1. Mark Jit Modelling and Economics Unit Health Protection Agency, London Case study: Structural uncertainty in human papillomavirus vaccination modelling M Jit 1,2, Y Choi 1, N Gay 1 and WJ Edmunds 1,3 1 Centre for Infections, Health Protection Agency 2 University of Birmingham 3 London School of Hygiene & Tropical Medicine

2. Outline Introduction to the problem Representing structural uncertainty Selecting optimal scenarios Presenting results

4. Gardasil TM Protects against HPV 6, 11, 16, 18. Doses at 0, 2, 6 months. Aluminium adjuvanted. Licensed widely, in use in several countries. Cervarix TM Protects against HPV 16, 18. Doses at 0, 1, 6 months. AS04 adjuvanted. Licensed in Europe and Australia. HPV vaccines

5. Female HPV prevalence in UK DNA studies

6. Epidemiological uncertainty Progression and regression of HPV- related neoplastic states. Duration of HPV infection. Prevalence of anogenital warts among HPV 6 and 11 infected people. Duration of vaccine protection. Coverage of vaccination programme. Uncertainty about disease burden Cost and QoL impact of screening. Cost and QoL impact of treating neoplasias, cancer, warts. Accuracy of cytological screening and DNA testing. Structural uncertainty Natural regression of HPV-related neoplasias. Existence of natural immunity to HPV. Pattern of sexual partnerships between age and risk groups. Vaccine protection against non- vaccine HPV types.

7. Sexual transmission stage Changes in disease end points after vaccine introduction Type specific HPV prevalence by cytological status (ARTISTIC trial) Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life) Cytological status at time of screening (Cervical screening programme) Sexual partnership patterns (Sexual lifestyle survey 2000) Cost and quality of life data (current literature) Diagnosed anogenital warts cases (GP and GUM clinic returns) Duration, QoL detriment of warts episode (York GUM study) Types of models used

8. Disease progression 162 models HPV type 3 scenarios (16, 18, others) Transition rates 8 parameters (in range [0,1]) Regression 2 scenarios (stepwise, immediate) Accuracy of cytological screening 3 scenarios (High, Medium, Low) Clinical specificity of DNA test 3 scenarios (100%, 99.75%, 99.5%) HPV prevalence 3 scenarios Sexual transmission stage Changes in disease end points after vaccine introduction Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life)

9. Jit M, Gay N, Soldan K, Choi YH, Edmunds WJ. Estimating progression rates for human papillomavirus infection from epidemiological data. Medical Decision Making (in press). Sexual transmission stage Changes in disease end points after vaccine introduction Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life)

10. Disease transmission 450 models Duration of natural immunity 5 scenarios (0, 3, 10, 20,  years) Parameters governing sexual partnerships 2 parameters (in range [0,1]) Duration of vaccine protection 3 scenarios (10, 20,  years) Risk of HPV transmission per partnership 2 scenarios (low, high) Assortativeness of sexual partnerships 3 scenarios (low, medium, high) Duration of infection 5 scenarios (6, 9, 12, 15, 18 months) Economic stage Changes in the economic burden (cost and quality of life) Sexual transmission stage Changes in disease end points after vaccine introduction Natural history stage Progression and regression rates

11. Economic stage Changes in the economic burden (cost and quality of life) Sexual transmission stage Changes in disease end points after vaccine introduction Natural history stage Progression and regression rates

12. Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life) Sexual transmission stage Changes in disease end points after vaccine introduction

13. Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life) Sexual transmission stage Changes in disease end points after vaccine introduction

14. Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life) Sexual transmission stage Changes in disease end points after vaccine introduction

15. Economic evaluation 50,000 samples Scenario variables 12 variables (integer coded) Distribution of incremental cost effectiveness ratios Duration of vaccine protection 3 scenarios (10, 20,  years) Model parameter variables 9 variables (continuous) QALY variables 14 variables (continuous) Cost variables 14 variables (continuous) Sexual transmission stage Changes in disease end points after vaccine introduction Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life)

16. Sexual transmission stage Changes in disease end points after vaccine introduction Natural history stage Progression and regression rates Economic stage Changes in the economic burden (cost and quality of life) Jit M, Choi YH, Edmunds WJ. Economic evaluation of human papillomavirus vaccination in the United Kingdom. BMJ 2008; 337:a769.

17. Some relevant issues that were raised during this work When is uncertainty structural rather than parametric? How should different structural scenarios be weighed or selected? How can we improve the computational efficiency of structural uncertainty analysis?