Dr E. S. J. M. de Bont, Division of Paediatric Oncology, Children's Cancer Centre, Beatrix Children's Hospital, University Hospital Groningen, P.O. Box 30001, 9700 RB Groningen, The Netherlands.
The standard therapy for patients with fever and chemotherapy-related neutropenia is hospitalization and infusion of broad-spectrum antibiotics. Early discharge of a defined group of patients at low risk for septicaemia would be of great advantage for these patients. In this study plasma interleukin-8 (IL-8) and interleukin-6 (IL-6) levels measured at start of fever (n = 72) could define a low-risk group of febrile patients with neutropenia due to chemotherapy. For this purpose we collected and analysed data on 72 fever episodes from 53 patients with chemotherapy-related neutropenia, aged between 1 and 66 years. Of the 72 episodes, 18 were classified as bacteraemia and/or clinical sepsis (sepsis group). The IL-6 and IL-8 plasma concentration were significantly increased in patients with chemotherapy-related neutropenia and fever due to bacteraemia versus fever of non-bacterial origin (P = 0.043 and P = 0.022 respectively). Logistic regression analysis, with sepsis as the outcome variable, revealed significant effects of age combined with either IL-6 or IL-8. Sepsis occurrence was lowest for patients <16 years and highest in patients between 16 and 50 years, and was higher in patients with increased IL-6 (P = 0.032) or IL-8 (P = 0.049). No significant effect of leucocyte count, C-reactive protein, sex or underlying malignancy at presentation was detected. The plasma IL-6 and IL-8 levels were fairly strongly correlated (Pearson r = 0.62). Using a cut-off value with 100% sensitivity, both IL-8 and IL-6 could define a low-risk group of neutropenic patients of 28% (CI 15–40%) at the start of the febrile period. Intervention studies are warranted to confirm this result and to investigate whether an early discharge based on IL-8 or IL-6 measurement is safe, increases the quality of life, and results in cost savings.
Patients with acute lymphoblastic or myeloid leukaemia (ALL and AML respectively) often suffer from infections with serious complications (Albano & Pizzo, 1988). The malignant disease and/or intensive chemotherapy may cause an impaired host defence to infection. Key factors are the intensity and duration of neutropenia, but a decreased function of granulocytes and disturbances of natural barriers may substantially add to the risk of serious infections (Albano & Pizzo, 1988).
Patients with chemotherapy-related neutropenia and fever are usually hospitalized and treated on empirical intravenous broad-spectrum antibiotic regimens until the patient is afebrile, the blood cultures are negative and absolute neutrophil count (ANC) has recovered to >0.5 × 109/l (Lee & Pizzo, 1993). The mean hospitalization duration varies between 6 and 7 d (Vellenga et al, 1996). For high-risk patients, hospital treatment is appropriate considering the possibility of rapid deterioration. However, prolonged hospitalization of patients with fever and neutropenia due to chemotherapy results in risk of selection for resistant flora and risk of exposure to nosocomial pathogens. Early discharge or no admission at all of low-risk patients would be of interest since the costs of prolonged hospitalization are substantial. The crucial question in this setting is, which test is a reliable diagnostic marker to identify a low-risk group of febrile patients with neutropenia.
In a retrospective study using clinical variables, no risk factors were found to define a population at low and at high risk for sepsis (Talcott et al, 1992). Recently several studies from one group have suggested that low-risk patients may not require extended hospitalization for intravenous antibiotics (Katz et al, 1992; Buchanan, 1993; Bash et al, 1994; Mustafa et al, 1996; Aquino et al, 1997a, b). Low-risk patients, defined at day 3 of hospitalization, had the following profile: signs of bone marrow recovery (ANC >0.1 × 109/l and platelets >75 × 109/l), negative blood culture after 3 d, afebrile for the last 24 h, improvement of a known local infection, no other reason to prescribe intravenous antibiotics and the possibility of returning to the hospital when there was recurrent fever. Unfortunately, the duration of hospitalization in these studies was still between 5 and 7 d (6.7 ± 0.3 d) for low-risk patients (Bash et al, 1994; Aquino et al, 1997b).
In non-neutropenic patients excessive release of cytokines by the host in response to bacterial infection causes the clinical picture of sepsis and septic shock, and is directly related to disease severity and outcome. In this cytokine cascade (pro)inflammatory cytokines, such as interleukin-6 and interleukin-8, play an important role (Hack et al, 1989, 1992; Ostermann et al, 1994; de Bont et al, 1993). In parallel to the production of (pro)inflammatory cytokines, inhibitory cytokines are also produced, such as interleukin-10, interleukin-1 receptor antagonist, soluble TNF receptor I and II (de Bont et al, 1995, 1996).
Some studies have suggested a possible role for IL-6 as a diagnostic test in the differentiation between fever due to sepsis or non-bacterial conditions in neutropenic patients (Engel et al, 1994; Heney et al, 1992; Bruserud et al, 1995), whereas others could not confirm this (Ostermann et al, 1994; Riikonen et al, 1992; Engervall et al, 1995a). CRP is also extensively studied in neutropenic patients with fever, but its usefulness remains questionable (Schimpff et al, 1972). Serial measurements of serum CRP levels may be helpful in determining the risk for unidentified infections or poor outcome when CRP stayed high for days (Manian, 1995; Santolaya et al, 1994). Another study preferred IL-6 instead of CRP in monitoring the acute response to infection (Engervall et al, 1995b). The present study was designed to determine the value of IL-8 and/or IL-6 as diagnostic markers in patients with fever and neutropenia due to chemotherapy at the start of the febrile episode. The results demonstrate that IL-8 as well as IL-6 plasma levels measured at the presentation of fever in patients with chemotherapy-related neutropenia can define patients at low risk of septicaemia.
PATIENTS, MATERIALS AND METHODS
This study was performed at the Division of Haematology and Paediatric Oncology/Haematology at the University Hospital Groningen from January to November 1998. Patients with ALL, AML, chronic myeloid leukaemia (CML), non-Hodgkin lymphoma (NHL), Hodgkin's disease, or paediatric solid tumours were included in the study and were treated with standard chemotherapy regimens dependent on the disease state. Eligible patients with chemotherapy-related neutropenia (granulocytes < 0.5 × 109/l or leucocytes <1 × 109/l) and fever (body temperature of >38.0°C over a 6 h observation period or 38.5°C once) were entered. Patients already receiving antibiotics for the suspected infection were excluded as were patients who had undergone autologous or allogeneic bone marrow transplantation. Fever periods clearly associated with the administration of potential pyrogenic agents, as well as fever after blood components were excluded. The use of selective gut decontamination or Pneumocystis carinii pneumonia prophylaxis (oral trimethoprim–sulphamethoxazole 3/15 mg/kg/d, 3 d a week) was not an exclusion criterion. The study protocol was approved by the medical ethics committee and all patients and/or parents, dependent on age, gave informed consent.
Design and treatment
This was a prospective study. From enrolled patients a full medical history was taken and physical examination was performed. Baseline investigations included full blood cell counts with differential white blood cell counts, sinus and chest radiographs. Cultures of blood and other suspicious sites were collected. In addition, EDTA blood specimens and serum were collected before the start of antibiotic treatment and frozen (−80°C) for CRP and cytokine analysis.
Clinical sepsis was defined by the following criteria: systolic blood pressure < 90 mmHg in adults or < −2 SD for age in children, heart frequency >100/min in adults or > +2 SD for age in children adapted from the criteria of systemic inflammatory response syndrome (SIRS) (Bone et al, 1992). Fever episodes without bacteraemia as shown by negative blood cultures and without signs of clinical sepsis were included in group I; no sepsis (i.e. fever due to non-bacterial conditions). Fever episodes with a positive blood culture or with signs of a clinical sepsis were included in group II sepsis. To minimize the risk of false-positive bacteraemias, coagulase-negative staphylococci were included only if both bottles of one blood culture set were positive. All patients were treated with Imipenem (4 × 500 mg/d for adults) or ceftazidime (3 × 50 mg/kg/d for children) as broad-spectrum antibiotic treatment. After 48 h of persistent fever vancomycine or teicoplanin was added (Lee & Pizzo, 1993).
Measurement of plasma IL-6 and IL-8 levels
EDTA blood specimens were used to measure IL-6 and IL-8 plasma concentrations with a continuous random access chemiluminescence immunoassay system (Diagnostic Product Corporation, Los Angeles, U.S.A.) according to the manufacturer's instructions. Duration of the test was 35 min.
Measurement of CRP level
Serum was used to measure the CRP concentration on a nephalometer according to the manufacturer (Dade-Behring, Amersfoort, The Netherlands).
Data on IL-8, IL-6 and CRP were summarized as medians and ranges. For crude comparisons of these parameters in various groups we used the Mann-Whitney U test. The correlation between these parameters was studied by means of the Pearson correlation coefficient calculated on a log scale. To investigate the relation between sepsis and age, sex, underlying malignancy, leucocyte count, CRP, IL-6 and IL-8 we used logistic regression analysis. For several patients more than one fever episode was recorded. As repeated observations within the same subject cannot be considered as independent, we used logistic regression model with a random intercept due to patients (Hedeker & Gibbons, 1996). We also carried out the logistic regression analysis with the first episode of each patient. Statistical tests were carried out at a significance level of 0.05.
Fifty-three patients were included in the study and their characteristics are depicted in 1Table I. In these 53 patients 72 episodes of chemotherapy-related neutropenia and fever were observed. 37 patients experienced one episode, 12 two episodes, one three episodes and two four episodes of fever during neutropenia. Among 15 patients with more than one episode, four experienced both an episode with and without bacteraemia. Two groups of febrile episodes could be distinguished. Group I consisted of 54 febrile episodes without a positive blood culture and without clinical signs of shock or a microbiologically documented local infection. Group II consisted of 16 blood culture positive febrile periods or fever periods with a documented clinical sepsis (n = 2). Three other febrile periods due to a documented local infection (one skin infection, one sinusitis on sinus radiographs, and one pneumonia on chest radiographs), were excluded because the number of documented local infections was too low for analysis as a separate group. Baseline characteristics of group I and group II are shown in 1Table I. Positive blood cultures were present in 16/18 febrile periods (89%) of the patients in group II (Table II). Bacteraemia due to coagulase-negative staphylococci occurred in 56% (9/16) of the positive blood cultures. Only in 19% (3/16) of the blood culture positive febrile periods were Gram-negative micro-organisms cultured.
Table 1. Table I. Baseline characteristics (malignancy, age and sex at the first febrile episode only) of 53 different study patients. Group I: febrile episodes during neutropenia without a documented bacterial infection or clinical sepsis; group II: febrile episodes during neutropenia with positive blood culture or clinical sepsis. AML: acute myeloid leukaemia, ALL: acute lymphoblastic leukaemia, CML: chronic myeloid leukaemia.
Table 2. Table II. (A) Plasma values of febrile episodes (n = 72) for IL-8, IL-6 and CRP (median (range)). Group I: febrile episodes during neutropenia without a documented bacterial infection or clinical sepsis; group II: febrile episodes during neutropenia with positive blood culture or clinical sepsis. (B) Plasma values of febrile episodes for IL-8, IL-6 and CRP (median (range)) included in group II. * P-value < 0.05: significant differences between group I versus group II, or between bacteraemia due to Gram-positive versus Gram-negative micro-organisms.
Leucocyte count, CRP, IL-6 and IL-8 plasma levels
The leucocyte count at presentation with fever was significantly higher in group I versus group II (median (range): group I: 0.3 × 109/l (0.0–1.8) versus group II: 0.15 × 109/l (0.0–0.7) respectively, P < 0.05). At the time of admission for fever during chemotherapy-related neutropenia the CRP levels were significantly lower in group I versus group II (P = 0.022) (Table IIa). The same pattern was observed for IL-6 and IL-8 (P = 0.043 and P = 0.022). In 2Table II(b) the relation between the cytokine levels and the different microbial agents of group II are enlisted. Within group II we compared the CRP, IL-6 and IL-8 levels from the febrile periods due to Gram-positive micro-organisms to the levels from the febrile periods due to Gram-negative micro-organisms. Although the number of documented Gram-negative septicaemias was low (n = 3), the CRP, IL-6 and IL-8 levels were significantly increased in febrile periods documented with Gram-negative blood cultures versus febrile periods documented with Gram-positive blood cultures (P < 0.05) (Table II). Furthermore, there was a significant increase of the CRP, IL-6 and IL-8 levels during the febrile periods with documented Gram-positive blood cultures versus the febrile periods from group I (P = 0.015 resp. 0.028 and 0.023). The lowest values for IL-8 in the sepsis group belonged to patients with an indolent progressing coagulase-negative staphylococci sepsis (Fig 1).
The random effect logistic regression analysis revealed a significant effect (P < 0.01) of IL-6 and of IL-8 in combination with age, where age was classified as <16 years, 16–50 years, >50 years on the prediction of sepsis. Addition of IL-6 to a model containing IL-8, or vice versa, did not significantly improve the fit. IL-6 and IL-8 were highly correlated (Pearson r = 0.62, P < 0.001). Other investigated variables, such as sex, underlying malignancy, CRP and leucocyte count, were not found to be related to sepsis (P > 0.05).
The simpler analysis based upon only the first febrile episodes in each patient produced similar results. These results are summarized in 3Table III. The final model contained both age and IL-6 or age and IL-8. The table shows that the risk of sepsis was lowest for young patients <16 years and highest for patients between 16 and 50 years (P < 0.05), but the risk of sepsis was higher in patients with increased IL-6 (P = 0.032) or IL-8 (P = 0.049).
Table 3. Table III. Results of logistic regression analysis for the variables (first measurements only) related to bacteraemia or clinical sepsis. Results represent the regression coefficient (Regr. Coeff.), standard error (SE), P-value and Odds ratio for age and IL-6 corrected for age (upper part) and for age and IL-8 corrected for age (lower part). Ln(IL-6) are log-transformed IL-6 plasma levels and Ln(IL-8) are log-transformed IL-8 plasma levels.
Using a cut-off value with 100% sensitivity, IL-8 as well as IL-6 plasma levels could define a low-risk group of neutropenic patients with fever of 28% (95% CI 15–40%) at the start of the febrile episode.
It is common practice to hospitalize patients with chemotherapy-related fever and neutropenia as long as the neutropenia persists and to administer broad-spectrum intravenous antibiotics. This strategy was endorsed in a 1990 consensus statement by the Infectious Disease Society of America (Hughes et al, 1990). However, this does not seem to be the most effective way to treat this complication, especially for patients with low-risk factors who might have limited benefit from a long hospital stay. Early discharge of this low-risk group would be an advantage for these patients. A diagnostic test to define the low-risk group of neutropenic patients with fever would be of great value.
In the present study we evaluated the value of CRP, IL-6 and IL-8 in 72 included fever episodes with neutropenia due to chemotherapy. 26% of the febrile periods were actually due to sepsis. This is consistent with previous studies in which the number of positive blood cultures was reported to be 20–30% which is mainly due to bacteraemia with coagulase-negative staphylococci (Blay et al, 1996; Vellenga et al, 1996).
In this study, IL-6 and IL-8 plasma concentrations together with age were the strongest and best predictors for bacteraemia in patients with fever and neutropenia due to chemotherapy.
In this study, age was a predictor of sepsis; patients at the age of 16–50 years had the highest occurrence of sepsis. The type of malignancy could not reach significance when analysing variables influencing bacteraemia. It is known that most malignancies are age-dependent and that malignancy and age determines the intensity of the chemotherapy given. Age is strongly influenced by the combination of type of malignancy and its standard chemotherapy. The influence of age in the prediction of bacteraemia is lower together with IL-8 plasma levels than combined with IL-6 plasma concentrations. The results demonstrated that IL-8 and IL-6 plasma levels varied in the sepsis group. These variations seemed to reflect the differences in the micro-organisms, but also genetic factors regulating the cytokine response might play a role, also called cytokine polymorphisms. Recently, it was shown that TNFB2/TNFB2 allotype was related to an increased risk of sepsis in polytrauma patients (Flach et al, 1999) or an increased risk of death and severe disease in children with meningococcal disease (Nadel et al, 1996). Specific polymorphisms for IL-8 are still unknown. A polymorphism in the IL-6 gene (−174) showed that a genotype with low IL-6 responses seemed protective in systemic-onset juvenile chronic arteritis (Fishman et al, 1998).
There are several reasons why IL-8 could be a reliable marker. Our own data showed that significantly elevated IL-8 plasma levels in the group with bacteraemia persisted for at least 2 d, which make it attractive for clinical practice. The risk of missing a small transient increase seems impossible. Furthermore, at diagnosis and remission of acute leukaemia (n = 14), IL-8 levels were 3-fold lower than the plasma concentrations measured at presentation with fever in the non-septic group, supporting the possible role of IL-8 as a diagnostic marker in patients with fever and neutropenia due to chemotherapy (unpublished observations; Ostermann et al, 1994).
It seems unlikely that neutrophils and monocytes are the main source of IL-8 and IL-6 during febrile episodes and neutropenia (Ostermann et al, 1994). The number of neutrophils and monocytes is more than 5-fold lower in these patients and neutrophil activation markers, such as neopterin and elastase, were not increased (Ostermann et al, 1994). Therefore other cell types must be considered as a major source of IL-8 in neutropenic patients. It is known that endothelial cells can produce IL-8, but also IL-6 upon stimulation with tumour necrosis factor α or endotoxin (Huber et al, 1991; von Asmuth et al, 1993). Recently, murine blood fibrocytes were reported to secrete cytokines such as IL-8 (Chesney et al, 1998).
From this study it would seem feasible to define a low-risk group of febrile neutropenic patients based on clinical parameters and plasma IL-8 or IL-6 level at presentation. Choosing a cut-off value with 100% sensitivity and a negative predictive value of 100%, IL-6 as well as IL-8 plasma levels defined a group of low-risk patients of 28% (95% CI 15–40%) who might be discharged earlier. Furthermore, age had less influence on the prediction of bacteraemia when used together with IL-8 than together with IL-6 plasma levels. The factors used so far to define a low-risk group were determined at day 3 after admission (Bash et al, 1994). The absolute neutrophil count, the absence of comorbidity such as hypotension and respiratory compromise, the duration of fever (at least 24 h afebrile), the type of malignancy and disease status were the most important indicators (Bash et al, 1994). A prospective randomized study under strict conditions (e.g. hospitalization) and with well-defined boundaries for the IL-6 or IL-8 plasma levels will ultimately answer the question whether the observed parameters are of clinical relevance and might be used, in the future, for early discharge of a selected group of patients and result in cost reduction. Additional studies are necessary before a clinical ward can safely discharge febrile neutropenic patients.
We thank the medical and nursing staff of the Haematology and Paediatric Oncology/Haematology units and E. Blomsma, G. Altena and Y. Hettinga for data management.