Does One Voriconazole Breakpoint Suit All Candida Species?

Abstract

Non-species-dependent interpretative breakpoints for voriconazole and Candida species were proposed by the Antifungal Susceptibility Subcommittee of the Clinical and Laboratory Standards Institute in January 2005 (susceptible, ≤1 μg/ml; susceptible dose dependent [SDD], 1 to 2 μg/ml; and resistant, ≥4 μg/ml) (11). The proposed breakpoints were based on the MIC distribution profile, pharmacokinetic and pharmacodynamic parameters, and the relationship between in vitro activity and clinical outcome in six phase III studies. A clinical success of 65% (binomial range, 55 to 74%) was predicted with a breakpoint of ≤1 μg/ml. We are concerned that a 65% target for successful therapy is too modest, especially with the overall better results seen with the echinocandins and liposomal amphotericin B in large, recently completed studies (75 to 80% response rates) (7, 9, 12). Included in the phase III clinical studies were 1,630 cases of infection caused by the following species: Candida albicans, 54%; C. tropicalis, 15% (both MIC90s, 0.25 μg/ml); C. parapsilosis, 11% (MIC90, 0.06 μg/ml); C. krusei, 3% (MIC90, 1 μg/ml); and finally, C. glabrata, 17% (MIC90, 4 μg/ml). Thus, the vast majority of cases caused by isolates defined as susceptible were cases of C. albicans, C. tropicalis, and C. parapsilosis infection, with MICs 4 to 5 logs below the suggested breakpoint, and those not fully susceptible were isolates from C. glabrata and, to a minor extent, C. krusei infections. As the authors point out, the infecting species was also important in determining the outcome, along with the MIC. This raises the question about whether different breakpoints are needed for different species. C. glabrata, which was an uncommon pathogen before the introduction of fluconazole, lacks the capability to form pseudohyphae and has in some animal experiments been shown to be less virulent than C. albicans and C. tropicalis (1). In later clinical studies, C. glabrata has been associated with a higher mortality than other Candida species (4, 14); however, the possibility cannot be ruled out that a higher age and the fact that many patients initially receive fluconazole, which has inferior activity against C. glabrata, and not the greater virulence of the organism itself, could explain this higher mortality (2, 6, 10, 13). We therefore question if a non-species-dependent breakpoint for voriconazole susceptibility of ≤1 μg/ml is appropriate. We certainly feel that a resistance breakpoint of ≥4 μg/ml is too high, as clinical response rates are inadequate. Until further experience is achieved for such isolates, we recommend the use of microbiological cutoff values for C. albicans and other species for which the voriconazole MICs for wild-type isolates are very low. We are also not comfortable with the terminology “susceptible dose dependent” for voriconazole. This antifungal agent has nonlinear kinetics, and reporting SDD to clinicians invites dose escalation. An argument can be made for the therapeutic drug monitoring of voriconazole in any case, partly because low-dose drug exposure may predict clinical failure and partly because high-dose drug exposure may lead to additional toxicity, as clearly demonstrated with respect to liver function tests, and possibly for other clinical manifestations (3, 5, 8). Simple dose escalation on the basis of an MIC in the SDD range and a lack of clinical response could lead to toxicity, when a switch of therapies would be most appropriate.