Payer/HTA Perspective

Transcription

1 Can there be a common methodology for comparative effectiveness in the absence of randomised controlled trials (RCTs)? Payer/HTA Perspective Rod Taylor Professor of Health Services Research 1

2 Declarations of Interest Contribute PenTAG programme of STAs and MTAs for NICE, commissioned by UK NIHR Programme Former member of National Institute of Health & Clinical Excellence (NICE) Technology Appraisal committee & current member Interventional Procedure Advisory Committee Consultant for healthcare industry Payer s/hta Preference Randomised controlled trial evidence... collected in the appropriate population conducted in the appropriate setting the appropriate comparator using the appropriate patient-relevant outcome(s) 2

3 NICE TA Reference Case v.2012 Design Population Head-to-head RCTs provide the most valid evidence of relative treatment effect The population for whom the technology is being appraised... highlight potential subgroups Comparator All relevant comparators Outcomes Clinical endpoints that reflect how a patient feels, functions, or how long a patient survives are regarded as more informative than surrogate endpoints Presentation Whilst head-to-head RCTs with final outcomes regarded as most valid source of comparative effectiveness evidence, such evidence may not always be available... Alternative evidence how acceptable? Use of non RCT data Use of surrogate outcomes 3

4 JAMA. 2001;286: Can J Surg, Vol. 55, No. 3, June

5 2009;339:b4538 Use of non-rct evidence Inferences about relative treatment effects drawn from non-rct evidence will necessarily be more circumspect than those from RCTs with properly controlled evidence [NICE, 3.3.6] The use of non-randomized studies as proof of the causality of an intervention...requires particular justification or specific preconditions and special demands on quality [IQWiG, v.4, 2011] 5

6 PBAC Guideline for Preparing Submissions. v Dec 2008 Frequency of Surrogates in HTA Elston & Taylor (2008) [IJTAHC:2008;25:6-13] 4/99 (4%) UK HTA reports has economic modelled based on surrogate outcome Garrido & Mangiapane (2009) [IJTAHC:2009:25:314] 6/140 (4%) international used solely surrogates PBAC [Andrew Mitchell personal communication] 37% (53/143) economic submissions over the past 5 years to PBAC inferred the transformation of a surrogate outcome 6

7 Surrogate Outcome Bias Ciani et al (2012) BMJ in press Intervention more effective Control more effective Final outcome More effective Surrogate outcome more effective *Adjusted f or interv ention, patient population, journal, sponsor, sample size and f ollow up Surrogate Validation Johnson et al, J Clin Epidemiol,

8 Validation Quantification Framework for use of surrogate outcomes in HTA Health Technology e.g. statins Cost-effectiveness model Disease e.g. CHD Final outcome e.g. mortality or HRQoL Surrogate outcome measure e.g. LDLchol Costeffectiveness ratio e.g. Cost per QALY Source e.g. Meta-analysis of RCTs from Elston and Taylor, Use of Surrogate Outcomes When the use of final clinical end points is not possible and surrogate data on other outcomes is used to infer the effect of treatment on mortality and health-related quality of life, validation of the surrogate-to-final endpoint outcome relationship must be provided, together with details of the information on which the relationship is based and an explanation of how this relationship is quantified for use in modelling. [NICE 5.7.5] 8

9 Conclusions With the timing of HTA often close to product launch, growing recognition by payers that their preferred evidence i.e. direct RCTs vs appropriate comparator(s) final clinical endpoints, may not be always available Whilst there remains a fundamental reluctance to (entirely) depend on non-rct evidence, other approaches to evidence are actively being considered Use of validated surrogates in RCTs Use of indirect comparison of RCTs Analyst Perspective Beth Woods, Oxford Outcomes 9

10 Declaration of interest I have worked within the Oxford Outcomes Health Economics team for six years. During this time I have conducted numerous projects on behalf of pharmaceutical, device and diagnostic manufacturers. The analyst Data available at launch Cost-effectiveness analysis always requires estimation of comparative effectiveness A base case is presented, often in the face of concerns regarding bias and uncertainty 10

11 Where do we go in the absence of RCTs? Network meta-analysis (NMA) Non-randomised evidence Surrogates Final endpoints Nonrandomised NMA What do we do when 1. We don t have a connected network 2. A new technology receives regulatory approval based on a single arm trial 3. The patient group of interest are a subpopulation of the patients included in the current network 11

12 1. Disconnected network Example rituximab as first line treatment for follicular lymphoma (review of TA110 1 ) Segment 1 Segment 2 Segment 3 1. Technology Assessment Report, Review of TA110, Disconnected network - options Search for more distant connections May not exist, increased uncertainty Use an endpoint where a connected network exists, then use as surrogate for endpoint of interest Estimate the missing parameter: Assumptions / clinical opinion Individual trial arms Observational data Observational data as an alternative to NMA 12

13 Similar problem 2. Single arm trial Missing link is with technology of interest Not yet in use Much less likely to be able to make comparisons using observational data Single arm trial - options Adjust for differences between single arm trial and other data (observational/trial) Confounding due to study-level differences Sample size available may challenge use of such methods Use of systematic review, meta-analysis and meta-regression Reliance on relationship between study-level differences in prognostic factors and study level outcomes 13

14 3. Subgroup of wider network Focused pivotal trial or reimbursement target Subgroup data may not be available for studies in the network Subgroup of wider network - options Use ITT data for comparator trials Analysis of subgroup data Proportion of subgroup as covariate in metaregression IPD for all trials 14

15 Conclusions Difficult cases like these are common and not covered adequately by existing guidance Analysts face a large number of (often highly imperfect) options Barriers to a common methodology Lack of research on alternative methods Detailed guidance would be required How to make the choice not the choice Would this ever be enough? Or would cross-hta body scoping processes also be required? Alternative designs for evidence generation postlaunch Pr Isabelle Durand-Zaleski APHP URCeco Paris 15

16 Declarations of Interest Scientific advisor and co-investigator for: Medtronic Janssen MSD GSK Hexacath Behring Abbott Air liquide Sanofi Aventis French response to insufficient evidence pre launch: «oui, mais» Conditional reimbursement Evidence generation post launch Re assessment with economic evaluation 3-4 years post launch 16

17 The constraints No individual randomisation Comparators that reflect current practice Relevant endpoints Some responses and a case study Controlled non randomised studies Stepped wedge 1. Ref: Cochrane review, EPOC group 2. Bonell CP, Hargreaves J, Cousens S, et al. Alternatives to randomisation in the evaluation of public health interventions: design challenges and solutions. J Epidemiol Community Health (2009) doi: /jech Brow n C A and Lilford RJ The stepped w edge trial design: a systematic review BMC Medical Research Methodology 2006, 6:54 17

18 Controlled non randomised Before-after with a control group Interrupted time series Not easy for drugs or devices Case control with propensity score matching Very widely used Data from registries or claims database Several methods Issue with the completeness of data for risk adjustment Bäumler M, Stargardt T, Schrey ögg J, Busse R. Cost ef f ectiveness of drug-eluting stents in acute my ocardial inf arction patients in Germany : results f rom administrativ e data using a propensity score-matching approach. Appl Health Econ Health Policy Jul 1;10(4): Le Ray I, Dabakuy o S, Crehange G, Bardou M, Arnould L, Fraisse J, Fumoleau P, Coudert B, Causeret S, Arv eux P, Maingon P, Bonnetain F. Neoadjuv ant therapy f or breast cancer has no benef its on ov erall surv iv al or on the mastectomy rate in routine clinical practice. A populationbased study with a median f ollow-up of 11y ears using propensity score matching Eur J Cancer Oct;48(15):

19 Single arm study: the case of aortic graft Coverage with requirement for evidence of effectiveness No control arm Sang Thrombose Vaisseaux. Volume 20, Numéro 2, , février 2008, Mini-revue 19

20 Participants /clusters Aortic graft, 2 Controls from claims database Problem of ICD1O codes vs. prospective CRF data collection for risk adjustment Record linkage for 2-year survival Identification of coding biases Alternative to individual randomisation: Stepped wedge Shaded cells represent intervention periods Blank cells represent control periods Each cell represents a data collection point Time periods 20

21 Stepped wedge design Political Economical in the deployment phase Complex statistical analysis Complex data collection Conclusion on a common methodology Difficult because of differences in: reference strategies Endpoints of interest Data availability Agreement possible on the risks and benefits of each method 21

22 Questions and discussion Can we reach a common methodology, within and across jurisdictions? Is this something that EUnetHTA can consider? 22