AMedicare advisory panel in May expressed little support for two types of genetic testing—those to determine cancer of unknown primary site (CUP) with metastatic tumors and those to identify patients at higher risk of cervical cancer. The vote of low confidence, experts say, has to do more with the lack of evidence demonstrating impact on clinical outcomes than the validity of the tests. Test developers throughout the diagnostics industry need to work to plug this evidence gap in order gain recognition by professional societies in clinical practice guidelines and by payers with reimbursement and coverage. “This [process] basically establishes what evidence exists behind a test and shows there is a big hole in clinical utility. Does having information from this test make patient outcomes better? No one made that leap,” says Diane Allingham-Hawkins, Ph.D., senior director, genetic test evaluation program at the consulting firm Hayes (Lansdale, Pa.). “Evidence is very much at the front and center of the discussion right now. There is awareness that this is the barometer tests will be judged by and companies must realize that. . . . Given the fact that organizations like CMS now know what they are being asked to pay for, there will be an increasing use of evidence.” As part of the review process, the Centers for Medicare and Medicaid Services (CMS) commissioned technology assessments on the two types of genetic tests. The 12-member Medicare Evidence Development and Coverage Advisory Committee voted with an average score of 2.08 (using a 1 to 5 scale of low to high confidence) for the clinical utility of DNA- or RNA-based testing to predict the origin of CUP. The panel had higher confidence (average 3.25 score) that the test results were reliable (clinical validity). Commercially available tests in this category include CancerTypeID (bioTheranostics), miRview (Rosetta Genomics), and PathworkDx (Pathwork Dx). The panel had even less confidence in the tests aimed at identifying high-risk patients for cervical cancer based on uncertain Pap smear findings. The panel voted 1.67 for the test’s reliability to detect cancer. “Manufacturers need to look critically at trial results. They say it shows clinical utility, but really it shows clinical validity. It detects what it is supposed to, but it is not showing a difference in patient outcomes. They need to go one step further,” Allingham-Hawkins tells DTTR
. “There is the ‘it should make a difference’ argument, but you cannot infer it will make a difference in actual practice. This is especially important for very, very expensive tests. A $100 test is not going to receive the same scrutiny as a $5,000 test. With that expense, you can understand why CMS wants to see evidence.” Acknowledgement of this evidence void is growing more widespread, and in May the Center for Medical Technology Policy (CMTP; Baltimore) released an effectiveness guidance document (EGD) aiming to close the gap between the presumed benefits of tests undergoing technology assessments and the actual evidence needs of payers, clinicians, and professional societies. “Although the calls for better evidence are frequent, there is considerable debate about how much evidence is needed and how it can be generated in an efficient and timely way,” wrote Patricia Deverka, M.D., senior research director at CMTP and lead author of the EGD. “The purpose of this initiative was to close the gap between the presumed benefits of tests undergoing technology assessments and the information needs of payers, clinicians, and patients. Our overarching goal is to bring greater clarity and predictability regarding the evidence requirements of all interested stakeholders.” With the increased reimbursement transparency that came with the initiation of 114 new molecular pathology Current Procedural Terminology codes this year and the threat of increased regulation of laboratory-developed tests looming, the need to build consensus on evidence requirements is great as future coverage and reimbursement of molecular tests is at stake for test developers. Sidebar: Clinical Validity, Utility Recommendations
The Center for Medical Technology Policy has issued a set of recommendations covering how to best assess the clinical validity and utility of molecular cancer diagnostic tests. The multiyear project involving a broad spectrum of key stakeholders sets out specific evidence goals and study design recommendations to improve the quality of these studies to ensure their relevance to reimbursement, policy, and clinical decisions. The recommendations include these:
- The potential therapeutic actions or decisions based on test results must be specified in advance.
- Both beneficial and harmful outcomes of testing must be measured, including validated patient-reported outcomes and robust end points such as survival and downstream health care utilization. Insufficient end points for demonstrating clinical utility include changes in physician behavior and intended care plans, which do not necessarily link clinical management decisions with outcomes.
- Clinical utility of a molecular biomarker should be assessed with randomized controlled trials. Randomizing patients to genomics-guided treatment versus usual care reduces statistical power and is not optimal because it requires larger sample sizes to demonstrate an effect of the test.
- Prospective-retrospective study is adequate to generate evidence of clinical utility if a clinical trial with banked biospecimens exists.
- A molecular diagnostics test’s clinical utility can be established with single-arms studies if it is Food and Drug Administration-approved based on pivotal trials, adequate archived biospecimens are not available, complete or overall response is a feasible end point, and comparable response data in a noncontemporaneous comparative cohort exists.
- Longitudinal observational study designs (prospective cohort studies, patient registries that explicitly include comparators, and multiple group, pretest/posttest designs) can be acceptable, but not retrospective observational studies.
- Decision-analytic models are useful if there is no direct evidence of clinical utility, if clinical validity is established.