Panitumumab vs Bevacizumab Added to Standard First-line Chemotherapy and Overall Survival Among Patients With RAS Wild-type, Left-Sided Metastatic Colorectal Cancer

Key Points

Question

Among patients with RAS wild-type, left-sided metastatic colorectal cancer, does adding panitumumab (an anti–epidermal growth factor receptor monoclonal antibody) vs bevacizumab (an anti–vascular endothelial growth factor monoclonal antibody) to standard first-line chemotherapy improve overall survival?

Findings

In this randomized clinical trial that included 802 patients with colorectal cancer (604 with left-sided tumors), adding panitumumab, compared with bevacizumab, to first-line chemotherapy significantly improved overall survival among patients with left-sided tumors (median overall survival, 37.9 months vs 34.3 months; hazard ratio for death, 0.82), and in the overall population (median overall survival, 36.2 months vs 31.3 months; hazard ratio for death, 0.84).

Meaning

Among patients with RAS wild-type metastatic colorectal cancer, adding panitumumab, compared with bevacizumab, to standard first-line chemotherapy significantly improved overall survival in patients with left-sided tumors and in the overall population.

Abstract

Importance

For patients with RAS wild-type metastatic colorectal cancer, adding anti–epidermal growth factor receptor (anti-EGFR) or anti–vascular endothelial growth factor (anti-VEGF) monoclonal antibodies to first-line doublet chemotherapy is routine, but the optimal targeted therapy has not been defined.

Objective

To evaluate the effect of adding panitumumab (an anti-EGFR monoclonal antibody) vs bevacizumab (an anti-VEGF monoclonal antibody) to standard first-line chemotherapy for treatment of RAS wild-type, left-sided, metastatic colorectal cancer.

Design, Setting, and Participants

Randomized, open-label, phase 3 clinical trial at 197 sites in Japan in May 2015–January 2022 among 823 patients with chemotherapy-naive RAS wild-type, unresectable metastatic colorectal cancer (final follow-up, January 14, 2022).

Interventions

Panitumumab (n = 411) or bevacizumab (n = 412) plus modified fluorouracil, l-leucovorin, and oxaliplatin (mFOLFOX6) every 14 days.

Main Outcomes and Measures

The primary end point, overall survival, was tested first in participants with left-sided tumors, then in the overall population. Secondary end points were progression-free survival, response rate, duration of response, and curative (defined as R0 status) resection rate.

Results

In the as-treated population (n = 802; median age, 66 years; 282 [35.2%] women), 604 (75.3%) had left-sided tumors. Median follow-up was 61 months. Median overall survival was 37.9 months with panitumumab vs 34.3 months with bevacizumab in participants with left-sided tumors (hazard ratio [HR] for death, 0.82; 95.798% CI, 0.68-0.99; P = .03) and 36.2 vs 31.3 months, respectively, in the overall population (HR, 0.84; 95% CI, 0.72-0.98; P = .03). Median progression-free survival for panitumumab vs bevacizumab was 13.1 vs 11.9 months, respectively, for those with left-sided tumors (HR, 1.00; 95% CI, 0.83-1.20) and 12.2 vs 11.4 months overall (HR, 1.05; 95% CI, 0.90-1.24). Response rates with panitumumab vs bevacizumab were 80.2% vs 68.6%, respectively, for left-sided tumors (difference, 11.2%; 95% CI, 4.4%-17.9%) and 74.9% vs 67.3% overall (difference, 7.7%; 95% CI, 1.5%-13.8%). Median duration of response with panitumumab vs bevacizumab was 13.1 vs 11.2 months for left-sided tumors (HR, 0.86; 95% CI, 0.70-1.10) and 11.9 vs 10.7 months overall (HR, 0.89; 95% CI, 0.74-1.06). Curative resection rates with panitumumab vs bevacizumab were 18.3% vs 11.6% for left-sided tumors; (difference, 6.6%; 95% CI, 1.0%-12.3%) and 16.5% vs 10.9% overall (difference, 5.6%; 95% CI, 1.0%-10.3%). Common treatment-emergent adverse events were acneiform rash (panitumumab: 74.8%; bevacizumab: 3.2%), peripheral sensory neuropathy (panitumumab: 70.8%; bevacizumab: 73.7%), and stomatitis (panitumumab: 61.6%; bevacizumab: 40.5%).

Conclusions and Relevance

Among patients with RAS wild-type metastatic colorectal cancer, adding panitumumab, compared with bevacizumab, to standard first-line chemotherapy significantly improved overall survival in those with left-sided tumors and in the overall population.

Trial Registration

ClinicalTrials.gov Identifier: NCT02394795

This randomized clinical trial assesses the effect of adding panitumumab vs bevacizumab to standard first-line chemotherapy on overall survival among patients with RAS wild-type, left-sided, metastatic colorectal cancer.

Introduction

Colorectal cancer was the second most common cause of cancer death worldwide in 2020.¹ An estimated 20% to 30% of patients with colorectal cancer have metastatic tumors at diagnosis.² Following surgery with curative intent, approximately 10% to 25% of patients develop metachronous metastases in studies of patients treated between 2000 and 2015.^3,4 Systemic chemotherapy is standard treatment for patients with unresectable metastatic or recurrent disease.^2,5 Combining chemotherapy with either anti–epidermal growth factor receptor (anti-EGFR) monoclonal antibodies, such as cetuximab or panitumumab, or the anti–vascular endothelial growth factor (anti-VEGF) antibody bevacizumab prolonged median overall survival up to approximately 30 months.^6,7

Sequence variants in KRAS (NG_007524) exon 2, which were reported in approximately 35% to 45% of patients with colorectal cancer in a systematic review of studies published through 2014, were associated with poorer survival with anti-EGFR antibodies compared with KRAS wild type.⁸ The phase 3 CALGB/SWOG 80405 and FIRE-3 trials comparing cetuximab vs bevacizumab combined with first-line chemotherapy showed inconclusive results in patients without KRAS exon 2 sequence variants.^6,7 Subsequent exploratory analyses that excluded patients with other, less frequent variants in KRAS (exons 3, 4) or NRAS (NG_007572; exons 2, 3, 4) reported improved survival with anti-EGFR antibodies in patients without any KRAS or NRAS variants (RAS wild type).^6,8 More recently, the benefit of anti-EGFR antibodies plus first-line chemotherapy was suggested to be enhanced in patients with RAS wild-type primary tumors originating in the left side of the colon, which includes the rectum.^9,10,11 Prevailing practice guidelines support use of either anti-EGFR antibodies or bevacizumab in patients with RAS wild-type, left-sided tumors, and these guidelines prompted evaluation of the optimal treatment approach for this group.^12,13

To our knowledge, this trial (PARADIGM) is the first randomized clinical trial to test the superiority of an anti-EGFR antibody, compared with an anti-VEGF antibody, in combination with standard fluorouracil and oxaliplatin combination chemotherapy in patients with RAS wild-type, left-sided metastatic colorectal cancer.

Methods

Trial Oversight

The steering committee members and investigators employed by the industry sponsor designed the trial. The protocol was reviewed and approved by the Certified Review Board of the National Cancer Center Hospital East and registered with Japan’s Ministry of Health, Labor, and Welfare (Japan Registry of Clinical Trials) and ClinicalTrials.gov. The trial was conducted in accordance with the ethical principles of the Declaration of Helsinki,¹⁴ the Clinical Research Act (Act No. 16 of 2017 [Japan]), the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH)–Good Clinical Practice, and all applicable laws and regulations. All participants provided written informed consent.

Trial Design and Participants

The design of this randomized, open-label, multicenter, phase 3 trial was described previously¹⁵; the trial protocol is available in Supplement 1 and the statistical analysis plan in Supplement 2. Briefly, eligible patients had KRAS/NRAS wild-type unresectable adenocarcinoma originating in the colorectum and had received no prior chemotherapy for metastatic colorectal cancer. Participants previously treated with oxaliplatin in the adjuvant or neoadjuvant setting were ineligible. Those treated with adjuvant or neoadjuvant fluoropyrimidine therapy alone were eligible provided that this treatment was completed at least 24 weeks prior to enrollment. KRAS/NRAS tests were performed using approved in vitro diagnostic tests,¹⁶ and all of the following exons/codons for both KRAS and NRAS were required to be wild type: exon 2, codons 12, 13; exon 3, codons 59, 61; and exon 4, codons 117, 146.¹⁷ Other inclusion criteria were age 20 to 79 years; Eastern Cooperative Oncology Group performance status score of 0 or 1; presence of at least 1 evaluable lesion according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1; adequate kidney and liver function; and life expectancy of at least 3 months after enrollment. Patients were excluded if they had received radiotherapy within 4 weeks or if they had brain metastasis or synchronous or metachronous cancers. Inclusion and exclusion criteria are described in the trial protocol (Supplement 1). Patients were enrolled regardless of primary tumor location.

Protocol Revisions

In May 2019, after enrollment completion (last participant enrolled in June 2017) but before data were analyzed for the first interim analysis (November 2019), the trial protocol was revised to assess the primary outcome of overall survival in participants with left-sided primary tumors, defined as those originating in the descending colon, sigmoid colon, rectosigmoid, and rectum. This change was made to evaluate the superiority of anti-EGFR therapy for RAS wild-type, left-sided tumors, as suggested by results of post hoc analyses of other randomized clinical trials reported after completion of enrollment but before analysis of this trial.^9,10,11 An independent data monitoring committee reviewed data for the prespecified interim analysis in November 2019 without sharing any results with the steering committee, the lead statistician, or employees of the sponsor. The trial protocol was further revised in July 2020 to adopt a hierarchical testing procedure that first compared the 2 treatments in participants with left-sided tumors followed by the overall population. The steering committee, the lead statistician, and employees of the sponsor were blinded from all study results when they made revisions to the statistical analysis plan.

Randomization

Participants were randomized in a 1:1 ratio to receive panitumumab plus mFOLFOX6 (combination fluorouracil, l-leucovorin, and oxaliplatin) or to receive bevacizumab plus mFOLFOX6 by minimization method, with a random component to balance groups on the basis of institution, age (20-64 vs 65-79 years), and presence or absence of liver metastases. The highest allocation ratio (2/3) was adopted to minimize imbalance between the 2 groups.

Treatments

Panitumumab, 6 mg/kg, was administered on day 1 as a 60-minute intravenous infusion. Bevacizumab, 5 mg/kg, was administered on day 1 as an intravenous infusion over 30 to 90 minutes as specified at each institution. mFOLFOX6 consisted of oxaliplatin, 85 mg/m², on day 1; l-leucovorin, 200 mg/m², on day 1; a bolus of fluorouracil, 400 mg/m², on day 1; followed by a continuous intravenous infusion of 2400 mg/m² on days 1 through 3. All treatments were administered every 14 days. Criteria for dose reduction or suspension and discontinuation of treatment and supportive care measures are provided in the trial protocol (Supplement 1). Participants received treatment until progressive disease or unacceptable toxicity occurred, participant withdrawal, physician decision to discontinue treatment for reasons other than disease progression or toxicity, or curative-intent surgery.

Primary Outcome

The primary end point was overall survival in participants with left-sided tumors and in the overall population. Overall survival was defined as the period from the day of randomization (day 1) until death due to any cause.

Secondary Outcomes

Secondary end points included progression-free survival, response rate, duration of response, and rate of successful curative-intent resection (defined as a postoperative R0 status). For participants who underwent curative-intent resection, the progression-free survival period ended on the final day when preoperative diagnostics confirmed no progressive disease. For participants who discontinued study treatment due to an adverse event or other reasons without disease progression or death, the progression-free survival period was the time from randomization to progressive disease or death or until the participant was censored at the last follow-up date.

Exploratory Outcomes

Exploratory end points included percentage of early tumor shrinkage, best percentage change from baseline in target lesions (ie, depth of response), disease control rate, and time to treatment failure. Imaging tests (eg, computed tomography, magnetic resonance imaging) were performed at enrollment, every 8 weeks for the first 2 years, and every 12 weeks thereafter. Adverse events were graded according to the Common Terminology Criteria for Adverse Events (Japanese edition, Japan Clinical Oncology Group version 4.03). After discontinuation of study treatment, information on follow-up treatment was collected every 6 months.

Sample Size Calculation

Initially, the trial was designed to compare overall survival in all participants and planned to enroll 800 participants, with 570 expected deaths to ensure 80% power with a 1-sided significance level of .025. To assess differences in overall survival, a piecewise exponential distribution was assumed, in which the survival rate of the 2 treatment groups would be the same until 18 months, and then after 48 months the survival rate would be 22% and 32% for the bevacizumab and panitumumab groups, respectively. In May 2019, the primary analysis was amended to detect a significant difference in overall survival between treatment groups in the left-sided tumor and overall populations with a 1-sided type I error of .0125 for each group. The interim analysis of overall survival was planned when approximately 70% of the overall targeted number of events were observed (399 events for all participants and 280 events for participants with left-sided tumors). In November 2019, at the prespecified interim analysis, the independent data monitoring committee recommended continuation of the trial. The primary analysis was further amended in July 2020 to adopt a hierarchical testing procedure comparing the 2 treatments, first in the population with left-sided tumors followed by the overall population. Based on Monte Carlo simulations, 420 deaths among participants with left-sided tumors would provide 80% power to detect an overall survival hazard ratio (HR) of 0.74 at a 1-sided significance level of .02101. The HR of 0.74 was assumed based on median survival reported in the TRICOLORE study (33.6 months).¹⁸ Because the interim analysis had already been performed and it was deemed impractical to explicitly calculate the correlation between the test statistics for the left-sided tumor population in the final analysis and the overall population in the interim analysis, the significance level of .0125 to .00308 was adopted, a conservative value. After considering the α spending that was prespecified for the left-sided tumor population in the interim analysis (1-sided α of .00308), the 1-sided significance level was set as .01159 for the left-sided tumor population in the final analysis, yielding a 1-sided significance level of .02101 for the primary analysis. The resultant value (.02101, corresponding to a 2-sided value of .04202) considers the test statistic based on the α spending function approach after interim analysis of both the left-sided tumor and overall populations and the test statistic for the final analysis in participants with left-sided tumors.

Statistical Analysis

Primary and secondary outcomes were evaluated in randomized participants who received at least 1 dose of study treatment and satisfied eligibility criteria and had left-sided tumors or had tumors located in either the right or left side of the colorectum. This analysis population was consistent with the definition of a full analysis set per the ICH Harmonised Tripartite Guideline: Statistical Principles for Clinical Trials.¹⁹ The incidence and severity of adverse events were assessed in all randomized participants who received study treatment.

Overall survival was analyzed using a stratified log-rank test, with randomization factors of age and presence of liver metastases, for participants with left-sided tumors at a 2-sided α = .04202. Survival in the overall population was only to be tested if participants with left-sided tumors showed a statistically significant difference (using a 2-sided α = .05). A Cox analysis with step function for a time-varying coefficient was conducted because a violation of the proportional hazards assumption (tested using the method of Grambsch and Therneau²⁰) was observed in post hoc analyses. Kaplan-Meier methods were used to generate survival curves, and intergroup HRs and confidence intervals were calculated using a stratified Cox model with the same randomization factors. The point-estimate difference (and 95% CI) between treatment groups in response rate was calculated with the Cochran-Mantel-Haenszel test using randomization factors other than study site. A threshold for statistical significance was not preestablished for group comparisons of secondary outcome measures because the primary outcome (overall survival) was considered the most clinically relevant outcome based on the results of the FIRE-3 and CALGB/SWOG 80405 trials, which showed no differences in progression-free survival or response rate between anti-EGFR and anti-VEGF therapies.^7,21 No adjustments were made for multiple comparisons in the analysis of secondary end points. Because of the potential for type I error due to multiple comparisons, findings for analyses of secondary end points should be interpreted as exploratory.

Subgroup analyses of prespecified participant characteristics were performed for overall survival and progression-free survival using a Cox proportional hazards model with no adjustment for multiplicity in which allocated treatment group, subgroup, and the interaction of allocated treatment group by subgroup were fixed effects. For the subgroup of participants with right-sided primary tumors, analyses of end points were predefined. A post hoc analysis evaluated overall survival in participants in whom curative resection was and was not achieved. No imputation of missing data was performed.

All statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc) and R version 4.0.3 (R Foundation).

Results

Participants

From May 2015 to June 2017, 823 participants were randomized at 197 centers in Japan. Twelve participants did not receive study treatment and 9 were excluded for not meeting inclusion criteria. The overall population included 400 participants who received panitumumab plus mFOLFOX6 and 402 participants who received bevacizumab plus mFOLFOX6; of these, 312 (78%) and 292 (73%) participants, respectively, had a left-sided primary tumor and 84 (21%) and 103 (26%) had a right-sided primary tumor; 4 (1%) and 7 (2%) had multiple primary lesions in both the left and right sides (Figure 1). Demographic and baseline clinical characteristics were balanced between treatment groups for left-sided tumor, overall, and right-sided tumor populations (Table 1). Among participants with right-sided tumors, more participants were female, especially in the panitumumab group, and fewer participants had liver-only metastases vs the overall population and the population with left-sided tumors. No participants were missing data for the analyses of overall survival and progression-free survival.

Figure 1. Participant Flow in the PARADIGM Trial.

mFOLFOX6 indicates modified fluorouracil, l-leucovorin, and oxaliplatin.

^aThe total number of patients approached for participation in the trial was not recorded.

^bThe total number of participants overall is not the sum of the participants with left-sided tumors and participants with right-sided tumors because 11 participants had lesions in both the left and right sides (4 in the panitumumab group and 7 in the bevacizumab group).

Table 1. Baseline Participant Characteristics.

Characteristics	Participants with left-sided tumorsa		Overall population		Participants with right-sided tumorsb
	Panitumumab plus mFOLFOX6	Bevacizumab plus mFOLFOX6	Panitumumab plus mFOLFOX6	Bevacizumab plus mFOLFOX6	Panitumumab plus mFOLFOX6	Bevacizumab plus mFOLFOX6
No. of patientsc	312	292	400	402	84	103
Age, y
Median (range)	65.5 (35-79)	65.5 (28-79)	66.0 (32-79)	66.0 (28-79)	68.0 (32-79)	67.0 (39-79)
No. (%)
20-64	138 (44.2)	127 (43.5)	164 (41.0)	168 (41.8)	26 (31.0)	39 (37.9)
65-79	174 (55.8)	165 (56.5)	236 (59.0)	234 (58.2)	58 (69.0)	64 (62.1)
Sex, No. (%)
Female	104 (33.3)	91 (31.2)	148 (37.0)	134 (33.3)	43 (51.2)	42 (40.8)
Male	208 (66.7)	201 (68.8)	252 (63.0)	268 (66.7)	41 (48.8)	61 (59.2)
ECOG performance status score, No. (%)d
0 (Fully active with no performance restriction)	261 (83.7)	231 (79.1)	328 (82.0)	319 (79.4)	65 (77.4)	82 (79.6)
1 (Ambulatory but strenuous physical activity restricted)	51 (16.3)	61 (20.9)	71 (17.8)	83 (20.6)	18 (21.4)	21 (20.4)
2 (Capable of self-care but unable to carry out work activities)	0	0	1 (0.3)	0	1 (1.2)	0
Primary tumor location, No. (%)
Left side	312 (100.0)	292 (100.0)	312 (78.0)	292 (72.6)	0	0
Right side	0	0	84 (21.0)	103 (25.6)	84 (100.0)	103 (100.0)
Both sides	0	0	4 (1.0)	7 (1.7)	0	0
No. of organs with metastasis, No. (%)
1	155 (49.7)	147 (50.3)	196 (49.0)	194 (48.3)	40 (47.6)	44 (42.7)
≥2	157 (50.3)	145 (49.7)	204 (51.0)	208 (51.7)	44 (52.4)	59 (57.3)
Metastasis site, No. (%)
Liver	225 (72.1)	206 (70.5)	275 (68.8)	278 (69.2)	49 (58.3)	66 (64.1)
Liver as only site of metastasis	90 (28.8)	89 (30.5)	105 (26.3)	113 (28.1)	14 (16.7)	21 (20.4)
Prior treatment, No. (%)
Primary tumor resection	185 (59.3)	193 (66.1)	239 (59.8)	272 (67.7)	51 (60.7)	73 (70.9)
Radiotherapy	2 (0.6)	2 (0.7)	2 (0.5)	3 (0.7)	0	0
Adjuvant chemotherapye	17 (5.4)	16 (5.5)	22 (5.5)	20 (5.0)	5 (6.0)	3 (2.9)

Abbreviation: mFOLFOX6, modified fluorouracil, l-leucovorin, and oxaliplatin.

^a Descending colon, sigmoid colon, rectosigmoid, and rectum.

^b Primary tumors originating on the right side of the colon, defined as cecum, ascending colon, or transverse colon.

^c The total number of participants overall is not the sum of the participants with left-sided tumors and participants with right-sided tumors because there were 11 participants who had multiple primary lesions in both the left and right sides.

^d Eastern Cooperative Oncology Group (ECOG) performance status scores range from 0 (fully active) to 4 (completed disabled). This status was evaluated at the initiation of treatment, not at enrollment, and 1 participant was enrolled and before starting treatment was evaluated as having an ECOG performance status of 2.

^e Patients who experienced relapse more than 24 weeks (168 days) after the final dose of perioperative adjuvant chemotherapy with fluoropyrimidine agents were allowed to enroll in the study. Patients who had received perioperative adjuvant chemotherapy that included oxaliplatin were not eligible for enrollment.

Primary Outcome: Overall Survival

At the cutoff date, January 14, 2022, with a median follow-up of 61 months, 448 deaths were observed in the participants with left-sided tumors and 613 in the overall population. In participants with left-sided tumors, overall survival was statistically significantly prolonged with panitumumab plus mFOLFOX6 with a median survival time of 37.9 months (95.798% CI, 34.1-42.6 months) vs 34.3 months (95.798% CI, 30.9-40.3 months) with bevacizumab plus mFOLFOX6 (stratified HR, 0.82; 95.798% CI, 0.68-0.99; P = .03) (Figure 2). The difference was also statistically significant in the overall population, with a median overall survival time of 36.2 months (95% CI, 32.0-39.0 months) with panitumumab and 31.3 months (95% CI, 29.3-34.1 months) with bevacizumab (stratified HR, 0.84; 95% CI, 0.72-0.98; P = .03) (Figure 2). Overall survival probabilities at 3, 4, and 5 years were 53%, 42%, and 32% with panitumumab plus mFOLFOX6 vs 47%, 33%, and 21% with bevacizumab plus mFOLFOX6 in participants with left-sided tumors, respectively, and 50%, 38%, and 29% with panitumumab plus mFOLFOX6 vs 42%, 30%, and 20% with bevacizumab plus mFOLFOX6 in the overall population, respectively. The Kaplan-Meier curves for overall survival were not parallel and appeared to cross around 28 months. In such instances, interpretation of the observed HRs is unclear.

Figure 2. Overall Survival and Progression-Free Survival in the Overall Study Population and by Tumor Sidedness.

Median follow-up for overall and progression-free survival was 61.3 (IQR, 57.6-65.5) months for panitumumab and 60.6 (IQR, 57.0-66.1) months for bevacizumab in the overall population; 61.3 (IQR, 57.2-65.2) months for panitumumab and 61.0 (IQR, 57.1-66.2) months for bevacizumab for left-sided tumors; and 61.3 (IQR, 60.5-66.6) months for panitumumab and 60.2 (IQR, 55.7-65.5) months for bevacizumab for right-sided tumors. Data are shown ordered as left-sided tumors, right-sided tumors, and overall for clinical interest, but the order of primary statistical analyses was prespecified to evaluate left-sided tumors first, then the overall population. Analyses of right-sided tumors were predefined exploratory outcomes.

Secondary Outcomes

Progression-Free Survival

Median progression-free survival was 13.1 months (95% CI, 11.6-14.5 months) with panitumumab vs 11.9 months (95% CI, 11.3-13.5 months) with bevacizumab (HR, 1.00; 95% CI, 0.83-1.20) in participants with left-sided tumors (Figure 2) and 12.2 months (95% CI, 10.8-13.2 months) vs 11.4 months (95% CI, 11.2-13.2 months), respectively, in the overall population (HR, 1.05; 95% CI, 0.90-1.24) (Figure 2).

Response Rate

In participants with left-sided tumors, complete or partial response was observed in 80.2% (247/308; 95% CI, 75.3%-84.5%) of participants in the panitumumab group compared with 68.6% (197/287; 95% CI, 62.9%-74.0%) in the bevacizumab group (point estimate of difference, 11.2%; 95% CI, 4.4%-17.9%) (Table 2). In the overall population, complete or partial response was observed in 74.9% (295/394; 95% CI, 70.3%-79.1%) of participants in the panitumumab group vs 67.3% (267/397; 95% CI, 62.4%-71.9%) in the bevacizumab group (difference, 7.7%; 95% CI, 1.5%-13.8%).

Table 2. Secondary Outcomes in Participants With Left-Sided Tumors and in the Overall Population and Exploratory Outcomes in Participants With Right-Sided Tumors.

Outcomes	Panitumumab plus mFOLFOX6	Bevacizumab plus mFOLFOX6
Participants with left-sided primary tumors	n = 312	n = 292
Response rate, No./total (%) [95% CI]	247/308 (80.2) [75.3-84.5]	197/287 (68.6) [62.9-74.0]
Median duration of response (95% CI), moa	13.1 (11.1-14.8)	11.2 (9.6-13.1)
Curative (R0) resection, No./total (%) [95% CI]	57/312 (18.3) [14.1-23.0]	34/292 (11.6) [8.2-15.9]
Overall populationb	n = 400	n = 402
Response rate, No./total (%) [95% CI]	295/394 (74.9) [70.3-79.1]	267/397 (67.3) [62.4-71.9]
Median duration of response (95% CI), moa	11.9 (10.5-13.4)	10.7 (9.5-12.2)
Curative (R0) resection, No./total (%) [95% CI)	66/400 (16.5) [13.0-20.5]	44/402 (10.9) [8.1-14.4]
Participants with right-sided primary tumors	n = 84	n = 103
Response rate, No./total (%) [95% CI]	45/82 (54.9) [43.5-65.9]	65/103 (63.1) [53.0-72.4]
Median duration of response (95% CI), moa	8.8 (5.8-11.1)	9.7 (6.7-14.3)
Curative (R0) resection, No./total (%) [95% CI]	9/84 (10.7) [5.0-19.4]	10/103 (9.7) [4.8-17.1]

Abbreviation: mFOLFOX6, modified fluorouracil, l-leucovorin, and oxaliplatin.

^a Duration of response was evaluated in participants with complete or partial response.

^b The number of participants in the overall population is not the sum of the participants with left-sided tumors plus participants with right-sided tumors because there were 11 participants who had multiple primary lesions in both the left and right sides.

Duration of Response

Among participants with a complete or partial response, the median duration of response was 13.1 months (95% CI, 11.1-14.8 months) in the panitumumab group and 11.2 months (95% CI, 9.6-13.1 months) in the bevacizumab group in participants with left-sided tumors (HR, 0.86; 95% CI, 0.70-1.10) and 11.9 months (95% CI, 10.5-13.4 months) and 10.7 months (95% CI, 9.5-12.2 months) in the overall population (HR, 0.89; 95% CI, 0.74-1.06).

Curative Resection

In participants with left-sided tumors, 57 of 312 (18.3%; 95% CI, 14.1%-23.0%) participants receiving panitumumab and 34 of 292 (11.6%; 95% CI, 8.2%-15.9%) participants receiving bevacizumab were able to undergo surgery with excision of all colorectal cancer (curative resection) after study treatment (difference, 6.6%; 95% CI, 1.0%-12.3%). In the overall group, curative resection was achieved in 66 of 400 (16.5%; 95% CI, 13.0%-20.5%) receiving panitumumab and 44 of 402 (10.9%; 95% CI, 8.1%-14.4%) receiving bevacizumab (difference, 5.6%; 95% CI, 0.8%-10.3%).

Subgroup Analyses

Outcomes in Participants With Right-Sided Tumors

Median overall survival was 20.2 months (95% CI, 15.2-32.0 months) with panitumumab and 23.2 months (95% CI, 18.5-29.1 months) with bevacizumab, with a stratified HR of 1.09 (95% CI, 0.79-1.51) (Figure 2). Median progression-free survival was 7.2 months vs 9.4 months, respectively (HR, 1.43; 95% CI, 1.03-1.97) (Figure 2). Response rates were 54.9% (95% CI, 43.5%-65.9%) and 63.1% (95% CI, 53.0%-72.4%), respectively. Among participants with a complete or partial response, the median duration of response was 8.8 months (95% CI, 5.8-11.1 months) and 9.7 months (95% CI, 6.7-14.3 months), respectively. Curative resection was achieved in 9 of 84 participants (10.7%; 95% CI, 5.0%-19.4%) receiving panitumumab vs 10 of 103 (9.7%; 95% CI, 4.8%-17.1%) receiving bevacizumab (difference, 1.0%; 95% CI, −7.7% to 9.7%).

Subgroup analyses based on other participant characteristics are shown in eFigures 1 and 2 in Supplement 3 and did not reveal any important differences. Testing for interactions revealed no statistically significant interactions between subgroups.

Exploratory and Post Hoc Outcomes

Exploratory end points of early tumor shrinkage, disease control rate, time to treatment failure, and depth of response are reported in eTable 1 and eFigure 3 in Supplement 3.

Post hoc analyses of violation of the proportional hazards assumption showed that the HR for overall survival was different during the first 28 months (participants with left-sided tumors: HR, 1.04 [95% CI, 0.80-1.35]; overall population: HR, 1.00 [95% CI, 0.81-1.24]) than from 28 months until the end of follow-up (participants with left-sided tumors: HR, 0.63 [95% CI, 0.49-0.83]; overall population: HR, 0.67 [95% CI, 0.53-0.85]). At least 80% of participants in each treatment group received subsequent therapy, with 45.5% of participants in the bevacizumab group receiving subsequent panitumumab (eTable 2 in Supplement 3).

Among participants in whom curative resection was achieved, median overall survival was not reached (ie, less than 50% of participants had died) in either treatment group in participants with left-sided tumors (panitumumab vs bevacizumab: HR, 0.84; 95% CI, 0.41-1.69) and in the overall population (HR, 0.89; 95% CI, 0.48-1.66). Among participants in whom curative resection was not achieved, median overall survival was 33.7 months with panitumumab vs 31.1 months with bevacizumab in participants with left-sided tumors (HR, 0.89; 95% CI, 0.74-1.08) and 30.6 vs 29.7 months, respectively, in the overall population (HR, 0.90; 95% CI, 0.76-1.06).

Adverse Events

Among all participants, the median treatment duration was 6.2 months with panitumumab and 7.7 months with bevacizumab. Median duration of oxaliplatin treatment was 6.2 months in the panitumumab group and 7.6 months in the bevacizumab group. Adverse events occurred in 98.6% of all participants (800/811) (Table 3). Grade 3 or higher adverse events were reported in 71.8% of participants (290/404) in the panitumumab group and 64.9% (264/407) in the bevacizumab group (Table 3; eTable 3 in Supplement 3). Treatment-related deaths occurred in 10 participants receiving panitumumab, including 4 due to interstitial lung disease per investigator assessment, as well as 1 participant each due to lung disorder, pneumonia, pneumonitis, pneumonia and pancytopenia, sepsis and peritonitis, and cerebral hemorrhage per investigator assessment. Two participants receiving bevacizumab had a treatment-related death; 1 death was due to respiratory failure and the cause of the other death was not specified.

Table 3. Incidence of Adverse Events.

Adverse eventsa	No. (%) of participants
	Panitumumab plus mFOLFOX6		Bevacizumab plus mFOLFOX6
No. of participantsb	404		407
Any adverse event	402 (99.5)		398 (97.8)
Grade 3 or higher adverse events	290 (71.8)		264 (64.9)
Serious adverse events related to study treatment	72 (17.8)		44 (10.8)
Adverse events leading to discontinuation of study treatment	96 (23.8)		75 (18.4)
Common adverse events (≥20%)c	Grade 3 or higher	Any grade	Grade 3 or higher	Any grade
Nervous system disorders
Peripheral sensory neuropathy	35 (8.7)	286 (70.8)	40 (9.8)	300 (73.7)
Dysgeusia	0	125 (30.9)	0	94 (23.1)
Gastrointestinal tract disorders
Stomatitis	27 (6.7)	249 (61.6)	7 (1.7)	165 (40.5)
Nausea	6 (1.5)	160 (39.6)	12 (2.9)	161 (39.6)
Diarrhea	26 (6.4)	151 (37.4)	13 (3.2)	136 (33.4)
Constipation	0	93 (23.0)	2 (0.5)	107 (26.3)
Skin and subcutaneous tissue disorders
Acne-like dermatitis	69 (17.1)	302 (74.8)	0	13 (3.2)
Dry skin	31 (7.7)	186 (46.0)	1 (0.2)	38 (9.3)
Palmar-plantar erythrodysesthesia syndrome	9 (2.2)	94 (23.3)	1 (0.2)	57 (14.0)
Laboratory measurements
Decreased neutrophil count	129 (31.9)	202 (50.0)	144 (35.4)	225 (55.3)
Decreased platelet count	6 (1.5)	86 (21.3)	3 (0.7)	80 (19.7)
Metabolism and nutrition disorders
Decreased appetite	31 (7.7)	225 (55.7)	18 (4.4)	206 (50.6)
Hypomagnesemia	33 (8.2)	121 (30.0)	0	7 (1.7)
General disorders and administration site conditions
Fatigue	18 (4.5)	159 (39.4)	15 (3.7)	162 (39.8)
Infections and infestations
Paronychia	38 (9.4)	210 (52.0)	1 (0.2)	20 (4.9)

Abbreviation: mFOLFOX6, modified fluorouracil, l-leucovorin, and oxaliplatin.

^a Adverse events were graded by the National Cancer Institute Common Terminology Criteria for Adverse Events (Japanese edition, Japan Clinical Oncology Group version 4.03) and were coded by system organ class and preferred term in the Medical Dictionary for Regulatory Activities version 24.0.

^b Participants who were randomized and received at least 1 dose of study treatment.

^c Adverse events reported in 20% or more of participants in either treatment group; sorted in descending frequency of occurrence by system organ class in the total group (panitumumab plus bevacizumab).

Adverse events more common with panitumumab vs bevacizumab were acne-like dermatitis (74.8% vs 3.2%), paronychia (52.0% vs 4.9%), dry skin (46.0% vs 9.6%), and hypomagnesemia (30.0% vs 1.7%), whereas hypertension (1.7% vs 18.9%) and epistaxis (3.2% vs 19.9%) were more common with bevacizumab.

Discussion

In this phase 3 clinical trial, the addition of panitumumab, an anti-EGFR antibody, to standard first-line chemotherapy (mFOLFOX6) significantly improved overall survival in patients with unresectable, RAS wild-type, left-sided, metastatic colorectal cancer, compared with the addition of an anti-VEGF antibody (bevacizumab) to mFOLFOX6, confirming results of post hoc analyses of previous randomized trials.^9,10,11 After a median follow-up of 61 months, overall survival in the panitumumab group reached a median of 37.9 months.^10,21,22 In a post hoc analysis of the FIRE-3 trial (n = 306), overall survival in patients with RAS wild-type left-sided tumors was 38.3 months for cetuximab plus FOLFIRI (combination fluorouracil, l-leucovorin, and irinotecan) vs 28.0 months for bevacizumab plus FOLFIRI (HR, 0.63; 95% CI, 0.48-0.85), and the survival curves showed separation after 2 years of treatment.^10,21,22 In patients with RAS wild-type, left-sided tumors in the CALGB/SWOG 80405 trial (n = 325), median overall survival was 39.3 months with cetuximab and 32.6 months with bevacizumab in combination with chemotherapy (HR, 0.77; 95% CI, 0.59-0.99).¹⁰

Among participants with left-sided tumors, median progression-free survival was 13.1 months in the panitumumab group and 11.9 months in the bevacizumab group, consistent with the finding that progression-free survival was not significantly different with anti-VEGF vs anti-EGFR treatment in the CALBG/SWOG 80405 and FIRE-3 trials.^6,7 The panitumumab group had numerically improved response rates and early and maximum tumor shrinkage, which may contribute to improved survival outcomes.^22,23 Depth of response (ie, maximum percentage decrease in size of target cancer lesions) to first-line treatment may be a better predictor of postprogression survival than progression-free survival in this setting because even if progression-free survival with first-line chemotherapy was similar between the 2 groups, postprogression survival may be longer in the group with less tumor burden after first-line chemotherapy or before starting second-line chemotherapy.²⁴ Although no apparent differences between treatment groups were observed in proportions of participants receiving subsequent therapies, treatment sequence might affect outcomes on subsequent treatments and survival.^21,25,26 In FIRE-3, the duration of second-line treatment was substantially longer when the first-line treatment was bevacizumab than when the first-line treatment was anti-EGFR antibodies, which was supported by a preclinical study showing that cetuximab-resistant colorectal cancer cells upregulate VEGF concentrations and respond to antiangiogenic treatment.^27,28 The New EPOC trial in patients with resectable colorectal liver metastases showed that adding cetuximab to perioperative chemotherapy significantly reduced progression-free survival and overall survival.²⁹ However, those results are difficult to compare with the current study, which was restricted to patients with unresectable advanced disease. Study treatment was discontinued if surgeons made the decision to proceed to surgery with the intention to completely resect a tumor.

Chemotherapy sometimes enables conversion of tumors from unresectable to resectable. Of note, the percentage of patients who underwent surgery and had a complete resection with no residual tumor (curative resection rate) after study treatment was 18.3% with panitumumab vs 11.6% with bevacizumab in the left-sided tumor population. This may partially explain the late separation of the overall survival curves and the lack of an observable difference in progression-free survival. The CALBG/SWOG 80405 trial^30,31 also found that the rate of curative-intent surgery was higher for patients randomized to cetuximab vs bevacizumab.

Although this study showed significant improvement of survival in the overall population, subgroup analysis suggested no significant difference in overall survival between treatments in patients with right-sided tumors, suggesting that the survival benefit was mainly among participants with left-sided tumors. Left-sided and right-sided colonic tumors differ in their embryonic origin, histology, and sequence variation.^32,33 Right-sided tumors have increased frequencies of microsatellite instability, CpG island methylator phenotype, and BRAF variants.^32,34 Because survival curves showed no discernible difference between the 2 groups until 28 months, even in patients with left-sided tumors, additional biomarkers, such as BRAF, ERBB2 (HER2), and microsatellite instability, that may account for these differences warrant further investigation.

Limitations

This study has several limitations. First, because the primary population was redefined based on tumor sidedness during the trial, randomization stratification factors did not include primary tumor location. Second, the primary analysis was not conducted among all randomized participants. Third, although previous studies have reported similar perioperative safety profiles in patients treated with a VEGF inhibitor vs an anti-EGFR antibody,^35,36 there was no strict definition for investigators regarding the decision to proceed to curative-intent resection, which may have introduced bias if surgeons were reluctant to operate on patients taking VEGF inhibitors. Fourth, it is unclear if the results may be extrapolated to patients with previously resected colon cancer treated with adjuvant FOLFOX, because these patients were excluded from this study. Fifth, the results may not apply to patients receiving FOLFIRI.¹³ Sixth, no information about microsatellite instability–high cancer, ERBB2, or BRAF status was available at the time of this analysis. Seventh, this trial enrolled only patients from Japan, and generalizability to other regions is uncertain. However, baseline disease characteristics and outcomes in the bevacizumab group were consistent with previous studies in other countries.^18,37,38 SCRUM-Japan, GI-SCREEN/MONSTAR-SCREEN, and global next-generation sequencing data analyses have identified few differences in the genomic profiles between Japanese and non-Japanese patients with colorectal cancer.^39,40

Conclusions

Among patients with RAS wild-type metastatic colorectal cancer, the addition of panitumumab, compared with bevacizumab, to standard first-line chemotherapy significantly improved overall survival in those with left-sided tumors and in the overall population.

Supplement 1.

Trial Protocol

Supplement 2.

Statistical Analysis Plan

Supplement 3.

Paradigm Trial Committee Members and Investigators

eAppendix. History of Trial Protocol Revisions, Revisions to the Statistical Analysis Plan (SAP), and Supplemental Results

eFigure 1. Subgroup Analyses of Overall Survival

eFigure 2. Subgroup Analyses of Progression-Free Survival

eFigure 3. Depth of Response

eTable 1. Exploratory Endpoints

eTable 2. Subsequent Therapy

eTable 3. Grade 3 or Higher Adverse Events Reported in More Than 5% of Patients in Either Treatment Group

Supplement 4.

Data Sharing Statement