This article is based on previously conducted studies; the authors did not conduct any new studies with animals or human participants.

The target population for this cost-effectiveness analysis was pediatric patients with SBS in Japan. An average starting age for treatment of 6 years was used in the model, based on results from a 24-week phase 3 trial of teduglutide in pediatric patients with SBS (TED-C14-006) [17, 26]. Teduglutide was compared with SOC for SBS, which aims to alleviate symptoms through optimization of residual intestinal function using PS and enteral nutrition. We conducted analyses from three perspectives: (1) the public healthcare payer’s perspective, (2) the public healthcare and long-term care payer’s perspective (i.e., the utility of caregivers was accounted for in addition to perspective 1), and (3) the societal perspective (i.e., productivity loss for caregivers was accounted for in addition to perspective 2). We used a Markov model with a lifetime analysis period and an analysis cycle of 1 month. The discount rate was 2% per annum for both costs and effects [27].

The model structure is shown in Fig. 1. The model was based on the following four health states, defined by the number of days that PS was required per week: “No PS” (without PS–PS withdrawal state); “Low PS” (PS required 1–3 days per week); “Mid PS” (PS required 4–5 days per week); “High PS” (PS required 6–7 days per week); and “Death” (total of five states). Patients entered the analysis model based on the PS distribution at baseline in the TED-C14-006 trial [17, 26]. They could transfer reversibly between PS states each cycle or irreversibly transition to “Death”.

Model structure. IFALD intestinal failure-associated liver disease, PS parenteral support

The analysis model also considered the occurrence of IFALD as a complication of PS. Patients who developed IFALD were assumed to progress irreversibly to liver fibrosis and cirrhosis [28]. In the model, teduglutide treatment was discontinued if the reduction in PS from baseline to 24 weeks was less than 20%. Of those patients who were still alive 2 years after the start of teduglutide treatment, 15% continued treatment for the rest of their lives and 85% discontinued treatment. After discontinuation of treatment, there was no transition between PS states and only mortality was considered.

The main parameters used in the analysis are summarized in Supplementary Table 1. The patients’ health status at the start of the analysis and transition probabilities between PS states were set on the basis of the TED-C14-006 trial [17]. The probabilities of transitioning between states in the first 6 months of treatment (the duration of the TED-C14-006 trial) were calculated on the basis of the observed monthly PS administration days in patients from the teduglutide 0.05 mg/kg/day and SOC treatment groups. Between 7 and 24 months, teduglutide was assumed to repeat the average transition probabilities seen in the final 3 months of the TED-C14-006 trial [17]. From 25 months onwards, there was assumed to be no transition between PS states. SOC was assumed to revert to the baseline distribution at 6 months, with no transition between states from 7 months onwards.

The prevalence of IFALD in patients with SBS was based on results of a modified Delphi survey used in the cost-effectiveness evaluation of teduglutide by the National Institute for Health and Care Excellence (NICE) in the UK [28]. The progression of patients with IFALD to liver fibrosis and cirrhosis was cited from the report by Cavicchi et al. [29]. We considered adverse events (AEs) related to treatment with both teduglutide and SOC with occurrence based on AEs experienced by at least 20% of patients in either the teduglutide 0.05 mg/kg/day group or the SOC group in the TED-C14-006 trial [17]. The occurrence rate of AEs after discontinuation of teduglutide treatment was assumed to be the same as that seen in the SOC treatment group. Mortality was set according to the PS state. Mortality for patients not receiving PS was assumed to be the same as that of the general Japanese population [30]; mortality for patients in other PS states was estimated on the basis of the report by Fullerton et al. for pediatric patients with IF in the USA [31].

Costs related to teduglutide treatment include the medication cost of teduglutide and management cost associated with self-injection. On the basis of the average weight of the study population in a pediatric teduglutide clinical trial in Japan, the prescribed dosage of teduglutide was assumed to be one vial per day (at a reimbursement cost of 79,300 JPY) [26, 32, 33]. Colonoscopy costs were also included, as detailed in the teduglutide package insert [26, 32]. The cost for each PS state was calculated on the basis of results of a medical resource consumption survey (targeting physicians who treat patients with SBS) [34], and according to the medical service fee point summary and drug price standard [32, 35]. Treatment costs associated with AEs were estimated on the basis of the guidelines and the expected treatment, with costs set according to medical service fee point summary and drug price standard [32]. For AEs for which additional treatment is not necessary or that can be managed with over-the-counter drugs, no cost was assumed. The cost for treating IFALD was set on the basis of the study by Ikeda et al. [36].

As there are no reported utilities in the literature for pediatric patients with SBS based on the number of days of PS required, we used the mean utility scores from a study of adult patients in the UK [37] to set the following utility values for each PS state: No PS, 0.820; Low PS, 0.717; Mid PS, 0.545; High PS, 0.385. We set the utility value for IFALD on the basis of a study by Hirao et al. [38] and used the weighted average value (0.830) of utilities for chronic hepatitis (inactive and active) and liver cirrhosis (compensated and decompensated) for both hepatitis B and hepatitis C. The utility reduction due to AEs was set on the basis of referenced literature [38].

The main parameters related to caregiver utility are summarized in Supplementary Table 2. We considered caregiver utility based on the PS state of the care recipient, with caregiver utility derived from the results of the EQ-5D survey completed by caregivers of patients with SBS in the UK [39]. The utility for caregivers in the “No PS”, “Low PS”, “Mid PS”, and “High PS” patient states was accumulated over the analysis period.

Productivity loss for pediatric patients with SBS was not considered but productivity loss for caregivers was included. The main parameters related to caregiver productivity loss are summarized in Supplementary Table 2 and are based on the results of a Work Productivity and Activity Impairment (WPAI) survey conducted on caregivers of patients with SBS in the UK (the same source as used for caregiver utilities), taking into account the PS state of the care recipient [39]. The gender ratio of caregivers, employment rate, percentage of caregivers in regular employment, percentage of full-time stay-at-home parents of either sex, average wages, and the value of unpaid labor were set by extrapolating from Japanese national statistics for the population [40,41,42,43]. On the basis of the opinions of Japanese and global medical specialists, the number of caregivers per patient was assumed to be 1.5 caregivers for patients under 18 years of age, and 0.8 caregivers for patients aged 18 years and over. The gender ratio of caregivers was based on the ratio published by the Ministry of Internal Affairs and Communications [42], and the age of caregivers was assumed to be 50 years throughout the analysis period.

The QALYs and ICERs were calculated to assess the cost-effectiveness of teduglutide compared with SOC. For the base-case analysis, we conducted the cost-effectiveness analyses from (1) the public healthcare payer perspective, (2) the public healthcare and long-term care payer perspective, and (3) the societal perspective using the analysis conditions and base value parameters described above and in Supplementary Tables 1 and 2.

For the societal perspective (3), the most comprehensive analysis, we conducted a deterministic sensitivity analysis (DSA) and a probabilistic sensitivity analysis (PSA). The setting range and distribution of each variable are provided in Supplementary Tables 1 and 2.

For the DSA, the sensitivity analysis range was set on the basis of the 95% confidence interval (CI). For parameters without a 95% CI, a range of ± 20% was used. The discount rate was set within the range of 0–4% according to the cost-effectiveness evaluation guidelines [27]. Parameters of the survival function for mortality were assumed to follow a multivariate normal distribution; for PS state transition probabilities, the 2.5th and 97.5th percentiles of the Dirichlet distribution function were set as lower and upper limits, respectively. The analysis results were presented using tornado diagrams, highlighting the top 10 variables with the largest variation.

For the PSA, QoL values were assumed to follow a beta distribution for probabilities (excluding PS state transition probabilities), costs were assumed to follow a gamma distribution, PS state transition probabilities were assumed to follow a Dirichlet distribution, and survival curve parameters were assumed to follow a multivariate normal distribution. For parameters without reported or estimable standard errors (SE), a SE value of 10% of the set value was assumed. We generated random numbers following the specified probability distributions for each parameter, and the simulation was performed by conducting 10,000 data extractions.

Two different scenario analyses were conducted for perspective (2), the public healthcare and long-term care payer perspective. In scenario 1, the decreased utility (negative value; taken from the UK study) [39] of the caregiver from “No PS” to “Low PS”, “Mid PS”, and “High PS” was cumulatively added to the utility of the patient over the analysis period; the utility of caregivers in the “No PS” state was assumed to be zero (the decrement method). In scenario 2, the state-specific utility of caregivers of patients with SBS was obtained using the Delphi method (utility values for this scenario are shown in Supplementary Table 2) [44]. Similarly, two different scenario analyses were also conducted for the societal perspective (3). In scenario 3, the analysis combined the conditions of scenario 1 with the productivity loss of the caregiver (decrement method and productivity loss). In scenario 4, the analysis combined the conditions of scenario 2 with the productivity loss of the caregiver (Delphi panel and productivity loss).