Antibiotic exposure as a risk factor for fluconazole-resistant Candida bloodstream infection

Abstract

Recent exposure to azoles is an important risk factor for infection with fluconazole-resistant Candida spp., but little is known about the role of antibacterial drug exposure in the emergence of drug-resistant Candida. We did a prospective nationwide surveillance study of candidemia in Israel and analyzed the propensity score-adjusted association between antifungal and antibacterial drug exposure and bloodstream infection with C. glabrata and fluconazole-resistant Candida isolates. Four hundred forty-four episodes of candidemia (450 Candida isolates, 69 [15%] C. glabrata isolates, and 38 [8.5%] fluconazole-resistant isolates) from 18 medical centers in Israel were included. C. glabrata bloodstream infection was strongly associated with recent metronidazole exposure (odds ratio [OR], 3.2; P < 0.001). Infection with a fluconazole-resistant isolate was associated with exposure to carbapenems, trimethoprim-sulfamethoxazole, clindamycin, and colistin (odds ratio, 2.8; P = 0.01). The inclusion of antibacterial drug exposure in a multivariable model significantly enhanced the model's predictive accuracy for fluconazole-resistant Candida bloodstream infection. Our findings may be relevant to the selection of empirical antifungal treatment and broaden the scope of antibiotic-associated collateral damage.

Fig 1

Association of antibiotic and nonantibiotic covariates with C. glabrata, and fluconazole-resistant Candida bloodstream infection. The forest plots show the associations of antibiotic and nonantibiotic covariates with C. glabrata BSI (A) and FRC BSI (B). The individual graphs show significantly associated covariates by bivariable analysis and multivariable analysis. The plots show the odds ratio (symbol) and 95% confidence interval (whiskers) for each covariate. Solid symbols are used for nonantibacterial covariates and open symbols for antibacterial covariates. All covariates refer to exposure within 30 days prior to the onset of candidemia. CVC, central venous catheter. Antibacterial drug exposure denotes exposure to carbapenems, trimethoprim-sulfamethoxazole, colistin, or clindamycin.

Fig 2

Comparative accuracy of predictive models for fluconazole-resistant Candida sp. bloodstream infection. The predictive accuracy, as represented by the area under the ROC plot, is shown for 3 models. FRC BSI is the dependent variable for all models. Covariates for model 1 were neutropenia and exposure to fluconazole; those for model 2 were neutropenia and exposure to antibacterial drugs (carbapenems, trimethoprim-sulfamethoxazole, clindamycin, or colistin); those for model 3 were neutropenia, exposure to fluconazole, and exposure to antibacterials. The area under the ROC curve was significantly higher for models that included antibacterial covariates (2 and 3) than for model 1 (P = 0.003).