Lung ultrasound for assessing disease progression in UIP and NSIP: a comparative study with HRCT and PFT/DLCO

Abstract

Background

This study aims to compare Lung Ultrasound (LUS) findings with High-Resolution Computerized Tomography (HRCT) and Pulmonary Function Tests (PFTs) to detect the severity of lung involvement in patients with Usual Interstitial Pneumonia (UIP) and Non-Specific Interstitial Pneumonia (NSIP).

Methods

A cross-sectional study was conducted on 35 UIP and 30 NSIP patients at a referral hospital. All patients underwent LUS, HRCT, and PFT. LUS findings such as B-lines, pleural fragmentation, and pleural thickening were compared with HRCT-based lung involvement and PFT parameters.

Results

In UIP patients, B-lines > 18 and pleural fragmentation significantly differentiated between < 50% and > 50% HRCT involvement. A logistic regression model showed that B-lines > 18 (OR = 39, p = 0.04) and pleural fragmentation (OR = 22, p = 0.037) independently predicted > 50% HRCT involvement. ROC analysis of the model revealed 84.2% sensitivity and 84.5% specificity. Furthermore, the crude number of B-lines (OR = 1.2, p = 0.038) and > 50% HRCT involvement (OR = 9.5, p = 0.045) independently predicted severe DLCO impairment, with a sensitivity of 94.7% and specificity of 84.5%. Linear regression showed that each additional B-line was associated with a 0.4% decrease in DLCO (Beta = -0.377, p = 0.043), independent of patient diagnosis. In NSIP patients, no significant correlation was observed between LUS findings and > 50% HRCT involvement (p > 0.05), though B-line numbers and pleural thickening increased in cases with severe DLCO impairment (p < 0.05).

Conclusions

LUS shows promise as a sensitive, radiation-free alternative to HRCT in monitoring the severity of UIP. It is particularly valuable in predicting the extent of lung involvement and severe DLCO impairment in UIP patients but has limited application in NSIP.

Background

Interstitial lung diseases (ILDs) represent a diverse group of disorders characterized by inflammation and fibrosis of the lung parenchyma [1]. ILDs can be idiopathic or arise from known causes such as environmental exposures or connective tissue diseases. Early diagnosis is essential in determining the prognosis and guiding treatment strategies [2]. Diagnosis typically involves a multidisciplinary approach, combining clinical assessment, radiological findings, pulmonary function tests (PFTs), and histological analysis [1, 3].

Among the idiopathic forms of ILD, idiopathic pulmonary fibrosis (IPF) is one of the most common and severe, constituting about 20% of ILD cases. It is characterized by progressive fibrosis of the lung tissue, leading to chronic respiratory failure [4]. The hallmark radiological and pathological pattern of IPF is usual interstitial pneumonia (UIP). UIP is also observed in other settings, such as rheumatoid arthritis or hypersensitivity pneumonitis, but is most frequently associated with IPF [5, 6]. High-resolution computed tomography (HRCT) remains the gold standard for diagnosing UIP, although lung biopsy is often required for definitive confirmation in challenging cases [7].

Non-specific interstitial pneumonia (NSIP) is another common ILD pattern, frequently associated with connective tissue diseases. Compared to UIP, NSIP generally has a better prognosis [8]. Radiologically, NSIP is characterized by ground-glass opacities and subpleural sparing on HRCT, making it distinct from UIP [9, 10]. While HRCT remains a key diagnostic tool, the associated costs, limited availability, and radiation exposure pose challenges, particularly for long-term monitoring [8].

In recent years, lung ultrasound (LUS) has emerged as a valuable, non-invasive imaging modality for the assessment of ILDs [11]. LUS is particularly useful in identifying B-lines, which serve as sonographic markers of interstitial syndrome [1, 12]. B-lines are generated due to partial lung deaeration or interstitial space expansion resulting from fluid accumulation or collagen deposition [1]. The presence of diffuse bilateral B-lines is a hallmark of pulmonary interstitial syndrome, correlating with fibrotic changes observed on HRCT [1, 7]. Moreover, LUS can detect pleural irregularities, such as thickened or fragmented pleural lines, which are indicative of subpleural fibrosis [3, 7]. Pleural fragmentation on ultrasound in NSIP represent areas of discontinuity or irregularity in the pleural line, which on HRCT may correspond to ground-glass opacities and areas of subpleural sparing. Pathologically, these findings align with interstitial inflammation and mild fibrosis, which are characteristic of NSIP.

The ability of LUS to identify pleural line abnormalities, including pleural fragmentation, and detect subtle subpleural changes provides valuable information, particularly in the context of fibrotic ILDs like UIP [3]. The distinctive features of UIP, such as pleural thickening and subpleural fibrotic changes, are often detectable with LUS [13]. However, in NSIP, where ground-glass opacities and subpleural sparing are more common, LUS has shown limited visibility [13,14,15]. More research is required to assess whether LUS can be reliably used to monitor disease progression in NSIP [13,14,15].

The role of LUS in monitoring the progression of ILDs is still under investigation [16]. While HRCT and PFTs are more widely recognized modalities for this purpose [13, 16], LUS offers a cost-effective, radiation-free alternative that can be repeated frequently [1, 7, 17]. This study aims to explore the utility of LUS in monitoring disease severity in patients with UIP and NSIP, comparing its findings with HRCT and PFTs to evaluate its potential as a complementary tool for managing ILDs.

Methods

Study design and participants

This cross-sectional study was conducted at a referral respiratory hospital in 2023. The diagnosis of UIP and NSIP in this study was established through a multidisciplinary discussion, involving pulmonologists, radiologists, and pathologists, based on HRCT findings and, when required, biopsy results. Patients diagnosed through this process were subsequently enrolled in the study. A convenient sampling method was used, with 35 patients in the UIP group and 30 in the NSIP group. Only adults aged 18 years or older were included, and written informed consent was obtained from all participants. Exclusion criteria included other forms of ILD, significant comorbidities, and obstructive findings on PFT.

Data collection procedures

Demographic and clinical characteristics, including age, sex, smoking status, and past medical history (such as Non-Alcoholic Steatohepatitis [NASH], Diabetes Mellitus [DM], Hypertension [HTN], Cardiovascular Disease [CVD], Hypothyroidism, Hyperthyroidism, Sleep Apnea, Gastroesophageal Reflux Disease [GERD], and connective tissue diseases [CTD]), along with physical examination findings, were collected from medical records. Clinical symptoms such as dry cough, productive cough, sputum production, dyspnea, wheezing, chest pain, rales or crackles, and finger clubbing were also documented.

Diagnostic procedures

HRCT scans of patients with both UIP and NSIP were classified based on lung involvement as either less than 50% or more than 50%, confirmed by two independent radiologists. Initial ultrasound examinations were performed by a skilled radiologist, with further evaluations conducted by both intra- and inter-observers to minimize operator-dependent variations. A Philips InnoSight Compact ultrasound device with a curved probe (1–5 MHz) was used for these assessments. Ultrasound evaluations were performed on both lungs at three sites: Posterior Superior (PS), Mid-Posterior (MP), and Posterior Inferior (PI). Patients were positioned in a seated position during the lung ultrasound, with the right hand placed on the left shoulder and the left hand placed on the right shoulder [18].

Pleural thickness (> 3 mm [19]), pleural surface characteristics (regular, irregular, or interrupted), and pleural fragmentation (more than two sites in a single frame) were assessed. The total number of regions with pleural fragmentation, the number of B-lines, and the regions with more than three B-lines were recorded. These findings were then compared with HRCT and PFT results [20,21,22].

PFTs were conducted using body plethysmography, measuring parameters such as Forced Vital Capacity (FVC), Forced Expiratory Volume in One Second (FEV1), FEV1/FVC ratio, and Total Lung Capacity (TLC). DLCO values were also obtained. Disease severity was determined using the lung function test, following the 2005 guidelines of the American Thoracic Society/European Respiratory Society [23].

Statistical analysis

Data were analyzed using SPSS v16. Continuous variables were described as mean ± standard deviation (SD) or median with interquartile range (IQR). Differences between groups were analyzed using the t-test or Mann-Whitney U test for continuous variables, and the chi-square or Fisher’s exact test for categorical variables. Logistic regression models were constructed using variables with a p-value < 0.2. The Receiver Operating Characteristic (ROC) curve was used to assess test sensitivity and specificity, with a significance level of 0.05. Odds ratios (OR) with 95% confidence intervals (CI) were calculated.

Ethical cosnsiderations

All participants provided written informed consent prior to enrollment, and data confidentiality was strictly maintained. This study was registered with the Medical Ethics Board of Shahid Beheshti University of Medical Sciences under the code IR. SBMU. NRITLD.REC.1401.093.

Results

Baseline characteristics

The mean age of participants was 58.3 ± 9.5 years. UIP patients had a mean age of 60.5 ± 8.1 years, while NSIP patients had a mean age of 55.7 ± 10.5 years (p = 0.046). Males included the majority of the study sample, accounting for 61.5% (n = 40) of participants. Among those diagnosed with UIP and NSIP, the frequency of males was 80.0% (n = 28) and 40.0% (n = 12), respectively (p = 0.001). Approximately half of the patients (50.8%) reported a previous medical history (Table S1).

The pulmonary US findings of these patients revealed that there is a significant difference in UIP and NSIP patients between the number of B-lines in the right side (12.5 ± 3.9 vs. 10.7 ± 4.25, p = 0.036), and in total (24.3 ± 6.4 vs. 20.3 ± 8.5, p = 0.031) when adding the number of B-lines in the superior, middle and inferior LUS examination in the patients with UIP and NSIP. However, in the left side (11.8 ± 2.9 vs. 9.6 ± 4.7, p = 0.078), no significant difference was found. Those with the UIP were more likely to have more than 18 B-lines in total than those with NSIP (80.0% vs. 56.7%, p = 0.042) (Table 1).

Table 1 Pleural fragmentation, Thickening, and B-lines in UIP and NSIP patients by Lung Zone

The median number of B-lines in the upper, middle and lower US examination of the right side of the UIP patients was 3 [IQR 2,4], 4 [IQR 3, 5] and 5 [IQR 3, 6] respectively, which was not significantly higher than the NSIP patients with the number of 3 [IQR 1.5, 4] (p = 0.115), 4 [IQR 3, 5] (p = 0.285) and 5 [IQR 4, 7] (p = 0.207). The same was not applicable on the left side with the exception of the upper examination which was not significant (upper: 3 [IQR 2, 4] vs. 3 [IQR 1.75, 4] p = 0.792, middle: 4 [IQR 3, 5] vs. 3 [IQR 2, 4] p = 0.023, lower: 5 [IQR 4, 6] vs. 4 [IQR 2, 5.25] p = 0.014).

Univariate analysis

HRCT findings

The sub-grouped patients were analyzed separately based on the extent of lung involvement. In UIP patients, the crude number of B-lines, B-lines > 18, and pleural fragmentation were significantly different between those with less than 50% involvement and those with more than 50% ([20.9 ± 6.0 vs. 27.2 ± 5.3, p = 0.003], [62.5% vs. 94.7%, p = 0.018], and [12.5% vs. 68.4%, p = 0.001], respectively). A considerable correlation was found between the total number of regions with three or more B-lines and HRCT involvement (correlation coefficient = 0.630, p < 0.001). In the NSIP subgroup, no significant results were observed (Table 2).

Table 2 Demographics, medical history, and Ultrasound findings in UIP and NSIP by HRCT involvement extent

PFT / DLCO findings

PFT findings are summarized in Table S2. In UIP patients, pleural fragmentations were more likely to occur in severe cases based on FEV1 (76.9% vs. 22.7%, p = 0.002), and a similar pattern was observed for pleural thickening (84.6% vs. 36.4%, p = 0.006). TLC did not appear to be affected by any of the studied variables in UIP. Almost all cases with severe impairments in FEV1 had > 50% HRCT lung involvement (92.3% vs. 31.8%, p = 0.001), and similar observations were made for FVC (Table S2).

In NSIP patients, TLC was more likely to be impaired when B-lines exceeded 18 or when there was > 50% involvement on HRCT (p < 0.05). Increased pleural thickening was observed in patients with severely impaired FEV1 (17.4% vs. 71.4%, p = 0.014). The crude number of B-lines, B-lines > 18 and increased pleural thickening were increased in NSIP patients with severe FVC impairments (P < 0.05). The median number of total fragmentation regions in NSIP patients with mild or moderate DLCO impairment was 0 [IQR 0, 4.2], while in patients with severe DLCO impairment, it was 2 [IQR 0.25, 2.75] (Table S2).

The frequency of severe DLCO impairment was 47.7% (n = 31), compared to 52.3% (n = 34) with mild or moderate impairments. The crude number of B-lines, B-lines > 18, and the extent of lung involvement on HRCT > 50% correlated with the severity of DLCO impairment in UIP patients ([20.5 ± 5.5 vs. 27.5 ± 5.3, p = 0.001], [62.5% vs. 94.7%, p = 0.018], and [25% vs. 78.9%, p = 0.001], respectively). In NSIP patients, the crude number of B-lines was higher in those with severe DLCO impairment (24.0 ± 7.3 vs. 17.8 ± 8.5, p = 0.043), along with increased pleural thickening (58.3% vs. 11.1%, p = 0.013) (Table S2). The total number of regions with fragmentation was higher in NSIP patients with severe DLCO impairment (p = 0.038).

The DLCO, as a continuous variable, showed significant correlations with the patients’ diagnosis (UIP or NSIP) (CC = 0.295, p = 0.046), the crude number of B-lines (CC = -0.539, p < 0.001), > 50% HRCT involvement (CC = -0.378, p = 0.01), pleural fragmentation at more than two sites (CC = -0.548, p < 0.001), and the presence of pleural thickening (CC = -0.484, p = 0.001) within a single field of view.

Multivariable analysis

HRCT (model 1)

A predictive model was developed to evaluate the extent of lung involvement in UIP patients based on US findings. HRCT was included in the model as the dependent variable, while B-lines > 18, pleural fragmentation at more than two sites, increased pleural thickness, past medical history, sex, and age were incorporated into a logistic regression analysis. The results demonstrated that B-lines > 18 (OR = 39, p = 0.04) and pleural fragmentation (OR = 22, p = 0.037) were independent predictors of > 50% lung involvement on HRCT. None of the other variables showed statistical significance (p > 0.05). No suitable predictive model could be established for NSIP patients in this context (Fig. 1).

Fig. 1

ROC Curve for HRCT Involvement in UIP Patients with Pleural fragmentation and B-lines [Model 1, dependent variable: HRCT, independent variables: B-line numbers > 18, pleural fragmentation in more than 2 sites, increased pleural thickness and past medical history, sex and age were entered a logistic regression model.]

DLCO (model 2)

Another logistic model was developed to predict DLCO in UIP patients, with DLCO as the dependent variable. Independent variables included sex, age, extent of involvement on HRCT, crude number of B-lines, and pleural fragmentations. In a linear regression model treating DLCO as a continuous variable, the only significant predictor was the crude number of B-lines (Beta = -0.377, p = 0.043). The model was adjusted for age (Beta = 0.125, p = 0.351), sex (Beta = -0.101, p = 0.472), patient diagnosis (UIP or NSIP) (Beta = 0.081, p = 0.589), > 50% HRCT involvement (Beta = -0.162, p = 0.253), pleural fragmentations at more than two sites (Beta = -0.117, p = 0.455), and the presence of pleural thickening in a single field of view (Beta = -0.150, p = 0.294). The model’s adjusted R² was 0.352, with a constant of 75.2 (p = 0.027). These findings indicate that with every unit increase in the number of B-lines, DLCO decreases by 0.4%, independent of the patient’s diagnosis (UIP or NSIP). Due to high collinearity with the extent of involvement on HRCT, pleural fragmentations were excluded from the model. The crude number of B-lines (OR = 1.2, p = 0.038) and the extent of lung involvement on HRCT (OR = 9.5, p = 0.045) were identified as independent predictors of DLCO. Age and sex were not significant predictors (p > 0.05) (Fig. 2). A similar model was constructed for NSIP patients to predict DLCO, but it only yielded results consistent with the univariate analysis (Fig. 3).

Fig. 2

ROC Curve for Severe DLCO Impairment in UIP Patients with Pleural fragmentation and B-lines [Model 2, dependent variable: DLCO, independent variable: Sex, age, the extent of involvement in HRCT, the crude number of B-lines and the pleural fragmentation s were entered a logistic regression model.]

Fig. 3

ROC Curve for Severe DLCO Impairment in NSIP Patients with Pleural fragmentation and B-lines

Diagnostic values and ROC curve analysis

UIP

The primary outcome of this study was to assess the diagnostic accuracy of LUS in predicting HRCT and DLCO values. In UIP, the sensitivity and specificity for detecting > 50% lung involvement on HRCT with B-lines > 18 were 94.73% and 37.5%, respectively. Pleural fragmentations demonstrated a sensitivity of 68.42% and a specificity of 87.5% for detecting > 50% HRCT involvement (Table 3).

Table 3 ROC Curve Analysis for Model Predictions and total B-line numbers in UIP and NSIP

A ROC curve was generated using the predicted probabilities from Model 1, with > 50% lung involvement on HRCT as the outcome variable. The optimal cut-off point was 0.39, with a sensitivity of 84.2% and specificity of 84.5% (Fig. 1). The AUC, p-value, and 95% confidence intervals are presented in Table 3. A similar approach was used to assess severe DLCO impairment as the outcome variable (Fig. 2), and the crude number of B-lines was evaluated for its diagnostic value (Table 3; Fig. 2).

Discussion

This study highlights the important role of LUS in assessing disease severity in UIP and NSIP. Current guidelines recommend a multidisciplinary approach considering clinical, functional, and HRCT findings for the diagnosis and management of ILDs [24, 25]. Although HRCT is considered the gold standard, it provides a definitive UIP diagnosis in only 55% of cases [7, 26]. Moreover, routine follow-up HRCT scans are not recommended for stable patients due to radiation risks and a lack of consensus on their use [27].

LUS has emerged as a promising complementary tool, showing significant correlations with the extent and severity of fibrosis on HRCT [28]. This study revealed that B-lines > 18 and pleural fragmentations were significantly different between UIP patients with < 50% and > 50% lung involvement. Another study involving 31 IPF patients reported a strong correlation between the average number of B-lines and HRCT fibrotic scores but did not assess the diagnostic value of B-lines [7]. They identified a pleural line thickness of 2.4 mm as the optimal diagnostic threshold (sensitivity 0.958, specificity 0.994), but did not adjust for other variables. In contrast, the findings from this study suggest that the number of B-lines and the presence of pleural fragmentations are stronger predictors than pleural thickness.

This study’s HRCT model (Model 1) demonstrated that both B-lines > 18 and pleural fragmentations were independent predictors of > 50% lung involvement in UIP after adjustment. B-lines > 18 increased the odds by 39-fold, while pleural fragmentations at > 2 sites increased the odds by 22-fold. ROC analysis of the predicted probabilities from Model 1 showed superior sensitivity (84.2%) and specificity (84.5%) compared to B-lines > 18 alone (sensitivity 94.73%, specificity 37.5%) or pleural fragmentations alone (sensitivity 68.4%, specificity 87.5%). Using both LUS features is recommended to optimize sensitivity and specificity. This model accurately predicts the extent of lung involvement in UIP and provides valuable clinical insights.

A similar study found a correlation between median B-lines and HRCT fibrotic scores, consistent with the findings of this study. However, they did not develop a model to evaluate the independent predictive value of B-lines, an area where this study provides additional insight. While they suggested that pleural thickening predicts higher HRCT scores, the univariate analysis and HRCT model in this study did not support this claim. Pleural thickening appears to be an insufficient independent predictor of significant HRCT involvement in UIP patients [15].

Regarding PFTs, FVC% and DLCO% remain vital indicators of ILD progression and are strong predictors of mortality. Historical cohort studies have consistently shown that FVC% predicts survival and serves as a preferred endpoint in clinical trials [29, 30]. However, the correlation between LUS features and PFT in UIP is relatively unexplored. This study identified significant differences in FVC and LUS features (fragmentations > 2 sites, pleural thickening), as well as > 50% HRCT involvement, between patients with severe and mild/moderate UIP. This is consistent with findings from a large study that demonstrated FVC as a reliable and clinically significant measure of IPF status, with 2–6% declines in FVC being considered clinically important [29].

In the UIP subgroup analysis for DLCO, significant differences were observed in the crude number of B-lines and > 50% HRCT involvement between patients with mild/moderate and severe impairment. In a linear regression model, the crude number of B-lines was the only significant predictor of DLCO when treated as a continuous variable (Beta = -0.377, p = 0.043), indicating that each additional B-line corresponds to a 0.4% reduction in DLCO, independent of the patient’s diagnosis (UIP or NSIP). Furthermore, each additional B-line increased the odds of severe DLCO impairment by 1.2, while > 50% HRCT involvement raised the odds 9.5-fold. Previous studies have indicated that DLCO decline is predominantly dependent on the extent of lung involvement, positioning it as a more suitable marker of IPF progression [7]. This study extends these findings by quantifying the relationship between B-lines, HRCT involvement, and severe DLCO impairment.

ROC analysis of the DLCO model (Model 2) in this study demonstrated superior sensitivity (94.7%) and specificity (84.5%) in predicting severe impairment compared to using B-line count alone. This model holds valuable implications for intensive care patients who are unable to undergo DLCO testing, as well as for centers lacking PFT capabilities, where combining HRCT and LUS can serve as an alternative means to assess DLCO severity. Importantly, even a single excess B-line per posterior zone was associated with significant DLCO impairment, with good sensitivity and specificity, a finding further enhanced by concurrent HRCT use.

In NSIP, this study found no significant relationship between B-lines and > 50% HRCT involvement, in contrast to a study that pooled various ILDs [31]. However, the number of B-lines, fragmentation regions, and pleural thickening increased in cases of severe DLCO impairment, aligning with previous research showing inverse correlations between LUS findings and DLCO in ILDs [31,32,33]. Though, these studies did not evaluate their findings within the NSIP subgroup. FVC impairment was more likely in patients with B-lines > 18 or > 50% HRCT involvement. While most studies focus on correlations between B-lines and FVC, this study uniquely demonstrate that each additional B-line in posterior views correlates with FVC impairment and NSIP severity. TLC impairment was also more likely in patients with B-lines > 18 or > 50% HRCT involvement, consistent with findings in other ILD studies [31, 32].

The strengths of this study include the separate analysis of UIP and NSIP, which helps clarify the differences in LUS monitoring for ILD subtypes, and the involvement of an experienced radiologist, with intra- and inter-observer assessments, to minimize operator dependence. However, the small sample size in this study limits the generalizability of the findings and precludes the calculation of negative and positive predictive values. The high odds ratios in the HRCT model may be inflated due to this limitation. Larger cohorts are needed to validate and refine these models. Moreover, pleural thickness can be quantified, and it is recommended that future studies either measure this thickness or establish a cut-off for its assessment. This study’s cross-sectional design also limits its ability to track changes in ultrasound findings over time. Future research should adopt a longitudinal approach to track how LUS findings evolve. This would help clarify their implications for disease progression and management. A longitudinal design could also provide a more comprehensive understanding of the dynamic nature of interstitial lung diseases.

In summary, LUS shows significant potential as a sensitive, radiation-free alternative to HRCT for monitoring the progression of UIP. This study highlights the value of LUS in assessing lung involvement and disease progression in ILDs, particularly in patients with UIP. LUS proved to be both sensitive and specific in predicting the extent of lung involvement on HRCT and severe DLCO impairment. In UIP patients, the sensitivities and specificities were 84.2%, 84.5%, 94.7%, and 84.5%, respectively.

Conclusion

This study compared LUS findings with HRCT and PFTs in detecting the progression of lung involvement in UIP and NSIP. The results demonstrated that in UIP patients, B-lines > 18 and pleural fragmentations were independent predictors of > 50% HRCT involvement. Furthermore, the number of B-lines and > 50% HRCT involvement independently predicted severe DLCO impairment. However, no significant relationship was found between LUS findings and > 50% HRCT involvement in NSIP patients. These findings suggest that LUS holds potential as a sensitive, radiation-free alternative to HRCT for monitoring disease progression, although further research is necessary to validate these findings across larger cohorts and in NSIP populations.