Pooled Analysis of Clinical Outcome of Patients with Chemorefractory Metastatic Colorectal Cancer Treated within Phase I/II Clinical Studies Based on Individual Biomarkers of Susceptibility: A Single-Institution Experience

Colorectal cancer (CRC) remains a significant cause of morbidity and mortality worldwide and many patients present with metastatic disease despite large-scale screening programs in various countries [1]. With the only exception of patients presenting with oligometastatic lesions confined to the liver or lung amenable to resection [2, 3], metastatic disease is considered incurable. In cases where treatment with curative intent is excluded, patients are typically given a combination of cytotoxic chemotherapy, often in conjunction with a targeted therapy.

Although advances in systemic therapy have been made, the 5-year survival rate in this setting is still 12.5% [4], with acquired resistance to therapy being the main reason for treatment failure [5]. Indeed, resistance to targeted therapy, as displayed by disease progression, is observed after a median of 4 months on epidermal growth factor receptor (EGFR) antagonists [45]. The discovery of RAS mutations in CRC as a mechanism of innate resistance to these therapies has been an important advance and has ameliorated their clinical use. However, there is an unmet need for effective therapeutic strategies after secondary resistance.

We have previously demonstrated that different molecular alterations that drive resistance can occur simultaneously in the same patient [7]. Identifying relevant molecular subtypes within this heterogeneous disease and matching patients with appropriate single agents or combinations of targeted therapies at resistance is crucial to therapeutic progress [8]. Therefore, recruitment into precision oncology clinical trials based on selection according to individual tumor molecular characteristics is expected to provide added value.

We retrospectively collected data from patients with metastatic CRC (mCRC) resistant to standard therapies treated at the Niguarda Cancer Center (NCC) (Milan, Italy) in phase I/II clinical studies based on the presence of specific tumor molecular profiles conferring susceptibility to experimental drugs, and performed a pooled analysis for measuring results according to main clinical and other molecular variables.

From June 2011 to May 2016, 2044 patients with mCRC were referred to NCC for molecular screening for the inclusion into phase I/II trials encompassing the targeting of actionable molecular alterations or molecular contexts of susceptibility. In total, 80 patients (3.9%) were enrolled into six ad hoc studies: one phase I/II and five phase II trials.

The median age was 60 years (range 36–86) and the median number of previous treatment lines was five (range 2–8). The six studies included therapies based on MGMT promoter hypermethylation (48.7%), HER2 amplification (28.8%), BRAF ^V600E mutation (20%), and gene fusions involving ALK or TRKA (2.5%) (Table 1 and Fig. 1). Among the 78 of 80 patients evaluable for KRAS mutations, any KRAS (exon 2) mutation was found in 30 (38.5%) of patients.

According to RECIST 1.1 criteria, one patient (1%) had complete response (CR), 13 patients (16.5%) had partial response (PR), and 28 (35%) had stable disease (SD), accounting for a 52.5% disease control rate (DCR = CR + PR + SD). The DCR was higher in patients with KRAS wild-type tumors (66 vs. 38%, p = 0.02). The median PFS was 2.8 months (range 2.63–3.83) (Fig. 2a), and 24% of patients displayed a PFS >5 months. A multivariate analysis of clinicopathological factors including age, gender, performance status, carcinoembryonic antigen (CEA) levels, and KRAS status showed that KRAS status and age were significantly associated with PFS (p = 0.011 and 0.047, respectively), whereas CEA and age were significantly associated with OS (p = 0.0063 and 0.0279, respectively). Patients with KRAS wild-type tumors had longer PFS than those with mutated tumors (3.80 [95% CI 2.80–5.03] vs. 2.13 months [95% CI 1.77–2.87], respectively) (p = 0.001) (Fig. 2b). Median OS was 7.83 months (range 7.17–9.33) (Fig. 3a), and patients with KRAS wild-type tumors had longer OS than those with mutated tumors (7.83 [95% CI 7.33–10.80] vs. 7.18 months [95% CI 5.63–9.33], respectively, p = 0.06) (Fig. 3b). The median GMI was 0.85 (range 0–15.61) and 32.5% of patients had a GMI >1.33 (Fig. 4). The Spearman Rho between PFS_n and PFS_n–1 was 0.10 (p = 0.372).

In this single-institution retrospective analysis we found that 4.0% of mCRC patients refractory to standard therapies were eligible for enrollment into phase I/II clinical trials based on molecular target selection. These patients, who had experienced a median number of five lines of previous treatments, achieved a median PFS of 2.8 months (with approximately one-quarter of patients displaying a PFS >5 months) and a median OS of 7.8 months. These data should be evaluated in the context of advanced mCRC, where the median PFS with newer agents has been less than 2 months [20], and rechallenge with standard chemotherapy [18, 19] or further chemotherapy [17] showed only poor efficacy and was burdened by adverse effects [21]. With regards to objective response rates (ORRs), 14 patients (17.5%) had PR and 28 (35%) SD, accounting for a 52.5% DCR. These data compare favorably with those achieved with licensed third-line therapies such as the multikinase inhibitor regorafenib (ORR = 1–4%) and the trifluridine–tipiracil combination TAS-102 (ORR = 2%), both of which are administered without the selection of a molecular target, or salvage chemotherapy with mitomycin-C or oxaliplatin rechallenge [17, 22, 23].

Following the seminal work by von Hoff et al. [14], several reports investigating the use of molecularly targeted agents on the basis of the identification of molecular alterations provided conflicting results [43, 44]. Provocative findings from Janku et al. highlighted that in phase I studies the lack of matching between molecular alterations and targeted treatment provides only limited clinical benefit in CRC and other solid tumors [24]. Recently, the same authors reported findings of a prospective, single-center ‘navigation’ study conducted in 500 patients with several different refractory cancers who underwent comprehensive genomic profiling by sequencing of 236 genes, offering proof-of-concept for the utility of this approach in assigning therapy to patients with refractory malignancies, especially in those patients with multiple genomic aberrations for whom combination therapies could be implemented [25]. However, patients with CRC were only a small proportion of this cohort and therefore definitive conclusions could not be drawn for this histology. In mCRC, Dienstmann et al. reported a series of 68 patients treated with targeted agents in phase I trials based on molecular profiling, concluding that this approach did not confer a significant clinical benefit on the basis of a comparison between the median time to treatment failure n (TTF_n) of 7.9 weeks and TTF_n–1 of 16.3 weeks [26]. Similarly, in our present analysis of 80 patients, we observed that PFS_n was inferior to PFS_n–1 (12.0 vs. 15.6 weeks). However, there are substantial differences between these two cohorts. Our series included only 2.5% patients treated within a phase I study versus 100% of the patients included in the study by Dienstmann et al. [26], and there are differences in the molecular alterations on which the treatments are based (KRAS/BRAF/PIK3CA mutations, PTEN and MET expression vs. HER2, BRAF, MGMT, and gene fusions, respectively). Assessment of clinical benefit by means of the ratio between PFS with the actual targeted therapy and that achieved with the previous line of treatment, referred to as the GMI, has also been proposed [14]. When we applied this index to our cohort, we found that at least one-third of patients benefited from the enrollment in clinical trials with targeted agents matched to selected actionable molecular alterations, making our results meaningful.

Among the molecular abnormalities selected in our cohort, we included gene fusions involving the ALK and NTRK genes. These alterations have only recently been recognized as targets occurring at low prevalence in CRC [27]. In particular, in 2014 we reported the characterization of the TPM3-NTRK1 gene rearrangement as a recurring, although rare, event in CRC. The concomitant discovery of entrectinib (NMS-P626; RXDX-101) as a novel highly potent and selective pan-Trk inhibitor by the group of Ardini et al. [28] enhanced the interest in this target, with further development across different histologies and currently ongoing phase II studies, including in CRC [12, 27]. Interestingly enough, the two gene fusions involving NTRK and ALK described in the present cohort were previously unreported [29, 30], highlighting the challenge of a comprehensive molecular pathology that should not merely include gene fusions together with mutations and amplifications but also elucidate the exact fusion partner and oncogenic potential for treatment with specific inhibitors.

During the last decade, the assessment of KRAS and neuroblastoma RAS viral oncogene homolog (NRAS) gained increasing importance for an optimal management of mCRC [31]. Mutations in KRAS codons 12 or 13 (in exon 2) affect approximately 40% of patients with metastatic CRC [32]. About an additional 5% of CRC patients display mutations in KRAS exons 3 or 4, usually at codon 61 or 146, and another 5% of CRC patients have mutations in NRAS in exons 2, 3, or 4. Extended RAS status (KRAS and NRAS exons 2, 3, and 4 mutations) is the validated biomarker of response to anti-EGFR antibodies [31]. RAS mutant mCRC exhibit a significantly higher cumulative incidence of lung, bone, and brain metastasis [33‐35]. The prognostic role of RAS mutations is controversial, but increasing evidence shows a negative effect in the adjuvant and metastatic setting [36‐38]. In our cohort, KRAS mutations showed an overall negative prognostic value, since patients harboring KRAS-mutated tumors had shorter PFS and OS. Also, DCR was more frequent in patients with KRAS wild-type tumors. However, it should be taken into account that half of the selected phase I/II trials were designed for patients with KRAS wild-type tumors (Cmek2110 and  Cmek2116 for BRAF-mutated tumors, HER2 Amplification for Colo-rectaL Cancer Enhanced Stratification-HERACLES for HER2 amplified tumors) and most of the KRAS-mutated patients were treated in the two studies (DETECT and TEMECT with dacarbazine and temozolomide, respectively) that were not based on directly targeting an oncogenic driver but rather exploiting a molecular context of susceptibility.

Limitations of our study include the small sample size and potential sampling bias, the notion that the use of GMI as an index of clinical benefit in CRC has been criticized as there is a lack of correlation between PFS_n and PFS_n–1 that is the statistical assumption for applying this method [39, 40], and that KRAS-limited instead of extended RAS analysis was performed in our cohort.

In conclusion, our single-institution study indicates that in a heavily pretreated mCRC population about 4% of tumors display a potential actionable molecular context suitable for phase I/II trials with matched therapeutics. Comprehensive molecular selection is challenging, encompassing inclusion of actionable alterations occurring at low frequency in CRC such as HER2 amplification [10] and rare gene fusions involving the ALK and NTRK genes [41], but enhances therapeutic options to be exploited in the advanced setting. This knowledge, coupled with recent advancements in the understanding of acquired resistance mechanisms to targeted agents through liquid biopsy [8, 42] is expanding opportunities of precision oncology therapies for CRC.