Should we treat patients with only one set of positive blood cultures for extensively drug-resistant Acinetobacter baumannii the same as multiple sets?

Acinetobacter species are not considered skin commensals and under-treatment is an overriding concern when caring for critically-ill patients who are mostly at risk of extensively drug-resistant Acinetobacter baumannii (XDRAB) infections. Hence even a single blood culture yielding XDRAB will tend to prompt intervention. However, field observations suggest that patients with single-positive blood cultures had milder disease and were more likely to be recruited in interventional studies than those with multiple-positive blood cultures, yet no distinction is made in current clinical or trial recruitment practices. To our knowledge, this is the first study to compare the clinical characteristics and outcomes of patients with single-positive versus multiple-positive blood cultures for XDRAB. In this multicenter prospective cohort study of XDRAB bacteremic patients from July 2010 to June 2015, only patients with at least two simultaneously drawn blood cultures were included. The patients were classified as having single-positive or multiple-positive blood cultures according to the number of positive blood cultures yielding XDRAB. The primary end-point was the 28-day mortality. Of a total of 155 patients enrolled, 69 had a single-positive and 86 had multiple-positive blood cultures. Leukopenia (37.2% vs. 16.2%; P = 0.004), thrombocytopenia (56.0% vs. 26.5%; P < 0.001), higher Pitt bacteremia scores (6.6 vs. 5.5, P = 0.03) and higher 28-day mortality rates (70.9% vs. 43.5%; P = 0.001) distinguished patients with multiple-positive from those with single-positive cultures. Multivariate logistic regression showed that multi-positivity independently predicted 28-day mortality (adjusted odds ratio, 2.34; 95% confidence interval (CI), 1.03–5.28; P = 0.04) and the Cox regression confirmed that multi-positivity (adjusted hazard ratio, 1.80; 95% CI, 1.13–2.85; P = 0.01) predicted rapid mortality. Patients with multiple versus single positive blood cultures yielding XDRAB had greater morbidity and mortality. Investigators and clinicians should be aware that the blood culture positivity rate impacts outcomes of XDRAB bacteremia.


Introduction
Extensively drug-resistant Acinetobacter baumannii (XDRAB) bacteremia usually occurs in severely ill patients in the intensive care unit with limited therapeutic options [1,2]. The associated crude and attributable mortality rates are high, ranging from 41% to 69% and 10% to 43%, respectively [3][4][5][6][7]. Several complicated risk factors relating to initial disease severity, antimicrobial resistance, host factors, and ineffective antimicrobial therapy have been identified [7][8][9][10][11][12][13][14][15]. However, few studies have focused on the impact of bacterial factors. Currently identified factors such as the Acinetobacter genospecies and serum OXA-51 DNA load require molecular tools too sophisticated for daily clinical practice [16,17]. In contrast, the blood culture positivity rate is an easy observation that has proven to be a semi-quantitative measure of the magnitude of Staphylococcus aureus bacteremia and to predict mortality in such patients [18]. However, the prognostic validity of the blood culture positivity rate is not universally applicable, as demonstrated by the similar clinical outcomes of patients with single versus multiple-positive blood cultures of enterococci or coagulase-negative staphylococci (CoNS) [19,20].
For XDRAB bacteremia, data on the impact of the blood culture positivity rate on outcomes are lacking. Furthermore, the significance of a single positive blood culture with XDRAB is not known [21]. While physicians may clinically suspect some patients to have pseudobacteremia if only one of at least two sets of blood cultures is positive for XDRAB, the lack of diagnostic criteria that has been validated in the setting of XDRAB, and the overriding concerns of under treatment when caring for high-risk patients, makes withholding treatment difficult. In clinical practice, this may result in unnecessary therapy, further selection of resistant pathogens, longer hospitalization and increased costs. In clinical research, the lack of distinction between contamination and true bacteremia may contribute to misleading mortality rates and inferred treatment responses.
For these reasons, we compared the clinical characteristics and outcomes of patients with a single versus multiple sets of blood cultures yielding XDRAB in this prospective multicenter cohort study. We hypothesized that the mortality associated with XDRAB bacteremia is not only impacted by the patients' characteristics, the primary infection foci, disease severity, antimicrobial susceptibilities, appropriate therapy, but also by the culture positivity rate. Our aims were to explore the clinical significance of a single-positive blood culture for XDRAB and to establish whether the total number of positive blood cultures could serve as a simple predictor of mortality for patients with XDRAB bacteremia.

Hospital settings
This study was conducted at three hospitals in Taiwan: the National Taiwan University Hospital (NTUH), a 2200-bed medical center located in Taipei; the Far Eastern Memorial Hospital, a 1000-bed medical center in New Taipei City and the NTUH Yun-lin Branch, a 600-bed teaching hospital in Dou-liou. The Research Ethics Committees of the National Taiwan University Hospital Main Branch, Taipei, the National Taiwan University Hospital Yun-Lin Branch, Dou-Liou (201008047R) and the Far Eastern Memorial Hospital, New Taipei (100132-F) approved this study. Informed consent was waived for this observational study. except for colistin and tigecycline were included. Patients who were not admitted to hospital, patients who were less than 18 years old, or had only one set of blood cultures taken were excluded. One set of blood culture comprised an aerobic and an anaerobic bottle. Single-positive blood culture was defined as patients with multiple sets of blood culture drawn simultaneously, but only one set grew A. baumannii. Multiple-positive blood cultures were defined as 2 or more sets of blood culture yielding A. baumannii among the multiple simultaneous sets of blood culture.
Demographic data, underlying diseases, the site of infection, severity of illness, treatment, and outcomes were recorded. Prior antimicrobial use was defined by the use of the parenteral form of the antimicrobial agent in the 28 days preceding bacteremia. The sites of primary infection were identified according to definitions of the US Centers for Disease Control and Prevention (CDC) [22]. If no focus could be identified, the bacteremia was classified as primary bacteremia. The Charlson comorbidity index was used to adjust for underlying conditions [23]. Bacteremia severity was assessed using the Pitt bacteremia score at bacteremia onset [24]. Use of immunosuppressive agents was defined as the receipt of antineoplastic drugs or other immunosuppressive agents within 6 weeks or receipt of corticosteroids at a dosage !20 mg of prednisolone daily for !2 weeks or 30 mg of prednisolone daily for !1 week before bacteremia onset [16]. Early mortality was defined as death that occurred before or on the day of culture report. The primary outcome was the 28-day in-hospital mortality.

Microbiological studies
Blood cultures were processed by the clinical microbiology laboratory. The A. baumannii complex was identified by the Vitek 2 identification system (bioMérieux Inc.). The antimicrobial susceptibility were determined using Vitek 2 identification system (bioMérieux Inc.) and were interpreted according to the CLSI criteria [25] After collecting the isolates, the genospecies were identified according to the 16S-23S ribosomal RNA (rRNA) gene intergenic spacer region, as described [26]. Patients with bacteremia caused by A. nosocomialis or A. pittii were excluded [16]. Only the genospecies A. baumannii was included.

Statistical analysis
Mean and standard deviation (SD) were calculated for continuous variables and percentages for categorical variables. The Student's t test and Fisher's exact test were used to compare continuous and categorical variables, respectively. Multivariate logistic regression and Cox's proportional-hazards models were applied for outcome analysis. All variables with a P value 0.2 in the univariate regression were considered in the multivariate analysis. Multivariable models were developed by minimizing Akaike's information criterion [27]. A sensitivity analysis excluding polymicrobial-infected patients was conducted. Analysis was performed using Stata software (version 14, StataCorp, College Station, TX). Two-sided P values 0.05 were considered significant.

Results
A total of 155 patients were enrolled in the study period (July 2010-June 2015) (Fig 1). The mean (SD) age of the study cohort was 66.9 (17.4) years, the Pitt bacteremia score was 6.1 (3.3) points, the Charlson comorbidity index was 3.9 (3.0) points, and hospitalization duration before A. baumannii bacteremia onset was 31.7 (30.1) days. One hundred and four (67.1%) patients were male, 45 (29.0%) patients had underlying malignancy, and 30 (19.4%) patients received immunosuppressive therapy. The all-cause in-hospital mortality was 71.6%.
Among the 86 patients who had multiple-positive blood cultures, 12 had polymicrobial infections with the following co-pathogens (n): enterococci (8)

Univariate analysis of factors associated with 28-day mortality
The multiple-positive group compared with the single-positive group had higher 28-day mortality by univariate logistic regression (odds ratio (OR) 3.17; P = 0.001) and by the Kaplan-Meier survival analysis (hazard ratio (HR) 2.29; log-rank P < 0.001). There were no major differences in the characteristics of patients who died within 28 days of bacteremia and those who did not, except for underlying chronic lung disease (OR 4.01; P = 0.03), prior carbapenem use

Multivariate analysis of factors associated with 28-day and early mortality
We adjusted for the above covariates and also those variables in the univariate analysis with a P < 0.2 in the multivariate logistic regression analysis (Table 2). Excess 28-day in-hospital mortality was predicted by multiple-positive blood cultures (adjusted OR (aOR), 2.34; 95% confidence interval (CI), 1.03-5.28; P = 0.04) (

Sensitivity analysis of monomicrobial-infected patients
Excluding those with polymicrobial infections, there were 118 monomicrobial-infected patients included in the sensitivity analysis. Eighteen of the 44 (40.9%) patients with singlepositive culture died within 28-days compared to 50 of 74 (67.6%) with multiple-positive cultures (P = 0.007). Among monomicrobial-infected patients, multiple sets of positive blood

Discussion
To our knowledge, this is the first study to compare the clinical outcomes of patients with single-versus multiple-positive blood cultures of XDRAB. Patients who had multiple sets of positive blood cultures for XDRAB differed significantly from those who had only one set of positive cultures out of multiple samples. Although single-positive bacteremia patients did not have negligible mortality, multiplicity of positive blood cultures independently predicted greater and more rapid mortality. These findings imply that the blood culture positivity rate can be used as a simple prognostic indicator in the context of XDRAB bacteremia and further studies to examine more efficient methods to detect contamination versus true infection in the context of a single-positive blood culture are needed. The differences in mortality between single versus multiple-positive patient groups might be explained by higher bacterial load since semi-quantitative blood cultures have been shown to be a measure of the magnitude of S. aureus bacteremia [18] and Chuang et al. have shown that the bacterial burden measured by real-time PCR is associated with the in-hospital mortality of A. baumannii bacteremia [17]. Compared to the expensive molecular quantification of the latter study, the semi-quantitative blood culture has the advantages of routine accessibility without additional costs or labor and remains the gold standard for the diagnosis of bacteremia [28,29]. However, single-positive blood cultures do not necessarily predict better outcomes compared to multiple-positive blood cultures in the case of coagulase-negative staphylococci (CoNS) and enterococci (Table 3). In addition, the difference between the single versus multiple-positive patient groups might also be explained by the lack of distinction between contamination and true infection.
Despite its limitations, the blood culture remains the "gold standard" for the detection of bacteremia and at least 2 blood cultures should be sampled simultaneously before therapy according to the US Centers for Disease Control and Prevention/National Health Surveillance Network (CDC/NHSN) and Surviving Sepsis Campaign [22,30]. Suggested laboratory criteria for true bacteremia include growth within 48h and multiple cultures positive for the same organism [31][32][33]. In contrast, increased duration before positivity, polymicrobial growth of skin organisms, or growth during treatment suggest contamination [31,32]. In this study, we confirmed that polymicrobial infections were more common among single-positive rather than multiple-positive groups and that the former had less severe disease and mortality compared to the latter. However, the crude mortality rate of the single-positive group in this study was 65% which is much higher than the reported pooled crude mortality rate of 27% for nosocomial bacteremia from a concurrent surveillance study of 49 US hospitals or the reported rates of 19-27% and 12.5-17.5% for bacteremia with single-positive blood cultures for CoNS [20,34,35] and enterococci, respectively (Table 3) [19,36]. In addition, the mean Pitt score in our single-positive group was high (5.5) making it unlikely that the majority of single-positive patients had pseudobacteremia and thus, better outcomes. Unlike CoNS and enterococci, A. baumannii is not considered a skin contaminant [22]. Thus, growth of XDRAB from a single blood culture will tend to prompt treatment. However, given secular changes in the standards of care, it is plausible that A. baumannii could occasionally contaminate cultures drawn from central lines in contemporary ICUs. In Weinstein's landmark studies of 500 episodes, later expanded to 843 episodes of positive blood cultures in hospital inpatients published in 1983 and 1997, respectively [21,37], upon which current understanding of blood culture results that are incorporated in widely used CDC/NHSN definitions for laboratory-confirmed bloodstream infections are based, species thought to represent true infections when isolated from a blood culture included S. aureus, pneumococcus, E. coli and other Enterobacteriaceae, P. aeruginosa, and C. albicans, whilst species thought to represent contamination in a significant proportion of cases, included CoNS, Corynebacterium species, Bacillus species, Propionibacterium acnes, Micrococcus species, viridans group streptococci, enterococci, and C. perfringens. Yet at the time, Impact of culture positivity rate on the characteristics and outcomes of XDRAB bacteremia Acinetobacter species were rarely isolated, being only identified in 0-1% (16 of 1844 blood isolates) in Weinstein's studies, in 0-1% (4 of 348 blood isolates) among male veterans, and 1.3% (67 of 5058 blood isolates) among North American 1997 SENTRY participants [38][39][40]. Since then, the isolation rate of Acinetobacter species from blood cultures has increased significantly, particularly in the ICU setting, however it is not known what proportion represents true bacteremia or contamination. Furthermore, there have been few studies in the last two decades to document the clinical significance of single versus multiple blood cultures for organisms other than CoNS and enterococci (Table 3). To our knowledge, our study is the first to document the prognostic impact of the blood culture positivity rate for gram-negative species. Further studies on whether the culture positivity rate impacts the outcomes of microorganisms not traditionally considered as skin contaminants, to reflect the changing microbiology and epidemiology are warranted.
Although treated equivalently as true bacteremia, we found major differences in the clinical characteristics between patient groups, namely single-positive patients were less likely to have monomicrobial bacteremia, underlying malignancy, pneumonia, bicytopenia, high Pitt scores and fatal disease compared to multiple-positive patients. These findings question the uniformity of treating single-positive patients as multiple-positive patients and of enrolling such patients together in clinical trials.
Currently registered trials for multidrug resistant A. baumannii at ClinicalTrials.gov all employ culture-confirmed randomization without specifying the number of positive blood cultures. Such trials will miss a significant proportion (51.2%) of multiple-positive patients and preferentially enroll those with minor disease or pseudobacteremia (84.1%). Consequently, such trials cannot be relied on to give response and mortality rates that can be replicated in the real world. For example, colistin use was independently associated with survival in this study. While this has been attributed to its effectiveness, it may reflect the fact that those who had already survived to when cultures were reported were patients with mild disease. After excluding the early mortality cases, colistin use was no longer associated with survival (27/71 (38.0%) vs. 9/29 (31%), P = 0.65). Our findings therefore support the importance of empiric randomization for maximizing the likelihood of detection of true superiority of the investigational agent for XDRAB [41].
Given that multiple-positivity was associated with increased mortality that was greatest in the period before culture becomes known and interventions can be initiated, we analyzed whether there were associated factors that could be manipulated to decrease the rates of multiple-positive blood cultures. Considering the modifiable risk factors such as length of hospital stay, central venous catheter and ventilator use, immunosuppression and the cumulative use of antimicrobial agents within 28 days of bacteremia onset, we found that carbapenem exposure independently predicted multiplicity of positive blood cultures for XDRAB (aOR, 2.80; 95% CI, 1.03-7.64; P = 0.045). Therefore, judicious carbapenem use might decrease the rate of multiple-positive blood cultures rates and curb the associated early mortality.
The strengths of the current study include the prospective multicenter design and the use of a hard end-point, mortality, to avoid recall biases. We included only patients who had more than two simultaneous blood cultures to avoid the heterogeneity of previous studies that mixed single-positive out of a total of one set of blood culture sample with single out of multiple sets of blood culture and also avoided the mixing of multiple-positive cultures due to sequential positive cultures (representing persistent bacteremia). We also validated our findings by sensitivity analyses. However, the following limitations should be acknowledged. First, the cohort studied may not be sufficiently large to render non-observed confounders obsolete. Since randomization may be impractical, our findings should be corroborated by a larger study. Second, we excluded a significant portion (214/889) of patients who had only one set of blood cultures collected. Third, we did not compare the bacterial loads of XDRAB in the single and multiple-positive patients by quantitative blood culture or real-time PCR. We also did not measure the bacterial loads of other co-pathogens. Fourth, there was no control group for the patients with a single-positive blood culture. Hence, what proportion of single-positive patients had pseudobacteremia and whether the 28-day mortality rate of 43.5% for this group is in excess of the mortality of critically-ill patients without XDRAB bacteremia remain unanswered.

Conclusions
In conclusion, the number of positive blood cultures is a prognostic indicator of the severity of XDRAB bacteremia. Multiple-compared to single-positive blood cultures for XDRAB is associated with rapidly fatal disease. In order to curb the alarming early mortality rates, prevention of multiple-positive XDRAB bacteremia by carbapenem stewardship might be a better strategy than attempts at curing the infection after culture results become known. Lastly, more research is needed to discern true infection and contamination among patients with single-positive cultures.