A de novo dynamic budget-impact model (BIM) with an underlying Markov structure was developed in Microsoft Excel® to estimate the 3-year financial impact of ocrelizumab SC in patients affected by RRMS on the budget of Italian hospitals.

The model was based on the logic and dynamics of previous economic evaluations [11, 12], but it incorporated the most recent data on national-level drug utilization [8]. The analyses were conducted from the perspective of MS centers (hospital setting), considering only direct healthcare costs.

The BIM compared a current scenario, in which ocrelizumab SC was not yet available, with an alternative scenario, reflecting the expected changes in the treatment mix after its introduction. Budget impact (BI) was calculated as the difference in total costs between the alternative and current scenarios.

The BIM used a dynamic approach that simulated the evolution of patients over time (Fig. 1a). Each year, newly eligible patients were included and followed until the end of the analysis using a Markov model embedded within the BIM (Fig. 1b).

Model structure. a Patient flow. Per-patient costs for each newly treated cohort were estimated by tracking patients over the time horizon and then multiplied by the corresponding cohort size. Total costs were summed across all active cohorts each year. b Markov model structure. EDSS Expanded Disability Status Scale, RRMS relapsing–remitting multiple sclerosis, SPMS secondary progressive multiple sclerosis, Y year

According to current Italian Health authorities’ prescription criteria, the modeled population are patients with RRMS with high disease activity despite treatment with at least one DMT (HA RRMS), or with rapidly evolving severe disease (RES RRMS) [13]. The pool of eligible patients was estimated by application of local epidemiological data to the Italian population (Table 1) [14].

The RRMS population treated with DMTs was estimated using Italian prevalence and incidence data [2]. In the first year, the target population included patients switching from an existing DMT to another therapy or starting a HA DMT for the first time. From the second year onward, entry into the model was restricted to patients initiating a HA DMT, assuming an increase in their proportion from 3.5% to 7% over the 3-year period (data on file [15]). This reflects the expected increase in the use of high-efficacy DMTs from the earliest stages of the disease. Indeed, clinical evidence shows that early initiation of high-efficacy therapies significantly reduces long-term disability [2].

The treatments included in the BIM were the DMTs currently approved and reimbursed for treatment lines beyond the first, or for RES forms, as indicated in the AIFA (Italian Medicines Agency) paper prescription form [13].

Appendix Table 1 in Online Resource 1 of the electronic supplementary material (ESM) shows the market share adopted for the two compared scenarios—current and alternative—over the 3-year observation period. Values for the current scenario were elaborated from market research based on Italian real-world drug utilization data (data on file [15]), reflecting current treatment patterns in the target population. In the alternative scenario, based on internal company assumptions, ocrelizumab SC was expected to reach a market share of 14.8, 29.7, and 45.0% in years 1, 2, and 3, respectively. Its market uptake was assumed to be drawn mainly by ocrelizumab IV, ofatumumab, and cladribine, which are the most commonly used HA DMTs among target patients in the current scenario. The remaining market shares were distributed evenly among all other available alternatives, with the exception of biosimilars, which followed identical trajectories in both scenarios. This assumption reflects the expectation that ocrelizumab SC would primarily draw market share from other originator DMTs rather than from biosimilars.

A Markov model, whose structure is illustrated in Fig. 1b, was used to simulate patient clinical evolution over time and to estimate costs separately for each treatment. Health states were defined by the Expanded Disability Status Scale (EDSS) score, which guides clinical decision making in patients with MS [11].

The cohort entered the model distributed across RRMS health states according to ocrelizumab trials data [7, 16]. In each 1-year cycle, patients in the model could transit between EDSS scores within ‘RRMS, treated’; discontinue treatment; progress to SPMS (with an EDSS score at least 1 point higher than the last score before conversion to SPMS); or die. SPMS patients could progress to a health state with a higher EDSS score (between 2 and 9), but they could not regress to a lower EDSS score or return to an RRMS health state. Relapses were modeled as events occurring within health states. The probabilities of disability progression and relapse were held constant over time and depended on the disease form (RRMS or SPMS) and the EDSS score. In the RRMS states, DMTs reduced the risk of disability progression and relapse rate. Treatment discontinuation was assumed to occur when patients reached an EDSS score ≥7 or transitioned to SPMS states. Following discontinuation, patients were assumed to continue incurring disease-related costs but no longer received active DMT.

The probability of death in each cycle was calculated based on the general mortality rates of the Italian population by age and sex, adjusted using relative risks (RRs) specific to each EDSS level in both RRMS and SPMS, which take into account the increased mortality associated with the disease. No direct treatment effects on mortality were simulated.

The 3-year time horizon was divided into equal cycles of 1 year each. A half-cycle correction was applied in the model, assuming that transitions between health states occur, on average, halfway through each cycle.

Baseline characteristics, such as age, sex, and disability status distribution at baseline, were derived from the OPERA I–II clinical trials (as a proxy for the target population) [6] (Appendix Table 2 in Online Resource 1, ESM).

The natural history of the disease was simulated using transition matrices which define the rates of disability progression and evolution to the SPMS form. In the absence of HA/RES-specific data, estimates derived from broader RRMS populations were applied in the model. The probabilities of transition among EDSS levels in RRMS patients were derived from the British Columbia database [17] (Appendix Table 3 in Online Resource 1, ESM), while the progression rates for the SPMS form and the annual probabilities of conversion from the RR to the SP form were obtained from the London Ontario registry (Appendix Tables 4 and 5 in Online Resource 1, ESM) [18]. The annualized relapse rate (ARR) in the absence of treatment, specific to EDSS score and clinical form, was based on the pharmacoeconomic evaluation by Cortesi and colleagues [11] (Appendix Table 6 in Online Resource 1, ESM). Relative treatment effects versus placebo, expressed as hazard ratios (HRs) for the 12-week confirmed disability progression (CDP) and risk ratios (RRs) for the ARR, were derived from a network meta-analysis (NMA) (Table 2) and reflect evidence from general RRMS populations [19]. The efficacy of the SC formulations was considered equivalent to that of the IV formulations.

Following the methodology proposed by Brandes et al. [20], the impact of non-adherence (proportion of days covered [PDC] <0.8) on ARR was estimated based on data from a US retrospective study [21] (Table 2). The proportions of adherent patients and discontinuation risk were derived from the study by Moccia et al. [8], using ocrelizumab as a reference. For ofatumumab, the treatment discontinuation HR was derived from real-world data on treatment use with ocrelizumab and ofatumumab [9]. With the lack of specific data, adherence and persistence of ublituximab were assumed to be the same as those of ocrelizumab.

All-cause mortality in patients with RRMS was calculated based on age- and sex-specific general mortality rates for the Italian population [22], adjusted by the relative risk of mortality associated with MS by EDSS score, estimated through linear interpolation of data from Pokorski [23] (Appendix Table 7 in Online Resource 1, ESM). In line with previous NICE appraisals [24, 25], mortality was assumed to depend solely on EDSS and not to vary according to disease form or disease activity.

The economic analysis adopted the hospital perspective. Consequently, only direct healthcare costs were considered, namely for drug administration, monitoring, adverse events, disease management, and relapse. Drug acquisition costs were included in a scenario analysis, using the maximum hospital tender price. Unit costs, updated to 2024 values [26], were collected from Italian sources.

The infusion costs of DMTs were calculated by proportionally adjusting the IV and SC administration costs of natalizumab—estimated in the EASIER 1 and 2 time-and-motion studies [27, 28]—based on the infusion and monitoring durations (Appendix Table 8 in Online Resource 1, ESM), as well as premedication requirements, reported in the summary of product characteristics (SmPC) for each treatment [29,30,31,32,33,34,35,36,37,38]. The cost of oral administration was assumed to be zero. The cost of monitoring during the first administration of fingolimod was estimated based on Stanisic et al. [39]. For ozanimod and ponesimod, monitoring costs were assumed to be 30% of those for fingolimod, consistent with the assumptions used in the NICE submission for ponesimod [24].

Annual treatment costs (acquisition and administration) were calculated using adherence data [8]. Specifically, non-adherent patients were assumed to receive 80% of the doses.

Annual pre-treatment (additional visits and diagnostic tests prior to starting a new therapy) and monitoring costs were estimated per therapy following guidelines and SmPC data, using hospital tariffs and healthcare professionals time-based cost assumptions (Appendix Table 9 and 10 in Online Resource 1, ESM) [25, 40,41,42,43]. For ofatumumab and ublituximab, resource consumption of ocrelizumab was applied. Unit costs were derived from the official national tariff schedule, used as a proxy for hospital costs [44].

Adverse event (AE) management costs were calculated by combining incidence rates with unit management costs (Appendix Table 9 and 11 in Online Resource 1, ESM) [40, 45,46,47,48,49,50]. Non-serious AEs were assumed to require outpatient care, while serious AEs were costed based on inpatient hospital stays (€911.78/day) [51] with average duration from national Diagnosis-Related Group (DRG) data [52] (Appendix Table 9 in Online Resource 1, ESM).

Calculated annual costs per treatment are reported in Table 3.

Disease management costs for MS patients were based on Italian data by disability level, covering hospitalizations (excluding those due to relapses), specialist visits, diagnostics, and use of additional medications [11, 53] (Appendix Table 9 in Online Resource 1, ESM). Relapse-related healthcare costs were extrapolated from published estimates [54, 55], updated to 2024 at €1638 per relapse (Appendix Table 9 in Online Resource 1, ESM).

A univariate deterministic sensitivity analysis (DSA) was carried out to identify main drivers and uncertainties. Each parameter was varied individually within a plausible range while holding all the others constant. The ranges were defined based on 95% confidence intervals reported in the literature, when available, or by applying a ±20% variation to base-case values in the absence of specific evidence. Results were shown in a tornado chart highlighting the 10 most influential parameters.

In addition, the following scenario analyses were performed: (A) inclusion of drug acquisition cost; (B) no impact of non-adherence on ARR (RRnon-adherent vs. adherent = 1); and (C) varying market penetration of ocrelizumab SC.