In the August, 2010, issue of The Lancet Oncology, Wendy De Roock and colleagues1 analysis of almost 900 patients with chemotherapy-refractory metastatic colorectal cancer assessed the role of KRAS, BRAF, NRAS, and PIK3CA in resistance to epidermal growth factor receptor (EGFR)-targeted therapy. The investigators should be congratulated for the scale of their study and the selected chemotherapy-refractory population. Their findings raise the question of whether these results should be immediately implemented in clinical practice.

KRAS is currently the only biomarker used to select patients for anti-EGFR monoclonal antibody treatment in clinical practice. Although there are several phase 3 randomised trials evaluating EGFR antibodies in metastatic colorectal cancer, consensus has not been reached regarding the value of downstream mutations for assessing resistance in KRAS wild-type patients. The role of BRAF, NRAS, and PIK3CA might never be validated prospectively for several reasons: underlying chemotherapy response creates difficulty when interpreting results, and mutational rates are low and recruitment of sufficient patients harbouring a rare mutation can take many years. There are ethical implications for any prospective study when large-scale retrospective data suggest little or no efficacy. The decision to incorporate routine KRAS testing of metastatic colorectal cancer into clinical practice followed conclusive results from retrospective analyses of trial data.2, 3 Should we now be making the same decision for BRAF, NRAS, and PIK3CA exon 20?

De Roock and colleagues data support the current opinion of these mutations on the basis of smaller studies,4 and the increase in response rate from 36·3% in KRAS wild-type patients to 41·2% in patients who were quadruple negative (ie, KRAS, BRAF, NRAS, and PIK3CA exon 20 wild-type) suggests the value of these additional tests, at least in the chemotherapy-refractory population. However, although a formal cost—benefit analysis of the tests is yet to be done, with the risk of administering high-cost toxic drugs in the setting of futility, it is difficult to argue that a full molecular profile should not be incorporated into treatment paradigms. The response rates of only 8·3% (2/24 patients) in BRAF mutants, 7·7% (1/13) in NRAS mutants, and 0% in PIK3CA exon 20 mutants, compared with 36—38% in wild-type patients, are consistent with published data. The absence of a relation between PIK3CA exon 9 mutation and cetuximab response, but the presence of cetuximab resistance in patients with exon 20 mutations might explain the contrasting results in previous studies that evaluated PIK3CA. The response rate of the unselected population is similar to the 22·9% reported in the BOND study,5 and it would therefore be interesting to confirm whether responding patients harbouring mutations were truly chemotherapy-refractory, because even in the KRAS mutant population the response rate was 6·7%.

In view of this new information, we suggest that analysis of these downstream genes is integrated into clinical practice for patients with chemotherapy-refractory disease in whom cetuximab or panitumumab is being considered; if not to definitively select quadruple-negative patients for therapy, then to introduce early response assessment or alternative strategies prior to EGFR treatments. Selection and stratification of patients on the basis of these mutations should also be introduced into future trials. Furthermore, evaluation of factors possibly contributing to enhanced response such as increased EGFR gene copy number or increased mRNA expression of epiregulin and amphiregulin could lead to an increased refinement of anti-EGFR therapy.

DC has received research funding from Amgen, Merck, and Roche. EH declared no conflicts of interest.