Early inflammatory polyarthritis: results from the Norfolk Arthritis Register with a review of the literature. II. Outcome at three years

This paper is the second in a series of two reviews which focus on the work of the Norfolk Arthritis Register (NOAR), a register of primary care‐based inception cohort patients with inflammatory polyarthritis (IP). The first paper examined genetic and environmental risk factors for the development of IP and of its subset, rheumatoid arthritis (RA) [91]. This paper summarizes the outcome of those patients recruited in 1990–1993 and followed for 3 yr in terms of remission of synovitis, the development of physical disability and radiological erosions. We have also examined prognostic indicators. Most of these results have not been published previously and this is the first time we have looked at all three outcomes in the same group of patients. The NOAR data are presented in the context of a detailed review of previous studies of IP and RA. The ability to predict disease outcome in patients newly presenting with inflammatory polyarthritis (IP) would be useful from a variety of perspectives [1], not least that of the patient. For example, there is mounting evidence that, to achieve their maximum potential, disease‐modifying anti‐rheumatic drugs (DMARDs) should be given as soon as possible after disease onset, before the disease process has become established and irreversible damage has occurred [2–9]. In order to target treatment appropriately and to avoid exposing patients with mild disease to potentially toxic drugs, it would be helpful to know which patients with IP are likely to go on to develop severe rheumatoid arthritis (RA). Many studies are confined to patients who satisfy the 1987 American College of Rheumatology (ACR) classification criteria for RA [10]. These criteria do not perform well in identifying patients with early IP who will later develop RA [11]. Nevertheless, they do give some consistency in patient selection between studies and so we have used them in this paper. Thus RA is a subset of IP. Table 1 lists all prospective studies which have included at least 50 patients with early (disease duration <3 yr) IP and which have investigated clinical and/or genetic predictors of outcome. Most are from European centres, probably because differences in the delivery of health care make it difficult to perform such studies in North America. Over 75% of these reports used the ACR criteria for RA to select patients for study. An accurate picture of disease outcome in IP is best obtained by recruiting patients soon after symptom onset and following them prospectively [68]. The studies listed in Table 1 were all based in hospital clinics, although some tried to recruit patients more widely from the local population [29, 33] or by invitation to attend early arthritis clinics [50, 54]. Compared with community‐based studies, hospital‐based studies are easier to establish and maintain, and offer greater diagnostic certainty. However, left censorship is a major problem in hospital‐based studies, particularly those that recruit using classification criteria for RA, because they exclude patients with mild disease and who enter remission early. The Norfolk Arthritis Register (NOAR) aims to recruit all patients with a new onset of IP who present to primary care [69]. Details of the referral and assessment procedures were presented in the previous paper in this series [91]. In this paper we examine disease outcome and predictors of prognosis among patients newly presenting with IP/RA, with emphasis on data from patients referred to NOAR. Prospective studies investigating clinical and/or genetic predictors of outcome in patients with early (<3 yr) IP aGenetic = including HLA‐DR alleles. b1958 ARA classification criteria. b1987 ACR classification criteria. cClinical diagnosis of RA. dEarly Rheumatoid Arthritis Study (ERAS). eUndifferentiated IP. Prospective studies investigating clinical and/or genetic predictors of outcome in patients with early (<3 yr) IP aGenetic = including HLA‐DR alleles. b1958 ARA classification criteria. b1987 ACR classification criteria. cClinical diagnosis of RA. dEarly Rheumatoid Arthritis Study (ERAS). eUndifferentiated IP. Between January 1990 and February 1993, 635 patients were referred to NOAR. Of these, 56 were later excluded because of an alternative diagnosis. Of the remaining 579 patients, 486 (84%) were followed for 3 yr, 24 (4%) died, 41 (7%) declined further participation and 28 (5%) were lost to follow‐up. The baseline characteristics of the 486 patients who completed 3 yr of follow‐up are shown in Table 2. Almost half (47%) satisfied at least four of the six ACR classification criteria (excluding X‐rays) at baseline. Over the 3‐yr follow‐up period, 357 (73%) patients were referred to hospital for their arthritis. Baseline characteristics of the NOAR study population used in this review (n = 486) aOnly 406 (84%) patients agreed to have blood taken. bAccording to 1987 ACR criteria applied in list format. cAccording to 1987 ACR criteria applied in classification tree format. Baseline characteristics of the NOAR study population used in this review (n = 486) aOnly 406 (84%) patients agreed to have blood taken. bAccording to 1987 ACR criteria applied in list format. cAccording to 1987 ACR criteria applied in classification tree format. We assessed outcome in three ways: (i) remission of synovitis; (ii) functional disability according to the Health Assessment Questionnaire (HAQ) score [70]; and (iii) radiological damage, using the grading system of Larsen et al. [71]. To aid comparison with previous studies, we also assessed outcome in the subset of 231 patients who satisfied the classification criteria for RA at baseline. The first systematic approach to the definition of remission in RA was carried out on behalf of the American Rheumatism Association (ARA) [72]. To fulfil the ARA criteria for remission, a patient must satisfy at least five of the following criteria for at least two consecutive months: morning stiffness lasting ≤ 15 min; no fatigue; no joint pain by history; no joint tenderness or pain on motion; no soft tissue swelling in joints or tendon sheaths; and erythrocyte sedimentation rate (ESR) <30 mm/h for men and <20 mm/h for women. Although these criteria have been used in other studies, the proportion of patients satisfying them is usually very low and there is still no universally accepted definition of remission [66]. We could not use the ARA remission criteria in the NOAR study [66] since we did not see patients in two consecutive months and did not measure the ESR or fatigue. Only 6% of the NOAR patients satisfied the remaining four criteria cross‐sectionally at 3 yr, and only 11% satisfied them cross‐sectionally at any time point during follow‐up. We therefore classified patients as being in ‘remission off treatment’ (as opposed to being in ‘treatment‐induced remission’) if they had no soft‐tissue joint swelling and had not been treated with DMARDs or steroids within the previous 3 months. The proportion of patients classified as being in remission at 1, 2 and 3 yr is shown in Table 3. Only 55 (11%) patients were in remission at 1, 2 and 3 yr and could be classified as having complete resolution of their disease. These remission rates are similar to those from other studies of IP. Woolf et al. [42] found that 30% of 88 patients with early synovitis were in remission at 5 yr. Nissilä et al. [27] reported a remission rate of 38% at 3 yr in 268 patients with definite RA or undifferentiated arthritis. Higher remission rates (54% at a mean of 6.9 yr) were reported in a North American study of 532 patients classified as having undifferentiated IP when first seen [57]. Patients classified as having RA have much lower remission rates [27, 33, 57]. In the ERAS study, only 15% patients with a clinical diagnosis of RA were in ‘sustained clinical remission’ after 3 yr [40]. We found a significant association between remission and younger age at disease onset: 62% of patients aged 16–25 yr at disease onset were in remission at 3 yr compared with only 30% of patients over 25 yr of age. Remission rates at 3 yr were slightly higher in males (37%) than in females (29%); [8% difference: 95% confidence interval (CI) 0, 18]. There was no association between symptom duration at presentation and future remission. Proportion of patients categorized as in remission off treatment at 1, 2 and 3 yr in the NOAR cohort aAccording to 1987 ACR criteria in list format. Proportion of patients categorized as in remission off treatment at 1, 2 and 3 yr in the NOAR cohort aAccording to 1987 ACR criteria in list format. We defined ‘moderate’ functional disability as an HAQ score of 1. Disability scores were highest at baseline, with an improvement at 1 yr (Fig. 1). A similar pattern has been observed in some previous studies [39, 58] but not in others [35]. In the NOAR cohort, the median HAQ score was 0.63 after 3 yr in the whole NOAR group, and 0.88 in patients classified as having RA. The proportion of patients with a HAQ score 1 was 36% in the whole group and 47% in patients classified as having RA. Similar results are found in most other studies of early RA [30, 58]. In the ERAS study, Young et al. [39] reported that 33% patients had an HAQ score >1 after 3 yr. An exception is the study by van der Heijde et al. [46], in which patients were reviewed every 4 weeks and had a mean HAQ score of only 0.35 at 2 yr. In the NOAR cohort, baseline HAQ scores were related to the number of swollen joints (Spearman's ρ, 0.52) and the number of tender joints (Spearman's ρ, 0.50). There was a linear relationship between age at disease onset and HAQ scores for all years (test for linear trend, P < 0.01). Women had significantly higher HAQ scores at all time points. Distribution of HAQ scores from baseline to 3 yr (n = 477). Three hundred and ninety (80%) of the 486 patients in the NOAR study satisfied the study criteria to be X‐rayed. X‐rays were available for 335 patients. The median time from symptom onset to the latest available X‐ray was 22 months. Due to the selection criteria, the 335 patients X‐rayed had slightly more severe disease than the whole NOAR cohort. For example, 207 (61%) satisfied classification criteria for RA at baseline and 38% had a positive test for rheumatoid factor (RF). One hundred and thirty‐seven (41%) of the patients X‐rayed had developed erosions, with a slightly higher prevalence (46%) among those who satisfied classification the criteria for RA at baseline. The median Larsen score was 3 (range 0–73) in all patients X‐rayed, 15 (range 2–73) in the subset with erosions, and 4 [range 0–73] in those who satisfied the classification criteria for RA. The prevalence of erosions was slightly higher in men (46%) than in women (39%) (difference of 7 percentage units, 95% CI −5, 19). There was a significant relationship between the development of erosions and older age at disease onset (test for linear trend, P = 0.01). The prevalence of erosions was also higher in patients with a longer symptom duration before initial presentation (test for linear trend, P = 0.03). The overlap between disease persistence, functional disability and radiological erosions in the NOAR group is illustrated in Fig. 2. Among the 335 patients X‐rayed, 74 (22%) had erosions, persistent synovitis and a HAQ score 1, whereas only 44 (13%) had none of these. Patients with erosions did not necessarily have disability and These two outcomes become more related in later disease of of disease outcome in 335 patients with IP referred to NOAR who had X‐rays taken. Many have tried to predict disease outcome in IP and RA using clinical obtained when the patient is first Although some predictors have been the results have in terms of the of and which factors are with an The results of studies using are shown in Table The for the differences in study are and are of hospital‐based RA patients are likely to predictors from studies of primary care‐based patients with IP because factors such as disease duration and disease hospital For example, older patients with IP be more likely to have persistent disease than younger patients. However, patients are seen in a hospital with RA, age have no on disease age be as a of erosions in a study based in primary but not in a hospital The of the study is also studies are more likely to that they are and using to clinical prognostic in patients with early IP = = X‐ray = score morning and = early morning 1, no swelling and no group swelling but no group erosions. over the first 2 yr. using to clinical prognostic in patients with early IP = = X‐ray = score morning and = early morning 1, no swelling and no group swelling but no group erosions. over the first 2 yr. Although most studies have investigated similar they have used of For example, radiological outcome be as the of erosions, an score of radiological damage, or the number of In outcome is reported at after disease Most studies have used such as linear and These results on the are categorized and the used to from the The of a be reported as a P but this is not very is to the ability of the to patients with the outcome using a The Study is the only other study to do this of patients are for this and the of the in the is than in the if a is the still not be to other patient The results of a are on the which are included as possible the baseline HAQ score is not as a possible it will not be Most studies have presented their results in terms of the of or the of the However, the most for the is the that an presenting with clinical will or will not develop the outcome in This be as positive and and The the proportion of patients with a positive test who develop the outcome in and the the proportion of patients with a test who do not develop the outcome. The and are by the prevalence of the outcome used and so the results will only be to with similar We used to baseline with outcome at 3 yr. of the results have been reported previously but not using this cohort or with 3 yr of follow‐up. The following at were as possible (i) age at symptom onset by initial disease duration in months by (ii) soft‐tissue swelling of the joint swelling of joints of the 1987 ACR classification criteria (excluding number of swollen joints by and number of tender joints by HAQ score and (iii) were for use of outcome was were and in a if they had a of were using a of the study and in the remaining The results presented to the of in the Table 5 the by the as predictors of an of the the of the and the and We obtained a for patients as having an HAQ score of This is higher than that obtained in most previous studies The was also for identifying patients who would develop radiological erosions studies which to predict remission of disease 33, also obtained We have also developed which be in clinical This was for erosions, since only two predictors and initial disease were To predict functional we used to the number of in the from five to The clinical of a on the in which it is to be In primary it is that the is to and which are such a is used to patients should be referred to are more than For example, if patients who have a test are not referred to it is that of them develop an outcome. However, if are being used to select patients for more the to be higher in order to give an The results in Table 5 that, whereas were low and were higher and However, these still be low to be data be used as a in but should not be used to clinical of to predict disease outcome at 3 yr among patients referred to NOAR in order of the of the association of to predict disease outcome at 3 yr among patients referred to NOAR in order of the of the association to patients according to of (i) and (ii) HAQ at 3 yr to patients according to of (i) and (ii) HAQ at 3 yr The relationship between age and disease outcome is The median age of patients referred to NOAR yr) was similar to that reported in hospital‐based studies 58]. This that age not referral to We found that younger patients were more likely to enter ‘remission off However, age had no on ‘treatment‐induced remission’ and and Young higher remission rates in younger patients, whereas et al. and et al. did We found that older patients were more likely to be according to the Although we previously reported that age was an of disability at 1 yr in the present it was not as an of the HAQ at 3 yr. have also reported that age is not with disability at 1 yr 2 yr or 3 yr However, age has been with disability at and yr because older arthritis have higher HAQ we found that older patients were more likely to develop radiological erosions. This is in with some previous reports but not with others In in the NOAR study, older patients had a In to other study we found that the prevalence of with age (test for linear trend, P = be that age was not as an of outcome in the because of its association with Many previous studies reported an prognosis in and none reported that outcome is in In the NOAR cohort, men were more likely to enter remission. This is in with previous reports Disability scores were higher in women. This be related to differences in ability between males and since van et al. reported that males were more likely to their functional ability according to the However, et al. that higher HAQ scores in women were by disease had on the development of radiological erosions or the Larsen Only two previous studies have reported that females have a radiological outcome There are a number of females have an is also possible that men with disease are more likely to be referred to hospital for to DMARDs in males and females and In the NOAR study, a higher proportion of men had been treated with DMARDs than women studies in RA that two younger men and older have a particularly The on early RA that, most men with RA go remission, those who do not have a outcome. We found that the duration of at presentation was an of future disability and radiological Only a other studies have examined initial disease duration as a of outcome. reported that patients with a longer disease duration did whereas others did not 33, There are a number of possible for the association between disease duration and outcome. patients who present have disease. However, remission rates at 3 yr were not by the time to patients who have had the disease for longer have had a longer time in which to for example, erosions. However, the relationship in the we time from disease onset to X‐ray as of the patients who present later to care have a This could be because in presentation is a for other factors with a such as factors and of or because patients have the for early treatment with have reported that patients with an disease onset are likely to care later than those with an have a prognosis 33, However, this was not the in other studies We the that the association is to the for early In the NOAR study, the of 3 months from symptom onset was because we disease duration the of which was 3 months. However, others have also that this of 3 months is A number of clinical have been as predictors of outcome of these are The baseline HAQ score was by the most of future disability in NOAR. A including only the HAQ score as an was as accurate as the which also included all other baseline The of functional ability at baseline has been reported by other studies using [29, 33, with exception Baseline HAQ was also the most of future work disability in NOAR for However, there was no association between baseline and radiological outcome. In the NOAR study, a positive test was an of an particularly radiological The for erosions in patients who were at baseline was However, as of patients also developed erosions, it is not to treatment in this but of previous studies investigating radiological outcome in early RA have the of a positive test 33, The association with functional disability was and was only for at a higher of previous studies have that is an of early disability [30, but others have that it is not 33, 58]. However, there is greater consistency for studies investigating later disability yr after all but a positive test as being In the NOAR study, we a positive test as being an of disease persistence, with a of compared with for a studies have also that patients are more likely to enter remission. Many studies have investigated genetic in prospective studies of early RA. Of studies investigating radiological erosions in early RA, four a positive association and three no association There is also the of on functional disability and disease We detailed for using based in 532 patients who were followed for 2 yr, of had X‐rays In we found that had no on the development of persistent disease and only a on the development of functional However, there was a association with the development of radiological erosions. Patients with at least of any were as likely to develop erosions as those who were the risk was CI Of we found that were only in patients who were and did not outcome in patients of has over that of in identifying patients likely to develop erosions. data do not the genetic of patients with early IP to those at risk of erosions. of erosions according to baseline and in the NOAR cohort of erosions according to baseline and in the NOAR cohort Norfolk was for the study because of the low of and the between primary and However, of Norfolk not be of the whole being slightly and with at the of the results therefore not be to other However, this is for all studies, since there are in characteristics within and between The criteria for the NOAR study were in order to avoid left This that results are not necessarily to patients seen in hospital clinics, although of NOAR patients were also referred to The proportion of patients with RA is lower in the NOAR study than in hospital‐based We RA as a subset of or the of the IP than IP as a the NOAR study was we to avoid exposing patients with mild who would be to have erosions, to However, with it would have been to X‐rays for all patients, which we are at 5 yr. of patients who did not satisfy the criteria to have X‐rays had erosions after 5 yr. A second is that we have used the results of X‐rays a median of 22 months after disease A number of potentially patients not have developed The data presented to the of early erosions. We did not measure of such as and at baseline. studies have that these are predictors of outcome Although on was not included in the present the results of a that in a number of not the outcome. This is in to some other studies on and are to be only three initial disease duration and baseline we could patients who would develop erosions and/or disability by 3 yr after has been that higher the order of be possible with the of genetic and data and by following health and and new disease such as and [68]. is that, for clinical and higher are to be RA is a outcome on the of and disease are it that it will be possible to predict outcome with a of in an patient. it is to the in clinical such as X‐ray damage and joint This the and the reported of possible In with other investigating early we are to develop classification criteria for early RA which could be used to consistency between the present time we that all patients who have IP with a duration of between 4 and weeks should be as potentially having RA and for if they are or have a baseline HAQ score of a patient has had persistent disease for weeks or more an outcome is more likely and the for is This study was by the Arthritis We are to the and hospital of the Health for their in patients for study. In we Norfolk We also the and previous NOAR and for their van and We of the and