Patients with acute leukemia (AL) are at high risk for developing pulmonary complications including infection, inflammatory drug reactions, tumours, alveolar hemorrhage, and cardiogenic pulmonary edema.1,2 These entities may present with non-specific symptoms (cough, fever, shortness of breath) and radiographic findings and can rapidly progress to acute respiratory failure. In patients with cancer, acute respiratory failure is the leading reason for admission to the intensive care unit (ICU) and carries a high mortality rate of up to 50%, particularly when the underlying cause cannot be identified.3,4,5 Diagnosis is critical to guide treatment, but is often challenging in this patient population. Available diagnostic tools include non-invasive techniques, such as computed tomography scans, and more invasive techniques, including flexible bronchoscopy with bronchoalveolar lavage (BAL) and lung biopsy. Flexible bronchoscopy with BAL is less invasive than lung biopsy and is used commonly in the immunocompromised population in an attempt to guide management.

Flexible bronchoscopy with BAL allows direct visualization of the airways and sampling of alveolar cells and fluid.6,7,8,9 It is considered safe and is of high diagnostic yield even in patients supported by mechanical ventilation and with underlying malignancy.1,10,11 Past studies evaluating its use in the immunocompromised population report a diagnostic yield between 32% and 66%.1,2,12,13,14 A recent systematic review comparing bronchoscopy with BAL to lung biopsy in patients with cancer or hematopoietic stem cell transplant found that bronchoscopy with BAL was superior to lung biopsy in diagnosing pulmonary infections and was associated with a lower complication risk including procedural mortality.15 In 2,792 procedures reviewed, BAL led to a diagnosis in 53% of patients and changes in management in 31%, with a low complication risk of 8%, including five cases of procedure-related death.8,15 The technique can be especially useful in diagnosing pulmonary infections, specifically opportunistic pathogens such as cytomegalovirus and Pneumocystis jiroveci.1,2,16 Identification of pathogens is important, as it allows for timely initiation of more targeted antimicrobial therapy that might not otherwise be prescribed and allows for discontinuation of unnecessary antibiotics.

Although the utility of flexibility bronchoscopy has been studied in immunocompromised populations including bone marrow transplant (BMT) recipients,17,18,19 patients with HIV,20,21 and thoracic malignancy,22 the diagnostic utility and safety of the technique among critically ill AL patients remain unclear. The primary objectives of this study were therefore to: 1) identify whether flexible bronchoscopy with BAL leads to changes in medical management and 2) describe the risk of procedure-related complications in patients with AL admitted to the ICU. Secondary objectives were to determine whether patient characteristics including age, sex, Acute Physiology And Chronic Health Evaluation (APACHE) II score, and immunosuppression status (patients undergoing active chemotherapy) were associated with the development of complications post-bronchoscopy.

Methods

This was a retrospective cohort study conducted at two academic teaching hospitals affiliated with the University of Toronto: Princess Margaret Cancer Center of the University Health Network and Sinai Health System. The Research Ethics Boards at both hospitals approved the study in October 2013.

Participants

We included consecutive patients who were 16 yr of age or older with a diagnosis of AL undergoing active medical management who were admitted to the ICU between 1 January 2007 and 31 December 2012 and underwent diagnostic flexible bronchoscopy with BAL. There were no exclusion criteria.

Data collection

Data were collected from paper and electronic hospital records. Patient demographics, baseline characteristics, details of the ICU admission, bronchoscopy procedural details, and clinical outcomes were collected for the entire cohort. The respiratory therapist assisting in each bronchoscopy recorded the respiratory rate, heart rate (HR), blood pressure, oxygenation, and ventilator settings at set intervals throughout the procedure and recorded one value per time period (before, during, and after the procedure). We recorded changes in medical management following bronchoscopy including a diagnosis of alveolar hemorrhage and identification of an infectious organism resulting in a change in antimicrobials or a therapeutic intervention. We searched for specific procedural complications including hypoxia (oxygen saturation ≤ 92% immediately post-bronchoscopy), the need for noninvasive ventilation or intubation within 48 hr and other complications that were directly related to the procedure. No data were missing for the data points of interest.

Data analysis

Descriptive statistics were calculated for all variables. Continuous variables were not normally distributed (Shapiro-Wilk statistic < 0.05) and were expressed as median (interquartile range [IQR]). Categorical variables were calculated as proportions. Vital signs immediately before, during, and after the procedure were compared using Friedman tests given the paired nature of the data. We explored the relationship between patient characteristics and whether bronchoscopy changed management or caused complications by developing a multivariable logistic regression model. Predictors included in the regression model were age, sex, immunosuppression status, and APACHE II score. These were selected a priori based on clinical experience and previous studies. Included predictors were also limited by the smallest category of the outcome variable (change in medical management, n = 32) and were not collinear (tolerance > 0.9). The Hosmer-Lemeshow test was not significant for either regression model, indicating adequate model fit (P > 0.97 and 0.92, respectively).

All statistical analyses were performed using SAS® Studio 9.3 University Edition (SAS® Institute; Cary, NC, USA). The type I error probability was 0.05 for two-sided tests of statistical significance when comparisons were made.

Results

Seventy-one patients with AL underwent bronchoscopy in the ICU during the study period. Baseline characteristics are shown in Table 1. Patients had a median [IQR] age of 55 [46-63] yr, median [IQR] APACHE II score of 22 [17-29], and 37 (52%) were female. The types of AL during various stages of treatment were: acute myeloid leukemia (61%), acute lymphoid leukemia (21%), acute promyelocytic leukemia (6%), and myelodysplastic syndrome (4%), and 8% had an alternate malignant hematologic diagnosis.

Table 1 Baseline characteristics of acute leukemia patients undergoing bronchoscopy and bronchoalveolar lavage in the intensive care unit
Full size table

The most common indication for ICU admission was respiratory failure (51 patients, 72%), followed by sepsis (14 patients, 20%). On the day of ICU admission, patients had a median [IQR] absolute neutrophil count of 0.34 [0.02-1.80] × 109 cells·L−1 and platelet count of 26 [11-47] × 109 cells·L−1. Of the 71 patients, 28 (39%) were receiving steroids, and 32 (45%) were receiving chemotherapy. The median [IQR] ICU length of stay was 11 [6-22] days, and 36 (58%) patients survived their ICU admission. Survival at one year was 31%.

The median [IQR] day that bronchoscopy with BAL was performed was day 2 [2-6] of ICU admission, and 60 (85%) patients were intubated at the time of bronchoscopy. The median [IQR] oxygen saturation and F i O2 on the day of bronchoscopy were 96 [94-98] % and 45 40-60] %, respectively. Vital signs immediately before, during, and after bronchoscopy are represented in Table 5 and the Figure. We found there was a significant difference with HR during the procedure being the highest; the median [IQR] HR before, during, and after bronchoscopy was 107 [97-122] beats·min−1, 113 [98-124] beats·min−1, and 109 [93-121] beats·min−1, respectively (P = 0.03). Other vital signs including oxygen saturation and mean arterial pressure remained stable throughout the procedure (Table 5, Figure).

Figure
figure 1

Patient vital signs before, during, and after bronchoscopy. Data points represent the median and 95% confidence interval for the readings during each time period. Pre = vital signs immediately before bronchoscopy; Bronch = vital signs during bronchoscopy; Post = vital signs immediately after bronchoscopy; *heart rate was highest during the procedure (P = 0.03)

Full size image

Overall, bronchoscopy with BAL was associated with a change in medical management in 32 of the 71 cases (45%). Changes to management included changes to antimicrobial therapy (including starting, discontinuing, or changing antimicrobials) based on culture results, the establishment of an alternative diagnosis, or a local therapeutic intervention based on bronchoscopy findings. An organism was identified from BAL in 19 cases (27%). The most common organisms identified included Pseudomonas aeruginosa (n = 3), Herpes simplex virus-1 (n = 2), Klebsiella pneumoniae (n = 2), Aspergillus fumigatus (n = 2), Aspergillus flavus (n = 2), and Candida albicans (n = 2) (Table 2). Of these 19 cases, 17 (89%) had changes to their antimicrobials to target the causative organism, while in two cases the organisms identified were presumed to be contaminants and no changes to antimicrobials were made. In three cases, changes to antimicrobial therapy were made despite negative cultures—two cases were treated as bacterial pneumonia based on the presence of pus and thick secretions in the airways, while a third was treated as a fungal pneumonia based on the appearance of black lesions in the airways. Empiric antimicrobials were discontinued in an additional 12 (17%) patients with negative BAL culture results who had alternative diagnoses. Of these 12 patients, nine were diagnosed with alveolar hemorrhage, one was diagnosed with mucous plugging, and one was diagnosed with laryngeal edema. These final two cases underwent additional therapeutic interventions: pulmonary hygiene (suctioning of secretions from the airways) and intubation within 48 hr of the procedure.

Table 2 Microbiology results from bronchoalveolar lavage
Full size table

The final patient was a case of refractory leukemia undergoing active chemotherapy with known invasive pulmonary fungal lesions. This patient experienced a tracheal perforation as a complication of the procedure, likely as a result of friable mucosa in the airways, and later underwent an elective tracheostomy and was discharged from the ICU five days later in stable condition.

Complications were rare and reported in only nine (13%) patients, with four major complications. These complications included the need for intubation within 48 hr post-procedure (one of 11 non-intubated patients), endotracheal tube (ETT) change (one of 60 intubated patients), post-bronchoscopy hypoxia with an oxygen saturation of < 92% (six patients, three of whom required adjustment of ventilator settings immediately following the procedure), and tracheal perforation (one of 71 patients) (Table 3).

Table 3 Bronchoscopy outcomes and complications
Full size table

The indications for intubation and ETT change were laryngeal edema secondary to an anaphylactic reaction to a penicillin antibiotic and diffuse pulmonary hemorrhage, respectively. Both patients survived to ICU discharge, but were dead at one-month post-ICU discharge from other causes. Age, sex, APACHE II score, and active chemotherapy were not significantly associated with 1) changes in management or 2) the development of complications post-bronchoscopy in multivariable analyses (Table 4).

Table 4 Predictors of change in management and complications post bronchoscopy
Full size table

Discussion

In patients with AL admitted to the ICU who underwent bronchoscopy with BAL, the results were associated with changes to management in 32 cases (45%). An infectious pathogen was identified in 19 of 71 cases (27%) and resulted in a change in antimicrobial therapy in 17 of the 19 cases (89%). In cases of negative BAL cultures, empiric antimicrobials were discontinued in all cases when an alternative diagnosis was made (12 of 71 cases, 17%) such as pulmonary hemorrhage, mucous plugging, tracheal perforation, or laryngeal edema. Complications were documented in nine patients (13%) and included intubation within 48 hr of the procedure, ETT change, post-procedural hypoxia, and tracheal perforation. There were no clinically significant changes in patient vital signs during or immediately following the procedure. These results suggest that among leukemia patients admitted to the ICU, flexible bronchoscopy with BAL was relatively safe, had a high diagnostic yield, and helped to guide medical management (Table 5).

Table 5 Vital signs before, during, and after bronchoscopy
Full size table

There have been few studies evaluating the clinical utility and safety of bronchoscopy with BAL in patients with hematologic malignancies23,24,25,26 and fewer looking specifically at AL patients.27,28 Only one of these studies looked specifically at patients admitted to the ICU.28 In general, they reported a low risk of complications ranging from 0-10%,23,24,25,26,27,28 and a diagnostic yield for infectious pathogens of between 15-66%.

Similar to these studies, the overall risk of post-procedural complications in our cohort was low at 13%. This is in contrast to previous studies of bronchoscopy with BAL in ICU patients29 and BMT recipients,17,18,19 which quote a much higher risk of complications of 49% and between 0-27%, respectively. The relative safety of the procedure in our patients may be related to the mechanical ventilator support in 85% of cases. The relatively low risk of post-procedural hypoxia (oxygen saturation < 92%) may also be due to a selection bias regarding which patients underwent the procedure. Pre-procedural hypoxia is a known risk factor for respiratory decompensation,30,31 and none of our patients were hypoxic prior to the procedure, although many were supported with moderate oxygen therapy. Therefore, the low incidence of hypoxia may be a reflection of the clinicians’ selection of patients expected to tolerate the procedure without complications. Additionally, the cases requiring intubation and ET tube change secondary to laryngeal edema and pulmonary hemorrhage may not have been direct complications of the bronchoscopy itself and may in fact overestimate the actual procedure-related complications. Both oxygen saturation and blood pressure remained stable throughout the procedure. Although HR was significantly higher during the procedure, the increase was small and unlikely to be of clinical significance.

Despite the low risk of complications, it should be noted that complications such as the need for intubation and tracheal perforation may have serious repercussions for the patient, including prolonged time spent on the ventilator, the need for a tracheostomy, and even an increased risk of death. Correspondingly, consideration must be given to the utility of the procedure, and patients and families should be made aware of these potential complications as part of the informed consent process.

The diagnostic yield in our cohort is similar to previous studies evaluating bronchoscopy with BAL in immunocompromised hosts.1,2,8,12,13,14,15 Our results suggest that the procedure is most useful for the diagnosis of infectious complications, specifically opportunistic pathogens such as aspergillus and herpes simplex virus. Identification of these opportunistic pathogens is important as they may portend poor clinical outcomes when left untreated; however, because treatment can have toxic side effects, it may only be initiated once these organisms are identified. Furthermore, identification of the causative organism can lead to a more timely initiation of targeted antimicrobial therapy or discontinuation of unnecessary broad-spectrum empiric antibiotics. Negative BAL cultures can also lead to the discontinuation of empiric antibiotics and seem to be particularly useful in cases where an alternative diagnosis (e.g., pulmonary hemorrhage) is made on bronchoscopy.

Age, sex, APACHE II score, and active chemotherapy were not significantly associated with changes in management or the development of complications post-bronchoscopy (Table 4). Age may not play a large role in predicting outcomes of bronchoscopy in this population given that the extremes of age (very old and very young) are largely absent (median age 55, IQR 46-63). Furthermore, while the APACHE II score reflects severity of illness and predicts patient mortality, it may not predict patient morbidity such as complications associated with a given procedure. Finally, chemotherapy status alone may not impact clinical outcomes, as leukemia patients are already severely immunocompromised as shown by the relative degree of neutropenia amongst our patients and the use of steroids alone or in addition to chemotherapy in 39%. In a larger population, other patient or procedural characteristics may better predict adverse outcomes.

The strengths of this study are the homogeneous population of AL patients and complete procedural, microbiologic, and outcome data. Limitations include the relatively small number of patients who underwent bronchoscopy during our study period (n = 71) and our inability to comment on the diagnostic yield and safety of the procedure in non-mechanically ventilated patients, given that 85% were intubated prior to the procedure. Furthermore, because of the observational study design, results may be influenced by residual or unmeasured confounders. Finally, the diagnostic yield and clinical utility of bronchoscopy in our study may be overestimated secondary to: 1) confounding by indication—i.e., the total number of AL patients admitted to the ICU during our study period, including those that did not undergo bronchoscopy is not known, and the procedure is more likely to be performed in cases of diagnostic ambiguity; 2) we cannot be sure whether the changes in management following the procedure were based on the results alone or took other factors such as the patients’ clinical status into account.

Conclusion

Our study suggests that while bronchoscopy with BAL is a safe procedure in the critically ill leukemia patient, serious complications including the need for intubation and tracheal perforation do occur. Furthermore, the results of the procedure can alter patient management in some cases, particularly when an infectious pathogen is identified or an alternative diagnosis to infection is made. More work is needed to better understand patient and procedural characteristics that will help predict the diagnostic utility and risk of complications related to flexible bronchoscopy.