Study cohort
A total of 2,115 patients with IP from Beijing Chao-Yang Hospital were sequentially and prospectively included from January 2017 to September 2019. IP was diagnosed according to the 2013 American Thoracic Society (ATS) and European Respiratory Society (ERS) Consensus Classification of IIPs [23]. All patients with IP underwent clinical examinations, laboratory tests, high-resolution computed tomography (HRCT) of the chest, pulmonary function tests, and, if necessary, pathological examinations at their first clinical visits [23].
Among the enrolled patients, 30 patients were diagnosed with PM, 20 with DM, 33 with ADM and 41 with IPAF. IIM was diagnosed according to the European League Against Rheumatism (EULAR)/American College of Rheumatology (ACR) criteria for definite IIM, and subclassification was performed using the classification tree in the criteria [24]. IPAF was diagnosed using the ERS/ATS research statement [5]. Of the 2,115 patients with IP, 42 underwent pathological examinations of the lungs and no patients received a lung transplant.
Data collection and definitions
At the first clinical visit, the patient medical records were reviewed to uniformly extract clinical data, including demographics (age, sex, and smoking status), patient-reported information (date of IP-related symptom onset, including cough and dyspnoea), clinical manifestations, physical examinations and comorbidities.
Serological markers were obtained within one month of presentation to the clinic, including C-reactive protein, erythrocyte sedimentation rate, fibrinogen, immunoglobulin (Ig) A, IgG and IgM. Levels of autoantibodies, creatine kinase and cardiac troponin I were also recorded. MSAs, including anti-ARS (anti-Jo-1, anti-PL-7, anti-PL-12, anti-OJ, and anti-EJ), anti-SRP, anti-Mi-2, anti-MDA5, anti-TIF1γ, anti-NXP2, and anti-SAE antibodies were detected by immunoprecipitation, as previously reported [25-27].
IP was diagnosed by HRCT. All enrolled patients underwent chest HRCT with a 1-s scan time, 0.625-mm sections, and 10-mm intervals from the lung apex to the base including both lungs in the field of view. Each HRCT scan was reviewed independently by two experienced thoracic radiologists blinded to the clinical data. HRCT patterns were classified as usual interstitial pneumonia (UIP), nonspecific interstitial pneumonia (NSIP), organic pneumonia, or diffuse ground-glass opacity (GGO) according to the classification of IIPs [23]. The interobserver correlation was good. The kappa value was 0.83.
A pulmonary function test was performed for each patient. The test items included forced vital capacity (FVC) and the diffusing capacity of the lung for carbon monoxide (DLCO) using the single-breath method [28].
Smoking status was categorized into non-smokers, ex-smokers (quit smoking ≥12 months previously) and current smokers (currently smoking or quit smoking <12 months previously). Acute (or subacute) onset was defined as less than three months from symptom onset to the first clinical visit, and chronic onset was defined as a duration of more than three months. Malignancy was recorded if it occurred within three years before or after a positive detection of MSAs [29]. Pulmonary hypertension was considered if the tricuspid regurgitation velocity was ≤2.8 m/s and/or the systolic pulmonary arterial pressure was ≥37 mmHg on echocardiography [30].
Follow-up and endpoint of the study
The outcomes of this study were the progression of IP defined as a relative decrease of FVC% predicted ≥10%, a relative decrease of DLCO% predicted ≥15%, or death within 6 months of the IP diagnosis [31]. The follow-up interval was 3 or 6 months, and the follow-up period ended in September 2019. Survival time was calculated from the onset of IP-related symptoms to the outcome or end of the follow-up period.
Statistical analysis
Quantitative data are reported as the means±standard deviations or medians (interquartile ranges), and qualitative data are reported as numbers and percentages. With the TwoStep Cluster algorithm, the clustering criterion was the Bayesian Information Criterion, the distance measurement form was logarithmic likelihood, the number of clusters was automatically determined by the algorithm, and the maximum value was set as 15 clusters. The variables included in the cluster analysis were all categorical variables related to the patients’ clinical characteristics, myositis autoantibodies and imaging findings. The variables included pulmonary symptoms (including cough and dyspnoea), skeletal muscle symptoms (including proximal muscle weakness and dysphagia), UIP patterns, and MSA subtypes (anti-ARS, anti-SRP, anti-Mi2, anti-TIF1γ, anti-MDA5, anti-NXP2, and anti-SAE). These variables were available for all participants. Analysis of variance was used for comparisons of normally distributed quantitative data, the non-parametric Mann–Whitney U test was used to compare quantitative data with a non-normal distribution, and the chi-square test was used for comparisons of qualitative data. Multivariable logistic regression was applied to determine potential risk factors for acute-onset disease. Survival curves were obtained using Kaplan–Meier curves and compared with log-rank tests. A multivariable Cox proportional hazards model was constructed to identify prognostic factors. Statistical analysis was performed using SPSS software (version 23.0, IBM), and P<0.05 was statistically significant.