WIXLIB

12more markers called biomarkers that can affect13prognosis in a beneficiary with a personal14history of cancer.15For today's purposes, molecular16pathology testing to estimate prognosis in17cancer provides information about the18likelihood of recurrence after treatment, death19from cancer, or death due to another cause20after the diagnosis of a cancer.21What factors affect prognosis?22Several factors can affect prognosis and can be23used to inform physicians' decisions and to24advise patients about the likely courses of25their diseases. For example, cancer-related141factors include the type, stage and location of2the cancer, treatment-related factors include3the body's response to prior therapies, and4patient-related factors could include the5patient's overall health, comorbidities, as6well as the patient's personal values and7wishes regarding cure versus palliative care.8Prognostics reviewed. Tests for9genomic factors of cancer cells are being10actively investigated as estimators offile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

11prognosis. Some are designed to detect12recurrence at an early point of time. Others13offer views of cancer aggressiveness and may14alter the physician's management. The15molecular tests reviewed today estimate16prognosis in cancer types frequently17encountered in adult patients, colorectal,18breast and non-small cell lung cancers.19Genomic-based prognostic assessments of other20cancer types are currently under investigation.21The outcomes of interest for CMS.22Well, when CMS asks what effect do the results23of these tests have in patient outcomes, we are24especially interested in the outcomes shown25here, overall survival, mortality, avoidance of151harm of anticancer treatments, quality of life2and others.3So what is reasonable and necessary?4Well, CMS assesses the evidence to determine5whether it is adequate to conclude that the6item or device, or specifically in this case7the molecular test for cancer prognosis, leads8to improvement of clinically meaningful9outcomes in our beneficiaries. Today we willfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

10be asking the MEDCAC panel to listen to the11presentations of the evidence and advise CMS on12how best to proceed.13So the MEDCAC questions, I will read14for you to keep under consideration during the15subsequent presentations: For the panel's16reference, a crosswalk is supplied to help17relate the MEDCAC voting questions and the TA18key questions which will be discussed after my19presentation.20Here is a voting question grid. The21MEDCAC panelists will vote on each prognostic22test per question. This is a sample of the23voting grid to identify your confidence level24or score for each question regarding each25prognostic test. If the mean average panel161score for a certain prognostic test is above22.5, then we will continue the voting questions3for that particular prognostic test. If the4mean score is below 2.5 then we will stop the5voting for that particular test. The TA to6MEDCAC crosswalk may be used as a reference7during your voting.8Other acronyms used. Well, we all arefile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

9aware of FDA, CDC, and there are quite a few10other acronyms which are going to be used11throughout the meeting. For your reference12we've also supplied a working glossary of13several of these other terms that will also be14used during the meeting. For example, test15validity measures, today's voting questions16will focus on analytic validity, which is a17test's ability to measure the genetic trait of18interest; clinical validity, the test's ability19to identify or predict the disease or condition20of interest; and clinical utility, which is the21balance of benefits and harms when the test is22used to influence patient management. This23information is also available in more detail in24the reference glossary that was provided.25On to the questions. MEDCAC question1711(a). For each prognostic test listed, how2confident are you that existing evidence is3sufficient to confirm the analytical validity4of the molecular pathology test to estimate5prognosis for Medicare beneficiaries with that6cancer type?7Again, if the answer for question 1(a)file:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

8is at least in the intermediate range, a mean9score of 2.5 or more, then we will move on to10question 1(b). If not, we will save that11particular test for the discussion question12number 4.13Question 1(b). For each prognostic14test listed, how confident are you that15existing evidence is sufficient to confirm the16clinical validity of the molecular pathology17test to estimate prognosis in Medicare18beneficiaries with that cancer type?19As mentioned before, if the mean score20is 2.5 or above, we will move on to question 221for that particular prognostic test.22Number 2. How confident are you that23there is sufficient evidence to conclude that24using the molecular pathology test to estimate25prognosis affects health outcomes, including181benefits or harms, for Medicare beneficiaries2with cancer whose anticancer treatment strategy3is guided by the test's results? If the range4is above 2.5, we move on to question 3. If5not, we skip to question 4.6Voting question 3. How confident arefile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

7you that there is sufficient evidence to8conclude that using the molecular pathology9test to estimate prognosis has clinical10utility, meaning that it improves health11outcomes either due to increased benefits12and/or reduced harms for Medicare beneficiaries13with cancer whose anticancer treatment strategy14is guided by this test's result? Again, you15vote, and if the mean score is 2.5 and above,16that's great, because we're going to move on to17question number 4, the discussion question.18The discussion question: Please19discuss whether the following factors change20generalizability of evidence about molecular21diagnostic tests estimating cancer prognosis.22(A), regulatory status of the test, i.e., FDA,23or approved/cleared by the FDA versus LDT; (b),24performing laboratory type, i.e., academic25medical center laboratories, independent191commercial laboratories or other; (c),2demographic subgroups within the Medicare3beneficiary population; and (d), cancer genomic4characteristics.5This is the conclusion of myfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

6presentation. I will invite Sreelatha Meleth7to give the TA presentation. Thank you.8DR. MELETH: Good morning. My name is9Sreelatha Meleth, I am representing the RTI-UNC10EPC, evidence-based practice center, and we11conducted the technology assessment for the, to12look at the tests for the estimation of13prognosis of common cancers.14A lot of people helped us, this is a15fairly extensive review, and I just wanted to16start off by acknowledging that, also17acknowledging that the funding, was based18on the AHRQ contract, and none of the19investigators had any conflicts of interest.20For the actual technology assessment,21I understand that we've changed the focus of22the cancers a little bit, but the actual TA23looked at four cancers and we looked at 1124tests, and the objective was to conduct this25systematic review assessing the 11 molecular201pathology tests that might inform estimation2for prognosis. The overarching question was3whether there is direct evidence that the4addition of the results of the tests would givefile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

5you more information about the prognosis in6addition to the traditional prognostic markers7that are traditionally used, and would improve8clinical outcomes for adult patients.9So, the table there shows you the10tests that we looked at. For breast we looked11at MammaPrint and Oncotype Dx. For colorectal12cancer we looked at BRAF, KRAS, MSI, MLH1 and13Oncotype Dx. Then for the objective lung we14looked at EGFR, KRAS and ALK translocation, and15for bladder we also looked at UroVysion.16Just a brief background. It is17estimated that there will be approximately 1.6718million new cases of cancer in 2014. It's the19second leading cause of death in the United20States. However, the death rates in general21for cancers are declining. In the last five22years death rates decreased 1.8 percent a year23for men and 1.4 percent a year for wome n. The24declines are largely due to the declines in25death rates in the four major cancers. In the211last five years lung cancer death rates are2down 34 percent for men, nine percent for3women; 34 percent, 45 percent and 46 percentfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

4respectively for breast, prostate and5colorectal cancer.6And part of this improvement in7survival can be attributed to advances in8molecular pathology, which has resulted in9better understanding of cancer subtypes and10development of treatments based on the11subtypes. For example, the identification of12the human epidermal growth factor 2 receptor13resulted in targeted therapies for breast14cancer. Advances in molecular pathology have15also helped identify tumor characteristics that16help predict prognosis for a patient in17addition to traditional markers such as stage18and differentiation.19We would like to, before we go into20the matter and results, clarify that the two,21questions 2 and 3 in MEDCAC, the questions that22were just presented, ask about the anticancer23treatment therapy being guided by these tests.24It's important to know that these genetic tests25were used in two different contexts. In one221the tests are used in a specific context of a2test and therapy combination where the test isfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

3being used to predict response to that4particular therapy, so it's a way of looking at5the test with respect to a very specific6treatment.7In the second context these tests are8used to estimate a patient's prognosis and9physicians use this information in a variety of10ways, including informing choices from a11variety of different treatment options.12CMS requested this report to evaluate13the second context, not the first. Therefore,14studies that evaluate specific test/therapy15combinations were excluded from this review.16So, we'll just briefly go through the17methods. We started off with key questions18that were then refined through discussions19within, between the EPC, AHRQ and CMS,20developed an analytic framework, we searched21the databases, we looked, did a systematic22review of published evidence, summarized23evidence qualitatively and quantitatively, with24a meta-analysis when that was appropriate.25The methods used in the review were231based on the framework developed at the CDC,file:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

2the Evaluation of Genomic Applications in3Practice and Prevention, EGAPP, by that working4group which was established in 2005 to develop5a systematic process for evidence-based6assessment that is specifically directed7towards, focused on genetic tests and other8applications of genomic technology.9The methods developed and recommended10by this working group share many elements with11other existing processes, such as the USPSTF12and the AHRQ Evidence-Based Practice Center.13It also recognizes that the gold standard for14direct evidence, randomized clinical trials,15may not be available in the evidence base for16these new tests, and outlines the process for17building a chain of evidence.18And that chain of evidence is also19called the ACCE model, or the A-C-C-E, which20stands for the analytical validity, clinical21validity, clinical utility and, where22applicable, ethical, legal and social23implications, so those are the sort of pillars24that we use to build the evidence.25Analytic validity is the technical24file:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

1performance of the test, so does the test2actually measure what it's supposed to. The3common ways in which we assess that are4sensitivity and specificity.5Clinical validity is looking at the6strength of the association between a genotype7and a result of interest. The strength of this8association determines the test's validity to9diagnose a disorder, assess susceptibility or10risk, or provide information on prognosis or11variation in drug response.12Clinical utility looks at the evidence13that the test results can actually change a14patient's management decisions and improve net15health outcomes down the road.16So, how did we apply the ACCE model?17So based on the working group's recommendation18we first developed an overarching question that19sought to find direct evidence addressing our20primary question. And then we created a set of21sub-questions based on the ACCE model that22would help build the chain of evidence that23could help answer the overarching question24indirectly.25So what was the overarching question?25file:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

1The overarching question was, is there direct2evidence that the addition of the specified3molecular pathology tests used alone or in4combination with traditional prognostic factors5changes physician decision-making and improves6outcomes for adult patients with colorectal7cancer, breast, lung or bladder cancer,8compared with the use of traditional factors to9predict risk of recurrence for adults with10these cancers? Ideally we hoped to find11published evidence that directly answered this12key question.13In the absence of direct evidence, in14the event that there would be an absence of15direct evidence for the key question one, we16developed key questions to build a chain of17evidence that would help answer KQ1, and the18chain again was based on the ACCE model. So19the analytic validity question was, does the20existing evidence establish technical accuracy21and reliability of these tests for detecting22the relevant molecule analytes? The clinical23validity question was, does existing evidence24establish the prognostic accuracy of the test25for predicting recurrence? And clinicalfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

261utility was, does existing evidence support2clinical utility of the molecular pathology3tests?4Clinical utility was further refined5into the impact on physician decision-making6and patient centered harm.78I have the red light flashing, is thatactually right? Okay.9What is the evidence that prognostic10information, the first question for clinical11utility was, is there evidence that the12prognostic information provided by these tests13modifies physician decisions regarding use of14adjuvant antineoplastic chemo and/or15radiotherapy, enhanced diagnostic testing for16recurrence, and/or surgery among adult patients17with malignant tumors?18And the KQ4b was, what is the evidence19that modified decisions leads to improved20outcomes, including patient-centered outcomes21such as improved quality of life, reduced22disease recurrence, increased overall survival23or disease-free survival, or reduced24therapeutic side effects?25And KQ5 looked at the harms associatedfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

271with treatment decisions informed by the2molecular pathology tests.3So based on that, we created an4analytic framework where, we are looking at5those, so that's how we expected to come to6our conclusion, to assess our evidence and go7throught the results. So we have the cancer,8we have the tests, and then you're looking at9whether the test is actually doing a good, you10know, detecting what they wanted to detect, and11then we're looking at whether it actually predicted12risk of recurrence of prognosis, and then looked13at treatment decisions, and then looked at health14outcomes for the patients, and also whether patients15were harmed.16We searched PubMed, the Cochrane17Library and EMBASE for English-language studies18published through November 2013. We also19searched the reference list of pertinent review20articles and studies that met our inclusion21criteria. We searched as well for unpublished22studies relevant to this review, for which we23used relevant websites, clinicaltrials.gov, the24FDA website, Health Services Research Projectsfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

25in Progress, and the European Union Clinical281Trials Register. We also requested information2from the College of American Pathologists and3from relevant companies, asking for data that4they believe should be considered for the5review.6A very brief look at the population7and intervention, so the population includes8adult patients with one of the cancer types of9interest that evaluated an eligible test. The10comparators for KQ1, 4 and 5 included studies11that compare at least one of the tests plus12standard prognostic factors with the standard13prognostic factors alone to determine whether14the test adds independent prognostic value or15benefit, or introduces additional harms.16We did not include studies focused on17patients with advanced or metastatic cancer or18studies focused on predicting response to19treatments. And just in case there's a20question, because the task of the technology21assessment was focused on risk of recurrence,22so it didn't seem relevant to look at advanced23metastatic cancers, and we already talked aboutfile:///F /MEDCAC/pg032415.txt[04/27/2015 1:35:39 PM]

2425response to treatments.The comparators for KQ2, which is the291analytic validity question, we included studies2of test performance, including intra/inter -lab3reproducibility for the included tests. For4KQ3, which is clinical validity, we included5studies comparing patients with different test6results, example, those with a mutation versus7those who are wild-type, to establish8prognostic value, with a multivariate analysis9to adjust for known factors. We required that10the results were either adjusted for known11factor using a model, or were specifically12addressed in other ways which could be either13the inclusion/exclusion criteria for the study14or something like stratification.15So, all studies, for everything we had16independent dual review to assess for17eli

21 been updated to reflect the state-of-the-art 22 activities related to this topic. 23 And now I will turn things over to 24 Dr. Henderson. 25 DR. HENDERSON: So, I'm Craig 9. 1 : Henderson, from University of California; 2 : . from the National Institute on Aging at: 14 . NIH, representing myself, and I have no : file:///F /MEDCAC/pg032415 .