Introduction

Idiopathic pulmonary fibrosis (IPF) is the most prevalent and lethal form of progressive lung scarring, with an average survival of only 2–5 years after the diagnosis1,2,3 and acute exacerbations in IPF (AE-IPF) being the main cause of death4. This is especially true in Japan, and thus the adequate treatment for AE-IPF is crucial5. AE-IPF is characterized by an acute, clinically notable respiratory deterioration marked by the emergence of new alveolar opacities. The diagnostic criteria for AE-IPF are as follows: (1) established or concurrent diagnosis of IPF, (2) acute onset or progression of shortness of breath within one month, (3) high-resolution computed tomography scans revealing new bilateral ground-glass opacities and/or consolidation superimposed on a usual interstitial pneumonia pattern, and (4) exacerbation that cannot be solely attributed to cardiac failure or fluid retention6. Patients who develop AE-IPF have a median survival of approximately 3–4 months7,8. There are no established blood biomarkers for AE-IPF, although Krebs von den Lungen-6 (KL-6) has been used9,10,11. Thus, the differential diagnosis of AE-IPF remains challenging due to the lack of specific biomarkers and the need to differentiate it from other conditions with similar presentations, such as pulmonary infections, thromboembolism, edema and cardiac events. The diagnostic criteria for AE-IPF rely on subjective clinical criteria with considerable inter-observer variability, imaging findings are not necessarily exclusive to AE-IPF, and the need to exclude other potential causes of acute respiratory deterioration, which is not always feasible in clinical settings. Due to their non-specific nature, these diagnostic criteria can lead to either an overdiagnosis or underdiagnosis of AE-IPF.

High-mobility group box 1 protein (HMGB1) is a nuclear non-histone chromosomal protein that has multiple functions depending on its subcellular location12. Within the nucleus, HMGB1 functions as a DNA chaperone that maintains the structure and function of chromosomes. Extracellularly, HMGB1 is damage-associated molecular pattern molecule that induces inflammatory and immune reactions through its interaction with various receptors or direct cellular uptake12. Previous studies have identified elevated HMGB1 levels in the bronchoalveolar lavage fluid (BALF) of patients with IPF13 and elevated serum levels during AE episodes11,14. These suggest the potential of serum HMGB1 as a diagnostic biomarker for AE-IPF. This study aimed to evaluate the utility of serum HMGB1 levels as a diagnostic marker for AE-IPF.

Methods

Study population and design

Study participants were selected from the clinical data obtained from the foundational research project “A Prospective Cohort Study on Fukuoka tobacco-related lung disease (FOLD) registry study”15. This prospective, multicenter observational study involved 29 major hospitals within the Fukuoka Prefecture region (Fukuoka tobacco-related lung disease registry study group). Patients diagnosed with IIPs, including CPFE and COPD, registered from September 1, 2013 to April 30, 2016 were included. Overall, 1016 patients (524 with IIPs, including 145 CPFE and 492 with COPD) were enrolled. Follow-up surveillance was performed to assess progression and acute exacerbation of the diseases and death. Whole blood samples were collected at the time of the 1st registration, annually follow-up, and acute exacerbation. From the planning stage of FOLD registry study, it was decided that (1) all enrolled cases (including IPF) were assigned to have blood samples collected at enrollment, annually, and at the time of AE-IPF, and (2) HMGB1 would be measured in blood samples and analyzed to determine if it could be used as a biomarker for AE-IPF15.

Patients with IPF were identified from the clinical data, and their serum HMGB1 measurement data were matched accordingly. The diagnosis of IPF was based on a central review with multidisciplinary discussion (MDD) using the American Thoracic Society/European Respiratory Society/Japanese Respiratory Society/Latin American Thoracic Association (ATS/ERS/JRS/ALAT) IPF 2011 guidelines and the ATS/ERS idiopathic interstitial pneumonias (IIPs) 2013 guidelines2,3. As reported in the FOLD registry study, all diagnoses of the enrolled patients were re-evaluated by a central diagnosis committee comprised pulmonologists, subspecialist radiologists, and histopathologists15. A total of 543 patients was diagnosed with IIPs, while 461 were diagnosed with chronic obstructive pulmonary disease. IPF and non-IPF IIPs were classified based on the ATS/ERS/JRS/ALAT guidelines after the MDD16.

Through MDD at each facility, AE-IPF was diagnosed when all of the following were present in patients with IPF within 1 month after excluding alternative diagnoses (i.e., pulmonary infection, pneumothorax, malignancy, pulmonary embolus, and heart failure): (1) worsening dyspnea, (2) new bilateral ground-glass opacities and/or consolidation superimposed on a background of a reticular or honeycomb pattern, (3) no evidence of infection or other identifiable cause and (4) a decrease in arterial partial pressure of oxygen of greater than 10 mmHg6. Samples from patients with AE-IPF were collected 0–5 days after admission.

Patient involvement

Patients or the public people were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

HMGB1 measurement

Serum HMGB1 levels were quantified using enzyme-linked immunosorbent assay performed by the Shino-Test Corporation (Sagamihara, Japan). The assay employed monoclonal antibodies to HMGB1 as previously described13, with a lower detection limit of 2 ng/L. All samples were analyzed twice, and the mean value was used for analysis.

Diagnostic approaches

Four distinct diagnostic approaches were evaluated to assess the utility of HMGB1 in diagnosing AE-IPF. The schema for each diagnostic approach is detailed in Fig. 1.

Fig. 1
Fig. 1
Full size image

Schematic representation of the four different diagnostic methods used to evaluate HMGB1 levels for AE-IPF diagnosis. The symbols show the HMGB1 level during the AE phase (●) and the HMGB1 level during the stable phase (○). Patients with and without AE separately were connected. The arrow with the solid line shows the diagnostic value of the AE phase and the arrow with the dotted line shows the diagnostic value of the stable phase. (A) Test-1. Diagnosis comparing the HMGB1 levels of AE and stable phases ignoring the intra- and inter-patient variations. (B) Test-2. Diagnosis by the difference from the baseline HMGB1 level shown by the two dot chain lines. The lengths of red and blue lines were compared. (C) Test-3. Diagnosis by the difference between the levels of the consecutive HMGB1 measurements. Lengths of red and blue lines were compared. (D) Test-4. Diagnosis by comparing the differences from the mean HMGB1 levels of three or more measurements during stable phase, shown by horizontal dashed lines. Lengths of red and blue lines were compared. HMGB1: high-mobility group box 1; AE: acute exacerbation.

Test-1: Diagnosis by simple comparison

HMGB1 levels were compared between during the stable phase and during the AE. Samples at stable phase were collected at enrollment and yearly intervals for the patient with AE or without AE during observation period, and data for the AE were taken at the time of AE for the patient with AE. Furthermore, all HMGB1 levels at AE-IPF were compared with all HMGB1 levels at the stable phase, disregarding baseline difference and intra-patient variation in HMGB1 level.

Test-2: Baseline difference analysis

The intra-personal differences in HMGB1 levels from the baseline value were compared between samples from the stable phase and at AE. This test considers the baseline difference between patients.

Test-3: Sequential measurement analysis

A diagnosis was made by comparing two types of differences; the difference between the HMGB1 levels measured at AE and that measured immediately before AE, and the difference between all sets of two consecutive HMGB1 levels during the stable phase.

Test-4: Mean-based analysis

Difference in HMGB1 levels between the mean levels in the stable phase and the level at the diagnosis were compared. In patients with AE, the HMGB1 level at the time of AE was considered as the level at the diagnosis. In patients without AE, the last HMGB1 level at the stable phase was considered as the level at the diagnosis. The mean HMGB1 level was determined by averaging three or more measurements during the stable phase except the last measurement for the patient without AE; this accounted for variability in HMGB1 levels due to the small sample size.

Calculation of positive and negative predictive values

The positive predictive value (PPV) and negative predictive value (NPV) are parameters of significant interest in clinical utility. These values are influenced by sensitivity and specificity, as well as the prevalence of the condition in the target population (i.e., the proportion of patients with AE out of the total number of patients). This prevalence can vary depending on the diagnostic test used (Tests-1 to 4) and the method of test administration. For example, higher prevalence is expected when tests are selectively administered among patients with symptoms of AE versus when tests are conducted routinely for all patients. Thus, PPV and NPV were calculated across various prevalence rates to provide material for future consideration. The formulas used to calculate these values are as follows:

PPV = (Sens × Prev)/(Sens × Prev + (1 − Spec) × (1 − Prev)).

NPV = (Spec × (1 − Prev))/((1 − Sens) × Prev + Spec × (1 − Prev)).

where Prev = prevalence; Sens = sensitivity; and Spec = specificity.

Statistical analysis

Diagnostic accuracy was evaluated using the area under the receiver operating characteristic curve (area under the curve [AUC]-receiver operating characteristic [ROC]). Poor, moderate, and high accuracy was defined using AUC-ROC values of 0.50–0.70, 0.70–0.90, and > 0.90, respectively. An AUC-ROC of 0.50–0.70 represented poor accuracy, 0.70–0.90 represented moderate accuracy, and values higher than 0.90 represented high accuracy.

All data management, aggregation, and statistical analyses were conducted using Stata version 17.1 (Stata Corp., College St. TX. USA). The ROC curves were plotted, the AUC and its 95% confidence intervals (CI) were estimated, and the difference between the ROC-AUCs were statistically tested using the rocreg and rocregplot commands utilizing the bootstrap method with 5,000 resamples17. During estimation, the multiple diagnostic values from a single patient were treated as a cluster. The CIs were estimated using the percentile method. The association between HMGB1 levels at the first exacerbation and subsequent overall survival was analyzed using Cox regression. Two-tailed P-values of < 0.05 were considered statistically significant.

Single-cell analysis for HMGB1

Single-cell RNA-sequencing (scRNA-seq) data were obtained from the publicly available dataset GSE135893, which included annotated lung cell populations from patients with IPF (n = 12) and healthy controls (n = 10)18. The raw Seurat objects provided in the original study were utilized. Data were processed in R (v4.3.1) with the Seurat package (v5.0). Cells annotated as either “control” or “IPF” were retained, while other diagnoses were excluded. Low-quality cells had already been filtered in the original dataset. Data normalization and scaling followed the standard Seurat workflow. Principal component analysis was conducted, followed by uniform manifold approximation and projection (UMAP) for visualization. Cell type annotations were used as provided in the original dataset. UMAP plots were generated to visualize clustering by cell type and diagnosis. HMGB1 expression was compared between IPF and control samples for each annotated cell type.

Statistical testing was conducted using the Wilcoxon rank-sum test as implemented in Seurat’s FindMarkers function. P-values were adjusted for multiple testing using the Benjamini–Hochberg method to control the false discovery rate (FDR). Differential expression was considered significant if FDR was < 0.05, and the absolute log2 fold change was ≥ 0.25. The analysis only included cell types with ≥ 20 cells per group. Dot plots were generated to summarize HMGB1 expression, with dot size representing the percentage of cells expressing HMGB1 (> 0 counts) and dot color indicating average expression. Significant comparisons (Wilcoxon, FDR < 0.05) were indicated by an asterisk (*).

Results

Study population and HMGB1 measurements

Out of 1016 individuals screened, 317 patients with IPF were identified, yielding 1311 observational data points. Serum HMGB1 levels were measured in 269 patients across 784 instances, and five measurements taken more than 3 months from the clinical observation date (all during the stable phase) were excluded. The final analysis included 779 HMGB1 measurements. The cumulative observation period amounted to 505.6 person-years, with an average follow-up duration of 1.88 years. The baseline characteristics of the patients are described in Table 1 and Table S1. During the study period, 46 measurements were obtained during AE events. The number of AE events and summary statistics for HMGB1 are presented in Table S2. The number of diagnostic measurements available for the four approaches (Tests-1 to 4) are different owing to the differences in the calculation methods. Comparing the median measurements, all diagnostic methods exhibit higher HMGB1 values in the AE phase than in the stable phase.

Table 1 Baseline characteristics of patients with idiopathic pulmonary fibrosis with HMGB1 measurements data are presented as mean (standard deviation) and median (25th percentile, 75th percentile).
Full size table

Diagnostic performance of HMGB1

Upon evaluating the diagnostic accuracy of HMGB1 across four different approaches (Tests-1 to 4), HMGB1 levels were consistently elevated during the AE phase compared to the stable phase in all tests. Increasing diagnostic accuracy from Test-1 to Test-4 was noted, with AUC values of 0.758, 0.773, 0.788, and 0.888, respectively. The complete AUC values and their CIs are presented in Table S3, with their corresponding ROC curves shown in Figure S1.

Specificity at fixed sensitivities for HMGB1

The specificities of each test and its 95% CIs at fixed sensitivities of 0.80, 0.85, 0.90, and 0.95 are presented in Table 2. Figure 2 depicts the ROC curves overlaid with the specificity and its 95% CIs at these sensitivities. Test-4 demonstrated the highest specificity of 0.919 at fixed sensitivities of 0.80 and 0.85, with a degree of uncertainty indicated by wide 95% CIs of 0.293–0.989 and 0.145–0.978, respectively. At a sensitivity of 0.90, the highest specificity was seen for Test-3 (0.398), followed by Test-1 (0.322), Test-4 (0.318), and Test-2 (0.259). At a sensitivity of 0.95, the highest specificity was seen with Test-2 (0.239), followed by Test-1 (0.229), Test-4 (0.156), and Test-3 (0.124). For Test-4, there was a sharp drop in specificity from sensitivities of 0.85 to 0.90, as shown in Fig. 2.

Fig. 2
Fig. 2
Full size image

ROC curves with 95% CIs for specificity at fixed sensitivities of 0.80, 0.85, 0.90, and 0.95 (indicated by red dots and horizontal error bars) for each HMGB1 diagnostic method. (A) Test-1: Single measurement comparison showing varying specificity ranges at different sensitivity thresholds. (B) Test-2: Baseline difference method demonstrating specificity CIs at fixed sensitivity levels. (C) Test-3: Consecutive measurement comparison displaying specificity ranges at fixed sensitivities. (D) Test-4: Mean comparison method showing notably wide CIs for specificity, particularly at higher sensitivities, despite having the highest overall AUC of 0.888. The diagonal line in each panel represents random chance (AUC = 0.5). AUC: area under the curve; CI: confidence interval; ROC: receiver operating characteristic; HMGB1: high-mobility group box 1.

Table 2 Statistical performance of HMMGB1 at fixed specificities for each diagnostic test.
Full size table

From Test-1, HMGB1 value of 4.15 ng/mL had a sensitivity of 0.80 and specificity of 0.508 with a single measurement. From Test-4, HMGB1 higher than the mean HMGB1 level at the stable phase by 1.02  ng/mL had a sensitivity of 0.85 and specificity of 0.919.

PPV across various prevalences for HMGB1

The prevalence of AE-IPF can vary due to several factors, and this ranged from 0.06 to 0.14 as shown in Table S2. At sensitivities of 0.80 to 0.85, Test-4 showed higher PPVs compared to the other tests, with PPV values > 0.3 and > 0.52 expected for a prevalence of 0.05 and ≥ 0.10, respectively. Although the PPVs for Tests-1 to 3 were relatively lower, they still showed PPV > 0.14 at prevalence ≥ 0.10. Meanwhile, at prevalence ≤ 0.10, all tests demonstrated NPV > 0.95, which is considered high. For sensitivities of 0.90–0.95, the PPVs were relatively lower, with PPV of < 0.14 even at a prevalence of 0.10.

Comparison with KL-6 in AE-IPF

The diagnostic performance of KL-6 was evaluated across 1307 measurements using the approach in Test-1. Direct comparisons with HMGB1 were performed using 764 paired measurements. The AUC of KL-6 was 0.583 (95% CI: 0.518–0.651). In the paired analysis, HMGB1 had an AUC of 0.771 (95% CI: 0.680–0.854), whereas KL-6 had an AUC of 0.631 (95% CI: 0.535–0.719). The diagnostic performance of HMGB1 was significantly better than that of KL-6 (p < 0.01). The summarized statistics and ROC curves are presented in Table S4 and Fig. 3, respectively.

Fig. 3
Fig. 3
Full size image

Comparison of diagnostic performance between KL-6 and HMGB1 for detecting acute exacerbation of IPF. (A) ROC curve for KL-6 alone, showing limited diagnostic value with an AUC of 0.583 (95% CI: 0.518–0.651). (B) Comparative ROC curves of HMGB1 (red line) and KL-6 (black line) using 764 measurements in which both markers were simultaneously assessed. HMGB1 demonstrated superior diagnostic performance with an AUC of 0.771 (95% CI: 0.680–0.854) compared to KL-6’s AUC of 0.631 (95% CI: 0.535–0.719). The diagonal line in each panel represents random chance (AUC = 0.5). AUC: area under the curve; CI: confidence interval; ROC: receiver operating characteristic; HMGB1: high-mobility group box 1; KL-6: Krebs von den Lungen-6; IPF: idiopathic pulmonary fibrosis.

HMGB1 expression in the lungs

To explore the potential cellular sources of HMGB1, we analyzed the single-cell RNA-sequencing dataset (GSE135893) of lung tissue from patients with IPF (n = 12) and healthy controls (n = 10). The dataset included 31 annotated cell types (Fig. 4A and B). HMGB1 transcripts were broadly detected across nearly all cell types, consistent with its role as a widely expressed nuclear protein. Quantitative analysis revealed significantly higher HMGB1 expression in selected IPF cell populations compared to controls (Fig. 4C). In particular, proliferating epithelial cells, T cells, and macrophages exhibited HMGB1 upregulation. These findings indicate that although HMGB1 expression is largely ubiquitous, specific epithelial and immune subsets demonstrate increased expression in IPF lungs.

Fig. 4
Fig. 4
Full size image

Single-cell analysis of HMGB1 expression in lungs of patients with idiopathic pulmonary fibrosis (IPF) and controls. (A) Uniform manifold approximation and projection (UMAP) of single-cell RNA-sequencing data (GSE135893), annotated by cell type. (B) UMAP colored by diagnosis (red, control; cyan, IPF), demonstrating overlap of groups across the cellular landscape. (C) Dot plot showing HMGB1 expression across cell types stratified by diagnosis. The size of each dot represents the percentage of cells that express HMGB1 within the indicated cell type, and color intensity represents the average expression level. Asterisks (*) denote significant differential expression between IPF and control samples as determined by the Wilcoxon rank-sum test with Benjamini–Hochberg false discovery rate correction (FDR < 0.05, |log2 fold change| ≥0.25).

Discussion

IPF is characterized by repeated injury to alveolar epithelial cells and a subsequent dysfunctional repair process that results in excessive deposition of extracellular matrix (ECM). Based on these mechanistic pathways, biomarker research in IPF can be categorized into alveolar epithelial dysfunction, immune dysregulation, ECM remodeling, epigenetic regulation, and metabolic alterations. Several biomarkers have been reported in each category; however, most are markers of the stable phase of IPF, with few studies on AE-IPF19. High levels of markers reflecting epithelial injury, such as osteopontin and KL-6, have shown promise as AE-IPF biomarkers20. As acute exacerbations significantly worsen prognosis in IPF, future research should focus on precision medicine approaches, biomarker development, and novel targeted therapies21.

Our single-cell analysis confirmed widespread expression of HMGB1 across diverse lung cell populations in IPF group and controls, with notable upregulation in proliferating epithelial cells, T cells, and macrophages in IPF. These results are consistent with the biology of HMGB1, a highly conserved nuclear protein that acts as a damage-associated molecular pattern when released into the extracellular space. Although HMGB1 is present in nearly all mammalian cells, its extracellular release is typically triggered by necrosis, apoptosis, or immune activation. In IPF, repetitive epithelial injury and immune dysregulation provide conditions that may facilitate HMGB1 release12. During acute exacerbations, widespread epithelial injury and increased immune activation are likely to amplify extracellular HMGB1 levels. Stressed or necrotic alveolar epithelial cells, along with activated macrophages and lymphocytes, represent plausible major contributors. Importantly, our single-cell analysis reflects transcript expression in lungs with stable IPF and does not directly demonstrate HMGB1 release during acute exacerbations. Thus, although the results highlight candidate cell types, further mechanistic studies—including functional assays and cell-type-specific release analyses—are needed to clarify the dominant cellular sources and triggers of HMGB1 release in AE-IPF.

This study investigated the diagnostic accuracy of serum HMGB1 levels for AE-IPF. Four different methods were used, all of which demonstrated high accuracy (AUC > 0.75). The highest accuracy was exhibited by Test-4, which compared the mean of multiple HMGB1 measurements during the stable phase against HMGB1 levels at the time of diagnosis within the same individual (AUC = 0.888). Test-4 also had a remarkably high specificity of 0.919 at sensitivities of 0.80 and 0.85. However, the specificity CIs for Test-4 were considerably wider compared to the other diagnostic tests, and the specificity significantly decreased when the sensitivity was set above 0.90.

From a practical perspective, Test-1 is useful because AE-IPF can be diagnosed with a single measurement of HMGB1. In line with this, a previous study reported that higher serum HMGB1 levels were found in AE-IPF compared to stable IPF22. In contrast, Test-2 needs at least one HMGB1 measurement during the stable phase before AE, whereas Tests 3 and 4 require multiple measurements. Notably, Test-4 requires three measurements during the stable phase before a diagnosis can be established. Thus, selecting the most appropriate diagnostic method depends on the availability of HMGB1 measurements for each patient.

Determining the optimal cutoff value is crucial when applying these diagnostic methods in clinical practice. However, the ideal cutoff value or sensitivity setting cannot be statistically ascertained due to the variabilities in PPV with prevalence and the consequent changes in the rates of false positives and false negatives. Additionally, the costs and benefits in cases of misdiagnosis must also be considered. Although the need to correctly diagnose an AE is extremely high and the risks associated with false positives may be negligible, treatment based on a false positive diagnosis could cause undue burden on the patient. These caveats should be carefully considered when setting the sensitivity and determining the cutoff value. Thus, we calculated PPVs at various sensitivity settings and prevalences (Table S2). However, the degree of false positives and negatives that can be tolerated must be discussed from a clinical perspective. Nevertheless, the PPV remained relatively high even at a low prevalence, and the decrease in PPV is limited even at a high sensitivity. Therefore, the diagnosis of AE based on HMGB1 levels is believed to have sufficient practicality for clinical use. We propose the cutoff value of HMGB1 at AE-IPF to be 4.15 ng/mL. This value was proposed based on the results of Test-1, which has an advantage on diagnosing AE-IPF over other tests examined in this study, in that AE-IPF can be diagnosed with a single HMGB1 measurement. As an additional criterion, AE-IPF can be diagnosed if the HMGB1 level is elevated by 1.02 ng/mL or more compared to the mean HMGB1 level at the stable phase. This value was proposed based on the Test-4, which needs three or more measurements during the stability phase.

Compared to other biomarkers, HMGB1 exhibited a remarkable diagnostic ability for AE-IPF. Previous reports have found that KL-6 was elevated in AE-IPF compared to stable IPF, suggesting its potential as diagnostic biomarker for AE-IPF9,10,11, but its diagnostic accuracy has not been thoroughly investigated. Several reports have suggested the usefulness of KL-6, HMGB1, and others as biomarkers, these were all small-scale, retrospective studies. A retrospective study by Collard et al. included 87 cases of IPF and 47 cases of AE-IPF, but no central diagnosis was performed9. Yamaguchi et al. reported that serum HMGB1 levels were significantly higher in patients with stable IPF versus healthy controls, while these were higher in patients with AE-IPF versus both patients with stable IPF and healthy controls. However, their study only included 17 cases of AE-IPF, and HMGB1 levels at the time of diagnosis had an AUC of 0.57 for identifying AE-IPF11, making it insufficient for diagnosis.

The present study directly compared the diagnostic performance of HMGB1 and KL-6 by simultaneously measuring both markers in the same patient cohort. HMGB1 demonstrated a significantly higher diagnostic accuracy versus KL-6 (AUC 0.771 vs. 0.631), strongly suggesting the superiority of HMGB1 as a superior diagnostic biomarker for AE-IPF. Another study by Kata et al. among patients with IIP reported that the serum levels of heme oxygenase (HO)‑1, a heat shock protein expressed exclusively on anti-inflammatory M2 macrophages, were higher in AE versus the stable phase23. The AUC for serum HO‑1 to detect an AE was 0.87, which is comparable to the diagnostic accuracy of HMGB1 found in our study. These findings highlight the potential of HMGB1 as a reliable and efficient diagnostic biomarker for AE-IPF, which can outperform other currently available biomarkers. To examine the association between HMGB1 levels and prognosis, Cox regression was performed in 29 patients with HMGB1 measured during the initial exacerbation. The hazard ratio per 1 ng/mL increase in HMGB1 was 0.97 (95% CI: 0.93–1.03, P = 0.41). This indicates that HMGB1 levels and prognosis did not show significant association.

The limitations of this study must be acknowledged. Since this was a study conducted on a single ethnic group, the generalizability of the findings may be limited. Additionally, including HMGB1 measurements from patients with IPF experiencing acute events other than AE, such as heart failure or bacterial pneumonia, would have allowed for a more comprehensive evaluation of the diagnostic capabilities of HMGB1.

In conclusion, this prospective cohort study demonstrated the high diagnostic utility of HMGB1 in AE-IPF. Additionally, HMGB1 had significantly better diagnostic value than KL-6, and its use as a biomarker can potentially facilitate the earlier diagnosis and treatment of AE-IPF.