Incidence and mortality of interstitial lung disease in rheumatoid arthritis: A population-based study

Authors


Abstract

Objective

Interstitial lung disease (ILD) has been recognized as an important comorbidity in rheumatoid arthritis (RA). We undertook the current study to assess incidence, predictors, and mortality of RA-associated ILD.

Methods

We examined a population-based incidence cohort of patients with RA and a matched cohort of individuals without RA. All subjects were followed up longitudinally. The lifetime risk of ILD was estimated. Cox proportional hazards models were used to compare the incidence of ILD between cohorts, to investigate predictors, and to explore the impact of ILD on survival.

Results

Patients with RA (n = 582) and subjects without RA (n = 603) were followed up for a mean of 16.4 and 19.3 years, respectively. The lifetime risk of developing ILD was 7.7% for RA patients and 0.9% for non-RA subjects. This difference translated into a hazard ratio (HR) of 8.96 (95% confidence interval [95% CI] 4.02–19.94). The risk of developing ILD was higher in RA patients who were older at the time of disease onset, in male patients, and in individuals with more severe RA. The risk of death for RA patients with ILD was 3 times higher than in RA patients without ILD (HR 2.86 [95% CI 1.98–4.12]). Median survival after ILD diagnosis was only 2.6 years. ILD contributed ∼13% to the excess mortality of RA patients when compared with the general population.

Conclusion

Our results emphasize the increased risk of ILD in patients with RA. The devastating impact of ILD on survival provides evidence that development of better strategies for the treatment of ILD could significantly lower the excess mortality among individuals with RA.

Is there a true association between rheumatoid arthritis (RA) and interstitial lung disease (ILD), and if so, what impact does ILD have on the survival of affected patients? Data suggesting a connection between the most common inflammatory joint disease and this rare pulmonary condition are based primarily on case series and referral center–based patient cohorts. There have been no population-based studies investigating the presence of ILD in the full spectrum of patients with RA and comparing incidence in individuals with and without RA.

The first description of pulmonary involvement in RA was provided by Ellman and Ball (1), who in 1948 reported 3 cases in which the classic manifestations of RA and extensive pulmonary disease appeared to be associated with the same underlying process. In 2 cases, autopsy was available, and the lungs in both showed a chronic fibrosing type of pneumonitis.

Since then, several studies based on referral cohorts have provided estimates of prevalent ILD among patients with RA; those estimates have ranged between 1% and 58% (2–7). The great variation in estimates of occurrence is not surprising, given the plethora of definitions and means of detection employed to diagnose ILD and its analog.

“RA-associated ILD” has served as a general term for a variety of different conditions affecting predominantly the pulmonary parenchyma, based on radiologic and/or histologic assessment. Importantly, this term has been used interchangeably with terms such as “RA-lung,” “RA–fibrosing alveolitis,” “RA–diffuse parenchymal ILD,” and “RA-pulmonary fibrosis” or “connective tissue disease–associated ILD.” As a result, studies of ILD in patients with RA have been hampered by a lack of acknowledged terminology and validated classification criteria. Furthermore, incidence or prevalence data based on referral samples may overestimate the incidence of ILD, since extraarticular disease is thought to be more frequent in patients with more severe RA (7–9).

Given the limitations of the existing evidence, we aimed to better define the incidence, risk factors, and mortality of ILD in patients with RA, in a population-based setting, using strict and reproducible classification criteria for ILD.

PATIENTS AND METHODS

The Rochester Epidemiology Project.

The Rochester Epidemiology Project, a medical records linkage system in Rochester, MN, allows access to the complete (inpatient and outpatient) records of the local population from all health care providers. This data system ensures virtually complete access to clinical and vital status information of all clinically recognized cases of RA among Rochester residents (10).

Cohort of patients with RA.

Using the Rochester Epidemiology Project as a data resource, a population-based incidence cohort of all cases of RA among Rochester residents ≥18 years of age, who were first diagnosed between January 1, 1955 and January 1, 1995, was assembled as described previously (11–13). All cases were diagnosed according to the 1987 American College of Rheumatology (ACR; formerly, the American Rheumatism Association) classification criteria for RA (14). Incidence date was defined as the date that a patient's disease first fulfilled 4 of the 7 ACR diagnostic criteria. This RA incidence cohort consisted of 603 subjects.

Matched control cohort of patients without RA.

For each of the 603 subjects with RA, a non-RA control patient, matched for birth year (±3 years) and sex, was randomly selected from the same source population. Non-RA subjects were also matched for the prior duration of their medical records history. All subjects in this cohort were assigned an index date corresponding to the date of RA incidence in the RA patient to whom they were matched.

Data collection.

The data abstraction process has been described in detail previously (11–13). Briefly, all subjects (both those who had RA and those who did not have RA) were followed up longitudinally through their complete medical records beginning at age 18 years (or date of migration to Rochester, MN, for those who became residents after age 18) and continuing until death, migration from Rochester, or January 1, 2006. Demographic and clinical characteristics (observed at any time after RA incidence/index date, unless stated otherwise) were abstracted by 4 trained nurse abstractors who were blinded with regard to the study hypothesis. These characteristics included smoking status (categorized as current, former, or never at RA incidence date); rheumatoid factor seropositivity (≥40 IU/ml); erythrocyte sedimentation rate (ESR); swelling, erosions, and/or destructive changes in large joints, as revealed on radiographs; the presence or absence of rheumatoid nodules; extraarticular disease; use of disease-modifying antirheumatic drugs (DMARDs) and/or corticosteroids; functional capacity (according to the Steinbrocker criteria (15) at RA incidence/index date); and sustained elevation of the ESR (3 recorded ESR values of ≥60 mm/hour, with a minimum interval of 30 days between 2 measurements).

Definition and ascertainment of ILD.

The prospectively determined criteria used to classify ILD were the result of consensus-forming discussions among 2 pulmonologists (JHR, RV) and 2 rheumatologists (TB, ELM). For the purpose of this study, ILD was divided into 2 levels of diagnostic certainty: “probable ILD” and “definite ILD.” Criteria were based on clinical data, pulmonary function test (PFT) results, radiologic studies, and lung biopsies (Table 1).

Table 1. Classification criteria for ILD in RA*
  • *

    ILD = interstitial lung disease; RA = rheumatoid arthritis; CT = computed tomography; PFT = pulmonary function testing; TLC = total lung capacity.

Probable ILDChest radiography/CT report containing terms such as “pulmonary fibrosis,” “fibrotic changes,” “fibrosis,” “RA-lung,” “fibrosing alveolitis,” and presence of nonspecific abnormalities that can be observed in ILD
and
Treating physician's diagnosis of “pulmonary fibrosis,” “RA-lung,” “fibrosing alveolitis,” or other terms in the medical record consistent with ILD
Definite ILDDiagnosis of ILD by a pulmonologist
and
Two of the following 3 criteria:
 ILD observed on CT or chest radiograph
 Restrictive pattern observed on PFT (TLC ≤80% predicted)
 Bronchoscopic or surgical lung biopsy results consistent with ILD

The medical records of all 1,206 individuals in the RA and non-RA cohorts were re-reviewed for parameters of pulmonary disease by 2 physician scientists and 1 trained nurse abstractor. All pulmonary diagnoses, results of PFTs, chest radiography data, computed tomography (CT) results, and pulmonary biopsy results (including autopsy reports) were abstracted into number/identifier codes. After completion of the abstraction process, a computer-based algorithm was applied in order to identify patients who met our ILD classification criteria. To minimize incidence–prevalence bias, RA patients and non-RA subjects were excluded from further analysis if they met the classification criteria for ILD prior to the incidence of RA/index date.

Exploration of accuracy of ILD classification criteria.

To investigate the validity of our criteria in classifying probable and definite ILD, we examined the medical records of all RA patients who met the criteria for probable or definite ILD, as well as the records of 30 randomly selected patients from the RA cohort who did not meet the criteria. These records and radiologic images (when available) were reviewed by a pulmonologist (JHR) and a rheumatologist (TB) who were blinded with regard to the criteria-based classification of the subjects. Based on this expert review, a reference diagnosis of “no ILD,” “probable ILD,” or “definite ILD” was assigned to each individual.

Kappa statistics were used to quantify agreement between criteria-based and expert-based diagnosis. In addition, we calculated the test characteristics of our classification criteria using “chart review–based expert opinion” as a gold standard.

Statistical analysis.

Descriptive statistics were used to summarize the data. Demographics were compared by 2-sample t-tests and chi-square tests, using SAS (SAS Institute) and S-Plus (Insightful) software. Cumulative incidence of ILD was estimated, adjusting for the competing risk of death. Cox proportional hazards models, adjusting for age, sex, and smoking status, were used to compare estimates between cohorts and to investigate possible predictors of ILD. Time-dependent covariates (such as the presence of extraarticular disease, large joint swelling, use of DMARDs, etc.) were used to represent risk factors that could develop over time; a variable was considered absent from RA incidence/index date until the date during the disease course when it became present. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) were calculated.

The survival rates among RA patients with ILD were compared with the survival rates among RA patients in the population. Expected survival rates among RA patients were obtained by multiplying the survival rates among the white Minnesota population by the standardized mortality ratio (SMR) for RA survival, i.e., 1.3. Sensitivity analyses were performed, in which age group– and sex-specific SMR values were used to modify the population rate table, and similar results were obtained. Overall survival following ILD was estimated by obtaining the age-, sex-, and calendar year–specific survival rates among the RA patients with ILD from this modified rate table. A 1-sample log rank test was used to determine whether the observed mortality of the RA patients with ILD differed from the expected mortality for RA patients. Multivariable Cox models were used to examine the effect of ILD on survival of patients with RA, after adjusting for smoking status, age, and sex.

We also estimated the risk of mortality attributable to ILD in the RA cohort. The attributable risk is the proportion of disease in a population that could be prevented by elimination of an exposure, or risk factor. Attributable risk is commonly calculated according to the equation AR = (P[D] − P[D, no F])/P(D), where AR is the attributable risk, P(D) is the probability of disease (i.e., death), and P(D, no F) is the conditional probability of disease among individuals without the risk factor. The cumulative incidence of death was used to estimate the probability of death, and these estimates were obtained from Cox models to allow for adjustment for age and sex. The conditional probability of death for those without ILD was estimated from the same Cox models, but with a target cohort that matched the observed cohort except for the fact that the subjects did not have ILD.

RESULTS

Cohort of RA patients and matched cohort of non-RA patients.

The population-based RA incidence cohort and the matched non-RA cohort comprised 603 patients each. After excluding 21 patients with ILD diagnosed prior to the RA diagnosis/index date, 582 patients with incident RA and 603 control subjects were used in the analysis. The mean age at RA incidence/index date was 58 years. The mean duration of followup was 16.4 years for RA patients and 19.3 years for non-RA subjects, and 73% of the patients were women. Smoking was more common among RA patients; 28.2% of RA patients were current smokers and 24.7% were former smokers, compared with 23.9% of non-RA subjects who were current smokers and 19.6% who were former smokers (P < 0.01). Chest radiographs were available for 230 RA patients and for 352 non-RA patients. “Fibrotic changes” were observed in a similar number of RA and non-RA patients (27.0% and 25.5%, respectively). A minority of both RA and non-RA patients underwent PFT (19.1% and 17.2%, respectively). A detailed description of both cohorts is provided in Table 2.

Table 2. Demographic and clinical characteristics of the RA patients and controls*
 RA patients (n = 582)Control patients (n = 603)P
  • *

    Except where indicated otherwise, values are the number (%) of subjects. See Table 1 for definitions.

Age at RA diagnosis/index date, mean ± SD years57.7 ± 15.158.2 ± 15.20.58
Years of followup, mean ± SD16.4 ± 10.519.3 ± 11.1
Women427 (73.4)441 (73.1)0.93
Smoking status at baseline  
 Current164 (28.2)144 (23.9)<0.01
 Former144 (24.7)118 (19.6) 
Documented chest radiography230 (39.5)352 (58.4)<0.01
Fibrotic changes on chest radiography157 (27.0)154 (25.5)0.57
Documented PFT111 (19.1)104 (17.2)0.42
Restrictive pattern on PFT (TLC ≤80% predicted)27 (4.6)11 (1.8)<0.01
Documented CT78 (13.4)85 (14.1)0.73
Presence of parenchymal disease on CT38 (6.5)24 (4.0)0.05

Incidence of ILD in patients with RA.

Probable or definite ILD developed in a total of 46 of the 582 RA patients (7.9%). Of these, 23 patients (4.0%) met our strict criteria for definite ILD.

The 10-, 20-, and 30-year cumulative incidence rates for probable and definite ILD were 3.5%, 6.3%, and 7.7%, respectively (adjusted for the competing risk of death), conveying a lifetime risk of ∼10%. Among non-RA subjects, much lower 10-, 20-, and 30-year cumulative incidence rates for probable and definite ILD were observed (0.2%, 0.9%, and 0.9%, respectively).

The risk of developing ILD among RA patients was significantly higher than among non-RA subjects (HR 8.96 [95% CI 4.02–19.94]), after adjusting for age, sex, and smoking status (Figure 1).

Figure 1.

Incidence of interstitial lung disease in rheumatoid arthritis (RA) patients (solid line) and in control subjects (broken line).

Validation of ILD classification criteria in patients with RA.

The records of all patients identified as having RA-associated ILD according to our classification criteria, as well as the records of a random sample of 30 patients with RA but not ILD, were reviewed by a pulmonologist and a rheumatologist. Our composite criteria were in excellent agreement with chart review–based expert opinion across all 3 categories of ILD (none, probable, and definite). Kappa statistics yielded an index of 0.88 (95% CI 0.79 – 0.97). When considering chart review–based expert opinion as a gold standard, the sensitivity and specificity of these criteria were 1.0 (95% CI 0.92–1.0) and 0.91 (95% CI 0.76–0.97), respectively.

Characteristics of ILD in RA and non-RA patients.

In both cohorts, a higher proportion of subjects diagnosed as having ILD were men. On average, RA patients were 15 years younger when diagnosed as having ILD, as compared with non-RA subjects.

CT scans of the chest were available for only a limited number of patients with ILD (50.0% of RA subjects and 57.1% of non-RA subjects). Pulmonary biopsy was performed in 14 of 46 patients with RA-associated ILD. A summary of ILD characteristics in RA patients and in non-RA subjects is provided in Table 3.

Table 3. Characteristics of the RA and non-RA patients with ILD*
 RA patients with ILD (n = 46)Non-RA with ILD (n = 7)
  • *

    Except where indicated otherwise, values are the number (%) of subjects. UIP = usual interstitial pneumonia; NSIP = nonspecific interstitial pneumonia; OP = organizing pneumonia (see Table 1 for other definitions).

  • P < 0.01 versus non-RA patients.

  • According to the American Thoracic Society consensus classification of idiopathic interstitial pneumonias (16).

ILD  
 Probable23 (50.0)5 (71.4)
 Definite23 (50.0)2 (28.6)
Age at ILD diagnosis, mean ± SD years71.3 ± 12.286.5 ± 10.0
Men27 (58.7)4 (57.1)
Smoking status  
 Current15 (32.6)1 (14.3)
 Former15 (32.6)4 (57.1)
Chest radiography performed45 (97.8)7 (100)
Chest CT performed23 (50.0)4 (57.1)
Lung biopsy/autopsy performed14 (30.4)0 (0.0)
ILD subtype based on  biopsy and/or CT  
 UIP typical pattern9 (19.6)3 (42.9)
 NSIP typical pattern1 (2.2)0 (0.0)
 OP typical pattern3 (6.5)0 (0.0)
 Pulmonary fibrosis, not  further specified15 (32.6)0 (0.0)
No CT or biopsy documented15 (32.6)3 (42.9)
PFT (TLC ≤80% predicted)15 (32.6)2 (28.6)
No TLC documented21 (45.7)4 (57.1)

Risk factors for ILD in RA patients.

The risk of developing ILD was increased in patients who were older at the time of their RA diagnosis (per 10-year increase in age HR 1.41 [95% CI 1.11–1.79]) and among male patients (HR 4.37 [95% CI 2.43–7.88]). Other risk factors having a statistically significant association with development of ILD were related to markers of disease activity and severity, including erosions or destructive changes and rheumatoid nodules, high ESR levels, functional status (levels III and IV versus levels I and II, according to the Steinbrocker criteria), and treatment with corticosteroids or methotrexate (MTX) (Table 4).

Table 4. Risk factors for ILD in patients with RA*
 RA patients with ILD (n = 46)RA patients without ILD (n = 536)HR95% CI
  • *

    Except where indicated otherwise, values are the number (%) of subjects. Models for extraarticular disease, rheumatoid factor, erosions, destructive joint changes, erythrocyte sedimentation rate (ESR), large joint swelling, rheumatoid nodules, use of disease-modifying antirheumatic drugs (DMARDs), use of methotrexate (MTX), use of steroids, and functional capacity level were adjusted for age at RA onset, sex, and smoking status. See Table 1 for other definitions.

  • Data not available on all patients.

  • According to the Steinbrocker criteria. The hazard ratio (HR) and 95% confidence interval (95% CI) were calculated by comparing levels III and IV with levels I and II.

Male sex27 (58.7)128 (23.9)4.372.43–7.88
Age at RA onset, mean ± SD years56.8 ± 14.757.8 ± 15.21.411.11–1.79
Smoking (ever)30 (65.2)278 (51.9)1.600.87–2.94
Extraarticular disease (ever)6 (13.0)80 (14.9)1.480.58–3.79
Rheumatoid factor (ever)33 (73.3)341 (67.0)1.840.91–3.71
Erosions (ever)10 (25.0)83 (17.4)1.920.92–4.02
Destructive joint changes (ever)27 (67.5)271 (56.7)2.371.18–4.76
ESR with 3 values ≥60 (ever)21 (45.7)149 (27.8)3.521.94–6.38
Large joint swelling (ever)39 (84.8)448 (83.6)1.900.83–4.36
Rheumatoid nodules (ever)21 (45.7)168 (31.3)2.601.41–4.79
DMARD use (ever)27 (58.7)306 (57.1)1.820.97–3.41
MTX use (ever)12 (26.1)116 (21.6)2.311.15–4.63
Steroid use (ever)24 (52.2)287 (53.5)2.011.12–3.63
Functional capacity level    
 I or II33 (71.7)425 (79.9)2.171.14–4.16
 III or IV13 (28.3)107 (20.1)  

Mortality associated with ILD in RA patients.

The median survival of RA patients after a diagnosis of ILD was 2.6 years, which is significantly lower than the expected median survival (9.9 years) of RA patients of the same age and sex overall (P < 0.001). The survival rates among patients with RA and ILD compared with rates among RA patients overall are shown in Figure 2. Survival rates were significantly worse among patients with RA-associated ILD compared with RA patients who did not have ILD, after adjusting for age, sex, and smoking status (HR 2.86 [95% CI 1.98–4.12]). Although the cumulative incidence of ILD was significantly increased in patients with RA, the risk of death associated with ILD was similar among RA and non-RA patients, after adjusting for age, sex, and smoking status (HR 1.60 [95% CI 0.63–4.09]).

Figure 2.

Survival rates among rheumatoid arthritis (RA) patients with interstitial lung disease (ILD) (solid line) compared with expected survival rates among RA patients overall (broken line).

Excess mortality among RA patients attributable to ILD.

During followup, 382 patients with RA died, which was a significant increase over the 347 deaths among the non-RA cohort. Thirty years after RA incidence/index date, the cumulative incidence of death was 79.3% among RA patients compared with 63.4% among non-RA patients. Based on this, 30 years after RA diagnosis, the RA cohort had an excess mortality of 15.9%.

We further explored the question of how much of this excess mortality could be attributed to ILD. When removing the effect of ILD on patient survival, the cumulative mortality of patients with RA dropped by 2.1%. Thus, the excess deaths in RA patients would be reduced by 13% (i.e., 2.1 = 13% of 15.9), if the risk of ILD in RA patients were the same as in non-RA subjects. In other words, if the effect of the increased risk of ILD in RA could be eliminated, ∼1 of every 8 excess deaths among RA patients could be prevented.

DISCUSSION

According to our estimates, ∼1 in 10 patients with RA will be diagnosed as having ILD over the lifetime of their disease. This risk is significantly higher than in the general population.

An important advantage of our methodologic approach in the current study is the avoidance of referral bias, which inevitably plays a role when basing estimates on convenience samples in referral centers. Instead, we aimed to assess every patient diagnosed as having RA in a given geographic area, resulting in a better reflection of the full spectrum of disease. Nevertheless, there are some inherent methodologic limitations of a retrospective, chart review–based diagnosis of disease. For this reason, we preferred to establish a range of probabilities for our primary outcome (ILD) rather than a point value which may or may not approach reality. Hence, the “true” cumulative incidence of clinically manifest ILD in RA patients in our population is likely to lie somewhere between our estimates of probable ILD (7.7%) and definite ILD (3.7%). Since systematic cross-sectional or prospective screening was not performed in this study, it is likely that our estimates are the minimum for ILD incidence in this population. Additionally, patients who developed ILD prior to their RA diagnosis were excluded from our analysis. Therefore, the overall risk for individuals with RA to be affected by ILD either prior to or after their diagnosis of RA will be somewhat higher than our estimate.

Despite our efforts to develop a valid definition for ILD for this study, we acknowledge that this definition still represents a “collection bin” for many different types of parenchymal lung disease. Because of the changes in the availability of diagnostic tools over time, as well as the evolution of definitions used to characterize ILD, reliable assignment of ILD subtypes for every patient according to the most recent American Thoracic Society consensus classification (16) was not possible.

Other potential limitations of this study include reliance on medical record information that has been accrued in clinical care over several decades and not through systematic prospective data collection. Availability and use of technologies such as CT and PFT have greatly increased since 1955, the starting date for our incidence cohort. As a result, the number of unrecognized cases of ILD may be higher during the earlier years of our followup period, resulting in an underestimation of true incidence.

Differential medical assessment of patients with RA versus individuals without RA is another possible weakness of our approach. For RA patients, the average frequency of physician visits is likely to be higher than for non-RA individuals. This potentially results in a higher likelihood of identifying additional medical conditions such as ILD. However, because the majority of subjects in our 2 cohorts see a physician at least once in any 3-year period (10), differential identification of a major medical diagnosis such as ILD appears less likely.

The population of Rochester, MN is mainly white, and the socioeconomic characteristics largely resemble those of the white US population in general (10). Appropriate caution must be exercised when generalizing our findings to populations with a different demographic composition and geographic region. This may be especially true for a disease such as ILD, which is thought to be influenced in its occurrence and severity by gene–environment interaction (17).

How do our data on the incidence of RA-associated ILD compare with information in the existing literature? In a previous study, in which the same population-based cohort of Rochester patients was followed up for 5 fewer years than in the current study, a 30-year cumulative incidence of 6.8% for “pulmonary fibrosis” was observed (7). The classification of pulmonary fibrosis was based on “clinical judgment plus a DLCO [diffusing capacity for carbon monoxide] of <85% of normal.” This estimate lies well within the range between probable ILD and definite ILD in the current study. While we did not find additional published data on the incidence of ILD, several studies have explored the prevalence of interstitial pulmonary changes and restrictive patterns of pulmonary function in non–population-based referral cohorts. In a sample of 64 consecutive patients with longstanding RA and no respiratory symptoms, 21 (33%) were found to have “early ILD” based on “high-resolution CT features of ILD” (5). In that study, the authors noted that changes in most patients were minimal. No control data from non-RA patients was reported. Using a cohort of hospital outpatients in northwest England, Dawson et al reported that 19% of 150 patients with longstanding RA had “fibrosing alveolitis,” defined as “CT changes suggestive for UIP [usual interstitial pneumonia]” (3).

These estimates of prevalence are more than twice as high as the 30-year cumulative incidence for ILD found in our population-based study. Several differences between these studies and our approach offer possible explanations for this discrepancy. First, the above-mentioned studies were hospital-based and subject to selection bias. They were performed using samples from consecutive RA patients, which can result in accumulation of more serious cases of RA with a higher likelihood of extraarticular disease (7–9). Second, our retrospective approach, which had to rely on medical record data, may have underestimated the incidence of ILD in patients with RA, as compared with the above-mentioned prospective trials, which used a standardized assessment of every patient. While it is very likely that we did not detect subclinical disease, it is unlikely that we failed to detect clinically meaningful disease in a significant number of patients. Third, patients classified as having ILD based on CT and/or pulmonary function changes in the absence of respiratory symptoms may or may not progress to having clinically meaningful disease. Our approach of using a physician's diagnosis of ILD as an essential classification criterion condenses the definition of ILD to clinically more overt cases and results in lower estimates.

The differences in ILD definition among studies reveal a significant gap between the large number of asymptomatic individuals with radiographic or functional pulmonary changes and the small number of patients who will actually develop overt pulmonary disease. Importantly, this gap is revealed not only by contrasting studies using different definitions of ILD; a within-trial comparison of our data also demonstrated a significant discrepancy between the large number of individuals found to have fibrotic radiographic changes on CT/chest radiographs and the relatively small number of patients diagnosed as having ILD. Although not formally evaluated, factors that appeared to influence the treating physician's assessment of whether these changes indicated ILD included the extent of pulmonary involvement, progression over time, clinical symptoms (dyspnea, cough, etc.), and abnormal findings on PFT and physical examination (crackles on auscultation). In our study, the number of individuals with a documented finding of “fibrotic changes” on pulmonary imaging studies was similar between RA patients and non-RA patients (27.0% and 25.5%, respectively).

Our analysis of risk factors indicated that being male, being older at the time of RA onset, having a high “inflammatory burden,” having low functional capacity, and being treated with glucocorticosteroids or MTX, are all associated with a diagnosis of ILD in RA patients. The apparent association of parameters of high disease activity, such as persistently elevated ESR, low functional capacity, rheumatoid nodules, and glucocorticosteroid and MTX use, is consistent with the findings previously reported by Saag and colleagues (18), as well as with the existing evidence of a strong association between RA joint disease severity and occurrence of extraarticular disease.

It is unlikely that the ILD in our patients was due to treatment with MTX. MTX-induced ILD was offered as a diagnostic option on the abstraction forms, but no such cases were identified. Independent review by 2 investigators revealed no cases of MTX-induced alveolitis. Finally, since several surrogate parameters for high RA disease activity were associated with a diagnosis of ILD and since MTX is commonly used to treat patients with more severe disease, the observed association may be due to increased disease activity as a confounding variable rather than to MTX itself.

It is unclear if the identified predictors of ILD in patients with RA are also indicators of a high risk for disease progression among patients who have subclinical interstitial pulmonary changes. Dawson et al (19) previously noted some factors associated with an increased likelihood of disease progression in patients with “CT changes suggestive for UIP.” Among 15 patients with stable RA-associated ILD, who were compared with 10 patients with progressive disease, significant predictors for progression included the extent of interstitial changes on high-resolution CT, bibasilar crackles, and a reduced DLCO. Although the number of patients was small, the results of that study highlight the importance of integrating clinical, imaging, and functional information in order to decide whether an individual patient may be a candidate for therapeutic interventions.

The importance of identifying patients with RA who are at risk of ILD progression becomes evident when assessing the significantly higher mortality rates among these patients in our cohort. The risk of death almost tripled for RA patients with ILD, even after adjusting for age, smoking, and sex. Our findings did not confirm those of previous studies (20, 21), which suggested a lower mortality among patients with RA-associated ILD versus patients with idiopathic ILD. This discrepancy may be explained in part by our population-based approach, which allowed us to more completely observe the full spectrum of disease.

Patients with RA have premature mortality compared with individuals from the general population (12, 22, 23). Our findings extend prior observations about the increased risk of respiratory death in patients with RA (24, 25), as well as the significant impact of extraarticular disease on excess RA mortality (26, 27).

Much attention has been given to the increased risk of cardiovascular disease in patients with RA and its contribution to premature mortality (28–30). A recent study based on the same patient cohort we used for the current study identified congestive heart failure (CHF), but not ischemic heart disease, as an important contributor to the excess overall mortality among patients with RA (31).

Interestingly, the excess mortality in RA that was attributable to ILD appears to be very similar to the relative contribution of CHF. Hypothetically speaking, ∼1 of every 8 excess deaths in patients with RA could be prevented if the risk of ILD (or CHF) were the same in RA patients as it is in non-RA patients. Our findings suggest that prevention and treatment of RA-associated ILD could significantly improve the survival of patients with RA.

The risk of ILD is significantly increased in RA patients versus the general population. We identified several predictors of ILD in RA patients, but the pathophysiologic links between these factors and pulmonary parenchymal changes remain unclear. Our data provide evidence of the devastating impact of ILD on survival in patients with RA.

Challenges for future research in RA-associated ILD include the identification of predictors that indicate progressive disease and a need for treatment. The large gap between the number of patients with functional and radiographic abnormalities and the relatively small number of individuals with symptomatic disease highlights the importance of individualized clinical decisions in balancing the risks and benefits of treatment. These decisions are further complicated by the lack of both controlled therapeutic trials using patients with RA-associated pulmonary disease and anecdotal reports of the ILD-promoting effects of some RA therapeutics (32, 33).

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Bongartz had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Bongartz, Nannini, Ryu, Vassallo, Gabriel, Matteson.

Acquisition of data. Bongartz, Nannini, Medina-Velasquez, Ryu, Vassallo, Matteson.

Analysis and interpretation of data. Bongartz, Nannini, Achenbach, Crowson, Ryu, Vassallo, Gabriel, Matteson.

Ancillary