Retreatment with anti-EGFR based therapies in metastatic colorectal cancer: impact of intervening time interval and prior anti-EGFR response

This retrospective study aims to investigate the activity of retreatment with anti-EGFR-based therapies in order to explore the concept of clonal evolution by evaluating the impact of prior activity and intervening time interval. Eighty-nine KRAS exon 2-wild-type metastatic colorectal patients were retreated on phase I/II clinical trials containing anti-EGFR therapies after progressing on prior cetuximab or panitumumab. Response on prior anti-EGFR therapy was defined retrospectively per physician-records as response or stable disease ≥6 months. Multivariable statistical methods included a multiple logistic regression model for response, and Cox proportional hazards model for progression-free survival. Retreatment anti-EGFR agents were cetuximab (n = 76) or cetuximab plus erlotinib (n = 13). The median interval time between prior and retreatment regimens was 4.57 months (range: 0.46-58.7). Patients who responded to the prior cetuximab or panitumumab were more likely to obtain clinical benefit to the retreatment compared to the non-responders in both univariate (p = 0.007) and multivariate analyses (OR: 3.38, 95 % CI: 1.27, 9.31, p = 0.019). The clinical benefit rate on retreatment also showed a marginally significant association with interval time between the two anti-EGFR based therapies (p = 0.053). Median progression-free survival on retreatment was increased in prior responders (4.9 months, 95 % CI: 3.6, 6.2) compared to prior non-responders (2.5 months, 95 % CI, 1.58, 3.42) in univariate (p = 0.064) and multivariate analysis (HR: 0.70, 95 % CI: 0.43-1.15, p = 0.156). Our data lends support to the concept of clonal evolution, though the clinical impact appears less robust than previously reported. Further work to determine which patients benefit from retreatment post progression is needed.

on anti-EGFR therapy. Mathematical modeling of such resistance suggested that subclones harboring the KRAS mutations were present in low frequency in the tumor before treatment [3]. This finding supports the theory that the mechanism of resistance to anti-EGFR agents may be from intratumor heterogeneity and clonal evolution via drug-selection [4]. Based upon this theory a treatment break after developing acquired anti-EGFR resistance may allow the dominant clone that is KRAS-wt to repopulate and render a tumor sensitive to anti-EGFR therapy.
Success with a retreatment strategy utilizing targeted therapy has been reported with other agents in different types of cancer, such as trastuzumab in breast cancer [5] and sunitinib in gastrointestinal stromal tumor [6]. Recent small studies have suggested a benefit from a retreatment strategy in colorectal cancer with the use of anti-EGFR therapy [7][8][9]. In a phase II study by Santini et al. 39 patients with KRAS exon 2-wt metastatic CRC who had previously progressed following an initial clinical benefit to cetuximab-based therapy, were retreated with cetuximab and irinotecan. Results demonstrated an overall response rate of 53.8 %, stable disease rate was 35.9 %, and the median progression-free survival was 6.6 months [8]. Metges et al. (PANERB trial) prospectively treated 32 KRAS wild-type metastatic CRC patients with cetuximab and irinotecan followed by panitumumab monotherapy after progression. In 11 patients who had previously responded to cetuximab and irinotecan, an objective response rate of 22 % to panitumumab, including a disease control rate (objective response plus stable disease) of 73 % was observed [10]. In heavily pretreated patients without acquired resistance to prior cetuximab-based regimens, panitumumab obtained 67 % disease control rate and 30 % objective response rate, with median PFS of 4.2 and median OS of 9.6 months [11].
In this study, we reviewed 89 patients with advanced KRAS exon 2-wt CRC who had progressed on anti-EGFR therapy and were subsequently retreated on an anti-EGFR containing phase I/II clinical trial. Our goal was to evaluate the impact of both prior anti-EGFR response and interval length from prior anti-EGFR therapy upon the outcome of patients retreated with anti-EGFR therapy.

Patient selection
Patients with KRAS exon 2 (codons 12 + 13)-wt CRC who had progressed on their previous anti-EGFR-based therapy (cetuximab or panitumumab) and subsequently received at least two doses of an anti-EGFR monoclonal antibody in the context of a phase I or phase I/II clinical trial at MD Anderson Cancer Center were eligible for analysis on or before 2/27/2013. Progression on prior anti-EGFR based therapy prior to retreatment clinical trial was based upon retrospective review of the medical records. As this was a retrospective study informed consent was waived by the MD Anderson Cancer Center Institutional Review Board.

Tissue samples and mutation analyses
All histology was centrally reviewed at MD Anderson. All tissue samples were obtained and molecularly tested as part of standard of care. Mutational results for KRAS exon 2 (codons 12 and 13) and when available extended KRAS, NRAS, BRAF V600E, and PIK3CA were recorded from standard of care mutational results done in accordance with the Clinical Laboratory Improvement Amendment (CLIA)-certified Molecular Diagnostic Laboratory within the Division of Pathology and Laboratory Medicine at MD Anderson. DNA was extracted from macro-dissected, paraffinembedded tumor sections and over the time period studied three testing methodologies were utilized. , BRAF V600E, and PIK3CA exon 9 and 20 hotspots was used. The lower limit of detection is 10 % for the first two methodologies and 5 % for the third. In a subset of cases additional PCR-based DNA sequencing was conducted for BRAF V600E (n = 54), NRAS (n = 12), and PIK3CA (n = 24).

Data collection
Clinical information included age, race, the date of initial diagnosis and staging, KRAS, NRAS, BRAF, and PIK3CA mutational status of the tumor specimen, prior treatment history, baseline Eastern Cooperative Oncology Group (ECOG) performance status (PS), serum albumin, serum lactate dehydrogenase (LDH), and number of tumor metastatic sites were collected at the initiation of anti-EGFR retreatment. Two individual reviewers worked independently on reviewing patient electronic medical records and crosschecking the collected data.
For patients treated on more than one anti-EGFR-based regimen before retreatment (n = 18), data from the last anti-EGFR-based therapy were used for analysis. Response prior to the retreatment clinical trial was defined as a radiographic response or stable disease ≥6 months determined by the treating physician's records. Responses on the retreatment clinical trial were prospectively determined for each clinical trial and categorized per RECIST 1.0 [14] or 1.1 [15] criteria. Clinical benefit on anti-EGFR retreatment clinical trial was defined as complete response (CR), partial response (PR), or stable disease (SD). Progression-free survival (PFS) was calculated as the time from the start of therapy to the first observation of disease progression or death, whichever occurred first. Patients without progression were censored on February 27, 2013.

Statistical analysis
Univariable analyses for clinical benefit/response and PFS included chi-square and log-rank tests, respectively. Multivariable analyses for response and PFS utilized a multiple logistic regression and a Cox proportional hazards model, respectively. Covariates included in the multivariable models were response on prior anti-EGFR treatment, interval between conclusion of previous anti-EGFR treatment and initiation of anti-EGFR retreatment, age, race, gender, PS, and Royal Marsden Hospital (RMH) prognostic score comprising points for serum LDH and albumin levels, and number of metastatic sites [16]. The variables that were included in the multivariable model in the present study have been included in multivariable models in previous studies of phase 1 clinical trials [17,18]. All of these variables were included in the multivariable model so as to avoid confounding bias that could potentially result from exclusion of specific variables [19]. All statistical analyses were carried out using SPSS 19 (SPSS Chicago, IL) by our biostatisticians GG and KH.

Patient and treatment characteristics
We identified 97 KRAS exon 2-wt CRC patients who were treated on a phase I or phase I/II clinical trials containing anti-EGFR therapy and had progressed on prior cetuximab-or panitumumab-containing regimens from 5/2007 to 12/2012. An additional 8 patients (4 with NRAS mutations and 4 with BRAF V600E mutations) were excluded. The final analyzed dataset consisted of 89 patients, who were predominantly Caucasian (71 %), younger age (<60 years old, 64 %), and evenly distributed in gender, Table 1. At the initiation of the anti-EGFR rechallenge, they had good PS (ECOG < = 1, 94 %), normal albumin levels (81 %) and elevated LDH levels (79 %).
Prior anti-EGFR therapy was combined with chemotherapy in 90 % (80/89) and consisted of single agent anti-EGFR therapy in 9 % (8/89). Anti-EGFR therapy on the retreatment clinical trials utilized cetuximab in all cases. In the retreatment clinical trials, cetuximab was combined with chemotherapy and a targeted therapy in 73 % (65/89) and with a targeted therapy alone in 27 % (24/89), Table 2. Prior response, defined as response or stable disease ≥6 months, was seen in 41.6 % (37/89) of patients on prior anti-EGFR-based therapies. The median interval time between prior and retreatment anti-EGFRbased therapy was 4.6 months (range: 0.46-58.7). No additional RAS pathway mutations were identified in the 70 patients tested for KRAS exons 3 or 4, or for the 23 patients tested for NRAS. A BRAF V600E mutation was

Improved clinical benefit to cetuximab-based clinical trial retreatment in prior responders
Of the 86 patients with response information on the cetuximab-based retreatments, a clinical benefit, defined as a CR, PR, or SD, occurred in 58 % (50/86) of patients. Of the 13 patients who were treated with cetuximab and erlotinib, 1 had a PR, 4 SD, and 8 PD as best response. In univariate analyses, patients who responded to the prior anti-EGFR-based regimens were more likely to obtain a clinical benefit to cetuximabbased retreatment compared to the prior non-responders (p = 0.007), Table 3. In addition, a trend was noted where patients with longer (≥ median) interval length between prior and retreatment anti-EGFR therapy were more likely to respond to cetuximab-based retreatment compared with patients with shorter interval length (< median), p = 0.053. Other factors such as race (p = 0.14), age (p = 0.99), serum albumin (p = 0.95), LDH (p = 0.28), RMH prognostic score (p = 1.0), PS (p = 0.53) or number of metastatic sites (p = 0.49), were not statistically associated with obtaining a response to cetuximab-based retreatment, Table 3. In addition, patients were significantly more likely to respond to cetuximab-based retreatment if they were prior responders to anti-EGFR therapy and were retreated after a longer interval than prior non-responders after either a longer (p = 0.032) or a shorter interval, p = 0.003, Fig. 1.
A multiple logistic regression model confirmed that response on prior anti-EGFR therapy was a significant predictor for clinical benefit on anti-EGFR retreatment (OR: 3.38, 95 % CI: 1.27, 9.31, p = 0.019), Table 4. Despite the lack of statistical significance for the interaction of treatment response and interval length in the multivariate model, the markedly increased odds ratio for the combination of long interval length and prior treatment response (OR 6.7) in comparison to either long interval (OR 1.6) or prior treatment response alone (OR 2.3) suggests that such an interaction may exist but that we had insufficient data to detect it as statistically significant, Additional file 1: Table S1.

Discussion
This study shows anti-tumor activity with anti-EGFR retreatment in KRAS exon 2-wt CRC patients who had progressed on prior cetuximab-or panitumumab-based treatment. Prior responders were more likely to achieve  a clinical benefit on the cetuximab containing retreatment clinical trial. In addition a longer interval length between prior and retreatment anti-EGFR therapy demonstrated a non-significant trend favoring an increased likelihood of obtaining a clinical benefit from anti-EGFR retreatment. This study lends support to the notion of anti-EGFR retreatment in metastatic CRC, however the magnitude of benefit from retreatment appears less than previously reported [8,9].
In the phase II prospective study by Santini et al., cetuximab retreatment resulted in an overall response rate of 53.8 %, stable disease rate was 35.9 %, and the median progression-free survival was 6.6 months [8]. The median interval time between therapies was 6 months whereas in our study it was 4.6 months. Similar to Santini et al. a response to prior anti-EGFR therapy was defined as a response or stable disease lasting ≥ 6 months, however our definition was based upon clinical reports and not RECIST. Another phase II prospective study by Fora et al. reported benefit from EGFR retreatment with a higher dose of cetuximab (500 mg/m2 weekly) in combination with irinotecan in 20 KRAS-wt metastatic CRC patients who had previously progressed on both agents [9]. A clinical benefit was seen in 9 patients (1 PR and 8 SD) and in an exploratory analysis patients treated >2 months from prior cetuximab progression had an improved PFS, p = 0.02). In the PANERB trial that prospectively treated 32 KRAS wild-type metastatic CRC patients with cetuximab and irinotecan followed by panitumumab monotherapy after progression, an objective response rate of 22 % to panitumumab, including a disease control rate (objective response plus stable disease) of 73 % was observed in 11 patients who had previously responded to cetuximab and irinotecan [10]. Recent data has demonstrated that mutations in NRAS exons 2, 3 and 4, and KRAS exon 3 and 4, termed extended RAS testing, confer resistance to anti-EGFR therapy in metastatic CRC. [20][21][22] A recent meta-analysis has estimated the prevalence of KRAS exon 3 and 4 mutations to be 11 %, and NRAS mutations to be 9.1 % [23]. It has also been reported that a more sensitive technology may detect additional mutations that confer resistance to anti-EGFR therapies [24]. A fundamental limitation of this report is that due to the use of standard of care testing, we were not able to exclude all patients with innate anti-EGFR resistance due to extended RAS mutations, as testing was done for non-exon 2 KRAS mutations in 70 patients, 79 %, and for NRAS mutations in 23 patients, 26 %.
A number of recent reports utilizing circulating free DNA have correlated the occurrence of EGFR resistance with the acquirement of mutations in both KRAS and NRAS, as well as other acquired alterations such as EGFR mutation or MET amplification [3,25,26]. In Fig. 1 Prior responders with longer interval length were more likely to respond to anti-EGFR retreatment. Prior responders with longer interval length (longer intervening time between prior anti-EGFR therapy and anti-EGFR retreatment) were more likely to respond to anti-EGFR retreatment by analysis combining prior response to anti-EGFR retreatment and intervening time interval between anti-EGFR treatments: short (< median) or long (≥ median) Multivariable models were tested for a possible interaction between response on prior anti-EGFR therapy vs. non-response on prior therapy (1, 0) and interval length at or above the median vs. below the median (1, 0). The interaction term between response on prior anti-EGFR therapy vs. non-response on prior therapy (1, 0) and interval length at or above the median vs. below the median (1, 0) was not significant in either model. Thus, the interaction term was not included in the final multivariable models addition, the exact threshold for determining RAS mutational status is uncertain with data from the CRYSTAL study suggesting improved discrimination with a mutation threshold down to 5 % [27]. Within this report we are unable to address these various resistance mechanisms retrospectively as both tumor tissue and blood were inconsistently collected across the various clinical trials studied. However, these recent findings regarding acquired resistance mechanisms to anti-EGFR therapy provide clear insights into future studies to help refine and better predict which patients are truly benefiting from an anti-EGFR retreatment strategy. This study lends supports to the concept of intratumor heterogeneity and clonal evolution via drug selection as a mechanism of resistance to anti-EGFR agents. Clonal evolution generated by genetic instability and genetic drift is not a new concept [28]. Evidence of intratumor heterogeneity and branched evolutionary growth has been revealed in solid tumors via both tissue sections and multiregion sequencing [29][30][31][32]. Kreso et al., followed the repopulation dynamics of 150 single lentivirus-marked lineages from ten human CRCs through serial xenograft passages in mice using DNA copy number profiling, sequencing and lentiviral lineage tracking. This study showed that individual tumor cells within a uniform genetic lineage remained stable on serial transplantation, but were functionally heterogeneous with variable chemotherapy tolerance [33]. In metastatic CRC Morelli et al. analyzed circulating cell free DNA and tissue samples collected from EGFR refractory patients and demonstrated that the percent of acquired KRAS mutant alleles detected in plasma declined with greater time away from anti-EGFR therapy [26]. Repopulation of sensitive subclones after the cessation of treatment has been noted in a number of model systems [34,35]. Retreatment has demonstrated efficacy in multiple types of tumors in clinical trials [34,36].
In non-small cell lung cancer cell lines, evolutionary mathematical modeling of tumor behavior demonstrated that optimally timed sequential strategies yielded large improvements in survival outcome with anti-EGFR treatment [37,38].
Our study is limited by the use of heterogenous anti-EGFR retreatments. Effects of the combined regimens may obscure the differential effectiveness of the anti-EGFR retreatment. Due to the heterogeneity of included clinical trials and the sample size, the impact of each individual re-treatment regiment could not be determined. Although all anti-EGFR retreatment efficacy assessments were conducted on prospective clinical trials, the evaluations of prior anti-EGFR therapies were based on retrospective review, and due to the inability to review these scans response criteria were not based upon RECIST. In addition, the dose escalation of these phase 1 studies precluded a small subset of patients from receiving the P = 0.064 by log-rank test full dose of cetuximab. As this study utilized standard of care mutation testing the majority of patients in this study did not have extended RAS testing of NRAS and non-exon 2 KRAS. In addition highly sensitive mutational testing methodology were not utilized in these patients and thus the impact of low frequency RAS mutations could not be determined [24]. Thus, our results may be confounded by the presence of a small subset of patients with innate anti-EGFR resistance from preexisting mutations. Despite these limitations, this study represents the largest metastatic CRC anti-EGFR retreatment population published, and has attempted to evaluate both prior EGFR response and duration from prior therapy in contributing to anti-EGFR retreatment efficacy.

Conclusions
In conclusion, this study supports the further exploration of anti-EGFR retreatment in metastatic CRC, which is ongoing with both the CRICKET (NCT02296203) and REGAIN (NCT02316496) phase II clinical trials investigating retreatment with cetuximab following prior anti-EGFR progression. Understanding the mechanisms of acquired resistance to anti-EGFR therapy in CRC will enable the improved identification of patients who are likely to benefit from a retreatment approach. However, at the present time, rechallenge with an anti-EGFR therapy remains investigational and should be conducted in the context of a clinical trial.