Purpose:

The routine use of liquid biopsy is not recommended for the choice of initial treatment for patients with metastatic colorectal cancer (mCRC).

Experimental Design:

We included patients with left-sided, RAS/BRAF wild-type, HER2-negative, and microsatellite stable mCRC, treated with upfront panitumumab/FOLFOX-4 in the Valentino study. We performed amplicon‐based genomic profiling of 14 genes in baseline plasma samples and compared these data with tumor tissue ultra-deep sequencing results. Specific gene mutations in circulating tumor DNA (ctDNA) and their clonality were associated with progression-free survival (PFS), overall survival (OS), and radiological dynamics.

Results:

Ten and 15 of 120 patients had a mutation of RAS and PIK3CA in ctDNA, with a positive concordance with tissue deep sequencing of only 31.3% and 47.1%, respectively. Presence of RAS or PIK3CA mutations in baseline ctDNA was associated with worse median PFS [8 vs. 12.8 months; HR, 2.49; 95% confidence interval (CI), 1.28–4.81; P = 0.007 and 8.5 vs. 12.9 months; HR, 2.86; 95% CI, 1.63–5.04; P < 0.001] and median OS (17.1 vs. 36.5 months; HR, 2.26; 95% CI, 1.03–4.96; P = 0.042 and 21.1 vs. 38.9 months; HR, 2.18; 95% CI, 1.16–4.07; P = 0.015). RAS mutations in ctDNA were associated with worse RECIST response, early tumor shrinkage, and depth of response, while PIK3CA mutations were not. Patients with higher levels of RAS/PIK3CA variant allele fraction (VAF) in ctDNA had the worst outcomes (VAF ≥ 5% vs. all wild-type: median PFS, 7.7 vs. 13.1 months; HR, 4.02; 95% CI, 2.03–7.95; P < 0.001 and median OS, 18.8 vs. 38.9 months; HR, 4.07; 95% CI, 2.04–8.12; P < 0.001).

Conclusions:

Baseline ctDNA profiling may add value to tumor tissue testing to refine the molecular hyperselection of patients with mCRC for upfront anti-EGFR–based strategies.

Translational Relevance

In this prespecified exploratory analysis of the Valentino trial, next-generation sequencing of circulating tumor DNA (ctDNA) obtained from baseline liquid biopsies allowed the detection of new RAS or PIK3CA mutations in patients treated with upfront anti-EGFR therapy for metastatic colorectal cancer (mCRC). Patients with left-sided, hyperselected mCRC receiving panitumumab/FOLFOX-4 and with RAS/PIK3CA mutations in ctDNA had worse outcomes, particularly those with variant allele fraction ≥5%. Liquid biopsy may refine the molecular selection of patients with highest likelihood of benefit from EGFR blockade as compared with tumor tissue profiling alone and it may guide decisions on upfront therapy in patients with mCRC.

Anti-EGFR mAbs, cetuximab and panitumumab, are guideline-recommended treatments for patients with RAS/BRAF wild-type metastatic colorectal cancer (mCRC; ref. 1). Both primary tumor sidedness and molecular hyperselection beyond RAS and BRAF status allow to select patients with the highest likelihood of benefit from EGFR inhibition, that is, those with left-sided tumors and absence of rare genomic alterations associated with primary resistance [HER2/MET amplification, gene fusions, PIK3CA mutations, and microsatellite instability (MSI); refs. 2, 3]. Tumor tissue profiling still represents the mainstay for the detection of such alterations, but it is limited by spatial and temporal genomic heterogeneity.

In the setting of acquired resistance to EGFR blockade, liquid biopsy overcomes such limitations as it allows to capture the dynamics of clonal evolution with greater precision compared with tumor rebiopsies (4, 5). Specifically, the emergence of resistance alterations (more frequently involving RAS) under the selective pressure of therapy may be related to the expansion of preexisting resistant tumor subclones, while their decay during postprogression treatments sustains the use of liquid biopsy to select patients with potential benefit from anti-EGFR retreatment (6, 7).

Nevertheless, limited data are available on the role of liquid biopsy to predict the outcomes of patients clinically eligible for an anti-EGFR–based upfront treatment (ref. 8; Supplementary Table S1). Indeed, some mechanisms of primary resistance to anti-EGFR therapy, such as HER2 amplification or gene fusions, are usually clonal alterations with the role of oncogenic drivers and may be easily assessed in tumor tissue. Other alterations, such as RAS and PIK3CA mutations, may be subclonal, with a variant allele fraction (VAF) below 5% or even 1%, and therefore may be missed by several standard assays and only detectable by high-sensitivity techniques, such as next-generation sequencing (NGS) with high-depth coverage (3). On top of this, such mutations may also be highly heterogeneous from a spatial point of view, with the chance of being detected only by means of liquid biopsy. Therefore, genomic profiling of baseline circulating tumor DNA (ctDNA) may increase the detection of anti-EGFR resistance alterations and it may allow to quantify their clonality and VAF, with potential association to treatment outcomes (4, 8).

In this prespecified exploratory analysis of the Valentino trial (9), we aimed at assessing the added value of baseline ctDNA NGS profiling in patients with molecularly hyperselected, left-sided mCRC receiving a panitumumab-based upfront treatment strategy.

Patients’ cohort

The Valentino study (NCT02476045) was a multicenter, randomized, open-label phase II trial that investigated the progression-free survival (PFS) noninferiority of maintenance with panitumumab (arm B) versus panitumumab plus 5-FU/LV (arm A), following an induction treatment with panitumumab/FOLFOX-4 in patients with RAS wild-type mCRC (9). In patients signing an optional consent form, liquid biopsies were collected at baseline and every 8 weeks during treatment (independently of delays of treatment cycles) until disease progression, consent withdrawal, or death.

In this prespecified exploratory analysis, we included all clinically evaluable randomized patients with available baseline liquid biopsies, and molecularly hyperselected (RAS and BRAF wild-type based on standard assays, HER2-negative, and microsatellite stable status) and left-sided tumors, who were therefore highly enriched for EGFR-dependent cells in their tumors’ bulk.

RECIST 1.1 response, early tumor shrinkage (ETS), and depth of response (DoR) were assessed by blinded independent central review, as described previously (10). Available tumor profiling data included: HER2 status assessed by both IHC and silver ISH, MSI status by multiplex PCR, and presence of RAS mutations with low VAF (i.e., <5%, and therefore missed by several standard assays or usually filtered in standard NGS algorithms) and PIK3CA mutations by centrally performed ultra-deep NGS (3).

Institutional review board approval was obtained from all participating centers and all patients provided written informed consent. The study was conducted in accordance with the Declaration of Helsinki.

NGS of plasma samples

Cell-free DNA (cfDNA) was isolated and quantified from 1.5 mL of plasma using Maxwell RSC cfDNA Plasma Kit (Promega) and quantified using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific). Targeted libraries were performed using Oncomine Colon cfDNA Assay (Thermo Fisher Scientific) that amplifies 48 amplicons covering about 240 key hotspot mutations from 14 genes that are frequently mutated in colorectal cancers (AKT1, APC BRAF, CTNNB1, EGFR, ERBB2, FBXW7, GNAS, KRAS, MAP2K1, NRAS, PIK3CA, MAD4, and TP53). An amount of 2–50 ng of cfDNA was used to prepare targeted libraries; briefly, each cfDNA molecule was assigned with a unique molecular tag through a first PCR in a VeritiTM thermal cycler and subsequently, tagged library fragments were amplified in a second round of PCR to produce independent barcoded libraries. Sequencing was carried out on the Ion GeneStudioTM S5 prime system. For sensitive variant detection down to 0.1% VAF, we targeted a median read coverage > 25,000, median molecular coverage > 2,500, and targets > 0.8 median molecular coverage > 60%. Sequencing reads that shared the same molecular barcode information were grouped into a single family; to reduce potential false positives, a variant was called when at least two molecular barcode families shared the same mutation, corresponding to two independent mutated alleles.

Statistical analysis

No formal statistical hypothesis was done due to the exploratory nature of this study. PFS was defined as the time from randomization to documentation of progressive disease/death; overall survival (OS) was defined as the time from randomization to death from any cause. Both parameters were censored at the last follow-up for event-free subjects. χ2 and Fisher exact tests were used to test the distribution of categorical data. Mann–Whitney U test was used for the comparisons of continuous nonparametric data. Univariate Cox regressions were used to model right-censored variables. Unidimensional k-means algorithm was used to model the VAF cutoff of relevant resistance mutations and their correlation with outcomes. Data were imported and handled in R v3.6.1, using ggplot2, dplyr, survminer, survival, and finalfit packages (11).

Patients population

Supplementary Fig. S1 shows flow chart of the patients included in this study. A total of 120 patients with available and evaluable plasma samples had left-sided, RAS and BRAF wild-type, HER2-negative, and microsatellite stable (MSS) cancers. Figure 1A shows the heatmap of all the alterations detected in ctDNA in the final cohort. Regarding genes with putative role in resistance to anti-EGFR therapy, 10 and 15 patients had RAS and PIK3CA mutations in liquid biopsy, respectively, with only 4 patients among these showing concomitant RAS and PIK3CA mutations. Table 1 shows the specific RAS and/or PIK3CA mutations, as well as “founder”/clonal mutations, such as TP53 or APC, with the VAF of each mutation in ctDNA, and RAS and PIK3CA status in matching tumor tissue. Supplementary Fig. S2 reports the VAF distribution for each mutation, reaching the highest median value for TP53 and the lowest for RAS, while Supplementary Fig. S3 shows the VAF distribution of relevant resistance alterations in RAS and PIK3CA, which was used to build the VAF cutoff for survival analyses. Supplementary Table S2 shows the patients’ and disease baseline characteristics in the final study cohort, without significant differences according to RAS or PIK3CA mutational status in liquid biopsy.

Figure 1.

A, Heatmap representing the gene mutations found in liquid biopsy of patients with hyperselected mCRC included in the study cohort; bottom annotation shows the maximum VAF for each sample. B and C, The HRs for progression and death with respective CIs for each gene mutation (ns, nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001).

Figure 1.

A, Heatmap representing the gene mutations found in liquid biopsy of patients with hyperselected mCRC included in the study cohort; bottom annotation shows the maximum VAF for each sample. B and C, The HRs for progression and death with respective CIs for each gene mutation (ns, nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001).

Close modal
Table 1.

RAS mutations, PIK3CA mutations, and founder/clonal mutations with their corresponding VAF for each sample.

Patient IDPlasma status of KRAS exons 2,3, and 4Plasma status of NRAS exons 2,3, and 4Plasma status of PIK3CA (exons 9 and 20)Type of mutationPlasma KRAS, NRAS, or PIK3CA mA%Plasma-associated mutated genesType of mutationmA%Tumor tissue status of KRAS or NRAS (exons 2, 3, and 4) and PIK3CA (exons 9 and 20) by NGSSite of tissue sampleTime lapse between tissue and blood collection (months)
001–005 Wild-type Wild-type Mutated (exon 9) p.E545K (c.1633G>A) 0.87% TP53 (exon 8) p.R273H (c.818G>A) 8.52% Wild-type 1.3 
001–017 Wild-type Wild-type Mutated (exon 20) H1047R (c.3140A>G) 1.16% FBXW7 (exon 9) p.R505C (c.1513C>T) 12.58% Wild-type 1.5 
      TP53 (exon 8) p.C275Y (c.824G>A) 0.32%    
001–023 Wild-type Wild-type Mutated (exon 9) p.E542K (c.1624G>A) 0.30% None NA NA Wild-type 0.6 
001–028 Mutated (exon 2) Wild-type Mutated (exon 9) KRAS: G12V (c.35G>T) KRAS: 25.24% None NA NA KRAS: G12V (exon 2) PIK3CA: E545K (exon 9) 3.1 
    PIK3CA: p.E545K (c.1633G>A) PIK3CA: 23.23%       
001–031 Mutated (exon 2) Wild-type Wild-type G12D (c.35G>A) 17.80% None NA NA KRAS: G12D (exon 2) 0.9 
001–041 Wild-type Mutated (exon 2) Mutated (exon 9) NRAS: G13C (c.37G>T) NRAS: 0.02% TP53 (exon 7) p.T230fs (c.688dupA) 8.10% PIK3CA E545K (exon 9) 2.5 
    PIK3CA: p.E545K (c.1633G>A) PIK3CA: 53.18%       
001–043 Mutated (exon 4) Wild-type Wild-type p.A146T (c.436G>A) 1.12% APC (exon 15) p.Q1291* (c.3871C>T) 10.40% Wild-type 1.8 
      TP53 (exon 7) p.N235KfsTer12 (c.705_706delCTinsA) 30.58%    
004–010 Wild-type Wild-type Mutated (exon 9) p.E542K (c.1624G>A) 18.76% TP53 (exon 8) p.R267SfsTer73 (c.799_815delCGGAACAGCTTTGAGGTinsA) 66.76% PIK3CA: E542K (exon 9) 1.4 
      APC (exon 15) p.S932* (c.2795C>G) 19.26%    
004–025 Wild-type Wild-type Mutated (exon 20) H1047L (c.3140A>T) 34.80% TP53 (exon 5) p.R175H (c.524G>A) 58.64% PIK3CA: H1047L(exon 20) 1.7 
004–033 Wild-type Mutated (exon 3) Wild-type Q61H (c.183A>T) 3.86% TP53 (exon 7) p.S241del (c.721_723delTCC) 15.32% Wild-type 6.9 
007–002 Wild-type Wild-type Mutated (exon 9) p.E545K (c.1633G>A) 1.50% TP53 (exon 8) p.R282W (c.844C>T) 53.95% Wild-type 2.8 
      SMAD 4 (exon 9) p.D351V (c.1052A>T) 0.21%    
007–009 Wild-type Wild-type Mutated (exon 9) p.E542K (c.1624G>A) 17.05% APC (exon 15) p.Q1367* (c.4099C>T) 14.00% PIK3CA: E542K (exon 9) 1.9 
      TP53 (exon 5) p.P152L (c.455C>T) 17.11%    
008–008 Wild-type Wild-type Mutated (exon 9) p.E545K (c.1633G>A) 43.00% APC (exon 15) p.Q1406* (c.4216C>T) 22% PIK3CA: E454K (exon 9) 1.4 
      TP53 (exon 7) p.M246T (c.737T>C) 46.6%    
008–009 Wild-type Wild-type Mutated (exon 20) H1047R (c.3140A>) 26.23% BRAF (exon 15) p.V600E (c.1799T>A) 0.18% PIK3CA: H1047R (exon 20) 1.2 
      TP53 (exon 5) p.R175H (c.524G>A) 0.07%    
      TP53 (exon 8) p.R273C (c.817C>T) 25.11%    
008–010 Mutated (exon 4) Wild-type Wild-type A146T (c.436G>A) 0.25% ERBB2 (exon 8) p.L313I (c.937C>A) 0.22% Wild-type 2.4 
015–006 Mutated (exon 2) Wild-type Mutated (exon 20) KRAS: G12D (c.35G>A) KRAS:1% APC (exon 15) p.E1408* (c.4222G>T) 8% KRAS: G13S (exon 2) 20.7 
      TP53 (exon 5) p.V157F (c.469G>T) 1.43%    
    PIK3CA: H1047R (c.3140A>G) PIK3CA: 2% TP53 (exon 5) p.R175H (c.524G>A) 18%    
015–008 Wild-type Wild-type Mutated (exon 20) H1047V (c.3134A>T) and 10% TP53 (exon 8) p.R248Q (c.743G>A) 4.41% PIK3CA: D1045V (exon 20) and PIK3CA: H1047R (exon 20) 1.3 
    H1047R (c.3140A>G) 10%       
020–001 Mutated (exon 2) Wild-type Wild-type G12A (c.35G>C) 1.30% APC (exon 15) p. A1305QfsTer3 (c.3913delG) 43.20% Wild-type 2.5 
      TP53 (exon 7) p.G245S (c.733G>A) 38.92%    
024–001 Mutated (exon 2) Wild-type Mutated (exon 20) KRAS: G12D (c.35G>A) KRAS: 35.39% TP53 (exon 5) p.V157F (c.469G>T) 45.42% KRAS: G12D (exon 2); PIK3CA: H1047R (exon 20) 1.2 
    PIK3CA: H1047R (c.3140A>G) PIK3CA: 32.40%       
025–002 Mutated (exon 3) Wild-type Wild-type p.Q61H (c.183A>C) 1.19% None NA NA KRAS: Q61H (exon 3) 38 
025–006 Wild-type Wild-type Mutated (exon20) H1047R (c.3140A>G) 0.84% TP53 (exon 5) p.R158L (c.473G>T) 0.49% Wild-type 1.8 
      TP53 (exon 5) p.V157F (c.469G>T) 12.93%    
Patient IDPlasma status of KRAS exons 2,3, and 4Plasma status of NRAS exons 2,3, and 4Plasma status of PIK3CA (exons 9 and 20)Type of mutationPlasma KRAS, NRAS, or PIK3CA mA%Plasma-associated mutated genesType of mutationmA%Tumor tissue status of KRAS or NRAS (exons 2, 3, and 4) and PIK3CA (exons 9 and 20) by NGSSite of tissue sampleTime lapse between tissue and blood collection (months)
001–005 Wild-type Wild-type Mutated (exon 9) p.E545K (c.1633G>A) 0.87% TP53 (exon 8) p.R273H (c.818G>A) 8.52% Wild-type 1.3 
001–017 Wild-type Wild-type Mutated (exon 20) H1047R (c.3140A>G) 1.16% FBXW7 (exon 9) p.R505C (c.1513C>T) 12.58% Wild-type 1.5 
      TP53 (exon 8) p.C275Y (c.824G>A) 0.32%    
001–023 Wild-type Wild-type Mutated (exon 9) p.E542K (c.1624G>A) 0.30% None NA NA Wild-type 0.6 
001–028 Mutated (exon 2) Wild-type Mutated (exon 9) KRAS: G12V (c.35G>T) KRAS: 25.24% None NA NA KRAS: G12V (exon 2) PIK3CA: E545K (exon 9) 3.1 
    PIK3CA: p.E545K (c.1633G>A) PIK3CA: 23.23%       
001–031 Mutated (exon 2) Wild-type Wild-type G12D (c.35G>A) 17.80% None NA NA KRAS: G12D (exon 2) 0.9 
001–041 Wild-type Mutated (exon 2) Mutated (exon 9) NRAS: G13C (c.37G>T) NRAS: 0.02% TP53 (exon 7) p.T230fs (c.688dupA) 8.10% PIK3CA E545K (exon 9) 2.5 
    PIK3CA: p.E545K (c.1633G>A) PIK3CA: 53.18%       
001–043 Mutated (exon 4) Wild-type Wild-type p.A146T (c.436G>A) 1.12% APC (exon 15) p.Q1291* (c.3871C>T) 10.40% Wild-type 1.8 
      TP53 (exon 7) p.N235KfsTer12 (c.705_706delCTinsA) 30.58%    
004–010 Wild-type Wild-type Mutated (exon 9) p.E542K (c.1624G>A) 18.76% TP53 (exon 8) p.R267SfsTer73 (c.799_815delCGGAACAGCTTTGAGGTinsA) 66.76% PIK3CA: E542K (exon 9) 1.4 
      APC (exon 15) p.S932* (c.2795C>G) 19.26%    
004–025 Wild-type Wild-type Mutated (exon 20) H1047L (c.3140A>T) 34.80% TP53 (exon 5) p.R175H (c.524G>A) 58.64% PIK3CA: H1047L(exon 20) 1.7 
004–033 Wild-type Mutated (exon 3) Wild-type Q61H (c.183A>T) 3.86% TP53 (exon 7) p.S241del (c.721_723delTCC) 15.32% Wild-type 6.9 
007–002 Wild-type Wild-type Mutated (exon 9) p.E545K (c.1633G>A) 1.50% TP53 (exon 8) p.R282W (c.844C>T) 53.95% Wild-type 2.8 
      SMAD 4 (exon 9) p.D351V (c.1052A>T) 0.21%    
007–009 Wild-type Wild-type Mutated (exon 9) p.E542K (c.1624G>A) 17.05% APC (exon 15) p.Q1367* (c.4099C>T) 14.00% PIK3CA: E542K (exon 9) 1.9 
      TP53 (exon 5) p.P152L (c.455C>T) 17.11%    
008–008 Wild-type Wild-type Mutated (exon 9) p.E545K (c.1633G>A) 43.00% APC (exon 15) p.Q1406* (c.4216C>T) 22% PIK3CA: E454K (exon 9) 1.4 
      TP53 (exon 7) p.M246T (c.737T>C) 46.6%    
008–009 Wild-type Wild-type Mutated (exon 20) H1047R (c.3140A>) 26.23% BRAF (exon 15) p.V600E (c.1799T>A) 0.18% PIK3CA: H1047R (exon 20) 1.2 
      TP53 (exon 5) p.R175H (c.524G>A) 0.07%    
      TP53 (exon 8) p.R273C (c.817C>T) 25.11%    
008–010 Mutated (exon 4) Wild-type Wild-type A146T (c.436G>A) 0.25% ERBB2 (exon 8) p.L313I (c.937C>A) 0.22% Wild-type 2.4 
015–006 Mutated (exon 2) Wild-type Mutated (exon 20) KRAS: G12D (c.35G>A) KRAS:1% APC (exon 15) p.E1408* (c.4222G>T) 8% KRAS: G13S (exon 2) 20.7 
      TP53 (exon 5) p.V157F (c.469G>T) 1.43%    
    PIK3CA: H1047R (c.3140A>G) PIK3CA: 2% TP53 (exon 5) p.R175H (c.524G>A) 18%    
015–008 Wild-type Wild-type Mutated (exon 20) H1047V (c.3134A>T) and 10% TP53 (exon 8) p.R248Q (c.743G>A) 4.41% PIK3CA: D1045V (exon 20) and PIK3CA: H1047R (exon 20) 1.3 
    H1047R (c.3140A>G) 10%       
020–001 Mutated (exon 2) Wild-type Wild-type G12A (c.35G>C) 1.30% APC (exon 15) p. A1305QfsTer3 (c.3913delG) 43.20% Wild-type 2.5 
      TP53 (exon 7) p.G245S (c.733G>A) 38.92%    
024–001 Mutated (exon 2) Wild-type Mutated (exon 20) KRAS: G12D (c.35G>A) KRAS: 35.39% TP53 (exon 5) p.V157F (c.469G>T) 45.42% KRAS: G12D (exon 2); PIK3CA: H1047R (exon 20) 1.2 
    PIK3CA: H1047R (c.3140A>G) PIK3CA: 32.40%       
025–002 Mutated (exon 3) Wild-type Wild-type p.Q61H (c.183A>C) 1.19% None NA NA KRAS: Q61H (exon 3) 38 
025–006 Wild-type Wild-type Mutated (exon20) H1047R (c.3140A>G) 0.84% TP53 (exon 5) p.R158L (c.473G>T) 0.49% Wild-type 1.8 
      TP53 (exon 5) p.V157F (c.469G>T) 12.93%    

Abbreviations: M, metastasis; mA%, mutation load; NA, not applicable; NGS, next-generation sequencing; P, primary tumor.

Prognostic impact of mutations in ctDNA

Figure 1B and C resume the impact of each gene mutation detected in ctDNA in terms of PFS and OS, respectively. Regarding genes with putative role in resistance to anti-EGFR therapy, both PFS and OS were significantly worse in patients bearing a RAS mutation in ctDNA versus RAS wild-type subgroup [median PFS (mPFS), 8 vs. 12.8 months; HR, 2.49; 95% confidence interval (CI), 1.28–4.81; P = 0.007; and median OS (mOS), 17.1 vs. 36.5 months; HR, 2.26; 95% CI, 1.03–4.96; P = 0.042; Fig. 2A and B]. Similar results were observed in patients bearing a PIK3CA mutation in ctDNA versus wild-type subgroup (mPFS, 8.5 vs. 12.9 months; HR, 2.86; 95% CI, 1.63–5.04; P < 0.001 and mOS, 21.1 vs. 38.9 months; HR, 2.18; 95% CI, 1.16–4.07; P = 0.015; Fig. 2C and D). Despite the low numerosity, a trend for a worse PFS and OS could be observed in patients bearing simultaneously PIK3CA and KRAS mutations compared with patients carrying only one of them (Supplementary Fig. S4). Table 2 shows the impact of RAS and PIK3CA status in ctDNA in terms of tumor response dynamics; notably, the presence of RAS mutations in baseline ctDNA was significantly associated with overall response rate according to RECIST v1.1. and ETS, and trended to significance for association with DoR, while PIK3CA status was not associated with these parameters.

Figure 2.

Kaplan–Meier curves representing PFS and OS for, respectively, RAS mutated (mut) versus wild-type status in baseline ctDNA (A and B) and PIK3CA mutated versus wild-type status in baseline ctDNA (C and D).

Figure 2.

Kaplan–Meier curves representing PFS and OS for, respectively, RAS mutated (mut) versus wild-type status in baseline ctDNA (A and B) and PIK3CA mutated versus wild-type status in baseline ctDNA (C and D).

Close modal
Table 2.

Summary of overall response rate according to RECIST v1.1, ETS, and DoR according to RAS and PIK3CA mutational status in baseline ctDNA.

N (%)RECIST responseNo RECIST responseP§ETSNo ETSP§Median DoR % reduction (IQR)P*
RAS Wild-type 110 (91.7) 92 (86.0) 15 (14.0) 0.039 75 (75.0) 25 (25.0) 0.023 0.47 (0.38–0.60) 0.072 
 Mutated 10 (8.3) 5 (55.6) 4 (44.4)  3 (33.3) 6 (66.7)  0.17 (0.16–0.47)  
PIK3CA Wild-type 105 (87.5) 86 (85.1) 15 (14.9) 0.267 70 (74.5) 24 (25.5) 0.169 0.47 (0.38–0.59) 0.305 
 Mutated 15 (12.5) 11 (73.3) 4 (26.7)  8 (53.3) 7 (46.7)  0.43 (0.21–0.57)  
N (%)RECIST responseNo RECIST responseP§ETSNo ETSP§Median DoR % reduction (IQR)P*
RAS Wild-type 110 (91.7) 92 (86.0) 15 (14.0) 0.039 75 (75.0) 25 (25.0) 0.023 0.47 (0.38–0.60) 0.072 
 Mutated 10 (8.3) 5 (55.6) 4 (44.4)  3 (33.3) 6 (66.7)  0.17 (0.16–0.47)  
PIK3CA Wild-type 105 (87.5) 86 (85.1) 15 (14.9) 0.267 70 (74.5) 24 (25.5) 0.169 0.47 (0.38–0.59) 0.305 
 Mutated 15 (12.5) 11 (73.3) 4 (26.7)  8 (53.3) 7 (46.7)  0.43 (0.21–0.57)  

Note: RECIST response was available for 116 of the 120 patients; ETS and DoR were available for 109 of the 120 patients.

§χ2 test or Fisher test as appropriate.

*Mann–Whitney test.

Comparison of liquid biopsy and tumor tissue NGS

All 120 patients had available tissue NGS deep-sequencing data, as reported previously (3). Supplementary Fig. S5 shows the concordance between liquid biopsy and tissue NGS in terms of RAS (Supplementary Fig. S5A) and PIK3CA (Supplementary Fig. S5B) mutations. Overall concordance was 109 of 120 (90.8%) for RAS status and 111 of 120 (92.5%) for PIK3CA status; however, positive concordance in mutated samples (in liquid biopsy and/or tumor tissue) was only 5 of 16 (31.3%) for RAS mutations and 8 of 17 (47.1%) for PIK3CA mutations. Patients with RAS or PIK3CA mutations found only in tumor tissue and not in liquid biopsy were very few and their outcomes could not be investigated as a unique subgroup, although they did not seem to show poorer individual outcomes (Supplementary Fig. S6).

Role of RAS and PIK3CA VAF in liquid biopsy

Median VAFs of RAS and PIK3CA mutations in liquid biopsy were 1.3% [interquartile range (IQR), 1.1%–14.3%] and 17.1% (IQR, 1.3%–29.3%), respectively. Expectedly, the median VAF of such mutations in liquid biopsy was higher in patients with positive concordance with tissue NGS versus those with undetectable tissue mutations (median RAS VAF, 17.8% vs. 1.1% and median PIK3CA VAF, 27.8% vs. 1.2%; Mann–Whitney test P = 0.056 and P = 0.004; Fig. 3A and D). RAS VAF was also significantly associated with ETS (P = 0.011; Fig. 3B) and DoR (P = 0.073; Fig. 3C). Such associations for PIK3CA VAF status were not significant (Fig. 3E and F).

Figure 3.

A, Box plot depicting RAS VAF according to the concomitant detection of RAS mutation in the matched tumor tissue. B, Waterfall plot showing the relationship between ETS and RAS VAF. C, Scatter plot showing the correlation of RAS VAF with DoR. D, Box plot depicting PIK3CA VAF according to the concomitant detection of PIK3CA mutation in the matched tumor tissue. E, Waterfall plot showing the relationship between ETS and PIK3CA VAF. F, Scatter plot showing the correlation of PIK3CA VAF with DoR. mut, mutated; wt, wild-type.

Figure 3.

A, Box plot depicting RAS VAF according to the concomitant detection of RAS mutation in the matched tumor tissue. B, Waterfall plot showing the relationship between ETS and RAS VAF. C, Scatter plot showing the correlation of RAS VAF with DoR. D, Box plot depicting PIK3CA VAF according to the concomitant detection of PIK3CA mutation in the matched tumor tissue. E, Waterfall plot showing the relationship between ETS and PIK3CA VAF. F, Scatter plot showing the correlation of PIK3CA VAF with DoR. mut, mutated; wt, wild-type.

Close modal

Exploiting the VAFs cutoff derived upon their distribution, we then split patients with RAS and/or PIK3CA mutations with VAF of at least 5% in liquid biopsy (“clonal” subgroup) versus those with mutations with VAF < 5% (“subclonal” subgroup) and those without mutations in ctDNA. Compared with the all wild-type one, the “clonal” subgroup had the worst PFS (mPFS, 7.7 vs. 13.1; HR, 4.02; 95% CI, 2.03–7.95; P < 0.001; Fig. 4A) and OS (mOS, 18.8 vs. 38.9; HR, 4.07; 95% CI, 2.04–8.12; P < 0.001; Fig. 4B).

Figure 4.

A and B, Kaplan–Meier curves representing PFS and OS for, respectively, patients with at least one RAS/PIK3CA mutation VAF > 5% versus patients with RAS/PIK3CA mutation with VAF < 5% versus patients with RAS and PIK3CA all wild-type status in baseline ctDNA.

Figure 4.

A and B, Kaplan–Meier curves representing PFS and OS for, respectively, patients with at least one RAS/PIK3CA mutation VAF > 5% versus patients with RAS/PIK3CA mutation with VAF < 5% versus patients with RAS and PIK3CA all wild-type status in baseline ctDNA.

Close modal

Although genomic alterations of primary resistance to EGFR inhibition may be detected in ctDNA of patients with mCRC and potentially missed by genomic profiling of matched tumor tissue, their clinical significance and usefulness remain largely unknown. When focusing on RAS testing, pivotal cohort studies showed a good, but not perfect concordance between plasma and tumor genotyping (7, 12–19). In fact, despite the use of tumor profiling assays with high sensitivity for the detection of subclonal alterations, only the use of liquid biopsy may overcome the issue of spatial heterogeneity. On the opposite side, liquid biopsy may have meaningful limitations in patients with low tumor burden and poorly shedding cancers, such as those with peritoneal/lymph nodal lesions and mucinous histotype. Exploratory analyses of the aforementioned studies suggested that RAS testing in liquid biopsy results in similar clinical outcomes when compared with tissue testing in patients treated with anti-EGFR–based regimens. However, such evidence was limited by the small sample size, the variety of treatment lines and regimens, and, above all, the lack of survival comparison between subgroups (12, 16–18, 20). Moving forward from this background, Normanno and colleagues showed that, in a subgroup of 92 patients with KRAS exon 2 wild-type mCRC receiving upfront FOLFIRI plus cetuximab in the frame of CAPRI-GOIM trial, patients with RAS mutations detected in plasma or in tumor tissue had significantly worse outcomes versus wild-type counterparts, but without differences in the discriminative ability of plasma versus tissue testing (19).

With the aim of providing new evidence on the potential added value of baseline liquid biopsy genotyping in the first-line setting, we took advantage of the all-RAS and BRAF wild-type patients’ population enrolled in the Valentino study, who received panitumumab/FOLFOX-4 upfront strategy (9) and had been characterized previously by tissue deep sequencing (3). Moreover, with the aim of minimizing the potential confounding effects of the factors associated with primary resistance to EGFR inhibition, we focused only on hyperselected patients with left-sided, HER2-negative, and MSS cancers, who are currently considered as the optimal candidates for an upfront anti-EGFR–based regimen. We showed that patients with RAS or PIK3CA mutations in baseline ctDNA have worse outcomes in terms of PFS and OS compared with the wild-type subgroups, with RAS, but not PIK3CA, mutations being also associated with poorer response dynamics in terms of RECIST response, ETS, and DoR. It should be acknowledged that the lack of significant association of PIK3CA status with response parameters may be related to the low number of patients and to the confounding effects of the associated chemotherapy. Finally, although TP53 mutations are not an established marker of resistance to anti-EGFR therapy—they had a negative prognostic value in our cohort because these mutations showed the highest median VAF across all samples and the highest prevalence—we speculate that the detection of TP53 mutations in individual subjects may be at least partly correlated with the higher amount of ctDNA and/or the higher tumor burden.

Regarding the specific negative predictive role of PIK3CA mutations in mCRC, their association with other markers of primary resistance to anti-EGFR agents (such as RAS/BRAF mutations and right sidedness) has previously limited the clinical transferability of the results of initial investigations (21). Our study showed the potential clinical usefulness of liquid biopsy to assess PIK3CA status in mCRC, in light of the high degree of heterogeneity of PIK3CA mutations in tumor tissue and the available data in patients with other tumor types, including advanced breast cancer (22). Notably, the use of NGS assays for ctDNA profiling allowed us to concomitantly assess both PIK3CA and RAS status, as well as other genomic alterations with potential clinical actionability.

RAS or PIK3CA mutations with low fractional abundance detected only in tumor tissue by ultra-deep sequencing, but not in ctDNA, may be subclonal and, on top of this, highly heterogeneous from a spatial point of view. The very few patients with this profile did not seem to show poorer individual outcomes, but larger datasets are needed to investigate such patients as a unique subgroup. Consistent with these results, the ULTRA study showed that RAS mutations in tumor tissue with a VAF < 5% and only detectable through an ultra-sensitive assay (digital droplet PCR) were not associated with worse outcomes, further corroborating the current clinical use of validated assays with 5% limit of detection for RAS testing in tumor tissue (23). On the other side, mutations detected through ctDNA genotyping may mirror the presence of clinically important tumor subclones and could be associated with the rapid emergence of acquired resistance, rather than primary resistance to anti-EGFR drugs. In an attempt to further characterize this phenomenon, we showed that patients with RAS and/or PIK3CA VAF > 5% in cfDNA have the worst PFS and OS, while patients with RAS VAF > 5% also showed worse dynamics of response, including ETS and DoR. It should be pointed out that the VAF may be related to tumor burden and per-se prognostic. Moreover, VAF thresholds of resistance mutations in ctDNA, particularly after adjustment for the VAF of “founder” mutations (such as TP53 or APC) simultaneously identified by NGS, need validation through pooled analyses of clinical trials and larger datasets. Our 5% VAF cutoff, which warrants further validation in similar settings, is based upon the VAF intrinsic distribution and is close to the values reported in previous datasets (16, 24, 25).

Our study has the potential of clinical transferability in several settings. Prospective validation of our data may be achieved by randomized clinical trials investigating alternative upfront regimens or non-anti-EGFR–based maintenance strategies in patients with left-sided and molecularly hyperselected mCRC with baseline RAS mutations in ctDNA. Moreover, given the potential actionability of PIK3CA mutations when regarded as therapeutic targets rather than contributors to the puzzle of negative hyperselection, liquid biopsy–guided proof-of-concept trials or preclinical studies of anti-EGFR agents combined with alfa-selective PIK3 inhibitors, such as alpelisib, could be planned in the population with “isolated” PIK3CA mutations. Clear limitations of our study include its exploratory nature, the lack of data on longitudinal monitoring of serial liquid biopsy obtained during treatment, and the lack of prospective validation of the mutational VAF in ctDNA. Moreover, the study design did not allow to discriminate between the prognostic value of specific gene mutations in ctDNA and their potential predictive role, because both study arms were panitumumab based. On top of that, the low number of patients in the RAS/PIK3CA-mutated subgroups did not allow us to properly assess the interaction with the two treatment arms. Finally, in light of the poor prognostic role of APC wild-type status in mCRC (26, 27), we acknowledge that the low frequency of APC mutations in our dataset is clearly related to the relatively low coverage of the gene sequence by the hotspots included in this custom panel.

In conclusion, we showed that baseline ctDNA genomic profiling may add value to tumor tissue profiling to refine the molecular hyperselection of patients with the highest likelihood of benefit from upfront anti-EGFR–based strategies. Liquid biopsy data should be incorporated in future clinical trials with targeted agents conducted in patients with mCRC.

P. Manca reports grants from Amgen during the conduct of the study. S. Lonardi reports grants and personal fees from Amgen, Merck Serono, AstraZeneca, BMS, Lilly, and Roche; grants from Bayer; and personal fees from Incyte, Daiichi Sankyo, Servier, Pierre-Fabre, and GlaxoSmithKline outside the submitted work. A. Sartore-Bianchi reports personal fees from Amgen, Bayer, Sanofi, and Servier outside the submitted work. P. Racca reports grants from Merck and Amgen outside the submitted work. F. Morano reports personal fees from Servier outside the submitted work. F. Pietrantonio reports personal fees from Roche, Sanofi, Merck-Serono, Servier, Bayer, Lilly, and Amgen, and grants from BMS outside the submitted work. No disclosures were reported by the other authors.

P. Manca: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. Corallo: Conceptualization, resources, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. A. Busico: Data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. S. Lonardi: Data curation, validation, visualization, writing–review and editing. F. Corti: Data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. C. Antoniotti: Resources, data curation, visualization, writing–review and editing. L. Procaccio: Resources, data curation, visualization, writing–review and editing. M. Clavarezza: Resources, data curation, visualization, writing–review and editing. V. Smiroldo: Resources, data curation, visualization, writing–review and editing. G. Tomasello: Resources, data curation, visualization, writing–review and editing. R. Murialdo: Resources, data curation, visualization, writing–review and editing. A. Sartore-Bianchi: Resources, data curation, visualization, writing–review and editing. P. Racca: Resources, data curation, visualization, writing–review and editing. F. Pagani: Resources, data curation, investigation, visualization, writing–review and editing. G. Randon: Resources, data curation, investigation, visualization, writing–review and editing. A. Martinetti: Resources, formal analysis, investigation, writing–review and editing. E. Sottotetti: Resources, investigation, visualization, writing–review and editing. F. Palermo: Resources, software, visualization, project administration, writing–review and editing. F. Perrone: Resources, data curation, software, formal analysis, supervision, investigation, writing–review and editing. E. Tamborini: Resources, data curation, software, formal analysis, supervision, writing–review and editing. M. Prisciandaro: Resources, data curation, visualization, writing–review and editing. A. Raimondi: Resources, data curation, visualization, writing–review and editing. M. Di Bartolomeo: Resources, data curation, validation, visualization, writing–review and editing. F. Morano: Resources, data curation, funding acquisition, validation, investigation, visualization, project administration, writing–review and editing. F. Pietrantonio: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, writing–review and editing.

This was an academic study. A research grant and drug supply during panitumumab-based maintenance treatment was provided by Amgen. We thank all the patients who agreed to take part in the trial. We also thank the investigators and the study teams who participated: Alberto Zaniboni, Oncology Unit, Fondazione Poliambulanza (Brescia, Italy); Vincenzo Adamo, University of Messina and AO Papardo (Messina, Italy); Marco Tampellini, Department of Oncology, AOU San Luigi di Orbassano, University of Torino (Orbassano, Italy); Fausto Petrelli, Medical Oncology Unit, Department of Oncology, ASST Bergamo Ovest (Treviglio, Italy); Raffaella Longarini, Medical Oncology Unit, Azienda Ospedaliera San Gerardo (Monza, Italy); Monica Giordano, Medical Oncology Unit, Azienda Socio Sanitaria Territoriale Lariana (Como, Italy); Lorenzo Antonuzzo, Department of Medical Oncology, AOU Careggi (Florence, Italy); Alessandro Bertolini, Department of Medical Oncology, ASST della Valtellina e Alto Lago (Sondrio, Italy); Daniele Fagnani, Department of Medical Oncology, ASST di Vimercate (Vimercate, Italy); Enrico Cortesi, Department of Medical Oncology B, Policlinico Umberto I, “Sapienza” Università di Roma (Rome, Italy); Saverio Cinieri, Medical Oncology Unit, Ospedale Antonio Perrino (Brindisi, Italy); Nicla Maria La Verde, Department of Medical Oncology, Ospedale Fatebenefratelli e Oftalmico (Milan, Italy); Maria Giulia Zampino, Gastrointestinal Unit, Istituto Europeo di Oncologia (Milan, Italy); Mario Airoldi, Department of Medical Oncology, AOU Città della Salute e della Scienza di Torino (Turin, Italy); Graziella Pinotti, Department of Medical Oncology, Ospedale di Circolo (Varese, Italy); Samantha Di Donato, Department of Medical Oncology, Azienda USL Toscana Centro, Ospedale di Prato (Prato, Italy); Antonio Nuzzo, Department of Medical Oncology, Ospedale Civico Renzetti (Lanciano, Italy); Francesco Leone, Gastrointestinal Unit, Fondazione del Piemonte per l’Oncologia – IRCC di Candiolo (Candiolo, Italy); and Mario Roselli, Department of Medical Oncology, Policlinico Univeritario Tor Vergata (Rome, Italy).

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.

1.
Van Cutsem
E
,
Cervantes
A
,
Nordlinger
B
,
Arnold
D
,
ESMO Guidelines Working Group
. 
Metastatic colorectal cancer: ESMO clinical practice guidelines for diagnosis, treatment and follow-up
.
Ann Oncol
2014
;
25
:
iii1
9
.
2.
Cremolini
C
,
Morano
F
,
Moretto
R
,
Berenato
R
,
Tamborini
E
,
Perrone
F
, et al
Negative hyper-selection of metastatic colorectal cancer patients for anti-EGFR monoclonal antibodies: the PRESSING case-control study
.
Ann Oncol
2017
;
28
:
3009
14
.
3.
Morano
F
,
Corallo
S
,
Lonardi
S
,
Raimondi
A
,
Cremolini
C
,
Rimassa
L
, et al
Negative hyperselection of patients with RAS and BRAF wild-type metastatic colorectal cancer who received panitumumab-based maintenance therapy
.
J Clin Oncol
2019
;
37
:
3099
110
.
4.
Parikh
AR
,
Leshchiner
I
,
Elagina
L
,
Goyal
L
,
Levovitz
C
,
Siravegna
G
, et al
Liquid versus tissue biopsy for detecting acquired resistance and tumor heterogeneity in gastrointestinal cancers
.
Nat Med
2019
;
25
:
1415
21
.
5.
Pietrantonio
F
,
Vernieri
C
,
Siravegna
G
,
Mennitto
A
,
Berenato
R
,
Perrone
F
, et al
Heterogeneity of acquired resistance to anti-EGFR monoclonal antibodies in patients with metastatic colorectal cancer
.
Clin Cancer Res
2017
;
23
:
2414
22
.
6.
Cremolini
C
,
Rossini
D
,
Dell’Aquila
E
,
Lonardi
S
,
Conca
E
,
Del Re
M
, et al
Rechallenge for patients with RAS and BRAF wild-type metastatic colorectal cancer with acquired resistance to first-line cetuximab and irinotecan: a phase 2 single-arm clinical trial
.
JAMA Oncol
2019
;
5
:
343
50
.
7.
Siravegna
G
,
Mussolin
B
,
Buscarino
M
,
Corti
G
,
Cassingena
A
,
Crisafulli
G
, et al
Clonal evolution and resistance to EGFR blockade in the blood of colorectal cancer patients
.
Nat Med
2015
;
21
:
795
801
.
8.
Antoniotti
C
,
Pietrantonio
F
,
Corallo
S
,
De Braud
F
,
Falcone
A
,
Cremolini
C
, et al
Circulating tumor DNA analysis in colorectal cancer: from dream to reality
.
JCO Precis Oncol
2019
;
3
:
1
14
.
9.
Pietrantonio
F
,
Morano
F
,
Corallo
S
,
Miceli
R
,
Lonardi
S
,
Raimondi
A
, et al
Maintenance therapy with panitumumab alone vs panitumumab plus fluorouracil-leucovorin in patients with RAS wild-type metastatic colorectal cancer: a phase 2 randomized clinical trial
.
JAMA Oncol
2019
;
5
:
1268
75
.
10.
Manca
P
,
Corallo
S
,
Randon
G
,
Lonardi
S
,
Cremolini
C
,
Rimassa
L
, et al
Impact of early tumor shrinkage and depth of response on the outcomes of panitumumab-based maintenance in patients with RAS wild-type metastatic colorectal cancer
.
Eur J Cancer
2020
;
144
:
31
40
.
11.
R Core Team
.
R: a language and environment for statistical computing
.
Vienna, Austria:
R Foundation for Statistical Computing
;
2019.
12.
Vidal
J
,
Muinelo
L
,
Dalmases
A
,
Jones
F
,
Edelstein
D
,
Iglesias
M
, et al
Plasma ctDNA RAS mutation analysis for the diagnosis and treatment monitoring of metastatic colorectal cancer patients
.
Ann Oncol
2017
;
28
:
1325
32
.
13.
Thierry
AR
,
Mouliere
F
,
El Messaoudi
S
,
Mollevi
C
,
Lopez-Crapez
E
,
Rolet
F
, et al
Clinical validation of the detection of KRAS and BRAF mutations from circulating tumor DNA
.
Nat Med
2014
;
20
:
430
5
.
14.
Bachet
JB
,
Bouché
O
,
Taieb
J
,
Dubreuil
O
,
Garcia
ML
,
Meurisse
A
, et al
RAS mutation analysis in circulating tumor DNA from patients with metastatic colorectal cancer: the AGEO RASANC prospective multicenter study
.
Ann Oncol
2018
;
29
:
1211
9
.
15.
Schmiegel
W
,
Scott
RJ
,
Dooley
S
,
Lewis
W
,
Meldrum
CJ
,
Pockney
P
, et al
Blood-based detection of RAS mutations to guide anti-EGFR therapy in colorectal cancer patients: concordance of results from circulating tumor DNA and tissue-based RAS testing
.
Mol Oncol
2017
;
11
:
208
19
.
16.
Grasselli
J
,
Elez
E
,
Caratù
G
,
Matito
J
,
Santos
C
,
Macarulla
T
, et al
Concordance of blood- and tumor-based detection of RAS mutations to guide anti-EGFR therapy in metastatic colorectal cancer
.
Ann Oncol
2017
;
28
:
1294
301
.
17.
Thierry
AR
,
El Messaoudi
S
,
Mollevi
C
,
Raoul
JL
,
Guimbaud
R
,
Pezet
D
, et al
Clinical utility of circulating DNA analysis for rapid detection of actionable mutations to select metastatic colorectal patients for anti-EGFR treatment
.
Ann Oncol
2017
;
28
:
2149
59
.
18.
Thierry
AR
,
Pastor
B
,
Jiang
Z-Q
,
Katsiampoura
AD
,
Parseghian
C
,
Loree
JM
, et al
Circulating DNA demonstrates convergent evolution and common resistance mechanisms during treatment of colorectal cancer
.
Clin Cancer Res
2017
;
23
:
4578
91
.
19.
Normanno
N
,
Esposito Abate
R
,
Lambiase
M
,
Forgione
L
,
Cardone
C
,
Iannaccone
A
, et al
RAS testing of liquid biopsy correlates with the outcome of metastatic colorectal cancer patients treated with first-line FOLFIRI plus cetuximab in the CAPRI-GOIM trial
.
Ann Onco
2018
;
29
:
112
8
.
20.
Hsu
H-C
,
Lapke
N
,
Wang
C-W
,
Lin
P-Yi
,
You
JFu
,
Yeh
CY
, et al
Targeted sequencing of circulating tumor DNA to monitor genetic variants and therapeutic response in metastatic colorectal cancer
.
Mol Cancer Ther
2018
;
17
:
2238
47
.
21.
Perrone
F
,
Lampis
A
,
Orsenigo
M
,
Di Bartolomeo
M
,
Gevorgyan
A
,
Losa
M
, et al
PI3KCA/PTEN deregulation contributes to impaired responses to cetuximab in metastatic colorectal cancer patients
.
Ann Oncol
2009
;
20
:
84
90
.
22.
André
F
,
Ciruelos
E
,
Rubovszky
G
,
Campone
M
,
Loibl
S
,
Rugo
HS
, et al
Alpelisib for PIK3CA-mutated, hormone receptor-positive advanced breast cancer
.
N Engl J Med
2019
;
380
:
1929
40
.
23.
Santos
C
,
Azuara
D
,
Viéitez
JM
,
Páez
D
,
Falcó
E
,
Élez
E
, et al
Phase II study of high-sensitivity genotyping of KRAS, NRAS, BRAF and PIK3CA to ultra-select metastatic colorectal cancer patients for panitumumab plus FOLFIRI: the ULTRA trial
.
Ann Oncol
2019
;
30
:
796
803
.
24.
Janku
F
,
Zhang
S
,
Waters
J
,
Liu
Li
,
Huang
HJ
,
Subbiah
V
, et al
Development and validation of an ultradeep next-generation sequencing assay for testing of plasma cell-free DNA from patients with advanced cancer
.
Clin Cancer Res
2017
;
23
:
5648
56
.
25.
Pairawan
S
,
Hess
KR
,
Janku
F
,
Sanchez
NS
,
Mills Shaw
KR
,
Eng
C
, et al
Cell-free circulating tumor DNA variant allele frequency associates with survival in metastatic cancer
.
Clin Cancer Res
2020
;
26
:
1924
31
.
26.
Wang
C
,
Ouyang
C
,
Cho
M
,
Ji
J
,
Sandhu
J
,
Goel
A
, et al
Wild-type APC is associated with poor survival in metastatic microsatellite stable colorectal cancer
.
Oncologist
2020 Nov 23
[Epub ahead of print]
.
27.
Mondaca
S
,
Walch
H
,
Nandakumar
S
,
Chatila
WK
,
Schultz
N
,
Yaeger
R
. 
Specific mutations in APC, but not alterations in DNA damage response, associate with outcomes of patients with metastatic colorectal cancer
.
Gastroenterology
2020
;
159
:
1975
8
.