Analysis of systemic lupus erythematosus-related interstitial pneumonia: a retrospective multicentre study

Thoracic diseases in patients with systemic lupus erythematosus (SLE), especially interstitial pneumonia (SLE-IP), are rare and have been poorly studied. The aims of this multicentre study were to evaluate SLE-IP and elucidate its clinical characteristics and prognosis. Fifty-five patients with SLE-IP who had attended the respiratory departments of participating hospitals were retrospectively evaluated in this multicentre study. Clinical information, high-resolution computed tomography (HRCT), and surgical lung biopsy/autopsy specimens were analysed by respiratory physicians, pulmonary radiologists, and pulmonary pathologists. IP patterns on HRCT and lung specimens were classified based on the international classification statement/guideline for idiopathic interstitial pneumonias. The most frequent form of SLE-IP at diagnosis was chronic IP (63.6%), followed by subacute (20.0%), and acute IP (12.7%). Radiologically, the most common HRCT pattern was “Unclassifiable” (54%). Histologically, “Unclassifiable” was the most frequently found (41.7%) among 12 patients with histologically proven IP. Interestingly, accompanying airway diseases were present in nine of these patients (75%). In multivariate analysis, current smoking (hazard ratio [HR] 6.105, p = 0.027), thrombocytopenia (HR 7.676, p = 0.010), anti-double-strand DNA titre (HR 0.956, p = 0.027), and nonspecific interstitial pneumonia (NSIP) + organizing pneumonia (OP) pattern on HRCT (vs. NSIP, HR 0.089, p = 0.023) were significant prognostic factors. In conclusion, chronic IP was the most frequent form of IP in patients with SLE-IP, and “Unclassifiable” was the commonest pattern radiologically and histologically.

Onset forms of IP and frequency of SLE-related thoracic diseases other than IP. The forms of IP at onset in the 55 patients with SLE-IP are shown in Fig. 1A. The chronic form was the commonest (35 patients, 63.6%), followed by subacute IP (11 patients, 20%) and acute IP (seven patients, 12.7%). The most frequent thoracic disease other than IP (Fig. 1B) was pleuritis (six patients, 10.9%), followed by pulmonary hypertension (five, 9.1%), pericarditis (three, 5.5%), and pulmonary thromboembolism (two, 3.6%). Serositis, including pleuritis and pericarditis, was diagnosed in 16.4% of the patients with SLE-IP.

Relationships between form of IP at onset, serositis and activity of SLE. Activity of SLE was
assessed and relationships between disease activity and form of IP at onset examined. SLEDAI-2K scores were significantly higher in patients with acute/subacute IP than in those with chronic IP (Supplementary Fig. S2A, p = 0.046). In addition, patients with SLE-IP and pleural/pericardial effusion had significantly higher SLEDAI-2K scores than those without such effusions ( Supplementary Fig. S2B, p = 0.039).
www.nature.com/scientificreports www.nature.com/scientificreports/ Figure 2. Frequency of interstitial pneumonia (IP) patterns on high-resolution computed tomography (HRCT) and prognosis. IP patterns were re-evaluated in 55 patients with systemic lupus erythematosus (SLE)-IP. IP patterns on HRCT were classified according to the international classification statement/guideline for idiopathic interstitial pneumonia and idiopathic pulmonary fibrosis. (A) The most frequent IP pattern was "Unclassifiable" (30 patients, 54%). Of the patients with "Unclassifiable" SLE-IP, 25% had a nonspecific interstitial pneumonia (NSIP) + organizing pneumonia (OP) pattern, 12 (22%) an OP pattern, seven (13%) an NSIP pattern, and five an usual interstitial pneumonia (UIP) pattern (two definite and three possible UIP pattern; 9%). www.nature.com/scientificreports www.nature.com/scientificreports/ pattern had significantly better prognoses than those with NSIP (log-rank test, p = 0.042). UIP pattern was not associated with a worse prognosis than other IP patterns, unlike IPF/UIP 10 or rheumatoid arthritis-related UIP 13 . Even in patients with SLE but without other CTDs, those with NSIP + OP pattern still had significantly better prognoses than those with NSIP ( Fig. 2I, log-rank test, p = 0.021). The extent scores of lung fibrosis on HRCT were not significantly different between NSIP + OP and NSIP (p = 0.548).

Discussion
In the present study, we retrospectively studied data of patients with SLE-IP who had attended respiratory departments with a particular focus on radiologic and histopathologic patterns in these patients. The most frequent form of SLE-IP at onset was chronic IP (63.6%). Further, according to IIPs/IPF classification statement/guidelines, "Unclassifiable" was the commonest pattern on both HRCT and SLB/autopsy specimens, and NSIP + OP pattern (i.e., included in "Unclassifiable") on HRCT was associated with a better prognosis than NSIP pattern. To our knowledge, the present study includes the largest series of patients with SLE-IP thus far published and is the first in which pulmonary physicians, radiologists, and pathologists have precisely evaluated SLE-IP.
Previous studies have reported that IP is less common in patients with SLE than in those with other CTDs, comprising 4-10% of patients with SLE and being diagnosed mainly on chest radiographs 4,5 . In addition, chronic IP is reportedly found in 3-13% of patients with SLE 3 . The most frequently reported intrathoracic disorder in patients with SLE is pleuritis, which occurs in 16-60% of such patients 2,3 . However, many of the above-cited studies were conducted on patients who were attending rheumatology departments. In contrast, in the present study, we found that the commonest thoracic disease in 60 patients with SLE and thoracic diseases who visited or were referred to respiratory departments was IP (91.7%), pleuritis having been identified in only 18.3%. Furthermore, (F) NSIP + OP pattern (i.e., included in "Unclassifiable") showing both ground-glass and patchy air space consolidation. (G) DAD pattern showing extensive areas of ground-glass attenuation and mild reticulation with peribronchovascular predominance. Mild traction bronchiectasis is also suspected. Open arrowheads: reticular opacity, closed arrowheads: ground-glass opacity, and arrows: airspace consolidation. (H) Survival curves from the diagnosis of IP according to HRCT pattern are shown. Patients with NSIP + OP pattern had significantly better prognoses than those with NSIP (log-rank test, p = 0.042). UIP pattern did not have a worse prognosis than other IP patterns. (I) Even in patients with SLE but without other CTDs, those with NSIP + OP pattern still had significantly better prognosis than those with NSIP (log-rank test, p = 0.021).
www.nature.com/scientificreports www.nature.com/scientificreports/ chronic IP accounted for 63.3% of patients with SLE-IP. Thus, it seems that the characteristics of patients with SLE who visit rheumatology departments are quite different from those of patients who visit respiratory departments.
In the current study, we found unexpectedly high frequency of comorbid other CTDs (19 patients [34.5%]) in patients with SLE-IP. Several studies have suggested that SLE-Sjogren overlap syndrome phenotype may have contributed to increased risk of IP, especially in older patients with SLE 6 . Further, in the current study, the presence of comorbid other CTDs was found to be an independent prognostic factor. These results seem to be important and useful real-world information in a clinical practice in this rare disease.
As for the relationship between disease activity and prognosis, SLEDAI-2K scores of ≤4 14 or <3 15 are reportedly associated with better prognoses irrespective of the presence of thoracic diseases. However, in the present study, SLEDAI-2K scores were not associated with prognosis in patients with SLE-IP. Additionally, SLEDAI-2K scores were significantly higher in patients with acute/subacute IP or pleural/pericardial effusion. These clinical features may therefore be useful in predicting SLE activity in patients with SLE-IP.
In patients with IIPs, IP patterns are important for predicting prognosis and selecting therapies 9,10,16 . In the present study, the most frequent IP pattern in both HRCT and SLB/autopsy specimens was "Unclassifiable"; in contrast, a previous study indicated that the NSIP pattern on HRCT was frequent in patients with SLE 17 . This "Unclassifiable" in patients with SLE-IP may represent heterogeneity of lung inflammation and/or fibrosis; this being unlike other CTDs, which mainly have NSIP-predominant patterns 18 . Further, in the present study, patients with NSIP + OP on HRCT had better prognoses than those with NSIP alone regardless of lung fibrosis extent. This difference in prognosis may be associated with whether the predominant milieu in the lung is inflammatory or fibrotic. UIP pattern was not associated with a worse prognosis than other IP patterns, unlike rheumatoid arthritis-related UIP 13 or IPF/UIP 10 . The better prognosis of UIP in SLE than of IPF/UIP is consistent with previous reports on CTD-IP 19,20 .
Interestingly, in our study, mild airway diseases such as cellular bronchiolitis were found in most SLB/autopsy specimens from patients with SLE-IP (75%). Among patients with CTD-IP, those with rheumatoid arthritis or Sjogren syndrome reportedly frequently have airway diseases 1 ; however, few studies have documented histologically-proven airway diseases in patients with SLE-IP. Our observations suggest that involvement of small airways is a characteristic feature of SLE-IP.
Regarding prognostic factors, old age, male sex, renal damage, psychiatric involvement, and high disease activity are reportedly significant predictors of poor prognosis in patients with SLE 7,8,11,14,15,21 . Although many studies have not found lung involvement to be a significant prognostic factor, Haye Salinas et al. have reported that pleuropulmonary manifestations are predictors of significantly worse prognosis 11 . In our cohort, which comprised only patients with SLE and IP, current smoking, serum KL-6, and NSIP + OP pattern on HRCT (vs. NSIP pattern) were significant prognostic factors according to multivariate Cox proportional hazards analyses, as were comorbid other CTDs, thrombocytopenia, and anti-dsDNA antibody titre. However, the most frequent cause of death was infection (six patients) and respiratory failure caused by IP occurring in only two patients. The risk of death from infection is reportedly 4.98-fold higher in patients with SLE than in the general population 22 . In the present study, 51 patients (92.7%) received corticosteroids for their IP and 22 of them received additional immunosuppressants; thus, careful attention to immunosuppressive therapy should be paid in clinical practice.
The present study has several limitations. First, SLE-IP is rare; accordingly, our patient cohort is small. Second, the data were retrospectively collected. Third, the treatments for SLE-IP were not uniform; however, most patients had been treated with corticosteroids with or without immunosuppressants. Fourth, proportion of comorbid  www.nature.com/scientificreports www.nature.com/scientificreports/ other CTDs, mainly Sjogren syndrome, was relatively high (34.5%) and this condition may have influenced our observations. Fifth, we were unable to directly compare the features of SLE patients who were attending respiratory departments with those of patients who were attending rheumatology departments. The proportions of IP or other lung involvements may differ between these two settings. A larger and prospective study, including both respiratory and rheumatology departments, would be ideal for further evaluating SLE-IP.
In conclusion, in this multicentre study of patients with SLE-IP, we found that the most frequent form of IP was chronic IP (63.6%). Further, the most common pattern on HRCT and SLB/autopsy was "Unclassifiable" (54.0% and 41.7%, respectively). Additionally, histological examination revealed a high prevalence of accompanying www.nature.com/scientificreports www.nature.com/scientificreports/ mild airway disease in patients with SLE-IP. To more comprehensively evaluate this rare lung disease, larger and prospective studies across rheumatology and respiratory medicine departments are needed.

Methods
Study design and participants. In this multicentre study, respiratory physicians, pulmonary radiologists, and pulmonary pathologists retrospectively reviewed the data of 62 patients with SLE and thoracic diseases who had visited respiratory departments in nine hospitals in Japan between 1987 and 2016 ( Supplementary Fig. S1). Two patients were excluded from this study because their thoracic diseases were deemed to be attributable to infections. A further five without interstitial pneumonia were also excluded. Three of these patients had pleuritis, another pleuritis and pericarditis, and the fifth pulmonary hypertension. Therefore, 55 patients with SLE and IP were studied. All diagnoses of SLE had been made in accordance with the diagnostic criteria of the American College of Rheumatology (ACR) 1997 and/or Systemic Lupus International Collaborating Clinics (SLICC) 2012 in collaboration with specialists in other areas such as rheumatologists and dermatologists 23 . Nine of the participants had undergone SLB and three had been autopsied. Acute exacerbation of IP was diagnosed according to the diagnostic criteria of that in IPF 24   16103114]), and this study was carried out in accordance with the approved protocol. The need for patient approval and informed consent was waived due to the retrospective nature of the study.
Data collection and evaluation for disease activity of SLE. Clinical data, including symptoms, laboratory and pulmonary function tests, treatments, and period from the diagnosis of IP were obtained from the participant's medical records. These data were evaluated by 10 respiratory physicians. Chronic, subacute, and acute IP was defined as duration of ≧3 months, 3-1 months, and <1 month, respectively. These were periods from the onset of respiratory symptoms to the diagnosis of IP. Activity of SLE at the time of diagnosis of IP was evaluated using SLE-disease activity index 2000 (SLEDAI-2K) scores, which are derived from 23 items 25,26 . Review of chest HRCT and lung pathological specimens. Three pulmonary radiologists independently evaluated the HRCT features in 55 patients with SLE-IP and subsequently reached a consensus on diagnosis and IP pattern. The extent of lung fibrosis on HRCT was semi-quantitatively evaluated based on honeycombing and reticulation. The extent scores of lung fibrosis were as follows: score 0, none; 1, <25%; 2, 25-50%; 3, ≥50%. Scores of 0 and 1 were defined as low extent scores, and those of 2 and 3 as high extent scores. Where specimens obtained by SLB or autopsy were available, they were evaluated histologically. Four pulmonary pathologists evaluated histological features independently, and subsequently reached consensus diagnoses. IP patterns, lymphoid follicle with germinal centres, small airway disease, vasculopathy, and pleural lesions were identified and assessed. IP patterns on HRCT and lung specimens were classified based on the international classification statement/guideline for IIPs 9 and IPF 10 . The definition of "Unclassifiable" on HRCT and/or histological examination is as follows: (1) multiple HRCT and/or pathologic patterns; (2) new entity or unusual variant of recognized entity, not adequately characterized by the international IIPs classification statements 9,27 ; and (3) inadequate radiologic or pathologic data.