Incremental contribution of bronchoalveolar lavage fluid data to multidisciplinary discussion of interstitial lung disease

Abstract

Background

Multidisciplinary discussion (MDD) integrating clinical, radiological, and laboratory information is considered the reference standard for the diagnosis of interstitial lung disease (ILD). Bronchoalveolar lavage fluid (BALF) analysis is recommended in several diagnostic algorithms, but its incremental contribution during multidisciplinary evaluation remains incompletely characterized. In this study, we aimed to assess the impact of BALF findings on diagnostic consensus and confidence during MDD for suspected ILD.

Methods

In this cross-sectional study, patients with suspected ILD at a tertiary referral center who remained without a definitive diagnosis after clinical evaluation and high-resolution computed tomography (HRCT) underwent bronchoscopy with BALF collection (n = 29). Blinded pulmonologists and radiologists evaluated each case using a sequential four-step framework: (1) radiological assessment; (2) radiological assessment with clinical data; (3) multidisciplinary discussion without BALF information; and (4) multidisciplinary discussion including BALF results. Interobserver agreement was assessed using Fleiss’ kappa, and diagnostic confidence was recorded at each stage.

Results

Diagnostic agreement increased across the sequential evaluation steps (Fleiss’ κ: 0.056, 0.565, 0.905, and 0.965, respectively; p < 0.001). Consensus diagnoses were achieved in 24 cases after MDD without BALF and in 27 cases following incorporation of BALF findings. The proportion of evaluations with high diagnostic confidence (≥ 90%) increased from 3.4% before BALF availability to 20.7% after its inclusion.

Conclusion

Within a multidisciplinary diagnostic framework for ILD, BALF information were associated with improved diagnostic consensus and increased clinician confidence. BALF may therefore serve as a complementary diagnostic tool in the multidisciplinary assessment of ILD, particularly in cases that remained uncertain after clinical and radiological evaluation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-026-04412-x.

Background

Interstitial lung disease (ILD) comprises a heterogeneous group of more than 150 diseases characterized by lung inflammation and/or fibrosis, ultimately impairing gas exchange and reducing survival [1]. Accurate diagnosis is critical for guiding treatment decisions and prognostic assessment [2]. However, the diagnosis of ILD is often challenging and requires a multidisciplinary discussion (MDD), which is regarded as the gold standard [2]. MDD integrates clinical data, imaging, and laboratory testing with collaborative input from pulmonologists, radiologists, and other specialists.

When noninvasive data are inconclusive, additional procedures such as surgical lung biopsy, transbronchial cryobiopsy, or bronchoalveolar lavage fluid (BALF) analysis may be required [2]. The contribution of lung biopsies to MDD has been well studied using structured, stepwise models such as the Flaherty approach [3]. In contrast, the role of BALF analysis in MDD remains less well defined. Current guidelines recommend BALF analysis primarily for: (a) suspected hypersensitivity pneumonitis (HP) or sarcoidosis [4, 5]; (b) cases where clinical, radiologic, and serologic findings are inconclusive; and [3] patients with indeterminate radiologic features for usual interstitial pneumonia, when cellular profiles, particularly eosinophils and neutrophils, may support a diagnosis of idiopathic pulmonary fibrosis [6, 7].

Despite these recommendations, the extent to which BALF results influence diagnostic agreement, confidence levels, or final consensus in MDD has not been systematically evaluated. Addressing this knowledge gap is clinically relevant because BALF is a widely available, minimally invasive tool that may enhance diagnostic precision, particularly in settings where biopsy is not feasible. Therefore, we aimed to investigate whether BALF analysis: (a) improves diagnostic concordance among MDD participants, (b) increases confidence levels in ILD diagnosis, and (c) influences the final diagnostic consensus.

Methods

Study design and ethics

We conducted a cross-sectional study, with prospective allocation, at a tertiary ILD center in Rio de Janeiro, Brazil, between May 2023 and December 2024. The study was approved by the Research Ethics Committee of Clementino Fraga Filho University Hospital at Federal University of Rio de Janeiro (CAAE: 66958223.9.0000.5257), and all participants provided written informed consent before enrollment.

Participants

Patients with suspected ILD referred to the center underwent detailed clinical evaluation, including history, physical examination, HRCT, laboratory tests (autoimmune panel and inflammatory markers), pulmonary function testing, and echocardiography. Patients who did not receive a definitive diagnosis after this initial evaluation were considered eligible for bronchoscopy with BALF collection, in accordance with current guidelines [4, 5, 7]. Exclusion criteria included contraindications to bronchoscopy, established idiopathic pulmonary fibrosis or connective tissue disease-associated ILD, pregnancy, refusal to consent, insufficient BALF cellularity (< 400 cells/cytospin and < 5% epithelial cells), or prior disclosure of BALF results to the MDD team (Fig. 1).

Fig. 1.

Flowchart illustrating patient recruitment, exclusions, and inclusion in the multidisciplinary discussion MDD analysis. MDD = Multidisciplinary discussion

Bronchoscopy and BALF collection

All examinations were performed by a single experienced bronchoscopist (B.P.) under intravenous sedation with midazolam and fentanyl. BALF was obtained preferentially from the most affected lobe on the HRCT image, using aliquots of 20 mL of saline up to a maximum of 120 mL. The first aliquot was reserved for microbiological analysis; remaining aliquots were used for cytologic and immunophenotypic analysis. BALF was processed within 1 h as follows: centrifuged at 1200 × g at 4 °C, total cell count determined via hemocytometer, differential counts performed on Panotic-stained cytospin preparations, and lymphocyte immunophenotyping conducted by flow cytometry (CD3⁺, CD4⁺, CD8⁺, CD19⁺) (Supplemental Fig. 1).

Data collection

Clinical data included age, sex, symptom duration, smoking status, environmental exposures, medications, family history, and autoimmune features. Complementary examinations included HRCT, serologic analysis, spirometry, diffusion capacity, plethysmography, and echocardiography to assess pulmonary hypertension.

Multidisciplinary discussion

The MDD team included at least 2 expert thoracic radiologists and 2 ILD pulmonologists, all blinded to patient information. The discussion followed a four-step Flaherty-based approach [3, 7]: (1) independent computed tomography review; (2) computed tomography plus clinical data review; (3) group discussion without BALF information; (4) group discussion including BALF results. The final diagnoses recorded included idiopathic pulmonary fibrosis (IPF), HP, sarcoidosis, unclassifiable ILD, nonspecific interstitial pneumoniae (NSIP), and others. Diagnostic confidence level was graded as definite (> 90%), provisional high (70%–89%), provisional low (50%–69%), or unclassifiable (< 50%), following the ontological framework for standardizing the diagnostic classification, proposed by Ryerson et al. [8]. Consensus required agreement of all participants; otherwise, the case was recorded as “no consensus”.

Statistical analysis

Descriptive statistics are presented as counts (%) for categorical variables and medians (interquartile range [IQR]) for continuous variables. Interobserver agreement was assessed using Fleiss’ kappa: 0.00–0.20 (weak), 0.21–0.40 (intermediate), 0.41–0.60 (moderate), 0.61–0.80 (strong), and 0.81–1.00 (almost perfect) [9]. Receiver operating characteristic curve analysis was performed to evaluate whether the incorporation of BALF results into the MDD process increased the likelihood of reaching a consensus diagnosis during MDD. Analyses were conducted using R 4.1.0 (R Core Team, 2021). A convenience sample included all patients referred during the recruitment period.

Results

Baseline characteristics

Between May 2023 and December 2024, 44 patients were assessed for eligibility at a Brazilian ILD reference center. Three patients declined participation, and 5 had formal contraindications to bronchoscopy. BALF was collected from 36 patients; 2 were excluded due to inadequate samples (> 5% epithelial cells), and 5 because their final diagnosis was established before MDD. Thus, the final study cohort comprised 29 patients (Fig. 1). Baseline characteristics of the patients are presented in Table 1. Most participants were women (69%), and 58.6% reported environmental exposures. The median interval between symptom onset and first evaluation at the ILD center was 34 months. Current or former smokers accounted for 44% of patients. Median pulmonary artery systolic pressure was 24 mmHg (IQR, 20–33 mmHg), forced vital capacity was 74% of the predicted value (IQR, 65%–87.5%), and median diffusing capacity for carbon monoxide was 48.5% of the predicted value (IQR, 41%–87.5%).

Table 1.

Baseline characteristics of patients undergoing bronchoalveolar lavage fluid analysis

	All Patients (N = 29)
Age (years), median (IQR 25%–75%)	65.0 (55.5–70.0)
Sex, n (%)
Female	20 (69)
Male	9 (31)
Smoking history, n (%)	13 (44)
Symptom duration (months), median (IQR)	34.0 (16.0–57.2)
Environmental exposures, n (%)	17 (58.6)
Avian antigens	10 (34.4)
Mold	11 (37.9)
Fibrosis family history, n (%)	3 (10.3)
Gastroesophageal reflux disease, n (%)	18 (62.1)
SPAP (mmHg), median (IQR)	24 (20–33)
FVC (% predicted), median (IQR)	74 (65.0–87.5)
DLCO (% predicted), median (IQR)	48.5 (41.0–87.5)
Pulmonary lobes washed, n (%)
Right superior lobe	5 (17.2)
Left superior lobe	2 (6.9)
Lingula	6 (20.7)
Middle lobe	2 (6.9)
Right lower lobe	8 (27.6)
Left lower lobe	6 (20.7)
Instilled volume (mL), median (IQR)	80 (80–80)
Returned volume, mL, median (IQR)	30 (30–36)
Cells (cells/mL), median (IQR)	1,780,000 (714,000–2,590,000)
Epithelial (%), median (IQR)	0.25 (0–1.25)
Macrophages (%), median (IQR)	78.3 (64.5–87.7)
Lymphocytes (%), median (IQR)	5.7 (1.3–12.5)
Neutrophils (%), median (IQR)	10.0 (2.3–22.0)

DLCO  Diffuse lung carbon monoxide capacity measurement, FVC  Forced vital capacity, IQR  Interquartile range, SPAP  Systolic pulmonary artery pressure

BALF was most frequently collected from the right lower lobe (27.6%), followed by the lingula (20.7%) and left lower lobe (20.7%). The median BALF cellular profile was 78.3% macrophages, 10% neutrophils, 5.7% lymphocytes, and 1.25% eosinophils.

Diagnoses across the MDD steps

At step 3 (group discussion without BALF information), the most frequent diagnoses were HP (n = 8, 27.6%), unclassifiable ILD (n = 6, 20.7%), and sarcoidosis (n = 4, 13.8%); consensus was not reached in 5 cases. After BALF results were incorporated (step 4), the leading diagnoses were HP (n = 6, 20.7%), unclassifiable ILD (n = 7, 24.1%), and sarcoidosis (n = 5, 17.2%). Two cases remained unresolved (unclassifiable ILD versus NSIP; unclassifiable ILD versus HP) (Table 2, Fig. 2).

Table 2.

Diagnoses assigned during multidisciplinary discussion without (step 3) and with (step 4) BALF results

Diagnostic	Step 3		Step 4
	Patients (N = 29)	%	Patients (N = 29)	%
Idiopathic pulmonary fibrosis	2	6.3	3	10.3
Hypersensitivity pneumonitis	8	27.6	6	20.7
Sarcoidosis	4	13.8	5	17.2
Idiopathic NSIP	1	3.5	1	3.5
Unclassifiable ILD	6	20.7	7	24.1
Others	3	10.3	5	17.2
No consensus	5	17.2	2	6.3

ILD  Interstitial lung disease, NSIP  Nonspecific interstitial pneumoniae

Fig. 2.

Sankey diagram showing changes and discordance in diagnoses between step 3 (MDD without BALF data) and step 4 (MDD with BALF data). The width of the flow represents the number of cases transitioning between diagnostic categories. BALF = Bronchoalveolar Lavage Fluid; MDD = Multidisciplinary discussion

Diagnostic concordance and agreement

Step 2 (computed tomography plus clinical data) showed 45.5% agreement (κ = 0.565; 95% confidence interval, 0.425–0.705). Step 3 increased agreement to 92.4% (κ = 0.905; 95% confidence interval, 0.825–0.985; p < 0.001), and step 4 further improved agreement to 97.2% (κ = 0.965; 95% confidence interval, 0.915–1.000; p < 0.001). Steps 3 and 4 demonstrated almost perfect interobserver agreement (Table 3).

Table 3.

Interobserver agreement and concordance across multidisciplinary discussion steps

Step	Kappa	p Value	Agreement Strength	Concordance (%)
1	0.056	< 0.001	Weak	45.5
2	0.565	< 0.001	Moderate	64.2
3	0.905	< 0.001	Almost perfect	92.4
4	0.965	< 0.001	Almost perfect	97.2

Kappa values and corresponding p values are shown for each step of the multidisciplinary discussion. Agreement strength is classified according to Landis and Koch criteria [13]. Concordance (%) represents the proportion of cases with unanimous or majority consensus among evaluators. Step 1, computed tomography review alone; Step 2, computed tomography plus clinical data; Step 3, group discussion without bronchoalveolar lavage fluid data; Step 4, group discussion including bronchoalveolar lavage fluid results

Diagnostic confidence level

At step 3, a diagnostic confidence level > 90% was achieved in 1 case (3.4%), a level of 70%–89% in 10 cases (34.5%), 50%–69% in 7 cases (24.1%), and < 50% in 6 cases (20.7%). At step 4, a confidence level > 90% was reached in 6 cases (20.7%); confidence levels of 70%–89%, 50%–69%, and < 50% were each observed in 7 cases (24.1%). Sarcoidosis was the most frequent high-confidence level diagnosis (Table 4, Fig. 3).

Table 4.

Diagnostic confidence levels in multidisciplinary discussion steps 3 and 4

Confidence Level	Step 3		Step 4
	Patients (N = 29)	%	Patients (N = 29)	%
> 90%	1	3.5	6	20.7
70%–89%	10	34.5	7	24.1
50%–69%	7	24.1	7	24.1
< 50%	6	20.7	7	24.1
No consensus	5	17.2	2	6.9

Number and percentage of patients achieving each confidence level during group discussion without bronchoalveolar lavage fluid data (step 3) and with bronchoalveolar lavage fluid results included (step 4). “No consensus” indicates cases where the multidisciplinary discussion did not reach agreement

Fig. 3.

Sankey diagram showing changes in diagnosis confidence level between step 3 (MDD without BALF data) and step 4 (MDD with BALF data). The width of the flow represents the number of cases transitioning between confidence level categories. BALF = Bronchoalveolar Lavage Fluid; MDD = Multidisciplinary Discussion

Impact of BALF on MDD

Consensus was reached in 24 of 29 cases without BALF data (step 3). BALF findings contributed to diagnostic consensus in 3 additional cases (sarcoidosis, organizing pneumonia, and idiopathic pulmonary fibrosis). Two cases remained unresolved despite BALF information. BALF analysis also influenced cases that had previously reached a diagnostic consensus, leading to changes in the final diagnosis at step 4. BALF differential cellularity is presented in Table 5 for cases in which it had a positive impact. Receiver operating characteristic analysis demonstrated an area under the curve of 0.70 (95% confidence interval, 0.46–0.94), supporting a moderate discriminative role of BALF in achieving MDD consensus (Fig. 4).

Table 5.

BALF cellularity in cases that it had positive impact in the diagnosis

Patient	Final diagnosis (Step 3)	Final diagnosis (Step 4)	BALF differential cellularity
9	No consensus	Fibrotic organizing pneumonia	Neut 31%, Lymph 27%, Mac 32%, Eos 7.4%
11	HP	Unclassifiable ILD	Neut 15%, Lymph 1.9%, Mac 78%, Eos 1.2%
13	No consensus	IPF	Neut 20.2%, Lymph 3.5%, Mac 69.2%, Eos 5.75%
18	No consensus	Sarcoidosis	Neut 0.75%, Lymph 31.2%, Mac 67.5%, Eos 0.5%
27	Unclassifiable ILD	Eosinophilic pneumonia	Neut 1.8%, Lymph 0%, Mac 80.6%, Eos 18%
29	HP	Unclassifiable ILD	Neut 21.6%, Lymph 0.7%, Mac 74%, Eos 3%

BALF  Bronchoalveolar lavage fluid, Eos  Eosinophils, HP  Hypersensitivity pneumonitis, ILD Interstitial lung disease, Lymph  Lymphocytes, Mac  Macrophages, Neut  Neutrophils

Fig. 4.

Receiver operating characteristic curve comparing the Multidisciplinary Discussion (MDD) diagnosis with Bronchoalveolar Lavage Fluid (BALF) results (Step 4) with MDD without BALF data (Step 3, reference standard). The Area Under the Curve (AUC) indicates the ability of BALF inclusion to enhance consensus in interstitial lung disease diagnosis. CI =Confidence Interval

Discussion

Accurate diagnosis of ILD requires integration of clinical, radiological, and occasionally pathological data through multidisciplinary discussion, which is widely considered the reference standard for ILD classification [3, 10]. In this cross-sectional study, we evaluated the incremental contribution of BALF analysis to the MDD for the diagnosis of ILD within a structured, stepwise diagnostic framework. By sequentially introducing new input, we were able to quantify how additional information influenced diagnostic agreement and confidence among clinicians involved in ILD evaluation.

Our results demonstrated that diagnostic agreement improved progressively as additional data was incorporated into the decision-making process. A high level of agreement (κ = 0.905) was observed after multidisciplinary discussion, consistent with the established role of MDD in ILD diagnosis [3, 10]. Importantly, the addition of BALF findings resulted in a further increase in interobserver agreement (κ = 0.965) and diagnostic confidence. Although the numerical increase in kappa was modest, improvements at this level of agreement among expert clinicians are clinically relevant, as they occurred in cases that remained diagnostically indeterminate after standard clinical and radiological evaluation.

In our study population, BALF findings guided consensus in 3 unresolved cases, highlighting its value when initial agreement could not be reached. In these cases, the presence of marked lymphocytosis favored final agreement for diagnosing sarcoidosis and organizing pneumonia, and in parallel, neutrophil predominance supported the diagnosis of idiopathic pulmonary fibrosis. These findings highlight the potential clinical relevance of BALF in the dynamics of MDD.

Regarding diagnostic confidence, the proportion of evaluations with high diagnostic confidence (≥ 90%) increased from 3.4% to 20.7% after the inclusion of BALF data in the MDD. This increase was primarily observed in cases of sarcoidosis (stage II disease), largely due to the presence of CD4/CD8 ratio > 3.5, as previous studies have demonstrated that such ratio is associated with the presence of a granuloma in the lung parenchyma or mediastinal lymph node [5]. In certain clinical scenarios, particularly when biopsy is not feasible due to the patient’s condition, BALF analysis — a less invasive procedure with a lower complication rate — may be sufficient to support a definitive diagnosis of sarcoidosis.

Currently, international guidelines recommend BALF analysis in the context of HP, as the presence of lymphocytosis increases the level of diagnostic confidence when histopathology is not available [4, 5, 7]. However, lymphocytic pattern was not often observed in HP patients involved herein, and this had a direct impact in final diagnosis. In this study, two patients initially diagnosed as HP were ultimately identified as unclassifiable ILD in step 4, since BALF showed neutrophilic pattern associated with a low-confidence HP diagnosis in step 3. The lack of lymphocytosis, associated with no exposure history in one patient, and no clear sign of airway involvement on HRCT, rendered HP a less probable diagnosis in both cases, changing the final diagnosis. Additionaly, only two of the 6 HP patients, who had non-fibrotic disease, presented lymphocytosis. In contrast, the remaining four cases, all classified as fibrotic HP [lung fibrosis extent on HRCT of 20%, 45%, 50%, and 50% [11]], exhibited neutrophilia, which has been associated with fibrotic lung diseases [12]. This observation corroborates previous evidence indicating that the likelihood of BALF lymphocytosis decreases substantially when fibrosis exceeds 20% of lung parenchyma on computed tomography images [13]. These findings suggest that, in the context of fibrotic HP, BALF cellularity may have limited role in distinguishing it from other fibrotic diseases.

Our study has several strengths. The prospective design, a structured four-step MDD process with blinded radiologists and pulmonologists, and standardized BALF protocols enhance the methodological rigor, and allows each patient to serve as an internal comparator across different stages of evaluation. This approach makes it possible to assess the incremental contribution of specific diagnostic components, such as BALF analysis, beyond information obtained from clinical assessment and HRCT. Few studies have systematically examined how BALF findings influence the dynamic process of multidisciplinary ILD diagnosis [10, 14], and our results provide additional empirical data on how this commonly used diagnostic test contributes to real-world clinical decision-making in a tertiary ILD referral center, supporting external applicability. Moreover, the use of Fleiss’ kappa and structured confidence scales provides reproducible and quantifiable measures of agreement and certainty.

This study has some limitations. First, the population size was modest and derived from a single tertiary referral center, which restrict statistical power and may limit generalizability to other clinical settings. Second, convenience sampling strategy via referral-based enrollment may also have introduced selection bias. Third, the heterogeneous disease distribution prevented disease-specific subgroup analyses. Larger multicenter studies will be important to confirm the incremental diagnostic value of BALF across specific ILD subtypes. Finally, although our analysis focused on diagnostic agreement and confidence, future studies incorporating longitudinal clinical follow-up may help further clarify the relationship between multidisciplinary consensus and long-term diagnostic accuracy.

Conclusion

In conclusion, BALF analysis adds incremental but important value to the ILD diagnostic pathway. Its primary role lies in consolidating diagnostic confidence in certain scenarios, such as sarcoidosis and enabling consensus when clinical and imaging data are insufficient. MDD remains the cornerstone of ILD diagnosis, but BALF is a minimally invasive technique that can provide value information when there is no consensus in MDD. Future multicenter studies with larger cohorts are needed to refine the role of BALF in specific ILD phenotypes and to define biomarkers that may further increase its discriminative power.

Abbreviations

BALF

Bronchoalveolar lavage fluid

CD

Cluster of differentiation

DLCO

Diffuse lung carbon monoxide capacity measurement

FVC

Forced vital capacity

HP

Hypersensitivity pneumonitis

HRCT

High resolution computed tompgraphy

IPF

Idiopathic pulmonary fibrosis

IQR

Interquartile range

ILD

Interstitial lung disease

MDD

Multidisciplinary discussion

NSIP

Nonspecific interstitial pneumoniae

SPAP

Systolic pulmonary artery pressure

Authors’ contributions

Conception and design: BP, MCC, FCQM, and FFC; data acquisition: BP, NPG, NRGRV, JRFS, MMAS, MMM; data analysis: BP, NPG, NRGRV, MCB, RSR, MMB, DC, CCS, CMM, PRMR; data interpretation: all authors; manuscript draft: BP and FFC; all authors collectively revised the manuscript, approved its final version, and agreed to be accountable for all aspects of the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Funding

Supported by the Brazilian Coordination for the Improvement of Higher Education Personnel (CAPES, 001), the National Institute of Gene Therapy (INTERGEN), process no. 401811/2024-7, the National Institute of Advanced Technologies in Diagnosis and Prognosis of Chronic and Neglected Diseases (INATEC), process no. 408142/2024-3, the Brazilian Council for Scientific and Technological Development (CNPq 314603/2021-2, 496911/2024-0, 309337/2025-9 to F.F.C.), and the Rio de Janeiro State Research Foundation (FAPERJ: E-26/210.338/2022; E-26/200.222/2023; E-26/204.803/2024; E-26/200.013/2026; E-26/210.019/2026 to F.F.C.).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethical approval and consent to participate

We conducted a cross-sectional study, with prospective allocation, at a tertiary ILD center in Rio de Janeiro, Brazil, between May 2023 and December 2024. The study was approved by the Research Ethics Committee of Clementino Fraga Filho University Hospital at Federal University of Rio de Janeiro (CAAE: 66958223.9.0000.5257). All experiments were performed in accordance with the Declaration of Helsinki and all participants provided written informed consent before enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.