Utility of bronchoalveolar lavage in diagnosing respiratory tract infections in patients with hematological malignancies: are invasive diagnostics still needed?

Background Patients treated for hematological malignancies have an increased risk of serious infections. Diagnosis and prompt initiation of therapy are essential. Bronchoalveolar lavage (BAL) is a well-established investigation for identifying the cause of pulmonary infiltrates in immunocompromised patients. The aim of the study was to determine the diagnostic yield of BAL in patients treated for hematological malignancies and how often it contributed to a modification of the anti-infectious therapy. Methods We reviewed records from 151 consecutive BAL procedures in 133 adult patients with hematological malignancies, treated at a tertiary hematology unit from 2004 to 2013. Extensive microbiological work-ups on BAL samples had been performed according to a standardized protocol. Results A microbiological finding causing the infectious episode could be identified in 59 (39%) cases. In 44 (29%) of the cases, results from BAL had an impact on clinical management either by contributing to a specific diagnosis (25%) or by leading to cessation of ongoing microbiological therapy. The most common diagnoses were invasive pulmonary aspergillosis (IPA) and Pneumocystis jirovecii pneumonia (PJP). Diagnoses of IPA and PJP were based on results from BAL in 65% and 93% of cases, respectively. Several microbiological tests on BAL samples rendered no positive results. Complications were few and mainly mild. Conclusion BAL is still important for either verifying or excluding some of the most important respiratory tract pathogens in patients with hematological malignancies, particularly IPA and PJP. Standardized procedures for BAL sampling should be continually revised to exclude unnecessary microbiological tests.


Introduction
Patients treated for hematological malignancies, in particular those undergoing allogeneic (allo-) hematopoietic stem cell transplantation (HSCT), have an increased risk of serious infections. Neutropenia, decreased cellular immunity, hypogammaglobulinemia, chemotherapy-induced damage to mucosal barriers, and the frequent use of central venous lines predispose to such infections (1)(2)(3)(4). The panorama of pathogenic micro-organisms affecting this patient group is broad and includes such opportunistic micro-organisms as Aspergillus spp., Candida spp., Mycobacterium spp., Pneumocystis jirovecii, and respiratory viruses (1)(2)(3)(4). In patients with severe neutropenia related to hematological disease, or its treatment, 40% to 60% develop pulmonary infiltrates (1,5,6), and pneumonia is the primary cause of mortality not directly related to the hematological malignancy itself. However, in the majority of these cases the causative pathogen remains undetected (4,7,8).
The major aim of the present study was to determine the diagnostic yield of BAL in consecutive adult patients treated in a tertiary hematology unit over a 10-year period. Further aims were to evaluate the usefulness of a standardized diagnostic procedure and its safety.

Material and methods
Inclusion criteria and BAL procedure Indication for BAL was pulmonary infiltrates or fever with or without respiratory tract symptoms but no radiological findings, indicating infection but with no obvious etiology, in immunocompromised patients with hematological malignancies. Prophylactic platelet transfusions were given prior to BAL in cases of severe thrombocytopenia. BAL was performed according to a standardized procedure by a limited number of experienced specialists in the day-care unit of the Department of Pulmonary Medicine, Uppsala University Hospital. BAL samples were analyzed using a standard protocol, which was updated in 2007 and 2012 (see Appendix 1). All samples were analyzed locally with the exception of GM, PCR for Aspergillus spp., and tests for Pneumocystis jirovecii, which were sent to an external academic laboratory for investigation.

Study design
All BAL procedures in adult patients with hematological malignancies treated at the Department of Hematology, Uppsala University Hospital from 2004 to 2013 were retrospectively reviewed in this study. Patients were identified through the local bronchoscopy register. Each BAL procedure was considered as one case/patient. Patients' records were reviewed for demographics, clinical characteristics (Table 1), radiological findings, and test results from BAL and from 'non-invasive' microbiological studies ( Table 2). Changes in management of anti-infectious therapy following BAL and major procedure-related complications were also recorded. The study was approved by the Regional Ethic Review Board in Uppsala.

Definitions
Results from a BAL sampling were considered clinically important if they altered clinical management of antibacterial, antiviral, or antifungal therapy, as recorded in the patient record by the attending physician. Invasive pulmonary aspergillosis (IPA) was defined according to the 'Revised Definitions of Invasive Fungal Disease' of the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) Consensus Group (15). Complications of the procedure were considered if they occurred during the bronchoscopy or within the next 24 hours.

Statistics
Results were analyzed using descriptive methods and presented as median and range. Categorical variables and diagnostic yields were expressed as frequencies and percentages.

Findings in BAL
Most frequent positive findings were seen on general cultivation for different species of fungi (22% of performed tests), PCR for Candida (20% of performed tests), and PCR for Pneumocystis jirovecii and cytomegalovirus (CMV) with positive findings in 19% and 17% of the performed tests, respectively. However, the treating physician considered a majority of the findings as contaminations or otherwise clinically irrelevant not leading to any change of diagnosis or treatment. Several investigations such as PCR for Legionella, Chlamydophilia, Mycoplasma, and Pneumococci rendered no positive results. Duplicate samples obtained using a protected specimen brush (PSB) did not yield any additional diagnostic information (data not shown).

Final diagnosis and impact of BAL on clinical management
In59 cases (39%) at least one microbiological agent that most probably had caused the infectious episode was identified. In 92 (61%) cases no etiological diagnosis was obtained, and these patients were finally diagnosed with pneumonia, febrile neutropenia, or lower respiratory tract infection of unknown cause ( Table 2). The most common final diagnosis was IPA (23 cases), of which 2 were classified as proven, 8 as probable, and 13 as possible. The second most frequent diagnosis (14 cases) was Pneumocystis jirovecii pneumonia (PJP).
In 38 cases (25%), results from BAL had an impact on clinical management either by establishing or contributing to the final diagnosis. In another 6 cases (4%) negative findings from BAL led to cessation of ongoing microbiological therapy (Table 2). Importantly, the diagnosis of IPA was based on results from BAL in 15 (65%) of the cases and by other methods in 8 (35%) cases. In almost all patients (13 of 14 patients) with PJP, the diagnosis was based on findings obtained from BAL (Table 3).

Safety
Complications were reported in 20 (13%) of the cases and were mainly mild. No lethal complications and no cases of pneumothorax were observed. Minor bleeding occurred in four (3%) of the subjects.

Discussion
The clinical value of BAL as a routine investigation in immunocompromised patients with pulmonary infiltrates of unknown etiology has been questioned for several reasons. First, improvements in radiology (i.e. chest CT scanning) and non-invasive microbiological diagnostics performed on blood or sputum may reduce the need for BAL. Second, because of the invasive character of BAL, clinicians may be reluctant to order it in severely ill cancer patients prone to bleeding. Third, a substantial proportion of BAL procedures may result in findings with no or only marginal impact on the clinical management of an individual patient. These considerations prompted us retrospectively to study the diagnostic yield of 151 BAL investigations in 133 unique patients treated at our institution for hematological malignancies over a 10-year period.
Our main finding was that BAL had a direct impact on clinical management (i.e. altered antimicrobial therapy) in 25% of all cases, either by establishing or contributing to a specific etiological microbiological diagnosis or by leading to cessation of ongoing empirical antibacterial or antifungal therapy. The findings are consistent with those of previous studies (1,(5)(6)(7)(9)(10)(11). Our results support BAL as an important diagnostic method also in the future in patients with hematological malignancies with fever and pulmonary infiltrates of unknown etiology.
The diagnostic yield and clinical utility of BAL depend, among other factors, on the patient population. In our patient cohort acute leukemia was the most frequent diagnosis, and about one-third of the patients had undergone allo-HSCT. Therefore, we had two high-risk groups for opportunistic pulmonary infection with a higher likelihood of positive findings from BAL. An overall mortality rate of 59% at the end of the 10-year investigation reflects a severely ill population. About one-third of the patient population were neutropenic, and the majority had high MASCC scores, indicating high vulnerability for infectious complications (17).
Importantly, BAL contributed to the diagnosis in twothirds of the patients with IPA, and, in those diagnosed with PJP, BAL added diagnostic information in all but one case. The morbidity and mortality of IFD, especially if caused by Aspergillus spp. or Pneumocystis jirovecii, are high, and early diagnosis together with prompt initiation of antifungal therapy are critical for a favorable outcome (1,2,(18)(19)(20). Several non-invasive culture-independent methods for diagnosing IFD are available. GM has high sensitivity, and PCR arrays for Aspergillus spp. (yielding a faster result) are under development (2,19,21,22). BG, a cell wall component of most fungal species including Pneumocystis jirovecii, can be detected in blood for many different IFDs. Its usefulness is limited by low sensitivity, but a negative test is valuable for excluding PJP (23). Diagnosing IFD remains a challenge (4,19,24), but BAL is still a valuable tool in diagnosing this lethal condition in immunocompromised patients.
The low incidence of verified bacterial pneumonia might be attributable to ongoing treatment with empirical or preemptive broad-spectrum antibiotics in two-thirds of the cases. This suggests that it is crucial to perform BAL at early onset of symptoms to render meaningful results (3,9). The low incidence of viral pneumonia in our material could be because non-invasive methods, such as PCR for respiratory viruses on nasopharyngeal secretion, are used instead of BAL when this diagnosis is suspected.
Rather than an individualized sampling schedule, we used a standardized procedure with extensive microbiological testing. An important finding of our study is that several BAL analyses yielded no diagnostic information. For example, PSB is used routinely to avoid contamination (1). In the present study, samples from PSB (obtained in 97% of cases), yielded additional information in only one single case. This is consistent with findings reported by Boersma et al. (1). Moreover, the diagnostic impact varied greatly between analyses; a positive result for Pneumocystis jirovecii (immune morphology) and Aspergillus (GM) always contributed to the diagnosis of the infectious episode, whereas Candida PCR, for example, was positive in 12 (20%) of the cases, but never led to any change of diagnosis or treatment. Clearly, the usefulness and cost effectiveness of BAL-related procedures and tests need to be regularly reevaluated. No serious procedure-related complications were observed in the present study, which accords with previous reports (1,3,7,9,12,13). Interestingly, no major bleeding complications occurred, although many patients had severe thrombocytopenia. The practice of giving prophylactic platelet transfusions prior to BAL in severely thrombocytopenic patients has never been proven to be necessary, but continues to seem reasonable.
Our study has some limitations. Reflecting the retrospective nature and complexity of infection diagnostics in these patients (4,19), documentation in patient records was not always clear, making estimates of 'clinical impact' difficult. The absence of more precise criteria for performing BAL may have affected patient selection. Our standard protocol for microbiological sampling also underwent moderate changes during the study period. Nevertheless, the study describes real-life data from a well-defined Nordic cohort of consecutive patients undergoing BAL using a well-standardized procedure including extensive microbiological sampling.
We conclude that BAL still contributes to either verifying or excluding some of the important respiratory tract infections in patients with hematological malignancies, in particular IPA and PJP. Standardized procedures for BAL sampling should be continuously revised to exclude unnecessary microbiological tests. Improvements of non-invasive microbiological and radiological methods may reduce the need for performing BAL in the future.

Disclosure statement
No commercial relationships or potential conflicts exist.