FRED scoring system for interstitial lung disease: Predicting exacerbations requiring admissions need for long-term oxygen therapy and mortality

Dear Editor, Interstitial lung disease (ILD) refers to a heterogeneous group of disorders with varying etiologies and presentation.[1] In India, its incidence is approximately 15 per 100,000 population,[2] with a 4-year mortality rate reaching 46%.[3] While tools like the gender, age, and physiology (GAP)[4] score are validated for predicting mortality in idiopathic pulmonary fibrosis (IPF), they have limited applicability in non-IPF ILDs. A unified, accessible prognostic tool applicable across all ILD types is lacking. We developed and evaluated a novel composite scoring system, the FRED (F: FVC and Functional, R: Radiological severity, E: Etiology, and D: DLCO) ILD score, to predict outcomes such as exacerbations requiring hospitalisation, long-term oxygen therapy (LTOT), and mortality. An ambispective study (2020–2024) was conducted in the ILD clinic of a North Indian tertiary hospital. Patients with complete FRED ILD data, ≥1 year follow-up, and good treatment compliance were included. Institutional Ethics Committee approval was obtained. The FRED ILD score at the time of diagnosis (range 0–7) was constituted using variables as mentioned in Table 1.Table 1: Variables used in FRED ILD scoreThe rationale of using four components of the FRED ILD score is based on the multi-dimensional prognostic factors in ILD. Etiology plays a key role because idiopathic ILD generally carries a worse outcome than connective tissue disease-related ILD or sarcoidosis. Physiological impairment assessed through FVC <50% and DLCO <40% is strongly associated with poor survival[5,6] and often indicates the need for early transplant evaluation.[7] Functional status measured by a reduced 6-minute walk distance (6mWD <50%) predicts higher mortality,[8] while values above 80% correlate with better outcomes.[9] Radiological severity, such as involvement of more than 10 lung segments or the presence of honeycombing or traction bronchiectasis, reflects advanced, irreversible fibrosis and indicates a poorer prognosis.[10,11] All patients underwent HRCT, FVC, DLCO, and the 6-minute walk test (6MWT). 6mWD was expressed as % predicted using an Indian population reference formula: 491.93− (2.148 × age) + (107.07 × gender; female = 0, male = 1) with a variability of 47%.[12] Data were analysed using SPSS v25.0, with P < 0.05 considered significant. Categorical variables were expressed as frequencies and percentages. Correlations between the FRED ILD score and clinical outcomes (hospitalisation, LTOT, and mortality) were evaluated using Pearson’s or Spearman’s correlation coefficient (r) and coefficient of determination (R2). Receiver operating characteristic (ROC) curve analysis was performed to determine the score cut-off to obtain the highest area under the curve (AUC) for outcomes. Kaplan–Meier survival curves were constructed to evaluate survival differences between patient groups stratified by the FRED ILD score cut-off, and the log-rank test was used to assess statistical significance. In addition, multi-variable logistic regression analysis was performed to identify independent predictors of hospitalisation, LTOT, and mortality, with coefficients (B), odds ratios (OR), and model fit indices (McFadden’s and Nagelkerke R2). A subgroup analysis was performed for IPF using ROC to compare the FRED ILD score to the GAP score. A total of 102 ILD patients were studied (79 retrospective, 23 prospective). The demographics details and types of ILD are as shown in Table 2. During 1-year follow-up, 53 required hospitalisation, 47 LTOT, and 31 died, with the highest rates seen in the IPF group.Table 2: Demographic details and types of ILDFor predicting exacerbations requiring admissions, FRED ILD score ≥4 correlated strongly and significantly with the need for hospitalisation (r: 0.83, R2:0.69, P < 0.0001) and had a relative risk (RR) of 23.5 for admission within a year. ROC analysis yielded an AUC of 0.976 with sensitivity and specificity both >95% [Figure 1].Figure 1: Receiver operator curve for exacerbations requiring hospitalisation, need for long-term oxygen therapy, mortality, GAP index, and FRED ILD score. (a) ROC curve showing the predictive ability of the score for exacerbations requiring hospitalisation (AUC = 0.976). (b) ROC curve showing the predictive ability of the score for the need for long-term oxygen therapy (AUC = 0.952). (c) ROC curve showing the predictive ability of the score for mortality (AUC = 0.812). (d) ROC curves comparing the score and the GAP score for predicting mortality in idiopathic pulmonary fibrosis. The score demonstrated a higher area under the curve (AUC = 0.9438) compared to the GAP score (AUC = 0.8875), indicating superior discriminative ability. Abbreviations: ROC = receiver operating characteristic; AUC = area under the curve; FRED = F: FVC and Functional, R: Radiological severity, E: Etiology, and D: DLCO; ILD = Interstitial Lung Disease; GAP = Gender Age Physiology; LTOT = long-term oxygen therapyFor predicting the need for LTOT, FRED ILD score ≥4 correlated strongly and significantly with the need for LTOT (r: 0.79, R2:0.62, P < 0.0001), with RR of 9.94 for requirement of LTOT within 1 year with AUC of 0.952 with a sensitivity of 91.49% and a specificity of 81.82% with 86% of accuracy [Figure 1]. For predicting mortality, FRED ILD score ≥4 correlated moderately and significantly with the mortality (r: 0.5, R2:0.25, P < 0.0001) with RR of 6.24 for mortality within 1 year, AUC of 0.812, a sensitivity of 87.10%, a specificity of 63.38%, and an accuracy of 70.59% [Figure 1]. Among 102 patients, 21 had IPF. Mortality prediction was compared using GAP and FRED ILD scores. The AUC for GAP (cut-off ≥4) was 0.888, while FRED ILD (cut-off ≥4) was higher at 0.944 [Figure 1]. Kaplan–Meier analysis showed patients with FRED ILD ≥4 had significantly reduced survival versus those <4 (log-rank P < 0.0001), confirming its strong prognostic value [Figure 2]. On multivariable logistic regression, the models showed good overall fit for hospitalisation (McFadden’s R2 = 0.73) and LTOT (McFadden’s R2 = 0.91), though no individual parameter was independently significant. For mortality, the model fit was moderate (McFadden’s R2 = 0.44), with 6mWD emerging as the only independent predictor.Figure 2: Kaplan–Meier survival curves showing cumulative survival by FRED ILD score. Kaplan–Meier analysis illustrating cumulative survival over time in patients stratified by the FRED ILD score. Patients with a FRED ILD score ≥4 (blue line) demonstrated significantly reduced survival compared to those with a score <4 (red line) (log-rank P < 0.0001). The shaded areas around each curve represent the 95% confidence intervals for the survival function. Abbreviations: FRED = F: FVC and Functional, R: Radiological severity, E: Etiology, and D: DLCO; ILD = Interstitial Lung Disease; CI = confidence interval.Currently, no composite scoring system exists for ILDs of all etiologies. Multiple scores, such as GAP, TNMCGAP[13], and ACDS[14], predict mortality in IPF but are not applicable to non-IPF ILDs. While our score shares certain parameters, such as FVC and DLCO, with the ILD-GAP model, it additionally incorporates etiology, radiology, and functional parameters. The FRED ILD score was developed to predict prognosis across ILD types, integrating four domains, etiology, physiology, functional status, and radiology, for a comprehensive assessment. Strengths include inclusion of diverse ILD subtypes and 1-year follow-up. The limitations are its ambispective study, single-center design, lack of an external validation, and modest sample size. The relatively small sample size further restricts the statistical power and generalisability of the results. Future multicentric prospective studies with larger cohorts are needed to validate the proposed scoring system. A FRED ILD score ≥4 identifies high-risk patients, suggesting its utility for transplant referral, closer monitoring, and risk stratification in ILD management. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.