Boyd, Kathleen Anne (2012) Employing early decision analytic modelling to inform economic evaluation in health care: theory & practice. PhD thesis http://theses.gla.ac.uk/3685/ Copyright and moral rights for this thesis are retained by the author A copy can be downloaded for personal non-commercial research or study, without prior permission or charge This thesis cannot be reproduced or quoted extensively from without first obtaining permission in writing from the Author The content must not be changed in any way or sold commercially in any format or medium without the formal permission of the Author When referring to this work, full bibliographic details including the author, title, awarding institution and date of the thesis must be given. Glasgow Theses Service http://theses.gla.ac.uk/ [email protected] Employing early decision analytic modelling to inform economic evaluation in health care: theory & practice Kathleen Anne Boyd MSc Economic & Health Economics, BA (Honours) Economics Submitted in fulfilment of the requirements for the degree of Doctor of Philosophy Health Economics and Health Technology Assessment Institute of Health & Wellbeing University of Glasgow October 2012 II Abstract Decision analytic modelling (DAM) is a mathematical technique which is used to structure and synthesise evidence in order to inform decision making, given uncertainty. Decision models are an ideal tool for undertaking economic evaluations as they enable a wide range of data on costs and effects to be synthesised within the model in order to derive cost-effectiveness outcomes. The iterative framework for economic appraisal has been proposed as good practice for undertaking economic evaluations (1), and DAM plays a key role within this framework. In particular there is a role for early stage DAM prior to primary research, to provide an indication of the potential cost-effectiveness of a new health technology (2) given current evidence, and the use of value of information (VOI) techniques to help inform further research priority setting. In practice, support and funding for early stage DAM and full exploitation of VOI techniques is rare. The aim of this thesis is to examine the role for early decision analytic modelling in informing research priorities and the design of future studies in a health care setting. This thesis explores the feasibility, merits and drawbacks of undertaking early DAM and considers potential reasons as to why it has not been more widely implemented. This thesis demonstrates the value and importance of early DAM; in both an ‘ideal’ setting and also in a less desirable, time-constrained setting. Applying early DAM and VOI techniques enables researchers to provide relevant conclusions and recommendations to decision makers, who can make informed decisions as to whether a new intervention should be adopted (or rejected), or whether further information is required to help make the decision; as opposed to making decisions based on subjective reasoning. There is considerable merit with employing early DAM for health care research, such as reduced uncertainty, reduction of costs and efficiency gains, however, some drawbacks exist in terms of whether it is always viable to fully exploit VOI analyses, which may hinder widespread support both inside and out-with the health economics community. III Table of Contents Abstract.................................................................................... II Table of Contents ........................................................................III List of Tables..............................................................................VI List of Figures...........................................................................VIII Contributions.............................................................................. X Publications and Presentations........................................................XII Acknowledgements..................................................................... XIII Author’s Declaration...................................................................XIV Abbreviations ............................................................................XV 1 Introduction...........................................................................1 1.1 Rationale.........................................................................1 1.2 Economic evaluation in health care .........................................3 1.2.1 Modelling...................................................................6 1.2.2 Trial based .................................................................7 1.2.3 DAM in economic evaluation using an iterative approach.........10 1.3 Thesis outline .................................................................12 2 DAM for economic evaluation in health care...................................15 2.1 Introduction ...................................................................15 2.2 Economic evaluation in health care .......................................18 2.3 Decision Analytic Modelling.................................................20 2.3.1 Specifying the problem.................................................20 2.3.2 Defining boundaries.....................................................21 2.3.3 Structuring the problem................................................22 2.3.4 Synthesising evidence ..................................................23 2.3.5 Handling Uncertainty...................................................25 2.4 Assessing parameter uncertainty...........................................27 2.4.1 Distributional forms.....................................................28 2.4.2 Fitting parameters to distributions...................................30 2.4.3 Illustrating parameter uncertainty ...................................33 2.5 Decision uncertainty: CEAC .................................................36 2.6 Value of Information.........................................................40 2.6.1 EVPI per decision/patient..............................................41 2.6.2 EVPI population level...................................................42 2.6.3 EVPI for parameters (EVPPI)...........................................45 2.7 EVSI.............................................................................49 2.8 ENBS............................................................................52 2.9 An Iterative Approach to Economic Evaluation...........................54 2.9.1 Stages of the iterative approach......................................56 2.9.2 Benefits of the Iterative Approach ...................................61 2.10 Summary.......................................................................67 3 Early DAM: a case of good practice..............................................68 3.1 Introduction ...................................................................68 3.1.1 The decision problem...................................................68 3.2 Methods........................................................................71 3.2.1 Overview..................................................................71 IV 3.2.2 Existing evidence........................................................72 3.3 Model design and development.............................................75 3.3.1 The diagnostic pathway................................................75 3.3.2 Model 1: Primary rectal & colon cancer .............................82 3.3.3 Model 2: Recurrent colorectal cancer................................96 3.3.4 Model 3: Metastatic cancer...........................................107 3.4 Model Parameterisation and population .................................117 3.4.1 Model 1: Primary rectal and colon...................................117 3.4.2 Model 2: Recurrent disease...........................................131 3.4.3 Model 3: Metastatic disease ..........................................143 3.5 Deterministic Results........................................................154 3.5.1 Primary colorectal cancer results....................................154 3.5.2 Recurrent colorectal cancer results.................................163 3.5.3 Metastatic colorectal cancer results ................................167 3.6 Summary......................................................................169 3.6.1 Primary CRC.............................................................172 3.6.2 Recurrent CRC..........................................................175 3.6.3 Metastatic CRC .........................................................177 3.6.4 Next steps in the analysis.............................................177 4 Setting research priorities.......................................................179 4.1 Introduction ..................................................................179 4.1.1 Identifying the decision problem ....................................179 4.2 Probabilistic results .........................................................182 4.2.1 Primary CRC results....................................................182 4.2.2 Recurrent CRC results .................................................189 4.2.3 Metastatic results ......................................................192 4.3 Decision Uncertainty........................................................194 4.3.1 Primary CRC decision uncertainty ...................................195 4.3.2 Recurrent CRC decision uncertainty.................................200 4.3.3 Metastatic decision uncertainty......................................202 4.4 Value in further research: EVPI & EVPPI .................................204 4.4.1 Primary CRC: Value of Information..................................204 4.4.2 Recurrent CRC..........................................................211 4.4.3 Metastatic...............................................................221 4.5 Discussion.....................................................................226 4.5.1 Primary results..........................................................226 4.5.2 Recurrent results.......................................................228 4.5.3 Metastatic results ......................................................229 4.5.4 Model conclusions......................................................230 4.5.5 Early DAM & VOI........................................................231 5 Designing clinical trials with economic evaluations .........................236 5.1 Case study: fFN testing for pre-term labour.............................238 5.1.1 A pre-trial model for fFN testing.....................................240 5.1.2 Decision Tree ...........................................................241 5.1.3 Pre-trial model predictions...........................................246 5.2 Sample Size calculations....................................................250 5.2.1 Superiority Sample Size ...............................................250 5.2.2 Net Monetary Benefit Sample Size...................................254 5.2.3 Non-inferiority approach..............................................260 V 5.3 Discussion.....................................................................264 6 Designing trials following an iterative approach .............................269 6.1 Re-analysis of the fFN case study.........................................270 6.1.1 Probabilistic analysis of fFN model..................................270 6.1.2 EVSI for the fFN case study...........................................278 6.1.3 EVSI Results .............................................................280 6.2 Practicalities with employing the EVSI approach.......................289 6.2.1 Interpretation of the EVSI process...................................289 6.2.2 Computation............................................................297 6.3 Conclusion ....................................................................301 7 Summary & Conclusions..........................................................303 7.1 Introduction ..................................................................303 7.2 Overview......................................................................304 7.2.1 Chapter 1................................................................304 7.2.2 Chapter 2................................................................304 7.2.3 Chapter 3................................................................305 7.2.4 Chapter 4................................................................306 7.2.5 Chapter 5................................................................308 7.2.6 Chapter 6................................................................308 7.3 Practical limitations for early DAM & VOI................................310 7.4 Areas for further research..................................................312 7.4.1 Fully exploiting VOI within an iterative context...................312 7.4.2 Exploring uncertainty in sample size calculations.................312 7.4.3 A simple guide to undertaking EVSI..................................313 7.5 Conclusions...................................................................314 Appendix 1: Literature Search Histories ............................................315 Reference List...........................................................................322 VI List of Tables Table 2-1: Typical parameters and their distributional forms in a DAM.........31 Table 3-1: Staging colorectal cancer using AJCC and TNM classifications......69 Table 3-2: Model type and analysis undertaken ....................................72 Table 3-3: Economics literature search..............................................74 Table 3-4: Diagnostic test accuracy ..................................................77 Table 3-5: Inappropriate treatments for inaccurate staging......................95 Table 3-6: Inappropriate treatments for inaccurate diagnosis ..................106 Table 3-7: Inappropriate treatments for inaccurate diagnosis ..................116 Table 3-8: Primary CRC model prevalence parameter estimates ...............118 Table 3-9: Primary colon & rectal cancer DTA parameter estimates...........120 Table 3-10: Primary colon & rectal cancer Treatment parameter estimates .123 Table 3-11: Primary colorectal cancer Survival parameter estimates..........124 Table 3-12: Primary colon & rectal cancer Utility parameter estimates.......126 Table 3-13: Primary CRC model costs...............................................128 Table 3-14: Recurrent CRC model prevalence parameter estimates ...........132 Table 3-15: Recurrent CRC model Diagnostic Test Accuracy estimates........135 Table 3-16: Recurrent CRC model Optimal Treatment estimates...............137 Table 3-17: Recurrent CRC model Survival estimates.............................138 Table 3-18: Recurrent CRC Utility parameter estimates .........................139 Table 3-19: Recurrent CRC model costs.............................................141 Table 3-20: Metastatic CRC model Prevalence parameters......................143 Table 3-21: Metastatic CRC model Diagnostic Test Accuracy parameters .....147 Table 3-22: Metastatic CRC model Treatment Parameters.......................148 Table 3-23: Metastatic CRC model Survival Parameters..........................149 Table 3-24: Metastatic CRC model Utility Parameters............................150 Table 3-25: Metastatic CRC model costs............................................152 Table 3-26: Primary rectal cancer – cost per correct diagnosis .................155 Table 3-27: Primary rectal cancer – DTA outcomes under different strategies ............................................................................................156 Table 3-28: Primary rectal cancer – cost per QALY gain..........................158 Table 3-29: Primary colon cancer – cost per correct diagnosis..................159 Table 3-30: Primary colon cancer – cost per QALY gain ..........................160 Table 3-31: Primary rectal cancer scenario – cost per QALY.....................161 Table 3-32: Primary colon cancer scenario – cost per QALY .....................162 Table 3-33: Recurrent rectal cancer – cost per correct diagnosis...............163 Table 3-34: Recurrent rectal cancer – cost per QALY gain.......................164 Table 3-35: Recurrent colon cancer – cost per correct diagnosis ...............165 Table 3-36: Recurrent colon cancer – cost per QALY gain........................165 Table 3-37: Metastatic cancer – cost per correct diagnosis......................167 Table 3-38: Metastatic cancer – cost per QALY gain ..............................168 Table 4-1: EVPPI parameter groups and parameters..............................214 Table 4-2: EVPPI parameter groups and parameters..............................223 Table 5-1: Model parameters.........................................................245 Table 5-2: Model predictions .........................................................246 Table 5-3: Scenario model predictions..............................................249 VII Table 5-4: Sample size calculation parameters....................................251 Table 5-5: Superiority sample sizes for each endpoint ...........................252 Table 5-6: NMB sample sizes under different power specifications.............258 Table 5-7: NMB sample sizes under alternative monetary thresholds (λ)......259 Table 5-8: Non-inferiority sample size under different power specifications.263 Table 6-1: fFN model predictions....................................................271 Table 6-2: fFN Parameters for probabilistic model................................272 Table 6-3: Net Benefits from posterior calculation ...............................279 Table 6-4: EVSI calculations for a range of sample sizes.........................281 Table 6-5: Sample sizes under different approaches .............................287 Table 6-6: EVSI results for 1 and 10,000 binomial draws.........................291 Table 6-7: Alternative methods for calculating posterior net benefits in EVSI294 Table 6-8: Calculating ENB for EVSI in a non-linear model.......................296 Table 6-9: Duration to run linear simulation with alternative number of priors ............................................................................................299 VIII List of Figures Figure 2-1: The cost-effectiveness plane............................................34 Figure 2-2: CEAC for Treatment versus Control.....................................38 Figure 2-3: The iterative approach to economic appraisal........................56 Figure 3-1: Diagnostic pathway for a two test intervention ......................78 Figure 3-2: Decision tree for staging primary colorectal cancer .................83 Figure 3-3: Treatment pathways for primary rectal & colon cancers ...........88 Figure 3-4: Two-state survival model for each AJCC colorectal cancer stage..91 Figure 3-5: Overall survival for primary rectal cancer for each AJCC stage....92 Figure 3-6: Decision tree for staging recurrent colorectal cancer...............97 Figure 3-7: Treatment pathways for recurrent rectal & colon cancers ........100 Figure 3-8: Two-state survival model for each recurrence status...............103 Figure 3-9: Overall survival for patients suspected of recurrent rectal cancer ............................................................................................104 Figure 3-10: Decision tree for staging metastatic colorectal cancer ...........108 Figure 3-11: Treatment pathways for metastatic colorectal cancer............111 Figure 3-12: Two-stage survival model for metastatic disease status..........113 Figure 3-13: Overall survival for patients suspected of metastatic recurrence ............................................................................................114 Figure 3-14: Accuracy data of PET/CT in the detection of recurrent CRC.....134 Figure 3-15: Accuracy of PET/CT in detecting hepatic metastases.............145 Figure 4-1: The CE plane for PET/CT in primary rectal cancer..................183 Figure 4-2: The CE plane for PET/CT in primary colon cancer ..................185 Figure 4-3: The CE plane for cePET/CT in primary rectal cancer scenario....187 Figure 4-4: The CE plane for cePET/CT in primary colon cancer scenario.....188 Figure 4-5: The CE plane for PET/CT in recurrent rectal cancer................189 Figure 4-6: The CE plane for PET/CT in recurrent colon cancer ................190 Figure 4-7: The CE plane for metastatic cancer...................................192 Figure 4-8: The CEAC for primary rectal cancer...................................195 Figure 4-9: The CEAC for primary colon cancer....................................196 Figure 4-10: The CEAC for primary rectal scenario................................197 Figure 4-11: The CEAC for primary colon scenario ................................199 Figure 4-12: The CEAC for recurrent rectal cancer ...............................200 Figure 4-13: The CEAC for recurrent colon cancer................................201 Figure 4-14: The CEAC for metastatic cancer......................................202 Figure 4-15: EVPI for primary rectal cancer – population level..................206 Figure 4-16: EVPI for primary colon cancer – population level ..................207 Figure 4-17: EVPI for primary rectal cancer scenario – population level.......208 Figure 4-18: EVPI for primary colon cancer scenario – population level .......209 Figure 4-19: EVPI for recurrent rectal cancer – population level................212 Figure 4-20: EVPPI for recurrent rectal cancer – population level (λ=£30,000) ............................................................................................215 Figure 4-21: EVPI for recurrent colon cancer – population level................218 Figure 4-22: EVPPI for recurrent colon cancer – population level (λ=£30,000)220 Figure 4-23: EVPI for metastatic colorectal cancer - population................222 Figure 4-24: EVPPI for metastatic CRC – population level (λ=£30,000).........224 IX Figure 5-1: Decision tree for diagnosing pre-term labour........................242 Figure 5-2: Cost-effectiveness plane for fFN model prediction .................247 Figure 6-1: The CE plane for Treatment (fFN test) versus Control (no test) ..273 Figure 6-2: The CEAC for fFN.........................................................274 Figure 6-3: EVPI for fFN – population level.........................................276 Figure 6-4: EVSI with varying sample size – per patient/decision...............282 Figure 6-5: EVSI with varying sample size – population level....................283 Figure 6-6: EVSI, Cost & ENBS for fFN trial.........................................285 Figure 6-7: Alternative EVSI, Cost & ENBS for an observational study..........286 Figure 6-8: EVSI curves under one and 10,000 binomial draws..................292 Figure 6-9: EVSI curves under alternative number of prior draws...............298