PNA-FISH has been shown to be a useful and accurate method in the identification of Gram positive, Gram negative, and Candida sp. from blood cultures and peritoneal fluid cultures. When analyzing PNA-FISH probe performance for both bacteria and yeast combined in both sample types, we demonstrated an overall identification accuracy of 98.9% (96/97). Accuracy of PNA-FISH for bacteria was 100% (80/80), accuracy for Candida sp. was 94.1% (16/17). This high level of accuracy with PNA-FISH has been replicated in other studies [6–10, 12, 15]. We did not test identification accuracy in BACTEC-negative bottles, as PNA-FISH requires an organism concentration of at least 105 CFU/mL for detection.
Clinical isolates identified in this study are representative of the most commonly reported pathogens implicated in nosocomial bloodstream infections in the United States . In blood cultures, accuracy of PNA-FISH for bacteria was 100%. In this study, accuracy for yeast identification was 90.9% (10/11). A single weakly reactive C. albicans isolate may have been due to a mismatch between organism rRNA and PNA-FISH probe secondary to point mutations in sequence, as have been described by other authors [11, 12].
Mean time saved using PNA-FISH for blood cultures compared to traditional culture methods averaged 72.4 hours from the time the organism (bacteria or yeast) was detected by Gram stain from a positive blood culture bottle until final identification of species [Table 1]. The significant difference in time of C-ID as compared to F-ID may partially be attributed to conventional laboratory workflow with C-ID; after subculture of a positive blood culture bottle, plate cultures were assessed once daily for adequate growth. Panels used for automated bacterial identification were set up once daily if plate culture growth was sufficient followed by 24-hour incubation. An advantage of the PNA-FISH methodology is ease of use, in that multiple final identifications may be generated in a single laboratory shift directly from positive blood culture bottles, without need for additional incubation.
The use of blood culture media is a useful method of enhancing recovery of microorganisms from peritoneal fluids [16–18]. Although not specifically FDA-approved for this indication, we also assessed performance of PNA-FISH using peritoneal fluid specimens incubated in blood culture bottles, as proof of concept. To the best of our knowledge this study is the first application of this PNA-FISH method for this sample type. Identification accuracy for both bacteria and Candida sp. was 100%. Although mean time saved with PNA-FISH was not significant for Enterococcus sp., this was likely attributable to low sample numbers (Table 1).
Laboratory-spiked bottles were not used for turnaround time determination for Candida sp. in peritoneal fluids, as the intent of the study was to mimic actual clinical performance of PNA-FISH from blood culture media as much as possible. Although the turnaround time for Candida sp. could not be specifically determined due to the use of spiked bottles, based on blood culture data, it is reasonable to assume a similar reduction in time to F-ID in peritoneal fluid samples would also apply. Low numbers of peritoneal fluids tested is a limitation of this study. Although blood culture bottles containing peritoneal fluid spiked with Candida sp. were used to increase numbers to statistically meaningful levels, every attempt was made to apply PNA-FISH to situations found in actual clinical practice. Excellent performance of PNA-FISH with this sample type is encouraging for application to larger scale studies required to truly assess assay performance.
Accurate identification of pathogens is a primary driver of antimicrobial or antifungal selection. Delays in appropriate therapy clearly affect patient outcomes in a negative fashion. Among 492 intensive care patients studied by Ibrahim et al., 30% received inadequate antimicrobial therapy for bacteremia; hospital mortality rate was 62% compared to 28.4% of patients receiving appropriate antibiotics . In terms of yeast identification, both SENTRY and EIEIO sentinel surveillance programs demonstrated a shift in pathogenic Candida species over the last decade . These findings support the importance of rapid identification of yeast to species level as critical to direct appropriate antifungal therapy. Excellent performance by PNA-FISH in both bacteria and yeast will address these clinical needs.
In this study, based on a limited chart review of clinical cases, the more expensive antifungal caspofungin was used empirically for 5–6 days in 2 cases in which the organism was C. albicans or C. parapsilosis fully susceptible to fluconazole. Based on our estimated antifungal cost savings, and 2009 AWP data, this could potentially result in pharmaceutical cost savings of $1,888.70 over 5 days of therapy. Significant pharmaceutical cost savings due to de-escalation of therapy from an echinocandin to fluconazole in infections caused by fluconazole –susceptible C. glabrata have also been noted in other studies .
We also examined the costs associated with the use of vancomycin for CoNS infection. In four cases in which chart review was available, empiric therapy with vancomycin was continued for 2–5 days and not discontinued until CoNS was identified and deemed a contaminant, based on clinical presentation of the patient. Avoiding unnecessary use of vancomycin in these instances could save an estimated $20.00 daily per patient. Based on our study, and use of PNA-FISH, more appropriate use of vancomycin and caspofungin could reduce hospital pharmaceutical costs. As indicated in other studies, rapid organism identification and judicious use of antibiotics has overall broader implications in antibiotic stewardship and may positively affect reduction of antibiotic resistance and patient mortality [6, 8, 21].
Based upon chart review, a disparity was noted between time of C-ID or F-ID, and the time of therapy initiation until a change in therapy was made based on ID results (Table 2). We cannot specifically comment why, in some cases, antifungal or antibiotic therapy was not changed after organism ID was finalized. This obviously affects overall estimates of pharmaceutical cost. Based on traditional practice patterns, it can only be assumed that if patients are improving under a given course of therapy, clinicians are less likely to change therapy, even with the availability of ID results. Proactive management of identification results by stakeholders such as Infectious Disease or Pharmacy, as part of an active antibiotic stewardship program will contribute to the downstream benefits of rapid diagnosis by PNA-FISH.
A limitation of PNA-FISH is a requirement of an organism concentration of at least 105 CFU/mL for detection. This requirement may prove to be problematic for detection of slow-growing, or fastidious organisms. At the time this study was performed, a limited number of PNA-FISH probes were available. However, additional specific PNA-FISH probes are now available for Group B Streptococcus , GNR Traffic Light (FDA-approved), C. dubliniensis, C. parapsilosis, K. pneumoniae, and Acinetobacter (analyte-specific reagents). Recent FDA approval of a more rapid PNA-FISH protocol should prove to be advantageous in further decreasing turnaround time to organism identification .