Pediatr Blood Cancer 2015;62:471–476
Patient and Central Venous Catheter Related Risk Factors for Blood Stream Infections in Children Receiving Chemotherapy Joanne Yacobovich, MD,1 Tal Ben-Ami, MD,2 Tameemi Abdalla,2 Hannah Tamary, MD,1 Gal Goldstein, MD,3 Michael Weintraub, MD,2 Isaac Yaniv, MD,1 Amos Toren, MD,3 Gili Kenet, MD,4 and Shoshana Revel-Vilk, MD, MSc2* Background. The use of central venous catheters (CVCs) has greatly improved the quality of care in children receiving chemotherapy, yet these catheters may cause serious infectious complications. The aim of this prospective registry study was to assess the host and CVC-related risk factors for blood stream infections (BSIs). Procedure. Patients undergoing CVC insertion for chemotherapy were followed prospectively for CVC complications. At the time of enrollment demographic-, clinical- and CVC-related data were collected. Survival and Cox-regression analysis were performed. Results. A total of 423 CVCs were inserted into 262 patients for a total of 76,540 catheter-days. The incidence of BSIs was 1.95 per 1,000 patient-days (95% CI 1.66–2.29). Myeloid leukemia and younger age were associated with higher risk for BSI. At least one BSI occurred in
187 CVCs with an incidence of 2.84 per 1,000 catheter-days (95% CI 2.47–3.24). Externalized CVCs, that is, tunneled externalized catheters and peripheral inserted central catheters, were associated with higher risk for BSI in the group of diseases with relatively lower rate of infection. However, in diseases with high rate of infection no such association was found. The type of BSI was associated with the diagnosis and the CVC type. CVC occlusion was associated with higher risk for recurrent BSI and for coagulase negative staph BSI. Conclusions. Both patient and CVC-related factors are associated with higher risk of BSI in children receiving chemotherapy. The results of this study could be used in developing studies aiming to reduce the rate of BSIs in children with cancer. Pediatr Blood Cancer 2015;62:471–476. # 2014 Wiley Periodicals, Inc.
Key words: blood stream infection; central venous catheter; chemotherapy; children; risk factors
INTRODUCTION Blood stream infections (BSIs) are among the most serious complications in children receiving chemotherapy leading to increased hospitalization, intensive care admissions, extensive antibiotic treatment, and mortality. The introduction of intensified chemotherapy protocols and extensive use of central venous catheters (CVCs) increased the risk for BSI in the last decades [1,2]. Currently almost all children receiving chemotherapy for cancer or bone marrow transplantation (BMT) undergo insertion of a CVC. The type of CVC inserted for chemotherapy treatment in children is based on hospital experience, underlying disease, physician, and patient preference. Studies comparing the rate of complications between the available types of CVC are important for the decision process with patients and families. The risk for BSI in children with cancer was related to the type of CVC used and to the underlying disease [1,3]. Increased risk for BSI was found in external catheters, catheters with double lumen and in diseases who receive more intensive chemotherapy. However, as external, double lumen catheters are used mainly in children who receive more intensive chemotherapy it is important to assess the CVC-related risk separately from the host and underlying disease risk. A 2-year registry was established following all pediatric patients undergoing CVC insertion for chemotherapy in the three largest pediatric Hematology/Oncology centers in Israel. The incidence and risk factors for thrombotic complications were previously published [4]. The aim of this report is to describe the host, underlying disease and CVC-related risk factors for symptomatic BSI in pediatric patients treated for cancer or undergoing BMT for non-malignant disease.
25 months) in the three largest pediatric Hematology/Oncology centers in Israel. Each patient undergoing CVC insertion for treatment of a newly diagnosed cancer, relapsed cancer or for BMT was eligible for registration in this study. The patient and/or patient’s parents were recruited and gave informed consent for this study prior to or within the first 3 days from CVC insertion. The only exclusion criterion for this study was the lack of consent to participate. The study was approved by the local Institution Review Board of each participating center.
Data Collection Patients. An initial registration form was completed for each participating patient. This form included questions regarding demographic, clinical and CVC-related data. Patients were followed from the day of enrollment until removal of the last CVC, death, last follow-up or at study closure, if none of the above occurred.
Additional Supporting Information may be found in the online version of this article at the publisher’s web-site. 1
Hematology/Oncology Center, Schneider Children’s Medical Center, Petah Tikva, Sackler Faculty of Medicine, Tel-Aviv University, TelAviv, Israel; 2Pediatric Hematology/Oncology Department, HadassahHebrew University Medical Center, Jerusalem, Israel; 3Pediatric Hematology/Oncology Department, Tel-Hashomer, Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel; 4Pediatric Coagulation Service, Sheba Medical Center, Tel-Hashomer, Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel Grant sponsor: Israel Cancer Association
METHODS Study Design A registry of patients undergoing CVC insertion for chemotherapy treatment was maintained from 6/2006 to 7/2008 (total C
2014 Wiley Periodicals, Inc. DOI 10.1002/pbc.25281 Published online 18 October 2014 in Wiley Online Library (wileyonlinelibrary.com).
Conflict of interest: Nothing to declare.
Correspondence to: Shoshana Revel-Vilk, Pediatric Hematology/ Oncology Department, Hadassah-Hebrew University Medical Center, POB 12000, Jerusalem 91200, Israel. E-mail: [email protected] Received 24 March 2014; Accepted 29 August 2014
472
Yacobovich et al.
CVCs. For each CVC insertion the following data were collected: date of insertion, type of CVC, procedure setting, side of insertion, vein cannulated, and placement of the tip of the catheter. Decisions related to CVC insertion were made by the treating physician and/or surgeon and were based mainly on personal preferences, treatment protocol and underlying patient characteristics. The insertion and maintenance of CVCs was in accordance with institutional protocols. All CVC care was routinely done in the hospital. Tunneled externalized catheters and peripheral inserted central catheters (PICCs) were routinely flushed with heparin at least once a week in the participating hospital’s outpatients clinic; while totally implantable catheters were routinely flushed with heparin at least once a month in the participating hospital’s clinic. Routine antibacterial or antifungal prophylaxes were not used in the participating centers. If the CVC was removed during the study period, the date and cause of removal were recorded. If a new CVC was inserted during the study period, the new CVC-related data were recorded. Complications. For documentation of CVC-related complications, the following events were reported to the registry: occlusion, venous thrombosis and infectious events. Occlusion of the CVC was defined as an inability to infuse and/or withdraw blood that required medical attention, that is, local installation of thrombolytic therapy, or surgical intervention, that is, CVC removal. Venous thrombosis was generally suspected by the primary physician and confirmed by appropriate diagnostic imaging. The results of the thrombotic complications, both venous thrombosis and occlusion, were previously reported [4]. Infectious events included all cases of positive blood cultures from the CVC drawn due to fever or clinical signs suggesting infection. Concurrent peripheral blood cultures were not routinely drawn. For coagulase negative staph (CONS), a common skin contaminant, true BSIs were considered only when two or more blood cultures drawn on separate occasions were positive [5]. Common skin flora bacterial BSI was not recorded unless it was treated with antibiotic by the local physicians. CVC-related complications were captured for both in-patient and ambulatory patients.
Data Analysis Demographic and clinical characteristics of the study cohort with 95% confidence interval (CI) are reported. The association between the type of CVC and other variables in the analysis was analyzed using the ANOVA test for continuous data, that is, age, and Chi-squared test for categorical data, that is, underlying disease, place of procedure, side of insertion, vein cannulated and place of the tip. As the age variable was found to be skewed to the right, a z ¼ square root y transformation gave an approximately normal distribution. Groups of underlying diseases were predefined including acute lymphoblastic leukemia (ALL), myeloid leukemia (acute and juvenile myelomonocytic leukemia), brain tumors, sarcoma (Ewing sarcoma, osteosarcoma, rhabdomyosarcoma), neuroblastoma, lymphoma (Hodgkin and nonHodgkin), allogeneic BMT for non-malignant diagnoses and other (solid malignancy occurring in 5 children). Analysis of the risks of BSI complications was performed separately regarding host and CVC-related factors. The risk (with 95% CI) of a patient to develop at least one episode of the outcome was calculated, the numerator being all patients who had at least one episode of outcome and the denominator being patient-days at risk. Pediatr Blood Cancer DOI 10.1002/pbc
Patient-days were the sum of follow-up from time of enrollment to time of removal of the last CVC, death, last follow-up or study closure, if none of the above occurred. Second, the risk of each CVC (with 95% CI) to develop the outcome was calculated. The numerator was all CVCs that had at least one episode of the outcome and the dominator was catheter-days at risk. Catheter-days were the sum of follow-up from time of CVC insertion to the CVC removal, death, last follow-up or study closure, if none of the above occurred. Patient-days and catheter-days were counted whether or not the patient was in the hospital. For both analyses a Kaplan– Meier curve was computed and the 1-year CR to develop the outcome was established. The log-rank test was used to compare survival by categories of study variables. Finally, a Cox-regression model was computed to quantify the independent contribution of one or more factors of interest on survival, expressed as the hazard ratio (HR). For host-related factors the following variables were considered: age, gender, underlying disease, and hospital. For CVC-related factors the following variables were considered: type of CVC, procedure setting (operating theater, catheterization lab), side of insertion, vein cannulated and placement of the tip. The association between type of organism, that is, gram negative (GN), gram positive (GP), yeast, and coagulase negative staphylococcus (CONS), and other variables in the analysis, that is, underlying disease, procedure setting, side of insertion, vein cannulated, and placement of the tip was analyzed using the Chisquared test for categorical data. The association between BSI and CVC-related occlusion was also analyzed using the Chi-squared test. Statistical analysis were performed using the SPSS 19.0 for Windows (release 19.0.1, 2010) and the WINPEPI (PEPI-forWindows, Ver. 2.8, March 2007) [6]. Missing data were not included in the analysis. A 2-tailed P value of less the 0.05 was considered statistically significant. TABLE I. Characteristics of Patients and Central Venous Catheters Patients Total number Age, median (range), years Male gender Underlying disease Acute lymphoblastic leukemia Sarcoma Lymphoma Myeloid leukemia Brain tumor Neuroblastoma Other solid malignancies Bone marrow transplantation Central venous catheters (CVCs) Total number One CVC per patient Two CVCs per patient Three or more CVCs per patient Duration of CVC/pt, median (95% CI), months Peripherally inserted central catheters Tunneled externalized catheters Totally implantable catheters a
Median (range), months.
No. (%) 262 7.4 (0.08–28.3) 155 (59.1%) 73 (27.9%) 48 (18.3%) 36 (13.7%) 32 (12.2%) 25 (9.5%) 16 (5.7%) 26 (10.3%) 6 (2.3%) 423 162 (61.6%) 64 (24.7%) 36 (13.7%) 4.8 (95% CI 3.9–5.6) 3.1 (0–22.3)a 4.1 (0.13–23.6)a 8.5 (0.1–24)a
Risk Factors for Blood Stream Infections
473
Host Related Factors The incidence of first BSI in the different underlying diseases is presented in Table II. Myeloid leukemia was significantly associated with higher risk for BSI (P < 0.0001) (Fig. 1B). The hazard ratio (95% CI) for a first BSI for patients with myeloid leukemia compared to all other tumors was 3.09 (2.05–4.63). Younger age was associated with higher risk for BSI, hazard ratio 1.05 (95% CI 1.01–1.1) (P ¼ 0.022). Other variables such as gender and hospital were not associated with the risk of BSI. The number of BSIs ranged from 1 to 13 episodes per patient, the majority having one or two BSI, 78 (30%) and 45 (17%) patients, respectively. For the analysis of patient related risk factors for GP versus GN bacteria, only patients who had one type of bacteria were included (n ¼ 93). Patients with myeloid leukemia had higher risk for GP bacteria compared to other types of diagnosis, 10/16 (62%, 95% CI 37–83%) compared to 20/77 (26%, 95% CI 17–37%), respectively (P ¼ 0.017). Seven of the 10 GP BSIs in myeloid leukemia patients were with strep viridans. Infection with yeast and CONS were not associated with underlying disease and occurred in 5.8% (95% CI 3.3–9.4%) and 20.9% (95% CI 16.1–26.4%) of patients, respectively.
CVC Related Factors
Fig. 1. Cumulative risk of first blood stream infection (BSI) per patient for (A) all patients and (B) for patients with or without myeloid leukemia.
RESULTS A total of 262 children were included in this prospective cohort (Table I). BSI occurred at least once in 152 patients with an incidence of 1.95/1,000 patient-days (95% CI 1.66–2.29/1,000 patient-days). The 12-month cumulative incidence for the first BSI per patient from enrollment was 65% (Fig. 1A).
A total of 423 CVCs were inserted during the study period (Table I). At least one BSI occurred in 187 CVCs with an incidence of 2.84/1,000 catheter-days (95% CI 2.47–3.24/1,000 catheter-days). The 12-month cumulative incidence to develop the first BSI per catheter was 67%. The incidence of first BSI per type of CVC is presented in Table III. Type of CVC was significantly associated with BSI (P < 0.0001) (Fig. 2A). The hazard ratio (95% CI) for BSI in tunneled externalized catheters and PICCs compared to totally implantable catheters was 2.16 (1.5–3.14) and 1.43 (1.07–2.16), respectively. The number of BSI per CVC, median (range), was 1 (1–4) for totally implantable catheters, 1 (1–7) for tunneled externalized catheters and 1 (1–5) for PICCs (P < 0.001, Kurskal–Wallis test). Insertion of CVC to the right side was associated with a higher risk for BSI, with a hazard ratio (95% CI) of 1.39 (1.02–1.89) compared to insertion to the left side (P ¼ 0.038). The vein cannulated, the placement of the tip and the insertion setting were not associated with BSI. Occlusion of the CVC was not associated with occurrence of first BSI but was associated with recurrent BSI. Two or more BSI occurred in 21/102 occluded CVCs (20%, 95% CI 13.2–32.2%) compared to 37/321 (11.5%, 95% CI 8.3–15.5%) of non-occluded CVCs (P ¼ 0.03). Infection with CONS occurred in 15/21 (71%, 95% CI 48–89%) of occluded CVCs compared to 12/37 (32%, 18%, 95% CI 18–50%) of non-occluded CVC (P ¼ 0.005).
TABLE II. Incidence of Blood Stream Infections (BSIs) Per Underlying Disease Underlying disease
No.
BSIs
Patient-days
Sarcoma Neuroblastoma Hodgkin lymphoma Brain tumor Acute lymphoblastic leukemia (ALL) Other solid malignancies NonHodgkin lymphoma (NHL) Myeloid leukemia (ML) Bone marrow transplant (BMT)
48 16 15 25 73 26 21 32 6
21 9 5 14 41 16 12 30 4
15,937 6,616 3,270 8,741 23,419 7,356 3,873 8,023 934
Pediatr Blood Cancer DOI 10.1002/pbc
Per 1,000 patient-days 1.32 1.36 1.53 1.60 1.75 1.77 3.10 3.74 4.27
(95% (95% (95% (95% (95% (95% (95% (95% (95%
CI CI CI CI CI CI CI CI CI
0.82–2.01) 0.62–2.58) 0.50–3.57) 0.88–2.69) 1.26–2.38) 0.94–3.02) 1.60–5.41) 2.52–5.33) 1.16–10.9)
474
Yacobovich et al.
TABLE III. Incidence of Blood Stream Infections (BSIs) Per Type of CVC Type of CVC
No.
BSIs
Catheter-days
Per 1,000 catheter-days
126 188 104
56 75 63
33,437 23,837 18,779
1.67 (95% CI 1.27–2.17) 3.15 (95% CI 2.47–3.94) 3.35 (95% CI 2.58–4.29)
111 126 51
45 44 27
32,062 18,793 11,351
1.38 (95% CI 1.01–1.85) 2.34 (95% CI 1.70–3.14) 2.38 (95% CI 1.57–3.46)
6 29 43
5 17 28
603 2,436 5,523
8.29 (95% CI 2.69–19.4) 6.98 (95% CI 4.07–11.2) 5.07 (95% CI 3.37–7.33)
a
All underlying disease Totally implantable catheters Peripherally inserted central catheters Tunneled externalized catheters Acute lymphoblastic leukemia, solid tumors (group A)b Totally implantable catheters Peripherally inserted central catheters Tunneled externalized catheters Myeloid leukemia, bone marrow transplantation (group B) Totally implantable catheters Peripherally inserted central catheters Tunneled externalized catheters a
In five cases the type of central venous catheter (CVC) was not recorded. bTen patients with acute lymphoblastic leukemia (ALL) who underwent bone marrow transplantation (BMT) during the follow-up where excluded from this analysis.
For the analysis of the type of organism, that is, GP versus GN bacteria, only CVCs who had one type of bacteria were included (n ¼ 134); no significant association was found between GN versus GP and type of catheter (P ¼ 0.079). The incidence of CONS was significantly lower in PICC, 8.5% (95% CI 4.9–13.5%), compared to totally implantable catheters and tunneled externalized catheters, 15.3% (95% CI 9.5–22.9%) and 18.5% (95% CI 11.5–27.3%), respectively (P ¼ 0.036). Yeast was not associated with type of catheter and occurred in 4.1% (95% CI 2.4–6.5%) of catheters. In order to further study the risk of BSI per type of CVC we separated the cohort to two groups based on the distribution of CVC used (Table IV) and the incidence of BSI (Table II). Patients with ALL and solid tumors were grouped together (group A) since the incidence of BSI was relatively lower (1.32–1.75), the rate of totally implantable catheters use was around 40% and PICCs use around 45%. Ten patients with ALL undergoing BMT during the follow-up were excluded from this analysis. Patients with myeloid leukemia and BMT were grouped together (group B) since the incidence of BSI was relatively higher (3.7–4.3), the rate of tunneled externalized catheters was around 60% and PICCs use around 35%. Patients with NHL were not grouped in group B since the rate of tunneled externalized catheter use was only 19% and rate of PICCs use was 62%. In group A, the hazard ratio (95% CI) for tunneled externalized catheters and PICCs BSI compared to totally implantable catheters was 1.85 (1.15–2.97) and 1.42 (0.94–2.17), respectively (Fig. 2B, P ¼ 0.03). In group B, the hazard ratio (95% CI) for tunneled externalized catheters and PICCs BSI compared to totally implantable catheters was 0.71 (0.27–1.8) and 0.62 (0.27– 1.72), respectively (P ¼ 0.65). The incidence of first BSI per type of CVC in groups A and B are presented in Table III.
Type of Organism Fig. 2. Cumulative risk of first blood stream infection (BSI) per central venous catheter (CVC) for tunneled externalized catheters, peripherally inserted central catheters (PICCs) and totally implantable catheters for (A) all patients, (B) patients with acute lymphoblastic leukemia (ALL) and solid malignancy. Ten patients with ALL who underwent bone marrow transplant during the follow-up where excluded. Pediatr Blood Cancer DOI 10.1002/pbc
BSI was diagnosed in 290 cases. The distribution of the organisms is presented in Table SI. The CVC was removed within 14 days from of positive culture in 108 (37%) cases. During followup, 44 (16%) died. The cause of death was not registered in this study. Of those who died, 17/44 (39%) died within 14 days from the BSI. Type of organism was not associated with death within 14 days from the BSI.
Risk Factors for Blood Stream Infections
475
TABLE IV. Association Between the Age and Underlying Disease and the Type of Central Venous Catheter (CVC) Inserted
Total Age, median (range), years Mean SD, years Underlying disease Solid tumor Acute lymphoblastic leukemia Myeloid leukemia NonHodgkin lymphoma Bone marrow transplantation
TECs
TICs
PICCs
P for all types
418
104 (25%) 3.9 (0.08–18.1) 5.2 4.6
126 (30%) 5.45 (0.48–28.3) 8 6.5
188 (45%) 8.95 (0.31–28.3) 9.9 6.1
—
Patient and Central Venous Catheter Related Risk Factors for Blood Stream Infections in Children Receiving Chemotherapy Joanne Yacobovich, MD,1 Tal Ben-Ami, MD,2 Tameemi Abdalla,2 Hannah Tamary, MD,1 Gal Goldstein, MD,3 Michael Weintraub, MD,2 Isaac Yaniv, MD,1 Amos Toren, MD,3 Gili Kenet, MD,4 and Shoshana Revel-Vilk, MD, MSc2* Background. The use of central venous catheters (CVCs) has greatly improved the quality of care in children receiving chemotherapy, yet these catheters may cause serious infectious complications. The aim of this prospective registry study was to assess the host and CVC-related risk factors for blood stream infections (BSIs). Procedure. Patients undergoing CVC insertion for chemotherapy were followed prospectively for CVC complications. At the time of enrollment demographic-, clinical- and CVC-related data were collected. Survival and Cox-regression analysis were performed. Results. A total of 423 CVCs were inserted into 262 patients for a total of 76,540 catheter-days. The incidence of BSIs was 1.95 per 1,000 patient-days (95% CI 1.66–2.29). Myeloid leukemia and younger age were associated with higher risk for BSI. At least one BSI occurred in
187 CVCs with an incidence of 2.84 per 1,000 catheter-days (95% CI 2.47–3.24). Externalized CVCs, that is, tunneled externalized catheters and peripheral inserted central catheters, were associated with higher risk for BSI in the group of diseases with relatively lower rate of infection. However, in diseases with high rate of infection no such association was found. The type of BSI was associated with the diagnosis and the CVC type. CVC occlusion was associated with higher risk for recurrent BSI and for coagulase negative staph BSI. Conclusions. Both patient and CVC-related factors are associated with higher risk of BSI in children receiving chemotherapy. The results of this study could be used in developing studies aiming to reduce the rate of BSIs in children with cancer. Pediatr Blood Cancer 2015;62:471–476. # 2014 Wiley Periodicals, Inc.
Key words: blood stream infection; central venous catheter; chemotherapy; children; risk factors
INTRODUCTION Blood stream infections (BSIs) are among the most serious complications in children receiving chemotherapy leading to increased hospitalization, intensive care admissions, extensive antibiotic treatment, and mortality. The introduction of intensified chemotherapy protocols and extensive use of central venous catheters (CVCs) increased the risk for BSI in the last decades [1,2]. Currently almost all children receiving chemotherapy for cancer or bone marrow transplantation (BMT) undergo insertion of a CVC. The type of CVC inserted for chemotherapy treatment in children is based on hospital experience, underlying disease, physician, and patient preference. Studies comparing the rate of complications between the available types of CVC are important for the decision process with patients and families. The risk for BSI in children with cancer was related to the type of CVC used and to the underlying disease [1,3]. Increased risk for BSI was found in external catheters, catheters with double lumen and in diseases who receive more intensive chemotherapy. However, as external, double lumen catheters are used mainly in children who receive more intensive chemotherapy it is important to assess the CVC-related risk separately from the host and underlying disease risk. A 2-year registry was established following all pediatric patients undergoing CVC insertion for chemotherapy in the three largest pediatric Hematology/Oncology centers in Israel. The incidence and risk factors for thrombotic complications were previously published [4]. The aim of this report is to describe the host, underlying disease and CVC-related risk factors for symptomatic BSI in pediatric patients treated for cancer or undergoing BMT for non-malignant disease.
25 months) in the three largest pediatric Hematology/Oncology centers in Israel. Each patient undergoing CVC insertion for treatment of a newly diagnosed cancer, relapsed cancer or for BMT was eligible for registration in this study. The patient and/or patient’s parents were recruited and gave informed consent for this study prior to or within the first 3 days from CVC insertion. The only exclusion criterion for this study was the lack of consent to participate. The study was approved by the local Institution Review Board of each participating center.
Data Collection Patients. An initial registration form was completed for each participating patient. This form included questions regarding demographic, clinical and CVC-related data. Patients were followed from the day of enrollment until removal of the last CVC, death, last follow-up or at study closure, if none of the above occurred.
Additional Supporting Information may be found in the online version of this article at the publisher’s web-site. 1
Hematology/Oncology Center, Schneider Children’s Medical Center, Petah Tikva, Sackler Faculty of Medicine, Tel-Aviv University, TelAviv, Israel; 2Pediatric Hematology/Oncology Department, HadassahHebrew University Medical Center, Jerusalem, Israel; 3Pediatric Hematology/Oncology Department, Tel-Hashomer, Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel; 4Pediatric Coagulation Service, Sheba Medical Center, Tel-Hashomer, Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel Grant sponsor: Israel Cancer Association
METHODS Study Design A registry of patients undergoing CVC insertion for chemotherapy treatment was maintained from 6/2006 to 7/2008 (total C
2014 Wiley Periodicals, Inc. DOI 10.1002/pbc.25281 Published online 18 October 2014 in Wiley Online Library (wileyonlinelibrary.com).
Conflict of interest: Nothing to declare.
Correspondence to: Shoshana Revel-Vilk, Pediatric Hematology/ Oncology Department, Hadassah-Hebrew University Medical Center, POB 12000, Jerusalem 91200, Israel. E-mail: [email protected] Received 24 March 2014; Accepted 29 August 2014
472
Yacobovich et al.
CVCs. For each CVC insertion the following data were collected: date of insertion, type of CVC, procedure setting, side of insertion, vein cannulated, and placement of the tip of the catheter. Decisions related to CVC insertion were made by the treating physician and/or surgeon and were based mainly on personal preferences, treatment protocol and underlying patient characteristics. The insertion and maintenance of CVCs was in accordance with institutional protocols. All CVC care was routinely done in the hospital. Tunneled externalized catheters and peripheral inserted central catheters (PICCs) were routinely flushed with heparin at least once a week in the participating hospital’s outpatients clinic; while totally implantable catheters were routinely flushed with heparin at least once a month in the participating hospital’s clinic. Routine antibacterial or antifungal prophylaxes were not used in the participating centers. If the CVC was removed during the study period, the date and cause of removal were recorded. If a new CVC was inserted during the study period, the new CVC-related data were recorded. Complications. For documentation of CVC-related complications, the following events were reported to the registry: occlusion, venous thrombosis and infectious events. Occlusion of the CVC was defined as an inability to infuse and/or withdraw blood that required medical attention, that is, local installation of thrombolytic therapy, or surgical intervention, that is, CVC removal. Venous thrombosis was generally suspected by the primary physician and confirmed by appropriate diagnostic imaging. The results of the thrombotic complications, both venous thrombosis and occlusion, were previously reported [4]. Infectious events included all cases of positive blood cultures from the CVC drawn due to fever or clinical signs suggesting infection. Concurrent peripheral blood cultures were not routinely drawn. For coagulase negative staph (CONS), a common skin contaminant, true BSIs were considered only when two or more blood cultures drawn on separate occasions were positive [5]. Common skin flora bacterial BSI was not recorded unless it was treated with antibiotic by the local physicians. CVC-related complications were captured for both in-patient and ambulatory patients.
Data Analysis Demographic and clinical characteristics of the study cohort with 95% confidence interval (CI) are reported. The association between the type of CVC and other variables in the analysis was analyzed using the ANOVA test for continuous data, that is, age, and Chi-squared test for categorical data, that is, underlying disease, place of procedure, side of insertion, vein cannulated and place of the tip. As the age variable was found to be skewed to the right, a z ¼ square root y transformation gave an approximately normal distribution. Groups of underlying diseases were predefined including acute lymphoblastic leukemia (ALL), myeloid leukemia (acute and juvenile myelomonocytic leukemia), brain tumors, sarcoma (Ewing sarcoma, osteosarcoma, rhabdomyosarcoma), neuroblastoma, lymphoma (Hodgkin and nonHodgkin), allogeneic BMT for non-malignant diagnoses and other (solid malignancy occurring in 5 children). Analysis of the risks of BSI complications was performed separately regarding host and CVC-related factors. The risk (with 95% CI) of a patient to develop at least one episode of the outcome was calculated, the numerator being all patients who had at least one episode of outcome and the denominator being patient-days at risk. Pediatr Blood Cancer DOI 10.1002/pbc
Patient-days were the sum of follow-up from time of enrollment to time of removal of the last CVC, death, last follow-up or study closure, if none of the above occurred. Second, the risk of each CVC (with 95% CI) to develop the outcome was calculated. The numerator was all CVCs that had at least one episode of the outcome and the dominator was catheter-days at risk. Catheter-days were the sum of follow-up from time of CVC insertion to the CVC removal, death, last follow-up or study closure, if none of the above occurred. Patient-days and catheter-days were counted whether or not the patient was in the hospital. For both analyses a Kaplan– Meier curve was computed and the 1-year CR to develop the outcome was established. The log-rank test was used to compare survival by categories of study variables. Finally, a Cox-regression model was computed to quantify the independent contribution of one or more factors of interest on survival, expressed as the hazard ratio (HR). For host-related factors the following variables were considered: age, gender, underlying disease, and hospital. For CVC-related factors the following variables were considered: type of CVC, procedure setting (operating theater, catheterization lab), side of insertion, vein cannulated and placement of the tip. The association between type of organism, that is, gram negative (GN), gram positive (GP), yeast, and coagulase negative staphylococcus (CONS), and other variables in the analysis, that is, underlying disease, procedure setting, side of insertion, vein cannulated, and placement of the tip was analyzed using the Chisquared test for categorical data. The association between BSI and CVC-related occlusion was also analyzed using the Chi-squared test. Statistical analysis were performed using the SPSS 19.0 for Windows (release 19.0.1, 2010) and the WINPEPI (PEPI-forWindows, Ver. 2.8, March 2007) [6]. Missing data were not included in the analysis. A 2-tailed P value of less the 0.05 was considered statistically significant. TABLE I. Characteristics of Patients and Central Venous Catheters Patients Total number Age, median (range), years Male gender Underlying disease Acute lymphoblastic leukemia Sarcoma Lymphoma Myeloid leukemia Brain tumor Neuroblastoma Other solid malignancies Bone marrow transplantation Central venous catheters (CVCs) Total number One CVC per patient Two CVCs per patient Three or more CVCs per patient Duration of CVC/pt, median (95% CI), months Peripherally inserted central catheters Tunneled externalized catheters Totally implantable catheters a
Median (range), months.
No. (%) 262 7.4 (0.08–28.3) 155 (59.1%) 73 (27.9%) 48 (18.3%) 36 (13.7%) 32 (12.2%) 25 (9.5%) 16 (5.7%) 26 (10.3%) 6 (2.3%) 423 162 (61.6%) 64 (24.7%) 36 (13.7%) 4.8 (95% CI 3.9–5.6) 3.1 (0–22.3)a 4.1 (0.13–23.6)a 8.5 (0.1–24)a
Risk Factors for Blood Stream Infections
473
Host Related Factors The incidence of first BSI in the different underlying diseases is presented in Table II. Myeloid leukemia was significantly associated with higher risk for BSI (P < 0.0001) (Fig. 1B). The hazard ratio (95% CI) for a first BSI for patients with myeloid leukemia compared to all other tumors was 3.09 (2.05–4.63). Younger age was associated with higher risk for BSI, hazard ratio 1.05 (95% CI 1.01–1.1) (P ¼ 0.022). Other variables such as gender and hospital were not associated with the risk of BSI. The number of BSIs ranged from 1 to 13 episodes per patient, the majority having one or two BSI, 78 (30%) and 45 (17%) patients, respectively. For the analysis of patient related risk factors for GP versus GN bacteria, only patients who had one type of bacteria were included (n ¼ 93). Patients with myeloid leukemia had higher risk for GP bacteria compared to other types of diagnosis, 10/16 (62%, 95% CI 37–83%) compared to 20/77 (26%, 95% CI 17–37%), respectively (P ¼ 0.017). Seven of the 10 GP BSIs in myeloid leukemia patients were with strep viridans. Infection with yeast and CONS were not associated with underlying disease and occurred in 5.8% (95% CI 3.3–9.4%) and 20.9% (95% CI 16.1–26.4%) of patients, respectively.
CVC Related Factors
Fig. 1. Cumulative risk of first blood stream infection (BSI) per patient for (A) all patients and (B) for patients with or without myeloid leukemia.
RESULTS A total of 262 children were included in this prospective cohort (Table I). BSI occurred at least once in 152 patients with an incidence of 1.95/1,000 patient-days (95% CI 1.66–2.29/1,000 patient-days). The 12-month cumulative incidence for the first BSI per patient from enrollment was 65% (Fig. 1A).
A total of 423 CVCs were inserted during the study period (Table I). At least one BSI occurred in 187 CVCs with an incidence of 2.84/1,000 catheter-days (95% CI 2.47–3.24/1,000 catheter-days). The 12-month cumulative incidence to develop the first BSI per catheter was 67%. The incidence of first BSI per type of CVC is presented in Table III. Type of CVC was significantly associated with BSI (P < 0.0001) (Fig. 2A). The hazard ratio (95% CI) for BSI in tunneled externalized catheters and PICCs compared to totally implantable catheters was 2.16 (1.5–3.14) and 1.43 (1.07–2.16), respectively. The number of BSI per CVC, median (range), was 1 (1–4) for totally implantable catheters, 1 (1–7) for tunneled externalized catheters and 1 (1–5) for PICCs (P < 0.001, Kurskal–Wallis test). Insertion of CVC to the right side was associated with a higher risk for BSI, with a hazard ratio (95% CI) of 1.39 (1.02–1.89) compared to insertion to the left side (P ¼ 0.038). The vein cannulated, the placement of the tip and the insertion setting were not associated with BSI. Occlusion of the CVC was not associated with occurrence of first BSI but was associated with recurrent BSI. Two or more BSI occurred in 21/102 occluded CVCs (20%, 95% CI 13.2–32.2%) compared to 37/321 (11.5%, 95% CI 8.3–15.5%) of non-occluded CVCs (P ¼ 0.03). Infection with CONS occurred in 15/21 (71%, 95% CI 48–89%) of occluded CVCs compared to 12/37 (32%, 18%, 95% CI 18–50%) of non-occluded CVC (P ¼ 0.005).
TABLE II. Incidence of Blood Stream Infections (BSIs) Per Underlying Disease Underlying disease
No.
BSIs
Patient-days
Sarcoma Neuroblastoma Hodgkin lymphoma Brain tumor Acute lymphoblastic leukemia (ALL) Other solid malignancies NonHodgkin lymphoma (NHL) Myeloid leukemia (ML) Bone marrow transplant (BMT)
48 16 15 25 73 26 21 32 6
21 9 5 14 41 16 12 30 4
15,937 6,616 3,270 8,741 23,419 7,356 3,873 8,023 934
Pediatr Blood Cancer DOI 10.1002/pbc
Per 1,000 patient-days 1.32 1.36 1.53 1.60 1.75 1.77 3.10 3.74 4.27
(95% (95% (95% (95% (95% (95% (95% (95% (95%
CI CI CI CI CI CI CI CI CI
0.82–2.01) 0.62–2.58) 0.50–3.57) 0.88–2.69) 1.26–2.38) 0.94–3.02) 1.60–5.41) 2.52–5.33) 1.16–10.9)
474
Yacobovich et al.
TABLE III. Incidence of Blood Stream Infections (BSIs) Per Type of CVC Type of CVC
No.
BSIs
Catheter-days
Per 1,000 catheter-days
126 188 104
56 75 63
33,437 23,837 18,779
1.67 (95% CI 1.27–2.17) 3.15 (95% CI 2.47–3.94) 3.35 (95% CI 2.58–4.29)
111 126 51
45 44 27
32,062 18,793 11,351
1.38 (95% CI 1.01–1.85) 2.34 (95% CI 1.70–3.14) 2.38 (95% CI 1.57–3.46)
6 29 43
5 17 28
603 2,436 5,523
8.29 (95% CI 2.69–19.4) 6.98 (95% CI 4.07–11.2) 5.07 (95% CI 3.37–7.33)
a
All underlying disease Totally implantable catheters Peripherally inserted central catheters Tunneled externalized catheters Acute lymphoblastic leukemia, solid tumors (group A)b Totally implantable catheters Peripherally inserted central catheters Tunneled externalized catheters Myeloid leukemia, bone marrow transplantation (group B) Totally implantable catheters Peripherally inserted central catheters Tunneled externalized catheters a
In five cases the type of central venous catheter (CVC) was not recorded. bTen patients with acute lymphoblastic leukemia (ALL) who underwent bone marrow transplantation (BMT) during the follow-up where excluded from this analysis.
For the analysis of the type of organism, that is, GP versus GN bacteria, only CVCs who had one type of bacteria were included (n ¼ 134); no significant association was found between GN versus GP and type of catheter (P ¼ 0.079). The incidence of CONS was significantly lower in PICC, 8.5% (95% CI 4.9–13.5%), compared to totally implantable catheters and tunneled externalized catheters, 15.3% (95% CI 9.5–22.9%) and 18.5% (95% CI 11.5–27.3%), respectively (P ¼ 0.036). Yeast was not associated with type of catheter and occurred in 4.1% (95% CI 2.4–6.5%) of catheters. In order to further study the risk of BSI per type of CVC we separated the cohort to two groups based on the distribution of CVC used (Table IV) and the incidence of BSI (Table II). Patients with ALL and solid tumors were grouped together (group A) since the incidence of BSI was relatively lower (1.32–1.75), the rate of totally implantable catheters use was around 40% and PICCs use around 45%. Ten patients with ALL undergoing BMT during the follow-up were excluded from this analysis. Patients with myeloid leukemia and BMT were grouped together (group B) since the incidence of BSI was relatively higher (3.7–4.3), the rate of tunneled externalized catheters was around 60% and PICCs use around 35%. Patients with NHL were not grouped in group B since the rate of tunneled externalized catheter use was only 19% and rate of PICCs use was 62%. In group A, the hazard ratio (95% CI) for tunneled externalized catheters and PICCs BSI compared to totally implantable catheters was 1.85 (1.15–2.97) and 1.42 (0.94–2.17), respectively (Fig. 2B, P ¼ 0.03). In group B, the hazard ratio (95% CI) for tunneled externalized catheters and PICCs BSI compared to totally implantable catheters was 0.71 (0.27–1.8) and 0.62 (0.27– 1.72), respectively (P ¼ 0.65). The incidence of first BSI per type of CVC in groups A and B are presented in Table III.
Type of Organism Fig. 2. Cumulative risk of first blood stream infection (BSI) per central venous catheter (CVC) for tunneled externalized catheters, peripherally inserted central catheters (PICCs) and totally implantable catheters for (A) all patients, (B) patients with acute lymphoblastic leukemia (ALL) and solid malignancy. Ten patients with ALL who underwent bone marrow transplant during the follow-up where excluded. Pediatr Blood Cancer DOI 10.1002/pbc
BSI was diagnosed in 290 cases. The distribution of the organisms is presented in Table SI. The CVC was removed within 14 days from of positive culture in 108 (37%) cases. During followup, 44 (16%) died. The cause of death was not registered in this study. Of those who died, 17/44 (39%) died within 14 days from the BSI. Type of organism was not associated with death within 14 days from the BSI.
Risk Factors for Blood Stream Infections
475
TABLE IV. Association Between the Age and Underlying Disease and the Type of Central Venous Catheter (CVC) Inserted
Total Age, median (range), years Mean SD, years Underlying disease Solid tumor Acute lymphoblastic leukemia Myeloid leukemia NonHodgkin lymphoma Bone marrow transplantation
TECs
TICs
PICCs
P for all types
418
104 (25%) 3.9 (0.08–18.1) 5.2 4.6
126 (30%) 5.45 (0.48–28.3) 8 6.5
188 (45%) 8.95 (0.31–28.3) 9.9 6.1
—
