Introduction

Infectious complications in patients with hematological malignancies are still a major cause of morbidity and mortality despite significant advances in diagnostic techniques and antimicrobial therapy [1]. Infections may go undetected in cancer patients due to the absence of any obvious signs and symptoms, except for fever [2]. A comprehensive understanding of the characteristics of hematological cancer patients with infection, and identification of predictors of mortality in this group will help clinicians to undertake timely preventive measures in high-risk patients.

In this study, we describe the characteristics of patients with hematological malignancies who were referred to the Infectious Diseases (ID) team for suspected infection, and the predictors of mortality in these patients.

Materials and methods

The study was conducted at Hacettepe University Hospital, a 900-bed tertiary care university hospital in Ankara, Turkey. Hacettepe University hospital has an active hematology–oncology unit and approximately half of the inpatients have underlying malignancy. A strong collaboration exists among the ID, Hematology, and Oncology Units. All cancer patients suspected of an infection were routinely examined by the ID team.

A retrospective chart review was conducted for all episodes of adult patients (17 years or older) with hematologic malignancies who consecutively consulted with the ID Team for fever or any other signs or symptoms of infection between January 1, 1997–December 31, 2001. Data were collected on some demographic and disease characteristics, and hospital outcome of the study participants, using a standardized data collection form. Infectious complications identified by the ID team were categorized into three groups: “clinically documented infections,” “microbiologically documented infections,” and “fever of unknown origin” (FUO). In the group, in 59.2% (N=675) of the episodes, patients were neutropenic. Of the 239 FUO episodes, 209 (87.4%) were neutropenic.

Statistical analysis included frequency and percent distributions, and logistic regression modeling. Statistical significance of differences between groups was tested using chi-square test, and Student’s t test when appropriate. Odds ratios and 95% confidence intervals were calculated to evaluate the size of effect and its significance. Univariate and bivariate analyses were conducted for a total of 1,132 episodes, but multivariate analysis of predictors of mortality was limited only to the first episode of each participant (n=641). All statistical analyses were conducted using SPSS version 10.0 statistical software package.

Results

A total of 1,132 consultations were done over a period of 5 years for 641 patients. Of the patients, 59.4% were male and the mean (±standard deviation) age of the study participants was 47.9±1.4 years.

The most common underlying disease was non-Hodgkin’s lymphoma (30.9%), followed by acute myelogenous leukemia (26.2%), multiple myeloma (10.9%), and Hodgkin’s disease (8.1%) (Table 1).

Table 1 Baseline characteristics of study participants (N=641)
Full size table

In 30.7% of the episodes the underlying disease was in remission at the time of infection diagnosis, in 31.2% it was newly diagnosed, and in 35.6% it was either refractory to treatment or had relapse. The patient was neutropenic in 60% of the episodes.

In 116 (10.2%) consultations, no infection was detected, and in an additional 241 episodes (18.5%) the patients were febrile, although no cause was identified, and were considered as FUO. More than one (maximum 3) infection was detected in 278 (24.6%) episodes (Table 2).

Table 2 Infectious complications identified during the study period*
Full size table

Of 1,306 infections detected, 567 (43.4%) were documented clinically, and 498 (38.1%) microbiologically. Pneumonia was the most common clinically documented infection and was detected in 173 episodes.

Bloodstream infections were present in 134 episodes, 122 were bacteremias and 12 fungemias; 56.5% of bacteremias were caused by gram-negative microorganisms (Table 3).

Table 3 Pathogens isolated in 134 episodes of bloodstream infections
Full size table

In 22 of the episodes, we clinically detected herpes labialis, although no other viral infections were reported, possibly due to our inability to microbiologically detect viral infections.

Invasive mould infections were diagnosed in 51 episodes, including 20 pulmonary involvement.

A total of 163 patients (25.4%) died during the infectious episode. The fatality rate was even higher in the presence of invasive aspergillosis, and 13 of the 29 patients with invasive aspergillosis died, corresponding to a fatality rate of 44.8%. In logistic regression analysis, the presence of pneumonia (OR 7.56, 95% CI 4.84–12.49), invasive fungal infection (OR 4.12, 95% CI 1.78–9.55), relapse or recent diagnosis of underlying disease (OR 2.82, 95% CI 1.53–5.21), and neutropenia (OR 2.70, 95% CI 1.70–4.31) were identified as statistically significant predictors of mortality, controlling for age and gender (Table 4).

Table 4 Logistic regression model for predictors of mortality in patients with hematological malignancies and infectious complications
Full size table

Discussion

Major advances in medicine over the recent decades have prolonged the survival of cancer patients, although they can not always be provided with a disease-free life. The price paid for “death” of cancer cells is the major risk of compromised host defense mechanisms, leading to a patient population prone to infectious complications with high morbidity and mortality. Numerous studies have reported on infectious complications in these patients; these studies were based either on a certain type of infection (i.e., bacteremia, fungemia) [3, 4] or the efficacy of a certain antimicrobial regimen [5, 6]. There are a few reports on the overall distribution of infectious episodes in patients with hematological malignancies, the outcome and the impact of several factors on the outcome [7, 8]. In this study, we summarize the experience of an ID team in more than 1,000 consultations involving patients with hematological malignancies in an attempt to increase the awareness of infectious disease specialists and oncologists regarding potential infections and related risk factors they might face in the management of such patients.

There are several aspects of our findings that merit attention. First, pneumonia is a major infectious complication in this patient population. One third of our patients with a clinical documentation of infection had pneumonia. The lung appears to be highly vulnerable to damage by chemotherapy and is extensively susceptible to infection. In addition, the presence of pneumonia was an ominous sign of poor prognosis. Second, bloodstream infections occur in a substantial proportion of episodes (11.8%). Gram-negative bacilli were identified as pathogens in 52.5% of episodes with bloodstream infection. This finding parallels with the results obtained by Homsi et al. [7] in a population of nonneutropenic patients with solid tumors (48.0%) and by Feliu et al. [9], in a similar study that focused on febrile neutropenic patients with cancer. In contrast, Marie et al. [10] found in their population of severely neutropenic patients that 63 out of 105 isolated organisms were gram-positive in type. They linked the predominance of gram-positive infections in these patients to repeated courses of chemotherapy, extensive use of indwelling catheters, and long duration of neutropenia. In the mid-1980s, there was a steady increase in gram-positive infections. Although several factors—including widespread use of catheters, longer duration of neutropenia, mucositis caused by more aggressive chemotherapeutic agents and prophylaxis targeted against gram-negative organisms with the use of fluoroquinolones—have been proposed as contributors to this change in the pattern of causative agents, the underlying reasons are not absolutely clear [11]. However, in recent years the number of cases of gram-negative bacteremia has risen again [12].

In our study population, the observation that gram-negative bacilli is an important cause of bacteremia could at least be partially explained by our hospital’s ID policy of not using fluoroquinolones prophylactically.

Incomplete chart information limited our ability to detect frequency of central venous catheterization (CVC) use in the study population; thus, the potential role of CVC use on bloodstream infection could not be investigated.

The incidence of invasive fungal infections has increased substantially during the past 30 years [13]. Prolonged neutropenia is a major, well-known risk factor for fungal infection [14]. Other host-related factors include broad-spectrum antibiotic use, corticosteroid therapy, indwelling venous catheters, mucositis. Environmental exposure is also important in the case of invasive aspergillosis [15]. Several of such risk factors might have played a role in the occurrence of invasive fungal infections in 51 episodes in our study population. We believe that the ongoing renovation activities in our hospital might be a major contributor to the predominance of invasive mould infections compared to candidemia in our patient population, despite the fact that we do not use routine antifungal prophylaxis except for the allogeneic hematopoietic stem-cell transplantation setting.

Our findings on the prognostic factors of the presence of neutropenia, pneumonia and fungal infections and the absence of remission appeared as poor prognostic indicators in patients with hematalogical malignancies confirm the findings of previous research in this field. The mortality due to infection in our study is 25.4%, which is higher than the rate reported usually. The high fatality rate (44.8%) of invasive aspergillosis and high rate of pneumonia, which was detected in one third of the episodes, could at least partially explain the high mortality rate in our study population.

In conclusion, infection has a major impact in the management of patients with hematalogical malignancies. The presence of neutropenia, pneumonia and fungal infections and absence of remission should alert the physician to take a more meticulous approach in the treatment of these patients.