Verteporfin photodynamic therapy for neovascular age-related ...

34 Key changes to the protocolsites. The questionnaires elicited the use of HSS relating to the patient’s eye condition in thepreceding 3 months. This included unscheduled low-vision appointments, use of antidepressants,visits to the general practitioner (GP), visits from social services (mainly home carers) and timein nursing homes, residential care or sheltered housing. The patient-reported HSS use at eachtime point was combined with national unit costs to give an estimate of the costs in the 3 monthspreceding the visit that were ‘attributable’ to the patients’ vision. 54 All costs (inflated to 2007prices) were summed across years 1 and 2 to give total costs per patient and were reported in UKpound sterling. 55Estimating incremental quality-adjusted life-years, and incremental costs forverteporfin photodynamic therapy versus best supportive careThe incremental costs of VPDT versus BSC comprised the mean differences in both interventionand HSS costs between the VPDT and BSC groups. The association of BCVA with HSS cost wasestimated using regression models, following the same strategy used to measure the associationbetween BCVA and HRQoL. Estimates for the CEA were taken from a model which included justBCVA (or CS) as an independent variable because other covariates did not improve model fit.Because a substantial proportion of the sample of patients incurred zero HSS costs, we used a‘two-part’ model. 56 The first part modelled all observations in a logistic regression, with use (ornot) of any service in the 3 months preceding the visit as the dependent variable and BCVA as theindependent variable. The second part of the model included only those observations for whichHSS was used and fitted a linear regression with the HSS cost per user as the dependent variableand BCVA as an independent variable. The resultant conditional probabilities of HSS use andHSS costs per user were combined to predict overall HSS costs with varying BCVA.The CEA then combined the 3-monthly BCVA data from the TAP trial with the associationbetween BCVA and HSS cost to report the incremental HSS costs for VPDT versus BSC. Thesecosts were added to the incremental intervention costs of VPDT and BSC to give the overallincremental costs of VPDT at 1 and 2 years.This association between BCVA and SF-6D was combined with the differential decline inBCVA from baseline for the VPDT and placebo groups in TAP to derive differences in HRQoLbetween VPDT and BSC at 3-monthly intervals. The incremental QALYs for VPDT versus BSCwere calculated as the average HRQoL difference for each 3-monthly time point multiplied by0.25 years and summed over 1 or 2 years. The CEA reported incremental (mean VPDT – meanBSC) costs, QALYs and costs per QALY.Sensitivity analysesA probabilistic sensitivity analysis was undertaken to recognise the sampling uncertaintysurrounding the key parameters (BCVA, association between BCVA and HRQoL, interventioncosts and association between BCVA and HSS costs), and to report the probability that VPDT iscost-effective compared with BSC at different levels of willingness to pay for a QALY gain (e.g.£20,000 per QALY). 28,57Further sensitivity analyses assessed the robustness of the results to the main methodologicalassumptions and data sources used in the base case. Five alternative scenarios were considered:1. The treatment frequency was taken from TAP trials rather than the VPDT cohort study.2. BCVA data for the VPDT group were taken from the cohort study (post- vs pre-VPDT)rather than TAP.