Developmental pathways of pulmonary abnormalities in rheumatoid arthritis according to sequential HRCT findings

Background To determine the patterns and development pathways of pulmonary abnormalities in patients with rheumatoid arthritis (RA). Methods We conducted a retrospective cohort study of consecutive RA patients who underwent high-resolution CT (HRCT) before and during biologic therapy. The presence of and change in 20 pulmonary lesions were recorded. Patterns of preexisting and new lesions were examined by cluster analysis and tested for a correlation between pre-existing and new lesions. abnormalities Strong correlation found between new ground-glass and each and

4 on the present findings, we propose the following development pathway for pulmonary abnormalities in RA: the initial lesion is bronchiolitis, which leads to the formation of complex AD, from which a variety of pulmonary lesions arise, particularly ILDs.

Background
Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by destructive polyarthritis. As well as joints, RA affects numerous organs, including the lungs. Pulmonary involvement contributes to the morbidity and mortality of RA patients (1)(2)(3) and is one of the major causes of death in RA (4). The presence of pulmonary abnormalities restricts the treatment of RA. Thus, the management of pulmonary involvement in RA and the treatment of RA in patients with pulmonary involvement are important.
Pulmonary involvement occurs in 60-80% of patients with RA (5,6) and can affect any component of the lung. It presents as various forms of pulmonary abnormality, including interstitial lung disease (ILD), airway disease (AD), nodules, pleural disease, and vascular disease (1,2). It is not uncommon for multiple abnormalities to be present in the same individual. To better manage these conditions, it is necessary to understand how pulmonary abnormalities develop; however, little is known regarding their developmental pathways. In the present study, we attempted to determine the developmental pathways by evaluating the co-existence patterns of pulmonary abnormalities and the relationship between pre-existing and newly emerging lesions in consecutive patients with RA.
Material and methods 5

Study design
We conducted a retrospective cohort study to determine the developmental pathways in RA by evaluating the existence patterns of pulmonary abnormalities and the relationship between pre-existing and newly emerging lesions in RA patients.

Patients
The participants were 410 consecutive Japanese patients with RA who first received Of the 410 patients, 208 who had undergone sequential chest HRCT during biologic therapy were admitted to the study. Excluded were patients with a pulmonary infection such as bacterial pneumonia, tuberculosis, non-tuberculous mycobacterial infection and pneumocystis pneumonia or drug-induced pulmonary injury on the sequential HRCT. The data of these patients were selected for further analysis. This study was approved by the local Ethics Committee Bioethics Committee Dokkyo Medical University (#2141).

Clinical data collection
We reviewed the subjects' medical records and collected the following data: patient characteristics, including age, sex and smoking history; respiratory symptoms and disease duration; serum levels of anti-citrullinated protein antibody, rheumatoid factor (RF), and CRP,; disease activity as assessed by the Disease Activity Score-28/Erythrocyte sedimentation rate (DAS28-ESR) (9); medication for RA, including 6 glucocorticoids, conventional disease-modifying antirheumatic drugs such as MTX and biologics; and reasons for the sequential HRCT examination.

Evaluation of chest HRCT images
All subjects received HRCT of the chest before and during biologic therapy. HRCT images were evaluated independently by a rheumatologist (AT) with 10 years of experience who had also received training as a pulmonologist, and a pulmonologist (SS) with 11 years of experience.
The readers were blinded to all clinical information. They evaluated the existence and distribution of lung abnormalities on HRCT; abnormalities detected at the first HRCT were evaluated for any change at each subsequent scan. If the assessment differed between the two examiners, the final judgment was made by a pulmonologist (MA) with 30 years of experience, referring to reports by pulmonary radiologists.
Change in a pulmonary lesion was defined as follows: 1) when a lesion not seen in any lung fields on the first HRCT was detected on the sequential scan, the lesion was judged as new; 2) when a lesion found on the first HRCT showed expansion within the same lung field on the sequential scan, or was newly detected in another lung field, the lesion was judged as worsened, when a lesion worsened in one lung field and was unchanged or improved in other fields, the lesion was judged as worsened.; 3) when a lesion seen on the first HRCT scan decreased in size, with no new or worsening lesions in any other lung fields, the lesion was judged as improved, when a lesion improved in one lung field and was unchanged in other fields, the lesion was judged as improved.; and 4) when a lesion seen on the HRCT scan did not change in size or distribution in all lung fields, the abnormality was judged as unchanged.
A change in category and pulmonary abnormalities was similarly defined as follows: 1) when a category/abnormality had a new lesion, the category/abnormality was judged as having a new lesion; 2) when a category/abnormality had a worsened lesion and no new ones, the category was judged as worsened; 3) when a category/ abnormality had an improved lesion and no new or worsened ones, the category was judged as improved; and 4) when a category/abnormality had no new, emerging, worsened, or improved lesions, the category/abnormality was judged as unchanged.

Statistical analysis
All analyses were performed using JMP11 software (SAS Institute Japan, Tokyo, Japan). Normally distributed continuous data were analyzed using 2-sample t-test or one-way analysis of variance and were presented as the mean ± standard deviation (SD). Non-normally distributed continuous data were analyzed using Wilcoxon/Kruskal-Wallis test. Categorical data were analyzed using chi-square test or Fisher's exact test. For multiple comparisons, Bonferroni correction was applied.
Statistical significance was defined as p < 0.05.

8
Cluster analysis was conducted according to the existence of the lesions by Ward's method. The characteristics of clusters were determined by chi-square test or Fisher's exact test followed by residual analysis (12).
To detect the correlation between pre-existing and new lesions, we conducted 4 analyses, for the reason that the numbers in the cells were small and the power to detect correlation was not strong. First, we made a checkerboard of 20 pre-existing and new lesions. Preexisting and new lesions that numbered fewer than 10 and 4, respectively, were excluded from analysis.

Patient characteristics
The subjects were 144 females (69.2%), the age at entry was 59.25 ± 13.16 years, and 46.8% had a history of smoking ( Table 1). The age of RA onset was 51.31 ± 15.26 years and disease duration was 7.94 ± 9.31 years. Anti-CCP antibody and RF were positive in 85.9% and 84.1% of the subjects, respectively; and there was usage of MTX and glucocorticoids in 65.3% and 75.4% of the subjects, respectively.
The agreement rate and the kappa statistic between the two evaluators regarding detection of lesions were 0.90 and 0.52, respectively.

Cluster analysis of the patterns of existence of pulmonary abnormalities
Because many individuals had several pulmonary lesions, we examined the patterns of their occurrence by cluster analysis, in which clusters were determined according to the presence or absence of the 20 lesions. Six clusters were identified (Fig. 1A): cluster 1, no pulmonary abnormalities; cluster 2, bronchiolitis alone; cluster 3, AD + curved linear opacity; cluster 4, AD; cluster 5, AD + nodular lesions; and cluster 6, AD + ILD (reticular pattern). Of note, AD was found in most patients with pulmonary abnormalities (clusters 2-6) suggesting that AD is a shared pulmonary abnormality in patients with pulmonary lesions. As shown in Table 2 There was much variation in the frequencies of worsening and improving lesions among the 20 lesions (Suppl. Table 1). For GGO and consolidation, some lesions worsened whereas others improved; however, some reticular pattern lesions and worsened, but few improved. Some honeycomb lesions worsened, and none improved. Lesions with small nodular pattern and AD including bronchiolitis and bronchiectasis worsened, but rarely improved.

Change in pulmonary abnormalities during biologic therapy
Regarding evaluation of change in a lesion between the initial and subsequent scan, the agreement rate and the kappa statistic were 0.83 and 0.65, respectively.

Patterns of newly emerging pulmonary abnormalities
To examine the emergence pattern of the pulmonary lesions, we conducted a cluster analysis of newly developed lesions in all 208 patients. New lesions formed clusters rather than developing at random. As shown in Fig. 1B, we identified seven clusters: cluster 1, no pulmonary abnormalities; cluster 2, nodular pattern; cluster 3, curved linear opacity; cluster 4, bronchiectasis; cluster 5, consolidation; cluster 6, bronchiolitis; and cluster 7, GGO.

Impact of pre-existing pulmonary lesions on development of new abnormalities
We examined whether the pre-existence of pulmonary abnormalities affected the development of new lesions. As shown in Fig. 2, new pulmonary abnormalities (especially ILD) frequently developed in patients with pre-existing diseases, particularly in those with nodular lesions and AD. The incidence of pulmonary lesions was 4.86/100 PY , which was lower than the incidence of pulmonary lesions (13.5/100 PY) in patients with pre-existing pulmonary abnormalities (Table 2).
However, not all pulmonary lesions developed in patients with pre-existing lesions.
GGO and consolidation frequently developed in patients with pre-existing disease, whereas curved linear opacity and bronchiolitis occurred in patients without pulmonary lesions ( Table 2).

Development of pulmonary abnormalities in patients without pre-existing pulmonary abnormalities
We attempted to determine the developmental pathways of pulmonary abnormalities in RA. First, to identify which lesions appear as initial abnormalities in RA, we examined the pulmonary lesions that developed in patients with no preexisting abnormalities.
Of 62 patients without pulmonary abnormalities, 11 patients (17.7%) developed a total of 23 new lesions. The most frequent lesions were bronchiolitis (n = 6; 54.5% of patients with new lesions), curved linear opacity (n = 4; 36.4%) and bronchiectasis (n = 3; 36.4%) (Fig. 1C). A heat map of the lesions suggested that the initial pulmonary abnormalities were bronchiolitis with curved linear opacity/bronchiectasis (Fig. 1C). Taken together, bronchiolitis and curved linear opacity were the initial lesions.

Development of pulmonary abnormalities in patients with pre-existing pulmonary abnormalities
We also examined the pulmonary lesions that developed in patients with preexisting abnormalities. Of 146 patients with pulmonary abnormalities, 57 patients (39%) developed a total of 92 new abnormalities. A variety of pulmonary lesions developed in these patients. A cluster analysis of new abnormalities in patients with pre-existing lesions identified five clusters: cluster 1, no pulmonary abnormalities; cluster 2, AD with bronchiectasis/bronchial wall thickening; cluster 3, bronchiolitis; cluster 4, consolidation; and cluster 5, GGO (Fig. 1D).
The heat map shows that clusters 2 and 3 emerged in patients with pre-existing AD, which suggests that AD induces other AD as a complex form of AD. Cluster 4 (consolidation) and cluster 5 (GGO) also developed in patients with pre-existing AD.
Taken together, we consider that AD induces a variety of pulmonary abnormalities, including further progression of AD.

Correlation between pre-existing and newly emerging lesions
To examine the correlation between pre-existing and new pulmonary lesions, we performed a checkerboard analysis of pre-existing and newly emerging lesions (Suppl. Table 2). Note that many individuals had or developed several lesions. Figure 3A shows the results of the checkerboard analysis. There are several 13 correlations between the pre-existing and new lesions. There was a strong correlation of each of pre-existing honeycomb, reticular pattern, small nodular lesion, bronchiolitis, and bronchiectasis with new development of GGO.
The above relationships are summarized diagrammatically in Fig. 3B, which outlines the development pathways of pulmonary abnormalities in RA.

Discussion
The major findings of the present study are as follows: 1) ~70% of RA patients had It is known that bronchial wall thickening is associated with bronchiolitis and that bronchiolitis, particularly follicular bronchiolitis, causes bronchial dilatation and bronchiectasis (13)(14)(15); accordingly, AD lesions could cause progression to the complex form of AD as each lesion in turn induces other AD lesions.
In addition, interlobular septal thickening, reticular opacity, and honeycombing are radiological findings of fibrotic NSIP (16), although the latter two are also a frequent finding in usual interstitial pneumonia (UIP) (17). This suggests that these lesions 14 form fibrotic ILD complex similar to the AD complex.
Taking the present findings together, we propose the following developmental pathways of pulmonary abnormalities in RA (Fig. 4) This figure can also be supported by the clusters of pre-existing lesions (Fig 1A), because cluster 2 was bronchiolitis, cluster 3 was AD (particularly bronchiolitis) with curved linear opacity, cluster 4 was AD, and clusters 5 and 6 were AD with nodules and AD with fibrotic ILD, respectively.
To the best of our knowledge, this is the first study to describe the possible sequential development pathways of pulmonary abnormalities in RA. In addition, our findings indicate the importance of AD in the further development of pulmonary abnormalities in RA, both as an initial pulmonary abnormality and with regard to the variety of pulmonary lesions (particularly ILD) that arise subsequently.
We have identified bronchiolitis and curved linear opacity as the initial lesions in RA. However, we consider that bronchiolitis was a true initial lesion and curved linear opacity was a secondary lesion because the cluster analysis revealed a cluster of bronchiolitis (cluster 2) as well as a cluster of bronchiolitis and curved linear opacity (cluster 3) (Fig 1A).
The present results are in agreement with those of several previous studies regarding AD as initial lesions. Mori et al reported airway disease in 40% of patients with early RA (18). Demoruelle et al reported the subsequent development of RA in the 70% of their subjects who did not have arthritis but had airways disease and were positive for anti-CCP antibody (19). Similarly, a high prevalence of AD in preand early RA has been reported (20).
We have also shown that a variety of pulmonary lesions, particularly ILD, arose from AD. Tokuda and Takemura claimed that AD, particularly bronchiectasis, caused destruction of airways and dilatation of the airway tracts in RA, resulting in cystic lesions that resemble UIP. (21,22). In contrast, the present findings suggested that AD induced inflammatory ILD such as GGO, rather than UIP.
Although the role of AD in the development of ILD remains largely unknown, various theories have been proposed. Similar to our findings, the coexistence of AD with ILD has been reported in RA (23,24). In IIP, there is a type of ILD, respiratory bronchiolitis-associated ILD, which is associated with small airway inflammation (25). Moreover, recent studies found MUC5B promoter variant in patients with RA-UIP and IPF, but not in patients with RA without ILD (26). MUC5B is a mucin secreted by bronco-epithelial cells, hence a connection between AD and ILD in RA-ILD might be suggested. Furthermore, many studies have suggested the lung as a primary site where the immune response can cause RA (27). Thus, AD could induce ILD directly through the spread of inflammation at the site, and indirectly through generating the autoimmune response and systemic inflammation.
The present study showed a relationship between pre-existing honeycombing and new GGO, consistent with previous reports that RA patients frequently have UIP, which is associated with acute exacerbation of ILD and poor prognosis (28,29).
The prevalence of pulmonary abnormalities, ILD, and AD in the present study was similar to that in previous studies (30). The incidence of ILD in patients without pre-existing abnormalities was 3.1/100 PY, which was high compared with previously reported cumulative incidences of 5% at 10 years (31) and 6.3% at 15 years (32).
The incidence of AD in patients without pulmonary abnormalities was 3.1/100 PY, which is higher than the cumulative incidence of 4% at 10 years by the study of Nannini et al (33). The difference in incidence between the present and previous reports was due to the methods employed to detect the abnormalities. In the present study, HRCT examination was conducted regardless of the patients' respiratory symptoms, whereas in previous studies ILD was diagnosed clinically.
The present study had several limitations. First, the present study was a retrospective single-center study The follow-up HRCT examinations were conducted according to the decision of the physician rather than a planned schedule, and the interval between the scans varied. In addition, as 15% of patients had respiratory symptoms and received follow-up HRCT, there might be selection bias that caused the incidence of pulmonary abnormalities to be higher than the true incidence, although 52% of patients had no symptoms on follow-up HRCT. Second, the number of subjects was small. We did not clarify several pathways, including the development of nodular lesions. Third, our classification system of 20 lesions did not include all abnormalities because they are so numerous in RA; the glossary of the Fleischner Society describes 106 lesions (11). Fourth, because pulmonary lesions coexist with other lesions, we considered that the combination of preexisting lesions was more important than the individual ones. However, we did not find clear relationship between pre-existing combination patterns and new lesions, which might be due to the small number of patients and methods of analysis. Lastly, the present subjects were treated with biologic therapy, which could have modified the incidence and development of new lesions.
In conclusion, we have proposed developmental pathways of pulmonary abnormalities in RA, and additionally showed the important role of AD in the development of pulmonary abnormalities such as ILD. As well as helping to predict the onset of pulmonary abnormalities, which is important in the management of RA, the present study might enable development of a method for preventing the emergence and progression of pulmonary abnormalities in RA, including ILD, by addressing airway inflammation. To verify and confirm our findings, prospective multi-centered studies are required that include a large number of patients.

Availability of data and materials
The datasets generated and/or analyzed during the current study are not publicly available for ethical and privacy reasons.

Contributions
AT and KK contributed to the study design, data analysis, data interpretation and manuscript writing. SS and SA contributed to the data analysis and data interpretation. YT, TM, AH, TH and RY contributed to the data analysis. TO and RM contributed to the study design, and data interpretation. MA contributed to the study design, data analysis, and data interpretation. All authors read and approved the final manuscript Ministry of Health, Labor and Welfare, Japan. As this was a retrospective observational cohort study, participants were not required to provide informed consent.

Consent for publication
Not applicable NOTE Figure 3B was not included in this submission. Figure 1 Patterns of pre-existing and new pulmonary abnormalities Clusters determined according to 24 Figure 2 Cumulative incidence rates of new pulmonary abnormalities in patients with pre-existing abn 25 Figure 3 Correlation between pre-existing and new pulmonary abnormalities A: Summary of checkerbo