2.1

Model design

We developed a half-cycle corrected state-transition Markov model using TreeAge Pro Suite 2022 (TreeAge Software, Cambridge, Massachusetts) [ ] to estimate the direct health system costs and quality adjusted life years (QALYs) of MWA versus RA-PN for patients with unilateral early stage RCCs with tumor size ≤4 cm. The model adopted an Australian healthcare system perspective over a 10-year horizon with a 3% annual discount rate applied for both cost and health outcomes. A discount rate is applied in health economic models with long-term horizons to reflect the present value of future outcomes [ ], and a 3% was selected based on the latest evidence with 1.5% and 5% considered in sensitivity analyses [ , ]. The model structure ( Fig. 1 ) was based on a previously published model [ ] and tailored to fit the Australian setting based on clinical opinions from 3 interventional oncologists.

Clinical pathway diagram for the management of a 67-year old patient with a sporadic, unilateral, cortical, small renal mass (≤4 cm) with ablation versus partial nephrectomy. (A): Simplified schematic of decision trees. (B): Markov model with 4 health states simulating lifetime health outcomes and costs.

The analysis was based on a 2-part constructed model, decision tree ( Fig. 1 A) and Markov model ( Fig. 1 B ), designed to represent the 30-day after undergoing the intervention and the long-term periods of time. A decision tree was used to allocate patients to an initial management strategy, with MWA as the intervention arm and RA-PN as the comparator. We assumed that all tumors were completely removed after partial nephrectomy, whereas 80% of tumors were completely ablated after 1 MWA procedure, and the remaining 20% were ablated after undergoing a second procedure [ ]. The model included the probability of major complication rates, and associated costs, for each arm [ , , ]. Minor complications were excluded due to their tendency to self-resolve and unlikely associated management costs. Peri-operative death wasn’t considered since both procedures reportedly have similar, negligible mortality rates.

After the short-term decision tree model, patients would enter a 4-state Markov state transition model with a 12-month cycle length. All patients initially entered a “progression free survival” health state, with the potential to remain in this state at the end of each cycle or to experience a local recurrence, distant metastases, or death due to RCC or non-RCC reason. Patients with local recurrence were at risk of developing distant metastases or death, while those with metastatic RCC were subject to risk of death. Relevant treatment and follow-up therapies were assumed to be provided to patients once local recurrence or distant metastasis occurred. Supplementary Fig. 1 depicts the treatment pathway for patients after MWA or RA-PN in the Australian setting.

2.2

Model inputs and data sources

The data selection process followed the International Society for Pharmacoeconomics Outcomes Research Modelling Good Research Practices guidelines [ , ]. All input parameters are detailed in Table 1 and Supplementary Table 3 along with the plausible range of uncertainty tested within 1-way sensitivity analyses. Supplementary Table 4 outlines data inputs with respective statistical distributions as applied in the probabilistic sensitivity analyses.

2.3

Base case patient cohort

Characteristics of the base case cohort were modelled based on a prospective observational cohort study ( n = 41 patients with renal, lung or liver tumors) conducted in 2 metropolitan tertiary hospitals in Queensland, Australia, between July 1, 2021 to June 30, 2022. Further details of this study have been previously reported [ ], and a summary of the demographic and clinical characteristics of this cohort is provided in Supplementary Table 3 . Consistent with the characteristics of this cohort, the patient population was assumed to be males aged 67 years with unilateral early stage RCCs and tumor size ≤4 cm.

2.4

Transition probabilities

Recurrence-free survival rate and metastasis-free survival rate were derived from a propensity-matched cohort study focusing upon long-term follow-up of oncologic outcomes of percutaneous MWA ( n = 59) and RA-PN ( n = 59) for the treatment of T1a RCC [ ]. Metastasis-free survival rate after local recurrence was assumed the same for both arms and was extracted from a retrospective study investigating 54 patients who underwent open nephrectomy for locally recurrent RCC [ ] ( Table 1 ). Five-year RCC-related survival rates were collected from the Surveillance, Epidemiology, and End Results (SEER) database and related publications [ , ]. Annual state transition probabilities and mortality rates were then calculated using the formula: <SPAN role=presentation tabIndex=0 id=MathJax-Element-1-Frame class=MathJax style="POSITION: relative" data-mathml='1−(FutureValuePresentValue)(1/n)’>1−(𝐹𝑢𝑡𝑢𝑟𝑒𝑉𝑎𝑙𝑢𝑒𝑃𝑟𝑒𝑠𝑒𝑛𝑡𝑉𝑎𝑙𝑢𝑒)(1/𝑛)1−(FutureValuePresentValue)(1/n)
1 − ( F u t u r e V a l u e P r e s e n t V a l u e ) ( 1 / n )
. The background mortality rate was estimated from the 2020-2022 Australian Life Table [ ] ( Table 1 ).

2.5

Healthcare cost estimates

For cost parameters which are strongly influenced by the nature of local health systems, we selected Australian data sources so that our model could represent as closely as possible the patient heterogeneity and healthcare delivery within the Australian setting ( Table 1 ). The healthcare system perspective was limited to direct health expenditures for small renal masses management and excluded direct nonmedical costs (e.g., transportation) and informal care costs. The hospital admission costs for ablation were derived from a micro-costing analysis conducted alongside the prospective cohort study at the 2 study hospitals [ ]. The cost for RA-PN was sourced from average hospital admission cost data for robotic assisted nephrectomy, the most commonly conducted type of surgery for patients with small renal masses at 1 study hospital. Major perioperative complications were associated with increased costs, and associated costs were estimated based on AR-DRG codes. The assumed follow-up treatment pathway for patients after MWA or RA-PN is depicted Supplementary Fig. 1 . National health service reimbursement rates under the Pharmaceutical Benefits Scheme or Medicare Benefits Schedule per course of treatment were used to assign unit costs for follow-up care items such as immunotherapy, pathology, CT scans, and drug administration [ ]. All costs were reported in 2022 Australian Dollars.

2.6

Health state utility estimates

QALYs were calculated by weighting survival time (i.e., length of life) with utility values (i.e., quality of life) that reflect health related quality of life. Patients’ utility values within the progression-free survival health state were assumed to increase to the level of general population from the second month after intervention, and were based on the latest Australian population norms [ ]. This is supported by the small but similar overall peri-operative (minor and major) complications between MWA and RA-PN for small renal tumors [ ]. We adopted utility values for local recurrence and distant metastases from published studies that most closely reflected the patient population and health states in our model ( Table 1 ) [ , ]. The EQ-5D-5L population norms in Australia was derived from the latest cross-sectional survey of 9,958 general population aged 18 or older [ ]. These studies were sourced from PubMed and the Tufts Medical Centre Cost Effectiveness Analysis Registry [ ].