Interstitial lung diseases (ILDs) are a group of disorders that affect the lung parenchyma (1). Progression occurs in almost all patients with idiopathic pulmonary fibrosis (IPF), and a large proportion of those with non-IPF forms of ILDs may also develop a progressive phenotype, a condition characterized by self-perpetuating fibrosis, worsening respiratory symptoms, impaired lung function, and increased mortality despite conventional treatment (2–6). The estimated median survival time from symptom onset to death for patients with progressive pulmonary fibrosis (PPF) is 61–80 months, depending on the underlying etiology (4–6). The diagnosis of PPF requires evidence of lung function decline and a combination of physiological, radiological, and symptomatic deterioration over time (6).

Recent clinical trials have shown that antifibrotic agents, such as nintedanib and pirfenidone, can slow down the rate of lung function decline in ILD patients with progression phenotype (7–9). Nintedanib is a tyrosine kinase inhibitor that blocks multiple pathways involved in fibrogenesis and has been approved for the treatment of IPF and other fibrosing ILDs with a progressive phenotype (7). Pirfenidone is a drug with anti-inflammatory, antioxidant, and antiproliferative effects, shown to be effective in IPF and other fibrosing ILDs with a progressive phenotype (8, 9). The results of these trials suggest that there may be a common mechanism of fibrosis in ILD patients who progress to end-stage disease. However, the risk factors and prognostic indicators for PPF in ILD were not well established. Identifying these factors could help clinicians monitor and prevent the progression of PPF in ILD patients.

In clinical practice, monitoring disease progression includes various components. Clinical and laboratory data had proposed a relative decline in pulmonary function, progression fibrosis in HRCT, and elevated serum Krebs Von den Lungen-6 (KL-6), predicting progression in specific types of ILDs based on changes in variables over time (10–17). However, the risk factors and prognostic indicators of baseline variables for progression across non-IPF ILD subsets were not well established.

Therefore, in this single-center, retrospective study, we enrolled patients with non-IPF ILDs and assessed their clinical characteristics at baseline and outcomes over 1 year. We aimed to identify the features and potential risk factors associated with PPF in this population. Data on demographics, patient-reported outcomes, serial pulmonary function tests (PFTs), high-resolution computed tomography (HRCT) scores, and serum biomarkers were collected, and a multivariable logistic regression analysis was used to develop a prognostic nomogram model for predicting the likelihood of PPF in ILD patients.

We identified 336 patients from January 2015 to December 2022 (Figure 1). We excluded 29 patients who had missing data for at least one variable related to PFT. The final cohort consisted of 307 patients.

The concept of PPF arose from the observation that a substantial proportion of patients with non-IPF ILDs developed a progressive fibrosis phenotype similar to IPF, with a rapid decline in lung function and early mortality (4, 21). In this study, we developed and validated a nomogram model for predicting the risk of PPF in non-IPF ILD patients using a retrospective study. We found that combined pneumonia, low baseline DLco% pred, high mMRC marks, and a high HRCT score were significant predictors of PPF, and this model showed good performance in predicting the incidence of PPF, which enabled physicians to identify patients whose disease was likely to progress using baseline information.

Previous studies have proposed various predictors of survival in IPF or progression in specific types of ILDs based on changes in variables over time (10–13). However, the risk factors and prognostic indicators of baseline variables for progression across non-IPF ILD subsets were not well established. Following the example of Ley et al. (22), who constructed a multidimensional GAP (gender [G], age [A], and two lung physiology variables [P] [FVC% and DLco%]) index and staging system using variables that were commonly measured in clinical practice to predict mortality in IPF, we also developed and validated a model that used physiologic, radiologic, and symptomatic variables to identify patients with non-IPF ILDs who will develop a progressive fibrotic pattern. We finally derived a nomogram based on combined pneumonia, baseline DLco% pred, mMRC scores, and HRCT scores to predict ILD progression. Our study also showed that these predictors had different impacts on PPF, with DLco% pred, mMRC, and HRCT scores having more weight than combined pneumonia.

First, we found that smoking status was associated with a higher incidence of progression, as non-smokers had a lower odds ratio than ex/current smokers (OR=2.03, 95% CI, 1.19–3.49). This may be related to the smoking-induced changes in cellular function that contributed to the pathogenesis of IPF (11, 23). Although previous research suggested that reflux/dysphagia symptoms resulting from esophageal motility dysfunction and chronic microaspiration were strong predictors of FVC% pred decline over time (24) and could cause persistent alveolar epithelial damage and accelerate pulmonary fibrosis, we did not find a multivariate link between gastroesophageal reflux and PPF in this research. Moreover, no significant differences were observed in baseline age, sex status, hypertension, diabetes, TLC% pred, the level of LYM, or initial treatment between PPF and non-PPF patients.

Second, this study documented a strong association between combined pneumonia and PPF (OR = 4.57, 95% CI, 1.24–18.43, P = 0.02), which could be explained by the increased infiltration of pro-inflammatory and pro-fibrotic cells that induced the production of pro-fibrotic cytokines and progressive remodeling of the fibrotic tissue (25), indicating that the early stages of fibrotic disease may be characterized by complex inflammatory events involving both the innate and adaptive immune systems.

Third, our data corroborated recent studies indicating that the decline of the HRCT score should serve as the linchpin of PPF criteria (OR = 1.22, 95% CI, 1.07–1.42), as it could assess and quantify the range of parenchymal abnormalities, including ground-glass opacities, consolidation, and honeycombing (26). These features added prognostic information to the histopathological diagnosis, as shown in previous studies of IPF (27), RA-ILD (28), SSc-ILD (29), chronic HP (30), pulmonary sarcoidosis (31), and unclassifiable ILD (32). Furthermore, the HRCT score used in this study had been previously documented to be an independent prognostic factor in patients with acute respiratory distress syndrome (ARDS) secondary to pneumonia, as well as AE-IPF and acute interstitial pneumonia by Ichikado et al. (18, 19). In addition to our findings, their findings also supported the hypothesis that this HRCT scoring system was useful in determining the prognosis of patients with acute and progressive fibroproliferative lung disease. Of note, we observed the base of the lower lungs was the most involved lung field in non-IPF ILD, and GGA without TBE was frequently found there. Moreover, Lee et al. (33) proposed the hypothesis, elucidating that specific radiological features such as honeycombing and traction bronchiectasis were associated with a worse prognosis (33), we confirmed part of it, as honeycombing was the most prominent HRCT pattern in PPF patients in our study, while the distribution of traction bronchiectasis was similar in both groups. Other CT distributions did not show a significant correlation with the patient outcome.

Next, our data suggested that a low baseline of DLco% pred should be the strongest and most consistent predictor of PPF incidence in this non-IPF ILD cohort, which was in line with previous studies (34, 35). However, we acknowledged that isolated low DLco% could sometimes reflect a worsening of pulmonary vascular disease rather than a progression of ILD (36). Therefore, PPF should be considered when patients have a low DLco% with concurrent FVC decline (26). A lower baseline FVC% was also an established predictor of mortality in patients with progressive fibrosing ILDs, as evidenced by numerous studies spanning IPF (10, 11), RA-ILD (12), SSc-ILD (29), and chronic HP (13). In this study, we also demonstrated that PPF patients had significantly lower mean baseline FVC% pred than stable ILD patients (71.78% vs. 79.79%, P < 0.01). However, we did not observe a statistical difference between the two groups in TLC% pred.

The mMRC, a validated symptom questionnaire for various lung diseases, was used in this study to assess quality of life, disease severity, and prognosis. According to our results, the mMRC score was one of the strongest predictors of PPF. We did not use it as an independent feature, as it could introduce ascertainment bias without objective evidence of lung function decline or fibrosis progression on HRCT.

Finally, the mechanisms and differences of PPF in different ILD subtypes were poorly understood. Previous studies indicated that patients in each ILD subgroup had similar clinical phenotypes of reduced lung function, worsening symptoms, impaired quality of life, and increased mortality (37–42). Pugashetti et al. (43) confirmed a >10% decline in relative FVC and strongly predicted decreased survival in non-IPF ILD patients across different cohorts. In our study, CTD-ILD (39.09%) and non-IPF IIP (37.46%) were the most common patterns in our ILD cohort. Except for unclassifiable ILDs, there was no significant difference between the PPF and non-PPF groups. The AUC of the combined nomogram was 0.96 (95% CI, 0.92–0.99) in the CTD-ILD cohort and 0.94 (95% CI, 0.87–1) in the non-IPF IIP cohort, showing favorable accuracy and efficacy in the distinct ILD subsets. Though the sample size was insufficient to detect the efficacy of nomograms in different disease subgroups, our results supported the hypothesis that PPF may result from a common mechanism of fibrosis in various ILDs, regardless of the initial cause or association. Furthermore, our data also agreed with recent studies that suggested phenotypic variability in ILD subtypes even with PPF criteria. Without homogenization of the PPF phenotype, calibration plots showed that CTD-ILD progression rates were closer to the observed rates, while PPF in non-IPF IIP was underestimated, and a higher level of baseline PFT was observed in non-IPF IIP than CTD-ILD. These results may contribute to the hypothesis that fibrosing ILDs with a progressive phenotype had some similarities but differences also existed, and this situation may also be due to the overrepresentation of one ILD subtype.

Our study has some limitations that should be acknowledged. First, this was a single-centered, retrospective study, which was suboptimal compared to prospective trials and might have biased patient selection. Next, due to the retrospective nature of this study, the duration of follow-up varied by patient and cohort, precluding the assessment of each PPF feature over a standardized time frame. Furthermore, this study involved a small number of patients, and the number of each subtype of patients with ILD enrolled was relatively small, which meant it was not possible to draw definite conclusions regarding enrichment in certain subgroups. Prospective studies with larger samples were needed to further validate our findings. Second, we relied on the mMRC questionnaire for symptomatic worsening, which may have introduced some degree of subjective bias but was similar to the methodology used in a recent PPF clinical trial. Third, our predictive model was not validated externally by more multicenter studies with enlarged patient cohorts, leading to a lack of generalizability of this model. Fourth, since we assessed HRCT according to Ichikado et al. (18, 19), which was conducted in 2002 and 2006. Due to the increasing depth of modern studies on the image of PF-ILD, the reference value of older research methods was limited. Moreover, although reticulations were rarely observed in the cohorts enrolled in this study, when reticulations were present, we attributed them to the honeycombing type, neglecting that reticulation may be present without honeycombing, which may introduce some selection bias. Moreover, although baseline pulmonary function tests were useful for predicting prognosis, changes over time may improve predictive power, but we did not document the changes in PFT. Finally, clinical and laboratory data from recent studies have reported that serum levels of KL-6 are elevated in a variety of ILDs, including IPF and collagen vascular disease-associated interstitial pneumonia (14–17). While most patients included in this research had absent data associated with KL-6, large prospective research with comprehensive indicators should be done to further verify and optimize the predictive model.

We developed and validated a prognostic nomogram model for predicting the risk of PPF in ILD patients based on four clinical predictors: baseline DLco% pred, complicated pneumonia, mMRC scores, and HRCT scores. The nomogram showed good discrimination and calibration in both the training and validation cohorts, and it outperformed every single predictor in terms of accuracy and efficiency. We believe the nomogram could help clinicians stratify ILD patients into different risk groups and tailor their management accordingly. Our study also characterized some features of ILD patients who may develop a progressive fibrosis phenotype, such as CTD-ILD and non-IPF IIP. However, our study had some limitations, such as the retrospective design, the single-center setting, and the potential selection bias. Furthermore, longitudinal studies are needed to confirm and optimize the predictive model in different populations and settings. The nomogram is a useful tool for risk assessment and decision-making, but it is not a substitute for clinical judgment or individualized care.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

The studies involving humans were approved by the Committee on Human Research at China-Japan Friendship Hospital approved the study design (2022-KY-166-1). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

J-MG: Writing – original draft, Writing – review & editing. J-JF: Writing – review & editing. S-YX: Data curation, Methodology, Writing – review & editing. M-YJ: Methodology, Supervision, Writing – review & editing. G-LH: Visualization, Writing – review & editing.