Abstract
BACKGROUND:
Although mediastinal lymph node (MLN) enlargement is common in fibrotic interstitial lung disease (ILD), its dynamic behavior and prognostic value are unclear.
RESEARCH QUESTION:
Do MLN size, count, and progression on high-resolution chest CT imaging associate with physiological decline, fibrosis progression, and transplant-free survival, and do these factors improve risk models?
STUDY DESIGN AND METHODS:
This study included a prospective cohort of patients with fibrotic ILD from a tertiary center (N = 1,122) with independent replication (n = 181). MLNs were scored by using standardized radiologic criteria; fibrosis was quantified according to deep texture analysis. Associations with outcomes were evaluated by using Cox and linear mixed-effects models; model discrimination was assessed by using Harrell’s concordance index and net benefit by decision curve analysis.
RESULTS:
Baseline MLN enlargement (≥ 10 mm) occurred in 67% and was independently associated with worse transplant-free survival (hazard ratio, 2.44; 95% CI, 2.01–2.96; P < .0001). MLN progression was associated with excess mortality (adjusted hazard ratio, 4.09; 95% CI, 2.34–7.17; P < .001) and modified the association between fibrosis progression and survival (interaction hazard ratio, 3.02; P = .001). Fibrosis remained the stronger prognostic marker, but MLN metrics added complementary information within fibrosis strata and modestly improved discrimination when added to clinical models based on ILD-GAP (gender, age, physiology) model score or progressive pulmonary fibrosis status. Findings were directionally consistent in the independent replication cohort.
INTERPRETATION:
MLN burden and progression are accessible, quantifiable imaging biomarkers associated with adverse outcomes in fibrotic ILD. Incorporating MLN metrics into existing clinical models improves risk discrimination and supports further evaluation of their role in ILD risk stratification.
Keywords: interstitial lung disease, mediastinal lymphadenopathy, prognosis, pulmonary fibrosis, transplant-free survival
Pulmonary fibrosis (PF) encompasses a heterogeneous group of interstitial lung diseases (ILDs) marked by progressive matrix deposition, immune activation, impaired gas exchange, and increased mortality.1–5 Although high-resolution CT (HRCT) fibrosis and physiological indices (FVC and diffusion capacity of the lung for carbon monoxide [DLCO]) aid risk stratification, they incompletely capture disease biology. Mediastinal lymphadenopathy is frequent in fibrotic ILD,1–10 yet remains absent from current indices (ILD-GAP [gender, age, physiology] model and 2022 American Thoracic Society/European Respiratory Society [ATS/ERS] progressive pulmonary fibrosis [PPF] criteria).11–13
Prior studies suggest that mediastinal lymph node (MLN) enlargement correlates with disease severity, but these reports have been limited by cross-sectional designs, imaging heterogeneity, and incomplete adjustment for key covariates.2,5,14–16 Prior literature has also raised the possibility that MLN features may vary by treatment exposure,2 although this has not been evaluated in a prospectively followed, systematically phenotyped cohort. The aim of the current study therefore was to assess whether MLN size, count, and longitudinal change associate with transplant-free survival (TFS) in fibrotic ILD, with independent validation in a replication cohort. Secondary objectives were to examine associations of MLN burden with fibrosis progression, physiological decline, and treatment exposure, and to determine whether adding MLN metrics improves established clinical models, including ILD-GAP and PPF.
Study Design and Methods
Details regarding the study design and method are presented in e-Appendix 1.
This prospective observational cohort study used data from the University of Chicago Interstitial Lung Disease Registry, a longitudinal institutional cohort established in September 2005. The trial was registered at ClinicalTrials.gov.17 All participants provided written informed consent under protocols approved by the University of Chicago Institutional Review Board (IRB) (IRB numbers 14163A and 17–1617).
Patients evaluated between May 2006 and December 2018 were prospectively followed up for dynamic changes in MLN features, CT fibrosis extent, pulmonary function, clinical outcomes, and survival over the succeeding 6 years. Multidisciplinary diagnosis of ILD was performed in accordance with contemporary American Thoracic Society/European Respiratory Society (ATS/ERS) guidelines,12 using clinical, physiological, radiologic, and, when available, histopathologic data.
A multidisciplinary consensus diagnosis of fibrotic ILD was required for inclusion, encompassing idiopathic pulmonary fibrosis (IPF), connective tissue disease-associated ILD (CTD-ILD), fibrotic hypersensitivity pneumonitis, interstitial pneumonia with autoimmune features, and unclassifiable ILD. CTD-ILD included Sjögren syndrome-associated ILD when present; subgroup numbers were insufficient for separate analysis. Patients with sarcoidosis, drug toxicity-related ILD, or known malignancy were excluded. Eligible participants were required to have baseline HRCT imaging of sufficient quality for MLN assessment and at least 1 postbaseline clinical follow-up assessment. Survival time was measured from the date of the baseline CT scan; therefore, patients who died or underwent transplantation within the first year remained eligible for time-to-event analyses. Serial CT analyses were restricted to prespecified longitudinal imaging subsets.
Replication analyses were performed in independent cohorts of patients with multidisciplinary ILD diagnoses from the Royal Brompton Hospital (London, United Kingdom; May 2012-December 2019; IRB number 15/SC/0101) and the University of California Davis (November 2004-December 2018; IRB numbers 585448–7 and 875917–2), each with baseline HRCT imaging suitable for MLN and fibrosis quantification.
Baseline pulmonary physiology was assessed by using standard spirometry and DLCO, expressed as percent predicted. FVC, FEV1, and DLCO values were collected at baseline and at serial follow-up visits. PPF was defined by using 2022 ATS/ERS criteria based on serial pulmonary function testing decline, radiologic progression, and symptom worsening.12 ILD-GAP scores were calculated as a composite index incorporating age, gender, FVC, and DLCO, and PPF status served as an independent categorical factor in multivariable modeling. Antifibrotic therapy (nintedanib or pirfenidone) and immunosuppressive medications (azathioprine, mycophenolate mofetil, or corticosteroids) were tracked from baseline and throughout follow-up.
HRCT Acquisition and MLN Assessment
HRCT scans were acquired at full inspiration in the supine position using institutional multidetector-row scanners. MLN size, count, and location were assessed on axial HRCT reconstructions using standardized criteria based on International Association for the Study of Lung Cancer nodal station mapping,2 and MLNs across stations 1 to 9 were evaluated, as defined by the association. Hilar and interlobar nodes (stations 10–14) were not assessed. Nodes with short-axis diameters ≥ 10 mm were classified as enlarged. A board-certified thoracic radiologist blinded to clinical and outcome data measured and recorded baseline MLN size, station, and total MLN count for all participants. Longitudinal change in MLN size (ΔMLN) was assessed in patients with serial CT scans obtained at least 12 months apart and was defined as an increase in the short-axis diameter of any previously measured MLN on subsequent CT imaging.
Quantitative fibrosis was measured by using a deep texture analysis platform (Coreline Soft), a validated convolutional neural network algorithm for fibrotic ILD.18,19 Fibrosis extent was expressed as the proportion of total lung volume classified as fibrotic by the algorithm. Changes in fibrosis scores over time were assessed for patients with follow-up HRCT scans as previously described.18,19 Quality control procedures included voxel alignment and slice thickness harmonization prior to analysis. Baseline fibrosis extent was calculated for 900 participants, and serial deep texture analysis scores were computed for longitudinal imaging subsets. Annualized fibrosis progression was estimated from linear slopes across serial CT time points.
Statistical Analysis
We estimated restricted mean survival time to 5 years and absolute TFS differences to contextualize hazard ratios (HRs), assessed model calibration at 1, 3, and 5 years by using calibration intercepts/slopes, and locally estimated scatterplot smoothing. Because the temporal relationship among MLN burden, fibrosis extent, and survival is biologically uncertain, primary Cox models omitted baseline fibrosis extent, and sensitivity analyses additionally included fibrosis extent; estimates were directionally consistent. To minimize immortal time bias, baseline survival analyses were anchored to the baseline CT date, whereas MLN progression (ΔMLN) and serial fibrosis analyses were restricted to longitudinal imaging subsets and were interpreted separately. Decision curve analysis used threshold probabilities of 5% to 30% 5-year risk, aligned with therapy escalation, transplant evaluation, and trial referral. Stratified 10-fold cross-validation with 1,000 bootstrap replications produced 95% CIs for c-indices.
A prespecified interaction between ΔMLN (binary) and fibrosis progression (continuous slope) was tested by using a Wald test, with marginal effects plotted. The primary end point was TFS; secondary end points were longitudinal FVC/DLCO and quantitative fibrosis extent. Multiplicity adjustment for secondary and exploratory analyses used a 5% false discovery rate, whereas Bonferroni correction was applied only in prespecified sensitivity analyses. Missingness was handled with multiple imputation by chained equations (20 imputations) with concordant complete case results.
The primary outcome was TFS, defined as the time from baseline CT imaging to death or lung transplantation. Kaplan-Meier survival curves were constructed to estimate TFS across MLN-defined subgroups and were compared by using stratified unadjusted log-rank tests. Cox proportional hazards models were used to assess the independent association of MLN features, including diameter, count, and longitudinal change (ΔMLN) with TFS. Thresholds for MLN diameter, count, and longitudinal change were selected based on radiologic convention and prior validation in ILD cohorts.2 MLN subgroups were defined by using 3 independent binary features (diameter ≥ 10 mm, count ≥ 3, and presence of enlarging ΔMLN over time), each modeled separately as dichotomous covariates in multivariable Cox models. Multivariable models were adjusted for age, sex, ILD subtype, baseline FVC and DLCO, BMI, smoking status, and immunosuppressive and antifibrotic therapy, and they were further refined by using stepwise covariate selection.
Sensitivity analyses additionally controlled for the presence of radiographic honeycombing and pulmonary artery diameter as a surrogate for pulmonary hypertension. Proportional hazards assumptions were assessed by evaluating time-dependent effects and inspecting Schoenfeld residuals, and all models satisfied these criteria. Multicollinearity was evaluated by using variance inflation factors, and model robustness was verified with residual diagnostics.
Secondary outcomes included longitudinal changes in FVC, DLCO, and CT-derived fibrosis extent, which were assessed by using linear mixed-effects models. Generalized linear models were used to evaluate the association between MLN features and both baseline and progression of quantitative fibrosis burden as determined by deep texture analyses. Comparisons of baseline clinical and imaging characteristics were made by using t tests, χ2 tests, or analysis of variance, as appropriate, with 10,000 bootstrap replications to generate precise 95% CIs. Bonferroni adjustment and false discovery rate corrections were applied to account for multiple testing.
We evaluated the prognostic performance of MLN-informed models for 5-year TFS relative to established clinical indices, including the ILD-GAP score and PPF status. Classifiers were constructed by using prespecified thresholds (ILD-GAP score ≥ 4, PPF positive, MLN ≥ 10 mm, and ΔMLN present or absent), alone or in combination with > 10% fibrosis on HRCT imaging. Sensitivity, specificity, predictive values, and likelihood ratios were calculated for each classifier. Discrimination was assessed by using Harrell’s concordance (C) index; the area under the receiver-operating characteristic curve was used only where applicable for binary classifier comparisons. The additive prognostic value of MLN features to ILD-GAP and PPF was assessed by changes in the C-index and decision curve analysis, with performance additionally examined across major ILD subtypes.
Data management, regression modeling, survival analyses, and decision-curve analyses were performed in Stata version 17 (StataCorp), whereas cross-validation and calibration were performed in R version 4.2.2 (R Foundation for Statistical Computing). Two-tailed P values < .05 were considered statistically significant.
Results
A total of 1,122 patients met inclusion criteria and were analyzed in the primary study cohort (e-Figs 1–3, Table 1). The mean ± SD age was 64.1 ± 12.1 years, and 53% were male. IPF accounted for 32.9% of cases, followed by CTD-ILD (21.2%), fibrotic hypersensitivity pneumonitis (12.6%), and 33.3% comprising interstitial pneumonia with autoimmune features, unclassifiable, and other ILD subtypes. Baseline pulmonary function was impaired, with a mean FVC of 64.7% predicted (SD, 18.7%) and DLCO 52.2% predicted (SD, 21.4%). Antifibrotic therapy was prescribed in 8.6% of patients, and 27.4% were receiving immunosuppressive therapy at baseline. HRCT imaging suitable for algorithmic fibrosis quantification was available for 900 patients; all 1,122 had MLN data for baseline analyses. Findings were directionally consistent in the independent replication cohort (n = 181) (e-Fig 1, e-Table 1).
TABLE 1 ].
Baseline Characteristics of Study Population (Primary Cohort, N = 1,122)
| Characteristic | MLN < 10 mm (n = 323) | MLN ≥ 10 mm (Single CT Scan) (n = 556) | MLN ≥ 10 mm (Serial CT Scans) |
P Valuea | |
|---|---|---|---|---|---|
| Stable ΔMLN (n = 64) | Increased ΔMLN (n = 179) | ||||
|
| |||||
| Age, y | 61.7 ± 13.8 | 67.0 ± 10.5 | 60.6 ± 12.0 | 60.8 ± 11.4 | < .001 |
| Male | 120 (37.2) | 344 (61.9) | 27 (42.2) | 105 (58.7) | < .001 |
| White race | 224 (69.4) | 443 (79.7) | 32 (50.0) | 125 (69.8) | < .001 |
| BMI, kg/m2 | 29.3 ± 6.8 | 29.8 ± 6.7 | 32.5 ± 7.8 | 31.2 ± 6.2 | .28 |
| Tobacco, pack-years | 16.7 ± 23.5 | 23.0 ± 26.3 | 17.3 ± 20.2 | 21.5 ± 22.5 | .002 |
| GER | 126 (42.6) | 209 (41.2) | 29 (53.7) | 89 (52.7) | .71 |
| CAD | 34 (15.6) | 105 (24.4) | 4 (8.2) | 29 (20.0) | .01 |
| Diabetes mellitus | 40 (18.3) | 95 (22.0) | 7 (14.3) | 19 (13.1) | .27 |
| Hypothyroidism | 55 (18.6) | 93 (18.3) | 7 (13.0) | 19 (11.2) | .93 |
| FVC (% predicted) | 68.4 ± 19.4 | 63.5 ± 18.3 | 61.1 ± 17.6 | 63.2 ± 18.4 | < .001 |
| FEV1 (% predicted) | 77.7 ± 21.5 | 75.8 ± 20.7 | 71.8 ± 19.0 | 74.3 ± 20.9 | .22 |
| FEV1/FVC, % | 82.5 ± 9.0 | 83.7 ± 8.6 | 85.2 ± 7.4 | 84.2 ± 8.4 | .08 |
| DLCO (% predicted) | 60.6 ± 23.8 | 46.6 ± 18.9 | 52.5 ± 19.6 | 54.8 ± 20.1 | < .001 |
| Positive ANA titer | 124 (42.9) | 267 (52.1) | 27 (50.0) | 80 (47.6) | .01 |
| WBC | 8.5 ± 3.2 | 9.5 ± 5.9 | 9.8 ± 4.3 | 9.6 ± 8.1 | .02 |
| Granulocytes | 5.6 ± 2.5 | 6.3 ± 2.7 | 5.5 ± 2.7 | 5.7 ± 2.6 | .02 |
| Lymphocytes | 1.7 ± 0.9 | 1.8 ± 1.0 | 1.6 ± 1.0 | 1.8 ± 1.0 | .26 |
| Monocytes | 0.7 ± 0.5 | 0.7 ± 0.6 | 0.7 ± 0.6 | 0.8 ± 0.5 | .76 |
| Eosinophils | 0.1 ± 0.5 | 0.1 ± 0.7 | 0.0 ± 0.0 | 0.1 ± 0.4 | .39 |
| Basophils | 0.0 ± 0.0 | 0.0 ± 0.1 | 0.0 ± 0.0 | 0.1 ± 1.0 | .47 |
| CRP | 11.4 ± 27.2 | 11.6 ± 36.5 | 9.5 ± 9.9 | 8.1 ± 10.9 | .95 |
| NT-proBNP | 260 ± 504 | 549 ± 1,617 | 602 ± 2,747 | 363 ± 758 | .03 |
| Immunosuppressive therapy | 83 (28.7) | 117 (22.2) | 38 (63.3) | 69 (41.8) | .04 |
| Antifibrotic therapy | 17 (5.3) | 43 (7.7) | 7 (10.9) | 29 (16.2) | .16 |
| HRCT honeycombing | 83 (30.4) | 236 (48.6) | 15 (28.3) | 69 (42.3) | < .001 |
| HRCT emphysema | 66 (21.8) | 142 (27.1) | 14 (25.5) | 38 (21.7) | .09 |
| Pulmonary artery diameter | 28.5 ± 4.5 | 30.9 ± 4.8 | 29.7 ± 4.8 | 29.7 ± 4.6 | < .001 |
| Aorta diameter | 32.5 ± 4.0 | 34.4 ± 4.2 | 32.9 ± 3.4 | 33.4 ± 4.6 | < .001 |
| ILD subtype | |||||
| IPF | 78 (24.2) | 225 (40.5) | 13 (20.3) | 53 (29.6) | < .001 |
| IPAF | 7 (2.2) | 16 (2.9) | 0 (0.0) | 1 (0.6) | .53 |
| fHP | 49 (15.2) | 59 (10.6) | 9 (14.1) | 24 (13.4) | .05 |
| CTD-ILD | 88 (27.2) | 84 (15.1) | 23 (35.9) | 42 (23.5) | < .001 |
| Unclassifiable/other ILD | 101 (31.3) | 172 (30.9) | 19 (29.7) | 59 (33.0) | .92 |
Data are presented as mean ± SD or No. (%). All entries represent original data without imputed values. Exception for baseline characteristics: BMI, n = 925; GER, n = 1,029; CAD, n = 844; hypothyroidism, n = 1,026; diabetes, n = 846; FVC, n = 1,002; FEV1, n = 977; FEV1/FVC, n = 965; DLCO, n = 910; ANA, n = 1,024; WBC, n = 695; CRP, n = 646; immunosuppressive therapy, n = 1,041; HRCT honeycombing, n = 979; HRCT emphysema, n = 1,062. Pack-years are presented for patients who formerly and currently smoke only . Positive ANA titer: ANA > 1:320. ANA = antinuclear antibody; CAD = coronary artery disease; CRP = C-reactive protein; CTD = connective tissue disease; DLCO = diffusion capacity of the lung for carbon monoxide; fHP = fibrotic hypersensitivity pneumonitis; GER = gastroesophageal reflux; HRCT = high-resolution CT; ILD = interstitial lung disease; IPAF = interstitial pneumonia with autoimmune features; IPF = idiopathic pulmonary fibrosis; MLN = mediastinal lymph node; NT-proBNP = N-terminal pro–B-type natriuretic peptide.
P value for comparison of MLN <10 mm vs MLN ≥ 10 mm.
MLN Progression Marks Accelerating Fibrosis
At baseline, larger MLNs and higher MLN counts were associated with worse physiology and greater fibrosis extent. Patients with MLN ≥ 10 mm had higher baseline fibrosis (3.9%; P < .001) and greater longitudinal increases in CT fibrosis (Figs 1A–1C). Median fibrosis extent was higher with enlarged MLNs in both the primary (4.5% vs 1.0%; P < .0001) and replication (5.4% vs 1.0%; P < .0001) cohorts (e-Fig 4). These percentages represent algorithm-derived fibrosis as a proportion of total lung volume. MLN diameter correlated negatively with FVC (r = −0.15; P < .001) and DLCO (r = −0.29; P < .001) and positively with fibrosis extent (r = .34; P < .001). Patients with MLN ≥ 10 mm had lower baseline FVC (−4.85%) and DLCO (−1.4%) and faster FVC decline (−2.20% vs −0.74%; P = .02) and DLCO decline (−3.19% vs −1.10%; P < .001) (e-Fig 5, Figs 1D–1E). Similar patterns were seen for MLN count, with ≥ 3 enlarged MLNs associated with greater fibrosis extent (5.0% vs 3.1%; P < .001) (e-Fig 6, Figs 2A–2B). Among 243 patients with ≥ 3 longitudinal CT scans, 74% had increasing ΔMLN and greater fibrosis progression than those with stable MLNs (e-Table 2).
Figure 1 –
A-E, MLNs, fibrosis burden, and physiological decline. A-B, Total lung fibrosis on CT imaging (percent lung volume) according to baseline MLN burden. Boxplots show Year 0, Year 2, and Year 5 values for patients with MLN size < 10 mm (A), MLN count 1 to 2 (enlarged), and MLN count ≥ 3 (enlarged) (B). P < .0001 across groups. C, Longitudinal change in CT fibrosis over 5 years according to MLN status (≥ 10 mm vs < 10 mm). Baseline fibrosis was higher with MLN ≥ 10 mm (3.9%; P < .001). Annual increase: MLN ≥ 10 mm = 1.80% (95% CI, 0.39 to 3.22); MLN < 10 mm = 1.24% (95% CI, 0.62 to 1.86); Wald χ2 test, P < .0001. D, FVC% predicted over 5 years according to MLN status. Baseline FVC was 4.85% lower with MLN ≥ 10 mm (P < .001). Annual decline: MLN ≥ 10 mm = −2.20% (95% CI, −3.55 to −0.76; P = .002); MLN < 10 mm = −0.74% (95% CI, −1.36 to −0.12; P = .02); Wald χ2 test, P < .0001. E, DLCO percent predicted over 5 years according to MLN status. Baseline DLCO was 11.4% lower with MLN ≥ 10 mm (P < .001). Annual decline: MLN ≥ 10 mm = −3.19% (95% CI, −5.04 to −1.34; P < .001); MLN < 10 mm = −1.10% (95% CI, −1.90 to −0.26; P = .01); Wald χ2 test, P < .0001. Boxplots display medians (labels), interquartile ranges (boxes), and cohort variability (whiskers). Estimates were derived from mixed-effects regression models. DLCO = diffusion capacity of the lung for carbon monoxide; MLN = mediastinal lymph node.
Figure 2 –
A-E, Longitudinal (MLN) dynamics and fibrosis burden. A-B, Total lung fibrosis on CT imaging by number of enlarging MLNs at baseline (MLN size < 10 mm and MLN counts 1–2, and ≥ 3) expressed as percent lung volume (A) and absolute volume (B). Across-group differences, P < .0001. Patients with baseline MLN size ≥ 10 mm displayed greater subsequent increases in CT-quantified fibrosis than those with MLN < 10 mm. C, Mean MLN diameter from diagnosis to Year 4 stratified according to immunosuppressive therapy (solid line/circles) vs no immunosuppression (dashed line/squares). Shaded bands are 95% CIs; numeric labels indicate the mean at each time point. D, Mean MLN diameter over time stratified according to antifibrotic therapy (solid line/circles) vs no antifibrotic therapy (dashed line/squares); 95% CIs and numeric means as in panel A. E, Violin plots of MLN counts according to year following diagnosis (Yr 0-Yr 4), grouped according to antifibrotic exposure (left panel, none; right panel, antifibrotic) and, within each, according to immunosuppressive therapy (left cluster, none; right cluster, immunosuppressive). White dots indicate medians; black bars indicate the interquartile range. Across-time differences in MLN counts by therapy strata were significant (P < .0001). MLN = mediastinal lymph node; tx = therapy.
MLN Dynamics According to Treatment Exposure
Immunosuppressive therapy exposure was associated with slower MLN progression (1.8 mm/year vs 2.9 mm/year; P = .02) (Figs 2C–2E). Patients receiving combined immunosuppressive and antifibrotic therapy had fewer enlarged MLNs than other treatment groups (P < .0001) (Fig 2E). In contrast, antifibrotic therapy alone was not associated with lower MLN size or number over time. Longitudinal analyses showed that immunosuppressive therapy exposure was associated with slower FVC decline (e-Fig 7A), whereas antifibrotic therapy exposure alone was associated with steeper decline (e-Fig 7B), likely reflecting underlying disease severity and treatment by indication. Analyses stratified according to MLN status are shown in e-Figure 8. These treatment-exposure analyses are exploratory and should not be interpreted as evidence of treatment effect.
MLN Features and Survival
Compared with the reference group (MLN < 10 mm and fibrosis < 10%), TFS was lower for MLN ≥ 10 mm with < 10% fibrosis (HR, 1.95; 95% CI, 1.52–2.52), MLN < 10 mm with ≥ 10% fibrosis (HR, 3.20; 95% CI, 1.92–5.35), and MLN ≥ 10 mm with ≥ 10% fibrosis (HR, 3.95; 95% CI, 2.99–5.20) (e-Fig 9, Fig 3A). This high-burden subgroup also exhibited steeper physiological decline and faster radiologic progression. Within fibrosis strata (< 10% and ≥ 10%), greater MLN size or count remained associated with lower TFS. ΔMLN was associated with increased mortality in univariate analysis (HR, 2.54; 95% CI, 1.69–3.84; P < .001) and following adjustment for PPF status, age, sex, tobacco exposure, and therapy (HR, 4.09; 95% CI, 2.34–7.17; P < .001) (Figs 3B, 3C, Table 2). Increasing MLN count over time was likewise associated with worse outcomes (unadjusted HR, 1.84 [95% CI, 1.26–2.69]; adjusted HR, 2.13 [95% CI, 1.33–3.41]; P = .002) (e-Fig 9, Table 2). Results were directionally consistent in the independent replication cohort (e-Table 3) and across major ILD subtypes and fibrosis severity (e-Fig 10). Although fibrosis extent carried the larger hazard signal, MLN metrics provided additive prognostic information within fibrosis strata and improved discrimination when incorporated into composite models. In a prespecified interaction model, ΔMLN and fibrosis progression were each associated with higher mortality (HRs of 2.10 and 4.07, respectively), with a significant interaction (HR, 3.02; P = .001) indicating that MLN progression modified the hazard associated with increasing fibrosis (Fig 3D).
Figure 3 –
A-F, Progressive imaging changes and mortality risk in fibrotic ILD. A-B, Kaplan-Meier survival curves illustrating the associations of MLN enlargement and HRCT fibrosis progression with transplant-free survival. In panel A, curves compare MLN < 10 mm and fibrosis < 10% (reference), MLN ≥ 10 mm with < 10% fibrosis, MLN < 10 mm with ≥ 10% fibrosis, and *MLN ≥ 10 mm with ≥ 10% fibrosis. C, Mortality hazard ratios for patients with ILD stratified according to changes in CT fibrosis, MLN size, and MLN count, with the reference group defined as no increase in any of the three parameters. D, Marginal effects interaction model illustrating the joint association of ΔMLN and fibrosis progression with mortality. E-F, Decision curve analyses comparing net benefit across MLN-informed (E) and fibrosis-informed (F) risk models. Fib = fibrosis; GAP = gender, age, physiology; ILD = interstitial lung disease; PPF = progressive pulmonary fibrosis; MLN = mediastinal lymph node.
TABLE 2 ].
Associations of MLN Features With Transplant-Free Survival
| Characteristic | HR (95% CI) | P Value |
|---|---|---|
|
| ||
| Univariate Cox regression | ||
| Enlarging MLN (increase ΔMLN size) | 2.54 (1.69–3.84) | < .001 |
| Increasing MLN count | 1.84 (1.26–2.69) | .002 |
| PT MLN stationa (stations 1–7) | 1.51 (0.93–2.44) | .093 |
| LZ MLN stationa (stations 8 and 9) | 3.26 (1.11–9.53) | .031 |
| Multivariable Cox regressionb | ||
| Enlarging MLN (increase ΔMLN size) | 4.09 (2.34–7.17) | < .001 |
| Increasing MLN count | 2.13 (1.33–3.41) | .002 |
| PT MLN stationa (stations 1–7) | 1.13 (0.64–2.01) | .66 |
| LZ MLN stationa (stations 8 and 9) | 5.41 (1.04–27.98) | .04 |
Univariate and multivariable Cox regressions compare patients with vs without enlarging MLN characteristics. HR = hazard ratio; ILD = interstitial lung disease; MLN = mediastinal lymph node; PT = paratracheal; LZ = lower zone.
Compared with patients without enlarging MLN.
Multivariable models are adjusted for ILD-GAP score components (gender, age, FVC, diffusion capacity of the lung for carbon monoxide, ILD subtype), tobacco pack-years, immunosuppressive and antifibrotic therapy.
MLN Metrics Improve Model Discrimination and Net Benefit
Adding MLN features to models based on routinely available clinical variables yielded modest incremental improvement in discrimination (Fig 3E, Table 3). Cross-validated c-indices for models with and without MLN variables are presented in Table 3, and classifier characteristics are presented in Table 4. Among these, the combination of ≥ 10% fibrosis and ΔMLN showed the highest specificity (97.3%) and positive likelihood ratio (6.51) but low sensitivity, indicating usefulness primarily for identifying a smaller very-high-risk subgroup rather than broad screening. Decision curve analysis revealed higher net benefit for MLN-informed models across intermediate-risk thresholds (Figs 3E, 3F), supporting further evaluation of MLN metrics as adjunctive markers for risk stratification and trial enrichment.
TABLE 3 ].
Discrimination of MLN-Informed Risk Models
| Characteristic | Cross-Validated c-Index (95% CI) Excluding Enlarging MLN | Cross-Validated c-Index (95% CI) Including Enlarging MLN | P Value |
|---|---|---|---|
|
| |||
| Model 1 (ILD-GAP score + tobacco pack-years) | 0.699 (0.699–0.699) | 0.715 (0.715–0.715) | < .001 |
| Model 2 (ILD-GAP score + tobacco pack-years + med) | 0.719 (0.719–0.719) | 0.729 (0.729–0.729) | < .001 |
| Model 3 (ILD-GAP score + tobacco pack-years + med + MDC + MMP-7) | 0.663 (0.663–0.664) | 0.766 (0.765–0.768) | < .001 |
| Model 4 (PPF status + tobacco pack-years) | 0.648 (0.648–0.648) | 0.652 (0.652–0.653) | < .001 |
| Model 5 (PPF status + tobacco pack-years + med) | 0.679 (0.679–0.680) | 0.686 (0.685–0.686) | < .001 |
| Model 6 (PPF status + tobacco pack-years + med + MDC + MMP-7) | 0.754 (0.753–0.754) | 0.846 (0.845–0.847) | < .001 |
Multivariable models are adjusted for ILD-GAP score components (gender, age, physiology [FVC and diffusion capacity of the lung for carbon monoxide], ILD subtype]), tobacco pack-years, immunosuppressive and antifibrotic therapy. The presence of enlarging MLN independently had a cross-validated c-index of 0.605 (95% CI, 0.605–0.605), whereas the ILD-GAP score independently had a cross-validated c-index of 0.678 (95% CI, 0.678–0.679), and PPF status independently had a cross-validated c-index of 0.615 (95% CI, 0.615–0.616). Cross-validated c-indices reflect model discrimination for 5-year transplant-free survival, comparing models with and without inclusion of enlarging MLN. Twenty repetitions of 5-fold cross-validation were used to estimate mean c-index and 95% CIs. ILD = interstitial lung disease; MDC = macrophage-derived chemokine; med = antifibrotic and immunomodulatory medications; MLN = mediastinal lymph node; MMP-7 = matrix metalloproteinase-7; PPF = progressive pulmonary fibrosis.
TABLE 4 ].
Evaluation of Classifier Performance for 5-Year Mortality Risk (n = 416)
| Classification | High ILD-GAP Score | PPF Status | ILD-GAP + MLN Status | PPF + MLN Status | > 10% Fibrosis +ΔMLN | |||||
|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||
| Outcome (+) | Outcome (−) | Outcome (+) | Outcome (−) | Outcome (+) | Outcome (−) | Outcome (+) | Outcome (−) | Outcome (+) | Outcome (−) | |
|
| ||||||||||
| High risk | 104 | 126 | 140 | 90 | 84 | 146 | 111 | 119 | 19 | 89 |
| Low risk | 43 | 143 | 70 | 116 | 26 | 160 | 40 | 146 | 3 | 108 |
| Sensitivity (95% CI) | 45.2% (38.7–51.9) | 60.9% (54.2–67.2) | 36.5% (30.3–43.1) | 48.3% (41.6–54.9) | 17.6% (10.9–26.1) | |||||
| Specificity (95% CI) | 76.9% (70.2–82.7) | 62.4% (55.0–69.3) | 86.0% (80.2–90.7) | 78.5% (71.9–84.2) | 97.3% (92.3–99.4) | |||||
| PPV (95% CI) | 70.7% (62.7–78.0) | 66.7% (59.9–73.0) | 76.4% (67.3–83.9) | 73.5% (65.7–80.4) | 86.4% (65.1–97.1) | |||||
| NPV (95% CI) | 53.2% (47.0–59.2) | 56.3% (49.2–63.2) | 52.3% (46.5–58.0) | 55.1% (48.9–61.2) | 54.8% (47.6–61.9) | |||||
| Positive likelihood ratio (95% CI) | 1.96 (1.45–2.64) | 1.62 (1.31–2.00) | 2.61 (1.76–3.88) | 2.24 (1.65–3.05) | 6.51 (1.98–21.4) | |||||
| Negative likelihood ratio (95% CI) | 0.71 (0.62–0.82) | 0.63 (0.52–0.76) | 0.74 (0.66–0.83) | 0.66 (0.57–0.76) | 0.85 (0.77–0.93) | |||||
“High ILD-GAP score” (with following component predictors: gender, age, baseline percent predicted FVC, baseline percent predicted diffusion capacity of the lung for carbon monoxide, and ILD subtype) indicates a score ≥ 4; “Status” indicates present or absent; “10% Fibrosis” indicates the extent of fibrosis on chest high-resolution CT imaging > 10%; and “ΔMLN” indicates an increase in MLN count, assessed in subset with serial chest high-resolution CT scans (n = 219). Mortality outcomes assessed as 5-year occurrence of death or lung transplantation. ΔMLN was selected for composite classifier analyses because it captures longitudinal progression and showed stronger associations than baseline MLN status in serial-imaging subsets. ILD = interstitial lung disease; MLN = mediastinal lymph node; NPV = negative predictive value; PPF = progressive pulmonary fibrosis; PPV = positive predictive value.
Discussion
The current study found that MLN metrics are associated with fibrosis progression, physiological decline, and reduced TFS in fibrotic ILD. These associations were directionally consistent across major ILD subtypes and in an independent replication cohort. MLN metrics also added prognostic information when considered alongside radiologic fibrosis extent. Patients with both MLN ≥ 10 mm and > 10% quantitative fibrosis had substantially worse TFS than those with either feature alone. Importantly, fibrosis remained the stronger prognostic marker, but MLN burden provided complementary information within fibrosis strata and improved discrimination when incorporated into composite models.
The prespecified interaction between ΔMLN and fibrosis progression suggests that MLN progression modifies the association between worsening fibrosis and mortality, although the biologic directionality of this relationship remains uncertain. In contrast to prior IPF-focused studies that evaluated MLN features,5,14,20 the current investigation contextualized MLN dynamics within the broader framework of quantitative CT fibrosis, offering a more integrative view of immunofibrotic disease behavior across the spectrum of fibrotic ILD. Treatment-exposure analyses were exploratory and should not be interpreted causally. Although patients exposed to combined antifibrotic and immunosuppressive therapy had the most favorable observed outcomes, these findings do not establish that MLN burden identifies a distinct therapeutic phenotype. Rather, they support further study of MLN metrics as adjunctive markers for monitoring, risk stratification, and trial enrichment in fibrotic ILD.
Incorporating MLN features into established clinical models yielded modest incremental improvement in discrimination when analysis was restricted to routinely available clinical variables. These findings support MLN assessment primarily as an adjunctive imaging marker for risk stratification rather than as a stand-alone determinant of clinical decision-making. MLN burden and dynamics remain biologically plausible and operationally accessible imaging features. These data support further evaluation of MLN size, count, and longitudinal change as accessible adjunctive imaging markers that may refine risk stratification and enrich future therapeutic trials. Whether MLN-informed models should meaningfully alter routine treatment decisions will require additional validation and implementation studies.
Although this study is strengthened by a large prospective cohort, centralized radiologic assessment, and independent replication, several limitations merit consideration. First, MLN evaluation was confined to thoracic mediastinal stations, and extrapulmonary nodal involvement was not assessed. However, thoracic MLNs are anatomically proximal to fibrotic lung regions and represent the most clinically relevant and accessible lymphatic compartment for imaging-based assessment. Second, imaging techniques evolved over the enrollment period, although findings were directionally consistent in the replication cohort. This heterogeneity reflects real-world clinical practice and reinforces the external validity of the study results. Third, the replication cohort was modest in size, and larger, multicenter validation is warranted. Fourth, longitudinal ΔMLN analyses required serial CT acquisition and may therefore be susceptible to selection and survivor biases despite anchoring baseline survival analyses at the date of the baseline CT scan. Fifth, residual confounding by pulmonary vascular or cardiac comorbidity remains possible despite sensitivity analyses incorporating pulmonary artery diameter and N-terminal pro–B-type natriuretic peptide levels. In addition, treatment analyses were exploratory and vulnerable to treatment-by-indication bias, particularly given the limited antifibrotic uptake early in the cohort. Finally, some composite classifiers showed high specificity but limited sensitivity, which may constrain their role to identifying smaller very-high-risk subgroups rather than broad clinical screening.
Interpretation
MLN burden and progression are dynamic, quantifiable imaging biomarkers associated with fibrosis progression, physiological decline, and mortality in fibrotic ILD. The features provide prognostic information beyond established clinical indexes and improve model discrimination when incorporated into composite risk frameworks. Future studies should validate these findings in larger multicenter cohorts, clarify mechanisms linking MLN remodeling to fibrotic lung disease, and determine how MLN-informed approaches may best complement existing risk-stratification strategies.
Supplementary Material
Additional information: The e-Appendix, e-Figures, and e-Tables are available online under “Supplementary Data.”
Take-Home Points.
Research Question:
Do mediastinal lymph node (MLN) size, count, and longitudinal change on high-resolution CT imaging associate with fibrosis progression, physiological decline, and transplant-free survival (TFS) in fibrotic interstitial lung disease (ILD), and do they add value beyond the ILD-GAP (gender, age, physiology) score and progressive pulmonary fibrosis?
Results:
In a prospective primary cohort (N = 1,122) with independent replication (n = 181), MLN enlargement (≥ 10 mm) and ΔMLN were consistently associated with lower TFS and steeper declines in FVC and diffusion capacity of the lung for carbon monoxide. Survival worsened stepwise with increasing MLN count, and within fibrosis strata, higher MLN burden remained associated with worse TFS. MLN metrics provided modest incremental improvement beyond routinely available clinical variables and helped identify a smaller very-high-risk subgroup with combined MLN enlargement and substantial fibrosis.
Interpretation:
MLN metrics are accessible, quantifiable imaging features that capture clinically meaningful disease activity and enhance risk stratification across fibrotic ILD subtypes. Incorporating MLN size, count, and ΔMLN into routine high-resolution CT assessment can sharpen prognosis and inform trial enrichment and monitoring strategies. Multicenter validation and implementation studies are warranted to standardize measurement and refine thresholds for clinical decision-making.
Funding/Support
Funding support was provided by the National Heart, Lung, and Blood Institute (K23HL146942, A. A.; T32HL007013, J. V. P.; K23HL138190, J. M. O.; and R56HL158935, J. M. O.), the Asthma+Lung UK Chair of Respiratory Research (P. L. M.), and Genentech (A. A. and J. H. C.).
Financial/Nonfinancial Disclosures
The authors have reported to CHEST the following: A. A. has received consulting fees from Genentech, Inogen, Medscape, AbbVie, PatientMpower, Brainomix, PureTech, Gossamer Bio, and Boehringer Ingelheim; has received honoraria from Boehringer Ingelheim; and is supported by a career development award from the National Heart, Lung, and Blood Institute (K23HL146942). J. M. O. has received consulting and advisory board fees from Genentech and Boehringer Ingelheim; and is supported by awards from the National Heart, Lung, and Blood Institute (K23HL138190 and R56HL158935). A. D. has received personal fees from Boehringer Ingelheim, Brainomix, AstraZeneca, GSK, and Genentech; and holds stock options in Brainomix. J. V. P. has received grant support from the National Heart, Lung, and Blood Institute (K23HL171871); serves on the CHEST Editorial Board; and has served on an advisory board for B4. C. T. L. has received research funding from the Pulmonary Fibrosis Foundation (Scholar Award) and the Walder Foundation. S. B. has received consulting fees from Boehringer Ingelheim. I. N. has received advisory board honoraria from Boehringer Ingelheim, InterMune, and Anthera; speaking honoraria from GSK; consulting fees from ImmuneWorks; and study contracts with the National Institutes of Health, Stromedix, Sanofi, and Boehringer Ingelheim. M. E. S. has received institutional research funding from the National Institutes of Health and Boehringer Ingelheim Pharmaceuticals; honoraria and medical writing support from Boehringer Ingelheim; and participation on adjudication and data safety boards for FibroGen and Bristol Myers Squibb. P. L. M. has received research support and consulting fees from Boehringer Ingelheim and Genentech. None declared (K. A. A., D. S., A. G., I. B. V., I. U., K. T., A. N. H., A. I. S., L. C., J. H. C.).
Role of sponsors:
The sponsor had no role in the design of the study, the collection and analysis of the data, or the preparation of the manuscript.
ABBREVIATIONS:
- ATS/ERS
American Thoracic Society/European Respiratory Society
- CTD
connective tissue disease
- DLCO
diffusion capacity of the lung for carbon monoxide
- GAP
gender, age, physiology
- HR
hazard ratio
- HRCT
high-resolution CT
- ILD
interstitial lung disease
- IPF
idiopathic pulmonary fibrosis
- IRB
Institutional Review Board
- MLN
mediastinal lymph node
- PF
pulmonary fibrosis
- PPF
progressive pulmonary fibrosis
- TFS
transplant-free survival
Contributor Information
Ayodeji Adegunsoye, Section of Pulmonary & Critical Care, Department of Medicine, The University of Chicago, Chicago, IL.
Justin M. Oldham, Department of Internal Medicine, University of Michigan, Ann Arbor, MI.
Katherine A. Aracena, Section of Pulmonary & Critical Care, Department of Medicine, The University of Chicago, Chicago, IL; Center for Research Informatics, The University of Chicago, Chicago, IL.
Daniel Sickinger, Coreline North America, Suwanee, GA.
Ahmadreza Ghasemiesfe, Department of Radiology, University of California Davis, Sacramento, CA.
Anand Devaraj, Royal Brompton and Harefield Hospitals, Guy’s and St Thomas’ NHS Foundation Trust, London, United Kingdom.
Janelle Vu Pugashetti, Department of Internal Medicine, University of Michigan, Ann Arbor, MI.
Cathryn T. Lee, Section of Pulmonary & Critical Care, Department of Medicine, The University of Chicago, Chicago, IL.
Inemesit Udofia, Section of Pulmonary & Critical Care, Department of Medicine, The University of Chicago, Chicago, IL.
Iazsmin Bauer Ventura, Section of Rheumatology, The University of Chicago, Chicago, IL.
Shashi Bellam, Division of Pulmonary & Critical Care, Endeavor Health, Chicago, IL.
Kiran Thakrar, Department of Medicine, and the Department of Radiology, Endeavor Health, Chicago, IL.
Aliya N. Husain, Department of Medicine, Department of Pathology, The University of Chicago, Chicago, IL.
Anne I. Sperling, Division of Pulmonary & Critical Care & Sleep Medicine, Department of Medicine, University of Virginia, Charlottesville, VA.
Imre Noth, Division of Pulmonary & Critical Care & Sleep Medicine, Department of Medicine, University of Virginia, Charlottesville, VA.
Mary E. Strek, Section of Pulmonary & Critical Care, Department of Medicine, The University of Chicago, Chicago, IL.
Philip L. Molyneaux, Royal Brompton and Harefield Hospitals, Guy’s and St Thomas’ NHS Foundation Trust, London, United Kingdom; National Heart and Lung Institute, Imperial College London, London, United Kingdom.
Lydia Chelala, Department of Radiology, The University of Chicago, Chicago, IL.
Jonathan H. Chung, Department of Radiology, The University of Chicago, Chicago, IL; Department of Radiology, University of California, San Diego, CA.
References
- 1.Maher TM. Interstitial lung disease: a review. JAMA. 2024;331(19):1655–1665. [DOI] [PubMed] [Google Scholar]
- 2.Adegunsoye A, Oldham JM, Bonham C, et al. Prognosticating outcomes in interstitial lung disease by mediastinal lymph node assessment. An observational cohort study with independent validation. Am J Respir Crit Care Med. 2019;199(6):747–759. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Adegunsoye A, Hrusch CL, Bonham CA, et al. Skewed lung CCR4 to CCR6 CD4(+) T cell ratio in idiopathic pulmonary fibrosis is associated with pulmonary function. Front Immunol. 2016;7:516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lescoat A, Lederlin M, Ballerie A, et al. Interstitial lung disease and mediastinal lymph nodes: a computed tomography-based biomarker beyond nosological and etiological borders? Am J Respir Crit Care Med. 2019;199(8):1038–1040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Sgalla G, Larici AR, Golfi N, et al. Mediastinal lymph node enlargement in idiopathic pulmonary fibrosis: relationships with disease progression and pulmonary function trends. BMC Pulm Med. 2020;20(1):249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Johannson KA, Adegunsoye A, Behr J, et al. Impact of environmental exposures on the development and progression of fibrotic interstitial lung disease. Am J Respir Crit Care Med. 2025;211(4):560–568. [DOI] [PubMed] [Google Scholar]
- 7.Raghu G, Ghazipura M, Fleming TR, et al. Meaningful endpoints for idiopathic pulmonary fibrosis (IPF) clinical trials: emphasis on ’feels, functions, survives’. Report of a collaborative discussion in a symposium with direct engagement from representatives of patients, investigators, the National Institutes of Health, a patient advocacy organization, and a regulatory agency. Am J Respir Crit Care Med. 2024;209(6):647–669. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Adegunsoye A, Kropski JA, Behr J, et al. Genetics and Genomics of pulmonary fibrosis: charting the molecular landscape and shaping precision medicine. Am J Respir Crit Care Med. 2024;210(4):401–423. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lara AR, Cosgrove GP, Janssen WJ, et al. Increased lymphatic vessel length is associated with the fibroblast reticulum and disease severity in usual interstitial pneumonia and nonspecific interstitial pneumonia. Chest. 2012;142(6):1569–1576. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Sarah MH, Abigail L. Lymphangiogenesis and its role in physiologic wound healing and the pathogenesis of pulmonary fibrosis. J Vasc Med Surg. 2015;3(6):236–238. [Google Scholar]
- 11.Ryerson CJ, Vittinghoff E, Ley B, et al. Predicting survival across chronic interstitial lung disease: the ILD-GAP model. Chest. 2014;145(4):723–728. [DOI] [PubMed] [Google Scholar]
- 12.Raghu G, Remy-Jardin M, Richeldi L, et al. Idiopathic Pulmonary Fibrosis (an Update) and Progressive Pulmonary Fibrosis in Adults: an official ATS/ERS/JRS/ALAT Clinical Practice Guideline. Am J Respir Crit Care Med. 2022;205(9):e18–e47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Pugashetti JV, Adegunsoye A, Wu Z, et al. Validation of proposed criteria for progressive pulmonary fibrosis. Am J Respir Crit Care Med. 2023;207(1):69–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wallis TJM, Gudmundsson E, Pontoppidan K, et al. Temporal progression of mediastinal lymphadenopathy in idiopathic pulmonary fibrosis. Eur Respir J. 2022;59(4):2200024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Tyker A, Ventura IB, Lee CT, et al. High-titer rheumatoid factor seropositivity predicts mediastinal lymphadenopathy and mortality in rheumatoid arthritis-related interstitial lung disease. Sci Rep. 2021;11(1):22821. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Velez TE, Yoon YM, Upadhyay V, et al. Active antigen-specific adaptive immune responses are shared among patients with progressive fibrotic interstitial lung disease. Am J Respir Crit Care Med. 2025;211(10):1823–1834. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.A study of the natural progression of interstitial lung disease (ILD). ClinicalTrials.gov identifier: NCT00470327. Accessed June 9, 2026. https://clinicaltrials.gov/study/NCT00470327 [Google Scholar]
- 18.Chae KJ, Lim S, Seo JB, et al. Interstitial lung abnormalities at CT in the Korean National Lung Cancer Screening Program: prevalence and deep learning-based texture analysis. Radiology. 2023;307(4):e222828. [DOI] [PubMed] [Google Scholar]
- 19.Kim GB, Jung KH, Lee Y, et al. Comparison of shallow and deep learning methods on classifying the regional pattern of diffuse lung disease. J Digit Imaging. 2018;31(4):415–424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sin S, Lee KH, Hur JH, et al. Impact of mediastinal lymph node enlargement on the prognosis of idiopathic pulmonary fibrosis. PLoS One. 2018;13(7):e0201154. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.