum of longest diameter of target lesions. TTP and hTTP were calculated from randomization to TTP or hepatic progression, respectively. PFS and hPFS were calculated from randomization to TTP (or hepatic progression) or death. In all cases, imaging response or progression was determined by blinded independent committee review and assessed using RECIST 1.1. Time to subsequent therapy was defined as the time from randomization to subsequent liver cancer therapy, which included new systemic or locoregional therapy but excluded de-escalation of anti-cancer treatment, dose changes, or switching a doublet therapy component for a related doublet component. A CEA response was defined as ≥ 25% decrease from the baseline level.

For the subgroups, time to deterioration in QoL, PFS, hPFS, time to subsequent therapy, OS, best percentage change from baseline CEA, and corresponding ORR data in CEA responders are presented. Corresponding data from the inverse of the subgroups (data from patients not included in a subgroup) are also provided. The Safety population (all randomized patients receiving at least 1 administration of TARE or Chemo with patients analyzed according to the treatment they received) was used for all safety analyses. Safety events occurring until disease progression or 30 days after discontinuation of study therapy (whichever occurred first) were additionally computed for each treatment and expressed as events per 100 patient-years—a safety measurement which accounts for the differences in time to event endpoints, ie, longer PFS, resulting from differing lengths of systemic treatment exposure and TEAE collection time between treatments arms. These differences were inherent in the protocol design, which had a longer TEAE collection period for TARE/Chemo (experimental arm) than for SOC Chemo. To better assess the adverse event profile of the 2 treatments, a non-study Medical Oncologist classified the events into 1 of the 7 categories as part of the post hoc analyses, namely; liver-related events, angiography procedure-related events, anti-cancer drug events, gastrointestinal events, hematologic events, and miscellaneous events, using common terminology criteria of adverse events preferred terms.

Statistical analysis

Time-to-event endpoints were compared between study arms using log-rank test and HR and 95% CI were estimated using a Cox proportional hazards model. Kaplan-Meier (KM) plots and estimates were obtained. For CEA subgroup purposes, a 35 ng/mL cutoff for baseline CEA was used since this value represented the median value in the primary population, and medians and/or specific values of CEA, ie, 25 or 50 ng/mL, have previously been used to compare treatment outcomes.^3-5

Results

EPOCH randomized 215 patients in the TARE/Chemo arm and 213 in the Chemo arm between May 2012 and August 2020. Patient disease characteristics confirmed an advanced-stage disease study population; 82% of patients had bilobar disease, 71% had a maximum liver lesion ≥ 40 mm, and 65% of patients had ≥ 6 lesions. Additional analyses in the primary population demonstrated that the median time to deterioration in QoL was identical between the 2 arms (3.8 months; HR = 0.86, 95% CI: 0.65, 1.14; 1-sided P = .1513; Figure 1A), with separation between the 2 study arms occurring beyond the median in the KM curves. Because the primary population was heterogeneous and the treatment groups unbalanced, this observation prompted the question of whether a subgroup of patients existed who had a QoL benefit but whose benefit was not apparent in the primary population analysis.

Subgroups were identified using the pre-specified multivariable Cox proportional hazards model analysis, which showed that KRAS status, ECOG status, baseline CEA level, and extrahepatic metastases were prognostic factors for time to deterioration in QoL (Table 1). However, because extrahepatic metastases (either present or not present) did not demonstrate a trend for improved time to deterioration in QoL, KRAS status, ECOG status, and baseline CEA level were used to identify subgroups of interest. Two subgroups were identified: Subgroup A (N = 303), which excluded patients with both ECOG 1 and CEA ≥ 35 ng/mL, and subgroup B (N = 168) which further excluded patients with KRAS mutant status (Figure 1B). In post hoc analyses, subgroups A and B demonstrated a statistically prolonged median time to deterioration in QoL benefit with TARE/Chemo vs Chemo (Table 2, Figure 1C, D), whereas the primary population showed no difference in the medians between groups (Table 2, Figures 1A and 2). No statistical difference in time to deterioration in QoL for patients excluded from subgroups A and B, ie, inverse subgroups A and B was noted (Figure 2, Supplementary Table S1). Notably, the corresponding Chemo subgroups showed no statistical difference in time to deterioration in QoL for any Chemo subgroup population when compared to the primary population.

Select time-to-event endpoints demonstrated larger magnitude differences for subgroup populations. Subgroup A patients receiving TARE/Chemo (N = 143) had a superior benefit vs patients receiving Chemo (N = 160) in PFS (HR: 0.64, 95% CI = 0.47, 0.87; 1-sided P = .0020), hPFS (HR: 0.53, 95% CI = 0.39, 0.73; 1-sided P < .0001), time to deterioration in QoL (HR: 0.65, 95% CI = 0.46, 0.91; 1-sided P = .0063), and time to subsequent therapy (HR: 0.52. 95% 0.37, 0.74; 1-sided p < 0.0001; Table 2). Subgroup B patients receiving TARE/Chemo (n = 77) had superior benefit vs patients receiving Chemo (n = 91) in PFS (HR: 0.60, 95% CI = 0.39, 0.92; 1-sided P = .0089), hPFS (HR: 0.51, 95% CI = 0.33, 0.79; 1-sided P = .0011), time to deterioration in QoL (HR: 0.48, 95% CI = 0.30, 0.76; 1-sided P = .0008), and time to subsequent therapy (HR: 0.42, 95% CI = 0.27, 0.66; 1-sided P= < .0001). Notably, median PFS, hPFS, and time to deterioration in QoL were numerically greater for TARE/Chemo in both subgroups vs the primary population, with the magnitude of the difference greatest in subgroup B. This trend was less apparent across Chemo groups. Outcome data in the inverse subgroups, demonstrated statistical differences (ie, 1-sided P-value < .025) in PFS, hPFS, and time to subsequent therapy; however, no statistical difference was noted for time to deterioration in QoL (Supplementary Table S2). OS was not significantly different between treatments for the primary population, for either subgroup A or B or for inverse populations. CEA response and the corresponding tumor response in subgroups A and B with baseline CEA > 5 ng/mL are shown (Figure 3, Table 3). For subgroup A, a higher percentage of patients with a ≥ 25% CEA decrease was noted (18.5 percentage point difference (ppd), 95% CI = 0.4, 36.3, 1-sided P = .0223 by continuity adjusted Wald approach) as well as higher overall ORR response with TARE/Chemo (16.0 ppd, 95% CI = 4.0, 33.9; 1-sided P = .0039 by continuity adjusted Wald approach). Similar trends were noted in subgroup B.

Additional pre-specified analyses undertaken in the primary population concluded that the (Figure 1A), duration of overall response was not statistically different for responders with TARE/Chemo (N = 73, median 7.2 months) vs Chemo (N = 45, median 6.6 months; HR = 0.79, 95% CI: 0.48, 1.30; 1-sided P = .178). Duration of disease control was significantly longer with TARE/Chemo (N = 171, median 7.2 months) vs Chemo (N = 155, median 5.8 months; HR = 0.64, 95% CI: 0.49, 0.85; 1-sided P = .0009). Mean depth of response in the TARE/Chemo arm (N = 196) was −25.6%, compared to −13.0% in the Chemo arm (N = 182; mean difference −12.6 percentage points; 95% CI: −18.9, −6.3; 1-sided P = .0001 by 2-sample t-test). Median TTP was significantly longer for TARE/Chemo (median 9.7 months) vs Chemo (median 7.6 months; HR = 0.70, 95% CI: 0.53, 0.91; 1-sided P = .0033), as was hTTP (median 12.5 months with TARE/Chemo vs 7.6 months with Chemo, HR = 0.55, 95% CI: 0.42, 0.73; 1-sided P < .0001; Supplementary Figures S1-S4). Median time to subsequent therapy with TARE/Chemo was 21.0 vs 10.1 months with Chemo (HR = 0.49, 95% CI: 0.37,0.67; 1-sided P < .0001; Figure 4).

Additional safety analyses in the primary population demonstrated that the number of TEAEs in any category and any grade was higher in the Chemo arm than in the TARE/Chemo arm when expressed as events/100 patient-years (Table 4). For grade ≥ 3 TEAEs, the event rate was higher for TARE/Chemo in the primary population (Table 4). Liver events, angiography procedure-related events, gastrointestinal tract events, or hematologic events had a higher rate with TARE/Chemo and in the Chemo arm, and anti-cancer drug events, infection events, and miscellaneous events had higher TEAE event rates/100 patient-years (Table 4). Further details on TEAEs by category are provided in Supplementary Table S1. In the TARE/Chemo Safety population (N = 187), 15 fatal events (8.0%) were reported of which investigator evaluation concluded that 7 (3.7%) were unrelated to therapy, 2 (1.1%) were related to Chemo only, 2 (1.1%) related to all therapies and procedures (hepatic failure and blood bilirubin increase), 1 (0.5%) related to device and Chemo (portal hypertension), and 3 (1.6%) related to the device only (all REILD cases). In the Chemo safety population (N = 207), 5 fatal events (2.4%) were reported with 1 (0.5%) event considered possibly treatment-related (pulmonary embolism).

Safety trends seen in the primary population persisted for both subgroups. The magnitude of the difference in grade ≥ 3 event rate between arms was diminished in subgroup A vs the primary population with subgroup B demonstrating a lower grade ≥ 3 event rate with TARE/Chemo than Chemo alone (Table 4). The breakdown of fatal events by subgroups is found in Supplementary Table S3.

Discussion

EPOCH is the only positive randomized controlled study supporting the use of Y-90 glass plus Chemo to treat advanced colorectal liver metastases.¹ Additional analyses on the primary population presented here, substantiate the previously reported superior (h)PFS and tumor control (but no OS benefit) with TARE/Chemo by demonstrating superior duration of disease control, depth of response, (h)TTP with combination therapy, and reaffirm the durability of the response. Although study randomization was stratified based on KRAS status, unilobar/bilobar involvement, and first-line chemotherapy, and additional prognostic factors were not evenly distributed between the 2 study arms. KRAS status, ECOG status, and baseline CEA levels are known prognostic factors for mCRC and were determined to be prognostic factors for time to deterioration in QoL and led to the identification of 2 subgroups of interest: subgroup A (excludes ECOG 1 and CEA ≥ 35 ng/mL patients) and subgroup B (additionally excludes KRAS mutant patients). Post hoc analyses showed that both subgroups experienced improved time to deterioration in QoL that was not previously noted in the primary population. Quality of life is a critical consideration for cancer patients weighing treatment options, especially in the advanced mCRC population in which QoL is likely to deteriorate over time, especially in post-first-line therapies. Improved time to deterioration in QoL combined with prolonged PFS suggests TARE/Chemo may be an appealing option for some patients. Subgroup analysis demonstrated a larger difference in PFS, hPFS, and time to deterioration in QoL for the TARE/Chemo vs Chemo when the 2 treatment arms were well balanced with respect to these prognostic factors, with subgroup B demonstrating the greatest benefits. Additionally, both subgroups demonstrated improved PFS, hPFS, and time to deterioration in QoL with TARE/Chemo compared to Chemo only and when compared to the primary population, with subgroup B demonstrating the greatest benefits. No OS benefit of TARE/Chemo vs Chemo was demonstrated in either subgroup. These analyses stress the importance of well-balanced patient populations in randomized comparative trials, understanding that there is a practical limitation in the number of stratification factors included at randomization. More importantly, the data help to inform optimal patient selection for TARE/Chemo.

Time to subsequent therapy was consistently longer with TARE/Chemo vs Chemo across the primary population and subgroups. This observation, in addition to the extended PFS and hPFS previously reported, suggests some patients may benefit from extending TTP and delaying initiation of third-line therapy, thereby maintaining downstream treatment options for longer while maintaining QoL with TARE/Chemo. For some patients, the opportunity for a treatment-free interval may be desirable and may offset the potential increased number of adverse events often experienced with combination therapy. Clinical benefit may thus be obtained for patients from integration of TARE into SOC chemotherapy, despite a lack of overall survival advantage (a secondary study endpoint).

A higher TEAE rate is not unexpected with combination treatment, and in this study, the most prevalent events were expected and in line with known TEAEs associated with administration of either/and liver-directed TARE or Chemo or natural disease progression, with many events occurring in low patient numbers. When expressed per 100 patient-years, which takes in consideration differing lengths of systemic treatment exposure and TEAE collection time between treatment arms, the TEAE event rate was greater with Chemo. The grade ≥ 3 TEAE event rate was higher with TARE/Chemo in the primary population. However, the magnitude of the difference between treatment arms decreased in subgroup A and was greater for Chemo in subgroup B.

Careful patient selection remains critical to a positive outcome for patients. In patients with colorectal liver cancer metastases, studies have shown that the overall survival of patients with an ECOG performance status of 0 at radioembolization is prolonged compared to patients who had ECOG 1.^6-8 Additionally, carcinoembryonic antigen (CEA) and tumor volume < 25% of total liver volume have been found to be independent predictors of survival and of early progression.^3,8,9 This study confirms the prognostic value of CEA and its utility in patient selection for TARE treatment. This post hoc analysis demonstrated that KRAS status, baseline ECOG, and baseline CEA levels are important prognostic factors that should be considered when selecting patients who will benefit from TARE. Subgroups A and B demonstrated improved time to deterioration in QoL and efficacy with TARE/Chemo compared to Chemo or compared to the primary population. Time to subsequent therapy was consistently longer with TARE/Chemo vs Chemo not only in the primary population but also in the subgroups and in the inverse subgroups. Improved efficacy is substantiated by a larger percentage of patients with a CEA reduction and a correspondingly larger ORR in the TARE/Chemo group vs Chemo.

Limitations of these analyses include the post hoc identification of subgroups A and B and the analyses based on these subgroups. With future increase in knowledge around the use of multicompartment dosimetry in colorectal liver cancer metastatic disease, optimal dosing with Y-90 glass + Chemo in this population may further refine optimal dosing and patient selection.

In conclusion, select prognostic factors, ie, KRAS status, ECOG status, and baseline CEA ≥ 35 ng/mL, are predictive for time to deterioration in QoL, PFS, hPFS, and time to subsequent therapy in appropriately balanced subgroups treated with TARE/Chemo vs Chemo. Patient subgroups with a balanced prognosis, ie, subgroups A and B, demonstrated prolonged time to deterioration in QoL and further improvement in efficacy parameters (PFS, hPFS, and time to subsequent therapy) for TARE/chemo vs chemo relative to the primary population.

In clinical practice, careful patient selection can lead to improved outcomes when glass Y-90 is added to standard irinotecan or oxaliplatin-based Chemo for second-line treatment of colorectal liver cancer metastases.

Acknowledgments

We thank the patients who participated in the EPOCH trial and their families, the investigators, nurses, and site staff. We also thank the Boston Scientific EPOCH study team members for their key roles in generating these additional analyses.

Author contributions

Conception/design: All authors. Provision of study material or patients: R.S., E.G., E.B., M.L., S.P., W.P.H.. Collection and/or assembly of data: all authors. Data analysis and interpretation: all authors. Manuscript writing and final approval of manuscript: all authors.

Funding

This work was supported by Boston Scientific Corporation. We thank Evelyn Schnuerer, MSc (Boston Scientific Corporation) funded by Boston Scientific for medical writing assistance.

Conflicts of interest

Riad Salem, MD, MBA is a consultant for Eisai, Bard Medical, Cook Medical, Boston Scientific, Sirtex Medical, AstraZeneca, QED Therapeutics, Genentech/Roche, Siemens, and research funding from Boston Scientific. Marnix Lam, MD, PhD is a consultant for Boston Scientific, Terumo, and Quirem Medical. He receives research support from Boston Scientific, Terumo, and Quirem Medical. Kirk D. Fowers, PhD and Eveline Boucher, MD are employees of Boston Scientific. William P. Harris is a consultant for Neo Therma, Eisai, Exelixis, Bristol Myers Squibb, QED Therapeutics, Zymeworks, BD Medical, Merck and receives research funding from ArQule, Exelixis, Halozyme, Bristol Myers Squibb, MedImmune, Agios, Bayer, Merck, BTG, Boston Scientific, Koo Foundation, and Zymeworks.

Data availability

Boston Scientific may share patient-level data from registered clinical trials with qualified healthcare practitioners or academic researchers in response to a formal clinical research proposal, and when the request is in the same therapeutic area as the original study. Clinical trial data may be shared when not in conflict with all other applicable regulations, laws, or Boston Scientific policies and/or written agreements. Data may be provided for regulated and approved product 6 months following manuscript publication and after the posting of the study results on clinicaltrials.gov. If made available, data will be accessible for 12-18 months following the end of the trial. There will be limited data availability for trials prior to January 1, 2018. Boston Scientific will disposition requests consistent with these and other internal company criteria for data sharing. Data-sharing requests can be made at https://www.bostonscientific.com/en-US/data-sharing-requests.html

References

© The Author(s) 2024. Published by Oxford University Press.

Issue Section:

gastrointestinal cancer

Email alerts

Citing articles via

More from Oxford Academic