Search methodology of this narrative review

Given the narrative nature of this review about the advances in the validation of LUS for the detection of RA-ILD, the search strategy was designed to identify representative and clinically relevant studies rather than exhaustively capture all available literature. We performed a search of the published literature on this topic in PubMed until the 19th of October 2025, using the following keywords and search strategy: (connective tissue diseases [Title/Abstract] OR rheumatoid arthritis [Title/Abstract]) AND lung ultrasound [Title/Abstract]. Only original articles that focused exclusively on RA were selected after reading the abstract. Original manuscripts that included patients with additional connective tissue diseases, reviews of any kind, and meta-analysis were excluded, although these papers were read to search for relevant manuscripts that could not have been identified by the performed search.

What LUS findings suggest RA-ILD?

The identification of the interstitial involvement pattern by LUS is based on the detection of multiple vertical artifacts called B lines and/or irregularities of the pleural line [22]. The ultrasound lung pattern associated with normality is characterized by the presence of a thin, regular, hyperechoic pleural line that has a lung sliding movement, together with the visualization of horizontal artifacts called A lines, which appear as hyperechoic lines parallel to the pleural line in depth as equidistant replicas.

On the contrary, when lung aeration decreases at the subpleural level, either due to inflammatory, edematous, or fibrotic contents, B lines appear as hyperechoic vertical lines from the pleural line to the bottom of the screen without fading. They blur the A lines and move synchronously with lung sliding. Additionally, in ILD, the pleural line may show an irregular appearance with focal, diffuse, linear, or nodular thickening. Recently, these consensual definitions of the sonographic findings in ILD have reached a high level of agreement (84.2% for B lines and 82.6% for pleural line irregularity) in a Delphi process and web-reliability exercise by the Outcome Measures in Rheumatology (OMERACT) Ultrasound Working Group, using video-clips of LUS from patients with SSc with and without ILD that had been acquired by the experts [23].

Accuracy of LUS to detect RA-ILD

The investigation of the usefulness of LUS to detect the presence of ILD in AIRD began in SSc, a disease in which clinical guidelines advocate screening for SSc-ILD at diagnosis by HRCT due to its high prevalence and risk of progressive fibrosis. Published studies suggest the potential usefulness of LUS as a screening technique due to its high Se and negative predictive value (NPV) when compared with HRCT, with values close to 90–100% [24, 25].

Following these encouraging results, a growing number of studies have been published in recent years investigating the usefulness of LUS for the identification of RA-ILD [25–41]. These studies are heterogeneous in terms of the populations included, the ultrasound examination protocols used, the type of probe, the sonographic lesions assessed, their quantification, and the cutoff points chosen to define sonographic ILD. It must be considered that these methodological differences among studies, as well as the limitations of LUS reproducibility across centers, may affect reproducibility and generalizability of their results and must be seen as an important limitation regarding clinical implementation of LUS. Furthermore, they are observational studies, with limitations in their design and some of them with a small sample size, which gives them a level of evidence 4 according to the criteria of the Centre for Evidence-Based Medicine of Oxford [42], and a low quality of evidence according to the GRADE system [43]. Despite this, all the studies suggest the potential of LUS as an aiding tool in the diagnostic algorithm for RA-ILD, especially for identifying patients who are candidates for HRCT (currently the gold standard for its diagnosis), which provides consistency. Cross-sectional and retrospective studies on the accuracy of LUS to detect ILD are shown in Table 1, whereas case-control studies are displayed in Table 2.

B lines are the most studied elementary lesion, although irregularity in the pleural line is being progressively incorporated in the most recent manuscripts, which have also increased the sample size to about 200 patients. The discriminatory capacity of LUS has been analyzed in case-control studies, in which the presence and number of B lines in patients with RA-ILD were significantly higher than those found in both healthy controls and in RA patients without ILD [26–30]. Analogous results were found when evaluating pleural line abnormalities [26, 28]. Besides, pleural line abnormalities might have a greater discriminatory capacity than B lines because, while the latter have been identified in 7% of healthy controls, Moazedi-Fuerst et al. [26] did not find pleural line thickening in any of the healthy controls.

A significant positive correlation has been described between the detection of RA-ILD by LUS and HRCT, with concordances over 83%. Most works have defined the suspicion of sonographic ILD by the presence of B lines above a certain cut-off point calculated by ROC curve analysis. The high Se and NPV values found (60–100% and 70–99%, respectively), with cut-off points ranging from 5–12 B lines according to the LUS protocol used, make LUS an attractive tool for the screening or early diagnosis of RA-ILD (Tables 1 and 2).

Studies evaluating the accuracy of LUS have mostly included patients with known RA-ILD or consecutively recruited patients with RA regardless of respiratory symptoms, although most of them were symptomatic [27–40]. However, other researchers have focused their analyses on a context closer to real-world clinical practice, selecting patients with suspected RA-ILD [32, 34]. Santos-Moreno et al. [34] included 192 patients with clinical suspicion of RA-ILD (respiratory symptoms, Velcro crackles, and/or abnormal PFT). HRCT identified RA-ILD in 60.9% of patients, and LUS showed a Se of 98.3% and a NPV of 84.6%. However, the specificity (Sp) was low (14.7%), with discrepancies due to false positives of LUS conditioned by the existence of other pathologies related to RA or its comorbidities. Importantly, the data seem to suggest that LUS may be a more interesting tool for screening than chest X-ray, which showed worse Se and NPV values (29.9% and 46.4%, respectively), despite its Sp being higher (94.7%) [34].

This better performance of LUS compared to chest X-ray has also been described by Otaola et al. [33], who also found superiority of LUS over DLCO and Velcro crackles (Table 1). Along these lines, Vermant et al. [38] found that the number of B lines had a better predictive performance for ILD in HRCT than cough and dyspnea. Furthermore, Schneeberger et al. [37] observed that the association of a positive LUS (cutoff ≥ 5 B lines) with a 5-variable RA-ILD clinical risk score (sex, Velcro crackles, age ≥ 60 years, and positive RF or ACPA) improved its accuracy (Table 1). However, the combined LUS-clinical score did not exceed the accuracy of LUS alone (Table 1).

Di Carlo et al. [32] focused on high suspicion of RA-ILD defined by either the presence of respiratory symptoms or Velcro crackles, or the presence of at least two risk factors for RA-ILD, such as male sex, age > 65 years, smoking, positive RF or ACPA, in the absence of the former. RA-ILD was identified in 34.7% of patients, with favorable LUS accuracy values (Table 1). Furthermore, LUS showed a significant negative correlation with PFT, and the number of B lines was significantly higher in men [32].

Other authors have also analyzed the association of LUS with RA-ILD risk factors, describing a significant association between the number of B lines and disease activity, RF, ACPA, age, male sex, or the Health Assessment Questionnaire, and an inverse correlation with DLCO [27–29, 35, 44]. A significant correlation (r = 0.97) has also been found between LUS and KL-6, identifying that cut-off values of KL-6 and LUS of 277.5 U/mL and < 5.5, respectively, had a Se of 86.7% and 100%, and a Sp of 88% and 100%, respectively [27].

Due to the importance of early diagnosis of RA-ILD, it is key to discuss the performance of LUS as a screening tool in asymptomatic patients with RA, although existing data are still very scarce. Moazedi-Fuerst et al. [26] evaluated 64 asymptomatic patients with RA and identified B lines in 28%, subpleural nodules in 18%, and fragmented pleural lines in 4%. Eighty-nine percent of cases showed concordant findings on HRCT. Of great interest is the recent study by Otaola et al. [41], who evaluated 203 patients with RA, asymptomatic at the respiratory level, with LUS and HRCT, finding subclinical ILD in 26%, predominantly NSIP (45.3%), and with an extension < 20% (75.5%). The accuracy of LUS in this setting surpassed that of DLCO, although not all patients had PFT (Table 1). Furthermore, NPV was greater than 90% regardless of the number of RA-ILD risk factors in patients [41].

Usefulness of LUS to assess the severity of RA-ILD

B-line quantification, either numerically by adding the total number of B lines in each of the examined intercostal spaces (ICS) or by semi-quantitative scores, has been shown to be significantly associated with the severity of RA-ILD determined on HRCT using the Warrick or Goh scores. Correlation coefficients of 0.806–0.838 postulate the usefulness of LUS to estimate ILD severity [29–31, 38]. Additionally, Di Carlo et al. [32] reported a good positive correlation (r = 0.559, p < 0.0001) between the number of B lines and the percentage of HRCT fibrosis measured by a computer-aided method, selecting a cutoff value of 9 B-lines to identify significant ILD on HRCT.

Reliability of LUS in RA-ILD

Reliability is one of the aspects of LUS validation that is in a more preliminary phase, with still little evidence in the literature. Most studies have evaluated the intra- and inter-observer reliability of either B lines or pleural irregularity [26, 31, 38, 44]. For the evaluation of B lines, intra-observer kappa values of 0.76–0.98 and inter-observer values of 0.73–0.96 have been reported, with the corresponding intraclass correlation coefficients (ICC) of 0.76–0.90 and 0.86–0.97, respectively [26, 35, 38, 44]. The ICC values reported for pleural irregularities were 0.79–0.9 and 0.79–0.84, respectively [35, 44].

A first approach to assessing the impact of reader experience was made by Bandinelli et al. [44], who analyzed the inter-reader reliability of residents who had been trained by an expert for 3 months (ICC 0.86–0.94 for B lines and 0.82–0.84 for pleural line abnormalities). The authors argued that the favorable results could suggest a short learning curve for LUS, although this conclusion could be biased since the examinations had been recorded by the expert.

The type of score used could also impact reliability. In this regard, Vermant et al. [38] found higher kappa values when assessing the presence/absence of B lines than when quantifying their number.

More relevant to clinical practice are the intra- and inter-explorer studies. Cogliati et al. [31] described an inter-explorer kappa of 0.78 between assessments made by an expert sonographer using a standard machine versus a less experienced explorer using a pocket-sized machine, suggesting that the technological requirements for LUS may be less demanding than for musculoskeletal ultrasound.

Recently, Vicente-Rabaneda et al. [40] published a pilot study focused on the intra- and inter-explorer reliability of different B-line and pleural irregularity scores. In this study, three experienced sonographers blindly performed LUS in 14 patients with suspected RA-ILD. The intra-explorer kappa values for B-line (global 0.73–0.82 and binary 0.80–0.90) and pleural line (semiquantitative 0.88–0.91 and binary 0.77–0.84) scores, as well as the inter-explorer kappa values for B-line (global 0.93 and binary 0.90) and pleural line (semiquantitative 0.84 and binary 0.84) scores, were good to excellent. These findings suggest that LUS reliability can be high when performed by adequately trained experts, underscoring the need for proper operator training, which remains to be defined, as recognized in a recent consensus on the use of LUS [45].

Scanning protocols and feasibility of LUS

LUS studies investigating RA-ILD show considerable heterogeneity in examination protocols. Linear, curved, and cardiac sector probes have been used. B lines have been studied, and in some cases, pleural line irregularities have been included. The scores used have been binary (presence/absence), numerical (sum of individual findings in the different ICS), or semiquantitative, with different cutoffs depending on the protocol used.

To increase the feasibility of LUS, the accuracy of simplified protocols has been investigated. Initially, examinations included 72 ICS, although later studies have found similar results scanning 8 or 10 ICS [28, 31]. The most used protocol includes 14 ICS due to the appropriate balance between examination length and performance time (8 minutes) [27, 29, 32, 33, 35–37, 41].

At present, many aspects of LUS remain to be investigated, such as machine settings, and a consensus on the optimal protocol to use has not yet been reached. Recently, the OMERACT Ultrasound Working Group recommended that B lines and pleural irregularity should be the key lesions to assess by LUS following three Delphi consensus rounds [23]. Further research is warranted [45].

The role of LUS in monitoring RA-ILD

One of the a priori advantages of LUS for monitoring RA-ILD is its safety and accessibility, which allows for repeat examinations without subjecting the patient to ionizing radiation. Studies on the Se to change of LUS in RA-ILD and the ideal protocol for this purpose are still lacking. A recent international consensus suggests that the same probe and protocol should be used in serial evaluations [45].

Fotoh et al. [27] have suggested that LUS could be a prognostic indicator. In the one-year follow-up of patients with RA-ILD in their study, 33% died, mainly due to ILD progression. In these patients, LUS showed a higher number of B lines (> 28), with an odds ratio of progression of 1.5 (95% CI, 1.3–1.8). Despite LUS potentially being a promising tool to monitor RA-ILD, further investigation is required since Se to change has not been formally validated against HRCT progression.

Where to incorporate LUS into the RA-ILD assessment algorithm?

Various proposals have been made to incorporate LUS into the diagnostic and screening algorithm for RA-ILD. Wang et al. [46] have suggested screening RA patients from diagnosis using LUS and KL-6. In those patients with B lines > 10 or KL-6 ≥ 500 U/mL, regardless of respiratory symptoms, they recommend performing HRCT and PFT. If negative, they propose repeating LUS and KL-6 every 6 months. Below these thresholds for B lines and KL-6, the monitoring cadence would be determined by risk factors. Reichenberger et al. [36] proposed a radiation-free protocol with LUS, PFT, and screening for cardiac involvement with an echocardiogram and CPET on a treadmill. Gutiérrez et al. [29] suggested annual screening for RA-ILD from RA diagnosis, using LUS in asymptomatic patients, and with LUS and chest X-ray in symptomatic patients. HRCT would be performed if the results were positive.

Despite the current lack of consensus, considering the epidemiology of RA-ILD, it seems more reasonable to propose screening in high-risk patients, either due to symptoms or suspected signs or the presence of risk factors for RA-ILD in asymptomatic patients, as the recommendations of scientific societies are beginning to suggest [20, 21]. However, the role of LUS in these algorithms remains to be defined. It is important to emphasize that LUS should be considered a potential aiding tool to add to existing ones, and does not in any way replace HRCT, which remains the current gold standard for screening and diagnosis of RA-ILD. In Table 3, we display the main characteristics of LUS in comparison with the tools used in clinical practice to screen for ILD.