Cost-effectiveness Analysis of Caspofungin and Fluconazole for Primary Treatment of Invasive Candidiasis and Candidemia in Ethiopia

Background: Caspofungin was shown to be more effective than uconazole in treating patients with invasive candidiasis and/or candidaemia (IC/C). However, cost-effectiveness of caspofungin for treating IC/C in Ethiopia remains unknown. We aimed to assess the cost-effectiveness of caspofungin compared to uconazole as primary treatment of IC/C in Ethiopia. Methods: A Markov cohort model was developed to compare the cost-utility of caspofungin versus uconazole antifungal agents as rst-line treatment for adult inpatients with IC/C from the Ethiopian health system perspective. Treatment outcome was categorized as either a clinical success or failure, with clinical failure being switched to a different antifungal medication. Liposomal amphotericin B (L-AmB) was used as a rescue agent for patients who had failed caspofungin treatment, while caspofungin or L-AmB were used for patients who had failed uconazole treatment. Primary outcomes were expected quality-adjusted life years (QALYs), costs (US$2021), and the incremental cost-effectiveness ratio (ICER). QALYs and costs were discounted at 3% annually. Cost data was obtained from Addis Ababa hospitals while locally unavailable data were derived from the literature. Cost-effectiveness was assessed against the recommended threshold of 50% of Ethiopia’s gross domestic product/capita. Deterministic and probabilistic sensitivity analyses were conducted to assess the robustness of the ndings. Results: In the base-case analysis, treatment of IC/C with caspofungin as rst-line treatment resulted in better health outcomes (12.86 QALYs) but higher costs (US$7,714) compared to uconazole-initiated treatment followed by caspofungin (12.30 QALYs; US$3,217) or L-AmB (10.92 QALYs; US$2,781) as second-line treatment. Caspofungin as primary treatment for IC/C was not cost-effective when compared to uconazole-initiated therapies. Fluconazole-initiated treatment followed by caspofungin was cost-effective for the treatment of IC/C compared to uconazole with L-AmB as second-line treatment, at US$316/QALY gained. Our ndings were sensitive to medication costs, drug effectiveness, infection recurrence, and infection-related mortality rates. Probabilistic sensitivity analysis conrmed the stability of our ndings. Conclusions: Our study showed that the use of caspofungin as primary treatment for IC/C in Ethiopia was not cost-effective when compared with uconazole-initiated treatment alternatives. The ndings supported the use of uconazole-initiated therapy with caspofungin as a second-line treatment to treat IC/C in Ethiopia and other low-income countries.

better health outcomes (12.86 QALYs) but higher costs (US$7,714) compared to uconazole-initiated treatment followed by caspofungin (12.30 QALYs; US$3,217) or L-AmB (10.92 QALYs; US$2,781) as second-line treatment. Caspofungin as primary treatment for IC/C was not cost-effective when compared to uconazole-initiated therapies. Fluconazole-initiated treatment followed by caspofungin was costeffective for the treatment of IC/C compared to uconazole with L-AmB as second-line treatment, at US$316/QALY gained. Our ndings were sensitive to medication costs, drug effectiveness, infection recurrence, and infection-related mortality rates. Probabilistic sensitivity analysis con rmed the stability of our ndings.
Conclusions: Our study showed that the use of caspofungin as primary treatment for IC/C in Ethiopia was not cost-effective when compared with uconazole-initiated treatment alternatives. The ndings supported the use of uconazole-initiated therapy with caspofungin as a second-line treatment to treat IC/C in Ethiopia and other low-income countries. Background Invasive candidiasis and/or candidemia (IC/C) is an increasingly common fungal infection worldwide and has been associated with high rates of mortality, hospitalization, and healthcare cost [1][2][3]. Its incidence is estimated to be 3-5 per 100,000 persons in the general population and 1-2% of all medical and surgical ICU admissions [3]. IC/C attributed mortality rates vary widely, ranging from 10% to 47% [3], and patients' hospital stay is 22-34 days longer compared to those with non-invasive candidiasis [1]. The total healthcare cost per patient with IC/C infection was estimated to range from US$ 48,487 to $157,574, with an average cost of $10,216 to US$ 37,715 per hospitalization [2]. In a study conducted in the US, candidemia was associated with a 14.5% increase in mortality, 10.1-days increase in hospital stay, and attributable costs of∼US$40,000 per patient [4].
Data on IC/C from low-and middle-income countries (LMICs) are scarce, but the few available studies demonstrate a high incidence and very high mortality rate [2,5]. In Ethiopia, Fungi are estimated to infect approximately 8% of the population each year, and the number of persons at risk of IC/C is increasing, owing to the expansion of intensive care units (ICU), and high prevalence of HIV/AIDs, malignancies, chronic diseases and other risk factors [6]. The paucity of evidence on the burden of fungal infections, particularly IC/C, is challenging for healthcare planning [6,7]. In 2017, a comprehensive assessment of the literature on fungal infection in Ethiopia reported that no study on nosocomial fungal infections had been published [7]. A recent study estimated an annual incidence of 5,300 candidemia cases and over 3,600 associated death, assuming a 5 per 100,000 person-year annual incidence and that up to 5% of all hospital beds serve as ICU beds [6].
The burden of IC/C, however, can be reduced by timely treatments with antifungal medications [8,9].
International guidelines recommend using echinocandins as rst-line treatment over azole or polyene antifungals, due to their enhanced clinical outcomes and safety pro le [10][11][12]. Ethiopia has no local treatment guideline for IC/C, making its management a di cult task for clinicians. Furthermore, there is a lack of local evidence on fungal infection epidemiology and insu cient diagnostic options, with hospitals lacking basic yeast diagnostics as well as antifungal susceptibility testing procedures. As a result, antifungal treatment for IC/C has remained empiric [6,13].
Caspofungin and uconazole are the most commonly used antifungals for rst-line therapy of IC/C, and they are used interchangeably due to the improved therapeutic bene t and affordability, respectively.
Likewise, second-line antifungal agents are used mostly interchangeably. We therefore aimed to evaluate the cost-effectiveness of using caspofungin compared to uconazole as empiric rst-line therapy for the treatment of hospitalized patients with IC/C in Ethiopia, a low-resource setting country.
Methods A Markov cohort model was constructed to assess the cost-utility of using caspofungin or uconazole antifungal agents as empiric rst-line therapy,i.e., in the absence of microbiology workup or while awaiting for culture and susceptibility data, for Ethiopian adult inpatients with IC/C. This study was carried out from the Ethiopian health system perspective over a lifetime horizon. Consistent with the health system perspective, we included direct medical costs such as drug acquisition cost, hospitalization costs, cost of diagnosis and monitoring tests. Primary outcomes were expected life years (LYs), qualityadjusted life years (QALYs), costs (US$ 2021), and the incremental cost-effectiveness ratio (ICER) expressed in US$ per QALY gained. The ICER was calculated as the difference in cost between the strategies divided by the difference in effectiveness (QALYs). QALYs were determined by multiplying the years lived in a given health state with the utility weights of that state [14]. Costs and QALYs were discounted at an annual rate of 3%, as recommended for LMICs [15]. Ethiopia has not established a costeffectiveness threshold. The World Health Organization (WHO) recommends a cost-effectiveness threshold of 1-3 times GDP/capita [15]. However, in recent years, the use of this threshold has been widely questioned for a lack of scienti c underpinnings to guide resource allocation decisions [16,17]. Hence, we compared our ICER values against the recently recommended threshold of 50% of a country's gross domestic product (GDP)/capita for LMICs [17]. Ethiopia's GDP per capita at the time of the study was US$952 [18]. The study was designed, conducted, and reported following the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement [19].

Treatment strategies
We compare three treatment strategies: 1) Caspofungin treatment followed by L-AmB (CASPO -> L-AmB): patient received intravenous caspofungin (loading dose 70 mg on day 1, then 50 mg daily maintenance dose for 14 days) and those who had experienced treatment failure were switched to an additional 14 days of L-AmB (3 mg/kg per day for an average weight of 70 kg); 2) Fluconazole treatment followed by caspofungin (FLU -> CASPO -> L-AmB): patient received uconazole oral (800-mg loading dose, then 400 mg daily for 14 days) and those who had failed to respond to uconazole were switched to caspofungin, with L-AmB being used as a rescue agent if infection persisted; 3) Fluconazole treatment followed by liposomal amphotericin B (L-AmB) (FLU -> L-AmB): patient took uconazole oral (800-mg loading dose and 400 mg daily for 14 days thereafter) and if this treatment failed, L-AmB was used as the second-line therapy. In accordance with current practice in Ethiopia, we assumed that uconazole and caspofungin would be prescribed for 14 days on average, regardless of their use as rst-or second-line therapy. We consider the same treatment duration for L-AmB therapy. To evaluate each treatment separately, we assumed that patients who had failed therapy and/or those who had a recurrence would be managed with the same treatment as used for the previous episode. We assumed that patients were hospitalized throughout the treatment period and no patients had their medication dose titrated.
Model structure A Markov cohort model was constructed based on current clinical practice and treatment outcomes of hospitalized IC/C patients receiving different types of antifungal therapy in Ethiopia. We built the model using TreeAge Software (TreeAge Software, Inc., Williamstown, MA). Figure 1 shows a simpli ed illustration of the model structure. A patient in hospital with IC/C could die from infection or be cured and transition to a healthy state, which is de ned as the complete resolution of the infection (i.e. clinical and microbiological success) with no need for additional systemic antifungal therapy [20]. Patients who were rst treated and cured could either stay healthy or develop IC/C again. If the rst-line treatment failed, patients would be switched to second-line antifungal treatment ( Fig 1B). All patients could die from causes unrelated to IC/C. The simulated population re ects the Ethiopian inpatient with a mean age of 39 years (informed by hospital data). All patients were followed in a 3-months time step (cycle length) over their life expectancy.

PARAMETER INPUTS
Model inputs including probabilities, utilities, and costs are reported in Table 1.

Probabilities
In the absence of local data, the literature was used to inform health state transitions. A meta-analysis by Millis et al (2009) reported that caspofungin is superior, with favorable treatment response in 76.1% of patients, as compared to 63% for uconazole and 72.98% for L-AmB [21]. The attributable mortality associated with IC/C was 28.44% in patients who received uconazole and 33.83% with caspofungin.
Recurrent candidemia was found in 4.4% of patients [22] and we assumed the same infection recurrence rate for all treatment strategies. Life expectancy data from the WHO Global Health Observatory for Ethiopia was used to populate age-speci c mortality unrelated to IC/C [23].

ANALYSIS
In the base-case analysis, we consider hypothetical IC/C patients aged 39 years (based on the mean age of adult inpatients at TASH. We performed deterministic and probabilistic sensitivity analyses to assess the impact of parameter uncertainties and the robustness of our analysis. In the deterministic sensitivity analysis, we assessed parameter value uncertainty by varying each input variable within a plausible range of values presented in Table 1. We also perform probabilistic sensitivity analysis, running 10,000 Monte Carlo simulations, in which all input variables were allowed to vary simultaneously according to the prede ned probability distribution (i.e., gamma distributions for costs, and beta distributions for probabilities and utilities).

Base-case analysis
The discounted and undiscounted life years, QALYs, costs, and ICERs are presented in Table 2. Our basecase analysis showed that caspofungin-initiated treatment of IC/C was both more effective and more expensive than uconazole-initiated treatment. Caspofungin-initiated treatment followed by L-AmB as second-line treatment was associated with an expected 12.86 QALYs and a cost of US$7,714 (discounted). Fluconazole-initiated therapy with caspofungin used as second-line treatment resulted in an expected 12.30 QALYs and a cost of US$3,217 (discounted), while uconazole-initiated therapy followed by L-AmB produced 10.92 QALYs and a cost of US$2,781(discounted).
The use of caspofungin as rst-line treatment for IC/C was not cost-effective when compared to uconazole-initiated therapy. Compared to uconazole-initiated therapy with caspofungin as second-line therapy, the caspofungin-initiated treatment resulted in an incremental 0.56 QALYs gained (0.94 QALYs undiscounted) at an incremental cost of US$4,497 (US$7,419 undiscounted); translating to US$ 8,079/QALY. We also compared caspofungin-initiated therapy to uconazole-initiated treatment followed by L-AmB, and found that caspofungin-initiated therapy was more effective and more expensive, yielding an ICER of $2,545/QALY. In the base-case analysis, uconazole-initiated treatment with caspofungin as second-line treatment was cost-effective compared to uconazole-initiated treatment followed by L-AmB with an ICER of US$316/QALY. Sensitivity analysis

Deterministic sensitivity analysis
The results of the deterministic sensitivity analysis are presented in Figures 2 and 3. In the costeffectiveness analysis of uconazole-initiated treatment followed by caspofungin versus uconazoleinitiated treatment followed by L-AmB strategies, a tornado diagram showed that the cost-effectiveness was most sensitive to the cost of medications (L-AmB and caspofungin), probability of infection recurrence, and effectiveness of caspofungin. Varying these parameters over their plausible ranges resulted in ICER values exceeding the reference threshold of US$ 476/QALY (i.e., 50% of Ethiopia's GDP/capita) ( Figure 2). Further, we found that the probability of infection recurrence, L-AmB treatment e cacy, and mortality from infection during caspofungin therapy were the most in uential variables when comparing caspofungin-initiated therapy to uconazole-initiated treatment with caspofungin as second-line treatment (Figure 3). Changing these parameters values over their plausible range, however, did not result in ICER values below the cost-effectiveness threshold.

Probabilistic sensitivity analysis
The cost-effectiveness acceptability curve in Figure 4 shows the proportion of simulations where each treatment alternative was cost-effective at various cost-effectiveness thresholds. At a cost-effectiveness threshold of US$ 476/QALY (i.e. 50% of Ethiopian GDP/capita) and US$ 2,856/QALY (i.e. 3 times GDP/capita), our probabilistic sensitivity analysis showed that uconazole-initiated treatment followed by caspofungin was likely to be cost-effective in all simulations. Caspofungin-initiated therapy was more likely to be cost-effective when the cost-effectiveness criteria was > 8 times Ethiopian GDP/capita.

Discussion
Invasive candidiasis is associated with high morbidity and mortality in Ethiopia, though the exact incidence is unknown due to lack of comprehensive epidemiological data. Despite the growing burden of the disease, patient care remains challenging owing to a lack of diagnostic resources and context-speci c treatment protocols, as well as drug shortages [6]. Our study compared the cost-effectiveness of caspofungin and uconazole-initiated therapies for the treatment of hospitalized patients with IC/C. Caspofungin-initiated treatment was not cost-effective when compared to uconazole-initiated treatments in Ethiopia, which is consistent with studies from other (high resource setting) countries [29,30]. Ou [29] showed that as compared to uconazole, caspofungin was associated with an ICER of €27,339 per successful treatment, which was below the Spanish ICER threshold, suggesting that caspofungin is a cost-effective agent. In our sensitivity analysis, the results remain stable, con rming the robustness of our ndings.
Although caspofungin is more effective than uconazole, it has also been associated with high drug acquisition costs. The use of caspofungin as a second-line drug to uconazole, therefore, could be a viable alternative for Ethiopia and other low-resource setting countries, given their limited budgets. We could not identify any cost-effectiveness studies that assessed the use of caspofungin as a second-line alternative. Our study compared uconazole-initiated treatment alternatives and showed that uconazoleinitiated treatment followed by caspofungin was cost-effective compared to uconazole-initiated treatment followed by L-AmB. Therefore, if a targeted therapeutic decision cannot be made due to a lack of microbiological data, our ndings support the use of caspofungin as a second-line treatment option for hospitalized IC/C patients.
While our study utilized a robust model with relevant sensitivity analyses, it has some limitations. Our analysis is for empiric treatment in the absence of microbiology workup and targeted therapies. It is worth noting that uconazole, unlike caspofungin, has high resistance rates and is ineffective against Candida bio lm. We recommend that hospital microbiology departments improve their ability to isolate fungi and test susceptibility to antifungal medications in order to guide targeted therapies [10,31].
Due to lack of local data, some of our input parameter values such as health state utilities, disease incidence, and associated mortality were obtained from the literature, which might not be applicable to the Ethiopian population. Our ndings, however, remained stable in sensitivity analyses, con rming the robustness of our ndings. Nonetheless, local effectiveness studies of antifungal treatments and other health-related outcome data such as health state utilities are important to improve context-relevance for future cost-effectiveness analyses. Despite these shortcomings, to the best of our knowledge, this is the rst cost-effectiveness study of caspofungin and uconazole for primary treatment of IC/C in Ethiopia. The ndings of the study can inform Ethiopian guidance on antifungal agents for empiric treatment for hospitalized patients with IC/C.

Conclusion
Our study showed that using caspofungin as rst-line treatment for hospitalized patients with IC/C was not cost-effective when compared to uconazole-initiated treatment that includes either caspofungin or L-AmB as second-line treatment. Fluconazole as a rst-line treatment followed by caspofungin was found to be cost-effective compared to uconazole-initiated treatment followed by L-AmB medicine. Ethics approval and consent to participate This study was approved by the Institutional Review Board of the School of Pharmacy, Addis Ababa University, Ethiopia. Permission to collect cost data was sought from the study hospitals. The institutional ethics code doesn't require patient consent since cost data were collected from historical records.

Consent for publication
Not applicable.

Availability of data and materials
No additional data are available for this analysis, however, any model-related requests should be directed to the corresponding author.  Cost-effectiveness acceptability curves. CASPO: Caspofungin; FLU: Fluconazole; L-AmB: Liposomal amphotericin B