OBJECTIVES: To estimate the cost-effectiveness of caspofungin versus amphotericin B, voriconazole, and anidulafungin in the treatment of invasive candidiasis in a subgroup of patients who are infected by fluconazole-resistant Candida species and in patients who cannot be treated with fluconazole due to other reasons. METHODS: For this analysis a lifetime Markov model was used with following health states: success, failure, and death over two treatment lines. A randomized clinical trial comparing caspofungin and amphotericin B was used to estimate daily transition probabilities. The corresponding probabilities for voriconazole and anidulafungin were estimated based on indirect comparisons using amphotericin B as the common comparator. The analysis was conducted from a societal perspective. Direct costs inside the health care system were included. Outcomes were reported as cost per life-year gained. To determine the robustness of the model and the impact of uncertainty, uni- and multivariate sensitivity analyses were carried out. RESULTS: In the base case analysis the 95% confidence intervals of the incremental life years were estimated at 0.23(-0.29, 0.85), 0.21(-0.31, 0.88), and 0.11(-0.37, 0.71) compared to amphotericin B, voriconazole, and anidulafungin respectively. Corresponding intervals for incremental costs were €1825(-€3954, €5581), €40(-€5243, €3354), and -€415(-€4938, €2570). Cost per life year gained was €7942 and €189 compared to amphotericin B and voriconazole. Caspofungin was dominant compared to anidulafungin. Probabilistic sensitivity analyses showed that at a willingness to pay threshold of €20,000/QALY, caspofungin has a 62%, 72%, and 68% probability to be cost-effective relative to respectively amphotericin B, voriconazole and anidulafungin. Univariate sensitivity analyses showed that results were sensitive to changes in efficacy parameters and infection-related mortality. CONCLUSIONS: The present model suggests that caspofungin is cost-effective in the treatment of invasive candidiasis in this specific subgroup of patients. Potential limitation of the model is that efficacy and failure estimates are partly based on indirect comparisons.