Eligible patients were consecutively recruited on the day of their appointment at the Rheumatology Department of the Principal Military Teaching Hospital of Tunis. A rheumatologist collected RA-related clinical data, evaluated the presence or absence of dyspnea, and assessed exercise-induced desaturation during the 6-min walk test (6MWT) desaturation was defined as a ≥ 4% drop in baseline oxygen saturation. Spirometry was requested if no test had been performed within the previous 3 months; the median time to spirometry was 10 days (IQR: 5–20 days), ranging from 1 to 60 days. A chest HRCT was ordered for patients whose previous HRCT was older than 1 year. Each patient was then referred to the Pulmonology Department to undergo LUS on the same day, performed by a pulmonologist with 8 years of experience. The median interval between the chest HRCT and LUS was 30 days (IQR: 15–100), with a minimum interval of 3 days and a maximum of 300 days.

LUS was performed using the EDAN ACCLARIX AX3 device. A low-frequency convex probe (2.5–5 MHz) was used with the frequency set at 5 MHz.

The 12-zone scanning protocol was applied for the examination (11). The thorax was systematically divided, on both the right and left sides, into three regions: anterior, lateral, and posterior. Each region was then subdivided into upper and lower segments, resulting in a total of 12 zones, numbered from T1 to T12.

The 12-zone protocol was chosen as it ensures a balanced assessment of anterior, lateral, and posterior thoracic areas on both sides (12), these zones have been validated as predictive of changes related to interstitial involvement (8). Moreover, the operator who performed the LUS was experienced with this protocol.

Ultrasound signs suggestive of interstitial involvement were assessed, including B-lines and pleural irregularities.

B-lines are vertical, hyperechoic reverberation artifacts that arise from the pleural line and extend to the bottom of the screen without fading (12).

The presence of three or more B-lines per scan defines a B-pattern and suggests interstitial involvement (13).

We used a semi-quantitative scoring system to classify ultrasound-detected interstitial involvement (11). Based on the total number of B-lines, we applied the following semi-quantitative classification: score 0 (≤5) indicated no interstitial involvement, score 1 (6–15) mild involvement, score 2 (16–30) moderate involvement, and score 3 (≥30) severe involvement. Pleural line irregularities defined as a discontinuity of the bright horizontal pleural line were also assessed, as they are suggestive of subpleural fibrosis (12).

Chest HRCT was performed without contrast injection using a Siemens 128-slice CT scanner with a standard protocol. Acquisition parameters included a tube voltage of 120 kV and a current of 100–200 mA; slice thickness of 1 mm; and interslice spacing of 1 mm. Bone algorithm reconstructions were generated with a lung window setting. Images were acquired at full inspiration, covering the entire thoracic area, with the patient in the supine position. Additional prone slices were obtained to rule out gravity-dependent changes. An experienced radiologist conducted both qualitative and quantitative analyses. Qualitative image assessment focused on identifying elementary lesions suggestive of ILD, including reticulations, fissural distortion, septal thickening, traction bronchiectasis, honeycombing, and ground-glass opacities. The chest HRCT pattern was categorized based on lesion type and distribution as either nonspecific interstitial pneumonia (NSIP) or usual interstitial pneumonia (UIP) (14, 15).

Quantitative analysis of ILD was performed using the Warrick semi-quantitative scoring system, which takes into account both the severity and the extent of interstitial involvement (16). A Warrick score of 0 corresponds to a normal appearance. Interstitial involvement is considered mild for scores below 8, moderate for scores between 8 and 15, and severe for scores above 15. We used a Warrick score threshold of 7 to define significant interstitial involvement. This threshold served as the dichotomous outcome measure for analyzing the diagnostic performance of LUS in detecting ILD (17).

Data analysis was performed using the Statistical Package for the Social Sciences (IBM SPSS Statistics, version 25). Quantitative variables with a normal distribution were expressed as mean ± standard deviation, while non-parametric quantitative variables were presented as median and interquartile range. Qualitative variables were expressed as frequencies and relative percentages. Comparisons of means between groups were conducted using the student’s t-test for normally distributed variables. Non-parametric tests were applied for variables that did not follow a normal distribution. A p-value ≤ 0.05 was considered statistically significant. Spearman’s correlation coefficient (r) was used to assess the relationship between lung ultrasound data and continuous variables. Receiver Operating Characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance of LUS in predicting interstitial involvement, defined dichotomously based on chest HRCT results. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of LUS were calculated using an online calculator (18).

This study received a favorable opinion from the Local Committee for the Protection of Persons (CLPP) at Hôpital Militaire de Tunis (Meeting No. 68, Decision No. 32/2023/CLPP, dated June 19, 2023). The committee, coordinated by Prof. Haroun Ouertani (Head of Endocrinology-Nutrition Service), reviewed the protocol for the study titled “Early diagnosis of rheumatoid lung: contribution of thoracic ultrasound.”

Written informed consent was obtained from all participants prior to their inclusion in the study, in accordance with CLPP requirements and prevailing Tunisian ethical standards. Participants were informed of the study’s objectives, procedures, potential risks, and their rights, including the right to withdraw at any time without prejudice.

A total of 118 patients with RA presented to our rheumatology department for their routine follow-up of rheumatic disease. After screening for eligibility and exclusion criteria, 73 patients were included and completed the study, the study flowchart is presented in Figure 1.

Among the 73 patients, there were 18 men and 55 women. The mean age was 55 ± 12 years [range: 26–81]. The mean body mass index (BMI) was 26.51 ± 4.53 kg/m². The average disease duration was 8.48 ± 7.96 years. The mean Disease Activity Score based on 28 joints using erythrocyte sedimentation rate (DAS28-ESR) was 3.47 ± 1.40. Inflammatory markers showed a median erythrocyte sedimentation rate (ESR) of 12 mm/h [interquartile range, IQR: 4.5–32] and a median C-reactive protein (CRP) level of 8 mg/L [IQR: 0–20.5]. Pulmonary function tests revealed a median forced expiratory volume in 1 s (FEV₁) of 102% predicted [IQR: 100–104], a median forced vital capacity (FVC) of 99% predicted [IQR: 90.5–101], and a median FEV₁/FVC ratio of 101% [IQR: 96–110]. Main demographic, clinical, and treatment characteristics of the enrolled patients are summarized in Table 1.

Chest HRCT identified pulmonary emphysema in 9.6% of patients (n = 7) and ILD in 28.8% (n = 21). Among those with interstitial abnormalities, a typical nonspecific interstitial pneumonia (NSIP) pattern was observed in 9.6% of cases, and a cryptogenic organizing pneumonia (COP) pattern in 4.1% (n = 3). In 15.1% of patients, the distribution and nature of the elementary lesions did not correspond to any specific interstitial pattern. Ground-glass opacities were the most frequently observed elementary lesion, reported in 19.2% of patients (n = 14). Other radiological findings included traction bronchiectasis (15.1%), reticular opacities (15.1%), septal thickening (13.7%), pulmonary fibrosis (12.3%), honeycombing (9.6%), and fissural distortion (4.1%). The median Warrick score was 3 ± 7, ranging from 0 to 27. Significant interstitial involvement, defined as a Warrick score ≥ 8, was identified in 30.1% of patients (n = 22).

Pleural irregularities were observed in 12.3%, while B-lines were present in 41.1% of patients. Notably, all patients with pleural irregularities had more than five B-lines. According to the semi-quantitative ultrasound score, interstitial involvement was detected in 21.9% of patients (n = 16), including mild involvement in 8.2% (n = 6), moderate in 11% (n = 8), and severe in 2.7% (n = 2).

A statistically significant association was demonstrated between ILD patterns on chest HRCT and the following ultrasound abnormalities: B-lines, pleural irregularities, and a B-line count > 5 (Table 2).

A moderate to strong positive correlation was observed between the total number of B-lines on LUS and the Warrick score (r = 0.685; p < 0.01).

To evaluate the diagnostic accuracy of LUS, the number of patients with and without ILD detected by LUS was compared to those with or without involvement on chest HRCT, considered as the reference standard (Table 3). The sensitivity of LUS was 59.1% (95% CI: 36.35–79.29), and the specificity was 94.12% (95% CI: 83.76–98.77). The positive predictive value (PPV) was 98.27% (95% CI: 94.74–99.45), while the negative predictive value (NPV) was 28.88% (95% CI: 19.65–40.26).

The diagnostic performance of B-line count varies notably depending on the cut-off value used. A threshold of ≥1 B-line yields a high sensitivity of 77.3% but a moderate specificity of 70.6%. Increasing the cut-off to ≥5 B-lines significantly improves specificity to 94.1%, although sensitivity decreases to 63.6%. At a threshold of ≥7 B-lines, specificity further increases to 96.1%, with a corresponding drop in sensitivity to 50.0%. Finally, a cut-off of ≥19 B-lines provides maximal specificity (100%) but with very low sensitivity (22.7%). These findings suggest that an intermediate threshold, such as ≥5 B-lines, may offer a reasonable diagnostic balance, maintaining high specificity while preserving acceptable sensitivity.

ROC curve analysis demonstrated an area under the curve (AUC) of 0.813 (Figure 2), indicating good diagnostic performance of LUS in detecting ILD. This performance was statistically significant (p < 0.001), with a 95% confidence interval ranging from 0.690 to 0.936. Analysis of the ROC curve coordinates identified the optimal cutoff at 5 B-lines, yielding a sensitivity of 63.6% and a specificity of 94.1%. The highest sensitivity (77.3%) was observed at a threshold of ≥1 B-line, whereas the highest specificity (100%) was reached at a threshold of ≥19 B-lines.