We performed a multicenter prospective observational study of a cohort of patients with RA-ILD from 11 university hospitals in Spain. Patients were recruited between March 2015 and March 2023. The study was approved by the Research Ethics Committee of Hospital Regional Universitario de Málaga (HRUM) (code: 1719-N-15). All participants provided their written informed consent before entering the study.

The cases were included consecutively from a prospective cohort of patients with clinically significant RA-ILD. All the patients were adults and fulfilled the 2010 classification criteria for RA of the American College of Rheumatology/European League Against Rheumatism (21). ILD was confirmed based on pulmonary function testing (PFT) and high-resolution computed tomography (HRCT) or lung biopsy (22). We excluded patients with inflammatory or rheumatic disease other than RA (except secondary Sjögren syndrome), infection, primary pulmonary hypertension, congestive heart failure, and known exposure to fibrosing environmental agents. We also excluded pregnant women.

The patients selected were seen by a rheumatologist in line with the data collection protocol at baseline (V0) and every 6-12 months, as well as at each admission for infection. Clinical and laboratory evaluations were performed for joint, lung, and infection-related variables. PFT and HRCT were performed at V0, and at 12, 24, and 60 months of follow-up or at any other time if the patient’s clinical situation so required in the opinion of the attending physician. The methodology used for performing and evaluating the HRCT and PFT in the study cohort have been described elsewhere (9, 23).

According to the protocol, coordinating meetings were held with the participants at baseline (V0), month 24 (V24), month 60 (V60), and at the end of follow-up (Vfinal). The data collection methodology was reviewed at each visit, and sufficient time was left to clarify doubts and resolve queries. In order to ensure the quality of data after collection, potential inconsistencies were analyzed at HRUM and queries that had to be corrected by the persons in charge at each center were noted.

The main outcome was the incidence of serious and fetal infections. Fatal infection was defined as a death in which infection played a key role, either immediately or during the 30 days following the last admission for infection. Incident serious infection was defined as a condition caused by infectious agents that required antibiotic therapy and for which any of the following outcomes were recorded: death, a life-threatening condition, hospitalization (initial or prolonged), disability or permanent damage, congenital anomaly/birth defect, need for an intervention to prevent permanent impairment or damage, or other important medical events. The most common infectious sites in RA were recorded, especially those affecting the upper and lower respiratory tracts, the urinary tract, and skin and soft tissue (20, 24). The etiology of the infection was determined as follows (1): staining and isolation of the microorganism in blood culture, respiratory secretions, urine, stool, pleural or joint fluid, and tissue; and (2) molecular methods and antigen testing. The microorganism had to be compatible with the clinical findings in all cases (16). The presence of Staphylococcus species, Streptococcus viridans group, Corynebacterium species, Propionibacterium species, or Bacillus species in only 1 blood culture vial was considered contamination (25). The infection was considered to be of unknown origin when no pathogen was identified in culture or antigen tests (26). Patients’ vaccination status was also evaluated to specify whether they had received the full course of vaccination against pneumococci, influenza, COVID-19, and herpes zoster.

The secondary outcome measures included baseline demographic data (sex, age, ethnic origin) and comorbidities (arterial hypertension, diabetes mellitus, dyslipidemia, obesity, and smoking history). The characteristics of RA included duration of RA, diagnostic delay, rheumatoid factor (reference value, 20 U/ml; high titer, > 60 U/ml), anti–citrullinated peptide antibodies (reference value, 10 U/ml, high values, ≥ 340 U/ml), radiological erosions, inflammatory activity according to the mean 28-joint Disease Activity Score with erythrocyte sedimentation rate (DAS28-ESR), C-reactive protein (CRP), and physical function according to the Health Assessment Questionnaire (HAQ). Mean DAS28-ESR was calculated as the mean of all DAS28-ESR values during follow-up. Similarly, we recorded the treatment used, including antibiotics, conventional synthetic disease-modifying antirheumatic drugs (csDMARDs), biologic DMARDs (bDMARDs), targeted synthetic DMARDs (tsDMARDs), immunosuppressants (azathioprine, mycophenolate mofetil, cyclophosphamide and tacrolimus), corticosteroids, and antifibrotic agents. In the case of corticosteroids, we collected the doses and mean dose during follow-up (27).

ILD was defined and classified according to the standard criteria of the American Thoracic Society/European Respiratory Society International Multidisciplinary Consensus Classification of the Idiopathic Interstitial Pneumonias (22). PFT included full spirometry expressed as percent predicted and adjusted for age, sex, and height. DLCO was evaluated using the single-breath method (DLCO-SB). Progression to ILD was categorized into the following outcomes (1): Improvement (i.e., improvement in forced vital capacity (FVC) ≥ 10% or diffusing capacity of the lungs for carbon monoxide (DLCO) ≥ 15% and absence of radiological progression) (2); Nonprogression (stabilization or improvement in FVC ≤ 10% or DLCO < 15% and absence of radiological progression) (3); Progression (worsening of FVC > 10% or DLCO > 15% and radiological progression); or (4) Death (9).

We performed a descriptive analysis of the main characteristics of patients with RA-ILD and of the type and number of infections, the initial infection, and associated types of infection in patients who died. Qualitative variables were expressed as absolute number and percentage; quantitative variables were expressed as mean (standard deviation [SD]) or median (p25-p75), depending on the normality of the distribution according to the Kolmogorov-Smirnov test. Data were missing in 10/148 patients (6.7%) for DAS28-ESR and in 42/148 patients (28.4%) for the HAQ. The patients with missing data were not excluded in order to maintain a representative cohort; however, they were managed based on imputation of data using regression analysis for the descriptive analysis; this approach involves specifying predictor variables related to the missing values and running a regression analysis to predict and fill in the missing values (28). Quantitative variables were compared using the t test for independent variables or the Wilcoxon–Mann-Whitney test; qualitative variables were compared using the Pearson χ² test. A bivariate analysis was performed with the paired t test or Wilcoxon test, as applicable, between V0 and Vfinal for lung and joint function. The risk of infection or death was estimated using incidence rates and their respective 95% confidence intervals (95%CI), using the Poisson method (29). The incidence rate for infection and death was calculated by dividing the number of events detected by the “time at risk” of the cohort. The “time at risk” was calculated in person-years by summing the times each patient had remained in observation between V0 and death and the date of the end of the observation period (Vfinal, March 2023 – censored case), since no losses to follow-up were recorded. Factors associated with infection up to the time of the first severe infection were identified using uni- and multivariate Cox regression analysis. Survival was measured from V0 to Vfinal or death. Furthermore, we ran uni- and multivariate multiple linear regression models to identify factors associated with the number of infections. The statistical analyses were performed using IBM SPSS Statistics for Macintosh, Version 28.0 (IBM Corp., Armonk, NY, USA).

A total of 148 patients with RA-ILD were prospectively followed up between March 2015 and March 2023. Mean (SD) follow-up was 56.7 (22.4) months, that is, a total cumulative exposure of 699.3 patient-years. The main characteristics at baseline (V0) are shown in Table 1 . The average age was 70 years, and 57% were women. A total of 51.4% of the patients had never smoked, while the remaining 48.6% had a history of smoking. The most comorbid condition was arterial hypertension (48%) followed by dyslipidemia (37%) and osteoporosis (37%). Most patients had seropositive disease (98%) with only 3/148 patients (2%) seronegative disease, and approximately 60% had erosions.

At baseline (V0), the average duration of ILD among patients was 2.7 years. The most common radiologic pattern was UIP (90/148 patients [60.4%]), followed by NSIP (45/148 [30.2%]), fibrotic NSIP (11/148 [7.4%]), and other types of ILD (3/148 patients [2%]).

In terms of treatment, all the patients were receiving a DMARD at V0. Most were receiving csDMARDs and slightly more than half were receiving bDMARDs. At the initiation of the study, 67 patients (45%) were receiving csDMARD monotherapy, 58 (39%) were receiving a combination of csDMARDs and bDMARDs, 23 (16%) bDMARD monotherapy, 19 (13%) a combination of a csDMARD and an immunosuppressant, and a further 15 (10%) a combination of a bDMARD and an immunosuppressant. Only 1 patient (1.4%) was receiving a csDMARD plus nintedanib. The DMARDs prescribed at V0 are shown in Table 1 . Furthermore, we can see that more than half the patients were taking corticosteroids, with a median (p25-p75) dose of 5.0 (0.0-7.5) mg/d.

At the last evaluation, 100 patients were still in follow-up, whereas 48 patients had died. There were no losses to follow-up. Table 2 shows the initial data (V0) and final data (Vfinal) for lung and joint function. A significant deterioration was observed in forced vital capacity (FVC), forced expiratory volume in the first second (FEV₁), and DLCO-SB at Vfinal. Of the first 148 patients, 66 (44.6%) experienced radiologic progression according to their HRCT scan, 78 (52.7%) had stabilized, and only 4 (2.7%) improved. Evaluation of the joints revealed no significant differences in inflammatory activity (DAS28-ESR and CRP) or in physical function evaluated using the HAQ.

As can be seen in Supplementary Table 1 , at the end of follow-up (Vfinal), 45 patients had switched therapy owing to arthritis and/or lung disease. These changes were characterized mainly by an increase in the frequency of bDMARDs in monotherapy (10.8% vs 18%; p=0.019), together with reduced use of methotrexate (40.5% vs 30.2%; p=0.004), leflunomide (25% vs 20.1%; p=0.057), and etanercept (4.7% vs 1.3%; p=0.025) in favor of increased prescription of abatacept (26.4% vs 32%; p=0.060), rituximab (12.8% vs 15.4%; p=0.103), and nintedanib (1.4% vs 5.4%; p=0.014).

During the follow-up period, 142/148 patients (96%) had at least 1 infection, with a median (p25-p75) time to first infection of 21.2 (8.0-45.2) months ( Figure 1A ). As can be seen in the survival graph, the risk of infection remained constant throughout the follow-up period. We recorded a total of 368 infectious episodes in the 148 patients, that is, a median (p25-p75) of 2 infections (1.0-3.0) per patient and an incidence rate (95% CI) of 52.6 (47.2-58.0) per 100 person-years. A direct association was observed between age and the probability of infection during follow-up. Notably, the incidence rate of infections per 100 person-years increased notably among patients aged 50 and older. The rates were 49.0 (95% CI: 38.0-60.0) in the 50-59 years age group, 44.0 (95% CI: 34.0-52.0) in the 60-69 years age group, 55.0 (95% CI: 45.0-63.0) in the 70-79 years age group and peaked at 62.0 (95% CI: 44.0-80.0) in the 80-89 years age group ( Supplementary Table 2 ).

Table 3 and Figure 2 show details of the number and type of infections, as well as the vaccination status of the 148 patients with RA-ILD. Respiratory infections were the most prevalent (274/368 [85.8%]); other sites were involved in 94/368 infections (25.5%), the most frequent being the urinary tract, followed by the skin and soft tissue.

Most patients had been appropriately vaccinated against influenza, pneumococci, and COVID-19; only 5 patients had been vaccinated against herpes zoster virus. Patients who were not vaccinated experienced a higher incidence of infections compared to those who were vaccinated (50% vs 24.6%; p=0.015), as well as a higher number of deaths (50% vs 14.3%; p=0.034). Out of 32 patients with COVID-19 infection, 10 received corticosteroids, 5 were treated with azithromycin, 4 with levofloxacin, 3 with ceftriaxone, 2 received a combination of azithromycin and levofloxacin, 1 was prescribed amoxicillin, 1 received septrim/paxlovid, 1 was treated with azithromycin/paxlovid, and 1 received ceftriaxone/paxlovid.

A total of 48 patients died during follow-up. Of these, 31 (65%) died directly from infection as the main cause: 80% from respiratory infection and 20% from other types of infection ( Table 3 ). The remaining 17/48 patients (35%) died from noninfectious causes: 14/48 mainly due to progression of lung disease, 1 from brain hemorrhage, and 2 from cancer (lung adenocarcinoma and primary brain tumor).

Table 4 and Supplementary Table 3 provide a detailed description of the causative microorganisms, number of infections, incidence rate classed by totals, first infection, infections with an identified pathogen, and deaths from infection. Respiratory infection was the most common first infection (74%), followed by urinary tract infection (9.9%) and skin and soft tissue infection (9.1%). There were no cases of ENT infection or sepsis as the first infection. Of the 31 patients who died from infection, all had had a respiratory infection during follow-up, mainly affecting the lower respiratory tract. All patients who died from infection did so during their last admission for infection, except 1 patient, who died during the 30 days following discharge (readmission). Compared with the remaining patients (ie, those who were still alive or had died of other causes), those who died from infection had a higher median (p25-p75) for the number of previous infections (3.0 [2.0-5.0] vs 2.0 [1.0-3.0]; p<0.001) and a higher incidence of infection (incidence rate [95%CI] 84.39 [68.34-100.46] per 100 person-years vs. 45.74 [40.20-51.28] per 100 person-years; incidence rate ratio [95%CI] 1.84 [1.48-2.30]) ( Figure 1B ).

With respect to the causative microorganisms ( Table 4 , Supplementary Table 3 ), the pathogen was not identified in a significant number of cases (234/368 [63.5%]); this result was similar for all 3 scenarios: total number of infections, number with a first infection, and deaths from infection. Given the total number of infections in which the causative microorganism was identified (n=134), most were caused by SARS CoV-2 (33.5%), Streptococcus pneumoniae (11.9%), Escherichia coli (11.9%), and Pseudomonas aeruginosa (11.1%). These microorganisms were also the most common in the first infection, together with Klebsiella pneumoniae. Furthermore, in the 31 patients who died, the most fatal microorganisms were SARS CoV-2 (25.8%), P. aeruginosa (12.9), and pneumococci (9.6%), followed by K. pneumoniae (3.2%), Haemophilus influenzae (3.2%), influenza A virus (3.2%), and respiratory syncytial virus (3.2%). It is worth noting that of the 8 deaths caused by SARS CoV-2, 6 (75%) were during 2020 and 2021. While no patients died from an opportunistic infection, the causative agent was not identified in 38% of fatal infections. Finally, patients who died from infection received a much higher median dose of corticosteroids than patients who survived (5.0 [5.0-15.0] vs. 0.0 [0.0-5.0], p < 0.001).

In the multivariate Cox regression models adjusted for duration of follow-up ( Table 5 ), we assessed various factors in patients with RA-ILD to understand their impact on the risk of first infection and death by infection. Age, inflammatory activity, and daily corticosteroid dosage showed as significant predictors for both outcomes. For first infections, age (hazard ratio [HR]=1.030, p = 0.011), DAS28 (HR=1.211, P=0.037), and corticosteroid dosage (HR=1.036, p = 0.012) were significant. Regarding death by infection, age (HR =1.108, p < 0.001), DAS28 (HR=1.497, P=0.037), DLCO-SB (HR=0.975, p = 0.044), and corticosteroid dosage (HR=1.065, p = 0.001) showed significance.

On the other hand, Table 6 shows the results of a multiple linear regression model of predictors for the number of infections in RA-ILD patients. Notably, time with ILD (B = 0.012, p = 0.043), mean DAS28 (B = 0.268, p = 0.009), treatment with immunosuppressants (B = 0.670, p = 0.042), and the dose of corticosteroids (B = 0.095, p < 0.001) showed significant associations with the number of infections.