Development of a diagnostic algorithm for the differential differential diagnosis of interstitial lung disease: preliminary data from a multicenter retrospective case-control study

Background: Interstitial lung diseases (ILDs) represent a heterogeneous group of disorders with different treatment and prognosis. ILD may be the presenting or the dominant manifestation of a connective tissue disease (CTD). Multidisciplinary team (MDT) discussion is currently the diagnostic standard. However, there is no consensus on how MDT diagnosis is validated and on the core elements of discussion. Objectives: To explore the performance of a diagnostic algorithm for the differential diagnosis of ILD based on clinical, serological and radiological data, supporting clinician decision-making. Methods: In this retrospective study, analysis was performed on clinical, serological and radiological features at diagnosis and 1-year follow-up in 71 patients, including 41 with CTD-ILD and 30 with idiopathic interstitial pneumonias (IIPs). In order to identify robust hallmarks, we conducted the Receiver Operating Characteristic (ROC) curve analyses in logistic regression, to discriminate significantly different features between CTD-ILD and non-CTD-ILD groups. Results: Out of 71 patients 46% were women, with a mean age of 66±11 years. History of smoking (8.8% current and 39.8% former smokers), was more associated with IIPs. 54% of patients had dyspnea on exertion and 39% dry cough, both more frequently associated with IIPs (p = 0.016). Among radiological features, NSIP pattern was more frequent in CTD-ILD, while UIP was associated with IIP. Lung fibrosis extent was greater in IIP (p = 0.063), in which CT is generally performed in symptomatic patients at diagnosis and rarely for screening purpose. Baseline features with good performance (OR p-value ≤ 0.05) were eligible as potential candidate discriminators: age, sex, smoking habit, rheumatological signs and symptoms, autoantibodies, ILD patterns were selected, to build a multivariate model with high discrimination accuracy (AUC 0.971). The model has a sensitivity of 100% and specificity of 89.7%. The most relevant correlations between population features and CTD-ILD are presented in Table 1. Conclusion: Our study shows that the most important variables in the differential diagnosis between CTD-ILD versus IIPs include, as expected, autoimmune features (rheumatological symptoms and serological data). Questionnaire tool containing these specific hallmarks may be relevant during MDT discussion, limiting the number of misdiagnosed CTD-ILDs and potentially avoiding further unnecessary investigations. However, only prospective cohort studies of early onset ILD are needed to fully validate the relative importance of clinical, serological, functional and radiological data.