Capecitabine and Temozolomide versus FOLFIRI in RAS-Mutated, MGMT-Methylated Metastatic Colorectal Cancer

The alkylating agent temozolomide typically induces methyl adducts in the O⁶ position of the guanine, which are directly repaired by O⁶ methyl-guanine methyltransferase (MGMT) enzyme (1). MGMT promoter methylation has been associated with gene silencing and reduced DNA damage response to temozolomide (2). MGMT methylation is a validated positive predictive biomarker and it is associated with extended survival in patients with glioblastoma multiforme treated with radiotherapy and concurrent plus adjuvant temozolomide (3). Also, MGMT methylation is observed in up to 40% of colorectal cancers and has been associated with increase of G>A transitions in several cancer-associated genes such as RAS and TP53 (4). Therefore, the prevalence of MGMT methylation is significantly increased in the patients' subgroup with KRAS mutations, reaching an incidence of about 50%, which is further increased in presence of specific G>A transitions of KRAS gene (5).

In the setting of metastatic colorectal cancer (mCRC), the average response rate to single-agent temozolomide was less than 10% in nonrandomized trials carried out in patients with chemorefractory disease and MGMT methylation detected by qualitative assays, for example, methylation-specific PCR (MSP; refs. 6–11).

Because cross-resistance to several chemotherapeutic agents including temozolomide may have developed in heavily pretreated patients (12), the efficacy of temozolomide may be improved by its use in earlier treatment lines and/or its combination with other active agents, thanks to a synergic effect and to potentially retained chemosensitivity. Moreover, the molecular selection of patients for temozolomide treatment may be refined by the implementation of additional biomarkers beyond MGMT testing by means of MSP. In particular, the low specificity of MSP may be overcome by its integration with MGMT protein expression data and/or MGMT methylation as detected by quantitative assays. Consistently, we showed that lack of MGMT expression assessed by IHC and high MGMT methylation percentage quantitatively assessed by MethylBEAMing (MB) are positive prognostic factors in temozolomide-treated patients with MSP-detected MGMT methylation (13, 14).

Here, we aimed at investigating whether second-line treatment with capecitabine plus temozolomide (CAPTEM regimen) was superior to standard FOLFIRI regimen in patients with RAS-mutated, MGMT-methylated mCRC, after failure of one previous oxaliplatin-based treatment.

This study (ClinicalTrials.gov identifier: NCT02414009) was a multicenter, open-label, randomized, phase II trial conducted at 18 Italian centers.

Inclusion criteria were as follows: histologically or cytologically confirmed colorectal adenocarcinoma; age of 18 years or more; ECOG performance status of 0–1; locally assessed RAS-mutated status coupled with MGMT methylation centrally confirmed by means of MSP; RECIST-defined disease progression on or after a previous oxaliplatin-based regimen with or without bevacizumab, which had to be administered for at least 3 months (previous adjuvant oxaliplatin-based chemotherapy was allowed if less than 6 months had elapsed from the end of treatment and radiological disease relapse); measurable lesions according to RECIST criteria v1.1; and adequate bone marrow, liver, and renal function (neutrophils ≥ 1,500/μL, platelet count ≥ 100,000/μL, hemoglobin ≥ 9.0 g/dL, albumin ≥ 2.5 g/dL, total bilirubin, AST, ALT ≤ 1.5 × the upper limit of normal (ULN), or in case of liver metastases AST and ALT ≤ 5 × ULN, and alkaline phosphatase ≤ 2.5 × ULN or in case of liver or bone metastases ≤ 5 × ULN, creatinine ≤ 1.5 × ULN, or creatinine clearance ≥ 50 mL/min, INR and aPTT ≤ 1.5 × ULN). We excluded patients who had received more than one prior treatment line for advanced disease, previous irinotecan, or temozolomide therapy. Other exclusion criteria were as follows: known clinically significant dihydropyrimidine dehydrogenase deficiency; presence of Gilbert syndrome based on total and indirect bilirubin; clinically relevant cardiovascular disease; concurrent active malignancies excluding those disease-free for more than 3 years, or other significant comorbidities that could affect patients' outcome; inability to take oral medications; pregnancy or lactation or being of child-bearing potential; not using or not willing to use medically approved contraception. The study was conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent. Institutional review board and Ethics committee approval was obtained from all participating Centers. All the patients provided written informed consent before any study-related procedures.

During molecular prescreening, archival tumor tissue was centrally collected at the Coordinating Center and MSP was used to assess MGMT promoter methylation status, as described previously (7). Patients with centrally confirmed MGMT methylation were considered eligible and randomly assigned (1:1) to CAPTEM (ARM A) or FOLFIRI (ARM B). Treatment group allocation was done with a minimization algorithm implementing a random component. Stratification factors were time-to-progression of previous oxaliplatin-based chemotherapy (< vs ≥9 months) and prior bevacizumab in combination with oxaliplatin-based chemotherapy (yes vs no).

The schedule of CAPTEM regimen consisted of oral capecitabine 750 mg/sqm twice daily from days 1 to 14 every 28 days plus temozolomide 75 mg/sqm twice daily from days 10 to 14 every 28 days. The schedule of FOLFIRI regimen consisted of irinotecan 180 mg/sqm i.v. over 60 minutes on day 1, leucovorin 200 mg/sqm i.v. over 120 minutes on days 1 and 2, followed by 5-fluorouracil (5-FU) 400 mg/sqm i.v. bolus and then 5-FU 600 mg/sqm administered as a continuous intravenous infusion over 22 ± 2 hours, both on days 1 and 2, every two weeks. The use of supportive measures to treat specific AEs was allowed in both arms, whereas prophylactic treatment was mandated for vomiting according to guidelines. The two treatments were continued until progressive disease (PD), consent withdrawal, unacceptable toxicity, or death.

Disease evaluation was performed by thorax and abdominal CT or MRI scans at screening and every 8 weeks thereafter until PD, consent withdrawal, or death. Tumor response was classified according to RECIST v1.1 criteria as complete response (CR), partial response (PR), stable disease (SD) or PD. Patients who discontinued the study treatment without PD had tumor assessments every 8 weeks until PD or study withdrawal. Treatment safety was assessed according to NCI CTCAE vers.4.0. Quality of life (QoL) was investigated through European Organization for Research and Treatment of Cancer (EORTC) QLQ-C30 (version 3.0) and Functional Assessment of Cancer Therapy-Colorectal (FACT-C; Version 4.0) questionnaires at baseline and every 8 weeks until PD, consent withdrawal, or death. Archival tumor tissue samples were centrally collected and assessed for MGMT status. MGMT expression was assessed by IHC and categorized as negative or positive, and MGMT quantitative methylation was measured by MB using correction for tumor content (9, 14) and reported according the previously identified cutoff of 63% (13).

Activity and efficacy analyses [progression-free survival (PFS), overall survival (OS), overall response rate (ORR), disease control rate (DCR)] were performed on the intention-to-treat population (all randomized patients). Safety analyses were done in all patients who received at least 1 cycle of induction treatment. No interim analysis of the efficacy and safety data was planned. The sample size was calculated on the basis of median PFS. Taking into account the median PFS of 2.5 months reported by the GERCOR trial with the use of FOLFIRI in the second-line setting (15), a sample size of 82 patients, 41 per arm, recruited over a 2-year period, would achieve 90% power to detect an increase of median PFS from 2 months in arm B to 4 months in arm A, with a one-sided significance level of 0.05.

OS and PFS curves and median PFS and OS (including corresponding 95% CIs) were estimated in the two treatment arms by the Kaplan–Meier method; the curves were compared using the log-rank test. Exact (binomial) two-sided 95% CIs were calculated for ORR and DCR in the two arms. Univariate Cox models were fitted to estimate HRs. The subgroup analyses according to stratification factors and other prognostic variables were performed by estimating the OS and PFS Kaplan–Meier curves according to treatment arm in each subgroup. Univariate Cox models were fitted to estimate subgroup HRs and corresponding 95% CIs and interaction test between treatment and each subgroup variable was calculated.

The median follow-up time was calculated by the reverse Kaplan–Meier approach.

All tests were performed two-sided at a significance level of α = 0.05, with the exception of the one-sided test of superiority for PFS. The analyses were carried out using the SAS (version 9.1) and R software. QoL analysis was performed using SPSS for Windows (version 25.0).

Between November 6, 2014 and May 10, 2019, a total of 155 patients were screened for MGMT methylation status and 86 molecularly eligible patients were randomized to arm A (n = 43) or arm B (n = 43). Patients' demographics and disease characteristics were well balanced in the two arms (Table 1). In the medical history of randomized patients, there was an absolute or relative contraindication to the use of bevacizumab in 27 (31%) patients, unacceptable or unresolved cardiovascular adverse events (AEs) after first-line use in 20 (23%), primary refractoriness to bevacizumab-based treatment (PD within 3 months from its start) in 15 (17%). One patient was randomized but did not receive study treatment. The median treatment duration was 2.8 months [interquartile range (IQR), 1.1–4.9] for arm A and 2.8 months (IQR 1.7–5.6) for arm B. The main reasons for treatment discontinuation were PD in 73 (86%) patients, AEs in 5 (6%), and patient/investigator decision in 5 (6%; Fig. 1).

The data cut-off date for the analyses was July 30, 2019. At the time of analysis 79 PFS events were observed, 40 in arm A and 39 in arm B. The median follow-up duration was 30.5 months (IQR, 12.2–36.3). CAPTEM regimen failed to show superiority over to FOLFIRI, because the log-rank, one-sided P value was equal to 0.223. Median PFS was 3.5 months (95% CI: 2.0–5.0) in arm A versus 3.5 months (95% CI: 2.3–6.1) in arm B (HR = 1.19; 90% CI: 0.82–1.72; Supplementary Table S1; Fig. 2A). The overall number of deaths was 59, 28 in arm A, and 31 in arm B. Median OS was 9.5 months (95% CI: 8.2–25.8) in arm A and 10.6 months (95% CI: 8.5–20.8) in arm B (HR = 0.97; 95% CI: 0.58–1.61; P = 0.893; Supplementary Table S1; Fig. 2B).

The forest plots for PFS and OS showing the treatment HR and 95% CI according to stratification factors and other prognostic variables are shown in Supplementary Figure S1 and Supplementary Figure S2, respectively.

There was no significant difference between arms in terms of ORR and DCR (Supplementary Table S1). ORR was 11.6% (95% CI: 3.9–25.1) in arm A, and 11.6% (3.9–25.1) in arm B; the OR for ORR in arm A versus B was 1.03 (0.27–3.89; P = 0.966). DCR was 53.5% (37.7–68.8) in arm A, and 53.5% (37.7–68.8) in arm B; the OR for DCR in arm A versus B was 1.06 (0.43–2.57; P = 0.903). In the 153 patients achieving disease control, median duration of disease control was 10.9 months (IQR, 5.8–21.0) in arm A, and 9.0 months (5.9–14.7) in arm B (P = 0.156; Supplementary Table S1). The waterfall plots depicting best tumor response according to RECIST1.1 in both arms are depicted in Supplementary Figure S3. The postprogression therapies according to treatment arm are summarized in Supplementary Table S2.

In the safety population of 85 patients, any grade AEs were reported in 35 of 43 (81.4%) and 40 of 42 (95.2%) patients in arm A and B, while grade 3/4 AEs were observed in 7 of 43 (16.3%) and 20 of 42 (47.6%) patients in arm A and B, respectively. No treatment-related deaths or unexpected drug-related AEs were recorded. As shown in Table 2, there was an increased incidence of anemia, neutropenia, diarrhea, stomatitis, and fatigue in arm B, while an increased incidence of thrombocytopenia and vomiting was reported in arm A. A delay in treatment administration for drug-related AEs was necessary in nine of 43 (20.9%) patients treated in arm A and 22 of 42 (52.4%) patients in arm B, respectively. Permanent treatment discontinuation due to AEs was reported in one (2.3%) and four (9.5%) patients in arm A and B, respectively. Detailed data on dose intensity, dose reductions, or delays according to treatment arm are reported in the Supplementary Appendix S1 in Supplementary Data.

The analysis of QLQ-30 and FACT-C questionnaires showed a significantly better QoL in CAPTEM arm (Supplementary Appendix S2 in Supplementary Data and Supplementary Figures S4, S5, S6, and S7).

Regarding exploratory biomarkers, negative and positive MGMT IHC expression was observed in 50 (58%) and 36 (42%) out of 86 tumors. MGMT methylation assessed by MB and classified using the previously validated cutoff (12) was ≥63% in 25 (31%) and <63% in 56 (69%) out of 81 evaluable samples. Results on the predictive role of MGMT IHC and MB according to the two treatment arms are depicted in Fig. 3 (PFS and OS) and summarized in Table 3 (ORR, DCR, PFS, and OS). The waterfall plots depicting best tumor response according to MGMT IHC and MB are depicted in Supplementary Figure S8.

Currently available second-line treatment options of patients with mCRC achieve unsatisfactory outcomes and the drug development process in this setting is extremely challenging. After failure of a prior oxaliplatin-based regimen, tumor response to FOLFIRI regimen is modest and ranges from 4% to 11% (15, 16). Moreover, the continuation of antiangiogenic agents added to the doublet chemotherapy backbone has provided a modest, albeit significant, survival gain of less than 2 months after the first progression to bevacizumab-based first-line regimens (17). The limited therapeutic options available in the second-line setting represent a particularly relevant unmet need in patients with RAS-mutated tumors, whose treatment mainstay is still mostly based on chemotherapy combinations and their sequencing, because of the lack of effective targeted strategies in this patients' subgroup. On the other hand, in RAS wild-type populations, several oncogenic drivers—such as EGFR pathway functional activation, BRAF mutations, HER2 amplification—are matched with specific targeted strategies according to major guidelines (18–20).

MGMT promoter hypermethylation is found more frequently among RAS-mutated colorectal cancers (4, 5) and is associated with specific KRAS mutational profiles (such as G12/13D, G12S, and G12A) derived from G>A transitions, which may be used to guide patients' screening.

MGMT methylation has been used in nonrandomized trials as a biomarker for the selection of patients with mCRC for treatment with methyltriazen-imidazole-carboxamide prodrugs dacarbazine and oral temozolomide. The limited single-agent activity of these drugs in the chemorefractory setting and the potential improvement of their efficacy when used in earlier treatment lines and/or in combination with other active agents represented the rationale for the design of temozolomide-based combination trials. For instance, in pretreated patients with irinotecan-sensitive disease (i.e., irinotecan-free interval >3 months), MGMT promoter methylation and MSS status, the combination of temozolomide with irinotecan (TEMIRI regimen) achieved a promising ORR of 24%—thanks to a potential synergy between the two agents (21). The CAPTEM regimen has been previously investigated in patients with metastatic, low-grade neuroendocrine tumors and showed manageable toxicity profile and promising efficacy (22). Synergism between 5-fluorouracil and temozolomide was reported in preclinical models of neuroendocrine tumors and a mechanistic basis for this effect has been proposed (23). Both capecitabine and temozolomide induce deoxythymidine triphosphate depletion resulting, respectively, from thymidylate synthase inhibition and repetitive cycles of excision and synthesis mediated by MMR proteins upon O⁶ methyl-guanine–thymine mispairs (23, 24). Thymidine pool depletion might induce DNA double-strand breaks and eventually apoptosis in rapidly dividing cells. However, it should be pointed out that other potential ways to increase the efficacy of temozolomide may be represented by dose-dense or metronomic schedules, which may be able to induce MGMT depletion in cancer cells (9).

Despite the clear clinical and biological rationale, this randomized phase II trial failed to show the PFS superiority of CAPTEM compared with FOLFIRI as second-line treatment in patients with RAS-mutated mCRC bearing MGMT methylation detected by MSP. Although the two treatment regimens displayed clinically comparable efficacy and activity, the safety profile of CAPTEM was more favorable compared with FOLFIRI, based on a 3-fold increase of grade 3/4 treatment-related AEs, which occurred in arm B versus A. Quality of life was significantly deteriorated in patients treated with FOLFIRI, as compared with those receiving CAPTEM, according to global QoL score, functional scales and individual symptoms. We acknowledge that the total dose of fluoropyrimidine used in the CAPTEM regimen was selected on the basis of previous studies in neuroendocrine tumors and was relatively lower (capecitabine 1,500 mg/sqm/day for 14 days every 4 weeks) than in the standard biweekly FOLFIRI arm. Moreover, the FOLFIRI regimen in this study adopted the use of 5-FU bolus for two consecutive days and may have increased the rate of diarrhea as compared with what is reported in the literature when using 5-FU bolus only in day (15, 16, 25). However, we assume that the worse tolerability of FOLFIRI reported in the study was mostly related to the class-specific toxicity profile of irinotecan. In addition, the fully oral and monthly schedule of CAPTEM, not requiring continuous infusion, may significantly impact the manageability, compliance, and costs compared with FOLFIRI regimen.

Several post hoc analyses of phase II nonrandomized studies showed that loss of MGMT expression by IHC or mass spectrometry, as well as elevated MGMT methylation % measured by MB are prognostic for higher ORR and PFS in patients with mCRC with MSP-detected MGMT methylation and treated with single-agent temozolomide (13, 14, 26). However, the absence of a temozolomide-free control arm did not allow to generate hypotheses on the predictive value of such assays. Here, we exploratorily assessed the potential predictive role of MGMT protein expression and MGMT gene quantitative methylation with IHC and MB, respectively, showing that patients with positive MGMT expression by IHC do not seem to derive benefit from temozolomide-based treatment. However, the interaction test was statistically significant only when exploring the impact of MGMT expression by IHC on DCR, whereas the lack of significant interaction for ORR, PFS, and OS may be related to the small sample size. Therefore, prospective trials with patients' selection based on the implementation of IHC are warranted, even if IHC has not been validated in glioblastoma due to several limitations linked to operator sensitivity, poor concordance with methylation assays, and low tumor specificity due to MGMT staining in nonneoplastic cells. Moreover, further investigations on methylated circulating tumor DNA (27) may also be a useful real-time biomarker, which may allow detecting MGMT methylation longitudinal changes over the natural history of the disease and during treatments (10).

Collectively, these data suggest that MSP alone has still inadequate specificity to predict the benefit from temozolomide and that additional MGMT-centered analyses, or other biomarkers related to DNA-damage response [such as base excision repair and homologous recombination pathways (2)] should be further explored to improve the molecular hyperselection of patients with mCRC for temozolomide-based therapy beyond MSP. The investigation on predictive biomarkers of response to temozolomide in mCRC is particularly important because a hypermutant status and mutations in mismatch repair genes may emerge at the time of acquired resistance to temozolomide in initially responsive mCRC preclinical models and patients (28). On the basis of these findings, ongoing trials are aimed at assessing the role of temozolomide as a “priming” therapy for the sequential or concomitant use of immune checkpoint inhibitors in MSS, MGMT silenced mCRC (NCT03832621, NCT03519412, NCT03879811).

The main limitation of this trial refers to the lack of antiangiogenic agents (bevacizumab or aflibercept) combined with chemotherapy in both arms of treatments, as antiangiogenic blockade beyond progression to bevacizumab containing upfront regimens is recommended as a standard of care (17). The FOLFIRI regimen in this study adopted the use of 5-FU bolus for 2 consecutive days and is not commonly used in the current clinical practice. Above all, this study was a phase II trial with a limited sample size, which prevented drawing robust evidences.

In conclusion, temozolomide and capecitabine combination had not superior efficacy compared with a standard second-line chemotherapy option for patients with RAS-mutated and MGMT-methylated mCRC. Further studies are needed to refine the molecular selection for temozolomide treatment and to assess the potential role of this agent in the most up-to-date treatment scenario, preferentially from the third-line treatment setting.

F. Pietrantonio reports receiving speakers bureau honoraria from Amgen, Roche, Sanofi, Bayer, Servier, Merck-Serono, and Lilly. S. Lonardi reports receiving speakers bureau honoraria from Roche, Merck Serono, Lilly, Servier, and Bristol-Myers Squibb, and is an advisory board member/unpaid consultant for Amgen, Merck Serono, and Lilly. F. Morano reports receiving speakers bureau honoraria from Servier. L. Rimassa reports receiving other commercial research support from AbbVie, Amgen, ArQule, AstraZeneca, Baxter, Bayer, Celgene, Eisai, Exelixis, Gilead, Hengrui, Incyte, Ipsen, Italfarmaco, Lilly, MSD, Roche, Sanofi, and Sirtex Medical. A. Sartore-Bianchi reports receiving speakers bureau honoraria from and is an advisory board member/unpaid consultant for Amgen, Bayer, and Sanofi. F. de Braud is a paid consultant for Amgen, Ignyta, Roche, and Bristol-Myers Squibb, and reports receiving speakers bureau honoraria from Bristol-Myers Squibb and Servier. No potential conflicts of interest were disclosed by the other authors.

Conception and design: F. Pietrantonio, M.D. Bartolomeo, F. de Braud

Development of methodology: F. Pietrantonio, L. Barault, M.D. Bartolomeo, F. de Braud

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Pietrantonio, R. Lobefaro, M. Antista, S. Lonardi, F. Morano, S. Mosconi, L. Rimassa, S. Murgioni, A. Sartore-Bianchi, G. Tomasello, R. Longarini, G. Farina, S. Gori, G. Randon, S. Corallo, V. Guarini, A. Martinetti, M. Macagno, L. Barault, F. Perrone, E. Tamborini, G. Fucà, M.D. Bartolomeo, F. de Braud

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Pietrantonio, R. Lobefaro, M. Antista, A. Raimondi, G. Randon, V. Guarini, M. Macagno, L. Barault, F. Perrone, E. Tamborini, F.D. Nicolantonio, M.D. Maio, G. Fucà

Writing, review, and/or revision of the manuscript: F. Pietrantonio, R. Lobefaro, M. Antista, S. Lonardi, A. Raimondi, L. Rimassa, G. Tomasello, R. Longarini, G. Farina, S. Gori, G. Randon, F. Pagani, A. Martinetti, M. Macagno, L. Barault, F.D. Nicolantonio, M.D. Maio, G. Fucà, M.D. Bartolomeo, F. de Braud

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Pietrantonio, R. Lobefaro, M. Antista, G. Randon, F. Palermo, L. Barault

Study supervision: F. Pietrantonio, M. Antista, F. Petrelli, M. Milione, F.D. Nicolantonio, F. de Braud

The authors thank all the patients who agreed to take part in the trial. The authors also thank the investigators and the study teams who participated. A temozolomide drug supply was provided by Sunpharma. This research has received funding from FONDAZIONE AIRC under AIRC IG 2018 ID. 21407 project (to principal investigator, F. Di Nicolantonio). This study was also partially funded by Fondazione Piemontese per la Ricerca sul Cancro-ONLUS 5 per mille 2015 Ministero della Salute Project “STRATEGY” (to F. Di Nicolantonio).

Maria Giulia Zampino, Gastrointestinal Unit, Istituto Europeo di Oncologia, Milan, Italy; Daniele Fagnani, Department of Medical Oncology, ASST di Vimercate, Vimercate, Italy; Vincenzo Adamo, Medical Oncology Unit, A.O. Papardo & Department of Human Pathology, University of Messina, Messina, Italy; Rosa Berenato, Medical Oncology, Unit A.O. Papardo, Messina, Italy; Valeria Smiroldo, Medical Oncology and Hematology Unit, Humanitas Cancer Center, Humanitas Clinical and Research Center – IRCCS, Rozzano, Milano, Italy.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.