Introduction

Stenotrophomonas maltophilia has emerged as an important nosocomial pathogen with high morbidity and mortality, reflecting the increases in antimicrobial agent use and the immunocompromised population including cancer patients [1,2,3]. S. maltophilia was reported to be one of the most common multi-drug-resistant (MDR) gram-negative bacilli isolated from respiratory specimens, and it has a unique pattern of antimicrobial susceptibility originating from its intrinsic resistance against aminoglycoside and beta-lactams [1,2,3,4]. Trimethoprim/sulfamethoxazole (TMP/SMX) is the drug of choice for S. maltophilia infection, although levofloxacin may be considered an alternative drug which shows non-inferiority [5, 6]. Its unique antimicrobial susceptibility patterns compared to other gram-negative bacilli make it difficult to administer the appropriate antimicrobial agents for an S. maltophilia infection. According to previous studies, crude mortality rates of S. maltophilia bacteremia are high, ranging from 14 to 69% depending on the patient population [1]. Furthermore, when limited to patients with hematologic malignancy, mortality rates have been reported 32 and 64.5%, respectively [7, 8].

Several previous studies were conducted to differentiate S. maltophilia bacteremia from other gram-negative bacteremia (GNB) with the aim of administering proper antibiotics [8,9,10,11,12,13]. However, the numbers of enrolled cases with S. maltophilia bacteremia were relatively small (from 12 to 54 cases), and they compared S. maltophilia with only restricted population, such as patients with non-fermentative GNB (e.g., Pseudomonas and/or Acinetobacter) [8, 12, 13], Escherichia coli bacteremia [9, 11], or non-bacteremia [10, 11]. This may be quite different from the real clinical situation, given that physicians are usually first informed of the results of gram-staining. In addition, few studies have focused on clinical predictors of S. maltophilia bacteremia in adult patients with hematologic malignancy, which are likely to be the most vulnerable patient group. Thus, we performed a case-control study to identify clinical predictors of S. maltophilia bacteremia in adult patients with hematologic malignancy compared with other GNB.

Methods

Study population

A case-control study was designed to identify clinical predictors of S. maltophilia bacteremia among hematologic patients with suspected GNB. The electronic medical records were reviewed for individuals diagnosed with S. maltophilia bacteremia during the period of January 2006 through December 2016 at Samsung Medical Center, Seoul, South Korea. This is a 1950-bed tertiary-care university hospital and referral center combined with the 700-bed Samsung Comprehensive Cancer Center. Cases were defined as S. maltophilia bacteremia in adult patients with hematologic malignancy. Bacteremia was defined as either the isolation of S. maltophilia from more than two separate blood samples or the isolation of S. maltophilia from a single blood sample in patients with clinical symptoms and a concomitant focus of infection [14]. Each patient was included only once during the study period. The control group consisted of patients with hematologic malignancy that experienced other GNB caused by frequently isolated gram-negative bacilli as a bloodstream infection in prior study in South Korea: Escherichia coli, Klebsiella species, Acinetobacter species, Pseudomonas species, and Enterobacter species [15]. We matched the cases with the controls by location (whether events occurred in the general ward or in the intensive care unit) and date (GNB other than S. maltophilia bacteremia that occurred at the date closest to the case).

Variables and definition of clinical information

We collected the following data from electronic medical records: age, gender, length of hospital stay, polymicrobial infection, previously isolated pathogens, breakthrough infection during carbapenem therapy, underlying diseases, history of hematopoietic stem cell transplantation, Charlson co-morbidity index (CCI) [16], neutropenia, administration of chemotherapy, use of mechanical ventilation, use of central venous catheter, history of renal replacement therapy, history of operation, Sequential Organ Failure Assessment (SOFA) score [17], previous antibiotic use, and appropriate empirical antibiotic use. Polymicrobial infection was defined as the isolation of two or more bacterial or fungal pathogens in a blood culture sample collected within 72 h. Breakthrough infection during carbapenem therapy was defined as new-onset gram-negative bacteremia after receiving more than 48 h of administration of carbapenem (e.g., imipenem, meropenem, and doripenem), and some of the data included in this study have already been described in our previous study [18]. Every antibiotic administered for more than 48 h within a 30-day period via an intravenous route was recorded as previous antibiotic use. If S. maltophilia or an identical causative pathogen was isolated from a patients’ clinical specimen (including contaminated blood culture) within 30 days before the onset of bacteremia, it was recorded as a previous isolation of S. maltophilia or previous isolation of an identical pathogen. Empirical antimicrobial agents were considered appropriate, if administered within 72 h after obtaining the blood culture sample, and if they were appropriate compared to the in vitro susceptibility test results [12].

Microbiological tests

All blood samples were taken from peripheral veins and/or a central line. A Bactec-9240 system (Becton Dickinson, Sparks, MD) or a BacT/Alert 3D system (bioMérieux Inc., Marcy l’Etoile, France) was used for blood cultures. A Vitek II automated system (bioMérieux Inc.) was used to identify microbes and their antimicrobial agent sensitivity, with a standard identification card and the modified broth microdilution method.

Statistical analysis

All statistical analyses were performed using SPSS 23.0 for Window (IBM Corp., 2015). To determine the predictive factors for S. maltophilia bacteremia, a Student’s t test or Mann-Whitney test was used to compare continuous variables and the Chi-square test or Fisher’s exact test was used to compare categorical variables. We used a logistic regression model to control for confounding variables and to identify clinical predictors of S. maltophilia bacteremia. Variables with P < 0.1 in the univariate analysis were candidates for multivariable analysis and were included in the forward stepwise logistic regression model. All P values were two-tailed and P values < 0.05 were considered to be statistically significant.

Results

Study population

During the study period, we identified 118 cases of S. maltophilia bacteremia and matched them to 118 control patients with other GNB. All cases were hospital-acquired or healthcare-associated infections. The control group consisted of 86 cases of bacteremia with Enterobacteriaceae and 32 cases of bacteremia with non-fermenters. Forty and 13 cases of polymicrobial infection were included in the S. maltophilia group and the control group, respectively. E. coli was most common in the control group (n = 47), followed by Klebsiella species (n = 40), Pseudomonas (n = 17), Acinetobacter (n = 15), and Enterobacter (n = 7). The distribution of frequency of each species was similar to that of the reported frequencies of GNB isolated in South Korea [6]. When we compared appropriate empirical therapy, only 31 (26.3%) of the cases with S. maltophilia bacteremia received appropriate antibiotics, while in the control group, 108 (87.3%) received appropriate antimicrobial therapy (P < 0.001). As for the antimicrobial susceptibility results of isolates in the study, 96.6% of S. maltophilia isolates were susceptible to TMP/SMX and 83.1% were susceptible to levofloxacin. 78.8% of GNB in the controls were susceptible to carbapenem; however, only 36.1% of non-fermenters in the controls were susceptible to carbapenem. The overall 14-day mortality rates of the case and control groups were 54.2 and 25.4%, respectively (P < 0.001), and the overall 30-day mortality rates were 61.0 and 32.2%, respectively (P < 0.001), demonstrating that S. maltophilia bacteremia resulted in significantly higher mortality than other GNB. When we reviewed the sources of bacteremia, pneumonia was the most common focus of bacteremia in the case group (55/118, 46.6%), while central line-associated bloodstream infection was the most common source of bacteremia in the control group (62/118, 52.5%). There were no cases of urinary tract infection (UTI) in the S. maltophilia bacteremia group, while 13 episodes (11.0%) with UTI occurred in the control group.

Clinical predictors for S. maltophilia bacteremia

Risk factors for the development of S. maltophilia bacteremia in adult patients with hematologic malignancy were analyzed and are summarized in Table 1. The univariate analysis showed a significant association of S. maltophilia bacteremia with length of hospital stay ≥ 30 days, polymicrobial infection, previous S. maltophilia or identical pathogen isolation within 30 days, breakthrough bacteremia during carbapenem therapy, UTI, pneumonia, leukemia, history of allogenic stem cell transplantation, use of mechanical ventilation, SOFA score ≥ 7, platelet ≤ 10,000/mm3, previous receipt of antimicrobial agents, and the number of different antibiotics previously used ≥ 3 within 30 days (Table 1). In the multivariable analysis, polymicrobial infection (P = 0.002), previous S. maltophilia isolation within 30 days (P < 0.001), breakthrough bacteremia during carbapenem therapy (P < 0.001), and the number of previously used antibiotics ≥ 3 within 30 days (P = 0.002) were significantly associated with S. maltophilia bacteremia. In contrast, previous use of TMP/SMX showed a negative association with S. maltophilia bacteremia (P = 0.002) (Table 2).

Table 1 Clinical characteristics of Stenotrophomonas maltophilia bacteremia in patients with hematologic malignancy compared to gram-negative bacteremia other than Stenotrophomonas maltophilia
Full size table
Table 2 Factors associated with Stenotrophomonas maltophilia bacteremia among the gram-negative bacteremia: multivariate analysis
Full size table

In the subgroup analysis divided by fermenter and non-fermenter, two of five variables that showed an independent association with S. maltophilia bacteremia had inconsistencies of statistical significance between fermenters and non-fermenters: breakthrough bacteremia during carbapenem therapy (P < 0.001 in fermenter, P = 0.224 in non-fermenter) and previous TMP/SMX use (P = 0.003 in fermenter, P = 0.122 in non-fermenter). The other variables showed statistical consistencies of association with S. maltophilia bacteremia (Table 3).

Table 3 Factors associated with Stenotrophomonas maltophilia bacteremia compared with gram-negative bacteremia divided by fermenter and non-fermenter group
Full size table

Discussion

The present study identified potential risk factors for S. maltophilia bacteremia among GNB in adult patients with hematologic malignancy and revealed the high mortality from S. maltophilia bacteremia in these patients. In our review of previous studies, the mortality rates from S. maltophilia bacteremia in patients with hematologic malignancy were relatively high and have not changed substantially over 20 years [7, 8]. This finding is consistent with our finding that the overall 30-day mortality rate was 61.0% (72/118), and among these deaths, 88.9% (64/72) of patients died within 14 days. These findings suggest that early administration of appropriate empirical antibiotics for S. maltophilia bacteremia should be emphasized. Administration of appropriate empirical antibiotics improved the survival rate of patients with S. maltophilia bacteremia [8, 19], while inappropriate antimicrobial therapy was associated with a poor prognosis in previous studies [20, 21].

Several predictive factors for S. maltophilia bacteremia have been reported in previous studies [8,9,10,11,12,13] and summarized in Table 4. Previous use of carbapenem was the most frequently reported predictive factor for patient with S. maltophilia bacteremia and its association was also demonstrated in our study. However, breakthrough bacteremia during carbapenem therapy showed a higher odds ratio (OR) than previous carbapenem use within 30 days in our study (breakthrough vs previous use, [OR] 10.36 vs 9.56). One prior study reported that S. maltophilia was the most common pathogen of breakthrough bacteremia during carbapenem therapy, followed by Acinetobacter and Pseudomonas [18]. This finding may explain why the previous use of carbapenem was not associated with S. maltophilia bacteremia when compared to non-fermenters, given the high rate of carbapenem resistance of non-fermenters in general and in our study [22].

Table 4 Summary of prior studies comparing cases of SM bacteremia with other controls
Full size table

Among the variables associated with S. maltophilia bacteremia, previous S. maltophilia isolation within 30 days showed the highest association with S. maltophilia bacteremia (OR, 26.25; 95% CI, 6.67–104.83). The previous study by Go Hotta et al. reported that previous S. maltophilia isolation was related to S. maltophilia bacteremia when compared to a group with Pseudomonas bacteremia or a group with Acinetobacter bacteremia [12]. Interestingly, in a subgroup analysis comparing previous isolation of identical causative pathogens, both patients with S. maltophilia and the non-fermenter group showed frequent isolation of the same pathogen preceding bacteremia. This finding is consistent with studies on the association between previous colonization with a non-fermenter or a MDR gram-negative bacillus and bloodstream infection [18, 23, 24].

We also demonstrated two unique factors associated with S. maltophilia bacteremia in patients with hematologic malignancy. One was polymicrobial infection which showed a positive association with S. maltophilia bacteremia, and the other was the previous use of TMP/SMX which showed a negative association with S. maltophilia bacteremia. High rates of polymicrobial bacteremia, ranging from 19.8 to 37.7%, have been reported in patients with S. maltophilia bacteremia [5, 7, 12, 13, 25]. However, previous studies excluded cases of polymicrobial infection or excluded the variable “polymicrobial infection” when they analyzed the predictive factors of S. maltophilia bacteremia. Given that polymicrobial infection might be a unique feature of S. maltophilia bacteremia, especially in patients with hematologic malignancy, we classified polymicrobial infection with both S. maltophilia bacteremia and other GNB as the case group. As S. maltophilia bacteremia accounts for only a small portion of GNB compared to other gram-negative bacilli [15], the inclusion of polymicrobial infection with S. maltophilia bacteremia would not distort the real clinical situation.

TMP/SMX is a drug of choice for serious S. maltophilia infection. Both prophylactic and therapeutic doses of TMP/SMX administration within 30 days showed negative association with S. maltophilia bacteremia, especially when compared with the fermenter group. Colonization might be an important step to develop S. maltophilia bacteremia in patients with hematologic malignancy and TMP/SMX may suppress both S. maltophilia and fermenters. However, fermenters might be able to develop bacteremia easily, because of their characteristics that need not be colonized before developing bacteremia. Compared with non-fermenters, there was no statistical significance even though S. maltophilia showed negative association to TMP/SMX.

The risk factors presented in this study differed from the traditional risk factors identified in previous studies. Although neutropenia and mechanical ventilation have been suggested in previous studies, one study reported that infection with S. maltophilia also occurs in cancer patients without those traditional risk factors [26].

There are several limitations in our study. First, this study was conducted in a single medical institute and collected data retrospectively. Some of the medical records in charts may have not been complete, even though we tried to enroll patients who had more concrete information. This study was performed in a tertiary-care hospital, wherein conditions, patient populations, and outcomes may differ from those in non-tertiary-care centers. Second, the association of carbapenem and TMP/SMX with S. maltophilia bacteremia can be changed by local resistance rate of these agents. If carbapenem resistance rate of GNB is high, positive association between S. maltophilia bacteremia and use of carbapenem might be decreased. Similarly, if TMP/SMX resistance rate of SM isolates is high, negative association between S. maltophilia and TMP/SMX might be decreased. Finally, the unique epidemiological composition of microorganisms in specific local area might also affect predictive factor of S. maltophilia bacteremia. Hence, our results may not apply directly in other institutions or countries. Despite these limitations, the findings of this investigation may improve the ability of clinicians to identify hematologic patients who are at high risk for S. maltophilia bacteremia.

In conclusion, our data suggest that S. maltophilia bacteremia in adult patients with hematologic malignancy has become a severe and urgent problem. Its mortality is still high and early administration of appropriate antimicrobial agents is still challenging. Several clinical factors associated with S. maltophilia bacteremia identified in our study might be useful to predict cases of S. maltophilia bacteremia among suspected GNB in patients with hematologic malignancy. Further study may be warranted for external validation of these clinical factors.