A cost-consequence model was designed to compare the clinical outcomes (efficacy and safety) and costs associated with three approved treatment alternatives ([177Lu]Lu-DOTA-TATE, everolimus and sunitinib) for patients with advanced GEP-NETs (NET G1 and NET G2; progressed following SSA therapy) in Italy. The main metric of this study was the cost of each treatment option per progression-free month, taking into account the costs of drug acquisition, administration and toxicity.
Median PFS data for patients with advanced (unresectable or metastatic) PanNETs or GI-NETs were taken from the official summaries of product characteristics (SmPCs) [15, 22, 23], and were consistent with the key results of pivotal trials of each treatment option [10-12, 14, 16, 24]. Median PFS, defined as the time from randomisation to progressive disease or death, was chosen for several reasons. Firstly, it was the primary efficacy endpoint in each trial [10-12, 14, 16, 24], and secondly, it was considered to most accurately reflect the survival benefit of [177Lu]Lu-DOTA-TATE, particularly in patients with progressive disease. In contrast to median OS, median PFS is not confounded by post-progression treatment, including treatment crossover. Thus, median PFS is a more direct reflection of underlying antitumour efficacy, and, consequently, cost per progression-free month is a relatively unbiased outcome measure. Thirdly, analyses based on median PFS are straightforward, and not confounded by the accrual of costs over a longer duration during the post-progression state. Lastly, at the time of developing the present model, median OS for [177Lu]Lu-DOTA-TATE was not yet known, precluding its use in the model.
Model assumptions on the efficacy [177Lu]Lu-DOTA-TATE in GI-NETs were based on the results of the phase III NETTER-1 trial [14, 15]. This was a randomised, active-controlled comparison of [177Lu]Lu-DOTA-TATE, plus best supportive care (BSC; octreotide LAR 30 mg every 28 days for symptom control), versus high-dose (60 mg) octreotide LAR every 28 days in patients with progressive, somatostatin receptor–positive, advanced midgut NETs. Of the 231 patients recruited, 117 received [177Lu]Lu-DOTA-TATE plus BSC and 114 received high-dose octreotide LAR. The groups were balanced with respect to tumour grade, somatostatin receptor expression and previous treatment . For PanNETs, median PFS for [177Lu]Lu-DOTA-TATE was obtained from the single-arm, monocentric, phase I/II Erasmus study [15, 16].
Median PFS data for everolimus and sunitinib were based on the results of randomised phase III placebo-controlled trials in which patients received BSC in addition to assigned study treatment. The RADIANT-3 and RADIANT-4 trials compared everolimus with placebo in patients with PanNETs and GI-NETs, respectively [10, 11], while another trial (NCT00428597) compared sunitinib with placebo in patients with PanNETs [12, 24].
Costs: base-case scenario
Costs were calculated from the perspective of the Italian national health service (NHS) over a time horizon of 1 year, and are reported in 2020 euros (€) [25-27].
For [177Lu]Lu-DOTA-TATE, acquisition costs were calculated for a complete course of therapy, consisting of four intravenous infusions of 7400 MBq each . The analysis was conservative because, on average, patients in the NETTER-1 trial received 3.55 doses instead of the 4 doses recommended in the approved SmPC . For everolimus (both generic and originator products) and sunitinib, acquisition costs were calculated for 1 year of continuous therapy at the relevant approved dosage (everolimus: 10 mg/day orally; sunitinib: 37.5 mg/day orally) [22, 23], and assumed that adherence was 100%. This approach reflects the structure and design of the respective Agenzia Italiana del Farmaco (AIFA) registries created for the management of these drugs in Italy, and takes into account the periodic efficacy assessments that are required for ongoing reimbursement.
Patients with NETs routinely receive treatment with an SSA for the management of symptoms associated with functional tumours. Therefore, the acquisition costs of supportive treatment with a long-acting SSA were calculated for 1 year’s treatment at an ‘average’ dose (octreotide LAR: 30 mg every 28 days; lanreotide: 120 mg every 28 days), and an assumed adherence rate of 100%. For [177Lu]Lu-DOTA-TATE, additional costs were included for 4 weeks’ pre-treatment with a short-acting SSA, in line with the SmPC  and the design of the NETTER-1 trial.
Additionally, [177Lu]Lu-DOTA-TATE must be co-administered with an intravenous amino acid infusion to reduce the radiation dose absorbed by the kidneys, and, consequently, the risk of renal toxicity (see Supplementary Appendix 1 in Online Resource 1) . In clinical trials, some amino acid products were associated with nausea and vomiting that necessitated the use of anti-emetic treatment . Therefore, costs were included in the model for both amino acid and anti-emetic therapy. It was conservatively assumed that all patients would require such treatment; however, the amino acid infusion recommended by the manufacturer of [177Lu]Lu-DOTA-TATE (see Supplementary Appendix 1 in Online Resource 1) causes nausea and vomiting at a much lower rate. Given the relatively low costs of such therapies, it was also assumed that these costs were included in the basic national tariff associated with Diagnosis-Related Group (DRG) 409 (see below) .
[177Lu]Lu-DOTA-TATE can only be administered in a designated nuclear medicine facility by healthcare professionals authorised to handle radiopharmaceuticals . Although current Italian legislation (Art. 158 of Legislative Decree 101/2020) does not mandate that patients receiving lutetium-based radiopharmaceuticals are treated as in-patients, we conservatively assumed that each dose of [177Lu]Lu-DOTA-TATE would require hospitalisation for at least 1 day. Hospitalisation costs for [177Lu]Lu-DOTA-TATE were calculated using the National Tariff for Hospital Services for Acute Patients, DRG 409 . Everolimus and sunitinib are administered orally, and therefore no administration costs related to overnight hospitalisation were assigned to either drug in the model.
Management of adverse events
To calculate the costs associated with specialist appointments and/or hospitalisations resulting from grade 3 or 4 AEs, current tariffs for specialist ambulatory services and for DRG-related hospital services were used. For ambulatory service costs, the most recent versions of the National Tariff for Hospital Services for Acute Patients  and the Regional Tariff for Hospital Services for Lombardy (the Nomenclatore Tariffario di Regione Lombardia)  were used.
The base-case assumption was that any grade 3 or 4 AE would prompt at least one consultation at a specialist centre. The cost of these visits was calculated as the mean of the tariffs in the National  and Regional Tariff for Hospital Services (Lombardy) . We also assumed that, in practice, a specialist visit would lead to hospital admission, at least in a certain percentage of patients, either for a standard in-patient stay (≥1 day) or for day care (<1 day). For each treatment, we identified the DRG codes associated with the management of each of the grade 3 or 4 AEs documented in phase III clinical trials [10-12, 14], and calculated the costs for both types of hospitalisation.
Subsequently, to estimate the frequencies of each type of hospital admission (≥1 or <1 day), several remote working sessions were held between March and May 2020. These were attended by clinical experts from three Italian centres with extensive experience in the treatment of NETs (FS, DC, GL, RL and SB). Hospital admission frequencies were based on recent clinical experience at the selected centres, and were entered into the model only after consensus was reached.
The purpose of sensitivity analysis is to investigate the degree to which the conclusions are sensitive to changes in the assumptions that underpin the model. However, sensitivity analyses tend to be arbitrary, and may not reflect reality; for example, drug acquisition costs used in the base-case analysis are official prices minus mandatory discounts, but discounts in real-world practice may be greater. Thus, instead of presenting our own sensitivity analyses, conducted by adjusting key variables by predetermined amounts or percentages, we developed a simple, flexible tool that allows healthcare professionals and payers in Italy to input data relevant to their own institutions; this allows immediate generation of institution-specific cost-consequence data for each treatment option.
Given the flexibility of the model, several simulations were performed to address specific questions of potential interest to real-world users in the context of the Italian NHS. Following discussion in an ad hoc working session, the authors identified the following questions, based on expected changes in drug acquisition costs, real-world experience with regard to variance in dose exposure and adherence, and potential changes in hospitalisation modalities for patients receiving [177Lu]Lu-DOTA-TATE:
- modelled costs per progression-free month obtained in the base-case simulation for [177Lu]Lu-DOTA-TATE?
- How would modelled costs per progression-free month change if the dose intensity or adherence assumed for the three treatments were lower than 100%?
- How would the modelled costs per progression-free month change if patients receiving [177Lu]Lu-DOTA-TATE were treated in a day-hospital setting, instead of being hospitalised for at least 1 day?
In each simulation, all other variables were kept constant, in order to explore the impact of each specific factor on the end result.
 Legislative Decree no.101/2020 reorganised the sector legislation, establishing safety standards in order to protect people from the risks associated with exposure to ionising radiation in the industrial, medical and research environments. It also regulates the safety of nuclear plants, installations and activities in which radioactive materials are used, as well as the management of spent fuel and radioactive waste.