Methods
Computer based medical records of our ILD cohort were investigated. ILD diagnosis was confirmed by evaluation of high resolution computed tomography (HRCT) images. Diagnoses of CTDs (SSc, RA, pSS, DM/PM, IPAF) were made with their corresponding classification criteria [14-18]. Demographic features, smoking history, type of symptoms, time at onset of respiratory symptoms, time at diagnosis, laboratory results, serologic tests including anti-nucleolar antibodies, extractable nuclear antigen (ENA) panel, rheumatoid factor, anti-citrullinated peptide and anti-neutrophil cytoplasmic antibodies were recorded.
Our ILD cohort was established in 2010 jointly by members from rheumatology and pulmonary departments. Since then, team members meet weekly to evaluate CTD-ILD patients (including symptoms, physiologic tests and imaging studies) and make decisions for individual treatment plans. Functional assessment of patients were performed by Modified Medical Research Council (MMRC)[19]. Pulmonary function tests (PFTs) were performed at baseline and every 6-months periods thereafter including; forced expiratory volume in 1 min (FEV1), forced vital capacity (FVC), and DLCO, expressed as percentages of predicted values adjusted fora ge, sex, height and hemoglobin level, according to techniques accepted by the American Thoracic Society (ATS)[20]. HRCT was performed at baseline then annually, or in case of a new significant respiratory symptom that suggest acute exacerbation of lung disease or significant deterioration of physiological parameters. ILD patterns were classified as non-specific interstitial pneumonia (NSIP), usual interstitial pneumonia (UIP) according to HRCT findings including ground glass opacities, traction bronchiectasis and sub-pleural honeycombing. Pulmonary involvement in HRCT in SSc patients was graded as limited or extended based on extent of reticular pattern, as suggested per consensus of experts[21]. In this consensus report, experts suggest to evaluate HRCT at 5 different levels for total disease extent of reticulation, for other rheumatologic disease ILD disease there is not any other radiological scoring system , thus same scoring system was applied for those sub-types.  Modified Gender-Age-Physiology [22] index was used for prognosis [23].
Medications used, including systemic corticosteroids, synthetic and biologic disease-modifying anti-rheumatic drugs (DMARD), AZA, CYC and were meticulously reviewed. There was not any other immunsuppresive drugs for the treatment of ILD during the induction therapy, which took six months. Clinically significant side effects were nausea/vomiting/abdominal pain impairing oral intake, increased liver function tests≥3 timesof upper limit of normal, neutrophil count of<1500/mm3, thrombocyte count <150,000/mm3, and hemorrhagic cystitis.
Primary outcome measure was treatment response at six-month after introduction of induction therapy. Radiological progression was defined as , more than 10% increase in reticular pattern on HRCT, functional progression defined as worsening in FVC greater than 10% or DLCO greater than 15%and clinical progression defined as increase in MMRC score. The patients had to have detoration in at least two of these domains in order to have progressive disease.
Categorical variables were given as numbers and percentages. Continuous data were as mean ± standard deviation (SD) or median (interquartile range). Conformity of continuous variables to normal distribution was evaluated using visual (histogram and probability graphs) and analytical methods (Kolmogorov-Smirnov / Shapiro-Wilk tests). If variables were normally distributed Student t-test was used, if not, Mann Whitney U test was used for comparison. Comparison of categorical variables was made with Chi-square or Fisher’s exact tests. A p-value of 0.05 or less was considered as statistically significant in all analyzes. Propensity score was used to explain the differences of patient groups receiving AZA and CYC. The propensity score value for each patient was calculated using a logistic regression model containing the determined variables. Two to one (One to two (1: 2)) nearest-neighbor matching and caliper width of 0.2 (caliper width of 0.2) were performed.