Population-based Screening for BRAF^V600E in Metastatic Colorectal Cancer Reveals Increased Prevalence and Poor Prognosis

Worldwide, colorectal cancer is the third most commonly diagnosed cancer, and the second most common cause of cancer-related deaths (1). Between 8% and 10% of colorectal cancers have a BRAF^V600E-activating mutation (2), which leads to RAS-independent MAPK pathway activation, tumor cell proliferation, and survival. BRAF^V600E has been well-studied in other neoplasms (3), most commonly in melanoma, for which targeted therapies are available (4). In colorectal cancer, BRAF^V600E-targeted therapies have been less successful (5). The recent BEACON trial has shown the greatest activity in BRAF-mutant metastatic colorectal cancer (mCRC). In this phase III trial, triplet therapy using encorafenib (a BRAF inhibitor), binimetinib (a MEK inhibitor), and cetuximab (an anti-EGFR antibody), was compared with a control arm of FOLFIRI and cetuximab or irinotecan and cetuximab. Median overall survival (mOS) was improved from 5.4 months to 9.0 months (P < 0.0001) with the triplet and to 8.4 months (P < 0.0001) with the doublet of encorafenib and cetuximab (6).

BRAF^V600E mutations are associated with a worse prognosis in mCRC (7, 8). However, our understanding of the true prognosis of BRAF^V600E mutations is hindered by inherent referral bias within survival calculations: patients with BRAF^V600E-mutant mCRC included in prognostic studies had to be well enough and survive long enough to undergo sequencing of BRAF. Patients with rapidly progressive disease, as is often observed clinically with BRAF^V600E mutations, may not survive long enough to be identified for inclusion within these studies. IHC may be used for detection of BRAF^V600E; however, the best currently available antibody (clone VE1) shows variable sensitivity/specificity. Thus, next-generation sequencing (NGS) remains the gold standard for detection (3). The use of IHC provides an accessible test with quick turnaround that allows for easy screening for BRAF^V600E-mutated colorectal cancer, and although it has limitations, its performance can be improved by the use of on-slide controls (9).

The presence of BRAF^V600E mutations helps rule out an underlying diagnosis of Lynch syndrome (10, 11). This is because of the observed linkage of BRAF^V600E mutations with the CpG island methylator phenotype (CIMP) pathway, which causes somatic hypermethylation of the MLH1 promotor and resultant microsatellite instability (MSI) in a nonhereditary fashion (10). Institutions in the Vancouver Coastal Health Authority (VCH, Vancouver, British Columbia, Canada) have been reflexively testing all colorectal cancers for loss of the mismatch repair (MMR) proteins MLH1, MSH2, PMS2, and MSH6 and the presence of mutated BRAF^V600E with IHC since 2014 for triage of genetic counselling, which provides a rich, nonbiased population-based cohort of colorectal cancer with BRAF^V600E testing. We aimed to use this population-based cohort and sequencing information from two large tertiary centers to (i) determine the true prevalence of BRAF^V600E mutations in mCRC, (ii) establish the prognostic impact of BRAF^V600E mutations, (iii) assess the impact of MMR status on the prognosis of patients with BRAF^V600E mCRC, and (iv) describe treatment patterns and outcomes for patients with mCRC with BRAF^V600E to better inform sequencing of therapies.

Institutional review board approval was obtained from the University of British Columbia (Vancouver, British Columbia, Canada) and the University of Texas MD Anderson Cancer Center (MDA, Houston, TX) for this study, with a waiver of consent due to the retrospective nature. All research was conducted in accordance with the Declaration of Helsinki. A retrospective chart review of clinical and pathologic records was completed on all patients within VCH (Vancouver, British Columbia, Canada) who had colorectal cancer and MMR/BRAF^V600E IHC screening on biopsy or surgical resection between April 1, 2014 and May 1, 2018. Prevalence was determined from an all-stages cohort and subsequent analysis focused on patients with metastatic disease. Patients with positive BRAF^V600E IHC and synchronous or metachronous metastatic disease are referred to as BRAF_IHC.

Existing records were reviewed to identify patients with BRAF^V600E-mutant mCRC at BC Cancer (Vancouver, British Columbia, Canada, January 1, 2013 to May 1, 2018; n = 33) and MDA (Houston, TX, January 1, 2004 to September 1, 2016; n = 221) identified by standard-of-care NGS as a comparator and are referred to as the BRAF_NGS cohort (n = 254). Overall survival (OS) did not differ between the NGS cohorts from BC Cancer (Vancouver, British Columbia, Canada) and MDA (Houston, TX, P = 0.11; Supplementary Fig. S1A) so the patients were pooled.

An additional group of 2,150 consecutive patients with NGS and no BRAF^V600E mutations from MDA (Houston, TX, n = 1762) and BC Cancer (Vancouver, British Columbia, Canada, n = 388) were included to provide a BRAF wild-type (BRAF_WT) population for comparisons and are described elsewhere (12). As all anticancer treatments are delivered by a single agency in the province of British Columbia, treatment regimens were reviewed for patients at BC Cancer (Vancouver, British Columbia, Canada) to assess attrition between lines of therapy, allowing for robust follow-up of treatment received in each line.

Within VCH, BRAF^V600E mutation status was determined by mutation-specific IHC (VE1 Antibody, Spring Bioscience) with on-slide control as described previously (9).

MMR status was retrospectively reviewed from patients' charts, and was evaluated only in patients who had testing performed as part of their clinical care. At VCH and BC Cancer (Vancouver, British Columbia, Canada), this was exclusively through the use of IHC for MLH1, MSH2, MSH6, and PMS2. At MDA (Houston, TX), testing consisted of a mixture of DNA-based PCR testing for MSI and IHC. Tumors were defined as deficient MMR (dMMR) if either the IHC or PCR-based assay detected an abnormality.

Between group comparisons were performed using χ² or Fisher exact tests for categorical variables as appropriate and Mann–Whitney tests for continuous variables. Kaplan–Meier curves summarized survival characteristics and were compared using log-rank tests. OS was defined as the time from stage IV diagnosis to death of any cause. Progression-free survival (PFS) was defined as the time from first-line treatment in the metastatic setting until progression or death. Patients without an event at the time of last follow-up were censored. Treatment effect was assessed in a combined cohort of NGS and IHC cases.

For multivariate analysis, after satisfying the proportional hazards assumption, a Cox regression analysis was performed using a forward likelihood ratio method with variables entering the model if P < 0.05 and removed from the model when P > 0.1. A total of 2,467 patients were included with 558 patients subsequently omitted because of missing values for a variable (usually MSI status) and a total of 1,162 events. Variables assessed for the model included gender, age, MMR status, primary tumor location, histology, synchronous disease, and method of BRAF mutation detection. A subsequent model was performed excluding wild-type patients to assess for interaction between BRAF method of detection and MMR status. This model included 235 patients and 188 events.

Reported P values are two-sided with P < 0.05 considered statistically significant. Univariate analysis was performed with GraphPad Prism version 8.0.2 (GraphPad Software). Multivariate analyses used SPSS version 14 (IBM).

Of 1,977 patients diagnosed with colorectal cancer within VCH, 1,898 (96%) underwent BRAF^V600E testing by IHC with 238 (12.5%) patients having a BRAF^V600E mutation by IHC. Staging investigations were available for 1,882 patients, including all patients with BRAF^V600E mutations. Of patients with synchronous metastatic disease at diagnosis (n = 314), 37 (11.8%) were positive for BRAF^V600E. Within patients who had synchronous or developed metachronous mCRC (n = 575), 61 (10.6%) were positive for BRAF^V600E by IHC. Among those who never developed metastases, 177 of 1,307 (13.5%) had BRAF^V600E by IHC. There was no difference in prevalence between early- and late-stage cases at diagnosis (P = 0.61), or between those that were ever metastatic versus those who never developed metastases (P = 0.078) (see Supplementary Table S1).

The remaining results and analyses focus on patients with mCRC. Patient baseline characteristics are summarized in Table 1. BRAF_IHC and BRAF_NGS cohorts showed no difference in sex (P = 0.68), primary tumor location (P = 0.62), or MMR status (P = 0.057). BRAF_IHC patients were older at diagnosis (P < 0.0001), and less likely to have synchronous metastases [OR, 0.43; 95% confidence interval (CI), 0.24–0.77; P = 0.0053] or be of signet/mucinous histology (OR, 0.41; 95% CI, 0.19–0.91; P = 0.023) compared with the BRAF_NGS cohort. Compared with BRAF_WT patients, pooled patients with BRAF^V600E mutation detected with either IHC or NGS showed an older median age at diagnosis (63 vs. 56 years; P < 0.0001), female predominance (OR, 1.53; 95% CI, 1.20–1.94; P = 0.0006), increased right-sided occurrence (OR, 5.07; 95% CI, 3.91–6.58; P < 0.0001), higher proportion with synchronous metastatic disease (OR, 1.97; 95% CI, 1.50–2.58; P < 0.0001), higher likelihood of having mucinous/signet ring histology (OR, 2.11; 95% CI, 1.59–2.82; P < 0.0001), and a higher association with dMMR (OR, 5.79; 95% CI, 3.70–8.87; P < 0.0001).

While all patients within the BRAF_NGS cohort were referred to oncology, 51 of 61 (83.6%) of the BRAF_IHC patients were referred for oncology consultation. Patients not referred with mCRC and a BRAF^V600E mutation detected by IHC were older (median age 85 vs. 69; P = 0.0067) but had no other baseline characteristics that significantly differed. NGS testing was completed in 26 (42.6%) of the BRAF_IHC patients of which 10 (16.4%) included BRAF^V600E coverage with 10 confirmed BRAF^V600E mutations. An additional two patients were initially BRAF^V600E positive by IHC with negative NGS results for a false-positive rate of 16.7% (2/12 cases tested with both IHC and NGS). On expert review of the IHC, and with retesting of the original tissues, both specimens returned negative IHC results, and were deemed wild-type.

OS was significantly worse for patients with BRAF^V600E mutations detected by any method compared with the BRAF_WT group (17.4 vs. 38.3 months; HR, 2.21; 95% CI, 1.93–2.53; P < 0.0001; Fig. 1). Among patients with BRAF^V600E mutations, OS was worse among the BRAF_IHC cohort compared with the BRAF_NGS cohort (5.5 vs. 20.4 months; HR, 2.90; 95% CI, 1.89–4.45; P < 0.0001).

The 26 BRAF_IHC patients who received NGS testing showed trends toward better OS relative to the 35 BRAF_IHC patients who never received any NGS testing (10.2 vs. 4.1 months; HR, 0.64; 95% CI, 0.38–1.09; P = 0.086; Supplementary Fig. S2). Given the differences in rates of synchronous/metachronous presentation between the IHC and NGS groups, we also compared OS among the IHC population stratified by whether their cancers developed with synchronous or metachronous metastases. We found no difference in OS (P = 0.56) in this comparison and also performed a multivariate analysis to ensure prognostic differences were not due to differences in baseline characteristics between groups. Even after controlling for baseline characteristics associated with a worse prognosis, IHC-detected BRAF^V600E mutations demonstrated a significantly larger impact on OS (HR, 7.37; 95% CI, 5.56–9.78; P < 0.0001) than those detected by NGS (HR, 1.75; 95% CI, 1.44–2.12; P < 0.0001) relative to patients with wild-type tumors (Table 2).

Both BRAF^V600E-mutated cohorts demonstrated increased proportions of dMMR compared with the wild-type population (BRAF_IHC 24.6%, BRAF_NGS 14.0%, and BRAF_WT 3.0%; P < 0.0001). Among all combined patients with BRAF^V600E mutations, worse outcomes occurred in patients with dMMR [dMMR 8.9 months vs. proficient MMR (pMMR) 17.2 months; HR, 1.46; 95% CI, 0.96–2.27; P = 0.043; Fig. 2]. However, this statistical difference did not persist when the BRAF_IHC (n = 61; dMMR 3.0 vs. pMMR 6.0 months; HR, 1.61; 95% CI, 0.81–3.23; P = 0.092; Supplementary Fig. S3A) and BRAF_NGS (n = 178; dMMR 19.3 vs. pMMR 20.7 months; HR, 1.23; 95% CI, 0.74–2.05; P = 0.41; Supplementary Fig. S3B) groups were analyzed independently, which was done given the significant magnitude of difference in outcomes between the groups. When controlling for covariates in a multivariate model, IHC-detected mutations drove worse OS in dMMR tumors when compared with NGS-detected mutations (HR, 2.74; 95% CI, 1.97–3.83; P < 0.0001; Supplementary Table S2); however, the test of interaction was negative (P = 0.16) for impact of MMR being dependent on BRAF ascertainment method.

There was no difference in first-line PFS between patients with mCRC and BRAF^V600E mutations detected by NGS from either institution (HR, 1.38; 95% CI, 0.93–2.05; P = 0.14; Supplementary Fig. S1B). The first-line PFS for the combined cohort (n = 239) of IHC- and NGS-detected mutations was 5.9 months (Fig. 3A). There were not enough patients in the IHC-detected cohort to analyze first-line PFS independently. When stratified by treatment regimen, the PFS differed among treatment arms (single agent, 2.7; FOLFIRI, 6.1; FOLFOX, 6.0; FOLFOXIRI, 8.9; BRAF directed, 21.3, and other 2.0 months; P = 0.0088; Fig. 3B). There was no difference between FOLFOX and FOLFIRI regimens (6.0 vs. 6.1 months; HR, 1.13; 95% CI, 0.83–1.54; P = 0.45). Clinical characteristics for the treatment cohorts are summarized in Supplementary Table S3.

A multivariate model demonstrated that only BRAF category (IHC/NGS) was prognostic for first-line PFS, with BRAF_IHC mCRC having worse PFS (HR, 1.98; 95% CI, 1.22–3.19; P = 0.005), but no difference noted between regimens, likely due to small sample size. Only 170 patients had all variables and could be included in the PFS model. If regimen was forced into the model, the results are summarized in Supplementary Table S4, but regimen did not stay in the model without forced entry.

The 84 pooled BRAF_IHC and BRAF_NGS patients from British Columbia showed attrition across lines of chemotherapy: 51 (60.7%; 95% CI, 49.5–71.2) received first-line palliative chemotherapy; 22 (26.2%; 95% CI, 17.2–36.9) received second-line; nine (10.7%; 95% CI, 5.0–19.4) received third-line; and three (3.6%; 95% CI, 0.7–10.1) received fourth-line chemotherapy (Fig. 3C). Attrition was only assessed in the BC Cancer cohort as chemotherapy is provided by a single system in this population allowing robust assessment of attrition, while MDA has a large referral population which would bias the estimate.

In this population-based study, we noted significantly worse OS associated with BRAF^V600E mutations in patients with mCRC than previously reported. This difference appears due to ascertainment bias among unwell patients, as only 42.6% of the metastatic population–based cohort (BRAF_IHC) underwent standard-of-care NGS testing and only 60.7% of them received chemotherapy in the metastatic setting. In support of this, the patients within the BRAF_IHC cohort that received NGS testing showed a trend toward better mOS than the BRAF_IHC patients that were never NGS tested (Supplementary Fig. S2).

While baseline differences between the IHC and NGS cohorts may have caused some prognostic differences, the magnitude of the HR difference between IHC- and NGS-detected mutations in our multivariate model relative to other covariates highlights how important BRAF mutations are to prognosis relative to other poor prognostic markers. Although, we did see large differences in rates of synchronous versus metachronous metastases between IHC-detected and NGS-detected mutations (Table 1), this is likely explained by the fact that IHC testing was done reflexively on cancers and this population would have included asymptomatic patients undergoing screening colonoscopies, while the NGS population would have included fewer patients who were asymptomatic and in a screening program.

Treatment differences between groups are also important to consider. Four of the patients included in our study (all within the BRAF_NGS group) were enrolled in clinical trials of BRAF-directed therapy (Fig. 3), with only one being included in the recently successful BEACON study. When we omitted these patients from our analysis it did not change the HR or mOS of the NGS group. Future studies addressing the survival of BRAF^V600E-mutated colorectal cancer will surely be affected by the advances in therapy, and our study timeframe provides a reflection of the pretargeted therapy era.

The prevalence of BRAF^V600E mutations in our study was 12.5% in all patients and 10.6% in mCRC. This is lower than the 20% reported in a Scandinavian population-based study; however, it is higher than reports from clinical trials and The Cancer Genome Atlas (13–15). Another important finding was that patients with BRAF^V600E mutations and dMMR have worse prognosis than those with pMMR, highlighting a group of patients with two actionable alterations and extremely poor prognosis. This finding emphasizes the importance of early ascertainment of molecular subtypes so that patients with aggressive characteristics can be directed toward personalized therapies with greater possibility of benefit.

No consensus has been reached regarding the interplay of BRAF^V600E and dMMR on OS. Opposite to our findings, but in a predominantly nonmetastatic context, Toon and colleagues evaluated two population-based cohorts of 1,426 and 1,109 consecutive colorectal cancer cases with IHC for MMR and BRAF^V600E. They found a worse OS among patients with BRAF^V600E mutations and pMMR compared with those with dMMR (16, 17). However, when the authors controlled for age and stage at diagnosis, no differences were noted. Tran and colleagues assessed 524 patients with mCRC with known BRAF^V600E mutation status and found no difference in OS between dMMR (n = 12) and pMMR (n = 30) tumors with BRAF^V600E mutations (18). In our study, we looked at prognosis only among patients with mCRC and even after controlling for covariates, demonstrated worse survival among patients with dMMR and BRAF^V600E mutations than those with pMMR and BRAF^V600E mutations. While a small proportion of patients within our study period may have received immunotherapy for their dMMR status, we expect that this would have improved their survival, and would be lessening the OS difference that we are seeing. We anticipate that without immunotherapy, the OS difference we have reported would only increase.

The VE1 antibody for detection of BRAF^V600E shows variable sensitivity/specificity across prior reports, with sensitivity ranging from 71% to 100% and specificity of 68%–100% within colon specimens (3). Staining patterns depend on adequacy of fixation and tissue. In light of the variable sensitivity and specificity of IHC, BRAF^V600E NGS testing remains the gold standard; however, we have previously shown that on-slide Immunohistochemistry Critical Assay Performance Control (ICAPC) can improve interobserver variability and reduce discordance from expert pathologic interpretation (9). In the absence of easily accessible NGS testing, BRAF^V600E IHC provides a reasonable screening test with quick turnaround that can be confirmed by NGS testing, if a mutation is noted by IHC. IHC testing reflexively identified 57% more patients with mCRC with BRAF alterations than were identified by standard-of-care NGS testing. A limitation of our study is that we were not able to validate the whole BRAF_IHC cohort with NGS testing; however, in those with overlapping NGS testing showing a BRAF^V600E mutation, the VE1 antibody was positive in 100% of patients (10/10). While it is possible that IHC testing included false-positives in the cohort, the strikingly poor survival suggests that the identification of BRAF^V600E-mutated cases was successful. The two false-positive cases identified within our cohort highlight the importance of stringent quality control measures when using this antibody. IHC may not be a perfect test for assessing BRAF mutations; however, the magnitude of survival differences noted in our study demonstrates that it can lead to rapid ascertainment of important clinical information. Although, one could argue that some of the patients with BRAF mutations may not undergo therapy even with this knowledge, the excellent side-effect profile of combination BRAF-directed therapy in contrast to standard doublet or triplet chemotherapy may lend itself to rapid integration into care plans for some patients (6).

Given the large cohort of patients we assessed, we aimed to evaluate optimal first-line therapy for patients with BRAF^V600E-mutated mCRC. We first evaluated the proportion of patients in our population-based cohort undergoing therapy. Only 60.7% of these patients underwent systemic therapy, with 26.2% receiving second-line and 10.7% receiving third-line therapy. This is an important finding as the BEACON clinical trial evaluated encorafenib, binimetinib, and cetuximab in the second- and third-line setting, and our results suggest that using this algorithm, a large proportion of patients would not receive this highly effective and well-tolerated therapy (6). Although the TRIBE clinical trials have shown that the best OS for systemic therapy in patients with BRAF^V600E mCRC is with first-line FOLFOXIRI + bevacizumab, this regimen can be challenging to administer to unwell patients (19). We found that first-line PFS was 5.9 months for all patients with BRAF^V600E mCRC and did not differ between FOLFOX and FOLFIRI. This sets a bar for potential first-line BRAF^V600E-directed trials such as the currently accruing ANCHOR-CRC trial (NCT03693170) assessing encorafenib, binimetinib, and cetuximab in the first-line setting for BRAF^V600E-mutated mCRC (20).

Despite important findings, our study must be interpreted in the context of several limitations. BRAF coverage was added to the BC Cancer NGS panel part way into the study, and as such we were not able to provide a robust assessment of concordance with NGS testing among all IHC-detected mutations. This explains why there is a discrepancy between the proportion of patients with NGS coverage and those with BRAF sequencing. In addition, although we use on-slide ICAPC, false positives do occur and may affect the prevalence we have reported. This may bias our results, but would have improved the outcome of the IHC-detected patients, further supporting the differences we noted. We were also limited in our ability to compare outcomes between first-line regimens given that a small number of patients received something other than FOLFOX or FOLFIRI. This was due to the retrospective nature of our study, which limits certain analyses and can introduce bias into assessments. However, the retrospective population-based design also provides our study's strength, as we were able to answer important questions regarding real-world outcomes to help inform treatment planning for this aggressive molecular subgroup.

We demonstrate that BRAF^V600E-mutated mCRC is associated with a worse prognosis than previously known and many of these patients are too unwell to undergo NGS testing or first-line chemotherapy. First-line PFS was 5.9 months among all patients with BRAF^V600E mutations, providing a baseline for first-line BRAF^V600E-directed trials. Concurrent dMMR may be associated with worse outcome among BRAF^V600E-mutant mCRC, and merits further evaluation. With the emergence of well-tolerated and effective therapy targeting BRAF^V600E mutations and dMMR, our work highlights the importance of timely ascertainment of molecular subtypes for treatment planning, and earlier consideration of targeted therapies.

V.K. Morris reports personal fees and grants from Array Biopharma (advisory board, research funding) and Bristol Myers Squibb (research funding) outside the submitted work. H.J. Lim reports personal fees from Roche, Ipsen, Lilly, Eisai, Taiho, Merck, and BMS research funding (grant) outside the submitted work. D.J. Renouf reports personal fees from Celgene, Servier, Roche, Taiho, Ipsen, and Bayer outside the submitted work. S. Kopetz reports grants from Amgen (research funding) during the conduct of the study, as well as other from MolecularMatch (stock ownership interests) outside the submitted work. D.F. Schaeffer reports personal fees from Robarts Clinical Trials Inc and Amgen Canada (outside the submitted work), and stock ownership from Satisfai Inc. J.M. Loree reports grants and personal fees from Ipsen, personal fees from Amgen, Novartis, Eisai, and Bayer outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

J.E. Chu: Conceptualization, data curation, formal analysis, methodology, writing-original draft, writing-review and editing. B. Johnson: Conceptualization, data curation, writing-review and editing. L. Kugathasan: Data curation, writing-review and editing. V.K. Morris: Conceptualization, data curation, writing-review and editing. K. Raghav: Conceptualization, data curation, writing-review and editing. L. Swanson: Conceptualization, data curation, writing-review and editing. H.J. Lim: Conceptualization, writing-review and editing. D.J. Renouf: Conceptualization, writing-review and editing. S. Gill: Conceptualization, writing-review and editing. R. Wolber: Data curation, methodology, writing-review and editing. A. Karsan: Conceptualization, data curation, writing-review and editing. S. Kopetz: Conceptualization, writing-review and editing. D.F. Schaeffer: Conceptualization, supervision, writing-review and editing. J.M. Loree: Conceptualization, data curation, formal analysis, supervision, writing-original draft, writing-review and editing.

J.M. Loree was the recipient of a Michael Smith Health Professional Investigator Award that supported this work. S. Kopetz is the recipient of NIH R01 grants that supported this research (CA 172670 and CA 187238).

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.