Comparison of species-level identification and antifungal susceptibility results from diagnostic and reference laboratories for bloodstream Candida surveillance isolates, South Africa, 2009–2010

From February 2009 through August 2010, we compared species-level identification of bloodstream Candida isolates and susceptibility to fluconazole, voriconazole, and caspofungin between diagnostic and reference South African laboratories during national surveillance for candidemia. Diagnostic laboratories identified isolates to genus/species level and performed antifungal susceptibility testing, as indicated. At a reference laboratory, viable Candida isolates were identified to species-level using automated systems, biochemical tests, or DNA sequencing; broth dilution susceptibility testing was performed. Categorical agreement (CA) was calculated for susceptibility results of isolates with concordant species identification. Overall, 2172 incident cases were detected, 773 (36%) by surveillance audit. The Vitek 2 YST system (bioMérieux Inc, Marcy l'Etoile, France) was used for identification (360/863, 42%) and susceptibility testing (198/473, 42%) of a large proportion of isolates. For the five most common species (n = 1181), species-level identification was identical in the majority of cases (Candida albicans: 98% (507/517); Candida parapsilosis: 92% (450/488); Candida glabrata: 89% (89/100); Candida tropicalis: 91% (49/54), and Candida krusei: 86% (19/22)). However, diagnostic laboratories were significantly less likely to correctly identify Candida species other than C. albicans versus C. albicans (607/664, 91% vs. 507/517, 98%; P < .001). Susceptibility data were compared for isolates belonging to the five most common species and fluconazole, voriconazole, and caspofungin in 860, 580, and 99 cases, respectively. Diagnostic laboratories significantly under-reported fluconazole resistance in C. parapsilosis (225/393, 57% vs. 239/393, 61%; P < .001) but over-reported fluconazole non-susceptibility in C. albicans (36/362, 10% vs. 3/362, 0.8%; P < .001). Diagnostic laboratories were less likely to correctly identify Candida species other than C. albicans, under-reported fluconazole resistance for C. parapsilosis and over-reported fluconazole resistance for C. albicans.