Cost-effectiveness of the implementation of [68Ga]Ga-PSMA-11 PET/CT at initial prostate cancer staging – Insights into Imaging

Patient cohort

Data from the PEPPER-study (NTR6830) was used, which evaluated the diagnostic performance of [⁶⁸Ga]Ga-PSMA-11 PET/CT for initial prostate cancer staging in a prospective study. Patients with a positive skeletal scintigraphy were excluded (Fig. 1).

Flow chart of initial prostate cancer staging and treatment planning in standard of care, PEPPER-study and in the potential PSMA PET/CT strategy. A: Standard of care. B: PEPPER-study. C: PSMA PET/CT strategy (skeletal scintigraphy was replaced by [⁶⁸Ga]Ga-PSMA-11 PET/CT and no ePLND in case of positive iN1 and iM1 findings on [⁶⁸Ga]Ga-PSMA-11 PET but only ePLND in case of negative PSMA). ePLND Extended pelvic lymph node dissection, MRI magnetic resonance imaging, MSKCC Memorial Sloan Kettering Cancer Center, PET/CT positron emission tomography/computed tomography, PSMA prostate specific membrane antigen

For our cost-effectiveness analysis, patients were categorized as: no lymph node metastasis (N0), limited lymph nodes metastasis defined as ≤ 4 pelvic lymph node metastases (N1_lim), extended lymph nodes metastasis defined as > 4 pelvic lymph node metastases (N1_ext), distant metastasis (M1) defined as extra pelvic lymph node metastasis (M1a), or bone and/or visceral metastasis (M1b/c). Ground truth for N-status was always based on pathology results, but for distant metastases this was based on a combination of pathology, additional imaging and/or follow-up (Additional file 1: Fig. S1) [7].

Costs, utilities, disutilities and yearly probabilities

Yearly probabilities, costs and disutilities of treatment procedures were derived from the literature or from internal sources (Additional file 1: Table S1). QoL (utility) values were expressed as quality-adjusted life years (QALY). A QALY of one indicates one year in best possible health, a QALY of zero indicates death. Disutilities express QALY decrement. Costs were expressed as 2020 price levels using the Dutch national price index [11]. No exact data on the impact of ePLND on QoL was found. Using literature and expert opinion, the disutility of this procedure was estimated (Additional file 1: Table S1).

Model development

Decision table

Short-term costs and QoL of diagnosis and treatment of prostate cancer patients were calculated using a decision table (Table 1). This included all possible scenarios for both PSMA PET/CT strategy and standard of care. Diagnostic accuracy was calculated using the frequency outcomes from the PEPPER-study (Additional file 1: Fig. S1). Subsequently, the treatment scheme was obtained using the standard of treatment given the diagnostic outcomes (Table 1). After treatment, the patients transit towards one of four health stages: (NEOD (coming from N0 or N1)), palliative and false palliative (pN1_lim patients being falsely diagnosed as iN1_ext).

Lifetime state transition model

To calculate lifetime costs and QoL of treatment, a lifetime state transition model simulating patients’ follow-up was created, based on previous work of Scholte et al. [10] (Fig. 2). The health stages of the decision table are integrated in the lifetime state transition model together with two additional health states, namely salvage and (cancer) death. Yearly probabilities, cost and utility values of each transition state are shown in Table 2.

Lifetime state transition model used for the different scenarios. The model consists of six health states where patients can find themselves in during follow-up: ‘No evidence of disease after treatment of N0 disease’ (NEOD-N0), ‘No evidence of disease after treatment of N1 disease’ (NEOD-N1), ‘Salvage’, ‘Palliative’, ‘False Palliative’ and ‘Cancer death’. The NEOD states were used to reflect patients who were treated curatively. It was assumed that patients in the NEOD-N0 state would be fully cured and stay there till death. Patients in NEOD-N1 state were assumed to be at risk for biochemical recurrence (BCR), when BCR occurs they transfer towards salvage or directly towards palliative. The salvage state was designed to reflect the period of salvage initialized after BCR would occur. After salvage treatment, they either stay in salvage state or transit to palliative state. The palliative state reflects the long-term palliative period for prostate cancer patients. In this period, no curative treatments are initialized. The false palliative state was designed to mimic the palliative state of patients who are falsely being diagnosed for palliative treatment by [⁶⁸Ga]Ga-PSMA-11 PET/CT. Patients in the palliative state and the false palliative state would stay there until death. Prostate cancer-related death could only occur in the palliative state and the false palliative state. All patients could transit to non-prostate cancer-related death from every state (these lines are hidden)

Average age at model start was 69 years (consistent with the existing patient cohort) and the model ran until death (40 cycles; one cycle corresponded to one year). Yearly discounting percentage of 4% and 1.5% were used for costs and utility outcomes, according to Dutch guidelines [12].

Finally, total costs and QoL were calculated by adding the mean treatment cost and disutility outcomes to the lifetime model costs and QoL outcomes.

For optimal modelling, a number of assumptions were made. Firstly, during state transitioning, subjects in the NEOD-N0 state could not experience BCR. Secondly, patients in the salvage or (false) palliative state could not transit back towards NEOD. Thirdly, regarding diagnostic accuracy, it was impossible for [⁶⁸Ga]Ga-PSMA-11 PET/CT to diagnose pN0 patients as being iN1_ext and vice versa. Fourthly, our model assumed that diagnosis of patients suffering from iM1a/b/c disease by [⁶⁸Ga]Ga-PSMA-11 PET/CT was always correct.

And lastly, PSMA-M1 patients were assumed to be diagnosed as pN0 patients in standard of care, since M1 disease on conventional bone scintigraphy was an exclusion criterion in the PEPPER-study.

Outcomes

Our main outcome: cost-effectiveness was expressed as incremental: costs, QoL (utility), life years and incremental cost-effectiveness ratios (ICERs), for [⁶⁸Ga]Ga-PSMA-11 PET/CT versus conventional diagnostics (ePLND and skeletal scintigraphy), from a health care perspective. The ICER (€/QALY) represents the investment cost for adding one QALY. An ICER was “dominant” when the treatment increased QoL and saved costs. Conversely, an ICER was considered “dominated” when the treatment reduced QoL and increased costs. In other words, a dominant strategy is cost-effective, whereas a dominated strategy is not cost-effective. Net monetary benefit (NMB) was calculated using a willingness to pay (WTP) €80,000, according to Dutch standards [13]. The NMB translates utility values into euros, using the WTP to quantify the net worth of one incremental QALY in €. When the NMB is above zero, the intervention is more cost-effective compared to any given treatment with an ICER of €80,000/QALY, thus creating NMB.

As our main outcome is only based on a single prospective study cohort, additional cost-effectiveness analysis, using the probabilities of the Dutch population were performed. These analyses were added as Additional file 1.

Analysis was performed using R version 4.0.3 and Microsoft Excel version 16.35. An online accessible tool is available at: https://wrke.shinyapps.io/shiny_html_temp/. Technical validation was performed by peer review and by recreating the excel model in R. All inputs values were verified by experts.

Sensitivity analysis

Three types of sensitivity analyses were performed:

Firstly, deterministic sensitivity analysis (DSA) was performed to evaluate the impact of all input parameters individually on model outcomes. All input variables were varied by ± the reported standard error (SE) value (Table 2, Additional file 1: Table S1) and cost-effectiveness result measured in NMB (WTP €80,000) was plotted.

Secondly, probability sensitivity analyses (PrSA) using 10,000 iterations to evaluate combined impact of all parameters uncertainty on model outcomes was performed. PrSA outcomes were plotted on the cost-effectiveness plane, used to calculate the 95% credibility interval (€ NMB). For PrSA, the SE and distributions are shown in Tables 1 and 2 and Additional file 1: Tables S1 and S2.

Thirdly, threshold analysis was performed to evaluate threshold values of parameters until a certain model outcome was reached. This is performed by varying the values of the number of pN1_lim patients who were falsely diagnosed as iN1_ext by [⁶⁸Ga]Ga-PSMA-11 PET/CT (FP) and the disutility of ePLND until a QALY gain was observed.