Addressing the challenges of health economic modelling in the context of suboptimal evidence base - case study based on a comparison between photodynamic diagnosis and narrow band imaging in non-muscle invasive bladder cancer.
Standard economic models, which have been developed to reflect typical data availability for pharmaceutical products, often struggle with non-pharmaceutical interventions like diagnostics and medical devices, where clinical trial data is more likely to be short-term or non-comparative. This paper explores methodological challenges in evaluating Photodynamic Diagnosis (PDD) via blue light cystoscopy (BLC) versus Narrow Band Imaging (NBI) for non-muscle invasive bladder cancer (NMIBC).
A semi-Markov state-transition cost-utility model was developed for a health technology appraisal (HTA) by the Danish Treatment Council. The basis of the model was a differential effect of the two technologies on the risk of early local cancer recurrence. Due to a lack of direct comparative trials, an indirect hazard-function-based approach was used. Using TreeAge Pro software, baseline hazard curves were derived from ten-year survival data for conventional white light cystoscopy. Hazard ratios (HRs) from meta-analyses were then applied to these curves to derive simulated time-to-event curves for BLC and NBI.
The base-case analysis yielded an incremental cost-effectiveness ratio (ICER) of DKK 70,707/QALY for BLC-TURBT versus NBI-TURBT. Sensitivity analyses confirmed results remained robust and well below the DKK 500,000/QALY willingness-to-pay threshold. The model was most sensitive to HRs for time to first recurrence. Scenario analyses, including traditional parametric extrapolation, yielded consistent ICERs between DKK 36,775 and DKK 215,965/QALY.
The hazard-based workflow effectively integrated survival data from disparate sources, using a software-based method was quicker, simpler and more intuitive to use than conventional statistical methods. The approach used is equally applicable to a partitioned survival structure. The alignment between hazard-based and traditional parametric methods suggests this is a valid, efficient alternative for developing models in the face of evidence gaps.
A hazard-function approach provides a transparent, practical solution for building robust economic models when clinical data is limited or incompatible with standard approaches.
A semi-Markov state-transition cost-utility model was developed for a health technology appraisal (HTA) by the Danish Treatment Council. The basis of the model was a differential effect of the two technologies on the risk of early local cancer recurrence. Due to a lack of direct comparative trials, an indirect hazard-function-based approach was used. Using TreeAge Pro software, baseline hazard curves were derived from ten-year survival data for conventional white light cystoscopy. Hazard ratios (HRs) from meta-analyses were then applied to these curves to derive simulated time-to-event curves for BLC and NBI.
The base-case analysis yielded an incremental cost-effectiveness ratio (ICER) of DKK 70,707/QALY for BLC-TURBT versus NBI-TURBT. Sensitivity analyses confirmed results remained robust and well below the DKK 500,000/QALY willingness-to-pay threshold. The model was most sensitive to HRs for time to first recurrence. Scenario analyses, including traditional parametric extrapolation, yielded consistent ICERs between DKK 36,775 and DKK 215,965/QALY.
The hazard-based workflow effectively integrated survival data from disparate sources, using a software-based method was quicker, simpler and more intuitive to use than conventional statistical methods. The approach used is equally applicable to a partitioned survival structure. The alignment between hazard-based and traditional parametric methods suggests this is a valid, efficient alternative for developing models in the face of evidence gaps.
A hazard-function approach provides a transparent, practical solution for building robust economic models when clinical data is limited or incompatible with standard approaches.