Chest MRI with CT in the assessment of interstitial lung disease progression in patients with systemic sclerosis

Abstract

Objective

To describe the performance of CT and MRI in the assessment of the progression of interstitial lung disease (ILD) associated with SSc and demonstrate the correlations of MRI with pulmonary function test (PFT) and CT scores.

Methods

This prospective single-centre observational study included patients with SSc diagnoses, and magnetic resonance (MR) images were assessed visually using the Scleroderma Lung Study (SLS) I system. Differences in the median scores were assessed with Student’s t-test and the Wilcoxon rank-sum test. Pearson’s and Spearman’s rank correlation coefficients were calculated to correlate imaging scores and PFT results. Using disease progression as the gold standard, we calculated the area under the curve (AUC) of the CT and MRI scores with Harrel’s c-index. The best thresholds for the prediction of disease progression were determined by receiver operating characteristic curve analysis with maximum Youden’s Index (P < 0.05). The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the scores were calculated.

Results

The AUCs for MRI and CT scores were 0.86 (0.72–0.98; P = 0.04) and 0.83 (0.70–0.99; P = 0.05), respectively. CT and MRI scores correlated with Forced vital capacity (%FVC) (MRI: r = −0.54, P = 0.0045; CT: r = −0.44; P = 0.137) and diffusing capacity of the lung for carbon monoxide (MRI: r = −0.39, P = 0.007; CT r = −0.36, P = 0.006). The sensitivity, specificity, PPV and NPV were 85%, 87.5%, 88.34% and 86.11% (MRI score) and 84.21%, 82.35%, 84.14% and 82.4% (CT score), respectively.

Conclusions

MRI scores from patients with SSc may be an alternative modality for the assessment of ILD progression in patients with SSc.

Introduction

Lung involvement, including interstitial lung disease (ILD), is a leading cause of death among patients with SSc [1–3]. The initiation of treatment early in the course of SSc-ILD may lead to improved clinical outcomes and is a predictor of significant lung function improvement, regardless of the drug used [4]. For this reason, rigorous screening programmes to facilitate the early diagnosis and treatment of SSc-ILD are of paramount importance [5].

The diagnostic tools used most frequently for ILD diagnosis in patients with SSc are high-resolution CT (HRCT) and the pulmonary function test (PFT) [6]. Due to its high degree of sensitivity, HRCT can be used to identify mild or early interstitial abnormalities of unknown clinical significance, which should prompt heightened surveillance for signs of progression. MRI of the lung has been demonstrated to be a feasible means of determining the presence of activity in patients with lung diseases; however, further research is needed before it can be established as an alternative to CT and PFT for the diagnosis and management of ILD. MRI has recently been established as a valuable diagnostic modality for ILD [7]. Its ultrafast sequences and improved image quality, as well as the lack of ionizing radiation and versatile tissue characterization abilities, are among its advantages [8]. The primary aim was to describe the performance of CT and MRI in the assessment of the progression of ILD associated with SSc. The secondary aim was to demonstrate the correlations of MRI with PFT and CT scores.

Methods

Patients

This single-centre study conducted at Santa Casa de Misericórdia de Porto Alegre Hospital was approved by the institution's ethics committee (no. 1780557). Consecutive patients with SSc who were referred for chest HRCT between January 2015 and December 2019 were included. The inclusion criteria were age ≥18 years, confirmed diagnosis of SSc based on the 2013 American College of Rheumatology and EULAR criteria, and PFT and HRCT performed with standard parameters in the previous 1.5 months [6]. Patients with claustrophobia were excluded.

Informed consent was obtained from the participants after describing the benefits of the study, as well as the risks and all the researchers' responsibilities.

Imaging technique

MRI was performed with a Magnetom Aera device (Siemens Healthcare, Joinville, SC. Brazil) with an 18-channel anterior body coil and a 32-channel posterior spine coil. The patients were supine with their arms extended along the body and were moved into the device headfirst. Two datasets were acquired for this study. A T2-weighted turbo spin-echo sequence with fat suppression was obtained using the periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) technique, and more specifically the BLADE method, which is a PROPELLER-equivalent function of the Siemens Medical System. A T2-BLADE sequence was acquired during patients’ free-breathing, with a respiratory navigator placed on the right at diaphragm level at end-normal expiration, using the following parameters: repetition time/echo time = 2000/27 ms, flip angle = 150°, averages = 1, voxel size = 1.4 × 1.4 × 3 mm³, and acquisition time = 4–8 min. For patients unable to adequately hold their breath, we used nasal-cannula oxygen delivery, patient hyperventilation and fewer phase-encoding steps to reduce the sequence acquisition time.

HRCT was performed as part of diagnostic assessment using a 64-section scanner (Light Speed; GE Medical Systems, Milwaukee, WI, USA) during a single breath-hold with standard acquisition parameters: 200 mA with automated dose reduction, 120 kV, pitch = 1, 0.5 s rotation time, 0.625 mm collimation, 400 × 400 mm field of view and 512 × 512 acquisition matrix. Pulmonary CT images were reconstructed using a soft filter to yield contiguous 0.625 mm axial sections from the apex of the lung to the diaphragm. Thin-section (1.25 mm) CT images were reconstructed every 1 mm using a high-spatial-resolution filter.

Image post-processing

Two independent radiologists (B.H. and M.C.B.), each with 10 years of clinical experience, who were blinded to the CT and MRI findings reviewed the CT and magnetic resonance (MR) images using a standard workstation (GE Healthcare, São Paulo, SP). They also identified enlarged (short-axis transverse diameter >10 mm) mediastinal lymph nodes and pleural and pericardial effusions. Pulmonary fibrosis, bronchial disease and emphysema were evaluated using the CT images.

Scoring of CT findings

The severity of ILD was assessed visually and graded using the semi-quantitative Scleroderma Lung Study (SLS) I system [9]. According to this system, three zones of each lung were delineated using the aortic arch and pulmonary vein (upper, apex to the aortic arch; middle, aortic arch to the inferior pulmonary vein; lower, inferior pulmonary vein to diaphragm). For each zone, the severity of pure ground-glass opacity, lung fibrosis (including reticulation, bronchiectasis and bronchiectasis), honeycombing and emphysema were graded on a scale ranging from 0 to 4 (0 = absent, 1 = 1–25%, 2 = 26–50%, 3 = 51–75%, 4 = >75%). The two readers (B.H. and M.C.B.) performed this assessment for all HRCT slices from the lung apex to base, scoring each of the six thoracic zones (range, 0–16). Total CT scores ranged from 0 to 96 (Fig. 1).

Scoring of MRI findings

MR images were assessed visually using the SLS I system [9]. The extent of pulmonary T2 signal hyperintensity in the three lung zones was scored on a scale ranging from 0 to 4, as in the CT assessment [10, 11]. This is the MRI extent score. The lung/muscle ratio was calculated by dividing the mean higher lung lesion signal intensity [in a region of interest (ROI) of 600–700 mm³] by the mean paraspinal muscle signal intensity (in an ROI of 200–300 mm³). Final MRI scores (range, 0–39) were obtained by multiplying the MRI extent score and MRI muscle lung scores [12] (Fig. 1).

Definition of disease progression

An attending rheumatologist followed patients via regular outpatient clinic appointments. The scans (CT and MR) were evaluated at time 0 and spirometry and clinic evaluation at time 12 (12 months after). The progression of disease was defined as the worst of clinical symptoms and functional progression of disease [≥10% reduction of forced vital capacity (FVC) or ≥15% reduction of diffusing capacity of the lung for carbon monoxide (DLCO)]. According to a previous study, although a change of ≥10% %FVC predicted is clinically meaningful, exploratory analyses can be considered, as the lesser relative changes in FVC: marginal sustained reductions in FVC (5–10% as relative change) and changes in HRCT extent [13, 14]. Two rheumatologists, each with >10 years of experience, who were blinded to the lung MRI data, evaluated each case separately. Four patients did not perform control DLCO.

Statistical analysis

The statistical analyses were performed using SPSS software (v. 18; SPSS Inc., Chicago, IL, USA). The Kolmogorov–Smirnov test was used to determine whether variables were distributed normally. Continuous variables are expressed as means (s.d.) or medians with interquartile range. Discrete variables are expressed as frequencies with percentages. Differences in the median scores were assessed with Student’s t-test and the Wilcoxon rank-sum test, which was conducted with Bonferroni correction. The differences of averages were assessed with the paired t-test with Bonferroni correction. Pearson’s and Spearman's rank correlation coefficients were calculated to investigate correlations between the imaging scores and PFT results. Using disease progression as the gold standard, we calculated the area under the curve (AUC) of the CT and MRI scores with Harrel’s c-index. The best thresholds for the prediction of disease progression using the CT and MRI scores were determined by receiver operating characteristic curve analysis with maximum Youden’s Index. The results are expressed in terms of AUCs. AUCs were compared using the method of DeLong [15]. We used the intra-class correlation coefficient (ICC). The ICC can take a value from 0 to 1, with 0 indicating no agreement and 1 almost perfect agreement. P-values < 0.05 were considered significant. Cox’s proportional hazards model was used to describe the hazard ratio of CT and MRI score in relation to time to progression.

Results

The study sample comprised 36 patients [n = 32 (88.9%) women] with a mean age of 49 (8) years (Table 1). Patients were followed for 12 months; during this period 19 patients [19 women, age: 51 (6) years] were categorized as showing disease progression. Median, minimum and maximum CT scores were 5, 25 and 69, respectively. Median, minimum and maximum MRI scores were 3, 21 and 35, respectively. CT and MRI scores correlated with %FVC (MRI: r = −0.54, P = 0.0045; CT: r = −0.44, P = 0.137) and DLCO (MRI: r = −0.39, P = 0.007; CT r = −0.36, P = 0.006) (Table 2). CT and MRI scores differed significantly between patients with stable and worsening lung involvement (Fig. 2). Inter-reader agreement was substantial for CT and MRI scores [ICC = 0.66 (95% CI: 0.50, 0.82) and 0.65 (95% CI: 0.49, 0.73), respectively]. For multivariate analysis, the CT score and MRI score were assessed (Table 3). A Bonferroni-corrected P-value of <0.0083 (0.05/6) was considered significant. The AUCs for MRI and CT scores were 0.86 (0.72–0.98) and 0.83 (0.70–0.99), respectively (Fig. 3). There was no statistical difference between CT score and MRI score AUCs. Thresholds for optimal accuracy were 17 for the MRI score and 25 for the CT score. The sensitivity, specificity, positive predictive value and negative predictive value were 85%, 87.5%, 88.34% and 86.11%, respectively, for the MRI score and 84.21%, 82.35%, 84.14% and 82.4%, respectively, for the CT score.

Discussion

In this study, we demonstrated that MRI scores had similar to the CT scores in the prediction of SSc-ILD progression (according to functional test) and differed significantly between the stable disease and progression groups. In general, our findings suggest that MRI is a valuable modality for the evaluation of ILD in patients with SSc, with no radiation requirement.

Chest MRI can be useful for the diagnosis and assessment of thoracic pathologies due to technological advances such as the development of ultrafast sequences and improvement of acquisition speed and image quality [9, 11, 16, 17]. As it does not involve ionizing radiation exposure, MRI is an option for serial pulmonary parenchyma assessment. Moreover, immunomodulant treatment is indicated for active SSc-ILD, and the deleterious effects of radiation in immunocompromised patients have been well described [18, 19]. Recently, with the reduction in SSc-associated mortality related to renal crisis and lung fibrosis, experts and patients have become more concerned about long-term SSc-associated morbidities, especially possible associated malignancies, as the incidence of malignant tumours is elevated in patients with SSc [20].

The use of chest MRI for SSc-ILD diagnosis has been described [10, 11, 16–18]. Müller et al. [14] found that pulmonary MRI had 100% sensitivity, but only 60% specificity, in the assessment of patients with SSc relative to chest CT. Pinal-Fernandez et al. [16] reported that MRI could be used to detect ILD with high degrees of sensitivity and specificity in patients with >0.5% parenchymal involvement, but that it was less sensitive than HRCT and its use led to the underestimation of disease extent. The use of MRI features as biomarkers of SSc-ILD worsening also has been examined. Gargani et al. demonstrated that MRI can be used to detect SSc-ILD independently of HRCT appearance and that it may be used to predict worsening lung involvement [7]. The same study demonstrated a significant difference, in either STIR or T1 values, between normal and pathological ILD tissue areas, with HRCT as the gold standard and these results were supported by the significant correlations with other functional (%FVC, %TLC, %DLCO) and radiological (HRCT SLS I score) parameters reflecting SSc-ILD [7]. The present study confirmed these findings and highlighted the use of MRI in such patient settings. However, more studies are needed to evaluate SLS I sQCT score adaptation to ILD-SSc MR assessment.

The limitations of this study include the single-centre design and the inclusion of a limited number of patients. Further studies on larger patient samples are needed for a better MRI evaluation, especially in the assessment of lung fibrosis progression. The ILD assessment with the MRI is part of the diagnostic path in SSc patients of the authors’ centre because the patients did cardiac evaluation together. These same criteria were used in a previous study that compared lung MRI signal with HRCT and evaluated the role of MRI in predicting ILD progression [7]. Also, we used manufacturer-specific T2-weighted MRI sequences to assess ILD, limiting the comparability of the data with those obtained using other manufacturers’ devices. Another limitation in this study may be the non-use of the Goh sQCT score [21], which is an important clinical CT score to estimate ILD extent, [21]. In this study we did not find a statistically significant correlation between SLS1 CT score and %FVC and this may be due to the specific CT score. Also, our study demonstrated a non-statistically significant correlatio n of r = −0.44. Previous correlations of FVC of r = −0.50 [22] and r = −0.33 [23] have been described, but with a larger number of patients included. We believe that the absence of statistical correlation in part could be related with size of our sample.

The comparison between Goh sQCT and SLS I scores could be studied in the future. The Goh sQCT score is widely used and is a simple staging system for SSc-ILD as a limited or extensive disease, based on a simplified HRCT evaluation and FVC estimation, that provides more powerful prognostic information than either component in isolation [21]. And similar to the Goh sQCT score, the SLS I score also evaluates the severity of pure ground-glass opacity, lung fibrosis (including reticulation, bronchiectasis and bronchiectasis), honeycombing and emphysema. In our study, we also considered the baselines values of %FVC, %TLC, %DLCO. SLS I score was used to assess the severity and extent of SSc-ILD with a high degree of reliability in several previous studies [13, 24–31]. Similar methods were also used [31, 32]. Also, considering the limited DLCO data of this study, the OMERACT definition of ILD progression could not be used for definition of ILD progression in all cases. However, this study considered the other aspects of OMERACT to define ILD progression as the change of ⩾10% %FVC and marginal sustained reductions in FVC (5–10% as relative change) [13, 33]. Another limitation is adaptation of SLS I to MR data, but this assessment was previously performed with acceptable results [16]. Also, the difference of pixel size between MR and CT is a point that could have some bias, but as we are studying diffuse lung diseases in both methods the influence could be attenuated.

In conclusion, the quantitative MRI methods used in this study do not require radiation exposure and have the potential to predict SSc-ILD progress, which would improve patient stratification according to the likelihood of benefiting from immunomodulatory therapy. The MRI score has statistically significant correlations with PFTs.

Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this article.

Disclosure statement: The authors have declared no conflicts of interest.

Data availability statement

All data are incorporated into the article.

References