Effect of Combined Immune Checkpoint Inhibition vs Best Supportive Care Alone in Patients With Advanced Colorectal Cancer

1Colorectal cancer is the second leading cause of cancer death worldwide, accounting for approximately 880 000 deaths in 2018. According to the American Cancer Society, there are expected to be 51 020 deaths from advanced colorectal cancer in 2019 in the United States. For most patients with advanced colorectal cancer, systemic therapy is the main treatment modality with median overall survival (OS) approaching 30 months in clinical trials.,

2Despite recent approval of agents such as trifluridine/tipiracil (TAS-102) and regorafenib, outcomes for patients with advanced colorectal cancer remain poor and new treatments are needed., Monoclonal antibodies (mAb) against programmed death receptor-1 (PD-1), pembrolizumab and nivolumab, have shown considerable activity in patients with advanced colorectal cancer with DNA mismatch repair deficient/microsatellite instability-high (dMMR/MSI-H) tumors.,, However, single-agent immune checkpoint inhibitors have not shown meaningful activities in DNA mismatch repair proficient/microsatellite stable (pMMR/MSS) colorectal cancer. Recently, combined blockade with nivolumab and the cytotoxic T-cell lymphocyte antigen-4 (CTLA-4) antibody, ipilimumab, demonstrated additional benefit compared with nivolumab alone in patients with dMMR/MSI-H tumors.

3Durvalumab (Imfinzi, AstraZeneca) is a selective, high-affinity human IgG1 mAb against programmed death ligand 1 (PD-L1), whereas tremelimumab is a selective human IgG2 mAb against CTLA-4. Durvalumab and tremelimumab can be combined with manageable adverse events. We hypothesized that a combined blockade of PD-L1 and CTLA-4 would provide greater anticancer activity and conducted a randomized phase 2 clinical trial to evaluate the efficacy of durvalumab and tremelimumab in patients with advanced refractory colorectal cancer.

4The trial protocol is available in Supplement 1. Eligible patients (121 men [67.2%] and 59 women [32.8%]; median [range] age, 65 [36-87] years) gave written informed consent; had histologically confirmed adenocarcinoma of the colon or rectum; received all available standard systemic therapies (fluoropyrimidines, oxaliplatin, irinotecan, and bevacizumab if appropriate; cetuximab or panitumumab if RAS wild-type; regorafenib if available); were aged 18 years or older; had adequate hematologic, renal, and liver function; had Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1, and measurable disease according to Response Evaluation Criteria in Solid Tumors (RECIST, version 1.1). Patients were excluded if they received prior mAbs targeting PD-1, PD-L1, or CTLA-4, or had a history of autoimmune disorders or severe immune-mediated toxic effects.

5The study was approved by the institutional review board of each participating center, conducted according to the principles of the Declaration of Helsinki, complied with all applicable regulations, and was registered on ClinicalTrials.gov (NCT02870920).

6Patients were randomized, in a 2:1 ratio, to receive 75 mg of tremelimumab intravenously every 4 weeks for the initial 4 cycles only, durvalumab 1500 mg of intravenously every 4 weeks, and best supportive care (BSC) (the treatment group) or BSC alone. The randomization was dynamically balanced by ECOG performance status (0 or 1), and the site of primary tumor using the method of minimization. Randomization was performed centrally by the Canadian Cancer Trials Group (CCTG) central office. The study was open label, and investigators and patients were not blinded to treatment assignments. No crossover was allowed between treatment groups.

7Patients were evaluated clinically every 4 weeks while on study treatments, and every 12 weeks after disease progression. Radiological assessments with computed tomographic images were performed every 8 weeks until progression. Treatments continued until there was radiological or clinical evidence of disease progression, intolerable toxic effects, withdrawal of consent, or death. Adverse events were collected and classified according to the National Cancer Institute Common Toxicity Criteria for Adverse Events, version 4.0.

8Blood samples for circulating cell-free DNA (cfDNA) were collected prior to study therapy, at 8 weeks, and at the time of disease progression. Baseline samples were analyzed using the GuardantOMNI next generation sequencing 2.15 Mb, 500-gene panel (Guardant Health, Inc) to identify single nucleotide variants (SNVs), indels, fusions, copy number amplifications, MSI-high status, and tumor mutation burden (TMB). Plasma TMB was reported as variations per megabase (vts/Mb) by the GuardantOMNI algorithm, which includes all somatic synonymous and nonsynonymous SNVs and indels excluding germline, clonal hematopoiesis of indeterminate potential (CHIP), driver and resistance variations with statistical adjustment for sample-specific tumor shedding and molecular coverage. Validation of plasma TMB and MSI have been previously described.,

9Quality of life was assessed using EORTC QLQ-C30 at baseline, 4, 8, 12, 16, 24 weeks, then every 12 weeks until deterioration to ECOG PS 4 or death.

10The primary end point was OS, defined as the time from randomization to death from any cause. Secondary end points included progression-free survival ([PFS], the time from randomization to the first objective documentation of disease progression or death from any cause), objective response rate ([ORR], the proportion of patients with a documented complete response [CR] or partial response [PR]), toxic effects, and safety.

11This study was designed to observe 150 deaths to have a power of 80% and a 2-sided α of 10% to detect a 35% reduction in the continuous risk of death. This assumption translated to a hazard ratio (HR) of 0.65, and corresponded to an increase of the median OS from 4.5 months for BSC to 6.9 months for the treatment group. It was calculated that 180 patients had to be enrolled over 18 months and followed for 6 months to observe the required number of deaths.

12Overall survival and PFS were analyzed according to intention-to-treat. They were summarized by Kaplan-Meier method, and compared by a stratified log-rank test adjusting for ECOG performance status and site of the primary tumor. The HRs and 90% CIs were calculated based on a stratified Cox proportional hazard model. A Cochran-Mantel-Haenszel test was used to compare ORR between the 2 study groups adjusting for ECOG Performance Status and site of the primary tumor. Adverse events were analyzed according to treatment received and compared by Fisher exact tests among patients who received at least 1 dose of protocol treatments. Data were analyzed on October 18, 2018, and SAS statistical software (version 9.0; SAS Institute, Inc) was used for analysis.

13In an exploratory analysis, a minimum P value approach was used to derive an optimal threshold for TMB as a predictive biomarker. This derived threshold was assessed as a predictive biomarker for OS benefit using a test of interaction between treatment group and TMB status in a Cox model.

14The CCTG Data and Safety Monitoring Committee regularly evaluated the conduct and safety of the study. The CCTG central office performed randomization, study monitoring, and data verification. Durvalumab and tremelimumab were supplied by MedImmune/AstraZeneca.

15Between August 2016 and June 2017, 180 patients were randomized, with 119 patients assigned to the treatment group and 61 patients to BSC alone. One patient did not receive durvalumab and tremelimumab as assigned; this patient was included in the treatment group for efficacy analysis but not in the safety analysis (Figure 1).

16Baseline demographics are shown in Table 1. There were slightly higher proportions of female and Asian patients in the treatment group. There were no differences for presence of liver metastases or sidedness. All patients had received multiple lines of prior chemotherapy; 180 (100%) received at least 1 prior chemotherapy for advanced disease containing a fluoropyrimidine, 176 (98%) irinotecan, 154 (86%) oxaliplatin, 143 (79%) had received bevacizumab, and 47 (26%) regorafenib; 68 (97%) of patients with RAS wild-type tumors received cetuximab or panitumumab. No patients had received TAS-102 (trifluridine/tipiracil) during the enrollment period of this study.

17Patients in the treatment group received a median of 12 weeks of durvalumab (range, 4-84 weeks) and 12 weeks of tremelimumab (range, 4-24 weeks). Dose reduction or delay as management of adverse events was not permitted per design; however, at least 1 dose omission occurred in 50 (42.4%) patients for durvalumab and 38 (32.2%) for tremelimumab, with the most frequent reasons being investigator decision and hospitalization. In total, 102 (86.4%) and 104 (88.1%) patients received at least 90% of planned durvalumab and tremelimumab doses, respectively.

18After disease progression, 8 (6.8%) and 4 (3.4%) patients in the treatment group received TAS-102 and regorafenib, whereas 3 (4.9%) patients in the BSC group received TAS-102. No patients in either groups received immune checkpoint inhibitors postprogression.

19Of 180 patients enrolled, baseline blood samples were available in 169 patients (93.9%), and cfDNA analysis was successful in 168 (eTable 1 in Supplement 2). There were 2 patients with MSI-H detected in plasma, 1 in each study group. All 11 missing baseline blood samples were for patients from the BSC group, and 9 of 11 were MSS by tissue-based PCR assessment with 2 unknowns owing to a lack of tissue. There was a higher proportion of patients with KRAS mutation in the treatment group based on cfDNA.

20At a median follow-up of 15.2 months (range, 0.16-22.0 months), 154 deaths were observed, with 149 patients (97%) dying due to disease progression. The median OS was 6.6 months for the treatment group (90% CI, 6.0-7.4 months) and 4.1 months for BSC (90% CI, 3.3-6.0 months). The HR for death was 0.72 (90% CI, 0.54-0.97; P =.07) (Figure 2A). In patients with MSS/pMMR tumors, the HR for death was 0.66 (90% CI, 0.48-0.89; P =.02) (Figure 2B).

21The median PFS was 1.8 months in the treatment group (90% CI, 1.8-1.9 months), and 1.9 months in the BSC alone group (90% CI, 1.8-1.9) (eFigure 3 in Supplement 2). The HR for progression was 1.01 (90% CI, 0.76-1.34; P =.97).

22There was no CR. One patient with MSS tumors in the treatment group had a PR that lasted longer than 21 months. Stable disease (SD) as the best response was observed in 26 (21.8%) patients of the treatment group and 4 (6.6%) patients in the BSC group. Disease control (CR, PR, or SD) was seen in 27 (22.7%) and 4 (6.6%) of patients in the treatment group and BSC, respectively (odds ratio, 4.16; 90% CI, 1.40-12.3; P =.006).

23Plasma TMB was available for 165 patients based on baseline cfDNA. Excluding the 2 cases of MSI-H (TMB 74.7 and 247.1 vts/Mb), the median TMB was 15.3 (range, 0.96-85.4) in the treatment group and 20.9 (range, 1.9-114.9) for the BSC group, respectively (P =.07). Patients with TMB of 28 or more (35 of 163 MSS cases [21%]) had the greatest OS benefit (HR, 0.34; 90% CI, 0.18-0.63; P =.004) for the treatment group (interaction P =.07) (eFigures 1 and 2 in Supplement 2). In addition, high TMB was associated with a worse OS in the BSC group (HR, 2.59; 90% CI, 1.46-4.62; P =.007).

24Adverse events were assessed in 118 patients who received at least 1 study treatment and 61 patients in the BSC group. All patients in the treatment group experienced adverse events, and 75 (62%) had at least 1 grade 3 or higher adverse event, whereas 52 (85%) of patients in the BSC group experienced adverse events, and 12 (20%) had at least 1 grade 3 or higher adverse event (P <.001 (Table 2). Incidences of all grades were significantly higher in the treatment group for fatigue, nausea, constipation, insomnia, cough, diarrhea, and cutaneous eruption. However, grade 3 or higher adverse events were only higher for abdominal pain (P =.05) and fatigue (P =.06).

25Incidences of all grades of laboratory abnormalities were significantly higher for lymphopenia, hypoalbuminemia, and hyponatremia in the treatment group, but there was no increase in grade 3 or higher incidences except a borderline higher incidence of lymphopenia. The incidence of TSH elevation of at least 2 × ULN (upper limit of normal) was significantly higher in the treatment group (18% vs 2%, P =.02), and 11 (9%) patients in the treatment group had grade 1 and/or 2 hypothyroidism compared with 1 (2%) in the BSC group. No increased incidences in other immune-mediated adverse events were observed.

26Treatment with durvalumab and tremelimumab did not result in significant deterioration in physical function or global health status at 8 weeks or 16 weeks. Details of quality-of-life analysis will be reported separately.

27One novel aspect of this study is that cfDNA analysis was successfully incorporated into a colorectal cancer study to interrogate novel predictors of immunotherapy benefit. Baseline blood samples were collected for 169 (94%) of 180 patients enrolled, and cfDNA analysis was successful in 168 of 169 patients. The median TMB in patients with MSS tumors was 16.3 (95% CI, 14.4-20.1). Further analysis revealed that the subclonal TMB (defined as those with variant allele frequency <10% of the maximal allele frequency) accounted for a substantial portion of the total TMB. The median clonal TMB was 5.8 (95% CI, 4.8-5.8), similar to the median TMB of 6 (range, 0-361) in MSS advanced colorectal cancer based on next-generation sequencing of primary tumor DNA. In a recent study, George et al reported that 424 of 1934 (21.9%) patients with MSS tumors had TMB ranging from 8.8 to 43.1. The higher TMB observed in this study may be explained by the fact that patients enrolled underwent multiple lines of systemic therapy, leading to clonal evolution and changes in the variational landscapes that may not present in the primary tumor DNA. The rate of KRAS and NRAS variation was 64% and 2% based on tissue analysis, and 78% and 10% based on cfDNA analysis (eTable 1 in Supplement 2). This hypothesis is supported by the observation that the development of resistance to EGFR inhibition is associated with downregulation of mismatch and homologous recombination repair proteins resulting in error-prone DNA repair and increased tumor variagenic ability.

28Tumor variation burden has emerged as a potential biomarker for response to immune checkpoint blockade. High TMB is consistently associated with benefit from immune checkpoint blockade across different types of malignant diseases, such as melanoma, lung, and bladder cancers.,,, A TMB of 28 or more was found to be the optimal threshold as a potential biomarker. Patients with TMB of 28 or more receiving BSC alone had worse OS compared with those with a TMB of less than 28 receiving BSC (median OS, 3.0 vs 5.3 months; P =.007). As seen in eFigure 1A in Supplement 2, the HR fell gradually after a TMB of 20 until reaching a plateau. The statistical power was then gradually eroded above a TMB of 36, by which point most patients fell in the TMB less than 36 group. Similarly, the interaction P value stayed suppressed over an entire range of thresholds above the cut point (eFigure 1B in Supplement 2) until there were limited numbers of patients remaining in the high-TMB group, suggesting this is a real biologic phenomena and not statistical chance.

29When treated with durvalumab plus tremelimumab, patients with a TMB of 28 or higher achieved significant survival improvements compared with those receiving BSC alone (median OS, 5.5 vs 3.0 months; P =.004). It is possible that this improvement was owing to the inferior outcome of patients receiving BSC. An interaction test showed a P value of.07, indicating a possible interaction between TMB and treatment with durvalumab and tremelimumab. Those with TMB of 28 or more represented 35 (21%) patients enrolled in this study. Similar benefits were observed based on clonal and subclonal TMB analysis. Samstein et al reported that the highest 20% of patients with TMB derived better OS from treatment with immune checkpoint inhibitors across different types of cancers. In colorectal cancer, a cutoff value of 52.2 was used, which was the highest among different cancers. The variable responses to immune checkpoint inhibitors in MSI-H colorectal cancer have been attributed to different variational loads among these patients., Peters et al recently presented blood and tissue TMB analysis from a phase 3 study of first-line durvalumab plus tremelimumab vs chemotherapy in patients with metastatic non–small cell lung cancer. Based on cfDNA, 26% patients had TMB of 20 or more. Treatment with durvalumab and tremelimumab was associated with significantly improved PFS and OS compared with chemotherapy in these patients. These data suggest that TMB is a potential biomarker for response to immune checkpoint inhibitors in advanced cancer regardless of MSI status. Genomic analysis of archival tumor tissues and serial blood samples collected while receiving treatment are ongoing.

30Like other studies in advanced refractory colorectal cancer,, PFS was short, 1.8 months in the treatment group and 1.9 months in the BSC group. Grothey et al reported a median PFS of 1.9 months in patients treated with regorafenib and 1.7 months in the placebo group. Mayer et al reported a median PFS of 2.0 months with TAS-102 and 1.7 months in the placebo group. The short PFS in this patient population and the smaller number of patients in this study (180 patients vs 760 and 800 patients in the other 2 studies, respectively) may explain why a significant benefit in OS was seen although there was no apparent improvement in PFS.

31Although single-agent immune checkpoint inhibition has not shown meaningful clinical activity in MSS colorectal cancer, emerging data indicate that combining these agents with others with different mechanisms of action can potentially overcome resistance. Fukuoka et al reported a response rate of 29% in refractory MSS colorectal cancer with regorafenib and nivolumab. Combining atezolizumab with capecitabine and bevacizumab resulted in increased PFS in refractory MSS colorectal cancer. The results of this study further lend support to this strategy.

32Patients in the treatment group experienced more frequent adverse events, including more frequent grade 3 or 4 adverse events. In the treatment group, 75 (64%) patients experienced at least 1 grade 3 or higher adverse event, whereas only 12 (20%) patients experienced at least 1 grade 3 or higher adverse event in the BSC group. No new safety concerns were identified. In addition, there was no significant deterioration in physical function or global health status at 8 weeks or 16 weeks.

33All 11 patients with missing baseline blood samples were from the BSC group. Although the collection of baseline blood samples was mandated per protocol, collection was permitted after randomization but before commencement of study therapy. It is likely that these patients did not wish to undergo blood collection because of disappointments with treatment assignments. Most of these patients (9 of 11) were subsequently confirmed to be MSS based on analysis of archival tissues. The effect of these missing baseline blood samples on interpretation of our study finding is minimal.

34Other important differences between study groups include higher proportions of women (P =.046) and Asians (P =.099) in the treatment group, more patients with BRAF variations (P =.13) and fewer postprogression therapies (P =.25) in the BSC group. These features may favor the treatment group and need to be considered when interpreting our results.

35This randomized phase 2 study suggests that combined PD-L1 and CTLA-4 blockade with durvalumab and tremelimumab may prolong OS in patients with heavily pretreated MSS colorectal cancer. Exploratory analysis suggests that TMB from cfDNA analysis could be a potential biomarker for benefits from immune checkpoint inhibitors. Given the lack of treatment options for this patient population, confirmation studies for combined immune checkpoint inhibitors are warranted.