Study design

Patients meeting the IPAF criteria, as defined by the ERS/ATS research statement,¹ were recruited in this prospective, observational, single‑center study carried out in the Department of Pneumology, Medical University of Lodz, between March 2021 and November 2022. The control group recruited at the same time comprised patients with AIRD‑ILD. AIRD diagnosis was established according to the European Alliance of Associations for Rheumatology / American College of Rheumatology classification criteria.^20-24 Exclusion criteria comprised age under 18 years, lack of informed consent, a history of malignancy at baseline, current respiratory tract infection or infection within the past 4 weeks, COVID‑19 during follow‑up, and incomplete relevant medical data.

Each patient was assessed by both a pulmonologist and a rheumatologist at baseline and after 12 months. At both time points, medical history and demographics were taken, and each patient underwent physical examination, spirometry, a 6‑minute walk test (6MWT), high‑resolution computed tomography (HRCT), and their transfer factor for carbon monoxide (TL_CO) and laboratory test results were evaluated. In the IPAF group, the autoantibody panel was performed (antineutrophil antibody 2 [ANA2], ANA3 profile in all patients, myositis and scleroderma profiles if needed, based on a multidisciplinary team decision) at both time points to assess seroconversion. For the AIRD patients, baseline or historical autoantibody titers were considered. BALF analysis was based on current or historical results, provided they were obtained within the last 2 years. The duration from symptom onset was measured in years. Presenting symptoms included both respiratory and systemic manifestations, except for the Raynaud phenomenon. Dyspnea was assessed using the modified Medical Research Council scale. Gender‑age‑physiology (GAP) index and staging system were calculated.²⁵

Data were collected on immunosuppressive therapy administered during follow‑up. For glucocorticosteroids, clinically relevant use was defined as oral prednisone at a dose of at least 5 mg daily for a minimum duration of 3 months or administration of intravenous pulse therapy.

This study was a part of a larger research project approved by the Ethics Committee of the Medical University of Lodz (RNN/43/21/KE). Written informed consent was obtained from all the enrolled patients.

Procedures

Spirometry and TL_CO assessments were performed in accordance with the ERS and ATS recommendations using the Lungtest 1000 system (MES, Kraków, Poland).^26,27 Pulmonary function test (PFT) results were reported as a percentage of predicted values (%pred), based on the Global Lung Function Initiative reference equations. The TL_CO results were presented with a correction to hemoglobin concentration. The 6MWT was performed at both time points by the same person, following established guidelines.^28,29 Bronchoscopy with BALF analysis was performed for diagnostic purposes according to the Polish Respiratory Society (PRS) recommendations.³⁰ The site of lavage was selected based on HRCT analysis, targeting lung segments with the most pronounced interstitial involvement. BALF was processed manually by a trained investigator under sterile conditions to ensure sample integrity and reduce interoperator variability. Cellular analysis of BALF included total and differential cell counts, and was conducted promptly following collection.

HRCT scans were independently assessed and scored by 2 experienced pulmonologists and ILD experts (WP and EM). The quantitative analysis (%ILD) was conducted following the recommendations of the PRS experts.³¹ Radiological patterns were classified following the ATS/ERS guidelines.^17,32

Assessment of interstitial lung disease progression

Progression was determined using 3 distinct criteria: PPF following the 2022 ATS/ERS/JRS/ALAT guidelines, PF‑ILD based on the INBUILD criteria, and absolute FVC decline by at least 10%.^17,18 Each criterion was assessed separately. Death due to respiratory failure or lung transplantation was considered as disease progression. The patients who developed malignancy during follow‑up were excluded from the analysis. To ensure consistency of the follow‑up observations, a uniform 12‑month period was applied to each definition.

Statistical analysis

Continuous variables with a normal distribution, as verified by the Shapiro–Wilk test, were presented as means with SD. Variables not normally distributed were reported as medians with interquartile ranges. Continuous variables were compared between 2 independent groups using the t test or the Mann–Whitney test, depending on the distribution. Categorical variables were compared using the χ² test or Fisher exact test, based on the distribution of expected frequencies. ANA titers were treated as a categorical variable and classified into 3 groups: 1) titers below 1:100, 2) titers between 1:100 and 1:320, and 3) titers above 1:320.

A univariable analysis was performed initially, and the results were reported as odds ratios (ORs) with corresponding 95% CIs. A multivariable logistic regression analysis utilizing backward‑forward stepwise selection was subsequently conducted to determine independent predictors, with results also reported as OR with 95% CI. The predictive ability and model accuracy were assessed by constructing a receiver operating characteristic (ROC) curve and calculating the area under the ROC curve (AUC).

A 2‑tailed P value below 0.05 was considered significant. All analyses were performed using Statistica 13.3 package (Polish version; StatSoft Polska, Kraków, Poland).

Patient characteristics

A total of 69 white patients were recruited, including 31 with IPAF and 38 with AIRD‑ILD. The AIRD‑ILD group comprised patients with systemic sclerosis (SSc; n = 18), rheumatoid arthritis (RA; n = 5), idiopathic inflammatory myopathies (IIMs; n = 5), systemic lupus erythematosus (n = 2), Sjögren syndrome (SS; n = 1), and overlapping AIRDs (n = 7): RA‑SS (n = 2), SSc‑IIM (n = 2), SSc‑RA (n = 1), SS‑mixed connective tissue disease (n = 1), and RA‑IIM (n = 1).

The demographic and clinical characteristics of the participants are presented in detail in Table 1. The results of the qualitative HRCT analysis are provided in Supplementary material, Table S1.

There were no significant differences in PFTs or peripheral blood neutrophil cell counts in the IPAF and AIRD‑ILD groups. In the BALF analysis, numerically higher total cell count and eosinophil count were observed in the patients with AIRD‑ILD than those with IPAF. However, these differences did not reach significance (Table 1).

Results of the annual follow‑up analysis

During 1‑year follow‑up, 4 patients from the IPAF group and 3 from the AIRD‑ILD group died from nonmalignant causes. In the IPAF group, all patients died due to respiratory failure. In the AIRD‑ILD group, 1 patient died from respiratory failure, 1 from gastrointestinal complications of SSc, and the cause of death of the other patient remains unknown.

One patient in each group developed malignancy (IPAF, pancreatic cancer; AIRD‑ILD, lung cancer), and they were excluded from further analysis. Three IPAF patients progressed to a defined AIRD: 1 developed SS, 1 SSc, and 1 IIM (Figure 1). All developed new clinical symptoms without seroconversion.

Eight IPAF patients did not receive immunosuppressive therapy during follow‑up, while 23 were treated with various combinations of immunosuppressive drugs (glucocorticosteroids, azathioprine, mycophenolate mofetil, tocilizumab). All patients with AIRD‑ILD were treated with immunosuppressants. Two IPAF patients received nintedanib.

In the IPAF group, 13 out of 30 patients (43.3%) fulfilled at least 1 definition of ILD progression or died due to respiratory failure (excluding 1 patient who was removed from the analysis due to malignancy during follow‑up). Figure 2 presents the analysis of IPAF patients regarding disease progression based on different definitions (INBUILD, PPF, FVC decline ≥10% + deaths). A detailed characterization of individual patients with IPAF, including follow‑up outcomes, is provided in Supplementary material, Table S2, while Table S3 summarizes baseline pulmonary function parameters (FVC% and TL_CO%) in the patients classified as progressors and nonprogressors.

Regardless of the adopted definition, numerically more patients in the IPAF than AIRD‑ILD group developed disease progression. However, this difference was significant only according to the PPF criteria (Table 2).

Determinants of interstitial lung disease progression in patients with interstitial pneumonia with autoimmune features

The only analyzed nonparametric clinical characteristic differentiating the IPAF patients who progressed from those who did not (according to the INBUILD and PPF criteria) was the presence of honeycombing on the baseline HRCT scan (Table 3).

The univariable logistic regression analyses were performed to assess the potential baseline predictors of disease progression according to 3 different definitions. A higher baseline FVC% was associated with a lower likelihood of progression defined by both the INBUILD criteria (OR, 0.95; 95% CI, 0.9–0.996; P = 0.04) and FVC decline by at least 10% (OR, 0.94; 95% CI, 0.89–0.998; P = 0.04). In addition, a higher neutrophil count in peripheral blood was associated with progression according to the INBUILD criteria (OR, 1.69; 95% CI, 1.13–2.51; P = 0.01) and FVC decline by at least 10% (OR, 4.21; 95% CI, 1.09–16.27; P = 0.04; Table 4).

In the multivariable logistic regression analysis, higher neutrophil count was identified as an important predictor of disease progression according to the INBUILD criteria (OR, 2.06; 95% CI, 1.19–3.59; P = 0.01), indicating that each additional unit increase in neutrophil count doubled the likelihood of progression. Furthermore, the presence of honeycombing on HRCT, a hallmark radiological feature, strongly predicted the progression of IPAF using the INBUILD criteria (OR, 69.1; 95% CI, 2.16–2220; P = 0.02). Despite the broad 95% CI suggesting variability due to limited events, the presence of honeycombing emerges as a particularly critical predictor. The logistic regression model demonstrated excellent discriminative capacity with an AUC of 0.91 (95% CI, 0.72–1), underscoring its robust predictive utility (Table 5).

For progression according to the PPF criteria, higher neutrophil count remained a solid predictor (OR, 1.89; 95% CI, 1.08–3.31; P = 0.03), indicating a substantial increase in the risk with rising neutrophil levels. Additionally, the UIP pattern on HRCT strongly predicted disease progression (OR, 43.8; 95% CI, 1.52–1260; P = 0.03). The number of lymphocytes in BALF did not reach significance (OR, 1.1; 95% CI, 0.99–1.22; P = 0.07). This logistic regression model also demonstrated a strong discriminative capability, with the AUC of 0.87 (95% CI, 0.732–1; Table 5).

Due to the limited number of patients meeting the criterion of FVC decline by at least 10%, the multivariable logistic regression analysis was not performed in this category.

Strengths and limitations

The strengths of this study include its prospective design and the setting—a center with recognized expertise in ILD care (a member of the ERN‑LUNG network for rare diseases). The study was designed and conducted by an interdisciplinary team of pulmonologists and rheumatologists, reducing the risk of misclassifying patients with subclinical AIRD as IPAF. The inclusion of consecutive patients meeting the IPAF criteria minimized the selection bias. This study has several limitations that should be acknowledged. First, although it is one of the few prospective studies investigating IPAF, the overall sample size remains limited due to the rarity of the disease. As a result, the statistical power, particularly in the subgroup analyses and multivariable models, may be insufficient to detect modest effect sizes or weaker associations. However, this limitation reflects the rarity of IPAF as an orphan condition. Comparable sample sizes were reported in other studies, including some multicenter and retrospective analyses conducted over longer periods.^8,33,35,41 Second, BALF was not available for all participants, especially within the AIRD‑ILD group, which could introduce a selection bias and limit generalizability of the findings regarding BALF cytology. Third, the relatively short follow‑up period of 12 months may not capture longer‑term disease trajectories, particularly in patients with initially stable disease. Fourth, radiological assessments were conducted by 2 experienced pulmonologists but not formally adjudicated by thoracic radiologists, which may have affected the classification of HRCT patterns, such as UIP or nonspecific interstitial pneumonia. Quantitative HRCT scoring, although based on expert guidelines, remains a subjective method with potential inter‑reader variability. Fifth, specific treatment regimens were not analyzed due to the limited cohort size and heterogeneity of therapies. We consider such an analysis potentially biased. Future multicenter studies with longer follow‑up periods, larger cohorts, and standardized therapeutic approaches are needed to confirm and expand upon these findings.

Conclusions

This prospective study confirms that disease progression in patients with IPAF is common within the first year of follow‑up and varies depending on the applied definition. Baseline peripheral blood neutrophil count and fibrotic HRCT features (UIP, honeycombing) were identified as predictors of short‑term progression, highlighting their potential value in early risk stratification. In the comparison of BALF cellular profiles, no significant differences were observed between the IPAF and AIRD‑ILD patients. Together, these findings support the need for individualized monitoring strategies in IPAF, incorporating PFTs, blood biomarkers, and imaging features. Importantly, the variability in progression rates across current definitions highlights a key limitation in applying existing classification systems, originally developed for broader fibrosing ILDs, to IPAF.