“is a medical model emphasising the systematic use of information about an individual patient to select or optimise that patient's preventative and therapeutic care”.
Dept Health & Human Services. Realizing the Potential of Pharmacogenomics:Opportunities and Challenges. Report of the Secretary’s Advisory Committee on Genetics, Health, and Society. May 2008
Holmes MV, Shah T, Vickery C et al. Fulfilling the promise of personalized medicine? Systematic review and field synopsis of pharmacogenetic studies. PLoS ONE 4(12): e7960. doi:10.1371/journal.pone.0007960A primary pharmacogenetic study was defined as one in which the title of the study or the stated aims or purpose within the text of the abstract indicated that the primary intention of the study was to investigate the effect of genetic variation on drug response.Systematic reviews and field synopses previously exposed the obstacles to progress in complex disease genetics. These included: a focus on candidate genes rather than genome-wide analysis; inadequate sample size; suboptimal capture of genetic variation; and significance chasing and reporting bias; all of which led to a failure to replicate and validate genetic associations [14,15,16].These overviews [17,18,19] were followed by improvements in research design which made an important contribution to the recent success in the identification in genes for common disease . These considerations and the absence of a prior systematic, quantitative overview of pharmacogenetic research was the motivation for the current studyCheck limitations in paper
Co-dependent in this instance is drug/genetic testBackground: Since the mapping of the human genome in 2003, the development of biomarker targeted therapy and clinical adoption of ‘personalised medicine’ has accelerated. There is, however, increasing international recognition that current reimbursement models may not adequately assess the safety, effectiveness and cost-effectiveness of these medicines due to the nature of their evidence-base.
The information provided in each submission (PBAC) or assessment report (MSAC) for these five simple paired co-dependent technologies was categorised and tabulated, and an initial list of points was compiled. Specific information provided in the submission/report that was found to be unnecessary to inform the public funding decision was also identified. A gap analysis was then conducted to assess what key information was considered to be missing or unavailable during the evaluation of each simple paired co-dependent technology on the basis of (a) matters raised in the evaluation (mentioned in the PBAC Commentary or MSAC assessment report) and (b) matters raised concerning the evaluation (relevant MSAC or PBAC minutes or ESC advice). A few additional evidentiary requirements were included in the list on the basis of previous evaluation experience. The draft structure took into account:the availability of direct evidence, as opposed to the need to produce a linked evidence approach (MSAC, 2005). This was on the basis of the evaluators’ experience with MSAC evaluations whereby the linked evidence approach is the primary evaluation technique when assessing diagnostic tests because the evidence base rarely includes direct evidence; availability of a reference standard against which to determine the accuracy of the test; and a request for listing of one or both of the simple paired co-dependent technologies, including circumstances where one technology is already listed (ie listed for another patient indication, or listed for the same indication but is unpaired) – this was exemplified by the imatinib simple paired co-dependent technology which was included for analysis. Each simple paired co-dependent technology was then formally rated against the combined list of evidentiary requirements in terms of whether they had specifically addressed each requirement. A free text column was provided to allow commenting on whether difficulties were likely to arise as a consequence of addressing a specific evidentiary requirement. The aim was to capture issues that would need to be considered during the evaluation process but that are beyond the scope of this short project. This first version of the combined list of evidentiary requirements was provided to the Project Secretariat for comment.
a 67 information items (denominator) were collated from submissions at the completion of Stage 1. Evidence gaps (numerator) were defined as a complete absence of information in the submission; however, please note that frequently the information items were only partially/inadequately addressed and in some instances items were not applicable; b decision at the time the framework was being developed; c systemic mast cell disease, hypereosinophilic syndrome and chronic eosinophilic leukaemia; d PDGFR rearrangements and the KIT D816V mutation are mutually exclusive so, as the PDGFR test was funded, there was no need to fund the KIT D816V test.PBAC = Pharmaceutical Benefits Advisory Committee; MSAC = Medical Services Advisory Committee
multiple opportunities to explain the association between biomarker and drug treatment outcomes, even in the absence of generally accepted experimental evidence. The Bradford Hill criteria, namely strength, specificity and temporality of the association between the biomarker and drug treatment on health outcomes; consistency and coherence of effect; biological plausability and gradient (eg dose-response); producing the effect upon experimentation or by analogy, were addressed
1. Tests are usually not expensive when compared to the drug component of a co-dependent, so lack of cost-effectiveness unlikely to be due to test cost. However, if a highly prevalent mutation then if test has moderate cost then it may impact on cost-effectiveness. May be other prognostic factors, other than identified biomarker driving treatment outcomes. May be that test is inaccurate and people are getting misclassified and receiving comparator treatment inappropriately.
Methodological framework to support the decision framework.As Holmes has indicated pharmacogenetics evidence is often very poor – developed a linkage approach to get all of the information that the decision-maker needs
Often this sort of assessment will be impossible because the evidence is just not there…….It is beginning to emergePrognostic impact – all else being equal ie no other unequal distribution of potential confounders. Cost-effective because patient group are already responders. Would occur regardless of whether it is a new drug or existing therapy.
Envisioned trial design that would adequately capture all/most information needs of policy maker as suggested by Lord et al........ The authors advocate using the hypothetical RCT as a conceptual framework to identify what types of comparative evidence are needed for test evaluation. Evaluation begins by stating the major claims ....and determining whether it will ....achieve these claims.Using the Principles of Randomized Controlled Trial Design to Guide Test Evaluation Sarah J. Lord, Les Irwig and Patrick M. M. BossuytMed Decis Making 2009; 29; E1 originally published online Sep 22, 2009; DOI: 10.1177/0272989X09340584This design allows analysis of whether the relationship between biomarker and drug is treatment effect modification or prognostic impact or both; as well as the incremental benefit of the test. However, would rarely (if ever) see a double randomised controlled trial. Ethical issues / expenseIt is likely the case that industry submissions will be a cobbled linkage of evidence between the different “levels” (or components) of direct evidence of test and/or drug effectiveness. A hierarchy has been suggested because we want industry to provide the best quality evidence available (and reduce policy-maker uncertainty), however we expect to see all sorts of linkages operating, depending on what evidence is available.Transferability of linkages.... Determining (and conveying clearly to the sponsor) the exact combination of different evidence/designs to answer the policy question, with the least amount of uncertainty, will be the challenge when writing the Guideline. In biomarker stratified design – at drug randomisation point, (assuming a reasonable sample size) all variables other than biomarker status should be fairly evenly distributed in Drug A and Drug B groups. Limit effects of confounders on drug treatment effectiveness. Best we can do. This allows us to see both TEM and prognostic impact or combination of the two but not the incremental benefit of the test – that is there may be uncertainty as to whether the biomarker +ve/-ve is responsible for the effect or some other unmeasured variable – ‘cos cannot randomise to biomarker status. I would envisage that RCT evidence addressing treatment effectiveness would be used to fill this gap. This additional information would help determine if similar health outcomes are produced (eg survival, adverse events) through using the new drug as a replacement for the old drug on the basis of clinical judgement alone compared to targeting the use of the new drug vs old according to biomarker. If there is no difference in health outcomes then it is not clinically necessary (and an unnecessary expense) to test for the biomarker. This might occur when the biomarker suggests a prognostic effect but that information does not end up altering the management of the patient (and thus their health outcomes).
I think something similar may occur with the ‘enrichment’ design - if provided, it would also require supplementary evidence of test accuracy to be provided, as discussed in the linked evidence section of the draft. Discordant results trial design – Lord……Biomarker-stratified design is inefficient if the biomarker is not common in the diseased population being tested. Its usefulness would depend on the prevalence of the biomarker – some biomarkers are quite common (eg KRAS), some quite rare (eg EGFR). The ‘discordant-results’ trial design is a very good way of improving efficiency – and also getting over the hurdle of an imperfect reference standard. Drug can affect mutational status in some instances.Confounding introduced as cannot randomise by biomarker status and so potentially unequal distribution of biomarkers in each group; potentially unequal distribution of other confounders/imbalanceThere is still the residual problem with using archival tissue that the biomarker status will have changed (mutations are unstable and change from primary to metastasis) and/or resistance created as a consequence of using the drug or other drug treatments subsequent to biopsy.
Manuscript submitted to Journal of Medical Decision-Making
Economic evaluation, reimbursement and context. How to assess personalised medicines for reimbursement decisions?
Adelaide Health Technology Assessment (AHTA)Monday, June 25, 2012HOW TO ASSESS PERSONALISED MEDICINESFOR REIMBURSEMENT DECISIONS?Developing a framework for AustraliaTracy Merlin, Claude Farah, Camille Schubert, AndrewMitchell, Janet E Hiller, Philip Ryan Life Impact | The University of Adelaide
Assessing personalised medicine• To adopt pharmacogenetics/genomics into clinical practice, a PGx test should demonstrate analytical validity, clinical validity, and clinical utility (DHHS, 2008). – Analytical validity – how accurately and consistently a test detects the presence of a specific genotype. – Clinical validity – the accuracy with which a test detects or predicts a given phenotype (clinical disorder or outcome). – Clinical utility – the net balance of risks and benefits for use of the test in clinical practice eg impact on patient health outcomes of targeting drug treatment according to presence of biomarker → Needed for reimbursement decision
Problem• Systematic review of pharmacogenetic studies (Holmes et al 2009) » 1987-2007, Medline, k=1668 studies » Review:study ratio (25:1) - high expectation but limited translation of research » Small sample sizes » Primarily a candidate gene approach using common alleles rather than genome-wide analysis » Surrogate rather than clinical outcomes » Likely ‘significance chasing’ ie post hoc subgroup analysis or publication bias
National initiatives No system in place nationally for evaluating co-dependent technologies, specifically personalised medicines, for reimbursement decisions No formal systems in place internationally for evaluating personalised medicines for reimbursement decisions
Project outline. I.Aim• To develop an approach for evaluating a biomarker/test/ drug (‘co-dependent technologies’) package to inform a national reimbursement decision.Methods – Stage 1• Personalised medicine case studies identified from recent reimbursement applications in Australia – information commonalities / gap analysis• Relevant international regulatory and reimbursement guidance documents reviewed
Project outline. II.Methods – Stages 2 & 3• Collated information synthesised into a cohesive structure amalgamating PBAC and MSAC evidentiary needs• Clinical effectiveness information grounded in Bradford-Hill causality theory → explain association between biomarker and drug treatment outcomes – strength, specificity and temporality of association – consistency and coherence of effect – biological plausability and gradient (eg dose response) – producing effect upon experimentation or by analogy.• Trialled on case studies• Circulated to federal policy-makers and technical experts for input + national public consultation.
Results - decision framework1. the drug is cost-effective in untested population, but cost- ineffective when conditioned on the biomarker identified by the test – drug funded – test not funded. Biomarker is not explanatory. Or test is not accurate.2. the drug is cost-effective in untested population but even more cost-effective when conditioned on the biomarker identified by the test – drug funded – test may be funded......uncertainty surrounding association between biomarker and drug treatment effect will drive the decision to reimburse the test
Decision framework3. the drug is not cost-effective in untested population but is cost-effective when conditioned on biomarker identified by the test – public funding of both test and drug depends on uncertainty surrounding the association between biomarker and drug treatment effect4. the drug is not cost-effective in untested and tested population – neither drug or test should be reimbursed.
Methodological framework Adapted PBAC Application structure: Section A [Q1-19] – Context for submission – Rationale for submission – Proposed impact on current clinical practice Section B [Q20] – Clinical benefit • Test safety / effectiveness • Biomarker-drug relationship (treatment effect modification, prognostic impact) Option 1: Direct evidence (hierarchy) [Q21-24] and/or Option 2: Linked evidence support [Q25-39] Section C [Q40-42] – Translate biomarker/test/drug evidence to Australian clinical setting (applicability, extrapolation, transformation) Section D [Q43-71] – Economic model. (includes test/no test arms) • Section E [Q72-79] – Financial impact
Section B - the evidence baseTwo key methodological areas that impact on fundingdecision1. Likelihood that results are correct?Relates to policy-maker’s certainty regarding the validity of the results– effect of bias and confounding2. What is the association between biomarker and drug? • Treatment effect modification • Prognostic impact • Both • UnknownHow to elucidate this relationship from the evidence base?
Simplistic example Outcome Biomarker +ve Biomarker -ve Drug A Drug B Drug A Drug B N=100 N=100 N=100 N=100 1. Treatment effect modification 5 year OS 60% 30% 30% 30% RR=2.0 [1.4, 2.8] RR=1.0 [0.7, 1.5] 2x 2. Prognostic impact 5 year OS 70% 60% 35% 30% RR=1.2 [1.0, 1.4] RR=1.2 [0.8, 1.7] 3. TEM + Prognostic impact 5 year OS 60% 40% 20% 20% RR=1.5 [1.1, 2.0] RR= 1.0 [ 0.6, 1.7]
Hypothetical RCT as conceptual frameworkBiomarker-stratified design: Is there a difference inIs there a difference in Drug A vs B Randomise to Drug A vs B effectivenesseffectiveness when conditioned on test in unselected population?biomarker? Linkage 1 Linkage 2 Test No test Randomise to drug +ve -ve Randomise to drug Drug A Drug B Drug A Drug B Drug A Drug B +e +ve Health Outcomes
Linking evidence Randomise to test Test Test accuracy - No test Linkage 3 Randomise to drug +ve -ve Randomise to drug Drug A Drug B Drug A Drug B Drug A Drug B Post-treatment test or test archival tissue +ve -ve +ve -veEnrichment design:Is there a difference in Drug A vs Beffectiveness when conditioned on Linkage 1 Health Outcomesbiomarker +ve? Linkage 2
Project outline. II.Results• Developed a decision framework for policy makers• Developed a methodological framework to support the decision making – 79 item checklist must be addressed when personalised medicine submitted for funding through Medicare (test) and the Pharmaceutical Benefits Scheme (drug). http://www.health.gov.au/internet/hta/publishing.nsf/Content/co-1 – Forms basis of co-dependent technology submission guidelines (not yet released)
ConclusionMerlin T, Farah C, Schubert C, Mitchell A, Hiller JE, Ryan P. Assessing personalisedmedicines in Australia: A national framework for reviewing co-dependenttechnologies. Journal of Medical Decision Making, 2012. [In press].
• Co-authors - Claude Farah, Camille Schubert, Andrew Mitchell, Janet Hiller, Philip Ryan• Feedback on framework- Sarah Lord, Robyn Ward, Graeme Suthers, Brian Richards, Lloyd Sansom• Part-funding – Australian Government Department of Health & Ageing firstname.lastname@example.org