Individualised selection of left-sided breast cancer patients for proton therapy based on cost-effectiveness.

INTRODUCTION: The significantly greater cost of proton therapy compared with X-ray therapy is frequently justified by the expected reduction in normal tissue toxicity. This is often true for indications such as paediatric and skull base cancers. However, the benefit is less clear for other more common indications such as breast cancer, and it is possible that the degree of benefit may vary widely between these patients. The aim of this work was to demonstrate a method of individualised selection of left-sided breast cancer patients for proton therapy based on cost-effectiveness of treatment. METHODS: 16 left-sided breast cancer patients had a treatment plan generated for the delivery of intensity-modulated proton therapy (IMPT) and of intensity-modulated photon therapy (IMRT) with the deep inspiration breath-hold (DIBH) technique. The resulting dosimetric data was used to predict probabilities of tumour control and toxicities for each patient. These probabilities were used in a Markov model to predict costs and the number of quality-adjusted life years expected as a result of each of the two treatments. RESULTS: IMPT was not cost-effective for the majority of patients but was cost-effective where there was a greater risk reduction of second malignancies with IMPT. CONCLUSION: The Markov model predicted that IMPT with DIBH was only cost-effective for selected left-sided breast cancer patients where IMRT resulted in a significantly greater dose to normal tissue. The presented model may serve as a means of evaluating the cost-effectiveness of IMPT on an individual patient basis.

Patient selection for proton therapy: a radiobiological fuzzy Markov model incorporating robust plan analysis.

While proton therapy can offer increased sparing of healthy tissue compared with X-ray therapy, it can be difficult to predict whether a benefit can be expected for an individual patient. Predictive modelling may aid in this respect. However, the predictions of these models can be affected by uncertainties in radiobiological model parameters and in planned dose. The aim of this work is to present a Markov model that incorporates these uncertainties to compare clinical outcomes for individualised proton and X-ray therapy treatments. A time-inhomogeneous fuzzy Markov model was developed which estimates the response of a patient to a given treatment plan in terms of quality adjusted life years. These are calculated using the dose-dependent probabilities of tumour control and toxicities as transition probabilities in the model. Dose-volume data representing multiple isotropic patient set-up uncertainties and range uncertainties (for proton therapy) are included to model dose delivery uncertainties. The model was retrospectively applied to an example patient as a demonstration. When uncertainty in the radiobiological model parameter was considered, the model predicted that proton therapy would result in an improved clinical outcome compared with X-ray therapy. However, when dose delivery uncertainty was included, there was no difference between the two treatments. By incorporating uncertainties in the predictive modelling calculations, the fuzzy Markov concept was found to be well suited to providing a more holistic comparison of individualised treatment outcomes for proton and X-ray therapy. This may prove to be useful in model-based patient selection strategies.

Cost-effectiveness of proton therapy in treating base of skull chordoma.

While proton beam therapy (PBT) can offer increased sparing of healthy tissue, it is associated with large capital costs and as such, has limited availability. Furthermore, it has not been well established whether PBT has significant clinical advantages over conventional volumetric modulated arc therapy (VMAT) for all tumour types. PBT can potentially offer improved clinical outcomes for base of skull chordoma (BOSCh) patients compared with photon (X-ray) therapy, however the cost-effectiveness of these treatments is unclear. In this study, the cost-effectiveness of PBT in the treatment of BOSCh patients is assessed, based on an analysis of comparative radiotherapy treatment plans using a radiobiological Markov model. Seven BOSCh patients had treatment plans for the delivery of intensity modulated proton therapy and VMAT retrospectively analysed. The patient outcome (in terms of tumour local control and normal tissue complications) after receiving each treatment was estimated with a radiobiological Markov model. In addition, the model estimated the cost of both the primary treatment and treating any resultant adverse events. The incremental cost-effectiveness ratio (ICER) was obtained for each patient. PBT was found to be cost-effective for 5 patients and cost-saving for 2. The mean ICER was AUD$1,990 per quality adjusted life year gained. Variation of model parameters resulted in the proton treatments remaining cost-effective for these patients. Based on this cohort, PBT is a cost-effective treatment for patients with BOSCh. This supports the inclusion of PBT for BOSCh in the Medicare Services Advisory Committee 1455 application.

A radiobiological Markov simulation tool for aiding decision making in proton therapy referral.

PURPOSE: Proton therapy can be a highly effective strategy for the treatment of tumours. However, compared with X-ray therapy it is more expensive and has limited availability. In addition, it is not always clear whether it will benefit an individual patient more than a course of traditional X-ray therapy. Basing a treatment decision on outcomes of clinical trials can be difficult due to a shortage of data. Predictive modelling studies are becoming an attractive alternative to supplement clinical decisions. The aim of the current work is to present a Markov framework that compares clinical outcomes for proton and X-ray therapy. METHODS: A Markov model has been developed which estimates the radiobiological effect of a given treatment plan. This radiobiological effect is estimated using the tumour control probability (TCP), normal tissue complication probability (NTCP) and second primary cancer induction probability (SPCIP). These metrics are used as transition probabilities in the Markov chain. The clinical outcome is quantified by the quality adjusted life expectancy. To demonstrate functionality, the model was applied to a 6-year-old patient presenting with skull base chordoma. RESULTS: The model was successfully developed to compare clinical outcomes for proton and X-ray treatment plans. For the example patient considered, it was predicted that proton therapy would offer a significant advantage compared with volumetric modulated arc therapy in terms of survival and mitigating injuries. CONCLUSIONS: The functionality of the model was demonstrated using the example patient. The proposed Markov method may be a useful tool for deciding on a treatment strategy for individual patients.