Gur

Patients with hematological malignancies are at increased risk of opportunistic pulmonary infections.[1] Flexible bronchoscopy-facilitated bronchoalveolar lavage (BAL) is important in this population's diagnosis and therapy guidance.[2] Coincidentally, thrombocytopenia is a common occurrence in this population. Thrombocytopenia, combined with other comorbidities and hemostatic dysfunction, is a common concern when considering BAL's safety and cost-benefit ratios in hematological patients.[3] This is particularly true in patients with platelet counts below 50x10³/μL, prompting most providers to consider prophylactic platelet transfusions (particularly in very low platelet counts below 10x10³/μL) in an attempt to mitigate the risk of periprocedural complications.[4]
Nonetheless, there are no universally accepted platelet count thresholds for BAL or the decision to transfuse periprocedurally. Previous observational studies have described vanishingly low complication rates above 20x10³/μL[5] or even 10x10³/μL.[6,7] Current accepted guidelines, based on low-level observational data, agree on the general safety of BAL when the platelet count is above 20x10³/μL.[8,9].
In this study, we aimed to describe the incidence of various periprocedural complications of BAL in hospitalized hematological patients with significant thrombocytopenia, assessing potential predictors of increased risk, in an attempt to assess the safety of this invasive and yet essential procedure in such a frail population.

This retrospective cohort study was conducted in Rambam Health Care Campus (RMC), a tertiary 1000-bed medical center, the largest medical center in northern Israel. The Electronic Health Registry (EHR) files of all patients undergoing BAL between January 1, 2012, and December 31, 2021, were reviewed.
The study included all adult patients (18 years or older) undergoing BAL while hospitalized in our hematology ward with a platelet count below 50x10³/μL 24 hours before the procedure. Exclusion criteria were: 1) Active hemoptysis, epistaxis, or known upper gastrointestinal bleeding in the 24 hours prior to the bronchoalveolar lavage and 2) BAL performed while the patient is mechanically ventilated.
Physician's notes, admission and discharge reports, imaging interpretations, and background diagnoses were manually and individually reviewed for each patient in this study. Additional demographic, clinical, and laboratory data, including date of birth, vital signs, and laboratory results upon presentation, were mined using the MD-Clone® interface (version 4.25 or older). Machine-mined data was assessed for accuracy and relevance by the investigator reviewing the EHR.
The primary outcome was any major complication in the 24 hours after the initiation of BAL, including any of the following: 1) clinically significant major bleeding, either resulting in the premature termination of the BAL and / or necessitating packed red blood cells transfusion peri procedurally and / or deemed by any of the treating physicians as potentially life-threatening; 2) The need for ventilatory support other than low flow conventional supplementary oxygen therapy (COT), including high flow (>10 L/min) oxygen therapy or any positive pressure ventilation (both invasive and noninvasive) or 3) death from any cause. The secondary outcomes were defined as any non-major bleeding within the first 24 hours post-procedurally, including self-limiting bleeding visualized during bronchoscopy and any epistaxis, whether said bleeding resolved spontaneously, or hemostatic measures (such as nasal packing, intraluminal epinephrine, tranexamic acid or cold saline injections) were required. Additional secondary outcomes included a decrease of ≥5 mmHg in mean arterial pressure (MAP) in the lowest measurement 24 hours post-BAL (compared with MAP measured immediately prior to BAL); an increase of ≥10 beats per minute (bpm) in heart rate (similarly defined as the highest resting heart rate documented in the 24 hours after bronchoscopy minus the heart rate immediately before bronchoscopy), a decrease of ≥1 mg/dL in hemoglobin (similarly defined) and a decrease of ≥5% capillary hemoglobin saturation (SpO2) as measured by pulse oximetry within 24 hours (similarly defined). In addition, we recorded the diagnostic yield of the BAL, i.e., whether any pathogens were recovered and the type of infectious syndrome (e.g., invasive pulmonary aspergillosis, bacterial pneumonia, or atypical infection such as legionellosis) supported by the BAL results.
Per institutional protocols, BAL was performed transnasally unless technically infeasible, in which case the transoral approach was implemented. This study was reviewed and approved by the RMC institutional ethics committee (RMB-22-0017).
Statistical Analysis. Standard descriptive statistics were used to summarize population characteristics. We used a chi-square test for categorical variables, a Mann-Witney U test for nonparametric variables, and a student's unpaired t-test for normally distributed continuous variables. Tukey's correction was applied when applicable to adjust for multiple comparisons. Categorical variables were described using proportions and percentages, nonparametric variables with median and interquartile range (IQR), and normally distributed continuous variables as mean with standard deviation (S.D.).
Multivariate logistic regression modeling was performed using Pearl and Reed's method. We used the Pearson correlation coefficient to determine possible correlations between independent variables; only variables that were not co-related (Pv>0.1 on univariate analysis) were included in the model. A 2-sided Pv<0.05 was considered statistically significant for all tests. All calculations were performed using PASW software version 29.0 (IBM, Chicago, IL).

Of 720 adult (age 18 or older) hematological patients undergoing BAL while hospitalized during the study period, only 553 (76.8) had a platelet count below 50x10³/μL. Nine (1.6%) patients were excluded due to active hemoptysis, epistaxis, or upper gastrointestinal bleeding in the 24 hours prior to bronchoscopy, and 37 (6.7%) patients were mechanically ventilated while undergoing BAL. A Consolidated Standards of Reporting Trials (CONSORT) diagram summarizing the data mining and filtering process is presented in Figure 1.

Figure 1. Study Design. The study phases are presented in accordance with the CONSORT guidelines. Abbreviations: EHR - electronic healthcare registry, BAL - Bronchoalveolar Lavage.

Of the 507 patients included in the final analysis, 281 (55.4%) had a platelet count below 20x10³/μL, and 210 (41.4%) were females. The mean age was 55.3 years (median 59, Sd 15.1), and the mean platelet count was 20.4x10³/μL (median 18, Sd 12). Regarding the background hematological diagnosis, the most common diagnosis was acute myeloid leukemia, affecting 264 (52.1%) of patients. Overall 324 (63.9%) had acute leukemia, 6 (1.2%) chronic leukemia, 11 (2.1%) indolent lymphomas, 66 (13%) other lymphomas, 46 (9.1%) multiple myeloma, 40 (7.9%) myelodysplastic syndrome, 10 (1.9%) myeloproliferative disorders, and 7 (1.4%) aplastic anemia. Only the rates of myelodysplastic syndrome were significantly (p=0.017) associated with higher platelet counts.
When examining the bloodwork and vital signs obtained on the morning of the procedure, mean hemoglobin concentration was 0.3mg/dL lower (p=0.003), and mean prothrombin time 0.7 seconds longer in patients with platelet counts below 20x10³/μL. The mean activated partial thromboplastin time was 29.9 seconds (Sd 5.5), heart rate 89.4 bpm (18.3), SpO2 95.8% (Sd 3.4) or mean arterial pressure 82.4 mmHg (Sd 10.4) with no significant association with platelet count. There was no significant difference between patients with platelet counts below and above 20x10³/μL in any of the other baseline characteristics examined, with the notable exception of the rates of periprocedural platelet transfusion occurring in four times as many patients with platelet counts below 20x10³/μL (p<0.001). These data are summarized in Table 1.

Table 1. Study patients characteristics.

Clinically significant bleeding (2 patients, 0.9%) and need for ventilatory support with noninvasive bilevel positive pressure (one patient, 0.4%) occurred exclusively in patients with platelet counts above 20x10³/μL, and the only primary outcome to have been recorded in the <20x10³/μL group was one account of death due to cardiovascular collapse in a patient with severe septic shock and multiorgan failure (manifested prior to bronchoscopy) 21 hours after BAL (p=0.219 for the composite of primary outcomes).
The incidence of non-major bleeding was higher (13 (4.6%) vs 4(1.8%)) in patients with platelets below 20x10³/μL (RR 2.614, 95%CI[0.864,7.906]), a trend that approached statistical significance (p=0.076). Further analysis demonstrated spontaneously resolved minor bleeds to be responsible for this trend (RR 3.217, 95%CI[0.919,11.262]), with a virtual similar rate of one patient (0.4%) requiring nasal packing with topical tranexamic acid in both groups. A decrease in mean arterial pressure (169 patients, 33%), SpO2 (40 patients, 7.9%), hemoglobin (61 (12%)) or an increase in heart rate (94 (18%)) as defined in the secondary outcomes section above was similar between groups, as presented in Table 2.

Table 2. Primary and secondary outcomes .

When examining the microbiological diagnostic significance of these BAL samples, overall diagnostic yield and the types and counts of major pathogens recovered did not differ significantly between groups. These findings are presented in Figure 2.

Figure 2. Diagnostic Yield of Bronchoalveolar Lavage. Positive results of BAL microbiology, either by means or direct culture or nucleic acid amplification. Abbreviations: CMV - cytomegalovirus, GAS - group A streptococci, HHV6 - human herpesvirus 6, HMPV - human metapneumovirus, HSV - herpes simplex virus, NTB - nontuberculous mycobacteria, PCP - pneumocystis Jirovecii pneumonia, RSV - respiratory syncytial virus, S.A. - Staphylococcus aureus, T.B. - Mycobacterium tuberculosis, VZV - varicella-zoster virus.

To further illuminate the potential association between the degree of pre-procedural thrombocytopenia and potential patient-specific confounders, three multivariate logistic regression models were constructed, predicting major complications (as defined by the primary outcome), any bleeding or a composite of all recorded complications (both primary and secondary outcomes). A trend towards increased risk of any bleeding was observed with preprocedural platelet transfusion (aRR 3.55, 95%CI[0.83,15.18]), although this finding did not reach the accepted level of statistical significance (p=0.087). Age, activated partial thromboplastin time, prothrombin time, preprocedural hemoglobin concentration or platelet count, gender, the presence of acute leukemia, or history of hematopoietic stem-cell transplant did not significantly predict the development of major complications, any bleeding, or any of the recorded complications, as detailed in Table 3.

Table 3. Multivariate Logistic Regression. Multivariate logistic regression models predict the primary outcome (major bleeding, need for ventilatory support or death), any bleeding, or any outcome (a composite of all primary and secondary outcomes).

This study suggests a very low incidence (0.8%) of serious periprocedural complications when performing BAL in hematological patients, even in the presence of VST. Both cases of major bleeding (British Thoracic Society grade 3[10] for the twain) were recorded in patients with higher platelet counts. While the reported rates of bronchoscopy complications range between 1 percent to 12%, various retrospective studies have demonstrated the incidence of serious complications (i.e., major bleeding halting procedure, need for ventilatory support or death) to be consistently in the range of 0.7-0.9%.[4,6,7,11-13] Interestingly, the rates of both serious complications in general and major bleeding, in particular, are similar when examining large cohorts of BAL in the general population and in patients with various degrees of thrombocytopenia, which could serve as a weak, albeit ecological, corroboration to the lack of a direct link between platelet counts and major complications in bronchoscopy.[6,11,14]
The rates of non-major bleeding were higher in severely thrombocytopenic patients, approaching significance only due to very minor, spontaneously resolving bleeding. Echoed in some previous reports[4,7,12] and considerably lower than others,[11] this observation is consistent with our understanding of platelet function in primary hemostasis. Namely, superficial bleeding (e.g., petechiae and purpura) is the major clinical manifestation of isolated thrombocytopenia.[15]
The trend towards increased risk of any bleeding with platelet transfusion represents correlation rather than causation. When adjusted for other observed confounders, such as hemoglobin concentration and hematological history, our data showed no clear signal toward the direct benefit of platelet transfusion in preventing BAL complications. Previous attempts to establish a safety threshold of 30x10³/μL or even 10x10³/μL were hampered by their retrospective nature and high prevalence of platelet transfusions in the control group.[6,7] Our study demonstrates the very strong predilection of clinicians to administer platelet transfusion below the currently set threshold of 20x10³/μL, which is based on a very low level of evidence.[8]
More interestingly, there was no significant association between platelet count or coagulation studies and any of the BAL complications examined. Coherent with the limited role of the coagulation cascade in minor superficial injuries, this observation is also consistent with the mounting evidence of these limited coagulation studies' very low predictive ability in predicting bleeding risk.[16]
To our knowledge, this is the only study focused exclusively on BAL by primary design. We attribute the relatively low incidence of bleeding of all varieties observed in the study to mild tissue derangement when bronchoscopy is confined to visualization and BAL.[17] The shorter duration, low energy, smaller bore diameters, and minimal tissue disruption could all explain the general safety of this procedure, even in the sickest of patients.[18] This is particularly important given the recent departure from tissue sampling to diagnose infectious diseases. Particularly, most recent guidelines[8,19] have accepted BAL-based testing, such as galactomannan, as a preferable alternative to diagnosing invasive pulmonary aspergillosis - by far the most prevalent diagnosis resulting from bronchoscopy in our sample.[20]
Of note, examining the BAL aspirate most often yielded a diagnosis of Invasive Pulmonary Aspergillosis, followed by gram-negative Enterobacteriaceae, Herpes Simplex Virus, Cytomegalovirus, and Pneumocystis jirovecii. These rates and relative incidence are similar to previously reported cross-sectional studies of diagnostic BAL in hematological patients.[21] While the degree of thrombocytopenia seems to correlate with bone marrow dysfunction (as evidenced by the high correlation with white and red blood cell counts), the microbiological yield is similar when compared to patients without severe thrombocytopenia (and by conjecture, probably a lesser degree of myelosuppression). Actually, the underlying hematological disease, rather than peripheral counts themselves, can explain the majority of immune dysfunction and resultant, rather opportunistic infectious profile.[1] Alternatively, patient selection (i.e., those who undergo BAL are those manifesting clinical characteristics of opportunistic infections) and characteristics (prolonged hospitalization, exposure to multiple antimicrobial agents, etc.), rather than the degree of pancytopenia, may be the main mechanism explaining BAL microbiology.

Safety of Bronchoalveolar Lavage in Hematological Patients with Thrombocytopenia. A Retrospective Cohort Study