Introduction

In patients with febrile neutropenia, Gram-positive bacteria caused <30 % of bacteremia cases in the 1970s, about 50 % in the 1980s, and more than 60 % in the 1990s [1]. This increase in the incidence of Gram-positive bacterial infections has also been observed in children who did not receive fluoroquinolone prophylaxis during cancer treatment [2], and similar results were noted in our hospital [3, 4]. Among these increasing Gram-positive bacterial infections, bacteremia caused by viridans streptococci rather than Staphylococcus spp. has been increasing [3, 5]. In our hospital between 2004 and 2006, 47 % of bacteremia cases were due to Gram-positive bacteria and 5 % were due to viridans streptococci in children with hematology/oncology diseases [4]. In 2009, the proportions increased to 57 and 23 %, respectively [3].

Because viridans streptococcal bacteremia (VSB) has a higher rate of severe complications such as shock, acute respiratory distress syndrome (ARDS), and death compared to other Gram-positive bacteremias, early diagnosis and treatment are critical [5]. The clinical practice guideline for the use of antimicrobial agents in neutropenic patients with cancer by the Infectious Diseases Society of America (IDSA) and the Korean guideline for neutropenic fever recommend the use of glycopeptides when there is a risk of Gram-positive bacterial infection, even in the initial empirical antibiotic treatment [6, 7].

Although viridans streptococci are the most common causative pathogen of bacteremia in febrile neutropenic patients [8] and the antibiotic resistance of viridans streptococci has been increasing [9], few studies since the mid-2000s have reported the clinical characteristics, including severe complications, of VSB or the antibiotic susceptibility of viridans streptococci in children with febrile neutropenia.

We conducted this retrospective study to investigate the clinical characteristics of VSB and its severe complications, as well as to evaluate the antibiotic susceptibility of viridans streptococci in children with febrile neutropenia in the context of increasing antibiotic resistance.

Patients and methods

Patients and study design

We conducted a retrospective study by reviewing the medical records of children with febrile neutropenia and VSB selected from all patients with hematology/oncology diseases admitted to the Department of Pediatrics of Seoul St. Mary’s Hospital, Seoul, Korea, between April 1, 2009 and June 30, 2012. This study was approved by the Institutional Review Board of Seoul St. Mary’s Hospital, with a waiver of informed consent. Piperacillin/tazobactam and isepamicin were given as empirical antibiotics for febrile neutropenic children, and were changed to meropenem and teicoplanin (or vancomycin) if the fever lasted for 3–5 days. If the fever lasted for more than 5–7 days, respiratory symptoms deteriorated, or abnormal findings developed on chest X-ray, empirical antifungal treatment was considered.

Data collection

Clinical characteristics including age, gender, underlying disease, relapse status, post-stem cell transplantation (SCT) status, time after the first day of the preceding chemotherapy, use of high-dose cytarabine (HD-ARA) during the preceding chemotherapy, presence of oral mucositis, respiratory or gastrointestinal symptoms, and peak body temperature on fever day 1 were collected. Furthermore, laboratory characteristics including white blood cell (WBC) count, absolute neutrophil count (ANC), and C-reactive protein (CRP) level on fever day 1, and maximum CRP level during the bacteremia episode were investigated. The clinical and laboratory characteristics of children with and without severe complications of VSB were compared. Blood for culture was sampled using a sterile technique with one set from a peripheral vein and another set from a central catheter if the patient has a central venous catheter. Each 1–3 mL of blood was inoculated into a culture bottle (BD BACTEC™ Peds Plus Culture Vial, Becton–Dickinson, Sparks, MD, USA) and an automated culture system (BACTEC™ FX, Becton–Dickinson, Sparks, MD, USA) was used. Bacterial identification was performed using the VITEK®2 automated system (bioMériux, Marcy l’Etoile, France), and antibiotic susceptibilities to penicillin, cefotaxime, cefepime, erythromycin, clindamycin, vancomycin, linezolid, and quinupristin/dalfopristin were measured using the E-test on a Mueller–Hinton agar plate including 5 % sheep blood. According to the antibiotic susceptibility test results, ‘S’ was defined as susceptible, and ‘I’ and ‘R’ were defined as non-susceptible. Because antibiotic susceptibilities and clinical characteristics are not significantly different among viridans streptococcal species [10, 11], we did not identify the species of viridans streptococci.

Definitions

VSB was defined as the growth of viridans streptococci within 5 days in an automated culture system from at least one peripheral or central blood sample with accompanying symptoms or signs suggesting bacterial infection, such as fever, chills, or shock. Neutropenia was defined as an ANC of lower than 500/μL or an ANC expected to be lower than 500/μL within 2–3 days [7]. Fever was defined as a body temperature of higher than 38.0 °C with a tympanic thermometer or 37.5 °C with an axillary thermometer [7]. Severe complications included ARDS, shock, transfer to the intensive care unit (ICU), and death of all causes. ARDS was defined as PaO2/FiO2 <200 in arterial blood gas analysis performed in children with hypoxia of SpO2 <90 % on pulse oximetry and bilateral pulmonary infiltrates on chest X-ray [12]. Shock was defined as hypotension requiring an intravenous fluid bolus or inotropic agents to maintain normal blood pressure [13]. Death was attributed to VSB when patients showed no clinical improvement after the diagnosis of VSB, and there was no other definite cause of death, such as aggravation of the underlying disease or development of a new disease.

Statistical analysis

Statistical analyses were performed using SPSS Statistics 17.0 (SPSS Inc., Chicago, IL, USA). Fisher’s exact test was used to compare categorical variables between children with severe VSB complications and those without complications, while the Mann–Whitney test was used to compare numerical variables between these two groups. The CRP levels on fever day 1 and maximum CRP levels during the bacteremia episode were compared using a paired t-test, and the correlation between the maximum CRP levels and fever duration was evaluated using Pearson’s correlation analysis. The cut-off value of maximum CRP level predicting severe complications of VSB was determined using a receiver operating characteristic (ROC) curve. Statistical significance was defined as a two-sided p-value <0.05.

Results

During the study period, 57 cases of VSB were diagnosed in 50 children (37 cases in 31 boys and 20 cases in 19 girls). Three children each experienced two episodes of bacteremia and two children each experienced three episodes of bacteremia, and each episode was diagnosed during separate admissions. In two cases, multiple bacterial species were present in blood cultures (viridans streptococci and Corynebacterium spp. in one, viridans streptococci and Escherichia coli in the other). In two other cases, a bacterium other than viridans streptococci was identified in a second blood culture (Enterococcus faecium in one, methicillin-susceptible Staphylococcus aureus in the other), but the subsequent blood cultures were sterile.

Clinical characteristics

The median age of enrolled children was 8.2 years (range 0.6–20.0 years), and 37 children (64.9 %) were male (Table 1). Acute myeloid leukemia was the most common underlying disease (Table 1). Eight children (14.0 %) received SCT. VSB occurred after chemotherapy for relapsed underlying leukemia in six children who had received allogeneic SCT between 4 and 109 months prior. In another child who had received allogeneic SCT 7 months prior, VSB occurred after chemotherapy for lymphoma that developed after SCT. In the remaining child, VSB occurred after chemotherapy for hemophagocytic lymphohistiocytosis that developed 3 years subsequent to autologous SCT. Fourteen children (24.6 %) were in relapse without complete remission (Table 1). Fifty-two children (91.2 %) had a central venous catheter; the exceptions were two children who had had their central venous catheters removed after previous SCT, two children with very severe aplastic anemia, and one child newly diagnosed with acute lymphoblastic leukemia (ALL). Forty-one children (71.9 %) had a Hickman catheter and 11 children (19.3 %) had a subcutaneously implanted port (Table 1). Trimethoprim/sulfamethoxazole prophylaxis (150 mg trimethoprim/m2 once daily, three times weekly) was given to 55 children (96.5 %), with the exception of one child with aplastic anemia and one child newly diagnosed with ALL (Table 1).

Table 1 Clinical characteristics of neutropenic children with viridans streptococcal bacteremia
Full size table

VSB occurred at a median of 13 days (range 8–21 days) after the beginning of the preceding chemotherapy and after a median of 6 days (range 0–46 days) of neutropenia (Table 1). Fever duration was a median of 4 days (range 1–21 days) and intravenous antibiotics were given for a median of 13 days (range 6–29 days) (Table 1). Gastrointestinal symptoms such as vomiting, diarrhea, and abdominal pain were the most common accompanying symptoms with fever, followed by respiratory symptoms (Table 1).

The two children with mixed bacterial infections and one child with a subsequent methicillin-susceptible S. aureus bacteremia recovered without severe complications, and there were no significant differences in their clinical characteristics compared to the other children. VSB occurred 8–12 days after the preceding chemotherapy and fever lasted for 2–6 days in these three children. The other child with subsequent E. faecium bacteremia died from underlying chronic myeloid leukemia (CML) blast crisis.

Laboratory characteristics

Laboratory tests showed severe neutropenia (ANC <100/μL) on fever day 1 in 56 cases. The median WBC count, ANC, and CRP level on fever day 1 was 70/μL (range 10–2,250/μL), 0/μL (range 0–330/μL), and 3.4 mg/dL (range 0.2–34.1 mg/dL), respectively (Table 1). The median of the maximum CRP levels during VSB was 17.2 mg/dL (range 0.9–34.1 mg/dL), and this was significantly higher than the CRP level on fever day 1 (p < 0.001) and showed a positive correlation with fever duration (r = 0.362, p = 0.007, Fig. 1). All maximum CRP levels during VSB were checked within a week after the onset of fever, and the maximum CRP levels of 33 children (57.9 %) and 52 children (91.2 %) were checked within 3 and 5 days after the onset of fever, respectively.

Fig. 1
figure 1

The correlation between the maximum C-reactive protein (CRP) levels during viridans streptococcal bacteremia and fever duration (r = 0.362, p = 0.007)

Full size image

Children with severe VSB complications

Severe complications occurred in four children (Table 2). One patient visited the hospital in septic shock, experienced multifocal pneumonia, eventually recovered, and was discharged on hospital day 11. In the other three children, severe complications began with hypoxia requiring oxygen therapy between fever days 3 and 5. Two children experienced ARDS and shock, and one of the two children died from viridans streptococcal sepsis. The last patient died from underlying CML blast crisis during oxygen therapy. Overall mortality was 3.5 % (two children), and mortality attributable to VSB was 1.8 % (one child). Severe complications developed significantly more frequently in children who had received SCT (75.0 vs. 9.4 %, p = 0.007) and in those with respiratory symptoms (75.0 vs. 17.0 %, p = 0.026). The maximum CRP levels were significantly higher in children with severe complications compared to those without complications (32.5 vs. 16.5 mg/dL, p = 0.001). The cut-off value of the maximum CRP level predicting severe VSB complications was 29.0 mg/dL, with a sensitivity, specificity, positive predictive value, and negative predictive value of 100, 86.8, 36.4, and 100 %, respectively (Fig. 2).

Table 2 Characteristics of children with severe complications of viridans streptococcal bacteremia
Full size table
Fig. 2
figure 2

Receiver operating characteristic curve for the maximum C-reactive protein levels during viridans streptococcal bacteremia. Area under the curve (AUC) = 0.929

Full size image

Antibiotic treatment

Fifty-six children (98.2 %) received piperacillin/tazobactam and isepamicin as an initial empirical antibiotic treatment, and 38 children (67.9 %) and 27 children (48.2 %) subsequently received meropenem as the second line of antibiotic treatment and empirical antifungal therapy, respectively. The other child (1.8 %) received meropenem as an initial empirical antibiotic treatment. Teicoplanin was given initially to five children (8.8 %) and glycopeptides were not given at all during the entire course of antibiotic treatment to four children (7.0 %). Fifteen (26.3 %) of the remaining 48 cases received glycopeptides (teicoplanin in 14 children, vancomycin in one child) before the second blood culture test. The remaining 33 cases (57.9 %) received glycopeptides (teicoplanin in 27 children, vancomycin in six children) after the second blood culture, and 30 of these children (52.6 %) eventually had a negative result for the second blood culture.

Antibiotic susceptibility

Antibiotic susceptibility to penicillin, erythromycin, clindamycin, vancomycin, linezolid, and quinupristin/dalfopristin was evaluated for all isolates, and susceptibility tests to cefotaxime and cefepime were performed for 56 and 55 isolates, respectively. Susceptibilities to penicillin, cefotaxime, and cefepime were 29.8, 58.9, and 69.1 %, respectively (Fig. 3). Clindamycin susceptibility was 80.7 % and vancomycin susceptibility was 100 % (Fig. 3). Antibiotic susceptibilities and the development of severe complications were not significantly related (data not shown).

Fig. 3
figure 3

Antibiotic susceptibility rates of viridans streptococci in 57 neutropenic febrile children with viridans streptococcal bacteremia

Full size image

Discussion

We conducted this study to determine the characteristics of VSB and the antibiotic susceptibility of viridans streptococci in children with febrile neutropenia.

In accordance with previous studies [8, 14–16], VSB occurred at a median of 13 days (range 8–21 days) after preceding chemotherapy and at a median of 6 days (range 0–46 days) after the onset of neutropenia, and fever lasted for a median of 4 days (range 1–21 days). Similar to known risk factors for VSB [14–18], 71.9 % of cases received a chemotherapy regimen including HD-ARA, and 70.2 % of cases had AML as an underlying disease. Although oral mucositis is another risk factor for VSB, it was present in only 5 % of cases. While previous studies reported that severe complications occurred in 15–39 % of patients and death occurred in up to 20 % of patients [10, 13–15, 19, 20], our study revealed severe complications in 7.0 % of cases, with an overall mortality and mortality attributable to VSB of 3.5 and 1.8 %, respectively. The occurrence of severe complications and death was less frequent than 8–17 years ago, despite the increase in occurrence of VSB, and this finding is probably due to the lower virulence of viridans streptococci and improvements in the supportive care of neutropenic patients.

Laboratory tests revealed that the CRP level increased significantly within a week after the onset of fever, and that the CRP levels of children with severe complications were significantly higher than those of children without complications. Except in one child who presented with septic shock, hypoxia, which developed on fever days 3, 4, and 5 in the other three children with severe VSB complications, respectively, was the first sign of impending severe complications, and the maximum CRP levels were observed earlier, on fever days 1, 3, and 4, respectively. Based on these findings, high CRP level on fever days 3 or 4 in children with VSB may be helpful in predicting increased risk of severe complications. Because the positive predictive value in predicting severe complications of VSB was 36.4 % of the cut-off value of the maximum CRP level, 29.0 mg/dL, we cannot predict the development of severe complications exactly. However, we can exclude the development of severe VSB complications in children with maximum CRP level <29.0 mg/dL. Because CRP is produced by the stimulation of inflammatory cytokines in hepatocytes [21] and severe VSB complications are assumed to be caused by the effects of inflammatory cytokines rather than viridans streptococci themselves [22–24], the CRP levels might be higher in children with severe complications under the common influence of inflammatory cytokines. Fever is also caused by the effects of inflammatory cytokines on the hypothalamic thermoregulatory center [25], and the positive correlation between maximum CRP levels and fever duration might be due to inflammatory cytokines too.

Antibiotic susceptibility tests showed low susceptibility to penicillin, confirming previous knowledge. Penicillin susceptibility of viridans streptococci was 86 % in 2009 [4], and it decreased markedly to 29.8 % in 2012 at our hospital. Cefotaxime susceptibility was about 80 % before [4, 11, 19] and also decreased to 58.9 % in this study. Vancomycin susceptibility remained at 100 % [4, 11, 19, 26]. In the past, decreased susceptibility to several β-lactam antibiotics, as seen in our results, was a reason for the initial use of glycopeptide in the empirical antibiotic treatment of patients with febrile neutropenia [1, 27]. In this study, however, cefepime susceptibility was roughly 70, and 90.9 % of patients who received glycopeptides after a second round of blood cultures showed a microbiologic response to empirical antibiotic treatment not including glycopeptides. In addition, some investigators have reported that the initial empirical use of vancomycin did not influence mortality and morbidity when compared to the addition of vancomycin after the identification of Gram-positive bacterial growth in culture studies [28–30], and that severe VSB complications were not significantly related to antibiotic resistance [10, 11]. Thus, practical guidelines in Korea and America recommend the selective use of glycopeptides rather than their universal use in the empirical treatment of patients with febrile neutropenia, and suggest that viridans streptococcal infections should be considered, especially in patients with septic shock or severe mucositis [6, 7]. To add to these guidelines, we suggest using glycopeptides according to the results of blood culture and antibiotic susceptibility tests rather than risk factors only, because severe mucositis and septic shock only occurred in 5 % of febrile neutropenic children with VSB, and many clinicians should hesitate to stop glycopeptides having been administered for several days.

It is interesting that 67.9 % of children who received piperacillin/tazobactam with or without glycopeptides were eventually treated with meropenem due to prolonged fever, despite sufficient microbiologic response to empirical antibiotic treatment and cefepime susceptibility. Prolonged fever after the eradication of viridans streptococci from blood could be caused by factors other than the viridans streptococci themselves; hypercytokinemia might be a cause of prolonged fever, considering the significant increase in CRP levels after the diagnosis of VSB and the positive correlation between fever duration and maximum CRP levels.

One limitation of this retrospective study was that we could not determine the exact therapeutic effect of the first-line empirical antibiotics because adding glycopeptides and changing to second-line antibiotics were more influenced by clinicians’ preferences than systematic guidelines during the study period. This resulted in a high prevalence of glycopeptide use in our study population. We were also not able to evaluate risk factors for severe VSB complications due to the small number of children with severe complications; multicenter studies including a much greater number of cases are needed in order to evaluate these risk factors. However, the low incidence of severe VSB complications may support the proposition that an initial aggressive therapy with glycopeptides is not required.

In conclusion, severe complications and death attributable to VSB were rarer than those reported in the past, despite the increased incidence of VSB in neutropenic febrile children. We suggest using glycopeptides according to the results of culture studies and antibiotic susceptibility tests rather than according to risk factors only or routinely, because cefepime susceptibility and the therapeutic response to empirical antibiotic treatment not including glycopeptide were sufficient in our study. Prolonged fever after the eradication of viridans streptococci may be related to hypercytokinemia, and the control of hypercytokinemia may have a therapeutic effect on VSB and its severe complications. These are to be investigated in the future.