Does the Chemotherapy Backbone Impact on the Efficacy of Targeted Agents in Metastatic Colorectal Cancer? A Systematic Review and Meta-Analysis of the Literature

Importance The EGFR inhibitors (EGFR-I) cetuximab and panitumumab and the angiogenesis inhibitors (AIs) bevacizumab and aflibercept have demonstrated varying efficacy in mCRC. Objective To document the overall impact of specific chemotherapy regimens on the efficacy of targeted agents in treating patients with mCRC. Data sources: MEDLINE, EMBASE and Cochrane databases were searched to 2014, supplemented by hand-searching ASCO/ESMO conference abstracts. Study Selection Published RCTs of patients with histologically confirmed mCRC were included if they investigated either 1) chemotherapy with or without a biological agent or 2) different chemotherapy regimens with the same biological agent. EGFR-I trials were restricted to KRAS exon 2 wild-type (WT) populations. Data Extraction and Synthesis Data were independently abstracted by two authors and trial quality assessed according to Cochrane criteria. The primary outcome was overall survival with secondary endpoints progression free survival (PFS), overall response rate (ORR) and toxicity. Results EGFR-I added to irinotecan-based chemotherapy modestly improved OS with HR 0.90 (95% CI 0.81–1.00, p = 0.04), but more so PFS with HR 0.77 (95% CI 0.69–0.86, p<0.00001). No benefit was evident for EGFR-I added to oxaliplatin-based chemotherapy (OS HR 0.97 (95% CI 0.87–1.09) and PFS HR 0.92 (95% CI 0.83–1.02)). Significant oxaliplatin-irinotecan subgroup interactions were present for PFS with I2 = 82%, p = 0.02. Further analyses of oxaliplatin+EGFR-I trials showed greater efficacy with infusional 5FU regimens (PFS HR 0.82, 95% CI 0.72–0.94) compared to capecitabine (HR 1.09; 95% CI 0.91–1.30) and bolus 5FU (HR 1.07; 95% CI 0.79–1.45); subgroup interaction was present with I2 = 72%, p = 0.03. The oxaliplatin-irinotecan interaction was not evident for infusional 5FU regimens. For AIs, OS benefit was observed with both oxaliplatin-based (HR 0.83) and irinotecan-based (HR 0.77) regimens without significant subgroup interactions. Oxaliplatin+AI trials showed no subgroup interactions by type of FP, whilst an interaction was present for irinotecan+AI trials although aflibercept was only used with infusional FP (I2 = 89.7%, p = 0.002). Conclusion and Relevance The addition of EGFR-I to irinotecan-based chemotherapy has consistent efficacy, regardless of FP regimen, whereas EGFR-I and oxaliplatin-based regimens were most active with infusional 5FU. No such differential activity was observed with the varying chemotherapy schedules when combined with AIs.


Search strategy Eligibility criteria
Published randomized controlled trials of any language or year were eligible for inclusion. Participants included were patients with metastatic (or advanced, unresectable) colorectal cancer.
Interventions studied were EGFR-I or AIs. EGFR-I trials were restricted to KRAS exon 2 wild-type (WT) populations. Eligible comparisons were 1) chemotherapy with biological agent versus chemotherapy alone or 2) different chemotherapy regimens with the same biological agent.
Search results were evaluated independently by two authors (DC, NP/ES), with disagreements in eligibility resolved by consensus after reference to the full text of the article. Data was extracted into piloted forms and double-checked by another author to ensure accuracy.

Endpoints
The primary endpoint was overall survival (OS); secondary endpoints were progression free survival (PFS), overall response rate (ORR) and toxicity. Quality of life (QoL) data was extracted where available.
Other data extracted included PICOS, the quality/description of randomization, and any relevant funding sources. Risk of bias was performed at the study level, using the Cochrane risk of bias tool, with summary risk of bias as per Cochrane recommendations.
The principal summary measures were hazard ratio (HR) for OS/PFS and odds ratios for ORR and toxicity. Meta-analysis was carried out using the generic inverse variant method, with fixed-effects analysis and calculation of HR/OR as applicable with 95% confidence intervals (CI).
Trials were characterized by type of biologic and chemotherapy backbone. The two groups of biological therapy investigated were: 1. EGFR-I: with oxaliplatin (ox) backbone vs with irinotecan (iri) backbone.
2. AIs: with ox backbone vs with iri backbone vs FP alone.
Subgroup analysis was performed by type of FP: capecitabine, infusional or bolus. The mIFL regimen was considered in the bolus group.
Given the increasing literature on the improved efficacy of EGFR-I in extended RAS settings, we performed additional analysis for OS in trials that reported this outcome in extended RAS wildtype populations.
Heterogeneity was explored when I 2 >50% and p<0.10. Sensitivity analyses and funnel plots were undertaken to investigate possible bias.

Study selection
The literature search identified 256 potentially eligible citations from 2827 search results. Thirty-nine papers representing 23 studies comprising 10478 patients were eligible for inclusion (Table 1, Fig 1). The EPIC trial [30] was excluded, as analysis by KRAS exon 2 status was available for only 300/1298 patients, with incomplete OS and PFS data. Upon clarification with the lead author, we confirmed that insufficient data was currently available to enable metaanalysis and that there were no active plans for this analysis to be undertaken in the future. The PEAK trial, comparing FOLFOX + cetuximab to FOLFOX + bevacizumab in the first-line setting, was not included in quantitative analysis because it did not investigate the activity of either cetuximab or bevacizumab alone in addition to chemotherapy but rather compared its effects. Furthermore, both arms received the same chemotherapy backbone, meaning that it does not address the research question posed. The other studies comparing anti-EGFR to antiangiogenesis agents with the same backbone (SPIRITT, FIRE-3) are excluded for the same reason. The Ye study [20] (investigating the addition of cetuximab to FOLFOX/FOLFIRI) met the set requirements, but was excluded from analysis as no results were available separately for the FOLFOX and FOLFIRI arms. PACCE and CAIRO2 were excluded given that both arms contained at least one biological agent (bevacizumab).

Risk of Bias
The overall quality of the studies was good (Table 1) 1. The effect of chemotherapy partner on efficacy of EGFR-I. 1.1 Oxaliplatin backbone + EGFR-I. Five studies (COIN [12], OPUS [2], PRIME [21], NEW EPOC [27] and NORDIC VII [18]), involving 2061 patients, investigated the addition of EGFR-I to oxaliplatin-based  Fig 4), with the infusional 5FU group demonstrating a PFS benefit (HR 0.82 (95% CI 0.72-0.94)) in contrast to the capecitabine (HR 1.09, 95% CI 0.91-1.30) and bolus FP (HR 1.07, 95% CI 0.79-1.45) groups. Only two studies evaluating capecitabine (n = 529 patients) were included in the PFS analysis by FP, but only one study (COIN) was included in the OS analysis, as data from the NEW EPOC Study for OS was not available to include.
1.3 Interaction between oxaliplatin and irinotecan with EGFR-I. In comparing trials combining EGFR-I with ox to those combining EGFR-I with iri, significant interaction was present for PFS (I 2 = 71.2%, p = 0.06, Fig 2) and ORR (I 2 = 96.7%, p<0.00001) but not OS (I 2 = 0%, p = 0.32). When the analysis was restricted to those utilizing infusional FP regimens (i.e. FOLFOX and FOLFIRI), interaction for PFS was no longer present (PFS I 2 = 0%, p = 0.49, S4 Fig) although the ORR interaction persisted (I 2 = 90.5%, p = 0.001), suggesting that choice of FP may be responsible for the interaction between the oxaliplatin-containing v irinotecan-containing regimens. To highlight this point, one can see that the pooled HR for PFS with all oxaliplatin containing regimens is 0.92 (95% CI 0.83-1.02) as compared with irinotecan containing regimens (HR 0.77; 95% CI 0.69-0.86) (Fig 2). When only infusional 5FU regimens are considered (S4 Fig), the pooled PFS HR for oxaliplatin containing regimens is 0.82 (95% CI 0.72-0.94) as compared with irinotecan containing regimens (HR 0.77; 95% CI 0.67-0.88). Thus greater PFS efficacy and confidence is observed with infusional 5-FU regimens and oxaliplatin than with bolus or capecitabine based oxaliplatin combinations.
With respect to the secondary outcome of PFS, pooled analysis was also performed. The addition of EGFR-I to oxaliplatin-based chemotherapy improved PFS (HR 0.70, 95% CI 0.57-0.86, p = 0.0009, S5 Fig). The addition of EGFR-I to irinotecan-based chemotherapy also improved PFS (HR 0.64, 95% CI 0.52-0.78, p<0.00001). Again, no significant subgroup differences were detected (I2 = 0%, p = 0.52). No significant statistical heterogeneity was present for either of the above analyses.
We conducted additional analyses to determine whether the choice of cetuximab or panitumumab may have influenced the results of our analysis, and found that the results were not affected. When only trials investigating cetuximab were included (4 oxaliplatin, 2 irinotecan), addition of EGFR-I to oxaliplatin-based chemotherapy did not improve OS (HR 1.02, 95% CI 0.88-1.     1.07, 95% CI 0.79-1.45) and capecitabine (HR 1.09, 95% CI 0.91-1.30). Given that only 4 trials were involved overall in this analysis (OPUS, COIN, NEW EPOC, NORDIC VII), this analysis should be interpreted with caution.
With regards panitumumab, given that there was only one oxaliplatin and two irinotecanbased trials, meta-analysis was not performed.
2.1.1. Impact of FP type on oxaliplatin + bevacizumab: Analysis by type of FP was performed in the NO16966 and E3200 studies. TML was excluded as separate results for the multiple types of FP used (XELOX, XELIRI, FOLFOX and FOLFIRI) were not available. No significant subgroup differences by type of FP were present. For OS, HR for the infusional group was 0.77 (95% CI 0.65-0.90), for the capecitabine group 0.78 (95% CI 0.53-1.15) with subgroup interaction values I 2 = 0%, p = 0.93. For PFS, HR for the infusional group was 0.70 (95% CI 0.60-0.81) and for capecitabine 0.72 (95% CI 0.50-1.04) with subgroup interaction values I 2 = 0%, p = 0.387.
2.2.1 Impact of FP type on irinotecan + bevacizumab/aflibercept: Analysis by type of FP was performed in the AVF2107g (mIFL), ARTIST (mIFL), ITACA (FOLFIRI) and VELOUR (FOL-FIRI) trials. As in 2.1.1, TML was excluded. For OS, the HR for the infusional group was 0.81 (95% CI 0.72-0.91) and for the bolus group 0.71 (95% CI 0.61-0.83), with subgroup interaction values I 2 = 40.4%, p = 0.20. For PFS, the HR for the infusional group was 0.76 (95% CI 0.67-0.86) and for the bolus group 0.55 (95% CI 0.47-0.64). Although significant subgroup interaction was noted between infusional and bolus 5FU groups in PFS (I 2 = 90.3%, p = 0.001), we note that the bulk of the statistical power in the infusional 5FU group (50.3% out of 58.8% weight) was contributed to by the VELOUR study, evaluating aflibercept in the second-line setting.
3. Trials directly comparing different chemotherapy backbones with same targeted agent. Four trials (CELIM, KRK0104, CECOG, Schmeigel 2013 [34][35][36]) evaluating a total of 262 patients investigated combination of biological therapy (cetuximab in 3 studies, bevacizumab in Schmeigel) with different chemotherapy backbones. Limited outcome data were available for the four studies. For the three cetuximab studies, no significant differences were observed for OS (HR 1.20, 95% CI 0.85-1.70), PFS (meta-analysis not performed as only one trial), or ORR (OR 1.25, 95% CI 0.64-2.45). Meta-analysis was not performed for the single bevacizumab study, which showed no significant differences in OS or PFS between CAPOX+B and CAPIRI+B (although it was not specifically powered for these endpoints).

Sensitivity analysis
We investigated the impact of excluding the NEW EPOC study, which investigated the addition of perioperative cetuximab for resectable liver metastases, as this clinical setting involving curative attempt surgery was distinctly different to the metastatic setting of the other studies. PFS HR was improved somewhat for oxaliplatin regimens with EGFR-I (HR 0.88, 95% CI 0.80-0.98) but unchanged for irinotecan regimens.

Toxicity and quality of life
The addition of biologic agents resulted in increased overall rates of toxicity (S9 and S10 Figs). Only 7/22 trials reported quality of life outcomes using validated tools (S1 Table). The PIC-COLO and AVF2192g studies reported improved quality of life in the experimental arm with other trials showing no significant difference.
Considering toxicity outcomes according to chemotherapy partner, no significant subgroup interaction was observed (I 2 = 60.6%, p = 0.11) for addition of EGFR-I but less toxicity was found adding AIs to oxaliplatin-based trials compared to irinotecan-based trials (I 2 = 90.1%, p = 0.002). Discussion Whilst biologic agents have improved outcomes for patients with mCRC and are integrated into treatment guidelines, the issue of the optimal combination and sequencing of agents remains unclear. This study is the first to systematically examine the effect of chemotherapy backbone, including fluoropyrimidine choice, on the efficacy of biological treatment in mCRC.
Considering the addition of EGFR-I to chemotherapy in KRAS exon 2 WT patients, benefits in OS, PFS and ORR were found in combination with irinotecan-based but not oxaliplatinbased chemotherapy. Investigating the EGFR-I + oxaliplatin subgroup more closely, superior efficacy was observed in trials utilizing infusional 5FU over those using capecitabine. Subsequent analysis of infusional FP based trials alone demonstrated remarkably similar efficacy between the two backbones, pointing to the use of capecitabine as a possible cause for the lower efficacy of EGFR-I when used in combination with oxaliplatin.
This study expands on the meta-analysis by Vale et al [24] by including data from PIC-COLO and NEW EPOC, and confirms that FP choice may be responsible for differential efficacy of adding EGFR-I to ox chemotherapy. We also note the meta-analysis performed by Loupakis et al [37] of anti-EGFR agents in the first line setting. We build upon this by including anti-EGFR trials in all lines, trials investigating anti-angiogenesis agents and perform further subgroup analyses. Given this consistent and independent finding, the available evidence suggests that infusional 5-FU regimens combined with oxaliplatin and EGFR-I may be preferable to bolus 5-FU or capecitabine combinations, notwithstanding other factors affecting choice of regimen such as toxicity and patient preferences.
Two hypotheses may explain the apparent differential activity between type of FP and EGFR-I. One explanation may be increased toxicity from capecitabine-containing regimens with resultant decreased total dose intensity and hence efficacy. Patients in the XELOX arm of the COIN trial received a shorter duration of treatment, median 25.1 weeks in XELOX versus the FOLFOX arm (28.1 weeks). Diarrhoea (23% vs 16% in treatment arms), HFS (16% vs 4%) and stomatitis (4% vs 1%) were all increased in the XELOX arm and may have led the protocol amendment mid-study reducing the dose of capecitabine from 1000 to 850mg/m 2 bid (which also carried through to the NEW EPOC study).
Another hypothesis, albeit speculative, involves the fact that capecitabine requires metabolic activation within cells to its active form as opposed to 5-FU. Cetuximab leads to G1 arrest and thus decreased cell cycling might lead to less cytotoxic activity.
There is scant information as to whether capecitabine combined with irinotecan has deleterious effects on EGFR-I efficacy; the only trial identified investigating this combination was KRK-0104, directly comparing CAPIRI+C and CAPOX+C (cited above) which showed no significant differences in efficacy.
CALGB 80405 [39], comparing the use of cetuximab and bevacizumab, showed no OS efficacy difference in both KRAS exon 2 WT and extended RAS WT populations (although higher response rate-68.6% vs 53.6%, p<0.01 -was achieved with cetuximab in extended RAS WT populations).
The combination of the AIs bevacizumab and aflibercept with chemotherapy improved OS, PFS and ORR with benefit preserved across both oxaliplatin-and irinotecan-based backbones. Subgroup interaction testing favoured increased efficacy for irinotecan. This finding was reported previously but in a pooled analysis of 3763 patients only [26]. This systematic review confirms these findings and also includes additional trials (VELOUR and AVEX).
Restriction to trials of AIs using infusional-only FP in combination with either ox or iri showed a persistent significant PFS benefit but no further subgroup interaction. This interaction is difficult to interpret given the VELOUR contributed to the bulk of the statistical power in the FOLFIRI analysis. A Phase II RCT with FOLFOX+aflibercept has been incompletely reported [14] and we were unable to include it in the analysis. Whilst there was evidence for increased efficacy of bevacizumab added to single-agent FP compared to FP chemotherapy alone, the lesser activity of single-agent FP means that it is usually reserved for elderly or frail patients in routine clinical practice.
A separate question not explicitly addressed by the study is which biological agent optimally combines with which chemotherapy agent (i.e. chemo + EGFR-I first then chemo + AI or vice versa). Whilst FIRE-3 and PEAK point to the possibly increased efficacy of EGFR-I in RAS WT patients, their restriction to one chemotherapy regimen (FOLFIRI in FIRE-3, FOLFOX in PEAK) mean that they cannot definitively answer the questions posed by this paper about chemotherapy backbone choice. We note other studies recently published that address this question. [40] The strengths of this study include the systematic review of all relevant trials and the rigorous methodology. The large number of patients included in analysis helps draw top-level conclusions about the subject matter. The suggestion that FP choice may be responsible for negative interactions between oxaliplatin-based chemotherapies and EGFR-I provides scope for further research.
We recognize several limitations to this study, including restriction of analysis to publication-only results, statistical heterogeneity and the relatively small number of patients in direct comparison trials.
The above meta-analysis has several implications for practice in mCRC. Assuming the availability of all agents, it would seem best to combine EGFR-I with FOLFIRI or FOLFOX based regimens. Based on the available data, CAPOX partnered with EGFR-I appears to be the least effective.
In contrast to the above, AIs may be combined with either oxaliplatin-based or irinotecanbased options. The improved efficacy of AIs added to fluoropyrimidine monotherapy may reflect their greater effectiveness in less active regimens. This points to the importance of considering use of targeted agents even in frailer patients.
Whilst this study raises interesting possibilities of an interaction between cetuximab, oxaliplatin and capecitabine, the biological basis underlying the combination of agents has not been fully elucidated and this study points to the importance of ongoing research in this area.

Conclusions
EGFR-I are best used in combination with irinotecan based regimens or with infusional FP regimens when combined with oxaliplatin. Capecitabine-oxaliplatin combinations with EGFR-I appear less effective. No statistically significant difference in efficacy is seen when AIs are used with both irinotecan or oxaliplatin based regimens.