*4.3. Laboratory Evaluation*

A 15 mL venous blood sample was collected from each enrolled patient and from HC subjects. Serum was separated from blood samples and stored at−20 ◦C. After samples were defrosted, SPD serum levels were evaluated through a commercial ELISA kit (Biovendor, Modrice, Czech Republic), following the manufacturer's guidelines. Each sample from the same patient has been evaluated in duplicate, with a variation coefficient of 3.9%. The analytic limit for SPD detection was 0.01 ng/mL.

### *4.4. Functional Evaluation*

Spirometry and nitrogen wash-out for measurement of dynamic and static lung volumes, respectively, as well as single-breath determination of DLCO, were performed for each patient through an automated lung function testing system (Quark PFT, Cosmed, Rome, Italy) according to the standards recommended by the American Thoracic Society (ATS)/European Respiratory Society (ERS) [47,48]. All measurements were recorded as raw value and percentage of predicted value [49,50].

On the same day of PFT, incremental symptom-limited CPET was performed for every participant on an electronically braked cycle ergometer through an automated testing system (OMNIA, Cosmed, Rome, Italy), in accordance with international recommendations [19,51]. A fixed work rate increment of 10 W min−<sup>1</sup> was used. The test was continued until the point of symptom limitation (peak of exercise).

Oxygen uptake (V'O2), carbon dioxide output (V'CO2), minute ventilation (V'E), tidal volume (VT) and respiratory frequency (fR) were analysed breath-by-breath during the test. Heart rate (HR), ECG and haemoglobin saturation by pulse oximetry (SpO2) were continuously monitored whilst blood pressure was measured every two minutes from rest to peak of exercise. All measured and derived parameters [e.g., ventilatory equivalents for O2 and CO2 (V'E/V'O2 and V'E/V'CO2, respectively), end-tidal O2 and CO2 partial pressures (PETO2 and PETCO2, respectively)] were recorded and averaged every ten seconds. Lactate threshold (θL) was non-invasively estimated by the use of the dual methods approach (V-slope and ventilatory equivalents methods) [52].

V'O2 at peak exercise (V'O2 peak) was normalized for body weight and expressed also as percentage of predicted value. Peak V'E response (V'E peak) was expressed as a raw value and relative to estimated maximal voluntary ventilation (eMVV), which was defined as forced expiratory volume in the 1st second (FEV1) × 40 [19].

Ventilatory efficiency was evaluated through the analysis of the relationship between V'E (*y* axis) and V'CO2. The linear phase of the V'E/V'CO2 relationship was detected on the V'E (*y* axis) on V'CO2 (*x* axis) plot, between the beginning of loaded exercise and the end of the isocapnic buffering period, which was identified when V'E/V'CO2 increased and PETCO2 decreased. Linear regression was then applied and the V'E/V'CO2 slope and its intercept on the *y* axis were calculated. V'E/V'CO2 raw value at θ<sup>L</sup> was also recorded.

The subject's effort was considered maximal either if the respiratory exchange ratio reached ≥1.10 or if HR achieved ≥85% of maximal predicted value at peak exercise (f). CPET parameters were compared with the predicted normal values [53].

All PFT and CPET were executed and analyzed by two physicians blinded to patients' clinical and laboratory features.

#### *4.5. Imaging Evaluation*

All ACPA-positive participants underwent HRCT of the chest by the use of a multidetector scanner (Somatom Definition Siemens, Erlangen, Germany) with helical supine inspiratory contiguous acquisition (5 mm); images were reconstructed at 0.6 mm every 20 mm with high-resolution algorithms. HRCT images were reviewed by 2 radiologists who were blinded to each subject with regard to disease status, in order to evaluate the presence of subclinical parenchymal abnormalities (i.e., emphysema, fibrosis, ground-glass opacities, consolidations, nodules) and/or airways abnormalities (bronchiectasis,

airways thickening, air trapping) as described previously [10]. The occurrence of "definite UIP" or "probable UIP" pattern was considered an exclusion criterion [54].

#### *4.6. Statistical Analysis*

Continuous variables are shown as mean ± SD or as median (range) for normally and non-normally distributed data, respectively. Categorical variables are presented as frequencies. Comparisons of continuous variables between two groups were performed using an independent samples T test or Mann–Whitney U test, whilst comparisons between more than two groups were tested through the ANOVA (with Bonferroni's correction for post hoc adjustment) or Kruskal–Wallis test, according to data distribution. Chi-squared analysis tested the differences between categorical variables. Logistic regression analysis was performed to assess the strength of association between HRCT abnormalities and clinical, laboratory and functional features of interest. The predictive capacity of SPD serum levels for the presence of HRCT abnormalities was analyzed using ROC curves. Cut-offs with sensitivity and specificity to discriminate subjects with HRCT abnormalities from subjects without them were calculated. The significance of the correlations was evaluated with Spearman's rank correlation coefficient.

All statistical analyses were performed using the SPSS Statistics version 24.0 software package (SPSS Inc., Chicago, IL, USA), and a two-sided *p* value < 0.05 was considered statistically significant.
