Empirical anti-Candida therapy among selected patients in the intensive care unit: a cost-effectiveness analysis.

Background: Mortality from invasive candidiasis is high. Low culture sensitivity and treatment delay contribute to increased mortality, but nonselective early therapy may result in excess costs and drug resistance.

Objective: To determine the cost-effectiveness of anti-Candida strategies for high-risk patients in the intensive care unit (ICU).

Design: Cost-effectiveness decision model.

Data Sources: Published data to 10 May 2005, identified from MEDLINE and Cochrane Library searches, ICU databases, expert estimates, and actual hospital costs.

Target Population: Patients in the ICU with suspected infection who have not responded to antibacterial therapy.

Time Horizon: Lifetime.

Perspective: Societal.

Interventions: Fluconazole, caspofungin, amphotericin B, or lipid formulation of amphotericin B given as either empirical or culture-based therapy and no anti-Candida therapy.

Outcome Measures: Incremental life expectancy and incremental cost per discounted life-year (DLY) saved.

Results Of Base-case Analysis: Ten percent of the target population will have invasive candidiasis. Empirical caspofungin therapy is the most effective strategy but is expensive (295,115 dollars per DLY saved). Empirical fluconazole therapy is the most reasonable strategy (12,593 dollars per DLY saved) and decreases mortality from 44.0% to 30.4% in patients with invasive candidiasis and from 22.4% to 21.0% in the overall target cohort.

Results Of Sensitivity Analysis: Empirical fluconazole therapy is reasonable for likelihoods of invasive candidiasis greater than 2.5% or fluconazole resistance less than 24.0%. For higher resistance levels, empirical caspofungin therapy is preferred. For low prevalences of invasive candidiasis, culture-based fluconazole is reasonable. For prevalences exceeding 60%, empirical caspofungin therapy is reasonable. For caspofungin to be reasonable at a prevalence of 10%, its cost must be reduced by 58%.

Limitations: Less severe illness and limited use of broad-spectrum antimicrobial agents, typical of smaller hospitals, could result in a lower risk for invasive candidiasis.

Conclusions: In patients in the ICU with suspected infection who have not responded to antibiotic treatment, empirical fluconazole should reduce mortality at an acceptable cost. The use of empirical strategies in low-risk patients is not justified.