Development and Assessment of a Novel Lung Ultrasound Interpretation Criteria for the Detection of Interstitial Lung Disease in Systemic Sclerosis

Abstract

Objectives:

Interstitial lung disease (ILD) is a frequent complication of systemic sclerosis (SSc), and ILD screening, characterization, and monitoring are important for therapeutic decision-making and prognostication. Lung ultrasonography (LUS) is a potential alternative imaging modality for ILD detection. In this study, we develop and test a novel LUS examination technique and interpretation criteria for detecting SSc-ILD.

Methods:

LUS acquisition was performed by collecting short ultrasound movies at 14 lung positions. LUS interpretation criteria for SSc-ILD detection focused on visualized pleural changes. To assess the performance of our methodology for SSc-ILD detection, we prospectively enrolled SSc patients with high resolution computed tomography (HRCT) imaging within 3 months of LUS. LUS exams were scored independently by two blinded readers (one ultrasonographer and one non-ultrasonographer). The sensitivity and specificity for SSc-ILD detection was assessed and agreement was measured with Cohen’s Kappa statistic.

Results:

To test the performance of our LUS acquisition technique and interpretation criteria, 20 SSc patients were evaluated by LUS (278 lung zones) and HRCT. HRCT confirmed ILD in 9 patients (45%). LUS was positive for SSc-ILD in 11 patients (55%) with a sensitivity of 100% and specificity of 82% versus HRCT, with perfect agreement between the two readers (κ=1). Analysis by individual lung zones found excellent agreement between readers with 93.8% concordance and κ=0.82.

Conclusion:

We developed a novel LUS examination technique and interpretation criteria that are highly sensitive and specific for SSc-ILD detection in an SSc cohort, affording perfect agreement between ultrasonographer and non-ultrasonographer readers.

INTRODUCTION

Interstitial lung disease (ILD) is a frequent complication in systemic sclerosis (SSc), with up to 80% of patients showing lung fibrosis, and 25-30% having progressive ILD.¹ ILD is now the leading cause of death in SSc and early detection and therapeutic intervention improves outcomes.^{2, 3} High resolution computed tomography (HRCT) of the chest is currently the gold standard for SSc-ILD diagnosis. SSc-ILD presents as nonspecific interstitial pneumonia (NSIP), and less commonly, usual interstitial pneumonia (UIP) patterns of disease, typified by HRCT findings of bibasilar ground glass opacities, traction bronchiectasis, reticulation, and fibrosis.⁴ HRCT is often sought as an initial screen for ILD in new SSc diagnoses, when patients have suggestive clinical signs or symptoms, abnormalities on chest x-ray and/or pulmonary function tests (PFT), and to monitor disease progression over time.⁵ For this purpose, patients with SSc may be exposed to multiple HRCT scans during their lifetime resulting in significant cumulative radiation, cost, and time-to-imaging delays in diagnosis. Additionally, SSc patients with early or mild ILD may be overlooked for screening HRCT due to normal functional studies or subclinical disease, missing an opportunity for early intervention. Lung ultrasonography (LUS) is a potential alternative imaging modality for screening and diagnosis of SSc-ILD.⁶ Although lung aeration prevents sonographic evaluation of the lung parenchyma, the peripheral and basilar lung features of SSc-associated NSIP and UIP result in characteristic pleural and subpleural sonographic findings.^6,7 Currently, ultrasonographic findings of pleural irregularity (PI), pleural line thickening (PLT), and the presence of B-line artifacts are known to be associated with the presence of underlying ILD with high diagnostic accuracy compared to HRCT.^7-11 Although this technique shows great promise in SSc-ILD it suffers from a lack of a well-defined acquisition technique and ultrasound interpretation criteria. In this study we develop a novel LUS interpretation criteria and acquisition technique and assess its performance on an SSc cohort.

METHODS

LUS examination technique

We perform a complete LUS examination by evaluating 14 lung positions as previously described by Gutierrez et al. with the patient supine for anterior lung positions, and sitting up with their back to the examiner for posterior lung positions.¹⁰ The probe is oriented in a sagittal or pseudo-sagittal orientation (spanning adjacent ribs) rather than transverse to afford better tracking of the pleural surface. LUS examinations were performed on a GE Logiq E using a medium frequency linear probe with typical B-mode settings used in rheumatologic musculoskeletal ultrasonography. US frequency was typically 12 MHz but was adjusted downward if needed for increased signal penetration required in larger patients or lung positions with increased pleural depth. We collected 4-second movies at each of the 14 lung positions during normal respiration. For posterior paravertebral and subscapular views, the 8^th intercostal space was used as a rough guide for probe positioning; however, superior to inferior adjustments were made to visualize the most inferior portion of the lung pleura, which varied from patient to patient.

LUS interpretation criteria

Our goal in developing LUS interpretation criteria was to outline a simple set of criteria to combine with our LUS examination technique to detect SSc-ILD with high sensitivity, specificity, and inter-reader reliability. LUS B-lines, which are ring-down artifacts that originate at the pleural line and extend to the bottom of the screen,¹² can be difficult to quantify, non-specific, and their presence depends on several factors including frequency, machine settings, and technique.^8,13,14 LUS findings of pleural thickening and granularity, and defects in the pleural surface (pleural irregularity) are easy to visualize, reproducible, and strongly associated with the presence of underlying ILD (Figure 1).^9,13 Our clinical experience with LUS in SSc-ILD reflected the limitations of B-lines and benefits of assessing US features at the pleural surface. Based on these considerations, we developed a set of interpretation criteria focused on pleural irregularity, thickening, and granularity (Figure 2).

Figure 1. Interstitial lung disease findings on lung ultrasound.

(A) normal pleural surface with surrounding ribs, (B) granular pleural surface in ILD, (C) pleural discontinuities and cavitations in ILD. ILD=interstitial lung disease.

Figure 2. LUS interpretation criteria for SSc-ILD.

Preliminary LUS interpretation criteria validation

This study was approved by Stanford University’s Institutional Review Board and informed consent was obtained from all patients. We performed a prospective study enrolling adult SSc patients meeting 2013 ACR/EULAR classification criteria seen at Stanford University outpatient rheumatology clinics from February to December of 2018 with a planned or prior HRCT of the chest for any indication within 3 months of recruitment. It is our institutional practice to obtain chest HRCT for all new SSc diagnoses to screen for ILD if there are no contraindications. The presence of ILD on HRCT was confirmed by a radiologist and pulmonologist for all positive studies. The same experienced ultrasonographer (RF) with 4 years of LUS experience performed all LUS evaluations and was blinded to SSc subtype, prior HRCT findings, and presence or absence of ILD diagnosis. The 4-second clips from each of the 14 lung positions from each patient were compiled into a single movie for reader evaluation. Subsequently, two blinded readers (ultrasonographer RF and non-ultrasonographer DY) independently evaluated each of the 20 lung examinations for the presence or absence of ILD. Prior to reading the LUS examinations, the non-ultrasonographer underwent a 15-minute instructional session on the interpretation criteria using representative LUS images of patients external to the assessment cohort.

Statistical Analysis

The sensitivity and specificity for SSc-ILD detection on LUS was assessed for each reader using chest HRCT as the gold standard. Agreement between readers was assessed with percent concordance and Cohen’s Kappa statistic.

RESULTS

General clinical features

Twenty SSc patients (9 diffuse, 11 limited cutaneous disease) were enrolled. Three SSc patients were new to our rheumatology clinic and received new diagnoses of SSc around the time of enrollment. The indications for chest HRCT were: ILD screening in newly diagnosed SSc (n=3), clinical symptoms/suspicion (n=11), worsening PFTs (n=1), follow-up on known ILD (n=4), and pulmonary nodule surveillance (n=1). No patients in this cohort previously underwent lung transplantation.

278 of 280 possible lung zones were evaluated by LUS, with 2 anterior lung windows not assessed due to breast implants in one patient and an enlarged cardiac footprint in the second.

ILD on HRCT was observed in 9 SSc patients (45%, 8 NSIP, 1 UIP). Of these patients, the median time since ILD diagnosis was 1.0 years (IQR 0-2.5 years); four of these patients were newly diagnosed with ILD at the time of reference HRCT. Of patients with ILD, PFTs showed a mean forced vital capacity (FVC% predicted) of 59.1±29.1 and diffusion capacity for carbon monoxide (DLCO% predicted) of 75.6±23.4. SSc patients without ILD showed an FVC% predicted of 96.5±16.4 and DLCO% of 80.8± 5.6. Additional patient characteristics are highlighted in Table S1 in the supplementary material. All LUS examinations were well tolerated by all participants and completed in under 15 minutes.

Sensitivity and specificity of US-detected interstitial lung disease

Both readers independently read the 20 lung examinations identically, finding 11 (55%) positive and 9 (45%) negative for SSc-ILD with a sensitivity and specificity for ILD detection compared to HRCT of 100% and 82%, respectively, with perfect agreement between readers (κ=1.00) (Table 1). When comparing reader evaluations of the 278 individual lung zones, 93.8% concordance was observed with near perfect agreement (κ=0.82) (Table 1). Both LUS readers identified the same two false positives on LUS. In the first case, HRCT was indicated for dyspnea and revealed consolidative nodular opacities in a patient with suspected pulmonary infection. In the second, HRCT was indicated to screen for ILD in a patient with newly diagnosed SSc, and showed bibasilar atelectasis and possible scarring at the lung bases, but no evidence of ILD. Selecting milder disease from our cohort, HRCT showed a combination of fine reticulations and mild ground glass opacities (GGOs) in 3 patients (study HRCT to LUS time of 5, 20, and 28 days), and mild GGOs alone in one patient (LUS performed same day as study HRCT); all four of these patients were positive by our LUS interpretation criteria. Individual patient ILD characteristics, imaging timing, and imaging findings are highlighted in Table S2 in the supplementary material.

Table 1.

Detection of SSc-ILD via LUS versus HRCT using LUS interpretation criteria

	Reader 1 (ultrasonographer)	Reader 2 (non-ultrasonographer)
Patients, n	20
Lung zones examined, n	278
SSc-ILD on HRCT, n (%)	9 (45)
LUS-ILD detection per patient
Read as positive, n (%)	11 (55)	11 (55)
Read as negative, n (%)	9 (45)	9 (45)
Sensitivity vs HRCT	100%	100%
Specificity vs HRCT	82%	82%
Kappa	1.00
LUS–ILD detection per lung zones
Read as positive, n (%)	64 (23)	59 (21)
Read as negative, n (%)	214 (77)	219 (79)
Concordance	93.8%
Kappa	0.82

HRCT = high resolution computed tomography, LUS = lung ultrasound, ILD = interstitial lung disease, SSc-ILD = systemic sclerosis associated interstitial lung disease.

DISCUSSION

We developed a novel set of LUS interpretation criteria and LUS technique for the detection of SSc-ILD and performed a preliminary prospective validation in an SSc cohort showing 100% sensitivity, high specificity, and perfect agreement between two readers. To our knowledge, this is the first set of reading criteria developed for LUS detection of ILD. We believe this technique and reading criteria have the potential to replace HRCT as the initial screen for ILD in newly diagnosed SSc patients, or to screen for ILD in prevalent SSc cases when suggestive signs or symptoms arise.

In the course of developing the LUS technique highlighted here, we noted a multiplicity of LUS procedures reported in the literature. These technical descriptions often lack enough granularity and specificity to be reproduced by musculoskeletal ultrasonographers without specific training in LUS. Our goal was to provide a clear and concise LUS technique that could easily be performed with minimal training. We made several simple modifications to previously reported LUS techniques that improved pleural and subpleural interrogation. Most prior studies perform LUS with the probe oriented in a transverse orientation with lung movement that is orthogonal to the long axis of the probe for most lung positions.^9-11,13 This orientation and relative lung movement hampers visual tracking of pleural or subpleural lesions and is analogous to fixing one’s eyes on a point of page of text and moving a page upwards rather than from side to side; only the latter method affords the reader context. The prior preference for this orientation likely stemmed from the increased viewable lung surface afforded when the probe is oriented parallel to the ribs. Additionally, less specific artifactual B-lines, heavily investigated in early LUS work on ILD, do not require spatial resolution or pleural surface tracking for evaluation. We found that by spanning the ribs in a sagittal or pseudo-sagittal orientation, sufficient pleural surface evaluation at each lung position is maintained while gaining the ability to track pleural lesions over time, leading to greater specificity for lesions reflective of ILD versus LUS artifacts. As an extension of this concept, still images are insufficient for lung surface evaluation and we chose 4 second LUS movie clips at each of the lung positions to view movement of the lung surface through a portion of the respiratory cycle.

The LUS interpretation criteria presented in this work center on abnormalities observed at the pleural surface including a combination of discontinuities, cavitations, and/or granularity. In addition, these features must persist and track with lung motion, reducing false positive readings due to noise and artifacts that can occasionally be observed in patients without ILD. To reduce false positives that can occur from occasional pleural irregularity encountered in normal lungs, some degree of pleural thickening compared to normal is also required for a positive study. Our criteria were sufficient to discriminate SSc-ILD without requiring a pleural thickness cutoff, which may vary depending on machine and machine settings, and would add an additional level of complexity to LUS reading. Importantly, our LUS interpretation criteria demonstrated sensitivity even in patients with mild ILD on HRCT scan (supplementary Table S2).

Most early studies in LUS for ILD focused primarily on B-line quantification to detect underlying ILD. We found artifactually based B-lines difficult to quantify, variably present, and potentially less specific for ILD, in line with several prior publications.^7,8,13,14 The momentum behind B-lines in early LUS ILD work was likely due to less advanced US machines used at the time of these investigations. These earlier US machines have less sophisticated processing algorithms, lower frequency probes (often 5MHz or less), and poorer resolution of the pleural surface overall. Modern US equipment used by the typical rheumatology practitioner is capable of easily visualizing the pleural surface due to improved processing algorithms and higher frequency probes that improve resolution and reduce noise. These US advancements also have the effect of reducing or eliminating artifactual B-lines in certain cases. Indeed, B-lines can be made more conspicuous by turning off advanced processing settings, such as compound imaging.¹⁵ Additionally, there is significant machine-to-machine variability in B-line detection and adjustments to the number and location of focal zones, time gain compensation, and the dynamic range, are all known to affect B-lines.^13,15 Given the current advancements in US machines and variability in B-line visualization, we chose to avoid including B-lines in our criteria.

Our overarching goal in this work was to develop a simple LUS technique and interpretation criteria for SSc-ILD detection for research and clinical practice. To highlight the simplicity of our LUS interpretation criteria we included a second blinded reader with no prior US experience. The perfect agreement between readers in this study highlights the simplicity and effectiveness of our criteria. The interpretation criteria require only 1 of the 14 lung positions to be positive for an overall positive LUS study, giving preference to sensitivity rather than specificity, as needed in an effective screening tool. When evaluated by each lung zone, disagreements between readers were always in the setting of multiple positive lung zones, never resulting in disagreement in the overall study interpretation. The two false positive results in this study were encountered in patients with comorbid lung infection, atelectasis and lung scarring on HRCT. These findings suggest that LUS may not be the ideal screening modality for ILD in patients with known non-ILD comorbid lung disease in SSc.

There are several limitations in this study. The technique and criteria outlined here were empirically derived from experience and literature review and were not the result of a consensus study or regression analysis. However, in review of associated HRCT imaging and the LUS-ILD literature, these criteria have both face and construct validity and were simple and effective enough for SSc-ILD discrimination by even a non-ultrasonographer. The inclusion of HRCT within 3 months of LUS could allow for changes in lung disease between studies. Upon review of each enrollee (supplementary Table S2) LUS was performed in less than 30 days from HRCT in all but one of the HRCT positive patients who had fibrotic NSIP. Of the HRCT negative patients, 2 had LUS within 5 days, 7 had follow-up HRCT after the study which remained negative, 1 had no follow-up HRCT, but normal follow-up PFTs and was asymptomatic, and 1 patient developed new mild NSIP on HRCT 1.5 years later. Overall, these details suggest LUS findings to be reflective of HRCT findings within this study. We did not assess the ability of LUS to differentiate the underlying pattern of ILD, and our study was not designed to specifically evaluate the performance of this tool to screen for ILD in the setting of comorbid lung disease. Additional limitations include being a single center study using a single LUS proceduralist, lack of evaluation in other CTDs, a relatively small validation cohort, and a possible selection bias stemming from the inclusion of patients undergoing HRCT in the course of their clinical care.

In conclusion, we present a novel LUS technique and interpretation criteria for SSc-ILD detection, affording perfect agreement between ultrasonographer and non-ultrasonographer readers in an SSc cohort. Further validation in a larger SSc cohort and in other CTDs associated with ILD including polymyositis and dermatomyositis is ongoing.

KEY MESSAGES:

• This study outlines a novel lung ultrasound technique and interpretation criteria for systemic sclerosis associated interstitial lung disease.

• In a systemic sclerosis cohort, our lung ultrasound technique and interpretation criteria showed high sensitivity and specificity for interstitial lung disease with perfect agreement between blinded readers.