Background

The RAS protein is a member of the G protein family and involved in signal transduction within the mitogen activated protein kinases (MAPK) pathway. Genetic alterations lead to constitutive activation of the RAS protein with a high oncogenic potential in metastatic colorectal cancer (mCRC) (Benvenuti et al. 2007; Vogelstein et al. 1988). RAS mutations (MUT) are detected in about 50% of all patients (Cunningham et al. 2004; Jonker et al. 2007; Sobrero et al. 2008). Since 2013, RAS wildtype (WT) status is required for the use of anti-EGFR (epidermal growth factor receptor) agents like cetuximab or panitumumab (Douillard et al. 2013; Heinemann et al. 2014).

As EGF receptor inhibition is ineffective because of constitutive oncogenic signalling, (Benvenuti et al. 2007) systemic treatment option in patients with a RAS MUT tumor currently include chemotherapy (fluoropyrimidines, irinotecan, oxaliplatin) with or without anti angiogenic agents for two treatment lines, followed by later-line treatment such as trifluridine/tipiracil and regorafenib (Grothey et al. 2013; Kubicka et al. 2013; Van Cutsem et al. 2018). For maintenance strategies following induction treatment, a combination of fluoropyrimidine and bevacizumab is usually recommended (Goey et al. 2017; Hegewisch-Becker et al. 2015; Van Cutsem et al. 2016).

Unlike anti-EGFR treatment, predictive biomarkers for the use of cytotoxic and anti angiogenic agents are still missing. A comprehensive efficacy analysis of these treatment strategies in RAS MUT tumors is currently not available.

We, therefore, performed a systematic review and meta-analysis of randomized controlled phase III trials with EMA/FDA approved cytotoxic and anti angiogenic agents to evaluate efficacy of the addition of chemotherapeutics and anti angiogenic agents when distinguished for RAS status, treatment line and investigated agents.

Methods

Trial identification

Our search strategy included trial identification by systematic literature review using the following terms: “metastatic colorectal cancer”, “randomized”, “phase III”, “NOT phase II”, “NOT meta”, “NOT pooled”. First search was performed in February 2019 and last search in November 2019. Only trials with available molecular subgroup analysis regarding (K)RAS status (KRAS exon 2–4, NRAS exon 2–4) were included. Hence, we included randomized controlled phase III trials with available subgroup data for (K)RAS status in mCRC evaluating the addition of chemotherapeutic or anti angiogenic treatment to a randomised control arm with FDA/EMA approved agents. As treatment efficacy should be evaluated according to (K)RAS status, trials with anti-EGFR treatment requiring (K)RAS wildtype status (cetuximab, panitumumab) were excluded. Patients with BRAF mutations were excluded from this analysis if indicated.

Following trials were identified in Pubmed, EMBASE, Web Of Science and the Cochrane Central Register of Controlled Trials (CENTRAL): TRIBE (Cremolini et al. 2015), AVG2107g (Hurwitz et al. 2009), FOCUS (Richman et al. 2009), ML22011 (Modest et al. 2018), AGITG MAX (Price et al. 2015), ML18147 (Kubicka et al. 2013), RAISE (Obermannova et al. 2016), VELOUR (Wirapati et al. 2017), CORRECT (Grothey et al. 2013), CONCUR (Li et al. 2015), RECOURSE (Van Cutsem et al. 2018), AIOKRK0207 (Hegewisch-Becker et al. 2015), CAIRO3 (Goey et al. 2017), PRODIGE 9 (Aparicio et al. 2018). Data were based on publications and/or poster presentations at congress meetings.

Trials

TRIBE and ML22011 investigated chemotherapeutic (de-)escalation strategies on bevacizumab based treatment arms in previously untreated mCRC. The MRC FOCUS trial compared 5-fluorouracil monotherapy to the combination regimes irinotecan/5-fluorouracil (IrFU) and oxaliplatin/5-fluorouracil (OxFU) as first-line therapy of mCRC. In AVG2107g and AGITG MAX, bevacizumab was used additionally to chemotherapy in untreated mCRC. All second-line trials (ML18147, RAISE, VELOUR) investigated the role of additional anti angiogenic agents to chemotherapy in previously treated patients with mCRC. CORRECT and CONCUR compared regorafenib vs. placebo treatment in mCRC. The RECOURSE trial compared single-agent chemotherapy with TAS102 to best supportive care. In maintenance, most trials investigated treatment with angiogenic inhibition compared to no treatment (AIOKRK0207: bevacizumab ± fluoropyrimidine vs. no treatment; CAIRO3: capecitabine + bevacizumab; PRODIGE 9: bevacizumab).

Data items, data collection process and summary measures

Retrospective data (hazard ratio with confidence interval) regarding overall survival (OS) and progression-free survival (PFS) were collected to compare outcome of chemotherapeutic and non-chemotherapeutic treatment addition strategies by (K)RAS status in patients with previously untreated and treated mCRC and by treatment lines at trial level. Control arms were used as reference, meaning that hazard ratios smaller than 1 indicated benefit of the addition of the respective drug to the treatment protocol.

Risk of bias in individual studies

To ascertain the validity of eligible randomized trials, two authors (AS, DPM) determined independently the adequacy of trials regarding phase of trial, presence of molecular subgroup analysis and strategies of additional treatment.

Risk of bias across studies

Primary tumor sidedness, microsatellite instability and type of cytotoxic treatment (FOLFIRI or FOLFOX) were not considered in this analysis and might have affected the cumulative evidence.

Planned methods of analysis

Standard error estimates were deduced from the 95% confidence intervals. Meta-analyses and meta-regression analyses based on the log-hazard rate ratios were performed. Random effect meta-regression models were fitted for all trials, for each treatment line (first, second, later and maintenance lines) and treatment addition (chemotherapeutic vs. anti angiogenic therapy). Interaction effect of RAS mutation type (RAS WT vs. RAS MUT tumors) with treatment addition was assessed. Heterogeneity explained by mutation was assessed by a Wald chi-square test. Residual heterogeneity was determined by computing the Cochran’s Q test (chi-square test) and the I2 statistic with its 95% confidence interval. In case of three-armed trials, correlation of 0.5 was added to treatment effects to integrate repeated comparisons of the control group to different experimental treatment arms into results. Data analysis was structured to resolve complexity of different result layers. In a first step, benefit of therapeutic addition vs. control was investigated for all patients regardless of therapy strategy or molecular subgroups. Subsequent analyses across all treatment lines were performed separately for RAS WT patients and for RAS MUT patients, and then for RAS WT vs. MUT patients, respectively. Within the molecular subgroups, we first compared efficacy of therapeutic addition vs. control regardless of substance classes. Then, benefit of chemotherapeutic and anti angiogenic strategies were analyzed in detail. Finally, each treatment line including maintenance was stratified by RAS WT, RAS MUT and RAS WT vs. MUT and analyzed for treatment efficacy. Weight of the trials was respected by number of trial patients. All tests were two-sided and the significance level was set to 0.05. The analyses were performed using R 3.6.1, particularly packages forestplot and metafor.

Results

Study selection

Search terms identified 114 phase III trials in total, of which 60 trials had to be excluded due to anti EGFR treatment (22 trials), testing of substances not approved by FDA/EMA for treatment of metastatic colorectal cancer (22 trials) and trial designs which did not compare an additional anti-neoplastic drug to standard treatment (16 trials). Of these remaining trials, 40 trials did not provide molecular subgroup analyses for (K)RAS status. (Fig. 1).

Fig. 1
figure 1

Workflow of trial identification process; ASCO American Society of Clinical Oncology, ESMO European Society of Medical Oncology

Full size image

Patients

Fourteen trials comprising 10,748 patients were included into the analysis. 6 810 patients (63.3%) were evaluated according to molecular status. (Table 1) Detailed outcome results for each trial in each treatment line according to RAS status were presented in supplementary data. (Online resources 1–4).

Table 1 Trial characteristics of the analyzed randomized controlled trials
Full size table

Effect of additional treatment agent (chemotherapy and/or anti angiogenic agent)

Across all trials the benefit in overall survival (OS) (HR 0.83 (95% CI 0.78–0.89), p < 0.0001, p for heterogeneity = 0.25) and PFS (HR: 0.60 (95% CI 0.54–0.67), p < 0.0001, p for heterogeneity < 0.0001) was significant.

Efficacy analysis in RAS WT vs. MUT tumors across all treatment lines

The benefit in OS with the addition of chemotherapeutic and/or anti angiogenic agents was significantly greater in RAS WT tumors as compared to RAS MUT tumors when all studies were analysed together (p for interaction = 0.003). In detail, the effect of the addition of a chemotherapeutic agent was less pronounced in patients with RAS MUT mCRC (WT: HR = 0.74, 95%CI 0.64–0.87; MUT: HR = 0.89, 95% CI 0.78–1.02), p for interaction = 0.07) and the addition of anti angiogenic treatment was significantly less efficient in RAS MUT compared to WT tumors. Interaction of anti angiogenic treatment and RAS status was significant (WT: HR = 0.78, 95% CI 0.70–0.87; MUT: HR = 0.91, 95%CI 0.82–1.01; p for interaction = 0.039).

Regarding PFS, the effect of addition of chemotherapeutic and/or anti angiogenic agents was comparable in patients with RAS WT and MUT tumors.. However, heterogeneity was significant when analysing all trials and the subsets of additional chemotherapeutic or anti angiogenic agents (p < 0.0001).]. (Figs. 2, 3, Table 2).

Fig. 2
figure 2

a Forest plot of overall treatment effect of chemotherapeutic escalation regarding overall survival for patients with RAS wildtype (WT) and mutated (MUT) tumors b Forest plot of overall treatment effect of escalation by anti angiogenic escalation regarding overall survival for patients with RAS wildtype (WT) and (MUT) tumors. OS overall survival, HR hazard ratio, 95% CI 95% confidence interval, RE random effects model. FOCUS upfront IrFU/OxFU vs. FU. FOCUS 5-FU: sequential 5-FU then IrFU/OxFU vs. FU

Full size image
Fig. 3
figure 3

a Forest plot of overall treatment effect of chemotherapeutic escalation regarding progression free survival for patients with RAS wildtype (WT) and mutated (MUT) tumors b Forest plot of overall treatment effect of escalation by anti angiogenic escalation regarding progression free survival for patients with RAS wildtype (WT) and mutated (MUT) tumors. PFS progression-free survival, HR hazard ratio, 95% CI 95% confidence interval, RE random effects model. FOCUS: upfront IrFU/OxFU vs. FU. FOCUS 5-FU: sequential 5-FU then IrFU/OxFU vs. FU

Full size image
Table 2 Efficacy of escalation vs. non-escalation and escalation strategies for OS and PFS (adjusted for trial effect)
Full size table

Efficacy according to RAS WT or MUT tumors for each treatment line

OS was improved regardless of treatment line in RAS WT patients. In patients with RAS MUT mCRC, the relative improvement of additional treatments was greater in first and later-line treatment, while patients in second-line (p for interaction = 0.07) did not benefit from additional therapy. (Table 3) PFS was improved with the addition of agents in all treatment lines.

Table 3 Efficacy of therapeutic escalation in each treatment line in RAS WT vs. mut tumors
Full size table

Maintenance options did not improve OS, but PFS with a trend towards higher efficacy in patients with RAS WT compared to MUT tumors (p for interaction = 0.066) (Table 3).

Discussion

Our analysis was motivated by the limited evidence regarding the benefit of adding further treatment to standards (control arms) in RAS MUT mCRC. One prior meta-analysis focussed on the benefit of the addition of bevacizumab to first-line treatment and found significantly prolonged PFS but not OS in currently used treatment regimen containing infusional 5-fluoruracil and irinotecan. However, molecular subgroups were not analysed. (Baraniskin et al. 2019) Therefore, we analysed data from fourteen randomized controlled phase III trials with available molecular subgroup data for RAS testing in mCRC across several treatment lines.

Mutations in KRAS and NRAS genes constitutively activate the RAS G protein with a high oncogenic potential in the MAPK signaling pathway. (Benvenuti et al. 2007) Thus, RAS mutations were often associated with worse prognosis of mCRC—both due to different biology and due to lack of anti-EGFR targeted therapy. (Andreyev et al. 2001, 1998; Barault et al. 2008; Cremolini et al. 2015; Hegewisch-Becker et al. 2018; Modest et al. 2016; Richman et al. 2009).

Generally, the addition of chemotherapeutic and/or anti angiogenic agents demonstrated a significant benefit in patients RAS WT and MUT tumors in our meta analysis in terms of OS and PFS. However, in patients with RAS MUT tumors the benefit in OS with the addition of a new agent across all trials and treatment lines was a modest relative risk reduction for death of 12%. Although statistically significant, it might be argued if 12% can be regarded as clinically meaningful improvement. Overall, the addition of agents to the comparators was significantly more effective in patients with RAS WT tumors when compared to RAS MUT tumors in OS and PFS (see Table 2). This finding may suggest that RAS WT mCRC represents a more treatment sensitive entity in as compared to RAS MUT mCRC independently from anti-EGFR antibody therapy.

When studies investigating chemotherapeutic agents were analysed separately, a trend towards limited efficacy was observed in RAS MUT tumors for OS, but not for PFS. The relative risk reduction in RAS MUT tumors in this respective setting was only 11% for OS compared to 26% in patients with RAS WT mCRC. Importantly, OS benefit from anti angiogenic treatment was significantly smaller in patients with RAS MUT tumors as compared to RAS WT tumors (see Table 2). These results overlap with our findings of less meaningful benefit in second-line treatment, as included second-line trials investigated anti angiogenic treatment only. Overactivation of the MAPK signalling pathway was shown to stimulate angiogenesis VEGF-independently and might be a reason for low efficacy of anti angiogenic treatment in patients with RAS MUT tumors. (Mehta and Besner 2007).

With a detailed view on different treatment lines, later-line treatment (as compared to control) improved OS to a greater extent in patients with RAS WT compared to patients with RAS MUT tumors. Although a certain benefit of later-line therapy was also observed RAS MUT mCRC, the hazard ratio for OS was only 0.81 (see Table 3). This limited efficacy in this treatment setting needs to be considered carefully in the context of short observation time (absolute benefit is very moderate) and resulting adverse effects and their impact on quality of life in end-stage cancer patients.

Concerning maintenance therapy, our meta-analysis included only trials that compared bevacizumab or capecitabine plus bevacizumab to best supportive care (BSC). (Aparicio et al. 2018; Goey et al. 2017; Hegewisch-Becker et al. 2015) A significant effect on overall survival was seen in neither RAS WT nor RAS MUT patients, while PFS trended to be improved in RAS WT mCRC. When stratified by substances, addition of anti angiogenic therapy alone did not improve outcome in the maintenance setting, while the combination of capecitabine and bevacizumab improved OS in patients with RAS wildtype tumors. These results might again strengthen the hypothesis of limited benefit of anti angiogenic therapy in patients with RAS MUT mCRC. Therefore, our findings raise the question if maintenance strategies (instead of treatment holidays) should be promoted in patients with RAS MUT tumors. Compared to active therapy, careful observation may provide a more quality of life friendly approach without significant impairment of outcome in patients with RAS MUT mCRC.

With 6 810 patients, our meta- analysis represents the one of the largest analyses in this setting so far and only randomized trials investigating FDA approved drugs were included. However, several limitations need to be mentioned. As no individual patient data were available, published hazard ratios and confidence intervals had to be obtained from data extraction. Additionally, two trials contained old treatment regimen (IFL and IrFU/OxFU, respectively) that are not recommended anymore (Hurwitz et al. 2009; Richman et al. 2009). Our treatment subgroups contained more anti angiogenic-based studies than chemotherapy investigating trials. In particular, data for studies with chemotherapeutic agents beyond first-line therapy are rare (only one further line trial) (Van Cutsem et al. 2018). This clear relation of treatment lines and substance classes might have biased our observation. In AIOKRK0207, outcome was distinguished between between double wildtype mutational status and any mutation only. Therefore, patients with BRAF MUT tumors might have biased AIOKRK0207 results in our analysis, although the number should be limited. As most of our investigated trials did not distinguish for primary tumor side and microsatellite (in)stability, we were not able to conduct side-related subgroup analyses. Furthermore, potential treatment interaction might have occurred, since irinotecan and oxaliplatin were used for cytotoxic treatment. Lastly, significant heterogeneity was observed for PFS evaluation in some sub-analyses.

Summary

In summary, our meta-analyses suggests that the addition of chemotherapeutic and/or anti angiogenic agents optimizes outcome in RAS WT, but not necessarily in RAS MUT mCRC. Treatment efficacy in RAS MUT compared to WT mCRC was significantly less evident with advancing treatment lines. Furthermore, in this analysis, maintenance options improved neither OS nor PFS in patients with RAS MUT tumors. Although anti angiogenic therapy is available irrespective of RAS status, our overall analysis demonstrates meaningful efficacy predominantly in RAS wildtype mCRC.