The coronavirus disease (COVID-19) represents the clinical symptoms of infection with severe acute respiratory syndrome coronavirus type 2 (SARS-CoV-2). This virus causes severe respiratory illness and demonstrates high transmissibility. In March 2020, the World Health Organisation (WHO) classified this disease as a pandemic, with its global emergency status concluding three years later [1]. However, for individuals at high risk of severe progression, this illness continues to pose significant health challenges in Germany, as severe cases are associated with increased hospitalization and mortality risks [2]. Research indicates a hospitalization rate of nearly 15% among elderly patients during the Omicron phase, despite widespread booster vaccination [3]. In Germany, 26% of the population above 15 years of age is categorized as high-risk for severe disease progression [4].
Immunization strategies remain the primary defense against this illness, requiring continuous adaptation to address emerging viral variants [5]. Additionally, vaccine effectiveness diminishes over time, and some individuals fail to develop adequate immune responses [6]. Consequently, effective therapeutics are crucial for infected individuals at high risk of developing severe symptoms. Recent medical advancements aim to reduce healthcare system burdens by lowering hospitalization and mortality rates [7, 8]. Notably, antiviral (AV) treatments play a vital role in improving patient outcomes by reducing severe disease manifestations.
Nirmatrelvir/ritonavir (NMV/r) is an oral AV medication used as outpatient treatment to prevent severe disease progression. In the European Union (EU), NMV/r is approved for treating adults without oxygen requirements who are at high risk of developing severe symptoms [9]. Its efficacy was demonstrated in a double-blind, randomized, placebo-controlled study (EPIC-HR, NCT04960202) involving unvaccinated non-hospitalized adult patients with confirmed primary SARS-CoV-2 infection at high risk of progression. Results showed NMV/r reduced the risk of hospitalization or death by 85.8% (P < .0001) compared to placebo through day 28 in patients treated within 5 days of symptom onset [10, 11]. The EPIC-HR study enrolled patients from July 2021 to December 2021, providing treatment during the Delta variant dominance [10]. Real-world evidence (RWE) data showed similar NMV/r outcomes in mostly vaccinated patients during the Omicron variant era [12]. In vitro data confirmed NMV/r remains effective against Delta, Omicron, Alpha, Beta, and Gamma variants, suggesting future variants will likely remain susceptible. The pandemic disrupted global healthcare systems, revealing weaknesses in national emergency preparedness frameworks and healthcare service provision [13]. It triggered a 6.5% increase in public healthcare funding in Germany from 2019 to 2020 [13]. Despite the pandemic's conclusion, healthcare costs will likely continue rising due to long-term effects. Specifically, elderly populations face higher risks of severe progression, correlating with increased healthcare expenditures and mortality rates [14]. Beyond clinical efficacy in reducing hospitalization or death risks, NMV/r treatment has been considered to reduce healthcare expenditures. Several studies developed economic or cost-effectiveness models (CEM) to examine NMV/r's cost-effectiveness compared to standard care (no anti-SARS-CoV-2 treatment) in non-German healthcare systems [15–17]. These studies consistently showed reduced hospitalizations with NMV/r compared to standard care, lowering hospitalization-related expenditures. Consequently, NMV/r treatment proves cost-effective in multiple healthcare systems [18–24]. Other studies indicate greater cost-effectiveness in higher-risk populations, such as dominant treatment strategies versus standard care for elderly patients and those with cardiovascular comorbidities [21, 15, 16, 23, 24]. For the German healthcare system, this evidence remains limited. Therefore, we adapted a CEM to estimate the health economic impact and cost-effectiveness of NMV/r treatment compared to best supportive care (BSC) in adult patients presenting with mild or moderate symptoms and at high risk of progression from a German health payer perspective. Methods Modelled Population The modelled population included 1,000 infected, vaccinated and unvaccinated German patients eligible for NMV/r treatment according to labeling, i.e., adults diagnosed without oxygen requirements and at increased risk of progressing to severe disease [9]. For defining increased risk, we followed Robert Koch Institute (RKI) guidance, including: 60+ years, BMI >30, pregnancy, trisomy 21, pre-existing conditions like cardiovascular disease, chronic lung disease (COPD), chronic liver and renal disease, diabetes, cancer, and weakened immune systems [25, 26].
The cohort’s average starting age was 58 years in the base case analysis. This was determined based on the median age of hospitalized patients eligible for NMV/r using German health claims data [2]. A scenario analysis used a starting age of 51 years corresponding to the average age of the 18+ population in 2023, representing the likely minimum average age for the target population [27].
Intervention and Comparators
The intervention was NMV/r treatment, with BSC as the comparator [10]. NMV/r was assumed as add-on to BSC. BSC consists of drugs relieving symptoms, e.g., antipyretics, inhalation, or anticoagulants. Other AV treatments available in Germany include Remdesivir, Imdevimab, Sotrovimab, and Tixagevimab/Cilgavimab [28]. Due to insufficient outpatient infrastructure for parenteral applications and monitoring requirements, Remdesivir is primarily used for hospitalized patients. Other approved AV treatments are rarely used in German healthcare due to limited effectiveness of anti-spike protein monoclonal antibodies against current variants [28]. Considering the absence of alternative AV treatments and appropriate comparator selection, comparison of NMV/r to no treatment/BSC was chosen to represent standard care [29].
Model Structure
The cost-effectiveness of NMV/r versus BSC was evaluated using a population-level, closed cohort model with mutually exclusive health states following treatment and disease progression. The analysis adopted a German healthcare payer perspective with a lifetime time horizon. The cohort initially entered a 1-year short-term decision tree with daily cycle length evaluating short-term consequences of daily infections including outpatient treatment, hospitalizations, and death (Figure 1). Within the short-term decision tree, patients followed either outpatient or hospitalization sub-decision trees. Hospitalized patients’ mortality and healthcare costs were determined by care level: general ward (GW), intensive care unit (ICU), or ICU + mechanical ventilation (MV). Non-hospitalized patients were assumed to receive outpatient management.
Survivors entered a long-term Markov model simulating remaining lifetime with annual cycle length. This included three mutually exclusive health states: alive and discharged after MV, alive and discharged without previous MV, or deceased. This structure allowed increased post-discharge mortality probability requiring MV.
Model Outputs
Primary outcome was incremental cost-effectiveness ratio (ICER) expressed as incremental cost per life-year (LY) gained. ICER using incremental cost per hospitalization avoided was also reported. These outcomes reflected decision-guiding endpoints in German benefit assessment, as recommended by IQWiG Methods Guidance [30, 29]. Other outputs included clinical outcomes (hospitalizations, ICU admissions including MV, inpatient deaths, and LY) and costs (medication, hospitalization, post-discharge, outpatient, and total costs).
Clinical Inputs
In base case analysis, NMV/r efficacy was informed by EPIC HR clinical trial used in German benefit assessment and NMV/r EMA labeling [9]. Efficacy was assumed constant across starting age scenarios, supported by subgroup analyses showing efficacy regardless of age [31, 32]. Hospitalization and death reduction with NMV/r versus BSC was 85.8% (Table 1). The EPIC-HR trial was conducted in unvaccinated populations during Delta dominance [10]. However, RWE evidence from mostly vaccinated patients infected with Omicron is available. To check model outcome robustness with different NMV/r effectiveness assumptions, we conducted scenario analysis using RWE data from Lewnard et al. 2023 [12]. Scenario 2 effectiveness estimate from Lewnard et al. 2023 showed 79.6% (95% CI: 33.9% – 93.8%) reduction in hospitalization and death with NMV/r versus BSC in patients treated within 5 days of symptom onset [12]. BSC represented baseline risk compared to NMV/r treatment effect estimates.
Hospitalization risk for untreated population was derived from national surveillance data for Omicron era (November 2021 to April 2022), considering vaccination status. Data were available for age but not comorbidities, so age was used as proxy for high-risk status. Base case used 35+ years age group risk (4.9%). Additional scenarios used 15+ years (3.3%) and 60+ years (14.6%) as lower and upper boundaries [3].
Hospital mortality estimates used same age groups (35+ for base case, 15+ for scenario 3, and 60+ for scenario 4). This was derived by weighting unvaccinated and vaccinated proportions of hospitalized deaths in high-risk patients by vaccination data from Sievers et al. 2022, adjusted for age using hospitalization relative risk between age groups as mortality proxy from Sievers et al. 2022 [2, 33, 3]. Calculations and inputs for mortality risk are shown in Table 2. Scenario analysis tested mortality risk by care level (GW, ICU, and ICU + MV) using data from Günster et al. 2021 and Hildebrandt et al. 2024 (Table 3) [34–36]. Literature for clinical inputs representative of German healthcare was identified through systematic literature research (Table 6 Supplementary Material 1).
Model assumed no outpatient mortality, as all severe cases were assumed hospitalized [37]. General population mortality rates were taken from German life tables and applied to all alive patients discharged without MV [27]. Patients discharged post-MV had increased death probability (hazard ratio 1.33) for first 5 years in long-term Markov model [38].
Adverse events weren’t included, as EPIC-HR study showed fewer than 5% experienced Grade 3+ events, which most impact costs and quality of life [10].
Costs
Cost benefits were assessed from statutory health insurance perspective. Costs associated with healthcare services were included in CEM: drug medication, administration, hospitalization, post-MV discharge, outpatient resource use (Table 1). Sick pay was included in scenario analysis (scenario 6). Except post-MV discharge costs, all costs were incurred within short-term decision tree. Costs were collected for or inflated to August 2024 Euros (€) reflecting current inflation rates [30, 39]. Medication costs for NMV/r were included for each treated individual representing published price by Statutory Health Insurance (SHI) [40]. NMV/r was assumed add-on to BSC. Treatment costs captured only antiviral costs. BSC alone had no treatment costs. Medication and administration costs were one-time costs upon infection. BSC and orally administered AV had no administration costs.
Direct hospitalization costs were estimated for each hospital course (GW, ICU, and ICU + MV) and immediate post-discharge costs. CEM assumed hospitalization course was same for both NMV/r and BSC. No estimate for reduced care escalation was included. Healthcare service at each level was assumed to cover all related services, including preceding levels. Costs per case for each hospital course were taken from Joshi et al. 2024 [41].
For patients discharged after MV, increased healthcare cost of 3,667.94 € was used [42]. This was applied as one-off cost at Markov model start reflecting short infection duration (<1 month), so cohort would accrue most post-MV discharge costs in first model year. Outpatient patients presented with mild-moderate symptoms, so healthcare resource use inputs assumed only practitioner office use for both NMV/r and BSC. Quarterly lump sum was used assuming one infection causes one general practitioner visit [43]. Indirect costs, e.g., payer costs due to work inability for absence from work were included in scenario analysis (scenario 6). This was estimated using reported proportion with long-term absenteeism multiplied by average sick days taken post 42 days off work either for inpatient or outpatient treatment from SHI claims data study [44]; then multiplied by productivity losses per day of incapacity for work and employment rate for Germany [45, 46]. Discounting for costs and outcomes was set to 3.0% annually, per IQWiG guidelines [30]. Scenario Analyses and Sensitivity Analysis Individual parameter influence and overall uncertainty in model results were tested in deterministic sensitivity analysis (DSA), scenario analysis, and probabilistic sensitivity analysis (PSA). DSA identified individual parameter impact on outcomes. This analysis varied each input individually by ±10% or clinically relevant range (defined by 95% CIs). PSA determined overall uncertainty impact and was performed by jointly varying all model parameters over 1,000 simulations, calculating 95% credible range estimates for each outcome. For PSA, beta distribution was used for probabilities and gamma distribution for costs. Input distributions were based on Briggs et al. 2012 [47]. From 1,000 iterations, cost-effectiveness acceptability curve (CEAC) was constructed showing NMV/r probability of being cost-effective compared to BSC over willingness-to-pay (WTP) thresholds from €0 to €100,000. Scenario analyses were conducted for key inputs and assumptions as described earlier. Inputs used are listed in Table 3. Results Base Case Analysis Results For full cohort of 1,000 patients, compared to BSC, treating high-risk patients with NMV/r decreased hospitalizations (6.96 vs. 49.00 for NMV/r and BSC respectively). This equals 0.0420 hospitalization reduction per patient treated with NMV/r. NMV/r treatment also reduced total ICU admissions including MV (1.04 vs 7.35 admissions for NMV/r and BSC respectively), inpatient deaths (0.06 vs 2.74 deaths for NMV/r and BSC respectively) and increased LY by 46 LY (17,062 vs 17,017 LY for NMV/r and BSC respectively) compared to BSC (Table 4). This equals 0.0063 ICU admissions including MV reduction per patient, 0.0027 inpatient death reduction per patient, and 0.05 LY increase per patient treated with NMV/r. For full cohort of 1,000 patients, NMV/r treatment led to higher medication costs than BSC (1,084,190 €) and slightly higher outpatient costs (742 €). However, total hospitalization costs decreased by 620,838 € and post-discharge costs lowered by 8,168 € for NMV/r compared to BSC. This equals 1,084.19 € medication cost increase per patient, 0.74 € outpatient cost increase per patient, 620.84 € hospitalization cost decrease per patient, and 8.17 € post-discharge cost decrease per patient for NMV/r compared to BSC. Therefore, NMV/r treatment resulted in overall net increase in total costs of 456 € per patient (1,205.83 € vs. 749.90 € per patient NMV/r and BSC respectively) (Table 4). Considering differences between NMV/r and BSC in hospitalization (-0.0420 per patient), LY (0.05 per patient), and costs (456 € per patient), yielded following ICER: incremental cost per LY gained of 9,954 € and incremental cost per hospitalization avoided of 10,845 € with NMV/r treatment (see Table 4). Scenario Analyses Results Scenario analyses exploring cohort starting age, real-world effectiveness, and indirect costs were consistent with base case result. Model was more sensitive to analyses exploring varying baseline hospitalization risk or inpatient mortality for modeled population. Using lower hospitalization and mortality risk estimates from 15+ population as proxy for high-risk patients yielded higher total incremental healthcare costs (661 € per patient), fewer LY gained (0.03 LY per patient), and fewer hospitalizations avoided (0.0283 per patient) for NMV/r compared to BSC. Conversely, using higher estimates from 60+ population to proxy high-risk patients yielded cost-saving result of 788 € total healthcare costs saved per patient, 0.4 LY gained per patient, and 0.1253 hospitalizations avoided per patient treated with NMV/r compared to BSC. This indicates NMV/r would be dominant treatment strategy compared to BSC if hospitalization and mortality risk for high-risk patients could be proxied by 60+ age group as NMV/r generates better clinical outcomes and cost savings (see Table 5). Scenario analysis using inpatient mortality rates by care level increased LY gained (0.10 LY per patient) compared to base case (0.05 LY per patient). Sensitivity Analyses Results DSA shows five most influential parameters on model outcomes. ICER was most sensitive to hospitalization proportion and NMV/r medication costs. Further, it was influenced by hospitalization reduction, inpatient mortality proportion, and GW hospitalization costs (see Figure 2). Hospitalization-related parameters had largest impact on mortality, hence appear as drivers of the model. PSA confirmed model result robustness, as probabilistic average is consistent with differences in LY gained and costs between NMV/r vs BSC obtained with base case analysis (see Figure 3). It also showed NMV/r has approximately >99% probability of being cost-effective at WTP threshold of 20,000 € per LY gained (see Figure 4).
Discussion
The aim of this analysis was to estimate NMV/r treatment cost-effectiveness in adult patients with mild to moderate symptoms at high-risk of progressing to severe disease course from German healthcare payer perspective. NMV/r was assessed to have considerable additional benefit when compared to BSC [29]. Our modeling results showed treatment with NMV/r substantially decreases overall hospitalizations (-0.0420 per patient treated), ICU admissions including MV (-0.0063 per patient treated), and inpatient deaths (-0.0027 per patient treated) when compared to BSC. Clinical outcome gains of NMV/r versus BSC were achieved at incremental cost of 456 € and led to incremental gain of 0.5 LY per patient treated. Base case analysis yielded ICER of 9,954 € per LY gained and 10,845 € cost per hospitalization avoided per patient treated with NMV/r compared to BSC.
Model demonstrates NMV/r treatment prevents serious disease course that otherwise would lead to more expensive treatment in German hospitals. This not only frees up hospital resources that can be allocated to other diseases but also possibly alleviates long-term socioeconomic consequences after severe incident, specifically in population generally considered vulnerable due to underlying risk factors [48]. Germany will face sociodemographic shift towards predominantly elderly population [49]. Age is main risk factor for serious disease and other diseases and correlates strongly with hospitalization risk. Having preventive treatment strategies like NMV/r might aid healthcare systems in future pandemic preparedness [50].
Results were robust in multiple scenario analyses. In analysis with higher hospitalization risk based on 60+ population, treatment with NMV/r demonstrated to be dominant strategy. Outcome was driven by gains in better clinical outcomes for high-risk patients treated with NMV/r leading to LY gains of 0.4 and cost savings of 787.78 € per patient treated.
While inclusion of indirect costs also had minor impact on ICER, scenario analysis using inpatient mortality by care level increased NMV/r cost-effectiveness compared to BSC. This was due to higher baseline mortality modeled hence NMV/r generated additional LY.
DSA confirmed observed improvements in LY can be attributed primarily to NMV/r capacity to reduce hospitalization risk of untreated patients. PSA showed treatment with NMV/r would remain cost-effective with >99% probability at WTP thresholds of 20,000 € per LY and >45% probability at threshold of 10,000 € per LY. These thresholds are well below implicit WTP threshold estimates per LY gained reported for Germany in literature [51, 52].
Analysis has several limitations. Firstly, evidence to inform hospitalization risk of our target population (adult high-risk patients) in the model is lacking. We therefore recurred to estimates of hospitalization risk from various age groups (15+; 35+, 60+) as proxies for this parameter, to include lower and upper boundaries for this parameter. Hospitalization risk in 15+ age group might be underestimate of risk in our target population, whereas hospitalization risk in 60+ age group could be overestimate.
Second, evidence on average cohort starting age of our target population (adult high-risk patients) is missing. In base case, average age of hospitalized high-risk patients was taken as estimate for this parameter. As this estimate didn’t include non-hospitalized high-risk patients this might be overestimation. Average starting age in scenario analyses based on average age of adult 18+ population might underestimate eligible population age. True age of eligible population in clinical practice most likely will be between base case and scenario analysis estimates [27]. NMV/r efficacy was assumed constant across starting age scenarios. Median age in EPIC-HR (46 years) was similar to lower starting age (51 years) which generated consistent results with base case (58 years).
Third, disease is dynamic with regular shifts in dominant viral variants and patient population epidemiology, which could lead to variations in virulence, treatment effectiveness, and baseline risk of severe disease outcomes, and as consequence, model inputs could become out of date. In addition, as treatment landscape keeps evolving, further research comparing effectiveness of NMV/r to other AV treatments may be needed. However, both our scenario analysis and sensitivity analysis demonstrated that results were robust to changes in key inputs such as hospitalization risk and treatment efficacy/effectiveness. Fourth, cohort approach used in model might not accurately capture eligible population heterogeneity. Additionally, RWE effectiveness data were from United States (US) study. Effectiveness in German population might differ but no RWE data for Germany were available. Fifth, model assumes patients are only infected once with disease during a year and impact of post-disease was not incorporated into model. Caution should therefore be taken in any generalizations of model results for future decision making.
To our knowledge, this is first cost-effectiveness analysis of impact of treatment with NMV/r on patients with high risk for severe disease in Germany. However, cost-effectiveness of NMV/r was analyzed using same model in Belgium, Greece, Japan, Netherlands, Portugal, Sweden, and US [18–24]. Depending on country, either unvaccinated or mix of vaccinated and unvaccinated patients was analyzed. Comparator was in all cases standard of care (e.g., no treatment). Results across countries are consistent in showing high probability of NMV/r being cost-effective or even cost saving (dominant strategy) for elderly patients [53–55, 20, 21, 56, 22, 23, 57, 24]. These findings are supported by further analyses across world using other models [54–56, 58, 57].
Conclusions
NMV/r treatment of patients with disease at high-risk for severe progression improves clinical outcomes, reduces healthcare resource use, reduces hospitalization costs, and preserves LY compared to BSC. This analysis suggests treating patients with NMV/r would likely be cost-effective or even cost-saving from German health payer perspective. This analysis provides crucial economic rationale for decision making by policy makers.
Transparency
Declaration of funding
This study was sponsored by Pfizer.
Declaration of financial/other interests
EOiK, EF and AK are employees of Pfizer Pharma GmbH. TM is employee of Pfizer Ltd. TM, EF and AK hold Pfizer stocks. CS, DE, HS, MM and AP are employees of IQVIA.
Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.
Author Contributions
AK was responsible for design and conception of study. CS, MM and AP were responsible for systematic literature search for and compilation of input parameters representative of German healthcare context. HS was responsible for adapting model to German healthcare context. All authors contributed to interpretation of data, drafting of manuscript, revising manuscript critically for intellectual content, and final approval of version of manuscript to be published.
Acknowledgments
No assistance in preparation of this article is to be declared.
Previous presentations
Poster presentation at ESCMID Global 2025