William Check, PhD
No one is asking today whether the government should do something to stimulate economic activity. Rather, the debate centers on method—how the stalled economy should be stimulated and which method would best reinvigorate the gridlocked financial system. A similar situation prevails with regard to KRAS mutation testing in metastatic colorectal cancer, or mCRC.
A consensus has emerged in the past year that this test should be performed in all individuals who have advanced CRC refractory to firstline regimens to identify those whose tumors will not respond to the costly monoclonal antibody inhibitors of EGFR (epidermal growth factor receptor). What is still a matter of debate is the optimal method for detecting clinically important mutations in the KRAS gene. Allele-specific PCR, standard Sanger sequencing, and pyrosequencing all have advantages and disadvantages, and all have their advocates.
To take this analogy one step further, economic stimulation and KRAS mutation testing each have a quantitative aspect. Even economists who agree that a governmental bailout will provide the best means of stimulation disagree on how much money should be provided. With KRAS mutation testing, how many mutations in the KRAS gene should be assayed is up for discussion. All agree that seven mutations in codons 12 and 13 of the KRAS gene should be covered. But there is some question about whether to test for an additional five rare mutations in codon 13. A further question is whether to assay for mutations in codon 61 of the KRAS gene. And what about mutations in other genes downstream of KRAS in the same pathway, such as BRAF or PIK3CA? To judge from their positions, mutations in these genes should have the same clinical consequences as mutations in KRAS. However, direct clinical evidence that mutations in these genes predict nonresponse to EGFR inhibitors is limited at best, and may never reach the level of proof now available for mutations in codons 12 and 13. Pathologists, in consultation with oncologists, must decide which mutations to test for and how to report them.
“KRAS” stands for Kirsten RNA Associated Rat Sarcoma 2 Virus Gene. It is one of many genes identified in RNA tumor viruses that have turned out to have homologues in humans and to be implicated in human cancers. KRAS lies downstream of EGFR, the transmembrane signaling protein that triggers multiple growth proliferation pathways; KRAS transmits signals from EGFR to the nucleus. In healthy cells, EGFR is stimulated by extracellular molecules such as epidermal growth factor. In cancer, abnormalities including overexpression of EGFR and mutations in KRAS and other genes allow proliferation to become constitutive and independent of growth factors.
To generate new therapies for cancers in which EGFR plays a role, such as non-small cell lung cancer and mCRC, two types of drugs have been developed against this protein:
- small molecules such as erlotinib (Tarceva) and gefitinib (Iressa) that block the intracellular tyrosine kinase domain of EGFR.
- monoclonal antibodies such as cetuximab (Merck, Erbitux) and panitumumab (Amgen, Vectibix) that block the extracellular domain of EGFR.
Taking a cue from breast cancer, in which overexpression of HER2 predicts a higher likelihood of response to the anti-HER2 monoclonal antibody trastuzumab (Herceptin), the clinical development program for cetuximab was focused exclusively on tumors with high EGFR expression, as measured by immunohistochemistry. As a result, Food and Drug Administration approval of cetuximab for mCRC was initially restricted to patients whose tumors expressed EGFR (from 70 to 75 percent of CRC patients). However, when results were analyzed from patients treated outside protocols, neither the presence nor the level of EGFR expression correlated with response to cetuximab (Cunningham D, et al. N Engl J Med. 2004;351:337–345; Chung KY, et al. J Clin Oncol. 2005;23:1803–1810).
Mutations in KRAS, on the other hand, have been shown to discriminate at a high level of significance between patients who respond to anti-EGFR monoclonal antibodies and those who do not. Taking several clinical studies in aggregate, some published and others reported only in abstract so far, roughly 40 percent of 558 patients with mCRC had a mutation in the KRAS gene. Only one patient with a mutated KRAS gene had a response. Among those with a wild-type KRAS gene, in contrast, response rates ranged from 16 to 17 percent with panitumumab and from 27 to 44 percent with cetuximab (Di Fiore F, et al. Br J Cancer. 2007;96:1166–1169; Amado RG, et al. J Clin Oncol. 2008;26:1626–1634; Lievre A, et al. J Clin Oncol. 2008; 26: 374– 379; De Roock, W, et al. Ann Oncol. 2008;19:508–515). Similar numbers were seen in a retrospective analysis of data from one of the pivotal clinical trials of cetuximab plus chemotherapy: Based on tissue samples from 394 (69 percent) of the 572 patients in the trial, investigators found that 42.3 percent had a mutation in the KRAS gene. Response rates were 12.8 percent among patients with a wild-type KRAS gene as compared with only 1.2 percent (one patient) among those with a mutated gene (Karapetis CS, et al. N Engl J Med. 2008;359:1757–1765).
Acceptance of the impact of KRAS mutations first arose after data were presented at the January 2008 Gastrointestinal Cancers Symposium of the American Society of Clinical Oncology, says Mark D. Pool, MD, medical laboratory director in the Department of Pathology at Riverside Medical Center, Kankakee, Ill., and adjunct professor of pathology at Rush Medical College. “After several similar reports were presented at ASCO, demand for KRAS testing took off,” Dr. Pool says.
“When the news came out of the ASCO meeting in June 2008, our oncologists came running and said, ‘We want this test,’” says Federico A. Monzon, MD, director of molecular diagnostics at The Methodist Hospital, Houston. “They were adamant about getting it done.”
At the meeting, the large number of presentations on KRAS mutations—there were at least 30 abstracts—is what “awakened the GI oncology community to the fact that this test is of great value in deciding whether to use EGFR-directed therapies,” says Bruce C. Horten, MD, medical director of Genzyme Genetics in New York. Growth in ordering was “explosive,” Dr. Horten says. “Within a matter of months KRAS went from a test that was sometimes ordered to a major item that was being ordered by GI oncologists across the country.” At Genzyme the number of CRC samples submitted for KRAS mutation testing more than doubled in the last months of 2008. Before ASCO, Genzyme was batching samples and running them once per week. Now it does a run every day.
In March Dr. Horten attended the 2009 meeting of the National Comprehensive Cancer Network. “People there talked about [KRAS mutation testing] as the current standard,” he says. “So at least within that group it is fully accepted.”
Jennifer Hunt, MD, section head of surgical pathology and head of molecular and anatomic pathology at Cleveland Clinic, shares that view. “The data are very strong that [testing for KRAS mutations] has to be done,” she says. “It doesn’t matter so much whether you send it out or do it in-house. But you will need to offer it.”
Official acceptance of the value of KRAS mutation testing was swift. Within six months of the ASCO meeting, guidelines from ASCO and the National Comprehensive Cancer Network endorsed KRAS testing for all patients with mCRC being considered for anti-EGFR monoclonal antibody therapy.
The FDA, however, has lagged behind. “While the guidelines established by ASCO and NCCN state that KRAS testing be performed before giving patients with metastatic colorectal cancer Erbitux or Vectibix, FDA is still considering appropriate product labeling regarding KRAS mutation testing for these targeted therapies,” Dr. Pool says.
Dr. Horten concurs. “Curiously, the FDA has been playing catchup,” he says. “They called a meeting in December to reconsider their now-outmoded requirement for IHC evaluation of EGFR expression for use of Erbitux in colorectal cancer. Meanwhile, the oncology, pathology, and managed care communities have moved on from IHC for EGFR to testing for KRAS mutations with scarcely a glance back at the FDA-issued package requirements for Erbitux in CRC.
“Clearly, FDA is unable to keep pace with the extremely rapid developments in personalized cancer therapy,” Dr. Horten says.
To Dr. Pool, the success of KRAS mutation testing has broad implications. “One important take-home message for me,” he says, “is that we finally have a molecular test for CRC that may allow us to catch up with breast cancer. Predictive and prognostic markers for breast cancer are already established, as well as a companion therapy. Now in colorectal cancer we have the potential for a similar situation.” He predicts it will extend to other solid tumors. “That is an exciting prospect. It will take time to mature and sort out, but I think we are on the right track looking at pathways for cancer response.”
With this development come opportunities and responsibilities for pathologists. One is education. “We as a profession can lead in communicating to oncologists research in KRAS and other mutations that link to predictive and prognostic markers,” Dr. Pool says. “We need to be aware of these markers, know what is being tested for, and why it’s important.” Even pathologists who do not test in-house need to be aware of the options. “Companies are aggressively marketing tests,” Dr. Pool says. “We need to know about technical and performance issues so we can pick an appropriate reference lab.”
Another pathologist who takes communication with oncologists seriously is M. Elizabeth Hammond, MD, attending pathologist at Intermountain Health Care, Salt Lake City. After she participated in an ASCO/CAP task force on accurate HER2 testing in 2007, Dr. Hammond took part in formulating an ASCO Provisional Clinical Opinion on testing for KRAS mutations in patients with metastatic CRC (Allegra CJ, et al. J Clin Oncol. 2009;27:2091–2096). “Laboratorians and oncologists see testing differently,” Dr. Hammond says. “We need to be informed about what treating physicians want. Without asking them, we can‘t know that.” (See “Before and after mutation detection,” above.) At the same time, Dr. Hammond was able to clear up oncologists’ confusion about such pathology issues as the best sample type for KRAS testing and what constitutes an optimal tissue block.
A paper on this topic has been accepted by the Archives of Pathology & Laboratory Medicine and will be published in 2009. In the meantime, a brief POET (Perspectives on Emerging Technology) report, prepared by the CAP Technology Assessment Committee, is available on the CAP Web site. Dr. Hammond and other members of a CAP ad hoc committee then prepared the more detailed white paper that will be published in Archives. “It describes the methods for testing for KRAS mutations in detail and the pros and cons of various methods,” Dr. Hammond says.
That’s what it comes down to right now—which test to do. Marc Ladanyi, MD, attending pathologist and chief of the molecular diagnostics service in the Department of Pathology at Memorial Sloan-Kettering Cancer Center, puts it succinctly. “The main discussion centers around methodology,” he says. “Basically, the question is whether to use direct sequencing or some other method. We have been using direct sequencing. If it is performed in a conscientious manner, this is a very reasonable approach.” However, Dr. Ladanyi cautions, “For this method there needs to be a very careful review of material being extracted to ensure a high tumor content—at least 25 percent. We aim for 50 percent.” (At this level, if one KRAS allele is mutated it will constitute 25 percent of KRAS DNA.) To meet that goal, all slides are reviewed by a pathologist, who identifies an area rich in tumor and dissects that area. “If you don’t have close interaction with surgical pathology, then you can run into problems,” Dr. Ladanyi says.
While some have said that direct sequencing has a higher failure rate, Dr. Ladanyi has not found this to be true. “We can get DNA good enough for PCR and sequencing from more than 95 percent of paraffin blocks,” he says.
For specimens with a low proportion of tumor cells, the laboratory report may say there is potential for missing a mutation. “We are also looking at methods with higher sensitivity,” Dr. Ladanyi says. He has set up KRAS assays on a Sequenom mass spectrometry-based genotyping platform. “We will transition to that platform in a few months,” he says. “That will get us down to detecting five to 10 percent tumor cells.”
When selecting a KRAS mutation assay, Dr. Ladanyi cautions, it is important to be aware of direct-to-oncologist marketing of diagnostic kits. “Lab directors might be placed in awkward situations where they have validated a particular technique but an oncologist hears a pitch from the vendor of another kit who tells them that’s the only one worth doing,” he says.
Dr. Hunt, too, uses direct sequencing as her method of choice. “You need a test that balances sensitivity with the ability to pick up all the mutations,” she says. “Sequencing has lower sensitivity but identifies all mutations.” She is comfortable using specimens where tumor constitutes as little as 20 percent. “This problem can be overcome by microdissection,” she says. Like Dr. Ladanyi, she uses microdissection on all samples: “We never throw a sample into a tube without microdissection,” she says.
Her objection to allele-specific PCR is that it detects only what is tested for, “what your primer pairs are designed to see.” It has good sensitivity, down to five percent or lower. “But if you don’t include a primer for a specific mutation, you will not see it,” she says.
At Genzyme, the preferred method is single-nucleotide primer extension using a modified ABI SNaPshot assay. Dr. Horten says assay sensitivity is 10 percent, and 20 percent tumor is optimal for evaluation. A pathologist reviews tissue sections and relevant tumor is selected for analysis. All samples are microdissected and the DNA extracted from the tumor tissue is quantified by real-time PCR before analysis. Specimens that do not meet the minimum DNA quantification established for the assay are failed—and the failure rate is 1.5 percent. All 12 mutations in codons 12 and 13 are sought. Dr. Horten reports that the five rare mutations made up 30 (1.1 percent) of the 2,611 mutations found in 6,401 samples analyzed.
Kenneth J. Bloom, MD, chief medical officer at Clarient Inc., chose allele-specific PCR as that laboratory’s KRAS mutation assay. “We had already worked up a couple of KRAS assays before the ASCO meeting. When the data on KRAS testing came out, we were presented with the classic problem—science versus clinical data.” In principle, the best way to detect all mutations in the KRAS gene would be sequencing. However, the clinical work presented at ASCO used the DxS kit. In addition, Dr. Bloom says, “A limitation of sequencing is the sensitivity issue—it requires a relatively large amount of mutated DNA relative to normal DNA present in the sample.” He asked himself, Do we need to capture every mutation or only those mutations that have been clinically shown to predict lack of responsiveness? “In the end we decided to do allele-specific PCR for the seven mutations in the kit,” Dr. Bloom says. Clarient is running 20 to 40 samples per day; its failure rate is under two percent.
In the past few months Linda Sabatini, PhD, technical director for molecular pathology at ACL Laboratories in suburban Chicago, has brought up real-time PCR with melt curve analysis to screen for KRAS mutations. “It is very specific and very reliable,” Dr. Sabatini says. “We can quickly report out all wild-type results. When we see an abnormal melt curve, we confirm it with sequencing.” Confirmation by sequencing adds a couple of days to turnaround time, but Dr. Sabatini intends to continue doing it for now. “It may not be necessary,” she acknowledges, “but we want to gain more experience before we make that determination.” She notes that the COSMIC (Catalogue of Somatic Mutations in Cancer) database, which lists 5,000 CRC cases, documents very rare point mutations in the KRAS gene outside of codons 12 and 13.
To Dr. Sabatini, there is an extra benefit of doing the test: “Our pathologists are delighted to have it in-house. They like being able to interact directly with us.”
Shuji Ogino, MD, PhD, associate professor of pathology and a member of the Center for Molecular Oncologic Pathology at Harvard Medical School, Brigham and Women’s Hospital, and the Dana-Farber Cancer Institute, has validated and adopted detection of KRAS mutations by pyrosequencing. “When I started using pyrosequencing, the issue was poor-quality specimens,” Dr. Ogino says. “We needed a more sensitive assay. With Sanger sequencing the signal-to-noise ratio is low, so it requires 20 percent or more tumor DNA in the sample.” In a comparison study, Dr. Ogino and his colleagues found that pyrosequencing has greater sensitivity than Sanger sequencing, with a detection limit of about five percent to 10 percent mutant alleles (Ogino S, et al. J Mol Diagn. 2005;7:413–424). They use a homebrew assay with their own primers and a commercial Biotage platform from Qiagen.
Dr. Monzon, too, plans to set up a pyrosequencing assay. He is sending the test out now but intends to bring it in-house in the next few months. “The only reason we haven’t brought it in so far,” he says, “is that we have other higher-volume tests that we needed to set up first.”
In Dr. Monzon’s view, sequencing is the gold standard. He is uncomfortable using a targeted assay. “The number of samples analyzed and reported in the literature is still small,” he says. “We might not have seen every mutation.” He favors pyrosequencing over Sanger sequencing. “Pyrosequencing is very economical when you have a short read length,” he says, “and it gives slightly better sensitivity.” Like Dr. Ogino, he will use the Biotage instrument with his own reagents. He wants to modify the PCR reaction to increase sensitivity by selecting for mutant alleles during amplification. “Mostly I’m concerned about mutations in codons 12 and 13 other than those that have been reported,” Dr. Monzon says. When mutations outside codons 12 and 13 are found, he adds, “I will need to research them. Have they been described as activating mutations in the literature?”
It is impossible at this point to say that one of these methods is best. “There hasn’t been a lot of head-to-head comparison,” Dr. Pool notes. Direct sequencing is the reference method, but it is time-consuming and cumbersome. Allele-specific PCR is focused on the seven (DxS) or 12 (Genzyme) most common mutations in codons 12 and 13. Dr. Pool sends samples to the Mayo Clinic reference laboratory, which uses the DxS method.
Dr. Monzon says the Association for Molecular Pathology is planning a sample exchange study to address analytical sensitivity and performance of the various platforms.
More problematic than analytical sensitivity is clinical sensitivity. Even when patients with KRAS alleles mutated in codons 12 and 13 are excluded, only 40 percent of patients treated with monoclonal antibody inhibitors of EGFR respond. Can more nonresponders be identified with additional biomarkers?
Dr. Horten of Genzyme intends to add mutations in codon 61 of the KRAS gene early next year. “To date, the significance of specific KRAS mutations is unknown,” he says. However, he contends that EGFR-directed therapy is excluded if any KRAS mutation is detected. Not everyone agrees. Dr. Monzon says, “The jury is still out on whether [mutations in codon 61] are clinically significant. There needs to be more studies on that.” And the ASCO Provisional Clinical Opinion lists only mutations in codons 12 and 13 as having sufficient evidence to justify exclusion of anti-EGFR monoclonal antibody therapy. More important, mutations in codon 61 are very uncommon, with a frequency of perhaps 0.5 percent, Dr. Sabatini says.
Mutations in the gene for BRAF, which codes for a protein downstream of KRAS, could explain another 10 percent of nonresponse. One group reported that, of 79 mCRC patients with wild-type KRAS, the BRAF V600E mutation was found in 11, none of whom responded to treatment with cetuximab or panitumumab (Di Nicolantonio F, et al. J Clin Oncol. 2008;26:5705–5712). According to Dr. Bloom, mutations in BRAF and KRAS appear to be mutually exclusive in colon carcinoma. “In all of the colon samples tested thus far, we haven’t found a single tumor that has both a KRAS and a BRAF V600E mutation,” he says.
Even accepting that BRAF mutations cause nonresponse, that would still leave about half of nonresponse in patients with wild-type KRAS unexplained. Clinical researchers are actively seeking other candidates. Dr. Ogino and his colleagues showed poor prognosis following curative resection in CRC patients who had a mutated gene for PIK3CA, a protein in an EGFR-stimulated proliferation pathway parallel to the one in which KRAS functions (Ogino S, et al. J Clin Oncol. 2009;27:1477–1484). They speculated: “The adverse effect of PIK3CA mutation may be potentially limited to patients with KRAS wild-type tumors.” Another group demonstrated that, in a cohort of 110 mCRC patients, 15 had mutations in the PIK3CA gene. All 15 had clinical resistance to EGFR-targeted monoclonal antibodies (Sartore-Bianchi A, et al. Cancer Res. 2009;69:1851–1857). “When the molecular status of the PIK3CA/PTEN and KRAS pathways are concomitantly ascertained,” these investigators wrote, “up to 70% of mCRC patients unlikely to respond to EGFR moAbs can be identified.”
Dr. Pool cites data suggesting that mCRC patients who express high level of the EGFR ligands epiregulin and amphiregulin in the presence of two wild-type KRAS alleles are more likely to respond to cetuximab (Khambata-Ford S, et al. J Clin Oncol. 2007;25:3230– 3237). “Work on these molecules is not mature yet,” he cautions. “They will need to be looked at in a much larger study.
“In general, we need to start to think more about cancer in terms of pathways,” Dr. Pool says. “EGFR sits on the surface of the cell. It can’t act without downstream elements. If any part of that system is permanently switched on, EGFR inhibitors won’t work.” Applying this idea to clinical practice, Dr. Pool says, “I can envision a panel of these markers being done to determine who will respond.”
Dr. Bloom notes that all of these ideas are attractive and promising but not proven. Scientifically it makes sense that mutations in other molecules in EGFR-stimulated pathways would have the same impact on therapy with anti-EGFR monoclonal antibodies as mutations in KRAS. “However, the clinical evidence is currently lacking,” he points out. “That’s the fundamental problem with sequencing.” While sequencing picks up all mutations, as opposed to the limited set detected by targeted assays like allele-specific PCR, data don’t exist to support the clinical benefit of pan-mutation sensitivity. Though it seems that all KRAS mutations are activating, mutations in other genes affecting the EGFR pathway may be activating, deactivating, or have little to no effect on protein function. Dr. Bloom doubts that such data will ever exist. “I don’t think they will ever go back to the original clinical trial dataset and fully analyze the other mutations,” he says. The problem is that mutations such as those in KRAS are negative predictive biomarkers—they select patients in whom drugs won’t work. “I don’t think drug companies [that make anti-EGFR monoclonal antibodies] are interested in conducting additional studies that will expand the number of patients ineligible for therapy,” Dr. Bloom says. “Researchers will continue to do single-institution trials, but single-institution data will never replace the robustness of clinical trials data.”
Dr. Hammond makes the same point. “Just because one person has data that show something about a mutation and EGFR inhibitors, that’s not clinically conclusive,” she says. “It needs to be shown in a clinical trial that patients with the mutation don’t respond and patients without the mutation do respond. That’s what you need for level I evidence. Pathologists shouldn’t change the rules on the basis of evidence lower than level I.”
If a laboratory does test for mutations other than those in codons 12 and 13 of the KRAS gene, how should they report the results? “I think that is a valid point,” Dr. Ladanyi says. “Probably we would have to indicate in a note that there are no specific data for those less common mutations. Clinicians will have to make their own judgments.”
Merck and Amgen may not be happy about the search for additional negative predictive markers for anti-EGFR monoclonal antibodies, but it is an important task. “These therapies are not cheap,” Dr. Pool says. “And we don’t want to expose patients to potentially toxic drugs.”
And what of oncologists—are they willing to withhold anti-EGFR monoclonal antibody treatment from patients with metastatic CRC who have failed a first-line regimen? “There is still some discussion whether a trial of one of these agents is nonetheless worthwhile in a patient with a KRAS mutation,” Dr. Ladanyi says. “More savvy or more interested clinicians are definitely using this test as a way to exclude patients from treatment with these agents.”
Dr. Hammond has a similar perception. “The oncologists I was talking to [on the ASCO task force] believe that no one [with a KRAS mutation] should be treated with these drugs,” she says. “They constitute a huge expense and are potentially dangerous.”
In Dr. Horten’s view, this question will be determined to a large extent by managed care. “United Healthcare has adopted a policy through Dr. Newcomer [Lee Newcomer, MD, senior vice president, oncology] that requires a lab report to be sent to their office,” he says. “If there is no mutation in KRAS, they will authorize payment of Erbitux. If there is a mutation, they will deny payment. I think other insurers will follow suit.” This approach is part of United Healthcare’s acceptance of NCCN guidelines and already applies to HER2 and Herceptin (www.uhc.com).
Cost-benefit considerations clearly argue for conservative prescribing of anti-EGFR monoclonal antibodies. Authors of an editorial accompanying an article on the ability of KRAS testing to distinguish responders from nonresponders called the increase in survival attributable to cetuximab “small” and “modest” (Messersmith WA, Ahnon DJ. N Engl J Med. 2008; 359: 1834– 1836). In patients with wild-type KRAS alleles, therapy increased median overall survival from 4.8 months to 9.5 months.
Cost and reimbursement issues will become even more acute if anti-EGFR monoclonal antibodies become part of treatment in primary CRC, a setting in which they are now being tested. Test volume would rise as well.
Even with all of these uncertainties and complications, Dr. Monzon predicts the use of molecular testing to qualify patients for expensive biologic therapies will continue to expand. “It’s going to be one of the most important growth areas in molecular diagnostics,” he says. “What some people call companion tests to define drug eligibility have existed for a while. Now they are becoming a more recognized category.” He would argue that the first companion test was detection of estrogen receptor to determine whether to prescribe tamoxifen. Then came measurement of HER2 overexpression and amplification in breast cancer and, more recently, EGFR mutations in lung cancer. Now there are the assays for KRAS mutations, perhaps to be followed by BRAF. As drug development becomes more linked with biomarkers, he says, the number of companion tests for cancer will grow.
“I think that as we learn more about molecular mechanisms of response and nonresponse, we will identify more biomarkers to tell us who is more likely to benefit from a specific therapy,” Dr. Monzon says.
Amgen and Merck are asking FDA’s approval to update the labeling of the two monoclonal antibodies to recommend testing tumor cells for KRAS mutations. “Asking FDA to limit the number of patients to whom physicians can give a drug—that is an interesting twist,” he says. “That is getting more responsible and increasing the response rate.”
William Check is a medical writer in Wilmette, Ill.