By Turna Ray
Pharmacogenomic tests to guide treatment with Herceptin for breast cancer, Gleevec in chronic myelogenous leukemia, and Vectibix and Erbitux for colorectal cancer improve patient outcomes, an advisory panel for the Centers for Medicare and Medicaid Services decided last week. In its review of these and two other tests, however, the panel identified numerous evidence gaps that CMS and private payors would likely require before covering them more broadly.
Although CMS does not have a national coverage determination on pharmacogenomic testing in cancer, the Medicare Coverage and Evidence Development Advisory Committee, or MEDCAC, met last week in Baltimore, Md., to discuss whether five tests are clinically useful in the Medicare population.
Of the five tests — CYP2D6 for breast cancer patients who are candidates for tamoxifen; UGT1A1 for colon cancer patients who are candidates for irinotecan; HER2/neu for breast cancer patients who are candidates for trastuzumab; BCR-ABL for CML patients who are candidates for imatinib (Gleevec); and KRAS for metastatic colorectal cancer patients who are candidates for cetuximab (Erbitux) or panitumumab (Vectibix) — the panel imparted strong confidence that the latter three were clinically useful in certain cancer populations.
It seems for pharmacogenomic cancer testing, however, that MEDCAC's recommendations are unlikely to inspire any immediate coverage decisions from CMS, though the panel's conclusions and concerns will inform CMS' future considerations of such tests.
"We purposefully held this meeting ahead of a [national coverage decision] to try to get a sense of where the evidence stands," said Louis Jacques, director of CMS analysis and coverage group, at the end of the meeting. He quipped that stakeholders in the audience should "get some sleep," because CMS isn't going to make a coverage decision any time soon. "We intentionally chose topics with heterogeneity to see how the panel would react," Jacques explained.
The evidence on three of the five PGx cancer tests — CYP2D6, BCR-ABL, and KRAS testing — were gathered for CMS through a systematic review of the Agency for Healthcare Research and Quality's Technology Assessment Program. Such technology assessments are funded by an interagency agreement between CMS to AHRQ and are used to inform national coverage policies and MEDCAC reviews. The technology assessment team was not asked to review HER2 or UGT1A1 testing; MEDCAC was provided data on those tests separately.
Thomas Trikalinos, assistant director of Tufts-New England Medical Center's Evidence-based Practice Center, led the technology assessments on the three tests. Trikalinos' technology assessment team searched Medline abstracts published through August 2009 using key words and criteria in order to look for outcomes associated with the tests, such as mortality and progression-free survival. The review team considered in its assessment 13 studies on CYP2D6 testing, and 31 studies each for KRAS and BCR-ABL testing.
The team attempted to look for studies that could answer the following questions about the three tests: Does the genetic test result predict response to therapy? What patient- and disease-related factors affect the test results, their interpretation, and response? How does the gene testing impact the therapeutic choice? What are the benefits and harms or adverse effects for patients when managed via genetic testing?
However, the tech assessment team was only able to find studies to inform the first question: whether or not the genetic test predicts response to therapy. They did not identify studies that informed whether patients' disease factors impact test results, whether genetic testing can guide treatment choice, or if there are benefits and harms associated with testing.
According to Trikalinos, the overall takeaway from the tech assessment was that most studies to gauge whether genetic tests predict treatment response are often limited by small sample sizes; key markers tend to be rare genetic markers, and treatment-by-gene interactions were often not assessed.
"It may not always be necessary to conduct new studies of 'repurposed' [randomized controlled trial] drugs of interest against a suitable comparator," Trikalinos said. He added that 'repurposed RCTs,' which are essentially retrospective studies using samples collected from patients while they are undergoing an RCT, may be as good as a prospective RCT. This strategy may be particularly useful for studying pharmacogenetic effects of somatic mutations, Trikalinos noted.
However, "repurposed RCTs" cannot measure how testing affects patient outcomes, Trikalinos emphasized, adding that if this strategy is undertaken, independent validation to control false-positive findings will be necessary.
At the end of the technological assessment and discussion of the presented evidence, the 12 voting members of the MEDCAC panel voted on whether there was sufficient evidence to determine whether pharmacogenomic testing in each of the five examples simply affects health outcomes, in terms of benefits or harms, for cancer patients. Panelists were asked to rate their confidence level: 1 for "low confidence, 2-3 for "intermediate confidence," and 4-5 for "high confidence."
For those tests for which the advisory group imparted "intermediate" confidence (mean score greater than 2.5), they were then asked to indicate their confidence level that these tests actually improved outcomes in cancer patients. Finally, panelists were asked to vote on whether the conclusions about the tests are generalizable to the Medicare population.
After hearing all the presentations and expressing their confidence for various tests, MEDCAC members identified several major areas of evidentiary gaps for PGx cancer tests. Panelists noted the need for prospective randomized controlled trials when possible and rigorous retrospective trials when appropriate; more PGx studies in the Medicare-age population that factor in comorbidities that affect this population; and better data on functional outcomes.
Elizabeth Mansfield, director of personalized medicine at the Office of In Vitro Diagnostic Evaluation in FDA's Center for Devices and Radiological Health and a guest MEDCAC panelist, urged CMS and FDA to work together to develop evidence requirements for registration trials. Furthermore, she requested sponsors collect biospecimens that are well-annotated with pharmacy and outcomes data. This will facilitate more retrospective studies, she said, adding that "FDA is on the verge of requesting this."
Elaine Jeter, a guest panelist and a contractor medical director for Palmetto GBA, pointed out that the level of evidence MEDCAC reviewed for these PGx cancer tests should give sponsors an idea of how high the bar is for payors to provide coverage for a test. Palmetto GBA administers Medicare health insurance for CMS.
Jeter estimated that diagnostic developers are currently vying to garner coverage for as many as 1,500 tests, and the majority of them submit marginal to poor evidence on the clinical utility of these assays. "This gives you an expectation of what CMS and the community wants and needs for evidence-based decision making," Jeter said.
[ pagebreak ]
In looking through studies on CYP2D6 testing for breast cancer patients treated with tamoxifen, the tech assessment team found highly heterogeneous study populations, in terms of stage, age, ethnicity, and lymph node and estrogen-receptor distribution. Outcome assessment in these studies was usually retrospective, and reviewers found two studies that were "repurposed randomized-controlled studies."
Although 11 of 13 studies looked at breast cancer recurrence, most of the studies did not report an association between CYP2D6 status and disease recurrence, the review showed. Studies seeking a mortality endpoint did not report a significant impact from genetic testing on mortality, Trikalinos reported. The two repurposed RCTs did not report any interaction tests, and furthermore the studies did not analyze whether the predictive ability of the gene was restricted to those treated with the drug.
"There was an inconsistent association between CYP2D6 and outcomes," Trikalinos' team concluded based on their review, adding that it was "unclear" whether CYP2D6 can predict outcome to tamoxifen treatment.
Based on the tech assessment team's findings and data reported in a paper published in the Journal of the American Medical Association by Schroth et al. in October, the majority of the MEDCAC panel did not find that CYP2D6 testing improved outcomes for tamoxifen-treated breast cancer patients. As a result, CYP2D6 testing received an average confidence score of "2" from voting members.
"In the CMS population, the stakes aren’t that high, aromatase inhibitors are also good," Dan Hayes, MEDCAC member and director of the breast oncology program at the University of Michigan, Ann Arbor, said during the meeting. Hayes pointed out that the stakes are a lot higher in younger women who cannot be treated with aromatase inhibitors. Therefore, it may be more critical in the younger population to establish CYP2D6 status and tamoxifen response.
In reviewing 31 studies investigating KRAS testing in colorectal cancer patients treated with monoclonal antibodies, Vectibix and Erbitux, the tech assessment team found all studies reported that patients with KRAS mutations were less likely to experience a treatment benefit.
"The direction of effect was consistent among studies, and formal significance was achieved in the majority of individual studies that reported information on the clinically relevant outcomes of overall and disease-free survival," the review team reported.
Furthermore, the study findings corroborated treatment guidelines from the American Society of Clinical Oncology, the European Medicines Agency, and the US Food and Drug Administration.
Last year, the FDA updated the labels for Vectibix and Erbitux to inform doctors and patients that those with KRAS mutations in codon 12 or 13 do not respond to EGFR inhibitors. The labeling update marked the first time the agency used retrospective clinical trial data to update the label of a drug with PGx information [see PGx Reporter 07-22-2009].
During MEDCAC's discussion of the findings, FDA's Mansfield noted that since KRAS testing is already so firmly entrenched in clinical practice, it may not be possible, and may perhaps be unethical, to run a prospective trial in this setting.
Although there were RCTs among the studies reviewed, the tech assessment team noted that most studies were conducted in the second-line setting and recommended further studies in the first line setting.
Vectibix sponsor Amgen recently presented results from a Phase III study investigating the safety and efficacy of Vectibix in combination with chemotherapy in patients with non-mutated KRAS tumors in the first-line setting. In this study, researchers concluded that Vectibix "significantly improves PFS and is well tolerated when added to FOLFOX4 for first-line treatment of patients with wild-type KRAS in mCRC" and that "PFS was inferior in patients with mutated tumors who received" the drug [see PGx Reporter 01-27-2010]. An Amgen representative informed MEDCAC of this study at the meeting last week.
Based on the totality of data presented, KRAS testing garnered strong confidence (an average score of 4.36 from voting members) from MEDCAC that such pharmacogenomic testing improved outcomes in colorectal cancer patients treated with Vectibix and Erbitux.
Trikalinos' review team identified 31 eligible studies for BCR-ABL testing and concluded that "the presence of any BCR-ABL1 mutation does not appear to predict differential response to treatment in CML patients treated with imatinib-, dasatinib-, or nilotinib-based regimens."
The review team did identify "consistent evidence" from studies that relatively rare mutations, so called T315I mutations, predicted which patients would not respond to treatment with certain tyrosine kinase inhibitors.
"In contrast, there is no evidence that that presence of any BCR-ABL1 mutation can differentiate response to TKI therapies," the review team noted. "The latter result is emblematic of the complexity of this topic: different mutations may confer different resistance to each of the three drugs." Furthermore, reviewers found "sparse" data on overall or progression-free survival in this setting.
Trikalinos suggested that due to the complexities associated with BCR-ABL testing, outcomes should be studied through collaborative registries of CML patients, instead of looking for such data through systemic reviews of published aggregate data.
After the technology assessment, MEDCAC departed somewhat from the agenda, by considering separately two types of BCR-ABL testing: testing to diagnose and monitor CML and testing to identify point mutations that affect CML treatment response.
MEDCAC expressed high confidence (4.27 average score from voting members) that BCR-ABL testing to diagnose and monitor CML improved outcomes. However, experts imparted low confidence (2.18 average score from voting members) for BCR-ABL testing using point mutations, calling such technology premature.
UGT1A1 and HER2 Testing
For the two pharmacogenomic tests not reviewed by Trikalinos' team, MEDCAC imparted strong confidence (with a 4.55 average score from voting members) that HER2 testing improved outcomes in breast cancer patients receiving tamoxifen, but expressed low confidence (with a 2.00 average score from voting members) that UGT1A1 testing to dose irinotecan in colorectal cancer patients was clinically useful in avoiding adverse reactions.
With regard to UGT1A1 testing, MEDCAC referred to a review conducted by the Evaluation of Genomic Applications in Practice and Prevention Working Group. EGAPP conducted an review of available outcomes data on this type of testing, and reported last year that current evidence was "insufficient" to recommend routine UGT1A1 genotyping in irinotecan-treated colorectal cancer patients in order to modify the dose and avoid adverse reactions, such as severe neutropenia.
In HER2 testing, MEDCAC viewed positively the availability of FDA-approved immunohistochemistry and fluorescence in situ hybridization tests. Panelists discussed concerns about the ability of these tests to determine treatment benefit for patients deemed to be in the "intermediate" or "negative" category. However, overall, MEDCAC decided to issue high confidence that HER2 testing improves outcomes in tamoxifen-treated breast cancer patients.
For the three tests MEDCAC felt improved outcomes — HER2 in breast cancer, KRAS testing in colorectal cancer, and BCR-ABL testing for diagnosis and monitoring of CML — the panel expressed strong confidence (4.18 average score from voting members) that the findings were generalizable to the Medicare population and in a community-based setting.