Background

Patients with rheumatoid arthritis (RA) are at risk of developing interstitial lung disease (ILD). The reported prevalence of ILD in patients with RA varies widely depending on the population studied and the criteria used to define ILD, with estimates ranging between 2% and 41% [1,2,3,4,5,6]. RA-ILD confers a substantially increased risk of mortality [4, 7,8,9,10,11]. A recent meta-analysis of data from 15 studies in patients with RA-ILD found a mortality rate of 49% over 5 to 10 years [12]. Patients with RA may also have other respiratory complications such as pleural disease [13], bronchiectasis [6, 14] or emphysema [6, 15].

The high mortality associated with RA-ILD means that it is important that it is detected and treated promptly. Indeed, a delay in the diagnosis of RA-ILD has been shown to be an independent predictor of mortality [16]. Early detection of RA-ILD would require screening of asymptomatic patients, as substantial lung damage and loss of lung function may occur before patients report respiratory symptoms such as dyspnoea [17,18,19] and respiratory symptoms may be underestimated in patients in whom joint disease impairs exertion. However, in an international survey of 354 rheumatologists conducted in 2019, 44% did not consider screening for RA-ILD to be necessary in patients with risk factors but no respiratory symptoms [20].

The gold standard for radiological assessment of ILD is a high-resolution computed tomography (HRCT) scan [21, 22]. Evidence is emerging on the use of lung ultrasound as a screening tool [23,24,25,26,27], but more data are needed to establish its accuracy in the detection of RA-ILD. The high prevalence of RA and lower prevalence of RA-ILD means that it would be inefficient and likely not feasible for all patients with RA to undergo HRCT, but there is no consensus as to which patients should be screened. Many factors have been associated with the development of ILD in patients with RA, including older age, male sex, smoking history, seropositivity for rheumatoid factor (RF) or anti-cyclic citrullinated peptide (CCP), elevated erythrocyte sedimentation rate (ESR), moderate or high RA disease activity, and the MUC5B promoter variant [3, 7, 28,29,30,31,32,33]. The risk of ILD appears to be similar in patients with RA taking different types of disease-modifying anti-rheumatic drugs (DMARDs) [34], although recent analyses have suggested a lower risk in patients taking tofacitinib compared with adalimumab [35] and in patients taking abatacept plus methotrexate compared with methotrexate alone [36]. The extent to which treatments for RA modify the risk of ILD, and whether any reduction in risk is due to better control of RA disease activity or to specific effects in the lung, remains unclear.

The Autoantibodies, Non-articular manifestations of RA, Cigarettes, He/him, Older age at RA onset, RA high disease activity (ANCHOR)-RA study is a multi-national cross-sectional study that will develop a multivariable model for prediction of RA-ILD, which can be used to inform screening for RA-ILD in clinical practice. This study will also evaluate the prevalence of RA-ILD, features of RA-ILD on HRCT, and other pulmonary complications of RA, in patients with risk factors for RA-ILD. A sub-study will compare the performance of lung ultrasound and HRCT in the detection of ILD. This manuscript describes the design of the ANCHOR-RA study.

Methods/design

Study design and participants

The ANCHOR-RA study is a cross-sectional study that will be conducted at approximately 30 sites in the USA, UK, Germany, France, Italy, Spain. Eligible patients will be diagnosed with RA according to the 1987 American College of Rheumatology (ACR) [37] or the 2010 ACR/European Alliance of Associations for Rheumatology (EULAR) [38] classification criteria, with any duration of RA, and have ≥ 2 of the following risk factors for RA-ILD: male; current or previous smoker; age ≥ 60 years at RA diagnosis; high-positive RF and/or anti-CCP (titre > 3 x upper limit of normal); presence or history of certain extra-articular manifestations of RA (vasculitis, Felty’s syndrome, secondary Sjögren’s syndrome, cutaneous rheumatoid nodules, serositis, and/or scleritis/uveitis); high RA disease activity in the prior 12 months (Table 1). Patients will be excluded if they have previously been identified as having ILD, have had a chest CT in the prior 2 years, have received treatments known to induce ILD (e.g. radiation therapy to chest, bleomycin), have had a lung transplant, or have been diagnosed with an autoimmune disorder overlapping with RA, except secondary Sjögren’s syndrome. There is no restriction on the use of disease-modifying antirheumatic drugs (DMARDs).

Table 1 Disease activity measures used to ascertain high RA disease activity in the ANCHOR-RA study
Full size table

Investigators will enrol consecutive patients from RA outpatient clinics using one of three methods: (1) they may contact consecutive patients from their clinic database to ascertain interest and screen them for eligibility; (2) they may conduct a search of their clinic database to identify patients who have ≥ 2 of the specified risk factors for RA-ILD and then contact consecutive patients in that group to ascertain interest and screen for eligibility, or (3) they may screen consecutive patients for eligibility and interest during routine clinic sessions.

The study will be carried out in compliance with the protocol, the principles laid down in the Declaration of Helsinki, and in accordance with the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH), Harmonised Tripartite Guideline for Good Clinical Practice and Guidelines for Good Pharmacoepidemiology Practices. The Advarra Institutional Review Board has approved the study protocol (Pro00071380). All participants will provide written informed consent prior to entering the study. The study is registered on clinicaltrials.gov (NCT05855109; submission date: 3 May 2023).

Data collection

Participants will undergo a volumetric HRCT scan using a standardised multi-detector row CT acquisition protocol at their local site. Visual assessment of scans will be conducted centrally by two expert radiologists. Imaging features on HRCT will be assessed according to established recommendations [21, 22, 45]. The presence/absence of ILD, free-standing bronchiectasis and emphysema will be adjudicated. Features related to these diseases will be assessed (Table 2). ILD will be defined as non-dependent abnormalities, including ground glass or reticular abnormalities, lung distortion, traction bronchiectasis, honeycombing, or non-emphysematous cysts. If there is disagreement between the radiologists regarding the presence or absence of ILD, bronchiectasis, or emphysema, a third expert radiologist will make the decision. HRCT scans will also be read by a local radiologist and incidental findings reported back to the investigator, as per standard of care.

Table 2 Features on HRCT that will be assessed in the ANCHOR-RA study
Full size table

Data will be collected prospectively on demographic and RA-related characteristics (including disease activity based on Disease Activity Score with 28 joints [DAS-28]), extra-articular features, comorbidities, pulmonary function and patient-reported outcomes (Table 3). The investigator will be required to start these assessments within 90 days of the participants providing informed consent and collect the data within a 90-day period. Participants can opt in or out of MUC5B promoter variant testing at a central laboratory and of biobanking of DNA, plasma and serum samples for future analyses.

Table 3 Participant information to be collected prospectively in the ANCHOR-RA study
Full size table

Data collected from medical records will comprise: age at RA diagnosis; previous treatments for RA; presence of erosions in hands on plain film radiography; disease activity based on the measures listed in Table 1 other than DAS-28 (remission/low/medium/high; first measures and last four measures); anti-SS-A/Ro antibody status; anti-SS-B/La autoantibody status; anti-nuclear antibody status and titres; RF status; anti-CCP status; and highest values and last four values for RF, anti-CCP, CRP and ESR.

Outcomes

The primary outcomes will be the development of a probability score for RA-ILD, based on a multivariable model incorporating potential risk factors commonly assessed in clinical practice, and an estimate of the prevalence of RA-ILD and of radiological features of RA-ILD on HRCT in the study population.

Secondary outcomes will be the demographic and disease characteristics of participants with RA who do and do not have respiratory symptoms (cough and/or dyspnoea), the performance of the model in sub-populations with and without cough and/or dyspnoea, and correlations of pulmonary function tests and abnormal findings on auscultation with findings on HRCT.

Further outcomes include a multivariable model that additionally considers the MUC5B promoter variant as a potential predictor of RA-ILD. The performance of the model to predict RA-ILD defined using different extents of ILD and patterns on HRCT will be assessed in sensitivity analyses. The prevalence and HRCT features of free-standing bronchiectasis and emphysema will also be assessed.

Model development and validation

Model development and validation will follow recommendations for diagnostic models [50]. Candidate risk factors for RA-ILD will be identified based on expert opinion, supported by a systematic literature search. A logistic regression model using least absolute shrinkage and selection operator (LASSO) penalisation will be fit with the full set of candidate predictors. The model may include a set of pre-specified terms to consider potential non-linear and interaction effects. The penalty parameter will be determined by a cross-validation procedure. The model will then be re-fit via classical maximum-likelihood methods with only those predictors with coefficients not set to zero in the first fit. Some predictors may be forced into the model following consultation with the steering committee. The overall performance of the model will be assessed using the Brier Score, Nagelkerke R² and adjusted Nagelkerke R². Discrimination will primarily be assessed using the area under the receiver operating characteristic (ROC) curve (c-statistic). Calibration will be assessed by a plot. Internal validation will be performed using a bootstrapping approach, which will be used to correct for over-optimism in the performance of the model. The ability of the model to identify participants who have RA-ILD on HRCT will be assessed by determining a cut-off for predicted probability (decision threshold). Participants with a predicted probability above the threshold will be classified as positive (i.e. as having RA-ILD) and participants with a predicted probability under the threshold as negative (i.e. as not having RA-ILD). Further performance measures (including sensitivity, specificity, positive and negative predictive values) of the model will be assessed. A similar process will be used to derive a model that additionally considers the MUC5B promoter variant. Reporting will follow the Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) statement [51].

Sample size calculation

It is planned that up to 1200 participants will be enrolled. This sample size was calculated based on an assumed number of model parameters of 22 to 24, an assumed prevalence of RA-ILD in the study population (which is enriched for patients with RA-ILD) of 30%, an assumed c-statistic for the model of 0.75, and an assumed shrinkage factor associated with the predictors of 0.9 [52]. In this setting, the number of events per predictor parameter would be 15. A difference of 0.05 between the apparent and adjusted R2 of the model and a margin of error of 0.05 in estimating the risk of RA-ILD when the coefficients for all candidate predictors are set to zero is expected. After HRCT results from the first 600 participants are available, an interim analysis will be performed to determine whether the prevalence of RA-ILD in the study population will enable the modelling based on risk factors or if the sample size needs to be adjusted.

Ultrasound sub-study

To investigate the performance of lung ultrasound in detecting ILD compared to HRCT, including predictive values, sensitivity and specificity, a sub-study will be conducted at sites where rheumatologists have been trained on bedside lung ultrasound. A standardised approach to ultrasound probe, anatomical location and data collection will be provided. Images will be taken from 14 areas. The measures analysed will include the total number of B lines, pleural thickness (normal/abnormal), pleural regularity (normal/irregular) and the presence/absence of pleural effusion and consolidation. It is not intended that sites will refer participants to radiology for formal ultrasonographic assessment.

Discussion

The data collected in the ANCHOR-RA study will establish the prevalence of RA-ILD in a high-risk population and enable development of a multivariable model for prediction of RA-ILD. This model will be based on risk factors that are easily assessed in clinical practice, which can be used to decide which patients with RA should be screened for ILD using HRCT. Risk scores developed in previous studies have shown good specificity and sensitivity for the identification of patients with RA-ILD [32, 53,54,55,56,57]. However, these risk scores were developed in small or single-centre populations [32, 53,54,55,56] and some were based on patients with known ILD [53, 57] or in patients referred for HRCT because ILD was suspected [56]. A risk score developed based on sex, smoking status, CDAI, ESR and extra-articular manifestations in 118 patients with RA had 90% sensitivity and 64% specificity for identifying patients with RA-ILD on HRCT [53], while a risk score developed based on sex, age at RA onset, DAS-28-ESR score and the MUC5B promoter variant in 163 patients with RA had 75% sensitivity and 85% specificity [32]. A risk score developed based on age at RA onset, smoking status, RF titre, CCP titre and DAS-28 score in 430 patients with RA had 86% sensitivity and 58% specificity for identifying patients with RA-ILD on HRCT [56]. A nomogram developed using data on sex, smoking status, RF, CRP and matrix metalloproteinase-3 from 223 patients at a single centre had a c-index of 0.826 for identification of RA-ILD compared with an assessment by a multidisciplinary team [55]. We seek to build on these studies by developing a risk score for RA-ILD that could easily be applied in clinical practice. The results of the ANCHOR-RA study will also elucidate the prevalence of RA-ILD in patients with particular risk factors. Understanding the prevalence of RA-ILD in sub-populations of patients with RA is important given its significant impact on prognosis [4, 7,8,9,10,11] and the burden that it places on healthcare resources [58].

The most frequent radiological patterns seen in patients with RA-ILD are usual interstitial pneumonia (UIP) and non-specific interstitial pneumonia (NSIP) [4, 59,60,61]. A UIP pattern on HRCT has been associated with worse survival in patients with RA-ILD [62,63,64]. Greater extents of fibrosis, traction bronchiectasis, honeycombing, and reticulation, and the presence of emphysema, have also been associated with worse survival [60, 64,65,66]. The ANCHOR-RA study will provide further data on the prevalence of these features on HRCT in patients with RA-ILD.

Early diagnosis of RA-ILD provides the opportunity to improve its monitoring and management. The course of RA-ILD is variable, with some patients experiencing rapid progression while others remain relatively stable [67, 68]. Progressive pulmonary fibrosis in patients with RA-ILD is characterised by increasing radiological fibrosis, decline in lung function, worsening symptoms, and early mortality [67, 69, 70]. Early detection of RA-ILD, and of its progression, enables treatment to be initiated promptly. Data from retrospective or uncontrolled studies suggest that immunosuppression may slow the progression of RA-ILD [71,72,73,74], but in the absence of randomised controlled trials, the effect of these therapies remains unclear. Based on the results of the randomised placebo-controlled INBUILD trial [75], the tyrosine kinase inhibitor nintedanib has been licensed for the treatment of progressive fibrosing ILDs of any aetiology and was given a conditional recommendation for use in patients with progressive pulmonary fibrosis who have failed standard management for fibrotic ILD in a clinical practice guideline [22]. Data from the subgroup of patients with progressive fibrosing RA-ILD in the INBUILD trial provide further support for its use in these patients [76].

Strengths of the ANCHOR-RA study include that it is large and multi-national, that consecutive recruitment of patients will minimise the risk of bias (e.g. investigators preferentially putting forward patients suspected to have ILD), and that a wide range of clinical and lung imaging data will be collected. Limitations include that the results will be applicable only to patients with RA who have the specified risk factors for ILD.

In conclusion, the results of the multi-national cross-sectional ANCHOR-RA study will enable the development of a probability score for RA-ILD based on a multivariable model incorporating risk factors and will elucidate the prevalence of RA-ILD in patients with multiple risk factors. These data will add to the body of evidence that will support recommendations for screening for RA-ILD to improve detection of this important complication of RA and enable early intervention.