Association between peripheral blood/bronchoalveolar lavage eosinophilia and significant oxygen requirements in patients with acute eosinophilic pneumonia

All consecutive patients diagnosed with AEP at the AFCH (tertiary referral military hospital with 874 beds in South Korea) between May 2012 and May 2017 were retrospectively reviewed.

The AFCH is the highest ranking Korean military hospital where all military personnel in need of bronchoscopy are transferred from smaller hospitals including The Armed Forces Busan Hospital, The Armed Forces Cheongpyeong Hospital, and 14 other military hospitals.

Medical records from the electronic medical records (EMR) system and picture archiving and communication system (PACS) were reviewed to confirm AEP cases. Diagnoses were made according to previously proposed criteria as follows: acute-onset febrile respiratory manifestations < 1 month in duration, bilateral diffuse infiltrates observed in chest radiographs, > 25% eosinophils in BAL fluid or eosinophilic pneumonia according to a lung biopsy, and absence of a known cause of pulmonary eosinophilia, including infections, toxins, and drugs [5]. We also included patients who did not undergo BAL or a lung biopsy, and diagnosed AEP by clinical impressions, radiological findings, and peripheral blood test results [10].

We retrospectively collected data from the EMR including general baseline characteristics, detailed smoking history including smoking duration, hospital days, intensive care unit (ICU) admission days, oxygen requirements, initial treatments, and laboratory data including white blood cell (WBC) counts, eosinophil differentials (both blood and BAL fluid), C-reactive protein (CRP) levels, and PaO₂/FiO₂ ratio. We also reviewed the PACS to identify chest radiographic findings, including pulmonary infiltrates and pleural effusion.

The primary endpoint of our study was total days of oxygen supply during admission, and we further determined which baseline parameters were associated with significant oxygen requirements (total days of oxygen supply ≥2 days after admission) and also evaluated if blood eosinophil (%) and peripheral eosinophil count show significant correlation with total days of oxygen supply.

Considering the unique background of patients included in our cohort, we only counted recent smoking experience. All enrolled patients were military personnel who must undergo basic military training upon entrance to the military. Smoking is strictly prohibited during basic military training of at least 6 weeks. We only evaluated smoking experience after basic military training regardless of past smoking experience.

Oxygen was prescribed to achieve a target saturation of 94–98% [14]. If the patients’ oxygen saturation was not maintained at the optimal level (94–100%) despite oxygen supply by a nasal cannula or mask, oxygen was supplied by a high-flow nasal cannula (HFNC) [15].

Due to a risk of hypoxemia while obtaining BAL fluid, supplemental oxygen was given to the patients during bronchoscopy [16]. To exclude the possibility of including procedural- related oxygen supplementation, which we did not consider significant if the total duration of oxygen supply was only 1 day, and we defined “significant oxygen requirement” as total days of oxygen supply ≥2 days after admission.

Using receiver operating characteristic (ROC) curve analyses, BEP was categorised using the calculated cut-off after plotting whether a patient required an additional oxygen supply. Patients were stratified into three groups by the initial PEC and BEP. Patients who underwent BAL analyses were categorised into high and low BEP groups. Patients were categorised by PEC using a cut-off of 500 × 10⁹ cells/L according to a previous study [8]. Patients with both high BEP and high PEC were given a score of 2. Patients with either a high BEP or PEC were given a score of 1. Patients with no increase in either BEP or PEC were given a score of 0.

Statistical analyses were performed using SAS 9.3 software (SAS Institute, Cary, NC, USA). Quantitative variables are presented as means ± standard deviations, and categorical variables are shown as numbers and percentages. Categorical variables, including sex, presence of a smoking history, ICU admission history, mechanical ventilation, HFNC, bilateral lung infiltrates, pleural effusion, and use of corticosteroids were compared between the groups using a chi-square test. Continuous variables, including age, number of cigarette smoking days, WBCs, PEC/BEP, CRP, hospital days, ICU admission days, oxygen supply days, and corticosteroid treatment duration were assessed using Student’s t-test. We performed correlation analyses to demonstrate the correlations between BEP and the PaO₂/FiO₂ ratio, oxygen supply days, and PEC. We performed multiple logistic regression on significant oxygen requirement with two different models using age, C-reactive protein (CRP), smoking duration, and BBE score (model 1) and age, CRP, BEP, and PEC (model 2).

A total of 338 patients with AEP were evaluated. Of all, 253 (74.9%) patients underwent BAL and 221 (65.4%) patients had eosinophil percentage more than 25%. Beside these patients, 117 (34.6%) patients were diagnosed as AEP on basis of clinical and radiological context. As there was no patient who underwent lung biopsy, no patient was diagnosed by pathologic confirmation. Table 1 shows the general clinical characteristics of the patients. All but one patient was male (99.7%), and their mean age was 20.4 years. Only two patients were never smokers. Mean total number of cigarette smoking days was 26.2. Mean WBC level was 14,359. Mean PEC, BEP, and CRP levels were 4.1, 46.7%, and 9.4, respectively. About 37% of patients were initially admitted to the ICU, and mean hospital days, ICU admission days, and oxygen supply days were 10.5, 2.0, and 2.8, respectively. Three patients underwent mechanical ventilation and nine patients were oxygenated with HFNC during ICU care. The initial chest radiograph findings revealed 94.1% bilateral infiltrates and 51.2% bilateral effusion. As an initial treatment, 76.9% of patients underwent intravenous corticosteroid therapy, 18.9% underwent oral corticosteroid therapy, and 4.1% did not receive steroids. The mean duration of steroid treatment was 14.2 days.

The correlation analyses showed significant but weak correlations between the variables. BEP had a linear increasing relationship with the PaO₂/FiO₂ ratio (Pearson’s correlation coefficient r = 0.207, P = 0.008) (Fig. 1a). In addition, the number of oxygen supply days had a linearly decreasing relationship with BEP (Pearson’s correlation coefficient r = − 0.202, P = 0.005) (Fig. 1b). PEC had a linearly increasing relationship with BEP (Pearson’s correlation coefficient r = 0.276, P < 0.001) (Fig. 1c) and a linearly decreasing relationship with the number of oxygen supply days (Pearson’s correlation coefficient r = − 0.157, P = 0.014) (Fig. 1d).

A total of 244 patients had both an initial complete blood count and BAL results for evaluation. The optimal cut-off of BEP for prediction of significant oxygen supply requirement was calculated using ROC curve analysis. (Fig. 2). The area under the ROC curve was 0.577 with a P-value of 0.04. The calculated cut-off was 41.5% for BEP.

The baseline clinical characteristics of the 244 patients are shown in Table 2. Mean WBC, PEC, BEP, and CRP were 14,231, 3.8, 46.8%, and 9.5, respectively. The mean number of hospital days was 11.0 and ICU admission days was 2.0. The mean PaO₂/FiO₂ ratio at admission was 261.9. After stratification by BBE score, 78 patients were in group with score 0, 108 patients were in group with score 1, and 58 were in group with score 2. Male proportion and mean age were not different between the groups. The proportion of ever smokers and the mean number of smoking days were not significantly different.

WBCs were significantly different between the groups. The mean WBC decreased with an increase in the BBE score. The values were 16,248, 14,489, and 11,037 for groups with score 0, 1, and 2, respectively (P < 0.001). Both mean PEC and BEP increased significantly as the score increased from 0 to 2 (P < 0.001 and P < 0.001, respectively). CRP was inversely related with the score (P < 0.001). The mean number of hospital days decreased significantly as the score increased from 0 to 2 (P = 0.009) (Fig. 3a). The proportion of patients who were admitted to the ICU was significantly lower in group with score 2 than in group with score 0 (P < 0.001). The mean duration of ICU stay was not significantly different between the groups (P = 0.731).

A higher proportion of patients was given additional oxygen as the score decreased from 2 to 0 (P = 0.031). Two patients in group with score 0 and 1 patient in group with score 1 were mechanically ventilated, but no one in group with score 2 underwent mechanical ventilation. The mean total number of days of oxygen supply was 3.4, 2.8, and 2.3 in groups with score 0, 1, and 2, respectively (P = 0.002) (Fig. 3b), while the mean number of steroid treatment days was 15.5, 14.0, and 12.6, respectively (P = 0.006). The initial PaO₂/FiO₂ ratios were 249.7, 253.8, and 300.9, respectively (P = 0.003) (Fig. 3c).

Age, CRP, smoking duration before admission, BEP, PEC, and BBE score were entered into logistic regression analyses to determine the association with significant oxygen requirements (Table 3). CRP, BEP (categorical), PEC, and BBE scoring were significant in univariate analyses. Two models were applied for multivariate analyses. In model 1, BBE scoring was entered, while BEP and PEC were excluded. In model 2, BEP and PEC were included, while BBE score was excluded. In model 1, BBE score was the only significant factor associated with significant oxygen requirements (P = 0.018). Compared to group with score 2, the odds ratio (OR) for significant oxygen requirements was 2.25 (95% confidence interval [CI]. 1.112–4.574, P = 0.024) for group with score 1, while the OR was 3.014 (95% CI 1.378–6.593, P = 0.006) for group with score 2. In model 2, BEP was a significant factor (OR 1.939, 95% CI 1.072–3.505, P = 0.028).