Volume 101, Issue 3 p. 567-577
Original Article
Free Access

Risk assessment of patients with hematologic malignancies who develop fever accompanied by pulmonary infiltrates

A historical cohort study

Massimo Offidani M.D.

Corresponding Author

Massimo Offidani M.D.

Department of Medicine, Immunology, and Hematology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Fax: (011) 390712813448

Clinica di Ematologia, Azienda Ospedaliera Umberto I, University of Ancona, Via Conca 1, 60020 Ancona, Italy===Search for more papers by this author
Laura Corvatta M.D.

Laura Corvatta M.D.

Department of Medicine, Immunology, and Hematology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Lara Malerba M.D.

Lara Malerba M.D.

Department of Medicine, Immunology, and Hematology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Monica Marconi M.D.

Monica Marconi M.D.

Department of Medicine, Immunology, and Hematology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Elisabetta Bichisecchi M.D.

Elisabetta Bichisecchi M.D.

Department of Radiology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Sara Cecchini M.D.

Sara Cecchini M.D.

Department of Radiology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Esther Manso M.D.

Esther Manso M.D.

Department of Microbiology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Tiziana Principi M.D.

Tiziana Principi M.D.

Intensive Care Unit, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Stefano Gasparini M.D.

Stefano Gasparini M.D.

Division of Pneumology, Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
Pietro Leoni PH.D.

Pietro Leoni PH.D.

Department of Medicine, Immunology, and Hematology, University of Ancona/Azienda Ospedaliera Umberto I, Ancona, Italy

Search for more papers by this author
First published: 28 June 2004
Citations: 11

Abstract

BACKGROUND

The mortality rate associated with fever accompanied by pulmonary infiltrates after chemotherapy for hematologic malignancies remains higher than the corresponding rate associated with febrile neutropenia without pulmonary infiltrates. Nonetheless, few studies have focused on the factors that predict outcome for patients with lung infiltrates. The purpose of the current study was to construct a risk model for clinical use by assessing the factors that affect outcome for patients with fever and pulmonary infiltrates.

METHODS

A historical cohort of 110 patients with hematologic malignancies who developed fever and pulmonary infiltrates was examined. Using parameters for which data were available at the onset of lung infiltrates, univariate and multivariate analyses were performed to assess factors affecting outcome. After a value of one point was assigned to each significant variable, a prediction score was calculated for each patient; scores were used to generate a system for identifying patients with a low risk of death due to fever accompanied by pulmonary infiltrates.

RESULTS

The crude mortality rate associated with pulmonary infiltrates was 23%; factors associated with cure included a favorable change in white blood cell counts (odds ratio [OR], 5.6; 95% confidence interval [CI], 1.7–18.9; P = 0.001), C-reactive protein levels < 10 mg/dL (OR, 4.6; 95% CI, 1.6–13.8; P = 0.001), and serum albumin levels ≥ 3 g/dL (OR, 3.2; 95% CI, 1.4–7.3; P = 0.004). Low-risk patients (risk score, 2–3) and high-risk patients (risk score, 0–1) had survival rates of 95% and 46%, respectively (P < 0.0001). The risk model had a specificity of 88% and a positive predictive value of 95%.

CONCLUSIONS

The risk model tested in the current study accurately predicted the survival of patients with hematologic malignancies who developed fever with pulmonary infiltrates. Once prospectively validated, the model could be used to select patients for trials involving novel diagnostic and therapeutic strategies. Cancer 2004. © 2004 American Cancer Society.

Febrile neutropenia after chemotherapy for hematologic malignancies remains a frequent and potentially life-threatening complication, although the mortality rate associated with this complication has decreased to less than 10% over the last 3 decades.1 Nonetheless, several studies have demonstrated that patients with neutropenia are a heterogeneous group exhibiting variability in terms of risk of infection-related morbidity and mortality.2-8 At present, low-risk patients with febrile neutropenia are safely managed with oral antibiotics9, 10 or with early discharge from hospital on the basis of suitable risk models.6, 8, 11

Pulmonary infiltrates were noted in 15–25% of all episodes of febrile neutropenia following chemotherapy for hematologic malignancies,12, 13 and despite advances in diagnostic techniques and therapeutic strategies,14 the mortality related to this complication remains high.12, 15 Distinction between infectious and noninfectious pulmonary infiltrates is difficult, and etiologic diagnoses often cannot be established, although bacteria and fungi are known to be the pathogens that are most commonly responsible.16 Thus, for most patients, the cause of death remains unknown. In addition, it is difficult to assess the risk of death due to a febrile episode associated with fever and pulmonary infiltrates. Nonetheless, few studies have attempted to identify risk factors predicting outcome for patients experiencing fever with lung infiltrates in this setting, and many of the studies that have done so have only considered clinical features that were present during the episode or were assessable retrospectively.12, 15, 17-19 In the current study, we set out to identify characteristics that were evaluable up to the onset of fever with pulmonary infiltrates and to assess their effect on outcome. From there, our goal was to combine these features in a risk model that could be used in the clinical decision-making process.

MATERIALS AND METHODS

Study Design

In the current historical cohort study, we enrolled all inpatients with fever and pulmonary infiltrates that were assessed by chest X-ray or computed tomography (CT) and that occurred after chemotherapy for a hematologic malignancy between 1998 and 2002 at the Azienda Ospedaliera Umberto I (Ancona, Italy), a tertiary-care institution. After study endpoints and other definitions were established, records containing the phrase fever and pulmonary infiltrates were extracted from our prospective database, which was designed specifically to investigate infections in patients with hematologic malignancies. Medical charts associated with 143 records were reviewed to verify eligibility and fulfillment of outcome assessment criteria. Thirty-three patients were excluded because they had experienced more than one episode of fever with pulmonary infiltrates or because they had developed fever with pulmonary infiltrates before or during chemotherapy. The remaining 110 patients were eligible for the study.

The primary endpoints of the study were 1) descriptive analysis of the clinical characteristics, epidemiology, and outcome associated with episodes of fever accompanied by pulmonary infiltrates and 2) identification of factors affecting outcome for patients experiencing this complication, with the ultimate goal of constructing a risk assessment model. We considered the following categoric variables to be assessable up to the start of chemotherapy: age (≤ 60 years vs. > 60 years), gender (male vs. female), underlying disease (other vs. acute myeloid leukemia), disease status (complete remission vs. other), comorbidity (no vs. yes), treatment phase (other vs. induction), type of chemotherapy used (other vs. high-dose cytosine arabinoside), treatment type (chemotherapy vs. transplantation), previous colonization with virulent pathogens (no vs. yes), previous infection (no vs. yes), previous hospitalization (no vs. yes), days of previous neutropenia, and days of previous exposure to antibiotics. The following variables were deemed to be assessable after chemotherapy: mucositis WHO grade (grading system; < 2 vs. ≥ 2), neutropenia (no vs. yes), days of neutropenia before onset of pneumonia (≤ 7 vs. > 7), white blood cell count trend (favorable vs. unfavorable), growth factor administration (yes vs. no), immunosuppressive therapy (no vs. yes), installation of central venous catheter (no vs. yes), antibiotic prophylaxis (yes vs. no), antifungal prophylaxis (yes vs. no), antiviral prophylaxis (yes vs. no), radiologic pattern (alveolar vs. interstitial-alveolar), hemoglobin level (≥ 10 g/dL vs. < 10 g/dL), platelet count (≥ 20,000 per microliter vs. < 20,000 per microliter), C-reactive protein level (≤ 10 mg/dL vs. > 10 mg/dL), serum albumin level (≥ 3 g/dL vs. < 3 g/dL), serum immunoglobulin level (≥ 0.4 g/dL vs. < 0.4 mg/dL), body temperature (< 39 °C vs. ≥ 39 °C), associated sepsis (no vs. yes), hypotension (no vs. yes), shock (no vs. yes), disseminated intravascular coagulation (no vs. yes), renal failure (no vs. yes), respiratory failure (no vs. yes), and abnormalities in terms of mental status (no vs. yes). In the current study, with the aim of constructing a prognostic model that is clinically useful at the time of diagnosis of fever with pulmonary infiltrates, we analyzed only variables that were assessable before or at the onset of this complication.

At the beginning of univariate analysis, we included all of the variables listed above. Adverse outcome was found to be significantly associated with hypotension, shock, respiratory failure, acute renal failure, disseminated intravascular coagulation (DIC), and alteration of mental state. We chose to exclude these abnormalities from subsequent analyses, because their association with poor outcome is well known and because they were detected in < 5% of the study population at the onset of fever with pulmonary infiltrates.

Patient Management and Monitoring

Patients were hospitalized in rooms containing one or two beds, depending on the severity of their condition and on whether they were infected with a virulent pathogen. All rooms were equipped with high-efficiency particulate air filters. The need for antibacterial, antifungal, and antiviral prophylaxis was established based on the expected duration of neutropenia (> 7 days vs. ≤ 7 days) and on other well known risk factors, with consideration also given to treatment of the underlying disease. Granulopoiesis-stimulating factors were administered to patients for whom neutropenia was expected to last longer than 7 days and to those who had developed infection during previous neutropenic phases. Patients were monitored daily for the development of signs and symptoms of infection. At the onset of fever with body temperature > 38 °C, duplicate blood culture analyses were performed, and antibiotic therapy with a single broad-spectrum agent or with a double-antibiotic combination was initiated. Subsequent modifications of empiric therapy were dictated by microbiologic and clinical considerations. The antifungal agent amphotericin B was added to the therapeutic regimen when fever persisted after 5–7 days of adequate antibiotic therapy. All patients who presented with respiratory symptoms underwent chest X-ray and/or thoracic CT scanning. For asymptomatic patients, these procedures were performed when fever persisted for 48–72 hours in spite of appropriate antibiotic therapy. Patients with persistent fever without pulmonary infiltrates underwent chest X-ray twice weekly and CT scanning once weekly. For patients who developed respiratory failure, decisions regarding therapeutic and ventilatory strategies, as well as possible admission into the intensive care unit, were agreed upon with intensive therapy specialists.

Cultures from sputum were obtained whenever possible. Invasive procedures such as bronchoalveolar lavage (BAL), transbronchial biopsy, percutaneous needle aspiration, and open-lung biopsy were performed only in selected cases. BAL samples were submitted for microscopic examination and for bacterial, fungal, mycobacterial, and Legionella culturing, immunofluorescent staining for Pneumocystis carinii, and viral culturing. Standard culturing methods were used. In patients with a high risk of cytomegalovirus (CMV) infection, CMV pp65 antigenemia was assessed weekly. Stool, oropharyngeal wash, and nasal swab cultures were examined weekly. Complete blood counts and electrolyte level measurements were obtained daily. Liver and renal function assays, as well as coagulation tests, were performed twice weekly. Protein electrophoresis and measurement of C-reactive protein (CRP) levels were performed once weekly at the onset of fever with or without pulmonary infiltrates and twice weekly thereafter.

Definitions

We used the term fever with pulmonary infiltrates rather than pneumonia, because patients with hematologic malignancies can develop noninfectious pulmonary complications, and a microbiologic diagnosis often cannot be established. We considered only patients for whom the onset of fever with lung infiltrates occurred at least 1 day after the end of chemotherapy. Pulmonary infiltrates were classified as having been microbiologically documented when a microorganism was detected in BAL specimens, the lung, the blood, or another normally sterile site. Coagulase-negative staphylococci or Candida spp. cultured from blood or BAL fluid were not considered causative of pulmonary infiltrates, because such organisms are very rarely responsible for pneumonia. Patients for whom all cultures yielded negative results were classified as having clinically documented pulmonary infiltrates. Methicillin-resistant Staphylococcus aureus, cephalosporin-resistant Streptococcus viridans, and multiresistant Pseudomonas aeruginosa, as well as multiresistant Acinetobacter and Enterobacteriaceae spp., were considered virulent because of their in vitro antimicrobial resistance, and Aspergillus, Fusarium, and Zygomycetes spp. were considered virulent because of their pathogenicity in vivo.

Previous infection was defined as infection with a virulent microorganism during a previous neutropenic episode. Fungal pneumonia was classified as proven, probable, or possible according to the definitions of the European Organization for Research and Treatment of Cancer/National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG).20 Respiratory failure was defined by a partial pressure of arterial oxygen–to–fraction of inspired oxygen (PaO2/FiO2) ratio of 200 or less.21

Patients were considered to have had a response to initial antibiotic treatment when progressive defervescence and improvement on chest X-rays were observed without a change in antibiotics. Final response was defined as defervescence lasting for a minimum of 3 days accompanied by clearing on chest X-rays. Outcome was classified as being unfavorable for patients who died with fever and pulmonary infiltrates. Laboratory-measured quantities included in the analysis of factors affecting outcome were determined on the day of detection of abnormalities on radiograph or CT scan or within 2 days of detection and no later. Hypogammaglobulinemia was defined by serum immunoglobulin levels < 0.4 g/dL as measured using an electrophoretic assay. Neutropenia following chemotherapy was defined by neutrophil counts of < 500 per microliter. The trend in white blood cell (WBC) count was considered favorable when WBC counts were recovering or neutrophil counts were already > 500 per microliter at the onset of fever with pulmonary infiltrates. The trend was considered unfavorable when WBC counts were decreasing or when they had not yet begun to recover following aplasia. CRP levels were measured using an immunonephelometric assay (NBII; Dade Behring, Marburg, Germany), and serum albumin levels were assessed using a bromocresol green–based colorimetric assay (ADVIA 1650; Bayer HealthCare LLC, Tarrytown, NY).

Statistical Methods

Patients' baseline characteristics and data on risk factors, epidemiology, therapy, and outcome were summarized by frequency tabulation (for categoric variables) or by calculating median values with ranges (for continuous variables). Univariate analysis of the association between each covariate and outcome was performed using the chi-square test (or Fisher exact test) for contingency tables. An estimated odds ratio (OR) was calculated for each variable. All continuous variables were categorized on the basis of clinical judgment or empirically (i.e., according to calculated regression coefficients). Maximum likelihood values were estimated, and P ≤ 0.1 was considered sufficient for inclusion of a variable in multivariate stepwise logistic regression analysis; P > 0.05 led to the removal of a variable from the multivariate model. Estimated ORs and confidence intervals (CIs) were calculated for all significant variables; because these ORs had similar 95% CIs, the same weight (equal to 1) was assigned to each variable. Based on this model, a prognostic score was calculated for each patient with the goal of identifying low-risk individuals.

We went on to assess the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the model. True-positive results were defined as cases in which a low-risk patient survived an episode of fever with pulmonary infiltrates; this definition was chosen with the goal of minimizing the rate of false identification of low-risk patients. We also generated a receiver operating characteristic curve to estimate the discriminatory power of the model.

The Kaplan–Meier method was used to calculate the probability of survival from the time of diagnosis of pulmonary infiltrates to last follow-up (90 days postdiagnosis). Differences between survival curves were assessed using the log-rank test.

All reported P values are two sided, and all statistical analyses were performed using SPSS statistical software (SPSS Inc., Chicago, IL).

RESULTS

Patient Characteristics and Risk Factors

Between January 1998 and December 2002, 110 patients developed fever with pulmonary infiltrates following chemotherapy for a hematologic malignancy (Table 1). The median patient age was 54 years (range, 20–75 years), and there were 56 male patients (51%). Fifty-one patients (46%) had acute myeloid leukemia (AML), and 20 (18%) had acute lymphoblastic leukemia; in addition, 29 patients (27%) had non-Hodgkin lymphoma (NHL), 5 (4.5%) had multiple myeloma (MM), and 5 (4.5%) were affected by blast crisis of chronic myeloid leukemia. Eighty-eight patients (80%) received chemotherapy, whereas 13 (12%) and 9 (8%) underwent autologous stem cell transplantation and allogeneic stem cell transplantation, respectively. High-dose cytosine arabinoside was administered to 44 patients (40%).

Table 1. Patient Characteristics and Risk Factors
No. of patients (%) Median (range)
Age (yrs) 54 (20–75)
 ≤ 60 66 (60)
 > 60 44 (40)
Gender
 Male 56 (51)
 Female 54 (49)
Underlying disease
 AML 51 (46)
 ALL 20 (18)
 NHL 29 (26)
 MM 5 (4.5)
 BC-CML 5 (4.5)
Disease statusa
 Untreated 27 (24.5)
 Recurrent 28 (25.5)
 In CR 23 (21)
 Progressive 22 (20)
 In PR 10 (8)
Phase of therapy
 Induction 75 (68)
 Consolidation 29 (26)
 Maintenance 6 (6)
Type of therapy
 Chemotherapy 88 (80)
 Autologous SCT 13 (12)
 Allogeneic SCT 9 (8)
Type of chemotherapy
 Other 66 (60)
 High-dose cytosine arabinoside 44 (40)
Comorbidity
 None 82 (74.5)
 COPD 9 (8)
 Heart failure 7 (6)
 Diabetes 5 (4.5)
 Liver failure 3 (3)
 Chronic renal failure 2 (2)
 Other 2 (2)
Previous colonization with resistant pathogens
 No 89 (81)
 Yes 21 (19)
Previous infection
 No 104 (94.5)
 Yes 6 (5.5)
Previous hospitalization
 No 61 (55)
 Yes 49 (45)
Previous exposure to antibiotics
 No 64 (58)
 Yes 46 (42)
Days receiving antibiotics 14 (0–83)
Mucositis grade
 < 2 57 (52)
 ≥ 2 53 (48)
Neutropenia at onset of pneumonia
 No 30 (27)
 Yes 80 (73)
Days of neutropenia before onset of pneumonia
 ≤ 10 46 (42)
 > 10 64 (58)
Trend with respect to WBC
 Favorable 45 (41)
 Unfavorable 65 (59)
Growth factor use
 Yes 97 (88)
 No 13 (12)
Immunosuppressive therapy
 No 83 (75.5)
 Yes 27 (24.5)
CVC in place
 No 18 (16)
 Yes 92 (84)
Antibiotic prophylaxis
 Yes 69 (63)
 No 14 (37)
Antifungal prophylaxis
 Yes 87 (79)
 No 23 (21)
Antiviral prophylaxis
 Yes 9 (8)
 No 101 (92)
Hemoglobin level (g/dL) 8.6 (4.1–13.5)
 > 8.5 61 (55.5)
 ≤ 8.5 49 (44.5)
WBC count (per μl) 125 (0–26640)
 < 100 48 (44)
 100–500 32 (29)
 > 500 30 (27)
PLT count (per μl) 15000 (1000–292000)
 > 20,000 38 (34.5)
 ≤ 20,000 72 (65.5)
CRP level (mg/dL) 10.1 (0.3–66.0)
 ≤ 10 57 (52)
 > 10 44 (40)
 ND 9 (8)
Serum albumin level (g/dL) 3.2 (1.5–4.8)
 ≥ 3 69 (63)
 < 3 38 (34.5)
 ND 3 (2.5)
Serum immunoglobulin level (g/dL) 0.7 (0–3)
 ≥ 0.4 67 (61)
 < 0.4 35 (32)
 ND 8 (7)
Body temperature (°C)
 ≤ 39 63 (57)
 > 39 47 (43)
Associated sepsis
 No 78 (71)
 Yes 32 (29)
Hypotension
 No 98 (89)
 Yes 12 (11)
Shock
 No 105 (95.5)
 Yes 5 (4.5)
DIC
 No 104 (94.5)
 Yes 6 (5.5)
Respiratory failure
 No 105 (95.5)
 Yes 5 (4.5)
Renal failure
 No 99 (90)
 Yes 11 (10)
Mental status abnormality
 No 101 (92)
 Yes 9 (8)
  • AML: acute myeloid leukemia; ALL: acute lymphoblastic leukemia; NHL: non-Hodgkin lymphoma; MM: multiple myeloma; BC-CML: blast crisis of chronic myeloid leukemia; CR: complete remission; PR: partial remission; SCT: stem cell transplantation; COPD: chronic obstructive pulmonary disease; WBC: white blood cell; PLT: platelet; CVC: central venous catheter; CRP: C-reactive protein; ND: not determined; DIC: disseminated intravascular coagulation.
  • a Before chemotherapy regimen after which pneumonia developed.

Comorbidities, the most common of which were chronic obstructive pulmonary disease (COPD), heart failure, and diabetes, were assessed in 28 patients (25.5%). Before the start of chemotherapy, colonization with virulent pathogens had occurred in 21 patients (19%), 49 patients (45%) had been hospitalized, and 46 patients (42%) had been exposed to broad-spectrum antibiotic therapy. During neutropenia following treatment, antibacterial and antifungal prophylaxis were administered to 63% and 79% of patients, respectively. At the onset of fever with pulmonary infiltrates, neutropenia was present in 73% of patients, Grade ≥ 2 mucositis was present in 48% of patients, and hypogammaglobulinemia was present in 32% of patients. WBC counts exhibited a favorable trend for 45 patients (41%), serum albumin levels were < 3 g/dL for 38 patients (34.5%), and CRP levels were > 10 mg/dL for 44 patients (40%).

Etiology

For 51 of 110 patients (46%) who had fever with lung infiltrates, a microorganism considered to be the cause of this complication was isolated premortem. The most common sites of microorganism isolation were blood (59%) and lung (28%). Most lung infiltrates (74.5%) were caused by bacteria, which typically were gram negative, whereas fungi were isolated in 7 patients (13.7%); however, according to the EORTC/MSG criteria,20 3, 3, and 14 patients, respectively, were classified as having proven, probable, and possible fungal pneumonia. In six cases, it was likely that lung infiltrates were attributable to CMV infection, as CMV was isolated in four of these patients, and all six had positive antigenemia findings and a radiologic pattern that was consistent with CMV-related pneumonia. Data on pathogens isolated premortem in patients with lung infiltrates are summarized in Table 2. Comparison of patients with lung infiltrates in which gram-negative bacteria or fungi were isolated with all other patients revealed a significantly greater proportion of patients with CRP levels > 10 mg/dL in the former group (70% vs. 29%; P < 0.0001). In contrast, serum albumin levels were not associated with presumed etiology (P = 0.451). Furthermore, there was no association between low serum albumin levels and CRP levels > 10 mg/dL (P = 0.251). Finally, no difference in terms of CRP levels was found between patients with neutropenia and those without (P = 0.470), and CRP levels also were not associated with mucositis grade (P = 0.393).

Table 2. Presumptive Etiology of Pulmonary Infiltrates
Microorganism type No. of patients (%)
Gram-positive 10 (19.5)
Enterococcus spp. 7 (13.5)
Streptococcus mitis 2 (4)
Staphylococcus aureus 1 (2)
Gram-negative 24 (47)
Pseudomonas aeruginosa 11 (21)
Escherichia coli 7 (14)
Stenotrophomonas maltophilia 5 (10)
Klebsiella pneumoniae 1 (2)
Mycobacterium spp. 2 (4)
Chlamydia spp. 1 (2)
Mycoplasma spp. 1 (2)
Virus
 Cytomegalovirus 6 (11.5)
Fungus 7 (14)
Aspergillus spp. 4 (8)
Fusarium spp. 2 (4)
Mucor spp. 1 (2)

Clinical Course, Response to Treatment, and Outcome

One hundred six patients received chest X-rays before CT scanning; abnormalities were detected in 64 (60%) of these patients. The most common radiologic patterns for pulmonary infiltrates detected on chest CT were as follows: consolidative (71%), interstitial (44%), ground-glass (36%), and nodular (27%). Alveolar patterns were documented in 46% of patients, interstitial-alveolar patterns were noted in 40%, and interstitial patterns were found in 14%. In contrast, halo signs and cavitations were rarely detected (in 4.5% and 7% of patients, respectively). During the course of pneumonia, 18% of patients developed hypotension, 13% experienced shock, 10% experienced renal failure, 5.5% developed DIC, 23% experienced abnormalities in mental status, and 34.5% experienced respiratory failure, which had occurred in only 5 patients (4.5%) at diagnosis of lung infiltrates. Antibiotic therapy with a single, broad-spectrum agent (ceftazidime, cefepime, imipenem, meropenem, or piperacillin-tazobactam) was administered to 19 patients (17%), whereas 91 (83%) received one of the above mentioned agents in combination with amikacin. Empiric antibiotic therapy was successful for only 7 patients (6%). The remaining patients required modification of therapy (86%) and/or the addition of amphotericin B (60%; primarily as a part of empiric therapy) or ganciclovir (17%) to the antibiotic regimen. Overall, responses were achieved in 78 patients (71%). The crude rate of mortality due to fever with pulmonary infiltrates was 23% (n = 25), and the 90-day survival probability was 79% (Fig. 1). Eighteen patients died of respiratory failure, three died of heart failure, three died of multiorgan failure, and one died of DIC. Autopsy was performed for 10 patients (40%); all 7 patients in whom microorganisms had been isolated premortem had the same microorganisms found in postmortem lung specimens. Of the three patients in whom microorganisms had not been isolated premortem, only one had microorganisms (Aspergillus spp.) detected in a postmortem lung specimen.

Details are in the caption following the image

Ninety-day survival data for the 110 patients with fever accompanied by pulmonary infiltrates (90-day survival rate, 79%).

Prognostic Analysis and Risk Model

In the first step of univariate analysis, we found significant associations between adverse outcome and hypotension, shock, respiratory failure, acute renal failure, DIC, and alteration of mental state (data not shown); however, these conditions were noted in less than 5% of all patients at the onset of fever with lung infiltrates. Therefore, these conditions were not considered in subsequent analyses.

In the second step of univariate analysis, we tested potential associations between variables listed in Table 3 and favorable outcome following an episode of pulmonary infiltrates (which was chosen as the reference variable). We found that favorable outcome was statistically associated with age ≤ 60 years, male gender, complete remission achieved before the current round of chemotherapy, phase of therapy other than induction, absence of previous infection, a favorable trend with respect to WBC count, antifungal prophylaxis, CRP levels ≤ 10 mg/dL, albumin levels ≥ 3 g/dL, and immunoglobulin levels ≥ 0.4 g/dL; no other factor that was evaluated affected outcome for patients with pulmonary infiltrates (Table 3). Multivariate analysis identified only a favorable trend with respect to WBC count, CRP levels ≤ 10 mg/dL, and albumin levels ≥ 3 g/dL as factors associated with favorable outcome for patients with lung infiltrates (Table 4). These parameters were included in a risk model that identified 4 groups of patients with significantly different outcomes; in particular, patients with ≤ 1 favorable prognostic factor and those with > 1 favorable prognostic factor exhibited considerable differences in outcome (Table 5). Therefore, patients were divided into 2 groups—those with 0–1 favorable prognostic factors and those with 2–3 favorable prognostic factors (Table 5). Favorable outcomes were noted in 95% of patients with 2–3 favorable prognostic factors and in 46% of patients with 0–1 favorable prognostic factors (OR, 19; 95% CI, 5–70; P = 0.0001). The model had a specificity of 88%, a sensitivity of 75%, a PPV of 95%, and an NPV of 54%. In addition, the ROC curve indicated that the discriminatory power of the model was 80% (95% CI, 70–90%). Survival probabilities at 90 days according to risk group are summarized in Figure 2.

Table 3. Univariate Analysis of Outcome for Patients with Pneumonia
Characteristic No. of patients Survival rate (%) OR P
Age (yrs)
 ≤ 60 66 82 1.32 0.164
 > 60 44 70.5
Gender
 Male 56 84 1.53 0.090
 Female 54 70
Underlying disease
 Other 61 77 1.01 0.950
 AML 49 78
Disease statusa
 In CR 23 91 3.08 0.071
 Other 87 74
Phase of therapy
 Other 35 87 2.27 0.053
 Induction 75 72
Comorbidity
 No 82 79 1.12 0.393
 Yes 28 71
Previous colonization with resistant pathogens
 No 95 73 1.03 0.695
 Yes 15 78
Previous infection
 No 59 83 1.44 0.120
 Yes 51 71
Previous hospitalization
 No 61 76 1.06 0.569
 Yes 49 73
Previous exposure to antibiotics
 No 64 80 1.20 0.502
 Yes 46 75
Therapy
 Chemotherapy 88 77 1.02 0.722
 Transplantation 22 82
Type of chemotherapy
 Other 66 82 1.32 0.353
 High-dose cytosine arabinoside 44 74
Mucositis grade
 < 2 57 81 1.23 0.373
 ≥ 2 53 74
Neutropenia at onset of pneumonia
 No 30 83 1.47 0.353
 Yes 80 75
Days of neutropenia before onset of pneumonia
 ≤ 10 46 79 1.02 0.862
 > 10 74 76
WBC count trend
 Favorable 45 90 4.11 < 0.0001
 Unfavorable 65 53
Growth factor use
 Yes 97 76 1.21 0.501
 No 13 85
Immunosuppressive therapy
 No 83 70.5 1.14 0.324
 Yes 27 79.5
CVC in place
 No 18 78 1.05 0.576
 Yes 92 72
Antibiotic prophylaxis
 Yes 69 81 1.26 0.207
 No 41 71
Antifungal prophylaxis
 Yes 87 82 2.8 0.035
 No 23 61
Radiologic pattern
 Alveolar 54 80 1.15 0.562
 Interstitial-alveolar 56 75
Prechemotherapy chest X-ray findings
 Positive 64 79 1.12 0.393
 Negative 42 71
Hemoglobin level (g/dL)
 > 8.5 61 78 1.01 0.950
 ≤ 8.5 49 77
PLT count (per μl)
 > 20,000 38 84 1.57 0.207
 ≤ 20,000 72 74
CRP level (mg/dL)
 ≤ 10 57 91 8.61 < 0.0001
 > 10 44 54
Serum albumin level (g/dL)
 ≥ 3 69 85.5 3.85 0.003
 < 3 38 60.5
Serum immunoglobulin level (g/dL)
 ≥ 0.4 67 86 3.66 0.012
 < 0.4 35 63
Body temperature (°C)
 ≤ 39 63 79 1.08 0.754
 > 39 47 76
  • OR: Odds Ratio referred to the first categories; AML: acute myeloid leukemia; CR: complete remission; WBC: white blood cell; CVC: central venous catheter; PLT: platelet; CRP: C-reactive protein.
  • a Before chemotherapy regimen after which patient developed pneumonia.
Table 4. Multivariate Analysis of Outcome for Patients with Pneumonia
Characteristic β (SE) OR (95% CI) P
CRP levels ≤ 10 mg/dL 1.544 (0.553) 4.6 (1.6–13.8) 0.001
Serum albumin level ≥ 3 g/dL 1.176 (0.416) 3.2 (1.4–7.3) 0.004
Favorable WBC trend 1.724 (0.620) 5.6 (1.7–18.9) 0.001
  • SE: standard error; OR: odds ratio; CI: confidence interval; CRP: C-reactive protein; WBC: white blood cell.
Table 5. Risk Model of Outcome for Patients with Pneumonia
Prognostic score No. of patients No. of patients cured (%) P
Scoring System 1
 3 24 24 (100)
 2 36 33 (92) 0.268
 1 29 17 (59) 0.0231
 0 12 2 (17) 0.014
Scoring System 2
 2–3 60 57 (95)
 0–1 41 19 (46) < 0.0001
Details are in the caption following the image

Survival of patients with pulmonary infiltrates according to number of positive prognostic factors. (A) Patients with 0 positive factors vs. patients with 1 positive factor vs. patients with 2 positive factors vs. patients with 3 positive factors. P = 0.0061 for patients with 0 positive factors vs. patients with 1 positive factor; P = 0.024 for patients with 1 positive factor vs. patients with 2 positive factors; and P = 0.1514 for patients with 2 positive factors vs. patients with 3 positive factors. (B) Patients with 0–1 positive factors vs. patients with 2–3 positive factors. P < 0.0001.

DISCUSSION

Pneumonia is one of the most feared complications of chemotherapy for hematologic malignancies. Lung infiltrates attributable to fungi, gram-negative bacteria, and viruses can actually be responsible for mortality rates of up to 50%22-24; however, mortality rates for patients who develop this complication typically range from 17% to 45%.12, 15, 17-19 Like patients with febrile neutropenia, patients who develop fever with lung infiltrates constitute a heterogeneous population, and thus, the probability of death due to this complication varies within the population. In the current study, we found that definitive parameters that could be assessed at the onset of fever with pulmonary infiltrates (e.g., trend with respect to WBC count, CRP levels, serum albumin levels) were predictive of survival. Although hypotension, shock, respiratory failure, acute renal failure, DIC, and alteration of mental state also had prognostic value, these clinical features were rarely present at the onset of pulmonary infiltrates. Furthermore, the impact of these conditions on outcome could overshadow the statistical impact of other factors. When abnormal vital signs are present at the onset of lung infiltrates, the prognosis must be considered poor, regardless of the score yielded by our model. Nonetheless, the role of a prognostic model is to identify patients who are at risk for the development of such abnormalities, which typically lead to death.

Using three parameters that were assessable at the onset of lung infiltrates, we constructed a simple model for predicting the survival of patients with fever accompanied by lung infiltrates. This model had a high level of specificity and a high PPV in identifying low-risk patients, whereas it appeared to be less predictive of mortality for high-risk patients. Nonetheless, these results satisfied the goal of the study, which was to accurately identify a group of low-risk patients so that poor outcomes could be prevented. Our results suggest that patients predicted to have a low risk of death constitute a homogeneous population, whereas high-risk patients, who exhibit nonnegligible survival rates, do not.

CRP normally is produced by the liver in response to tissue injury or infection.25 Manian et al.26 found that in patients with neutropenia, CRP levels increased significantly during the 48–72 hours preceding the diagnosis of pneumonia. Therefore, as is also supported by our findings, patients with neutropenia may be able to produce an inflammatory response to appropriate stimuli. Furthermore, several studies have demonstrated associations between CRP levels and levels of inflammatory cytokines such as tumor necrosis factor, interleukin-6, and interleukin-1.27-31 Because, like others,32 we found a significant correlation between high CRP levels and pathogens such as gram-negative bacteria and fungi, we posited that the increase in CRP levels may be a manifestation of the ‘cytokine storm’ that is triggered by these microorganisms and is responsible for tissue damage. Nonetheless, we believe that it is more likely that high CRP levels reflect the severity of tissue injury and, consequently, the severity of lung damage caused by these microorganisms. In fact, cell deficiencies such as those occurring in patients with unfavorable WBC count trends, either due to the temporary aplasia that often develops following chemotherapy or due to the failure of the underlying disease to respond to therapy, are suggestive of impairment of the phagocytic defense (as carried out by macrophages and neutrophils) against infectious agents, which could prevent a suitable inflammatory response. This type of impairment, along with the impairment of B-cell and T-cell function, prevents innate and adaptive immune responses against infectious microorganisms,33 which encounter few obstacles in producing harmful pulmonary and systemic effects. Although WBC recovery is a significant prognostic factor for patients with all types of infections, it is most significant for patients with infectious complications such as lung infiltrates caused by gram-negative bacteria or fungi. This is the likely explanation for why the trend with respect to WBC count has prognostic significance in our study.

Although a recent study revealed that severe inflammatory responses can cause glomerular damage followed by more dramatic decreases in serum albumin levels,31 we did not find a significant correlation between high CRP levels and low serum albumin levels, the latter being a potential marker of nutritional and performance status according to other scoring systems.34, 35 Alternatively, low serum albumin levels may serve as a risk factor, as they are associated with an increased risk of hematologic toxicity following chemotherapy.36 Published risk models generated by studies involving the general population of individuals with febrile neutropenia2-8 did not identify the same factors that were identified in the current study. This finding indirectly strengthens the hypothesis that the current model is a relatively specific one.

Outcomes for patients with hematologic malignancies who develop fever accompanied by lung infiltrates depend on interactions among several factors, such as the etiology of the lung infiltrates, the status of patient, the immune response, and the toxicity and efficacy of treatment for the underlying disease. The results of the current study indicate that all of these factors can be accurately represented by parameters such as CRP levels, serum albumin levels, and trend with respect to WBC count.

The current prognostic model suggests that the absence of a given favorable prognostic factor at the onset of fever with lung infiltrates is not significantly associated with poor outcome; however, when two favorable factors are absent, outcomes may be poor. Outcomes are consistently poor when all three favorable prognostic factors are absent, although it is likely that the observed differences in outcome among high-risk patients depend on the use (or lack thereof) of effective supportive therapy to prevent death before recovery from aplasia.

Few studies have addressed the assessment of factors predicting outcome for patients with hematologic malignancies who develop fever accompanied by pulmonary infiltrates.12, 15, 17-19 Many of these studies found the achievement of complete response to be strongly associated with favorable outcome12, 15, 18; however, data regarding this parameter are rarely available at the onset of fever with lung infiltrates, and therefore, complete response is not a prospectively useful parameter. Two studies found that the trend with respect to WBC count was associated with outcome for patients with pneumonia, as was also reported in the current study.15, 17 Unfortunately, in the first of those two studies,15 the term trend in WBC count was not defined, and in the second study,17 this parameter was difficult to assess. In the prognostic scoring system developed by Ewig et al.,19 a heart rate (HR)-to–systolic blood pressure (SBP) ratio ≥ 1.2 and a radiographic score ≥ 3 were the only risk factors associated with poor outcome for patients with pulmonary infiltrates; however, at the onset of pneumonia, an HR-to-SBP ratio ≥ 1.2 was observed in only 10 of 53 patients. Furthermore, SBP and HR depend strongly on the time of measurement and on the operator, and radiographic score depends on the time at which radiologic assessment is performed. In contrast to the scoring system of Ewig et al.,19 the current model was constructed on the basis of data obtained from a larger number of patients and included parameters that were prospectively available and were neither time dependent nor operator dependent. Nonetheless, we are aware that the current study is also retrospective and that its results require confirmation by larger prospective studies.

In conclusion, we proposed a novel risk model to predict outcome for patients with hematologic malignancies who develop pulmonary infiltrates. In this model, the prognostic score is based on objective variables and is easy to calculate and apply in clinical practice, thereby allowing the selection of patients for various diagnostic and therapeutic strategies. Thus, we believe that prospective validation of our model may be worthwhile, as it has the potential to lead to improvements in the management of fever and pulmonary infiltrates in patients with hematologic malignancies.