Lack of Correlation between Epidermal Growth Factor Receptor Status and Response to Panitumumab Monotherapy in Metastatic Colorectal Cancer

Although most patients with colorectal cancer will be cured, nearly 50,000 deaths were estimated to have occurred in 2009 in the United States from this disease (1). The development of therapeutic antibodies to epidermal growth factor receptor (EGFR) such as panitumumab, a fully human IgG2 monoclonal antibody, and cetuximab, a chimeric monoclonal antibody, has resulted in new options for patients with refractory disease. EGFR is a transmembrane receptor tyrosine kinase with multiple ligands that promotes cell growth and survival in both normal and malignant cells (2). EGFR expression has been observed in numerous types of cancer, including colon, lung, head and neck, ovarian, and renal cell carcinomas (3).

Multiple trials have shown response rates of 5% to 12% with these agents used as monotherapy for advanced disease in unselected patients with metastatic colorectal cancer (4–9), but the majority of patients derive no clinical benefit. From the earliest development of anti-EGFR therapies, there have been attempts to identify those patients most likely to respond. Retrospective evidence suggests that EGFR expression is prognostic, with high levels of EGFR expression associated with shorter survival in patients with colorectal cancer (10). A rational approach, therefore, was to select patients whose tumors expressed higher EGFR levels for treatment with EGFR inhibitors. In fact, almost all early trials of such drugs excluded patients whose tumors did not express detectable EGFR by immunohistochemistry, as did the studies that led to Food and Drug Administration approval of panitumumab and cetuximab. As a result, both antibodies are approved only for patients whose tumors are EGFR positive by immunohistochemistry (11, 12). Most colorectal tumors, however, have some detectable EGFR and there seems to be no clear correlation between level of expression and response in early trials of anti-EGFR antibody therapies (5, 13). Therefore, we designed these linked phase II studies to examine the safety and efficacy of panitumumab monotherapy in chemorefractory metastatic colorectal cancer patients whose tumors expressed all levels of EGFR by immunohistochemistry.

After completion of these studies, analyses of data from the pivotal phase III trial of panitumumab have shown that the mutational status of KRAS, an oncogene encoding a GTP-binding protein involved in the EGFR signaling pathway, is predictive of response in patients with refractory metastatic colorectal cancer (14). Therefore, we also retrospectively evaluated the association of KRAS mutational status and efficacy in our studies.

Two linked, phase II, multicenter, single-arm, open-label studies evaluated the safety and efficacy of panitumumab administered as monotherapy to patients with metastatic colorectal cancer. A key enrollment criterion of both studies was the percentage of EGFR-positive cells detected by immunohistochemistry. The High EGFR study sequentially enrolled patients whose tumors expressed EGFR on ≥10% of tumor cells (high EGFR). Patients who were ineligible to participate in this study because of low (1-9%) or negative (<1%) EGFR expression could be consented and screened for the Low/Negative EGFR study. Aside from EGFR status, eligibility for the two studies was identical.

Patients received panitumumab 6 mg/kg every 2 wk until disease progression or intolerance. Patients who discontinued treatment underwent a safety follow-up visit 4 wk after the last infusion of panitumumab. All patients were followed for up to 24 mo after the date of the first panitumumab infusion to assess disease status and survival.

The study protocols were approved by the institutional review boards, and all patients signed a written consent form before initiation of study-specific screening procedures. The trials were registered under the U.S. NIH ClinicalTrials.gov identifiers NCT00083616 and NCT00089635.

Eligible patients were ≥18 y old, capable of providing informed consent, and had the following: a pathologic diagnosis of colorectal adenocarcinoma; Eastern Cooperative Oncology Group (ECOG) status of 0, 1, or 2; documented evidence of progression of metastatic disease during or following treatment with a fluoropyrimidine and a predefined minimum dose-intensity of both irinotecan (≥65 mg/m²/wk over ≥8 consecutive wk) and oxaliplatin (30 mg/m²/wk over ≥6 consecutive wk); and measurable disease. Patients had received two to three prior chemotherapy regimens for metastatic colorectal cancer. Adequate hepatic, renal, and hematologic function was required. Prior radiotherapy was allowed; however, index lesions could not have been chosen from a previously irradiated field. Receipt of bevacizumab prior to 6 wk before enrollment was allowed. Key exclusion criteria included receipt of systemic chemotherapy or radiotherapy within 30 d of enrollment, prior anti-EGFR antibody therapy, or evidence of a prior cancer within 5 y.

EGFR expression was assessed on tumor samples prior to enrollment. Immunohistochemistry to detect EGFR was conducted at a central laboratory (LabCorp Research Triangle Park, Research Triangle Park, NC) using the DakoCytomation EGFR PharmDx kit (DakoCytomation).

KRAS status assessments were done by HistoGeneX on DNA extracted from archived paraffin-embedded tumor tissue. Mutant (MT) KRAS sequences were detected using a K-RAS Mutation Test Kit (DxS Ltd) that used allele-specific real-time PCR (14).

Panitumumab was administered i.v. at 6 mg/kg once every 2 wk over approximately 60 min. Patients who developed mild to moderate skin toxicity were allowed to receive concomitant medications for skin symptoms (e.g., topical agents or oral antibiotics) at the discretion of the investigator. Panitumumab was withheld in patients who developed severe skin toxicity including the following: symptomatic skin-related toxicity requiring narcotics, systemic steroids, or felt to be intolerable by the patient; skin infection requiring systemic i.v. antibiotic or i.v. antifungal treatment; need for surgical debridement; any skin-related serious adverse event (AE). Panitumumab therapy was reinitiated at 50% of the original dosage after a 2-wk break and patients could receive the original dosage if improvements in the skin-related event were maintained. Patients with persistent, severe, skin-related toxicities were discontinued from the studies.

The primary efficacy end points of both studies were the effect of panitumumab on the objective response rate (complete response or partial response) through week 16 and the duration of response. Key secondary efficacy end points included best objective response, progression-free survival (PFS), and overall survival (OS). An exploratory analysis of outcomes by KRAS status was carried out.

Key safety end points included the incidence of AEs, changes in laboratory values, and the incidence of anti-panitumumab antibody formation.

Tumor response was assessed using a modification of the WHO criteria (bidimensionally measurable disease was required to be ≥20 mm in at least one dimension) at weeks 8, 12, 16, 24, 32, 40, and 48, and every 3 mo thereafter until progression of disease (responses were confirmed at least 4 wk after response criteria were first met). Scans evaluated for disease response were read centrally by a panel of at least two blinded independent radiologists (RadPharm).

The primary analyses were done on the efficacy set of each study, which included all consented and enrolled patients who had been determined to be eligible by the Independent Eligibility Review Committee (IERC). The intent-to-treat (ITT) set included all consented and enrolled subjects. The KRAS ITT set comprised the subsets of patients for whom KRAS status was known. The KRAS efficacy set was the subset of patients in the KRAS ITT set who were determined to be eligible by the IERC. The safety analysis set included all consented and enrolled patients who received at least one dose of panitumumab.

The planned sample size for patients in the efficacy set was 300 patients for the High EGFR study and 150 patients for the Low/Negative EGFR study, which was calculated to produce a 9% objective response rate point estimate, based on an analysis of a prior phase II study (8). The planned sample sizes allowed a 95% probability of observing at least one specific type of any AE if the true incidence rate was 2% (Low/Negative EGFR study) or 1% (High EGFR study) for that event. The Low/Negative EGFR study completed accrual. When data became available from the randomized study comparing panitumumab and best supportive care to best supportive care alone (7), it was felt that there were enough data to warrant early closure of the High EGFR study.

Descriptive statistics are provided for patient demographics and baseline characteristics. Exact two-sided 95% confidence intervals (95% CI) were calculated (15) for rate of objective response through week 16. Kaplan-Meier methodology with 95% CIs (16) estimated time to event and duration of event outcomes. Patients still alive on study or lost to follow-up were censored at their last contact date.

The studies were conducted in the United States from March 2004 through December 2006. These two studies were designed to be conducted in parallel and the majority of study sites recruited patients into both studies. The accrual to the Low/Negative EGFR study was slightly more brisk than that to the High EGFR study. Of 632 patients screened for the High EGFR study, 185 patients were enrolled and 182 patients received panitumumab. The most frequent reason for screen failure in the High EGFR study was the inability to meet the requirement of ≥10% EGFR tumor expression as measured by immunohistochemistry (284 screen failures). In the Low/Negative EGFR study, 252 patients were screened and 203 patients were enrolled and received panitumumab. This included 11 patients (none were responders) for whom EGFR staining results were missing (n = 8) or for whom EGFR staining was ≥10% (n = 3).

In the High EGFR study, of the 185 patients enrolled, there were 182 (98%) in the safety analysis set, 142 (77%) in the efficacy set, 168 (91%) in the KRAS ITT set, and 130 (70%) in the KRAS efficacy set. Eligibility for inclusion in the efficacy set of the High EGFR study was not confirmed by the IERC for 43 patients because there was a lack of radiographic progression on prior therapy (n = 21), last chemotherapy failure occurred >6 months prior to study start (n = 4), or receipt of the prespecified dose and/or overall exposure of the defined prestudy chemotherapy regimen was not confirmed (n = 23). In the Low/Negative EGFR study, of the 203 patients enrolled, there were 158 (78%) in the efficacy set, 171 (84%) in the KRAS ITT set, and 134 (66%) in the KRAS efficacy analysis set. Forty-five patients in the Low/Negative EGFR study were not confirmed for inclusion in the efficacy set by the IERC because there was a lack of radiographic progression (n = 19), last chemotherapy failure occurred >6 months prior to study start (n = 2), or receipt of the prespecified dose and/or overall exposure of the defined prestudy chemotherapy regimen was not confirmed (n = 29).

Patient demographics and baseline disease characteristics are shown in Table 1. Demographics and baseline disease characteristics were balanced across groups with negative, low, or high EGFR expression levels, with a notable difference of more patients in the High EGFR study with a baseline ECOG status of ≥1 than in the Low/Negative EGFR study (68% versus 49%, respectively). Age, sex, and race were also balanced across both studies according to KRAS status (data not shown). Fewer patients in the High EGFR study with MT KRAS had rectal cancer (13%) compared with patients with MT KRAS in the Low/Negative EGFR study (25%). ECOG status was also slightly better in the patients with wild-type (WT) KRAS in both studies.

For both studies, the most common reason for ending treatment was disease progression [130 patients (72%) in the High EGFR study and 160 patients (79%) in the Low/Negative EGFR study]. Four patients in the High EGFR study and 10 patients in the Low/Negative EGFR study ended treatment because of an AE. The remaining patients ended treatment because of the following reasons in the High EGFR/Low/Negative EGFR studies: ineligibility determined (n = 1/1), consent withdrawn (n = 1/3), patient request (n = 2/4), administrative decision (n = 2/2), protocol-specified criteria (n = 11/2), death (n = 12/5), and other reasons (n = 18/16).

By blinded central radiology review, the percentage of subjects responding through week 16 was 4.2% (95% CI, 0.9-11.9) in patients with negative EGFR, 3.8% (95% CI, 0.8-10.7) in patients with low EGFR, and 3.5% (95% CI, 1.2-8.0) in patients with high EGFR. The overall response rate was 5.7% (95% CI, 2.6-10.5) in patients with low/negative EGFR and 4.2% (95% CI, 1.6-9.0) in patients with high EGFR. By investigator assessment, there were seven patients with low EGFR and one patient with negative EGFR with a partial response for an overall response rate of 5.1% (95% CI, 2.2-9.7) in the Low/Negative EGFR study. In the High EGFR study, there were six patients with a partial response identified by the investigators, for a response rate of 4.2% (95% CI, 1.6-9.0). The median duration of response was 22 weeks (range, 12-46) for the seven responders in the Low/Negative EGFR study and 14 weeks (range, 12-102) for the five responders in the High EGFR study.

All responders had tumor samples evaluable for KRAS testing, and all were WT for KRAS (Table 2). The response rate in patients with WT KRAS was 8.5% (95% CI, 3.7-16.1) and 5.8% (95% CI, 1.9-13.0) in the Low/Negative EGFR study and the High EGFR study, respectively. By local review, responses were also seen only in patients with WT KRAS.

A total of 81 patients (efficacy set) in both studies achieved stable disease as the best objective response, which included 50 (33%) patients in the Low/Negative EGFR study and 31 (22%) patients in the High EGFR study.

PFS was similar among all patients regardless of EGFR expression levels. Median PFS was 7.3 weeks (95% CI, 7.1-7.6) in patients in the High EGFR study and 8.1 weeks (95% CI, 7.4-11.0) in patients in the Low/Negative EGFR study. There was no difference in PFS in patients with negative (<1%) or low (1-9%) levels of tumor EGFR expression (Fig. 1). Median OS was 30.7 weeks (95% CI, 26.4-34.7) in patients in the High EGFR study and 37.6 weeks (95% CI, 35.1-45.0) in patients in the Low/Negative EGFR study. Median OS was also similar among patients with <1% tumor EGFR expression and patients with 1% to 9% tumor EGFR expression.

PFS and OS were longer in patients with WT KRAS than those with MT KRAS (Fig. 2). Median PFS was 15.0 weeks (95% CI, 11.3-20.1) and 9.6 weeks (95% CI, 7.1-15.0) in patients with WT KRAS in the Low/Negative EGFR and High EGFR studies, respectively. In contrast, median PFS was 7.1 weeks (95% CI, 7.0-7.4) and 7.1 (95% CI, 7.1-7.4) in patients with MT KRAS in the Low/Negative EGFR and High EGFR studies, respectively. Similarly, median OS was 54.0 weeks (95% CI, 38.7-60.0) and 32.7 (95% CI, 27.9-46.3) in patients with WT KRAS in the Low/Negative EGFR and High EGFR studies, respectively, compared with 29.1 weeks (95% CI, 23.9-35.9) and 27.0 (95% CI, 20.7-34.7) in patients with MT KRAS. Decreases in the size of index lesions compared with baseline were more frequent and greater in magnitude in patients with WT KRAS status in both studies by waterfall plot analysis (Fig. 3).

Of the 203 patients in the Low/Negative EGFR study safety analysis set, 197 (97%) had at least one AE, and the incidence of grades 3 or 4 AEs was 25% (n = 51) (Table 3). Serious AEs were reported in 61 patients (30%). Of the 182 patients in the High EGFR study safety analysis set, 169 (93%) had at least one AE, and the incidence of grade 3 or 4 AEs was 27% (n = 49). Serious AEs were reported in 61 patients (34%).

Treatment-related AEs were primarily skin related (Table 3). There were 192 patients with skin-related events in the Low/Negative EGFR study, with 27 (13%) grade 3, and 164 patients with skin-related events in the High EGFR study, with 30 (16%) grade 3. There were no grade 4 skin-related events. Hypomagnesemia of any grade was reported in 10% of all patients, but events of grades 3/4 were rare (<1%). There were two serious infusion-related reactions (grade 3) in the Low/Negative EGFR study and none in the High EGFR study.

Most deaths on study in both studies were directly attributed to the underlying malignancy (19 of 23 in the High EGFR study and 16 of 20 in the Low/Negative EGFR study). Of the remaining eight on-study deaths across both studies, most occurred in the setting of documented or suspected tumor progression. There were two on-study deaths in patients in whom there was no evidence of disease progression (one death was due to pneumonia in a patient with a history of obstructive lung disease, the second death was due to diffuse ischemic stroke and myocardial infarction in a patient with a history of thrombotic events).

The results of these two studies are the most definitive data to date that EGFR staining does not predict efficacy of the EGFR inhibitor panitumumab. In fact, the results of both studies are consistent with those of other trials of single-agent panitumumab (7–9). The drug was generally well tolerated and, as expected, the most common side effects were skin related. Severe infusion reactions were rare (occurring in approximately 1% of patients) despite the fact that many sites, including that of the principal investigator of the High EGFR study, were within a geographic region that seems to have an excess risk of severe cetuximab infusion reactions (17).

The response rates in both trials were slightly lower than in other panitumumab studies, although confidence intervals were overlapping. Furthermore, the tumor control rates of 26% and 36% in the High EGFR and Low/Negative EGFR studies are similar to other studies (7, 8), including a subsequent randomized phase III trial of panitumumab monotherapy with similar inclusion criteria and in which EGFR staining on ≥1% of tumor cells was required. That study reported a response rate of 10% and a disease control rate of 37% in unselected patients (7).

The efficacy results also seem similar to published reports of cetuximab monotherapy (4, 5). The overall response rates of 5.7% in patients with low/negative EGFR and 4.2% in patients with high EGFR with median PFS time of 7.3 to 8.1 weeks and median OS time of 30.7 to 41.8 weeks for single-agent panitumumab are similar to results of the BOND trial, in which single-agent cetuximab had a response rate of 8.5% in patients who had received both irinotecan and oxaliplatin with median PFS and OS times of 6.5 weeks and 28 weeks, respectively, for the overall group (5). In that study, the percentage of EGFR-positive tumor cells also did not correlate with clinical response rate to an EGFR-inhibitor.

The relationship between EGFR expression and treatment and outcome is complex. There is clearly no major difference in response rates to treatment with panitumumab by level of EGFR staining. In addition, responses were seen in patients who had no detectable staining for EGFR by immunohistochemistry. These results confirm, in a standardized fashion, the anecdotal reports of efficacy of anti-EGFR antibodies in EGFR "negative" patients (18). Furthermore, EGFR testing should not be used to deny patients treatments with these antibodies. Unfortunately, EGFR staining does not identify a group of patients more likely to benefit from therapy. In fact, the overall PFS and OS in the High EGFR study were somewhat lower than those seen in the Low/Negative study, although the confidence intervals overlap. Potential explanations for this include imbalances between the two trials in patient characteristics (particularly in performance status), as well as biological differences between tumors that express different levels of EGFR. EGFR expression has been shown to be an adverse prognostic factor in colorectal cancer (19).

Although EGFR testing has no clinical utility, these data instead confirm recent reports that KRAS is an important predictive marker for patients with colorectal cancer treated with EGFR-inhibiting monoclonal antibodies (14, 20–24). As in these studies, all patients with objective responses and the majority with stable disease had tumors expressing WT KRAS. In addition, among patients with stable disease, many with WT KRAS tumors had minor tumor reductions, whereas almost no patients with MT KRAS had tumor shrinkage. Our studies also support the need for collection of tissues for biomarker testing in clinical trials conducted in cancer patients.

In conclusion, the findings of these two studies of single-agent panitumumab confirm the activity and toxicity profiles presented in previous reports (7–9). The main toxicities were rash and hypomagnesemia, both of which are manageable. However, these studies provide further evidence that anti-EGFR antibody therapy has efficacy in tumors with high, low, and no expression of EGFR by immunohistochemistry, and should be limited to those patients whose tumors express WT KRAS. Future studies of panitumumab should focus on improving management of cutaneous side effects of EGFR inhibitors as well as increasing efficacy in patients with tumors expressing WT KRAS. Patients with tumors that express MT KRAS need new approaches to later-line therapy of metastatic colorectal cancer.

J.R. Hecht, M.A. Neubauer, P. Swanson, T. Lopez, and G. Buchanan have no disclosures. E. Mitchell is a consultant for Pfizer and Amgen and has received honoraria from Pfizer, Roche, Sanofi, and Amgen. H.A. Burris III has received honoraria from Bristol-Myers Squibb. M. Reiner and J. Gansert are compensated employees and shareholders of Amgen Inc. J. Berlin has received honoraria from Bristol-Myers Squibb and Imclone.

We thank the patients and their families and friends for participating in the study and the clinical study staff at all participating institutions, and the following individuals at Amgen Inc.: Shanna Stout, Jody Brown, and Paul D'Avirro for study management; Justin Ney and Nathan Koons for clinical data management; Kim Musgrave for programming; and Dr. Julia R. Gage, for editorial assistance on behalf of Amgen Inc.

Grant Support: Funding: Amgen Inc.

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.