We developed a decision tree cost-effectiveness model (CEM) in Microsoft® Excel® from a US commercial payer perspective, using inputs from a real-world US administrative claims database study [23] and from the published literature. This cost-effectiveness model was conducted based on analyses of a secondary-use database and did not recruit any human participants or involve animals.

The base case CEM compared antiviral treatment with baloxavir versus oseltamivir or no antiviral treatment for symptomatic individuals with influenza in an insured US population. A lifetime time horizon with 3.0% discounting of costs and quality-adjusted life-years (QALYs) was used [24]. Results were reported for a hypothetical population of 100,000 individuals in terms of QALYs and incremental cost-effectiveness ratios (ICERs) over one influenza season. The CEM decision tree pathways were mutually exclusive at the individual level for both overall and high-risk populations, where an individual may or may not experience infection, receive treatment, develop a complication, and recover or die. Individuals with a complication could have outpatient care, or inpatient care with or without admission to the intensive care unit (ICU) (Fig. 1). A summary of key CEM inputs is provided in Table 1.

Decision tree structure of the cost-effectiveness model. Bolded sides indicate terminal nodes. ICU intensive care unit

Clinical trials have shown significantly faster declines in viral load by Day 2 with baloxavir treatment compared with oseltamivir or placebo [13, 14]. The ongoing CENTERSTONE study (NCT03969212) is exploring whether faster cessation of viral shedding with baloxavir treatment would lead to a reduction in viral transmission. Since the CENTERSTONE study was still ongoing when this CEM was developed, the base case CEM conservatively assumed there would be no additional transmission reduction benefit provided by treatment with baloxavir. A scenario analysis was then conducted to explore the potential impact of reduced viral transmission with baloxavir treatment in an insured US population.

A retrospective observational study was conducted using the Merative™ MarketScan® Research Databases from the 2018–2019 and 2019–2020 influenza seasons (October 1 through May 31, based on the Centers for Disease Control and Prevention [25]), which included the Commercial Claims and Encounters, MarketScan Medicare, and Medicaid Supplemental Research Databases [23]. In line with the US indication for baloxavir when the study was designed, patients in the real-world administrative claims study had to be ≥ 12 years of age at the time of influenza diagnosis, defined by International Classification of Diseases, 10th edition Clinical Modification (ICD-10-CM) codes (J09.xx, J10.xx, J11.xx), with continuous medical and pharmacy health plan enrollment for ≥ 6 months before and 30 days after the treatment date (or proxy treatment date for untreated patients). Treatment arms were defined by the presence of a prescription fill for baloxavir or oseltamivir within 2 days after the influenza diagnosis date (treated patients), or no record of any influenza treatment during the study period after influenza diagnosis (untreated patients). Patients were excluded if they received antiviral medication for prophylaxis, had a diagnosis code for coronavirus disease 2019 (COVID-19) during the 2019–2020 influenza season, received treatment with any other influenza antiviral medication (i.e., peramivir, zanamivir), or received baloxavir or oseltamivir > 2 days after influenza diagnosis.

The real-world US administrative claims study included a total of 753,721 patients (baloxavir, n = 14,868; oseltamivir, n = 466,420; no antiviral treatment, n = 272,433) [23]. Approximately 50,000 patients were excluded for having a COVID-19 diagnosis code and 28 patients were excluded for use of peramivir or zanamivir. The real-world study provided inputs for the CEM population including the mean age of 32 years and the proportions of individuals with influenza categorized as OwH (56%; n = 425,589/753,721) or high-risk (44%; n = 328,132/753,721) [23]. Definitions of OwH and high-risk were based on the Centers for Disease Control and Prevention [26] and the CAPSTONE-2 clinical trial [14]. CEM outcomes were modeled separately for the OwH and high-risk populations then combined to provide results for the total population (OwH + high-risk).

The rates of complications in the OwH and high-risk populations were based on patients who developed a complication within the 30 days after receiving baloxavir, oseltamivir, or no antiviral treatment in the real-world US administrative claims study [23]. Prespecified complications included pneumonia, bronchitis, upper respiratory tract infection, acute respiratory distress syndrome, sepsis, otitis media, gastrointestinal bleeding, sinusitis, exacerbation of asthma or chronic obstructive pulmonary disease, or influenza-related cardiovascular, renal, or central nervous system complications (Supplementary Table S1). The prevalence of complications was determined regardless of any history of the condition in the 6 months prior to influenza diagnosis. Complications where ≤ 1% of patients were managed in an inpatient setting (with or without an ICU stay) were redistributed to an outpatient care setting to avoid the influence of outliers (where very few observations would increase the uncertainty and possibly indicate exceptionally high or low costs for a particular care setting) [23]. Since very few observations were redistributed (acute respiratory distress syndrome, 0.07%; sepsis, 0.15%; renal, 0.79%; gastrointestinal bleeding, 0.60%), the impact of the redistributions on the model estimates was considered to be negligible.

In the OwH population, the proportions of individuals with a complication in the baloxavir, oseltamivir, and no treatment arms were 26.2% [95% confidence interval (CI) 25.3, 27.1], 28.1% (95% CI 27.9, 28.2), and 31.8% (95% CI 31.6, 32.0), respectively, based on the real-world US claims study [23]. In the high-risk population, the proportions of individuals with a complication were 30.5% (95% CI 29.3, 31.6), 31.3% (95% CI 31.1, 31.5), and 37.8% (95% CI 37.5, 38.1) in the baloxavir, oseltamivir and no treatment arms, respectively [23]. Proportions of patients with each complication (Supplementary Table S2), management of complications in the outpatient, inpatient (no ICU), or inpatient with ICU stay (Supplementary Table S3), and management costs for complications in each care setting (Supplementary Table S4) were based on the real-world US administrative claims study [23]. The proportions of cured patients in each care setting (Supplementary Table S5) were based on the published literature and on a published National Institute for Health and Care Excellence (NICE) technology appraisal (NICE TA168) of antiviral treatments that included oseltamivir [4, 27, 28]. The cost of an outpatient office visit ($68.10; code 99,213) was taken from the 2023 Centers for Medicare & Medicaid Services Physician Fee Schedule [29].

The duration of influenza symptoms was assumed to be the same for treated patients receiving either baloxavir or oseltamivir based on the generally similar reductions in median TTAS and TTIS observed in the CAPSTONE-1 [13] and CAPSTONE-2 [14] trials. Median TTAS was similar for baloxavir (53.7 h) compared with oseltamivir (53.8 h) among OwH individuals in CAPSTONE-1 (placebo, 80.2 h; P < 0.001 vs. baloxavir) [13], and median TTIS was numerically shorter than oseltamivir (73.2 vs. 81.0 h) among high-risk individuals in CAPSTONE-2 (placebo, 102.3 h; P < 0.0001 vs. baloxavir) [14]. The input values for total mean duration of symptoms were based on NICE TA168 for treated patients (reported for oseltamivir in NICE TA 168 and assumed to be the same for baloxavir) and untreated patients (reported for placebo in NICE TA168) [4]. Mean duration of symptoms for treated individuals was thus 6.9 days for the OwH population and 12.3 days for the high-risk population [4]. Mean duration of symptoms for untreated individuals was 8.9 days for the OwH population and 13.8 days for the high-risk population [4]. All inputs related to duration of symptoms are provided in Supplementary Table S6.

Mean baseline utility score (0.96), the utility associated with influenza (0.81, with or without complications), and the utility associated with recovery from an ICU stay (0.90) were the same as those used in an oseltamivir cost-effectiveness model by Kamal et al. [27, 30,31,32]. Utilities associated with specific complications and the durations of complications are provided in Supplementary Table S7 and Table S8, respectively, based on management in the outpatient, inpatient, or inpatient with ICU setting.

The disutility associated with adverse events was assumed to be 0.21 for both the OwH and high-risk populations [33]. The proportions of patients receiving baloxavir or oseltamivir who might experience a treatment-related adverse event were based on adverse events reported in the CAPSTONE-1 trial for the OwH population (baloxavir, 4.4%; oseltamivir, 8.4%) [13] and the CAPSTONE-2 trial for the high-risk population (baloxavir, 5.6%; oseltamivir, 7.9%) [14]. The most common treatment-related adverse events in both the baloxavir and oseltamivir groups were nausea and diarrhea in both CAPSTONE trials [13, 14]. A sensitivity analysis for this CEM was conducted to explore how identical occurrences of treatment-related adverse events between baloxavir and oseltamivir groups might impact the results. The durations of adverse events were 7 days (assumption) with a disutility of 0.21 [33]. The cost of treating the adverse events was assumed to be $10 for over-the-counter medication.

Drug costs used in the model were calculated per episode of care, including one tablet for baloxavir [20] and 75 mg twice daily for a 5-day course of oseltamivir [34]. The model inputs used the wholesale acquisition costs (as of July 1, 2023) per episode of care of $159.14 for baloxavir and $50.40 for oseltamivir (calculated from the average of all generic oseltamivir drugs available in the US) [35].

One-way deterministic and probabilistic sensitivity analyses were conducted to explore the robustness of the primary findings. Deterministic sensitivity analyses explored a range of complication rates among baloxavir-treated patients (10–50%), the cost of baloxavir treatment ($0 to $300) while the cost of oseltamivir remained constant, and an identical rate of adverse events with baloxavir and oseltamivir (8.4% each for the OwH subset; 7.9% each for the high-risk subset). The probabilistic sensitivity analysis (1000 simulations) generated a cost-effectiveness plane (incremental costs by incremental QALYs) with an expected ICER and a probability of baloxavir being cost-effective at a willingness-to-pay (WTP) of $100,000/QALY gained.

Scenario analysis 1 evaluated the cost-effectiveness of baloxavir using the base case assumptions applied separately to the OwH and high-risk subpopulations. Since the base case analysis assumed no (zero) reduction in viral transmission with antiviral treatment, scenario analysis two explored the potential impact of a range of potential values for reduced viral transmission rates with baloxavir compared with oseltamivir or no antiviral treatment in terms of ICERs and net monetary benefit (NMB). NMB is a function of effectiveness (QALYs) and a WTP threshold of $100,000/QALY gained minus the cost of the intervention, such that an NMB > 0 indicates cost-effectiveness of the intervention [36].