PATIENTS AND METHODS

Patients

The population of Rochester, Minnesota, is well suited for an investigation of the epidemiology of RA and associated extra-articular features because comprehensive medical records for all residents seeking medical care are available. A record linkage system allows ready access to the medical records from all healthcare providers for the local population, including the Mayo Clinic and its affiliated hospitals, the Olmsted Medical Group, the Olmsted Community Hospital, local nursing homes, and the few private practitioners. The potential of this data system for use in population based studies has been described previously.^33,34 This system ensures virtually complete ascertainment of all clinically recognised cases of RA among the residents of Rochester, Minnesota.

In previous surveys^21,22 all diagnosed cases of RA aged ⩾18 years between 1 January 1955 and 31 December 1994, (n=609) were identified using the computerised diagnostic index and a review of the complete medical record in each potential case, as previously described.²² The incidence date was defined as the earliest date at which the patient fulfilled four or more of the 1987 American College of Rheumatology (ACR) classification criteria for RA.³⁵ This study was undertaken as part of further characterisation of the RA cohort.

Incidence estimates

One of the authors (CT) used a structured protocol to review the complete medical records of all patients in the cohort. Extra-articular manifestations were identified according to predefined criteria (table 1) as previously described.¹¹ To ensure completeness and accuracy of the diagnosis of ExRA manifestations, a subsample was reviewed by another author (ELM), with concordant results.¹¹ The criteria for severe disease manifestations (table 1; criteria 1–8) were based on those used in a previous study of the occurrence of ExRA performed in Malmö, Sweden,¹⁰ where they were found, as a group, to identify patients with poor survival. To assess the occurrence of rheumatoid pericarditis during the time before echocardiography was available, the criteria for pericarditis were modified as previously described.^4,11 In the present study further ExRA manifestations (table 1; criteria 9–17) were also investigated. The criteria for Sjögren’s syndrome (SS) have been used previously to identify cases of primary and secondary SS in epidemiological surveys.^36,37

The patients were followed up from 1 January 1955 until 31 December 2000, or until death or loss to follow up (median follow up 11.8 years; interquartile range 6.8–18.9). For patients who moved away from the area, the date of the last physical examination noted in the case record was used as the date of loss to follow up. The date of fulfilment of the criteria for ExRA was noted, as well as the date of death according to the death certificate. The cumulative incidence of ExRA was calculated and adjusted to the follow up time, which was calculated using the actuarial method. The cumulative incidence at 30 years’ follow up time was estimated.

The time to first presentation of ExRA was compared in patients with RA by decade of diagnosis (1955–64, 1965–74, 1975–84, and 1985–94). To ensure comparable follow up time in each subgroup in this analysis, patients with RA diagnosis during the period 1955–64 were followed up until 31 December 1970, those diagnosed in 1965–74 were followed up until 31 December 1980, those diagnosed in 1975–84 were followed up until 31 December 1990, and those diagnosed in 1985–94 were followed up until 31 December 2000. ExRA-free survival curves were estimated using the Kaplan-Meier product limit method,³⁸ and the subgroups were compared using the log rank test. This was done for all ExRA and for ExRA according to the Malmö criteria (table 1; criteria 1–8) separately.

Predictors

Possible ExRA risk factors were identified in case record reviews. The results of all rheumatoid factor (RF) and ANA tests were noted. Disability at diagnosis was measured by estimating the Steinbrocker functional class³⁹ within three months of RA diagnosis (defined as the time of fulfilment of the 1987 ACR criteria for RA³⁵). Data on the smoking history at the time of RA diagnosis were gathered, and the patients were classified as ever versus never smokers (table 2). Investigators collecting smoking and disability data were unaware of the hypothesis that this might influence the risk of ExRA manifestations.

The effects of age, sex, and other possible predictors on the development of ExRA were examined using Cox proportional hazards. Positive RF tests and positive ANA tests were examined in these models as time dependent covariates. As some patients never had an ANA test done, the presence of a completed ANA test was also entered into the models as a time dependent covariate. Variable selection for the multivariate analysis and possible interactions among the significant main effects examined were validated using bootstrap sampling.⁴⁰

RESULTS

The RA cohort consisted of 445 women and 164 men. The median age at diagnosis was 58.1 years. The median follow up period was 11.8 years (range 0.1–42.8; interquartile range 6.8–18.9). During the follow up period, 247 (40.6%) patients developed ExRA. The 30 year cumulative incidence of ExRA was 46.0% (standard error (SE) 3.4%). Table 3 shows the incidence of various ExRA manifestations. The most common ExRA manifestations were rheumatoid nodules (n=172; 30 year cumulative incidence 34.0 (3.4)%), secondary Sjögren’s syndrome (n=58; 30 year cumulative incidence 11.4 (2.6)%), and pulmonary fibrosis (n=34; 30 year cumulative incidence 6.8 (1.9)%). A subgroup of 78 patients (12.8%) fulfilled the Malmö criteria for severe ExRA (table 1, criteria 1–8)¹⁰ (30 year cumulative incidence 16.7 (3.2)%).

Overall, ExRA-free survival by decade of RA diagnosis decreased significantly during the period (p=0.001), indicating that ExRA occurred more frequently for each new decade. The increasing overall incidence of ExRA manifestations was mainly accounted for by an increase in the number of patients with rheumatoid nodules (data not shown).

For ExRA according to the Malmö criteria (ExRA Malmö) there was no significant trend (p=0.165) over the decades, and although there were fewer ExRA Malmö manifestations among patients with onset of RA in 1955–64, the survival curves of the three other subgroups were virtually identical up to 13 years of follow up (fig 1).

In the multivariate analysis, ExRA manifestations of any type (criteria 1–18) were predicted by the presence of a positive ANA test (relative risk (RR)=1.58; 95% confidence interval (CI) 1.12 to 2.21), RF (RR=1.56; 95% CI 1.19 to 2–04), smoking (RR=1.52; 95% CI 1.15 to 2.01), and Steinbrocker class III-IV at diagnosis (RR=1.42; 95% CI 1.04 to 1.94), but not by age and sex (table 4).

The main predictors of ExRA Malmö manifestations in the multivariate analysis were smoking (RR=2.94; 95% CI 1.68– to 5.13) and early disability (Steinbrocker class III-IV at diagnosis) (RR=2.45; 95% CI 1.51 to 4.00). The occurrence of ExRA Malmö manifestations also increased significantly with age at RA diagnosis (RR=1.019/year; 95% CI 1.00 to 1.038). When these factors were taken into account, male sex did not predict ExRA (table 5). Univariately, RF was weakly associated with ExRA Malmö, and this association did not withstand bootstrap validation (data not shown), indicating that smoking, early disability, and old age are stronger predictors, with a contribution which overwhelmed the contribution of RF.

DISCUSSION

In this community based cohort study of RA, which included patients with incident RA during the period 1955–94, there was no decrease over the decades in the incidence of extra-articular disease manifestations. ExRA was predicted by previous smoking and early disability. Age, RF, and ANA were significant predictors of ExRA manifestations overall, but not specifically of severe ExRA organ manifestations.

As has been discussed previously,¹¹ estimates of the incidence of ExRA manifestations tend to be lower in community and population based studies than in clinic based series, as the latter are selected for disease severity and complications. The occurrence of severe ExRA manifestations can be more reliably estimated, as such manifestations are less likely to go unnoticed. The recognition of mild disease manifestations, such as rheumatoid nodules, depends to a greater degree on the doctor’s inclination to record such observations in the case history. When using the criteria for Sjögren’s syndrome used in this and other studies,^36,37 the frequency of the diagnosis depends heavily on the availability of specific tests for the evaluation of possible xerostomia and keratoconjunctivitis sicca.

The apparently increased incidence of ExRA manifestations as a whole was mainly due to an increased number of patients with rheumatoid nodules in later decades. This may be due to detection bias (that is, increased likelihood of nodules being noticed and reported in case records) or possibly due to changes in other time dependent factors. Methotrexate has been shown to be associated with nodules in patients with RA,⁴¹ and the use of methotrexate in the treatment of RA in the USA expanded greatly during the relevant time period.

There are few data on the incidence of ExRA, and apart from a survey of systemic rheumatoid vasculitis in Norfolk, UK,⁴² no major studies have been published on incidence trends over time. Despite this, there is a widespread belief that severe ExRA manifestations are becoming less common. This study does not support such a concept. Our data are in accordance with those of Watts and coworkers,⁴² who did not find any major decrease in the incidence of rheumatoid vasculitis during the 1990s. However, our study is not large enough to evaluate with certainty every single type of ExRA manifestation, and, possibly, the incidence of some manifestations is decreasing, while that of others is stable or increasing. Also, as the study only includes patients diagnosed up to 1995, we do not know how subsequent changes in antirheumatic treatment in recent years, including the introduction of biological agents since 1995, may affect the occurrence of ExRA. Treatment with tumour necrosis factor blocking agents seems to have a major impact on disease outcome in many patients,^43,44 but their specific effect on ExRA has not yet been evaluated.

Smoking has been shown to be associated with RA and RF.^45–48 In a series of patients with early polyarthritis, smokers were more likely to develop rheumatoid nodules and less likely to develop erosions.⁴⁹ In longstanding disease, smoking may predict nodules and joint damage,^50–52 but not joint counts, pain, or erythrocyte sedimentation rate.⁵¹ Taken together with our findings on the relation between smoking and ExRA, and similar findings by others,³² this may indicate that smoking in some way drives the rheumatoid process towards extra-articular involvement. Vascular abnormalities may be important not just in rheumatoid vasculitis but also in other ExRA manifestations, such as pericarditis,⁵³ and rheumatoid nodules.⁵⁴ Smoking is a well known predictor of cardiovascular disease^55–57 and, possibly, smoking also has specific effects on blood vessels in patients with RA. This is of particular interest, as patients with RA,^18,58,59 in general, and patients with ExRA, in particular,¹⁰ have an increased mortality due to cardiovascular disease, and recent data indicate that patients with RA also have increased cardiovascular morbidity.⁶⁰

The association between smoking and ExRA was independent of RF, indicating that other mechanisms may play a part. Suggested pathogenic factors in ExRA include complement binding and activating circulating immune complexes,^61,62 clonally expanded CD4+ CD28(null) T cells⁶³ and generalised endothelial activation.⁶⁴ It is unknown to what extent these mechanisms are directly or indirectly influenced by smoking. Environmental factors such as smoking are likely to interact with the individual genetic background in shaping disease outcome.

Early disability was also associated with the later development of extra-articular disease. Although functional class may be difficult to assess in a retrospective case record review, these findings are in agreement with the generally held view that patients with a delayed diagnosis or a very aggressive disease usually have a complicated disease course. The inclusion of cervical myelopathy, which is a consequence of articular pathology, among ExRA manifestations (but not in the severe ExRA subset) may contribute to the association between disability and overall ExRA, although the impact of this is likely to be limited given the small number of patients with cervical myelopathy. Severe ExRA manifestations also tend to become more common with older age. RF and ANA were also associated with a moderately increased risk of ExRA, and this association seems to be stronger with mild disease manifestations. Male sex is not a strong predictor for ExRA.

The main limitation of this study is the retrospective method, which limits the analysis to the data which are found in the case record. At the same time, this method probably includes the vast majority of clinically important manifestations. Missing and inaccurate data may decrease the likelihood of correctly identifying important associations in such surveys. For example, it is impossible to state that other factors, such as diabetes or alcohol abuse, did not in any single case contribute to neuropathy in the investigated patients with RA. Such misclassification may potentially bias the association between ExRA and smoking, although this is unlikely to be of major importance, as we excluded unclear cases and the number of patients with neuropathy was small. The major strengths are the community based cohort and the data record linkage system, which enable virtually complete ascertainment of all cases in the area. The findings in this study are thus not likely to be explained by selection bias.

In conclusion, we have demonstrated that extra-articular disease manifestations are common and have not become less frequent among patients with RA during a period of 46 years. The main clinical predictors of extra-articular manifestations are smoking and early disability. As we have no evidence indicating that extra-articular RA is a vanishing phenomenon, it is of major importance to try to elucidate the underlying mechanisms behind these associations.