Skip to content

Technology Evaluation
Center (TEC)

Special Report: Companion Diagnostics: Example of BRAF Gene Mutation Testing to Select Patients with Melanoma for Treatment with BRAF Kinase Inhibitors

Executive Summary


Drug discovery and development generally follows one of two different paradigms: physiology-based discovery or target-based discovery. Physiologic-based drug discovery screens a large number of compounds using a physiologic readout such as an animal model or a cellular assay. Target-based drug discovery begins by identifying the function of a possible therapeutic target, such as an enzyme or a cell surface receptor, and its role in disease. Particularly in the latter case, patients often must be selected for treatment using a laboratory test (“companion diagnostic”) that identifies which patients have the target.

Examples of target-based discovery include imatinib, originally designed to inhibit the constitutively activated kinase product of the abnormal fusion gene BCR-ABL found in chronic myelogenous leukemia (CML); and trastuzumab, an antibody designed to bind to HER2 receptors overexpressed on the cell surface of some types of breast cancer. In the case of imatinib, however, as nearly all patients with CML harbor the BCR-ABL gene, a companion test to initially select patients for treatment is largely unnecessary. The HER2 companion test initially commercialized was not used in the clinical trials of trastuzumab and the subjectivity component of test readout was not initially appreciated, resulting in the design of improved HER2 tests and extended analyses of the quality and utility of HER2 companion testing. In some cases, companion test development and evaluation is retrospective, as in the case of KRAS mutation testing to select colorectal cancer patients likely to respond to cetuximab or panitumumab treatment. Thus, these are not the best examples of treatment-test co-development.

Recent examples of targeted drug design and companion test co-development (Table) include crizotinib (Pfizer, Inc.), which targets an abnormal fusion gene product in some patients with nonsmall cell lung cancer, and vemurafenib (Hoffmann-La Roche Ltd.) and dabrafenib (GlaxoSmithKline plc), both BRAF inhibitors that target a mutated form of the BRAF kinase, derived from the mutated BRAF gene (BRAFV600E) occurring in 40–60% of melanoma tumors. This mutation results in a change from valine to glutamic acid at amino acid position 600 in the BRAF protein. As these products were in development, the U.S. Food and Drug Administration (FDA) was developing a draft guidance for in vitro companion diagnostics, the major principles of which were likely applied to products nearing FDA submission, including crizotinib and vemurafenib. This guidance is important not only for its applicability to targeted drug development, but also for its impact on a broader category of “personalized” treatments that use a companion test to select patients most likely to respond, or to avoid treating patients likely to have serious adverse reactions.

Table.  Examples of Drugs Requiring Companion Diagnostic Tests



Molecular Indicator of response/no response

Companion Test


Chronic myelogenous leukemia

(other indications not listed here)

BCR-ABL kinase

Qualitative detection of BCR-ABL fusion transcript


Breast cancer


Protein staining by immunohistochemistry; Gene amplification by FISH*

Cetuximab, Panitumumab

Advanced colorectal cancer

(other indications not listed here)


Mutation detection, various methods


Advanced non-small cell lung cancer

ALK fusion proteins

ALK gene rearrangements by FISH


Advanced melanoma

BRAF V600 mutation

Real-time PCR BRAF mutation detection

 FISH: fluorescence in situ hybridization; PCR: polymerase chain reaction


An important question is whether the co-development process ensures sufficient validation of the companion diagnostic test, such that an independent evaluation becomes unnecessary. This Report will address that question using a targeted drug discovery example: the recently approved vemurafenib for certain patients with melanoma and its co-approved companion diagnostic test.

Search Strategy

The MEDLINE® database was searched (via PubMed) for articles using the terms “PLX4032,” “vemurafenib,” “V600E,” and “BRAF inhibitor,” all coupled with the term “melanoma.” The reference lists of relevant study publications and review articles were also examined. The meeting abstracts for the 2011 annual meeting of the American Society of Clinical Oncology were searched using the MEDLINE® search terms. If available, virtual presentations and slides were reviewed for key abstracts. The “grey literature” was consulted in the form of drug and laboratory test approval information released by the FDA, ongoing clinical trials from, and online searches for status and ancillary information.

Selection Criteria

For primary clinical data, we looked for trials of vemurafenib that used the final version of the companion test to select patients. Trials that used a prototype assay were also considered.

Main Results

The cobas® 4800 BRAF V600 Mutation Test real-time polymerase chain reaction (PCR) test is intended to detect the BRAFV600E mutation in formalin-fixed, paraffin-embedded (FFPE) tumor specimens from patients with advanced or metastatic melanoma and return a positive (mutation detected) or negative (mutation not detected) result. Once the test was co-approved with vemurafenib, extensive and detailed evidence regarding the analytic validity (technical performance) of the assay was available in the FDA Summary of Safety and Effectiveness and from the kit insert. Analytic validation data support a sensitive and robust assay for the detection of the V600E mutation in FFPE melanoma specimens.

One limitation was lack of complete data for calculating sensitivity and specificity. Sanger sequencing was used as a gold standard for FDA submission studies, but due to insensitivity at lower percentages of mutated alleles is an inappropriate standard. For the Phase III trial, samples from patients screened for the trial were sequenced, including those which were BRAFWT; for the Phase II trial, only enrolled patients were sequenced, thus only percent agreement between the two methods could be calculated. Subsequent studies indicated a high percent agreement between the cobas 4800 BRAF V600 Mutation Test and more sensitive sequencing methods, but insufficient information was available to allow recalculation of the Phase III sample data. Comparison with sequencing showed that in addition to very high percent agreement for the V600E mutation, the companion test also identified some, but not all, V600 variant mutations (e.g., V600K) as positive. Tumors with V600 variant mutations have a different amino acid substitution in the BRAF kinase.

When a therapeutic treatment is developed for a specific biological target that characterizes only some patients with a particular disease, and a test is co-developed to identify diseased patients with that target, studies supporting clinical validity (defined as the ability of the test to predict the disorder or phenotype of interest) and clinical utility (evidence of improved measurable clinical outcomes with test use) are no longer separate and sequential. Rather, the clinical studies of treatment benefit, which use the test to select patients for enrollment, provide evidence of both clinical validity and clinical utility. The primary evidence of clinical validity and utility for the cobas 4800 BRAF V600 Mutation Test is provided by the Phase III clinical trial of vemurafenib, which also supported the FDA approval of the drug. In addition, evidence from Phase I and Phase II trials was reviewed. All trials used enrichment trial designs, in which patient enrollment was based on a positive result for a V600 mutation in a tumor specimen.

The Phase I single-arm clinical trial of vemurafenib used a prototype assay to detect BRAFV600E mutations in enrolled patients. After dose determination, the extension phase of the study resulted in 81% of 32 patients responding according to Response Evaluation Criteria in Solid Tumors (RECIST); nearly all were partial responses.

The Phase II single-arm clinical trial is currently ongoing; interim results presented at a meeting showed a 53% objective response rate, median progression-free survival of 6.7 months, and median overall survival not reached at the time of analysis. Patients were selected for enrollment based on a finalized version of the cobas 4800 BRAF V600 Mutation Test.

The Phase III comparative trial of vemurafenib versus standard chemotherapy (dacarbazine) also enrolled patients based on the results of the finalized companion test. At a planned interim analysis, the results met the specified criteria for primary endpoints, and patients in the dacarbazine treatment arm were allowed to cross over to vemurafenib. At this time, median survival had not been reached; the hazard ratio for death was 0.37 (95% CI: 0.26–0.55). At 6 months, overall survival was 84% (95% CI: 78 to 89) for vemurafenib-treated patients and 64% (95% CI: 56 to 73) for dacarbazine-treated patients. Progression-free survival was evaluable in 549 patients; the hazard ratio for tumor progression was 0.26 (95% CI: 0.20–0.33). The median progression-free survival was estimated to be 5.3 months for patients treated with vemurafenib and 1.6 months for patients treated with dacarbazine. Tumor response was evaluable in 439 patients; the objective response rate was 48% in patients treated with vemurafenib versus 5% in those treated with dacarbazine. Only 2 patients treated with vemurafenib had a complete response.

A very small number of patients with V600 tumor mutations other than V600E were included in the Phase II and III clinical trials. No subgroup analyses were performed, although it was noted that some of these patients had at least partial responses.

Author’s Comments and Conclusions

The targeted drug-companion diagnostic test co-development process is a subset of a broader category of “personalized” treatments that use a companion test to select patients most likely to respond, or to avoid treating patients likely to have serious adverse reactions. For example, an evaluation of HLA-B*5701 testing as a companion test for abacavir resulted in the exclusion of test-positive patients from abacavir treatment and the elimination of immunologically confirmed, serious hypersensitivity reactions. In this case as in other examples, the companion test was developed and evaluated after drug approval. Going forward, as the molecular mechanisms and genetic influences of treatment actions are increasingly understood, companion tests are more likely to be developed in tandem. Targeted drug development, which depends on such an understanding, offers early examples of this process. In recognition, the FDA has provided draft guidance for a parallel and coordinated review and approval process in which both therapeutic treatment and diagnostic test must be approvable, or neither will be approved. As justification, the FDA notes that when use of the treatment depends on the results of the companion test, then there are important concerns regarding the safety and effectiveness of both the therapeutic treatment and the companion diagnostic test.

This Report examined the data supporting the co-review and approval process using vemurafenib and its companion test, the cobas 4800 BRAF V600 Mutation Test, for the treatment of advanced/ metastatic melanoma. Based on this example, general caveats to the co-development process require discussion. First, the vemurafenib Phase II and III clinical studies were all of enrichment design, enrolling patients based on a positive result from the companion test. The feasibility of this trial design relies on the knowledge from targeted drug design and from preclinical studies that melanoma tumors with a V600 mutation are likely to respond to vemurafenib, whereas tumors without a V600 mutation are unlikely to respond. As a result, smaller, more efficient trials can be conducted to determine drug efficacy. The trade-off, however, is the loss of information regarding the effect of vemurafenib on V600 wild-type melanoma tumors. Although preclinical data suggested that vemurafenib may potentiate tumor growth in V600 wild-type tumors, the comparable effect in patients remains unknown. Thus, the co-development process results in a narrow indication for treatment.

A limitation of the vemurafenib clinical studies was the absence of data on tumor tissue heterogeneity or homogeneity for the BRAFV600E mutation. The cobas 4800 BRAF V600 Mutation Test provides qualitative mutation detection using extracted DNA from a melanoma tissue specimen. Although it is highly sensitive for samples with low percent mutation in studies of analytic validity, clinical trials did not evaluate percent mutation within tumors, i.e., tumor heterogeneity, and correlate that result with response/nonresponse, or degree of response. Studies published by other groups suggest that BRAF mutations are early events in the development of melanocytic tumors, but alone are insufficient to cause primary tumor formation and progression. Tissue from primary tumors positive for BRAF mutations appears to be heterogeneous at a cellular level, but related metastases are consistently BRAF positive if the primary tested positive. Immunohistochemistry studies suggest that metastases are homogenous in terms of mutated protein reactivity, but intensity of staining can be variable, and a more quantitative correlation of BRAF mutation with treatment outcome may be useful. Although this information is preliminary, it is conceivable that improved companion tests or tests that enhance the information provided by the basic companion test may be developed; whether the co-development review process allows for such development is unclear.

In recent years, molecular tests have been carefully evaluated by groups like the EGAPP Working Group1 and the Agency for Healthcare Research and Quality’s various evidence-based assessment programs.2 This is because FDA review of diagnostic tests does not usually encompass clinical utility and, in fact, most molecular tests are laboratory-developed tests not submitted to FDA for review.

The companion test development process results in the selection of patient groups that are either highly likely or highly unlikely to respond, seemingly a special case of a predictive test. And while targeted drug development does represent a new era of structural drug design that is more likely to elicit strong and predictable responses, general considerations of test evaluation are not unique. In general, predictive tests attempt to separate a patient population into those with a higher versus a lower likelihood of response to a treatment. Often the separation of groups is insufficient for clinical utility for individual patients, although separation of groups may be statistically significant (clinical validity). Targeted drugs increase the likelihood of response such that clinical utility is more often shown.

This Report specifically asked whether the companion diagnostic co-review and approval process generated sufficient evidence of clinical utility in the example reviewed, such that additional evaluation would not be necessary for future companion tests approved via this process.

The cobas 4800 BRAF V600 Mutation Test was developed sufficiently early in the overall drug-test validation process such that a finalized version of the test was used in both Phase II and key Phase III clinical studies, the latter of which resulted in the clinical data submitted to the FDA. The companion test was co-submitted as a Class III device, requiring a higher level of support data. This was likely required because of potentially higher risk to the patient based on test results than for other types of laboratory tests that are classified as Class I or II devices. Extensive data support analytic validity, the technical performance of the test.

The results of the Phase III trial, supported by the results of the earlier trials, support the clinical validity and clinical utility of the cobas 4800 BRAF V600 Mutation Test, the companion diagnostic test for vemurafenib. Using the test to select patients for treatment results in improved outcomes compared to the usual standard of care, dacarbazine. In addition, comparison of these results with the trial results of the recently approved ipilimumab, suggests that treatment with vemurafenib results in improved outcomes compared to ipilimumab. Ipilimumab is notable as the first therapy to show a survival advantage in a Phase III trial for patients with advanced melanoma, and while vemurafenib was in clinical trials, may have become the new treatment standard for late stage disease and thus is an important comparator. Currently, a trial is underway testing the combination of vemurafenib and ipilimumab ( Identifier NCT01400451).

Some important limitations and gaps in knowledge were identified:

* As noted, the vemurafenib companion test identifies some melanoma tumors carrying BRAF V600 mutations other than V600E as positive; follow-up on vemurafenib response in this patient population will be important.

* Use of commercially available, non-FDA-approved laboratory developed tests for the BRAFV600E mutations to select patients for vemurafenib treatment is not straightforwardly addressed by the FDA guidance on in vitro companion diagnostics or by other in vitro diagnostic or clinical laboratory regulations. It is possible that several different tests could be used for this purpose; impact on treatment outcomes is unknown.

* Knowledge of the impact of tumor heterogeneity with respect to the BRAF mutation on vemurafenib treatment outcomes may inform patient selection.

* In general, it should be remembered that the development process, using enrichment trial designs, is an efficient process that provides information regarding patients with the molecular target, but provides no information regarding all other patients, who may have received all other treatment possibilities, and for whom any new available treatment may be attractive.

We conclude that the FDA process for coordinated review and co-approval of both therapeutic treatment and associated companion diagnostic test ensured sufficient validation of the companion diagnostic in the case of vemurafenib and its companion test, the cobas 4800 BRAF V600 Mutation Test. Important limitations and knowledge gaps specific to the example, and to the co-development process in general were identified. We further conclude that review of the final version of the FDA guidance document on in vitro companion diagnostics as well as additional specific examples will be important to determine whether the co-development process will be sufficient to ensure adequate evidence of clinical utility for all companion diagnostics approved via this regulatory review pathway.

1. Evaluation of Genomic Applications in Practice and Prevention, information available at

2. For example, Effective Health Care and Evidence-based Practice programs; U.S. Preventive Services Task Force; see

Full Study


Full studies are in PDF format. You will need Adobe Acrobat Reader to view all studies.
Adobe Acrobat Download Adobe Acrobat Reader here.

antimelanoma; B-Raf; carboplatin; cobas; co-developed; EGAPP: imidazole; ERK; genetic therapeutic index; ipilimumab; MAPK; MEK; melanoma. V600; metastatic; mutant; mutated; mutation; oncogenic; paclitaxel; PLX4032; RAF; sorafenib; transactivate; unreseactable; vemurafenib; wild-type; WT;