Ann Hematol DOI 10.1007/s00277-014-2273-z
ORIGINAL ARTICLE
Association between early peak temperature and mortality in neutropenic sepsis Robert Weinkove & Michael Bailey & Rinaldo Bellomo & Manoj K. Saxena & Constantine S. Tam & David V. Pilcher & Richard Beasley & Paul J. Young
Received: 11 August 2014 / Accepted: 2 December 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Fever is often the first sign of neutropenic infection, but its prognostic impact has not been established. We aimed to determine whether early peak temperature is associated with mortality in patients with neutropenic sepsis admitted to intensive care units (ICUs). We used a database of admissions to 157 ICUs in Australia and New Zealand between 2005 and 2013 to seek an association between peak temperature within the first 24 h in ICU and in-hospital mortality in neutropenic and non-neutropenic sepsis. Odds ratios for inhospital death were calculated for four temperature bands, adjusting for illness severity. Two patient cohorts were identified: neutropenic sepsis (N=4027) and nonneutropenic sepsis (N = 114,040). In-hospital mortality was higher in neutropenic sepsis than non-neutropenic
sepsis. In both cohorts, early peak temperature below 36.5 °C was associated with significantly increased mortality compared to normothermia. Among nonneutropenic patients, an early peak temperature of 37.5 °C or higher was associated with reduced mortality compared to normothermia. In contrast, in patients with neutropenic sepsis, fever was not associated with reduced mortality compared to normothermia. Similar findings were seen in a subgroup of the neutropenic sepsis cohort with a documented haematological malignancy. In neutropenic sepsis patients admitted to ICU, a temperature below 36.5 °C is associated with increased mortality compared with normothermia. In contrast to non-neutropenic sepsis, fever was not associated with a significant reduction in mortality in neutropenic patients.
Electronic supplementary material The online version of this article (doi:10.1007/s00277-014-2273-z) contains supplementary material, which is available to authorized users. R. Weinkove Wellington Blood & Cancer Centre, Capital & Coast District Health Board, Wellington, New Zealand R. Weinkove (*) Vaccine Research Group, Malaghan Institute of Medical Research, Wellington, New Zealand e-mail: [email protected]
R. Bellomo Intensive Care Unit, Austin Hospital, Melbourne, VIC, Australia M. K. Saxena St George Clinical School, University of New South Wales, Sydney, NSW, Australia C. S. Tam Peter MacCallum Cancer Centre, Melbourne, VIC, Australia
R. Weinkove Department of Pathology and Molecular Medicine, University of Otago Wellington, Wellington, New Zealand
D. V. Pilcher Intensive Care Unit, Alfred Hospital, Melbourne, VIC, Australia
M. Bailey Australian and New Zealand Intensive Care Research Centre, Monash University, Melbourne, VIC, Australia
R. Beasley : P. J. Young Medical Research Institute of New Zealand, Wellington, New Zealand
R. Bellomo : D. V. Pilcher Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation, Melbourne, VIC, Australia
P. J. Young Intensive Care Unit, Capital & Coast District Health Board, Wellington, New Zealand
Ann Hematol
Interventional studies are needed to determine whether physical or pharmacological measures to reduce fever influence outcomes during neutropenic infections. Keywords Neutropenia . Fever . Sepsis . Body temperature . Critical care
registry study including 118,067 admissions to intensive care units in Australia and New Zealand.
Materials and methods Study design and ethical approval
Introduction Neutropenia is a frequent complication of cytotoxic chemotherapy and carries the risk of life-threatening infection[1]. Neutropenic sepsis is defined as a systemic inflammatory response in the presence of infection and neutropenia[2, 3]. Although prompt antibiotic treatment improves outcomes, the mortality of patients admitted to intensive care units with neutropenic sepsis remains above 40 % [4–7]. Fever is commonly used to diagnose neutropenic infections [8–11] but is also an important component of the innate immune response to infection. Temperatures in the physiological febrile range suppress bacterial and viral growth and enhance bacterial sensitivity to antibiotics and to serum complement [12–15]. In animal models, fever is associated with prolonged survival in bacterial sepsis [16], and in human randomised controlled trials, the commonly used antipyretic paracetamol (acetaminophen) reduces clearance of falciparum malaria [17] and prolongs the duration of symptoms of rhinovirus infection [18]. The median duration of fever in neutropenic infections is 5 days, even with appropriate antibiotic therapy [19]. Optimal management of fever during this period has not been established, resulting in considerable variation in clinicians’ thresholds for physical and pharmacological cooling during neutropenic infections [20]. In critical care patients with infections, but not in patients without infections, the presence of fever is associated with improved survival compared with normothermia or hypothermia [21, 22], raising the possibility that permissive hyperthermia could be beneficial in sepsis. Interventional studies comparing fever management strategies are an area of active research [23–25]. The relationship between fever and mortality in neutropenic sepsis might plausibly differ from that observed in non-neutropenic sepsis. On one hand, fever could be of additional importance when innate immunity is otherwise compromised by neutropenia. On the other hand, the relationship between fever and mortality could be weaker because of the different spectra of infection and the contribution of malignancy-related fevers in neutropenic patients. To investigate this, we conducted a
A retrospective cohort study design was used to assess the association between peak temperature recorded during the first 24 h in intensive care and in-hospital mortality.
Data collection All patients over 16 years old admitted to an intensive care unit (ICU) at one of 157 centres in Australia and New Zealand between January 1st 2005 and December 31st 2013 were eligible for inclusion. Data were extracted from the Australian and New Zealand Intensive Care society Adult Patient Database (ANZICSAPD) [26], one of the three registries run by the ANZICS Centre for Outcome and Resource Evaluation (CORE). Presently, the APD receives data from approximately 110,000 ICU admissions each year, estimated to represent 90 % of all adult ICU admissions in Australia and New Zealand [27]. Patients were categorised into two cohorts using peripheral blood white cell count (WCC) and APACHE III (acute physiology and chronic health evaluation) codes [14] as follows: lowest WCC≥4.0×109/L and an admission diagnosis indicating infection and absence of an APACHE diagnostic or co-morbidity code indicating a haematological malignancy (non-neutropenic sepsis cohort); lowest WCC≤1.0×109/L and an admission diagnosis indicating infection or haematologic disorder (neutropenic sepsis cohort). A subset of the neutropenic cohort with an APACHE diagnostic or co-morbidity code indicating leukaemia, myeloma or lymphoma (‘neutropenic sepsis with haematological malignancy’) was also defined. Details of the diagnostic codes used to define each cohort are in Supplementary Appendix S1. Temperature was considered as a categorical variable, divided into four groups according to peak temperature during the first 24 h in intensive care: 39.4 °C. These categories were chosen based on physiologically relevant cut-offs and on the outcomes of prior analysis of the association between fever and outcome in ICU sepsis [28]. The primary outcome was in-hospital mortality determined at the time of discharge from the hospital containing the ICU to which the patient was admitted. The secondary outcome measure was ICU length of stay.
Ann Hematol
Statistical analysis Group comparisons were made using Chi-square tests for equal proportion, analysis of variance for normally distributed outcomes and Kruskal-Wallis tests otherwise. Results have been reported as N (%), mean (standard deviation) and median (interquartile range), respectively. Hospital mortality was analysed using logistic regression with results reported as odds ratios (95 % CI) referenced against the normothermia group (peak temperature 36.5 to 37.5 °C). Length of stay in ICU was found to be well approximated by a log-normal distribution so was log-transformed and analysed using generalised linear modelling with results reported as geometric means separately for survivors and non-survivors. To account for changes over time, variability between sites and illness severity, multivariate analysis was performed adjusting for year of admission, attending hospital and for illness severity using the Australian and New Zealand Risk of Death (ANZROD) model.[29] ANZROD is a mortality prediction model calibrated for use in Australian and New Zealand ICUs, is derived from components of the APACHE II and III scoring systems with additional diagnostic variables and combines eight risk adjustment algorithms, one for each major diagnostic group. To avoid confounding between illness severity, temperature and white cell count, the risk of death score was recalculated with temperature and white cell count removed. Analysis was performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). A two-sided p value of 0.05 was considered statistically significant. The study protocol was reviewed by the Australian and New Zealand Intensive Care Society (ANZICS) Centre for Outcome and Resource Evaluation (CORE) Management Committee and was prospectively registered (Australian New Zealand Clinical Trials Registry reference ACTRN12613000784718). Ethical approval for this study was obtained from the Alfred Hospital Human Research Ethics Committee (reference 333/13). The Quality Assurance Legislation of the Commonwealth of Australia (Part VC Health Insurance Act 1973, Commonwealth of Australia) allows use of data for research purposes without individual patient consent. In New Zealand, the use of anonymous data for outcome analysis is classified as a minimal risk observational study.
Results Baseline characteristics Over the study period, 965,745 patient episodes were included in the ANZICS-APD, of which 118,067 (12.2 %) fulfilled
inclusion criteria for one of the two cohorts and were included in the final analysis. A flow diagram of patients included in and excluded from the study is provided in Fig. 1. Supplementary Appendix S2 lists primary diagnoses for the nonneutropenic and neutropenic sepsis cohorts. Table 1 summarises baseline characteristics of the patient groups. Patients with neutropenic sepsis were significantly more likely to be male, were younger, were less likely to be intubated and had lower mean arterial pressures and higher peak temperatures than patients with non-neutropenic sepsis (p
ORIGINAL ARTICLE
Association between early peak temperature and mortality in neutropenic sepsis Robert Weinkove & Michael Bailey & Rinaldo Bellomo & Manoj K. Saxena & Constantine S. Tam & David V. Pilcher & Richard Beasley & Paul J. Young
Received: 11 August 2014 / Accepted: 2 December 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Fever is often the first sign of neutropenic infection, but its prognostic impact has not been established. We aimed to determine whether early peak temperature is associated with mortality in patients with neutropenic sepsis admitted to intensive care units (ICUs). We used a database of admissions to 157 ICUs in Australia and New Zealand between 2005 and 2013 to seek an association between peak temperature within the first 24 h in ICU and in-hospital mortality in neutropenic and non-neutropenic sepsis. Odds ratios for inhospital death were calculated for four temperature bands, adjusting for illness severity. Two patient cohorts were identified: neutropenic sepsis (N=4027) and nonneutropenic sepsis (N = 114,040). In-hospital mortality was higher in neutropenic sepsis than non-neutropenic
sepsis. In both cohorts, early peak temperature below 36.5 °C was associated with significantly increased mortality compared to normothermia. Among nonneutropenic patients, an early peak temperature of 37.5 °C or higher was associated with reduced mortality compared to normothermia. In contrast, in patients with neutropenic sepsis, fever was not associated with reduced mortality compared to normothermia. Similar findings were seen in a subgroup of the neutropenic sepsis cohort with a documented haematological malignancy. In neutropenic sepsis patients admitted to ICU, a temperature below 36.5 °C is associated with increased mortality compared with normothermia. In contrast to non-neutropenic sepsis, fever was not associated with a significant reduction in mortality in neutropenic patients.
Electronic supplementary material The online version of this article (doi:10.1007/s00277-014-2273-z) contains supplementary material, which is available to authorized users. R. Weinkove Wellington Blood & Cancer Centre, Capital & Coast District Health Board, Wellington, New Zealand R. Weinkove (*) Vaccine Research Group, Malaghan Institute of Medical Research, Wellington, New Zealand e-mail: [email protected]
R. Bellomo Intensive Care Unit, Austin Hospital, Melbourne, VIC, Australia M. K. Saxena St George Clinical School, University of New South Wales, Sydney, NSW, Australia C. S. Tam Peter MacCallum Cancer Centre, Melbourne, VIC, Australia
R. Weinkove Department of Pathology and Molecular Medicine, University of Otago Wellington, Wellington, New Zealand
D. V. Pilcher Intensive Care Unit, Alfred Hospital, Melbourne, VIC, Australia
M. Bailey Australian and New Zealand Intensive Care Research Centre, Monash University, Melbourne, VIC, Australia
R. Beasley : P. J. Young Medical Research Institute of New Zealand, Wellington, New Zealand
R. Bellomo : D. V. Pilcher Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation, Melbourne, VIC, Australia
P. J. Young Intensive Care Unit, Capital & Coast District Health Board, Wellington, New Zealand
Ann Hematol
Interventional studies are needed to determine whether physical or pharmacological measures to reduce fever influence outcomes during neutropenic infections. Keywords Neutropenia . Fever . Sepsis . Body temperature . Critical care
registry study including 118,067 admissions to intensive care units in Australia and New Zealand.
Materials and methods Study design and ethical approval
Introduction Neutropenia is a frequent complication of cytotoxic chemotherapy and carries the risk of life-threatening infection[1]. Neutropenic sepsis is defined as a systemic inflammatory response in the presence of infection and neutropenia[2, 3]. Although prompt antibiotic treatment improves outcomes, the mortality of patients admitted to intensive care units with neutropenic sepsis remains above 40 % [4–7]. Fever is commonly used to diagnose neutropenic infections [8–11] but is also an important component of the innate immune response to infection. Temperatures in the physiological febrile range suppress bacterial and viral growth and enhance bacterial sensitivity to antibiotics and to serum complement [12–15]. In animal models, fever is associated with prolonged survival in bacterial sepsis [16], and in human randomised controlled trials, the commonly used antipyretic paracetamol (acetaminophen) reduces clearance of falciparum malaria [17] and prolongs the duration of symptoms of rhinovirus infection [18]. The median duration of fever in neutropenic infections is 5 days, even with appropriate antibiotic therapy [19]. Optimal management of fever during this period has not been established, resulting in considerable variation in clinicians’ thresholds for physical and pharmacological cooling during neutropenic infections [20]. In critical care patients with infections, but not in patients without infections, the presence of fever is associated with improved survival compared with normothermia or hypothermia [21, 22], raising the possibility that permissive hyperthermia could be beneficial in sepsis. Interventional studies comparing fever management strategies are an area of active research [23–25]. The relationship between fever and mortality in neutropenic sepsis might plausibly differ from that observed in non-neutropenic sepsis. On one hand, fever could be of additional importance when innate immunity is otherwise compromised by neutropenia. On the other hand, the relationship between fever and mortality could be weaker because of the different spectra of infection and the contribution of malignancy-related fevers in neutropenic patients. To investigate this, we conducted a
A retrospective cohort study design was used to assess the association between peak temperature recorded during the first 24 h in intensive care and in-hospital mortality.
Data collection All patients over 16 years old admitted to an intensive care unit (ICU) at one of 157 centres in Australia and New Zealand between January 1st 2005 and December 31st 2013 were eligible for inclusion. Data were extracted from the Australian and New Zealand Intensive Care society Adult Patient Database (ANZICSAPD) [26], one of the three registries run by the ANZICS Centre for Outcome and Resource Evaluation (CORE). Presently, the APD receives data from approximately 110,000 ICU admissions each year, estimated to represent 90 % of all adult ICU admissions in Australia and New Zealand [27]. Patients were categorised into two cohorts using peripheral blood white cell count (WCC) and APACHE III (acute physiology and chronic health evaluation) codes [14] as follows: lowest WCC≥4.0×109/L and an admission diagnosis indicating infection and absence of an APACHE diagnostic or co-morbidity code indicating a haematological malignancy (non-neutropenic sepsis cohort); lowest WCC≤1.0×109/L and an admission diagnosis indicating infection or haematologic disorder (neutropenic sepsis cohort). A subset of the neutropenic cohort with an APACHE diagnostic or co-morbidity code indicating leukaemia, myeloma or lymphoma (‘neutropenic sepsis with haematological malignancy’) was also defined. Details of the diagnostic codes used to define each cohort are in Supplementary Appendix S1. Temperature was considered as a categorical variable, divided into four groups according to peak temperature during the first 24 h in intensive care: 39.4 °C. These categories were chosen based on physiologically relevant cut-offs and on the outcomes of prior analysis of the association between fever and outcome in ICU sepsis [28]. The primary outcome was in-hospital mortality determined at the time of discharge from the hospital containing the ICU to which the patient was admitted. The secondary outcome measure was ICU length of stay.
Ann Hematol
Statistical analysis Group comparisons were made using Chi-square tests for equal proportion, analysis of variance for normally distributed outcomes and Kruskal-Wallis tests otherwise. Results have been reported as N (%), mean (standard deviation) and median (interquartile range), respectively. Hospital mortality was analysed using logistic regression with results reported as odds ratios (95 % CI) referenced against the normothermia group (peak temperature 36.5 to 37.5 °C). Length of stay in ICU was found to be well approximated by a log-normal distribution so was log-transformed and analysed using generalised linear modelling with results reported as geometric means separately for survivors and non-survivors. To account for changes over time, variability between sites and illness severity, multivariate analysis was performed adjusting for year of admission, attending hospital and for illness severity using the Australian and New Zealand Risk of Death (ANZROD) model.[29] ANZROD is a mortality prediction model calibrated for use in Australian and New Zealand ICUs, is derived from components of the APACHE II and III scoring systems with additional diagnostic variables and combines eight risk adjustment algorithms, one for each major diagnostic group. To avoid confounding between illness severity, temperature and white cell count, the risk of death score was recalculated with temperature and white cell count removed. Analysis was performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). A two-sided p value of 0.05 was considered statistically significant. The study protocol was reviewed by the Australian and New Zealand Intensive Care Society (ANZICS) Centre for Outcome and Resource Evaluation (CORE) Management Committee and was prospectively registered (Australian New Zealand Clinical Trials Registry reference ACTRN12613000784718). Ethical approval for this study was obtained from the Alfred Hospital Human Research Ethics Committee (reference 333/13). The Quality Assurance Legislation of the Commonwealth of Australia (Part VC Health Insurance Act 1973, Commonwealth of Australia) allows use of data for research purposes without individual patient consent. In New Zealand, the use of anonymous data for outcome analysis is classified as a minimal risk observational study.
Results Baseline characteristics Over the study period, 965,745 patient episodes were included in the ANZICS-APD, of which 118,067 (12.2 %) fulfilled
inclusion criteria for one of the two cohorts and were included in the final analysis. A flow diagram of patients included in and excluded from the study is provided in Fig. 1. Supplementary Appendix S2 lists primary diagnoses for the nonneutropenic and neutropenic sepsis cohorts. Table 1 summarises baseline characteristics of the patient groups. Patients with neutropenic sepsis were significantly more likely to be male, were younger, were less likely to be intubated and had lower mean arterial pressures and higher peak temperatures than patients with non-neutropenic sepsis (p
