Abstract
Introduction
The mitogen-activated protein kinase (MAPK) signalling network aberrations in metastatic colorectal cancer (mCRC) generate intrinsic dynamic effects and temporal variations that are crucial but often overlooked in clinical trial populations. Here, we investigate the time-varying impact of MAPK pathway mutation genotype on each treatment line’s contribution to the overall clinical course.
Methods
The PROMETEO study focused on mCRC patients undergoing second-line treatment at 20 hospitals. We evaluated genotypes and employed flexible models to analyse the dynamic effect of each mutation.
Results
We examined data derived from 1160 patients. The effects of KRAS G12C or G12V, and BRAF V600E are clearly time-varying, with unexpected consequences such as the deleterious effect of BRAF V600E vs other genotypes dissipating over time when subjects receive antiangiogenics, or KRAS G12V and G12C showing increasing aggressiveness over time. Thus, contrary to expectations, the 12-month survival rate from the second line for those who survived >6 months was 49.9% (95% CI, 32.7–67.3) for KRAS G12C and 59% (95% CI, 38.5–80.6) for BRAF V600E.
Conclusions
The dynamic perspective is essential for understanding the behaviour of tumours with specific genotypes, especially from the second line onward. This may be relevant in patient monitoring and treatment decision-making, particularly in cases with distinct mutations.
Subject terms: Oncogenes, Colorectal cancer
Introduction
The tumourigenesis of metastatic colorectal cancer (mCRC) is driven by the accumulation of genetic alterations in molecular pathways that regulate cell proliferation, apoptosis, and angiogenesis. Two mutations, located in oncogenes that operate downstream of the epidermal growth factor receptor (EGFR) gene, have prognostic and predictive effects, and specific inhibitors targeting them are currently being developed. Thus, mCRC harbours Kirsten Rat Sarcoma Viral Oncogene Homologue (KRAS) mutations in 50% of cases, particularly in right-sided colon tumours, with hotspots in codons 12, 13, and 61 [1, 2]. Codon 12 mutations are particularly aggressive and account for about two-thirds of cases [3], with p.G12D and p.G12V being the most common in mCRC; in contrast, in non-small cell lung cancer (NSCLC), the predominant alteration is p.G12C [4]. Moreover, the discovery of a new allosteric regulation site mediated by the cysteine residue of KRAS p.G12C (c.34G>T) spurred the development of direct inhibitors for this variant [5–7]. Although present in only 3–4% of mCRC cases [5, 8, 9], these strategies are significant as they can restore the sensitivity to anti-EGFR therapies [10]. The use of sotorasib and adagrasib has been assessed in heavily pretreated mCRC patients with the KRAS p.G12C mutation [11, 12]. These agents are currently being investigated in combination with anti-EGFR antibodies in second-line randomised clinical trials (RCTs) (NCT04793958, NCT05198934). In contrast, KRAS proteins harbouring mutations other than p.G12C are devoid of the cysteine substrate and possess a diminished intrinsic rate of GTP hydrolysis, which precludes the use of the same covalent inhibition tactic [13].
Meanwhile, mutations in B-rapidly accelerated fibrosarcoma (BRAF), primarily in codon 600 (p.V600E), occur in less than 10% of metastatic colorectal cancers (mCRC) and are a marker of poor prognosis. This mutation is more prevalent in right-sided colon tumours and in elderly women, and is found in over 40% of tumours with high microsatellite instability [14]. In contrast to melanoma, mCRC does not exhibit a response to monotherapy using BRAF inhibitors, which can be attributed to the heightened activation of the EGFR pathway, leading to an increase in intracellular proliferative signals [15]. The effectiveness of EGFR inhibitors and chemotherapy remains debated [16, 17]. In contrast, blocking the mitogen-activated protein kinase (MAPK) network may be a useful strategy. Thus, in the phase 3 BEACON RCT, the combination of a BRAF inhibitor (encorafenib) and cetuximab increased overall survival (OS) compared to standard chemotherapy and cetuximab in patients with mCRC harbouring BRAF p.V600E mutation who had progressed on one or two prior regimens [18]. Additionally, the phase 2 SWOG S1406 RCT suggested that progression-free survival (PFS) associated with the combination of irinotecan, vemurafenib, and cetuximab was similar to that obtained in the BEACON study with encorafenib and cetuximab, questioning the role of chemotherapy.
Overall, aberrant activation of the MAPK pathway plays a central role in the pathobiology of mCRC, and while progress has been made in understanding this process and developing treatments to counteract it, questions remain about the clinical impact of these mutations.
Considering this prognostic effect, a crucial aspect in interpreting clinical trials with MAPK pathway inhibitors is the evolution of patient populations over time, which can give rise to considerable dynamic effects at the clinical trial level. For instance, an analysis of three clinical trials found that tumours with BRAF mutations had worse OS [19]. However, much of this poor prognosis came from patients who experienced rapid deterioration during first-line treatment, with only 33% progressing to a second line. Likewise, a series from MD Anderson Cancer Center has indicated that patients with KRAS p.G12C have a lower objective response rate (ORR), PFS, and OS compared to nonG12C KRAS [8]. This corresponds with the reality that merely 62% and 36% of patients are considered eligible for second- and third-line therapies, respectively [20], potentially resulting in a survivorship bias caused by the extensive baseline heterogeneity and the selection of favourable cases in subsequent line trials. In this sense, while an unfavourable prognosis linked to the BRAF mutation has been detected in global CRC, second-line clinical trials struggle to identify this effect, and study findings are not always consistent. In the context of third-line clinical trials, the predictive significance of the BRAF mutation is often challenging to validate due to its low prevalence in such cohorts [21].
Taking these observations into account, our team posited that the impact of mCRC mutational status, encompassing KRAS p.G12C and BRAF p.V600E, might be dynamic, shaping the cancer’s progression in a part of its trajectory (early or late). Understanding these chains of events could be crucial in developing effective therapies. With this framework in mind, we assessed whether employing flexible modelling strategies could help unveil these interactions, better capturing and interpreting the predictive and prognostic effect of KRAS/BRAF mutational status throughout the course of the disease.
Methods
Design and patients
The PROMETEO study is a multi-centre, observational research initiative. It enrolled adult patients (≥18 years) diagnosed with mCRC at 20 participating Spanish hospitals between 2014 and 2020. All subjects included in the study were required to have received a minimum of one cycle of standard second-line systemic therapy, in accordance with international clinical guidelines [22]. Enrolment occurred irrespective of patients’ KRAS, NRAS, or BRAF mutation status, while individuals on non-standard experimental treatments were excluded from participation. In focusing on second-line endpoints, we oriented our recruitment around incident cases to minimise potential biases, notably the survivorship bias [23]. This technique, although necessitating a non-inferential view of first-line outcomes, offers a more stringent backdrop to interpret the second-line findings. Building on this premise, the main objective of this study is to assess the potential dynamic impact of KRAS p.G12C and BRAF mutations on patients with mCRC undergoing second-line treatment. The focus on this stage is due to it being less extensively studied, the presence of significant ongoing drug development activity, and addressing a setting where patient populations have already shifted from first-line treatment, facilitating comparison with other studies in the literature.
Protocol and study approval were granted by the Spanish Agency of Medicines and Medical Devices (AEMPS) (code AFM-AFL-2019-01), a lead research ethics committee, and local committees at each participating centre. The study adheres to the Good Clinical Practice guidelines of the International Conference on Harmonisation, the principles of the Declaration of Helsinki, and relevant local laws and regulations. Written informed consent was obtained from all alive patients during data collection process.
Variables and treatment
Data management was conducted through an electronic registry, www.prometeostudy.com, which includes filters and an online monitoring system to ensure data reliability and address any missing or inconsistent data (DC, PJ-F, and AFM). Recorded variables included baseline patient characteristics (age, sex, Eastern Cooperative Oncology Group performance status (ECOG-PS), and comorbidities), tumour features (histology, location, metastatic site and burden, mutational status, presence of microsatellite instability, and Lynch syndrome), laboratory variables before the initiation of second-line treatment (haematologic data, liver function tests, lactate dehydrogenase, and CEA marker), treatment administered, and follow-up until death.
RAS/BRAF mutations were analysed according to each centre’s clinical practice, regardless of the technique employed. In all cases, the strategy involved molecular testing using Sanger sequencing, next-generation sequencing, FISH-based techniques, or Idylla. No centralised analysis was performed, since the goal was not to depict the mutational landscape but to illustrate the dynamic results of precisely characterised mutations.
Administered second-line treatment regimens were grouped into the following categories: FOLFIRI + aflibercept, FOLFIRI + antiEGFR, FOLFOX + antiEGFR, Irinotecan-based chemotherapy + bevacizumab, oxaliplatin-based chemotherapy + bevacizumab, mono-chemotherapy + antiEGFR, mono-chemotherapy + bevacizumab, poly-chemotherapy, mono-chemotherapy, and non-chemotherapeutic monotherapy. Dosing and schedules were determined by the investigator according to the drug’s technical specifications, local protocols, and individual patient characteristics. Dose adjustments and delays were carried out due to toxicity or other reasons, and second-line treatment was discontinued for progression, toxicity, or other reasons.
Objective response rate (ORR) was assessed using RECIST1.1 criteria on computed tomography scans according to the clinical practice of each participating centre.
To robustly address potential biases, our primary focus was on the endpoints of PFS-2 and OS-2, gauged from the initiation of second-line therapy to disease progression or death, respectively. Censoring was applied for those without events by the study’s conclusion. Additionally, PFS-1, spanning from the onset of first-line treatment to progression, death or censoring, was included for contextual insight. The duration from the start of each treatment line to progression defined the time to progression (TTP-1 and TTP-2) endpoint. In light of possible interpretive complexities due to immortal time bias, OS-1 is not presented in this analysis. Recognising the occurrence of survival function crossovers, both median survival times and the 12-month survival rate from the second line have been incorporated as relevant endpoints.
Statistical analysis
The main objective of this study calls for the application of “flexible” models that enable dynamic or segmented phase analysis [24]. Annex Table 1 provides a detailed description of these models, emphasising the complexity of addressing the complete scope of the hypothesis. The Royston–Parmar spline model (RPSM) was used to describe time-varying hazard ratios [25]. The RPSM is generally considered a useful descriptive tool in dynamic settings. Candidate predictors were selected after an exhaustive literature search and consultation with experts from participating centres, taking into account the degrees of freedom supported by the data (at least 15 events per degree of freedom) [26]. All models were adjusted for mutational status, age, tumour location, hepatic tumour burden, number of metastatic sites, ECOG-PS, and the use of antiangiogenics in first and second line; for PFS-2/OS-2, the models also included progression to the first line and the time between first and second line. No data-driven method was used in the final selection of variables. To achieve a balance between flexibility and parsimony, the RPSM used three terms to model the effect of mutational status and two knots to model the baseline hazard. The RPSM allows for the estimation of time-varying HR, such that dynamic effects can be represented through time-dependent plots, while fixed effects are reported in standard hazard ratio tables. To improve clinical understanding and considering that the log hazards inverted their sign around 6 months (with crossing of survival curves), it was considered that a clinically relevant estimate was the 12-month survival rate from the start of the second line (OS-2), conditional on surviving at least 6 months. All interactions were tested. A fixed sample was used, conditioned on the available patients, which requires considering the width of the confidence intervals. Finally, the relative contribution of each endpoint (PFS-1, PFS-2) to the total was assessed using partitioned survival analysis (PSA) [27, 28]. The partitioned survival analysis traces the journey from the start of the first-line therapy to the initiation of the second-line, the duration of the second-line up to its cessation, and from the end of the second-line to the eventual death. The analysis is conditioned on subjects receiving a second-line treatment. To describe the treatment sequences at each health stage, Sankey plots were used. All analyses were carried out using R version 4.2.3 [29], including the survival, networkD3, hesim, and mstate libraries [30–33]. The R code is presented in Annex Table 1B.
Results
Characteristics and outcomes of the full cohort
The database comprised 1160 patients. Table 1 displays the baseline characteristics of the sample at the start of the second line, stratified by mutational status. The median age was 66 years, with a range of 18–86 years. Roughly 48% of patients had left-sided colon tumours, while 26%, 22%, and 4% had rectal, right-sided colon, and transverse colon tumours, respectively. In 51% (n = 594) of cases, no RAS/BRAF mutations were detected, whereas 42%, 3%, and 3% had mutations in the KRAS, NRAS, and BRAF genes, respectively. Among the KRAS mutations, codon 12 mutations were the most frequent (72%, n = 355), with the most prevalent being KRAS p.G12D, p.G12V, and p.G12C (8%, 7%, and 3%, respectively, corresponding percentages to the total). The full list of mutations can be found in Table 2.
Table 1.
Baseline characteristics according to mutational status.
| Characteristics | Total N (%) | BRAF mutation | KRAS mutation | NRAS mutation | Wild type |
|---|---|---|---|---|---|
| Total | 1160 (100) | 37 (100) | 492 (100) | 37 (100) | 594 (100) |
| Age (years), median (range) | 66.3 (18–86.7) | 66.2 (50.2–85.9) | 67.5 (30.7–86.7) | 66.8 (48.6–82.9) | 65.5 (18–86.7) |
| Sex, male | 741 (63.88) | 21 (56.76) | 310 (63.01) | 19 (51.35) | 391 (65.82) |
| ECOG performance status | |||||
| 0 | 333 (28.71) | 8 (21.62) | 126 (25.61) | 13 (35.14) | 186 (31.31) |
| 1 | 717 (61.81) | 25 (67.57) | 323 (65.65) | 22 (59.46) | 347 (58.42) |
| 2 | 108 (9.31) | 4 (10.81) | 43 (8.74) | 2 (5.41) | 59 (9.93) |
| 3 | 2 (0.17) | 0 (0) | 0 (0) | 0 (0) | 2 (0.34) |
| ≥2 chronic comorbidities | 568 (48.97) | 24 (64.86) | 233 (47.36) | 21 (56.76) | 290 (48.82) |
| Primary tumour site | |||||
| Left | 558 (48.1) | 8 (21.62) | 213 (43.29) | 19 (51.35) | 318 (53.54) |
| Rectum | 298 (25.69) | 6 (16.22) | 134 (27.24) | 11 (29.73) | 147 (24.75) |
| Right | 257 (22.16) | 20 (54.05) | 128 (26.02) | 7 (18.92) | 102 (17.17) |
| Transverse | 42 (3.62) | 3 (8.11) | 15 (3.05) | 0 (0) | 24 (4.04) |
| Unknown | 5 (0.43) | 0 (0) | 2 (0.41) | 0 (0) | 3 (0.51) |
| Grade | |||||
| 1 | 300 (25.86) | 8 (21.62) | 125 (25.41) | 11 (29.73) | 156 (26.26) |
| 2 | 578 (49.83) | 13 (35.14) | 258 (52.44) | 17 (45.95) | 290 (48.82) |
| 3 | 116 (10) | 11 (29.73) | 44 (8.94) | 6 (16.22) | 55 (9.26) |
| Unknown | 166 (14.31) | 5 (13.51) | 65 (13.21) | 3 (8.11) | 93 (15.66) |
| Mucinous differentiation | 122 (10.52) | 7 (18.92) | 64 (13.01) | 1 (2.7) | 50 (8.42) |
| Alteration of mismatch repair proteins | 30 (2.59) | 6 (16.22) | 8 (1.63) | 1 (2.7) | 15 (2.53) |
| Not tested | 367 (31.64) | 5 (13.51) | 138 (28.05) | 7 (18.92) | 217 (36.53) |
| Number of metastatic sites (organs involved) | |||||
| 1 | 227 (19.57) | 8 (21.62) | 85 (17.28) | 8 (21.62) | 126 (21.21) |
| 2 | 136 (11.72) | 4 (10.81) | 55 (11.18) | 4 (10.81) | 73 (12.29) |
| ≥3 | 797 (68.71) | 25 (67.57) | 352 (71.54) | 25 (67.57) | 395 (66.5) |
| Tumour size of the largest metastasis | |||||
| Low (<2 cm) | 301 (25.95) | 11 (29.73) | 126 (25.61) | 11 (29.73) | 153 (25.76) |
| Intermediate (2–4 cm) | 512 (44.14) | 15 (40.54) | 231 (46.95) | 16 (43.24) | 250 (42.09) |
| High (>4 cm) | 347 (29.91) | 11 (29.73) | 135 (27.44) | 10 (27.03) | 191 (32.15) |
| Life-threatening metastases | |||||
| 0 | 1001 (86.29) | 29 (78.38) | 427 (86.79) | 35 (94.59) | 510 (85.86) |
| 1 | 159 (13.71) | 8 (21.62) | 65 (13.21) | 2 (5.41) | 84 (14.14) |
| Metastases sites | |||||
| Lung | 640 (55.17) | 17 (45.95) | 281 (57.11) | 25 (67.57) | 317 (53.37) |
| Liver | 822 (70.86) | 21 (56.76) | 354 (71.95) | 22 (59.46) | 425 (71.55) |
| Bone | 75 (6.47) | 5 (13.51) | 35 (7.11) | 1 (2.7) | 34 (5.72) |
| Peritoneum | 311 (26.81) | 14 (37.84) | 137 (27.85) | 11 (29.73) | 149 (25.08) |
| Ascites | 85 (7.33) | 4 (10.81) | 41 (8.33) | 1 (2.7) | 39 (6.57) |
| Distant lymph node metastases | 332 (28.62) | 12 (32.43) | 120 (24.39) | 9 (24.32) | 191 (32.15) |
| Brain | 12 (1.03) | 2 (5.41) | 3 (0.61) | 0 (0) | 7 (1.18) |
| Others | 80 (6.9) | 37 (100) | 33 (6.71) | 3 (8.11) | 44 (7.41) |
| Neutrophil–lymphocyte ratio, median (range) | 2.4 (0–923.1) | 2.5 (0.3–142.3) | 2.4 (0–923.1) | 2.2 (0.9–115.4) | 2.4 (0.3–307.7) |
| Albumin < LLN | 136 (11.72) | 9 (24.32) | 48 (9.76) | 2 (5.41) | 77 (12.96) |
| Alkaline phosphatase > ULN | 676 (58.28) | 17 (45.95) | 292 (59.35) | 20 (54.05) | 347 (58.42) |
| CEA ng/mL, median (range) | 19 (0–16,477) | 8.5 (2–137) | 19.5 (0–15,000) | 8 (0–2593) | 20 (0–16,477) |
| Surgery of primary tumour | 774 (66.72) | 29 (78.38) | 318 (64.63) | 26 (70.27) | 401 (67.51) |
| Surgery of metastasis | 36 (3.1) | 1 (2.7) | 11 (2.24) | 1 (2.7) | 23 (3.87) |
| First-line treatment | |||||
| Capecitabine | 29 (2.5) | 0 (0) | 17 (3.46) | 1 (2.7) | 11 (1.85) |
| Fluoropyrimidine + antiEGFR | 15 (1.29) | 0 (0) | 0 (0) | 0 (0) | 15 (2.53) |
| Fluoropyrimidine + bevacizumab | 17 (1.47) | 2 (5.41) | 10 (2.03) | 1 (2.7) | 4 (0.67) |
| FOLFIRI + antiEGFR | 59 (5.09) | 1 (2.7) | 2 (0.41) | 0 (0) | 56 (9.43) |
| FOLFIRI + bevacizumab | 44 (3.79) | 2 (5.41) | 24 (4.88) | 0 (0) | 18 (3.03) |
| FOLFIRI + aflibercept | 3 (0.26) | 0 (0) | 3 (0.61) | 0 (0) | 0 (0) |
| FOLFIRINOX + bevacizumab | 13 (1.12) | 4 (10.81) | 4 (0.81) | 0 (0) | 5 (0.84) |
| FOLFOX + antiEGFR | 221 (19.05) | 6 (16.22) | 2 (0.41) | 0 (0) | 213 (35.86) |
| Irinotecan-based polyCT | 51 (4.4) | 2 (5.41) | 25 (5.08) | 4 (10.81) | 20 (3.37) |
| Oxaliplatin-based + bevacizumab | 388 (33.45) | 11 (29.73) | 239 (48.58) | 22 (59.46) | 116 (19.53) |
| Oxaliplatin-based polyCT | 305 (26.29) | 9 (24.32) | 157 (31.91) | 8 (21.62) | 131 (22.05) |
| Others | 15 (1.29) | 0 (0) | 9 (1.83) | 1 (2.7) | 5 (0.84) |
| Second-line treatment | |||||
| CT monotherapy | 41 (3.53) | 2 (5.41) | 20 (4.07) | 0 (0) | 19 (3.2) |
| FOLFIRI + aflibercept | 343 (29.57) | 13 (35.14) | 169 (34.35) | 13 (35.14) | 148 (24.92) |
| FOLFIRI + antiEGFR | 107 (9.22) | 0 (0) | 0 (0) | 1 (2.7) | 106 (17.85) |
| FOLFOX + antiEGFR | 24 (2.07) | 1 (2.7) | 0 (0) | 0 (0) | 23 (3.87) |
| Irinotecan-based CT + bevacizumab | 232 (20) | 7 (18.92) | 128 (26.02) | 9 (24.32) | 88 (14.81) |
| MonoCT + antiEGFR | 36 (3.1) | 2 (5.41) | 0 (0) | 0 (0) | 34 (5.72) |
| MonoCT + bevacizumab | 19 (1.64) | 0 (0) | 5 (1.02) | 2 (5.41) | 12 (2.02) |
| Oxaliplatin-based CT + bevacizumab | 83 (7.16) | 6 (16.22) | 29 (5.89) | 2 (5.41) | 46 (7.74) |
| PolyCT | 255 (21.98) | 4 (10.81) | 134 (27.24) | 9 (24.32) | 108 (18.18) |
| Monotherapy, other | 20 (1.72) | 2 (5.41) | 7 (1.42) | 1 (2.7) | 10 (1.68) |
KRAS Kirsten rat sarcoma viral oncogene homologue, a gene that codes for a protein involved in cell signalling pathways, NRAS neuroblastoma RAS viral oncogene homologue, BRAF B-Raf proto-oncogene, serine/threonine kinase, G12D, G12V, G12C, G12A, G12S, G12R, G12X specific mutations of the KRAS gene that result in a change in the amino acid at position 12 of the protein, WT wild type, which refers to the normal, non-mutated version of a gene or protein, ECOG Eastern Cooperative Oncology Group, a performance status scale used to assess how well a patient is able to perform common activities of daily living, LLN lower limit of normal, ULN upper limit of normal, CEA carcinoembryonic antigen, antiEGFR anti-epidermal growth factor receptor, FOLFIRI a chemotherapy regimen that combines folinic acid (leucovorin), fluorouracil, and irinotecan, FOLFOX a chemotherapy regimen that combines folinic acid (leucovorin), fluorouracil, and oxaliplatin, polyCT polychemotherapy, which refers to the use of multiple chemotherapy drugs in combination, CT chemotherapy.
Table 2.
Specific mutation type.
| Mutation typesa | N (%) |
|---|---|
| Total | 1160 (100) |
| Wild type | 594 (51.21) |
| KRAS (all) | 492 (42.41) |
| KRAS p.G12D (c.35G>A) | 94 (8.1) |
| KRAS codon12 (mutation details not available) | 91 (7.84) |
| KRAS p. G12V (c.35G>T) | 85 (7.33) |
| KRAS p.G13D (c.38G>A) | 59 (5.09) |
| Mutant NRAS (all) | 37 (3.19) |
| Mutant BRAF (all) | 37 (3.19) |
| KRAS p.G12C (c.34G>T) | 33 (2.84) |
| KRAS p.G12A (c.35G>C) | 23 (1.98) |
| NRAS mutated exon3 (mutation details not available) | 21 (1.81) |
| KRAS p.G12S (c.34G>A) | 21 (1.81) |
| NRAS exon2 (mutation details not available) | 16 (1.38) |
| KRAS codon13 (mutation details not available) | 15 (1.29) |
| KRAS exon3 (mutation details not available) | 13 (1.12) |
| KRAS p.A146T (c.436G>A) | 13 (1.12) |
| KRAS exon4 (mutation details not available) | 12 (1.03) |
| KRAS exon2 (mutation details not available) | 10 (0.86) |
| KRAS p.G12R (c.34G>C) | 6 (0.52) |
| KRAS p.Q61H (c.183A>C) | 5 (0.43) |
| KRAS p.Q61L (c.182A>T) | 3 (0.26) |
| KRAS p.A146V (c.435C>T) | 2 (0.17) |
| KRAS p.G13C (c.37G>T) | 2 (0.17) |
| KRAS p.G12X | 1 (0.09) |
| KRAS p.Q61E (c.181C>G) | 1 (0.09) |
| KRAS p.Q61R (c.182A>G) | 1 (0.09) |
| KRAS p.G12R (c.34G>C) | 1 (0.09) |
| KRAS p.G13S (c.37G>A) | 1 (0.09) |
aMutation types have been listed in descending order of frequency.
Patients with mutated BRAF demonstrated poorer characteristics upon initiation of the second line, with only 22% ECOG-PS 0 compared to 31% in KRAS wild type. They also exhibited a higher frequency of life-threatening metastases (26% vs 14%), and a greater rate of peritoneal involvement (38% vs 25%). Detailed characteristics categorised by specific mutation are provided in Annex Table 2. Intriguingly, these disparities were observed despite an apparent selection of good prognosis cases during the first-line treatment. Indeed, an analysis detailing the distribution of variables based on the time span from diagnosis of advanced disease to the start of second-line therapy can be seen in Annex Fig. 1. Notably, tumours with extended intervals between these treatments displayed more favourable prognostic indicators.
Treatments
The most common first-line treatments consisted of oxaliplatin/fluoropyrimidine-based regimens with or without bevacizumab or antiEGFR. The first-line data is presented in Annex Table 4. Potential survivorship bias has been noted for these data, and they serve as a reference point for the second-line results. Either way, the mutations with the worst PFS-1 were KRAS p.G12C and BRAF p.V600E. Coherently, these identical mutations were also associated with the worst PFS-1 outcomes.
In the second line, the most prevalent regimens included FOLFIRI-aflibercept (30%), poly-chemotherapy without monoclonal antibody (22%), and irinotecan-based regimens plus bevacizumab (20%). Baseline characteristics according to treatment are shown in Annex Table 3A/3B.
The median number of treatment lines administered was 3, with a range of 2–7. Annex Fig. 2 features a Sankey plot that illustrates the treatment sequences for the complete sample. The distribution of therapeutic agents utilised beyond the second line, based on mutational subtype, and the number of subsequent lines of therapy are illustrated in Annex Figs. 3 and 4.
Efficacy and survival endpoints from the second line
In the full cohort, the median OS-2 was 13.6 months (95% CI: 12.8–14.9), and the median PFS-2 was 6.6 months (95% CI: 6.2–15.9), as shown in Annex Fig. 5. The outcomes broken down by mutational status are illustrated in Fig. 1. Specifically, patients harbouring KRAS p.G12C and BRAF p.V600E mutations experienced a median OS-2 of 10.6 (95% CI, 8–16.3) and 8.5 months (95% CI, 5.2–18) respectively. Correspondingly, PFS-2 durations for these mutations were 4.8 months (95% CI, 3.5–8.1) and 3.8 months (95% CI, 2.8–7.6). In contrast, those with wild-type tumours exhibited a PFS-2 of 6.7 (95% CI: 6.2–7.4) and an OS-2 of 15.8 (95% CI: 14.6–17.7) months. Of note, the Kaplan–Meier curves for the BRAF p.V600E subgroup demonstrated a crossover with other subgroups at the midterm, as illustrated in Fig. 1. This finding is associated with a higher rate of PD as the best response to the second line (54%), comparable to the result obtained in BRAF mutated tumours (PD rate of 54%), and higher than expected in the overall KRAS mutated and wild-type cases (39%, in both cases) (Annex Table 4A/4B).
Fig. 1.
Kaplan–Meier plots for progression-free survival and overall survival from first (a) and second-line treatment (b, c), stratified by genetic variant.
Given the small subgroup sizes, the curves stratified by mutational status are noisy and do not allow for the rejection of the null hypothesis that the endpoints for each scheme are indifferent to the genotype (Fig. 2 and Annex Fig. 6). However, the dynamic trends for BRAF mutations seem different in subjects treated with FOLFIRI–aflibercept. For example, in the FOLFIRI + aflibercept stratum, the 12-month results for BRAF mutated and RAS wild-type are comparable, with a 12-month survival rate (OS-2) of 53.8% (95% CI, 32.6–89.1) and 53.7% (95% CI, 46.2–62.5), respectively. In contrast, in the poly-chemotherapy stratum, the 12-month survival rate (OS-2) was 25.0% (95% CI, 4.5–1) and 50.0% (41.4–60.3) for these same groups. This trend did not occur for other mutations such as KRAS p.G12C, which maintained poor prognosis regardless of the treatment scheme (Annex Fig. 6). Annex Table 4A/4B presents the results broken down by strata and line, revealing the worse outcomes of KRAS G12C when compared to other scenarios.
Fig. 2. Kaplan–Meier plots for progression-free survival and overall survival from second-line treatment, stratified by genetic variant.
Each panel represents a therapeutic strategy or regimen in the second line.
Dynamic effect and contribution of mutational status
The partitioned survival analysis suggests a distinct pattern during the second line and beyond, with a higher contribution to the total from the second line observed in cases of BRAF p.V600E, NRAS, and KRAS p.G12D (c.35G>A) (Fig. 3). However, in the first two cases, mortality occurred more rapidly once the second line ceased to demonstrate efficacy. Regarding KRAS p.G12C (c.34G>T), the notably poor outcomes following the initial line of treatment manifested despite this subgroup going through more lines of therapy (median 2, range 2–6) when compared to 3 lines (range 3–7) for BRAF p.V600E and other treatment options (refer to sequences in Sankey plots; Annex Fig. 7A–C). To further explore the impact of subsequent lines of therapy, Kaplan–Meier curves were generated to stratify survival following the completion of the second line (OS-3). When comparing survival functions for patients receiving three or more lines of therapy, the data did not indicate a notable difference in survival from the onset of the third line (OS-3), depending on the mutational subtype (log-rank test, p = 0.3) (Annex Fig. 8A/B). Yet, the common transitions between the first and second lines of treatment nor substantially influenced the clinical outcomes of BRAF-mutated tumours. The treatments across subtypes were largely similar, and it is noteworthy that no patient with a BRAF mutation received immunotherapy at any point during the study (Annex Table 5A/5B).
Fig. 3. Partitioned survival analysis.
The areas describe the probability of an individual being in each health state, as described by the colour code, over time.
To shed light on these results, we investigated the potential presence of dynamic effects over time. Specifically, for PFS-2 and OS-2, the proportional hazards (PH) test rejected the notion that the effects of mutations such as KRAS p.G12C (c.34G>T), KRAS p.G12V (c.35G>T), and BRAF p.V600E were time-invariant (p value < 0.05 for each mutation). This suggests the presence of a dynamic effect. The Royston–Parmar spline model (RPSM) demonstrated that the harmful effect of KRAS p.G12C (c.34G>T) and KRAS p.G12V (c.35G>T) on OS-2 escalated for all contrasts from the commencement of the second line of treatment (Fig. 4 and Annex Fig. 9). For instance, the hazard ratio (HR) for OS-2 in the contrast of KRAS p.G12C (c.34G>T) vs RAS wild-type exhibits an HR of 1.18 (95% CI, 0.63–1.95), 1.58 (95% CI, 0.99–2.43), and 2.77 (95% CI, 2.02–3.52) at 6, 12, and 18 months, respectively. The dynamics of KRAS p.G12V (c.35G>T) vs RAS wild type were similar. Fixed effects were shown in Annex Tables 6A and 7.
Fig. 4. Time-varying hazard ratios for overall survival from second-line treatment (OS-2).
Results are obtained from a multivariable Royston–Parmar model. These hazard ratios represent the dynamic part of the model. The fixed part is reported in Annex Table 7. The ratio of the hazard rates of variable1 versus variable2 is computed as hazard(variable1)/hazard(variable2).
Strikingly, the HR for OS-2 associated with the BRAF p.V600E mutation exhibited an opposite trend. Thus, compared to RAS wild-type, patients with BRAF p.V600E had a higher risk of early mortality from the second line, which is evident from very early temporal cut-off points (i.e. HR for OS-2 of 2.22, 95% CI, 1.15–3.92 at the third month). Yet, by the sixth month into the second-line treatment, this effect diminished, so that from this point, the BRAF mutation did not influence OS-2 compared to the wild-type genotype. Using a standard Cox model, the average HR for OS-2 compared to RAS wild-type is 1.59 (95% CI, 1.11–2.25), which does not capture the notable dynamic effect observed.
A comparison of KRAS p.G12C (c.34G>T) and BRAF p.V600E highlights both situations, with the HR for OS-2 quickly escalating from the second line (Fig. 4 and Annex Fig. 10) within 6 months. The inversion of hazard ratios results in the crossing of OS-2 estimates. For example, among those surviving 6 months post-crossover in second-line survival functions (OS-2), the 12-month survival rate from the second line is 49.9% (95% CI: 32.7–67.3) for KRAS p.G12C (c.34G>T) and 59% (95% CI: 38.5–80.6) for BRAF p.V600E. At 24 months from the second line, the conditional survival rates (OS-2) are 1.6% (95% CI, 0.47–4.60) and 25.9% (95% CI, 38.5–80.6), respectively.
To address potential survivor bias from the first-line treatment, we carried out a sensitivity analysis, stratifying results based on the time elapsed between the initiation of the first and second lines of therapy. As anticipated, the results reveal a correlation between the duration leading up to the start of the second-line therapy and subsequent outcomes (see Annex Fig. 11). Yet, when we excluded patients with the most unfavourable prognosis, specifically those advancing to the second line within 6 months of commencing the first-line therapy, the impact of the mutational subtype on the OS-2 persisted (Annex Fig. 12). Similarly, the dynamic effect continued to manifest in a consistent manner (Annex Fig. 13).
The choice of second-line treatment might influence the dynamic effect in certain scenarios, altering the pattern of study attrition. For instance, in patients treated with regimens other than FOLFIRI-aflibercept, the hazard ratio (HR) for overall survival (OS-2) associated with the BRAF mutation compared to RAS wild-type was 2.53 (95% CI, 1.20–4.95) at 3 months, and 0.79 (95% CI, 0.50–1.10) at 18 months. In contrast, for those receiving FOLFIRI-aflibercept, the HRs at 3 and 18 months were 1.13 (95% CI, 0.12–4.28) and 0.51 (95% CI, 0.29–0.73), respectively. This observation may hint at a beneficial effect of FOLFIRI-aflibercept in BRAF-mutated tumours, potentially counteracting the mid-term adverse implications of the BRAF mutation. Nevertheless, the wide confidence intervals warrant a prudent interpretation. In line with these observations, a similar pattern is evident for PFS-2, underscoring the notion that the dynamic influence of the mutational subtype hinges significantly on second-line therapy outcomes (Annex Fig. 6). Similar trends are not observed for other mutational subtypes.
Discussion
Our analysis demonstrates that KRAS/BRAF mutational status influences distinct clinical profiles and differential responses to second-line treatment regimens, thereby manifesting dynamic effects in the study outcomes. The most unfavourable mutational statuses in the second line were observed for KRAS p.G12C (c.34G>T), KRAS p.G12V (c.35G>T), and BRAF p.V600E, aligning with previously reported data [34–36]. This poorer prognosis likely arises from differential attrition over time, potentially influenced by the unique aggressiveness conferred by the mutational subtype and the varying impacts of treatment regimens. Additionally, our results clearly indicate that proportional hazards models, which assume time-invariant effects, overlook significant dynamic influences that shape the clinical trajectory of these patients. Therefore, it is imperative for researchers to conduct these assessments in a descriptive section of their papers, as has been previously suggested [37, 38].
To contextualise with existing literature, the median PFS-2 and OS-2 for patients with BRAF mutations not treated with anti-BRAF therapy stood at 3.8 months (95% CI, 2.8–7.6) and 8.5 months (95% CI, 5.2–18), respectively, aligning with figures reported in other studies. For instance, a study from an Asian cohort reported second-line PFS and OS medians of 2.5 and 6.2 months, underscoring the poor prognosis of this patient subset [39]. In the phase 3 BEACON trial, the combination therapy of encorafenib with cetuximab demonstrated slightly improved PFS and OS medians in second-line treated subjects, at 4.2 and 7.9 months, respectively [40, 41]. However, it is noteworthy that the control arm in this trial exhibited poorer outcomes, with PFS and OS medians of 1.5 and 4.8 months in the second line, contrasting with our findings. Nonetheless, our data suggest that BRAF p.V600E-mutated patients who survive first-line therapy and do not exhibit rapid progression to the second line (i.e. <3 months) tend to manifest increased treatment sensitivity compared to those with other mutations. This is especially evident when treated with FOLFIRI + aflibercept, indicating a dynamic effect that might reflect the presence of a predictive factor yet to be elucidated.
Building on the previous insights, it is tempting to hypothesise that the outcomes for the control arm of the BEACON trial might have been improved by recognising this dynamic effect and by formally incorporating anti-angiogenic therapies to the control arm. Such a notion is further supported by the outcomes of pivotal second-line anti-angiogenic trials, which suggested that the subgroup with BRAF p.V600E mutations appears to derive greater benefit from these therapies [42]. Conversely, the TRIBE study underscored BRAF mutations as unfavourable prognostic markers for both progression-free survival and overall survival among patients administered irinotecan-based chemotherapy alongside [43, 44].
Owing to their poorer prognosis, dynamic effect, and molecular features that promote allosteric inhibition, individuals with KRAS p.G12C (c.34G>T) mutations represent another group of interest. Our findings verify that the clinical aggressiveness associated with this mutation is on par with that of BRAF p.V600E. Consistent with broader literature, this mutation is generally associated with a poorer prognosis. Our data align well with findings from other notable studies. For example, the Italian CRC-KRA GOIM study documented a median PFS and OS from the first line of 9 (95% CI, 7.5–10.5) and 21 months (95% CI, 17.4–24.6) for KRAS p.G12C tumours [20]. Similarly, the MD Anderson series indicated a median OS-1 of 21.2 months (95% CI, 16.7–25.7) for KRAS p.G12C tumours [8]. A Japanese series further echoed these findings, reporting median PFS-1 and OS-1 durations of 9.4 and 21.1 months, respectively [45]. Collectively, these outcomes are less favourable than what is typically observed with other mutations.
Crucially, our findings from the second line align with the broader literature. For instance, our PFS-2 results in KRAS p.G12C tumours (median of 4.8 months, 95% CI, 3.5–8.1) are consistent with data from the MD Anderson study (3.9 months, 95% CI, 2.9–5.9) [8], and the US clinical genomic database [46]. However, beyond these initial observations, our analysis emphasises the dynamic nature of the effect. Specifically, we observed notably worse outcomes for BRAF p.V600E between 12 and 18 months from the start of the second-line treatment. Therefore, a possible short-term solution to this issue could be intensifying first line treatment due to its poor prognosis, using FOLFOXIRI [20], especially if metastases are unresectable. Even though the supporting data for this strategy is limited, the TRIBE RCT did not find a subgroup effect based on RAS mutation, suggesting that the FOLFOXIRI triplet can counteract some of the negative prognosis [47]. Furthermore, a subanalysis of the JACCRO CC-11 trial revealed that the KRAS mutation disappeared in ctDNA in a notable percentage of patients receiving FOLFOXIRI plus bevacizumab [48]. While this hypothesis cannot be meaningfully examined in our study since only three subjects with KRAS p.G12C were given a triplet, our data do confirm the low activity of doublets with antiangiogenic agents in the first line and the marginal contribution to the overall second line in this context. The imperative to strengthen the MAPK pathway inhibition strategy, including the adoption of allosteric inhibitors, becomes even more pressing.
Third, mutations in KRAS p.G12V (c.35G>T) exhibit a better initial prognosis from the second line compared to BRAF and KRAS p.G12C (c.34G>T). However, medium-term sensitivity to second-line irinotecan + antiangiogenic regimens is worse for KRAS p.G12V than for BRAF p.V600E, leading to a reduction in the initial differences between the two types. This finding aligns with the US clinical genomic database, which reported a median OS-2 of 11.2 (9.9–12.5) months for KRAS pG12V tumours, compared to 10.3 (95%, 9.0–11.5) in BRAF-mutated tumours [46]. Notably, our data reveal that the medium-term outcome for KRAS p.G12V (c.35G>T) is still not as poor as KRAS p.G12C (c.34G>T), the worst group from the second line, further emphasising the need to advance the development of MAPK pathway inhibitors in this setting.
Our data, in addition to the aforementioned clinical implications, emphasise the need to critically examine the PH assumption for patients with RAS/BRAF mutational variants, capturing subtle yet significant time-dependent effects [24, 37, 38]. The dynamic perspective holds practical importance. For instance, when managing a patient with a BRAF mutation in second-line treatment, limiting the intensity of support during severe complications might be considered due to the dire prognosis, but some of these patients could be benefiting the most. Additionally, the need to delve deeper into the non-PH mechanism is clear. It is likely that an underlying potent predictive factor remains yet to be uncovered [49]. In this scenario, the dynamic effect could be explained by shifts in population composition over time, as patients exhibit varied treatment responses, and only those most responsive to therapy remain event-free in the midterm, or those with more favourable clinical profiles are selected. This notion does not conflict with the possibility that the inherent underlying mechanism could be genuinely dynamic at the individual level, linked to the time-dependent adaptability of the MAPK network and other resistance mechanisms [48, 50].
Finally, our study, like many others, has various limitations with the most notable being its retrospective design. However, death, treatment regimens, and line changes are accurately recorded in patient histories. Second, the analysis of mutational status was carried out according to each centre’s clinical practice. This included instances where the exact details of the variant were not available in a percentage of cases. Furthermore, the database does not include details on the type of testing performed (e.g. expanded codon testing), which currently appears relevant. Nevertheless, the main objective of our work is not so much the characterisation of the mutational landscape but rather the description of the dynamic effects of accurately characterised mutations, which was achievable with enough events. Readers should bear in mind the conditional interpretation of survival estimates since the study focused on subjects who had survived until the second line. For the interpretation and extrapolation of our findings, it is essential for readers to understand that our PFS-1 estimates are inherently descriptive. Given that some patients with poor prognosis presumably died early or experienced severe clinical deterioration during the first line, the PFS-1 estimated in this study is likely higher than that observed from the standpoint of patients initiating a first line. Nevertheless, these endpoints are valuable for description and comparison with other cohorts, and by providing insight into the tumour’s aggressiveness, they serve as useful prognostic factors during the second line. In the same vein, readers should note that our partitioned survival analysis is specifically tailored to patients who undergo second-line treatment. Consequently, this perspective might not fully encapsulate the entire journey of patients commencing with first-line therapy, particularly those who do not transition to a second line.
To sum up, our study underscores the significance of employing a dynamic approach, enhancing the simple characterisation of mutational landscapes in tumours, and promoting an appreciation of the clinical implications of each variant, which may diverge slightly from prior beliefs. The clinical and prognostic consequences of mutational status necessitate in-depth analysis from the second-line treatment due to the presence of dynamic effects. This holds importance for patient follow-up and therapy selection, especially for those with specific mutations. Ongoing research into the intricate world of mutations and their continuously changing clinical implications will contribute to the further development of precision medicine.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Supplementary information
Acknowledgements
We thank PROMETEO study researchers for their contribution to this study and Miguel Vaquero, Natalia Cateriano and the IRICOM S.L. team for the support of the website registry.
Author contributions
DC-T, PJ-F, AFM, CLL, and ACB conceived the project, monitored variables, analysed data, and wrote the initial draft of the manuscript. The remaining authors were involved in patient recruitment, supplying clinical information, and offering comments and suggestions for manuscript improvement. All authors contributed to data interpretation, discussion, and critical review of the manuscript.
Funding
The study has been funded by Roche and Sanofi Spain.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Competing interests
PJ-F discloses honoraria as speaker from Astellas, Astra Zeneca, BMS, Lilly, MSD, Novartis, Roche. CLL discloses consulting/advisory role for Amgen, Roche, Sanofi, MSD, Merck, Servier, Bayer; research grants from Amgen, Roche, MSD, Merck, Servier and travel grants and/or congress support from Servier, Sanofi, Amgen, Roche, MSD, Merck. AFM discloses honoraria as speaker from Amgen, Astra Zeneca, Astra Zeneca, Eisai, Lilly, MSD, Pierre Fabre, Servier. AMLM discloses consulting/advisory role for Amgen, Bayer, Roche, Eisai; honoraria as speaker from Eisai, Lilly, Amgen, Bayer, Sanofi, Merck Serono, Roche, Bristol, Servier, Pierre Fabre; travel grants and/or congress support from Amgen, Roche, Servier. FVR discloses consulting/advisory role for MSD; honoraria from Roche, Servier, Lilly, Merck, Astra-Zeneca; travel grants and/or congress support from Merck, Roche, Servier. VA discloses advisory role for Amgen, Roche, Merck, Servier, Ipsen pharma, Novartis; travel grants and/or congress support from Merck. JA discloses consulting/advisory role for Merck; honoraria as speaker from Amgen, Merck and Servier; travel grants and/or congress support from Roche, Amgen, Merck, Servier and Sanofi. FS discloses consulting/advisory role for Merck Serono, Amgen; honoraria and travel grants and/or congress support from Hoffman La-Roche, Sanofi Aventis, Amgen, Merck Serono, Servier, Bristol-Myers Squibb and Terumo corporation. MG discloses travel grants and/or congress support from Roche, Amgen and Servier. AMC discloses consulting/advisory role for BMS, Sanofi and Novartis; honoraria as speaker from Roche, BMS, MSD, Novartis, Merck and Servier; travel grants and/or congress support from BMS, Novartis, Merck and Servier. RJM discloses honoraria as speaker from Servier, Sanofi, Pfizer and Pierre Fabre; travel grants and/or congress support from Merck. SCG discloses travel grants and/or congress support from Bristol Myers Squibb, MSD, Rovi, Amgem. BGA discloses consulting/advisory role for Sanofi; honoraria as speaker from Roche, BMS, Amgen MSD, Merck and Servier; travel grants and/or congress support from Amgen, Merck and Servier. PGV discloses consulting/advisory role for Servier, Amgen; travel grants and/or congress support from Merck, Sanofi, Amgen. ACVM discloses travel grants and/or congress support from Amgen, Merck, MSD, Roche, Sanofi, Servier. MMS discloses consulting/advisory role: Merck, Servier; honoraria as speaker from Amgen, Merck, Servier and Bayer; travel grants and/or congress support from Roche, Amgen, Merck, Servier. CCL discloses consulting/advisory role for Merck, honoraria as speaker from Amgen, Bayer, Sanofi, Merck Serono, Roche, Servier; travel grants and/or congress support from: Amgen, Bayer, Sanofi, Merck Serono, Roche, Servier. JAp discloses consulting/advisory role for Amgen, Bayer, Merck, Merck Sharp & Dohme, Pierre Fabre and Servier. AC-B discloses honoraria as speaker from Amgen, Astellas, Bayer, BMS, Eisai, Lilly, MSD, Merck, Novartis, Roche, Servier. DC-T, MCR, EAM, ABF-D, BA, and ACC have no conflict of interest, financial or otherwise, in relation to the scope of this work.
Ethics approval and consent to participate
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Consent for publication
Informed consent was obtained from all individual participants included in the study.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Débora Contreras-Toledo, Paula Jiménez-Fonseca.
Contributor Information
Débora Contreras-Toledo, Email: ct.debora@gmail.com.
Alberto Carmona-Bayonas, Email: alberto.carmonabayonas@gmail.com.
Supplementary information
The online version contains supplementary material available at 10.1038/s41416-023-02563-w.
References
- 1.Ottaiano A, Normanno N, Facchini S, Cassata A, Nappi A, Romano C, et al. Study of Ras mutations’ prognostic value in metastatic colorectal cancer: STORIA analysis. Cancers. 2020;12:1919. doi: 10.3390/cancers12071919. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Bylsma LC, Gillezeau C, Garawin TA, Kelsh MA, Fryzek JP, Sangaré L, et al. Prevalence of RAS and BRAF mutations in metastatic colorectal cancer patients by tumor sidedness: a systematic review and meta‐analysis. Cancer Med. 2020;9:1044–57. doi: 10.1002/cam4.2747. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bazan V, Migliavacca M, Zanna I, Tubiolo C, Grassi N, Latteri MA, et al. Specific codon 13 K-ras mutations are predictive of clinical outcome in colorectal cancer patients, whereas codon 12 K-ras mutations are associated with mucinous histotype. Ann Oncol. 2002;13:1438–46. doi: 10.1093/annonc/mdf226. [DOI] [PubMed] [Google Scholar]
- 4.Zhu G, Pei L, Xia H, Tang Q, Bi F. Role of oncogenic KRAS in the prognosis, diagnosis and treatment of colorectal cancer. Mol Cancer. 2021;20:1–17. doi: 10.1186/s12943-021-01441-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Patelli G, Tosi F, Amatu A, Mauri G, Curaba A, Patane DA, et al. Strategies to tackle RAS-mutated metastatic colorectal cancer. ESMO Open. 2021;6:100156. doi: 10.1016/j.esmoop.2021.100156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Goebel L, Müller MP, Goody RS, Rauh D. KRasG12C inhibitors in clinical trials: a short historical perspective. RSC Med Chem. 2020;11:760–70. doi: 10.1039/d0md00096e. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ostrem JM, Peters U, Sos ML, Wells JA, Shokat KM. K-Ras (G12C) inhibitors allosterically control GTP affinity and effector interactions. Nature. 2013;503:548–51. doi: 10.1038/nature12796. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Henry JT, Coker O, Chowdhury S, Shen JP, Morris VK, Dasari A, et al. Comprehensive clinical and molecular characterization of KRASG12C-mutant colorectal cancer. JCO Precis Oncol. 2021;5:PO.20.00256. [DOI] [PMC free article] [PubMed]
- 9.Dienstmann R, Connor K, Byrne AT, Fridman WH, Lambrechts D, Sadanandam A, et al. Precision therapy in RAS mutant colorectal cancer. Gastroenterology. 2020;158:806–11. doi: 10.1053/j.gastro.2019.12.051. [DOI] [PubMed] [Google Scholar]
- 10.Amodio V, Yaeger R, Arcella P, Cancelliere C, Lamba S, Lorenzato A, et al. EGFR blockade reverts resistance to KRASG12C inhibition in colorectal cancer overcoming KRASG12C inhibitor resistance in colorectal cancer. Cancer Discov. 2020;10:1129–39. doi: 10.1158/2159-8290.CD-20-0187. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hong DS, Fakih MG, Strickler JH, Desai J, Durm GA, Shapiro GI, et al. KRASG12C inhibition with sotorasib in advanced solid tumors. N Engl J Med. 2020;383:1207–17. doi: 10.1056/NEJMoa1917239. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Yaeger R, Weiss J, Pelster MS, Spira AI, Barve M, Ou S-HI, et al. Adagrasib with or without cetuximab in colorectal cancer with mutated KRAS G12C. N Engl J Med. 2023;388:44–54. doi: 10.1056/NEJMoa2212419. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hunter JC, Manandhar A, Carrasco MA, Gurbani D, Gondi S, Westover KD. Biochemical and structural analysis of common cancer-associated KRAS mutations. Mol Cancer Res. 2015;13:1325–35. doi: 10.1158/1541-7786.MCR-15-0203. [DOI] [PubMed] [Google Scholar]
- 14.de la Chapelle A, Hampel H. Clinical relevance of microsatellite instability in colorectal cancer. J Clin Oncol. 2010;28:3380. doi: 10.1200/JCO.2009.27.0652. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Kopetz S, Desai J, Chan E, Hecht JR, O’Dwyer PJ, Maru D, et al. Phase II pilot study of vemurafenib in patients with metastatic BRAF-mutated colorectal cancer. J Clin Oncol. 2015;33:4032. doi: 10.1200/JCO.2015.63.2497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rowland A, Dias MM, Wiese MD, Kichenadasse G, McKinnon RA, Karapetis CS, et al. Meta-analysis of BRAF mutation as a predictive biomarker of benefit from anti-EGFR monoclonal antibody therapy for RAS wild-type metastatic colorectal cancer. Br J Cancer. 2015;112:1888–94. doi: 10.1038/bjc.2015.173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Pietrantonio F, Petrelli F, Coinu A, Di Bartolomeo M, Borgonovo K, Maggi C, et al. Predictive role of BRAF mutations in patients with advanced colorectal cancer receiving cetuximab and panitumumab: a meta-analysis. Eur J Cancer. 2015;51:587–94. doi: 10.1016/j.ejca.2015.01.054. [DOI] [PubMed] [Google Scholar]
- 18.Kopetz S, Grothey A, Yaeger R, Van Cutsem E, Desai J, Yoshino T, et al. Encorafenib, binimetinib, and cetuximab in BRAF V600E–mutated colorectal cancer. N Engl J Med. 2019;381:1632–43. doi: 10.1056/NEJMoa1908075. [DOI] [PubMed] [Google Scholar]
- 19.Seligmann JF, Fisher D, Smith CG, Richman SD, Elliott F, Brown S, et al. Investigating the poor outcomes ofBRAF-mutant advanced colorectal cancer: analysis from 2530 patients in randomised clinical trials. Ann Oncol. 2017;28:562–8. doi: 10.1093/annonc/mdw645. [DOI] [PubMed] [Google Scholar]
- 20.Ciardiello D, Chiarazzo C, Famiglietti V, Damato A, Pinto C, Zampino MG, et al. Clinical efficacy of sequential treatments in KRASG12C-mutant metastatic colorectal cancer: findings from a real-life multicenter Italian study (CRC-KR GOIM) ESMO Open. 2022;7:100567. doi: 10.1016/j.esmoop.2022.100567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Hochster H, Hager S, Pipas JM, Tebbutt N, Laurent S, Gravalos C, et al. O-010 KRAS and BRAF gene subgroup analysis in the phase 3 RECOURSE trial of TAS-102 versus placebo in patients with metastatic colorectal cancer. Ann Oncol. 2015;26:iv111. [Google Scholar]
- 22.Cervantes A, Adam R, Roselló S, Arnold D, Normanno N, Taïeb J, et al. Metastatic colorectal cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann Oncol. 2023;34:10–32. doi: 10.1016/j.annonc.2022.10.003. [DOI] [PubMed] [Google Scholar]
- 23.Humbert M, Sitbon O, Yaïci A, Montani D, O’callaghan DS, Jaïs X, et al. Survival in incident and prevalent cohorts of patients with pulmonary arterial hypertension. Eur Respir J. 2010;36:549–55. doi: 10.1183/09031936.00057010. [DOI] [PubMed] [Google Scholar]
- 24.Carmona-Bayonas A, Jimenez-Fonseca P, Garrido M, Custodio A, Hernandez R, Lacalle A, et al. Multistate models: accurate and dynamic methods to improve predictions of thrombotic risk in patients with cancer. Thromb Haemost. 2019;119:1849–59. doi: 10.1055/s-0039-1694012. [DOI] [PubMed] [Google Scholar]
- 25.Royston P, Parmar MKB. Flexible parametric proportional‐hazards and proportional‐odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects. Stat Med. 2002;21:2175–97. doi: 10.1002/sim.1203. [DOI] [PubMed] [Google Scholar]
- 26.Harrell F. Regression modeling strategies: with applications to linear models, logistic and ordinal regression, and survival analysis. 2nd ed. New York: Springer; 2015.
- 27.Woods BS, Sideris E, Palmer S, Latimer N, Soares M. Partitioned survival and state transition models for healthcare decision making in oncology: where are we now? Value Health. 2020;23:1613–21. doi: 10.1016/j.jval.2020.08.2094. [DOI] [PubMed] [Google Scholar]
- 28.Cranmer H, Shields GE, Bullement A. A comparison of partitioned survival analysis and state transition multi-state modelling approaches using a case study in oncology. J Med Econ. 2020;23:1176–85. doi: 10.1080/13696998.2020.1796360. [DOI] [PubMed] [Google Scholar]
- 29.R Core Team. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2014.
- 30.De Wreede LC, Fiocco M, Putter H. The mstate package for estimation and prediction in non-and semi-parametric multi-state and competing risks models. Comput Methods Prog Biomed. 2010;99:261–74. doi: 10.1016/j.cmpb.2010.01.001. [DOI] [PubMed] [Google Scholar]
- 31.Therneau T, Crowson C, Atkinson E. Multi-state models and competing risks. 2016. https://cran.r-project.org/web/packages/survival/vignettes/compete.pdf.
- 32.Allaire JJ, Ellis P, Gandrud C, Kuo K, Lewis BW, Owen J, et al. Package ‘networkD3.’ D3 JavaScript Netw graphs from R. 2017. https://CRAN.R-project.org/package=networkD3.
- 33.Incerti D, Jeroen G, Jansen P. hesim: Health Economic Simulation Modeling and Decision Analysis. 2021 Feb 18 [cited 2024 Jan 3]; Available from: https://arxiv.org/abs/2102.09437v2.
- 34.Jones RP, Sutton PA, Evans JP, Clifford R, McAvoy A, Lewis J, et al. Specific mutations in KRAS codon 12 are associated with worse overall survival in patients with advanced and recurrent colorectal cancer. Br J Cancer. 2017;116:923–9. doi: 10.1038/bjc.2017.37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Imamura Y, Morikawa T, Liao X, Lochhead P, Kuchiba A, Yamauchi M, et al. Specific mutations in KRAS codons 12 and 13, and patient prognosis in 1075 BRAF wild-type colorectal cancers. Clin Cancer Res. 2012;18:4753–63. doi: 10.1158/1078-0432.CCR-11-3210. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Modest DP, Ricard I, Heinemann V, Hegewisch-Becker S, Schmiegel W, Porschen R, et al. Outcome according to KRAS-, NRAS-and BRAF-mutation as well as KRAS mutation variants: pooled analysis of five randomized trials in metastatic colorectal cancer by the AIO Colorectal Cancer Study Group. Ann Oncol. 2016;27:1746–53. doi: 10.1093/annonc/mdw261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Castañon E, Sanchez-Arraez Á, Jimenez-Fonseca P, Alvarez-Manceñido F, Martínez-Martínez I, Gongora LM, et al. Bayesian interpretation of immunotherapy trials with dynamic treatment effects. Eur J Cancer. 2022;161:79–89. doi: 10.1016/j.ejca.2021.11.002. [DOI] [PubMed] [Google Scholar]
- 38.Castañon E, Sanchez-Arraez A, Alvarez-Manceñido F, Jimenez-Fonseca P, Carmona-Bayonas A. Critical reappraisal of phase III trials with immune checkpoint inhibitors in non-proportional hazards settings. Eur J Cancer. 2020;136:159–68. doi: 10.1016/j.ejca.2020.06.003. [DOI] [PubMed] [Google Scholar]
- 39.Mitani S, Taniguchi H, Honda K, Masuishi T, Narita Y, Kadowaki S, et al. Analysis of efficacy and prognostic factors in second-line chemotherapy for BRAF V600E mutant metastatic colorectal cancer. Ann Oncol. 2017;28:v180. doi: 10.1093/annonc/mdy107. [DOI] [PubMed] [Google Scholar]
- 40.Tabernero J, Van Cutsem E, Yoshino T, Yaeger R, Wasan H, Desai J, et al. SO-29 Clinical characteristics, efficacy, and safety in patients receiving second-or third-line encorafenib plus cetuximab (E+ C) vs control for metastatic colorectal cancer (mCRC): BEACON CRC post hoc analysis. Ann Oncol. 2023;34:S174. [Google Scholar]
- 41.Tabernero J, Grothey A, Van Cutsem E, Yaeger R, Wasan H, Yoshino T, et al. Encorafenib plus cetuximab as a new standard of care for previously treated BRAF V600E–mutant metastatic colorectal cancer: updated survival results and subgroup analyses from the BEACON study. J Clin Oncol. 2021;39:273. doi: 10.1200/JCO.20.02088. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Wirapati P, Pomella V, Vandenbosch B, Kerr P, Maiello E, Jeffery GM, et al. Velour trial biomarkers update: Impact of RAS, BRAF, and sidedness on aflibercept activity. J Clin Oncol. 2017;35:3538.
- 43.Loupakis F, Cremolini C, Masi G, Lonardi S, Zagonel V, Salvatore L, et al. Initial therapy with FOLFOXIRI and bevacizumab for metastatic colorectal cancer. N Engl J Med. 2014;371:1609–18. doi: 10.1056/NEJMoa1403108. [DOI] [PubMed] [Google Scholar]
- 44.Yoshino T, Portnoy DC, Obermannova R, Bodoky G, Prausová J, Garcia-Carbonero R, et al. Biomarker analysis beyond angiogenesis: RAS/RAF mutation status, tumour sidedness, and second-line ramucirumab efficacy in patients with metastatic colorectal carcinoma from RAISE—a global phase III study. Ann Oncol. 2019;30:124–31. doi: 10.1093/annonc/mdy461. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Chida K, Kotani D, Masuishi T, Kawakami T, Kawamoto Y, Kato K, et al. The prognostic impact of KRAS G12C mutation in patients with metastatic colorectal cancer: a multicenter retrospective observational study. Oncologist. 2021;26:845–53. doi: 10.1002/onco.13870. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Fakih M, Tu H, Hsu H, Aggarwal S, Chan E, Rehn M, et al. Real-world study of characteristics and treatment outcomes among patients with KRAS p. G12C-mutated or other KRAS mutated metastatic colorectal cancer. Oncologist. 2022;27:663–74. doi: 10.1093/oncolo/oyac077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Cremolini C, Antoniotti C, Lonardi S, Bergamo F, Cortesi E, Tomasello G, et al. Primary tumor sidedness and benefit from FOLFOXIRI plus bevacizumab as initial therapy for metastatic colorectal cancer. Retrospective analysis of the TRIBE trial by GONO. Ann Oncol. 2018;29:1528–34. doi: 10.1093/annonc/mdy140. [DOI] [PubMed] [Google Scholar]
- 48.Sunakawa Y, Satake H, Usher J, Jaimes Y, Miyamoto Y, Nakamura M, et al. Dynamic changes in RAS gene status in circulating tumour DNA: a phase II trial of first-line FOLFOXIRI plus bevacizumab for RAS-mutant metastatic colorectal cancer (JACCRO CC-11) ESMO Open. 2022;7:100512. doi: 10.1016/j.esmoop.2022.100512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Mok TS, Wu Y-L, Thongprasert S, Yang C-H, Chu D-T, Saijo N, et al. Gefitinib or carboplatin–paclitaxel in pulmonary adenocarcinoma. N Engl J Med. 2009;361:947–57. doi: 10.1056/NEJMoa0810699. [DOI] [PubMed] [Google Scholar]
- 50.Siena S, Sartore-Bianchi A, Garcia-Carbonero R, Karthaus M, Smith D, Tabernero J, et al. Dynamic molecular analysis and clinical correlates of tumor evolution within a phase II trial of panitumumab-based therapy in metastatic colorectal cancer. Ann Oncol. 2018;29:119–26. doi: 10.1093/annonc/mdx504. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.