Abstract

Background. It is 40 yr since the last age‐ and sex‐specific estimates of the prevalence of rheumatoid arthritis (RA) for the UK were published. Since then the classification criteria for RA have been revised and there has been evidence of a fall in the incidence of RA, especially in women.

Objectives. To estimate the age‐ and sex‐specific point prevalence of RA (defined as fulfilment of a modification of the 1987 ACR classification criteria for RA on the day of assessment). The estimate was made in the primary care setting in Norfolk, UK.

Methods. A stratified random sample was drawn from seven age and gender bands. The 7050 individuals selected were mailed a screening questionnaire. Positive responders were invited to attend for a clinical examination. The sample was matched against the names in the Norfolk Arthritis Register (NOAR), a register of incident cases of inflammatory polyarthritis which has been in existence since 1990.

Results. The overall response rate was 82%. Sixty‐six cases of RA were identified. Extrapolated to the population of the UK, the overall minimum prevalence of RA is 1.16% in women and 0.44% in men. A number of incident cases of RA previously notified to NOAR were not identified as cases in the survey because they had entered into treatment‐induced remission. In addition, some cases who failed to attend for examination had significant disability. These prevalence figures are therefore an underestimate.

Conclusions. The prevalence of RA in women, but not in men, in the UK may have fallen since the 1950s.

It is 40 yr since the last age‐ and sex‐specific estimates of the prevalence of rheumatoid arthritis (RA) in the UK were published [1]. Since that time, a new set of classification criteria has been introduced [2]. In addition, there is some evidence that there has been a decline in the incidence and prevalence of RA both in the UK [35] and elsewhere [69]. Data from the third and fourth national morbidity surveys in the UK [10, 11] show a decrease in the prevalence of ‘RA and allied conditions’ (which includes ankylosing spondylitis) between 1981 and 1991. These national morbidity surveys are based on visits to a general practitioner (GP) during a 1‐yr period for a GP diagnosis of ‘RA or a related condition’. During this period, the overall prevalence of RA appears to have fallen by 31% in women and 19% in men (although the methods used in the two surveys were not exactly the same) (Fig. 1). This apparent fall in the prevalence of RA could be real or could be the consequence of a fall in consultation rates by people with RA between the two surveys. These various UK estimates either use data from primary care, which do not use any standardized diagnostic criteria [10, 11], or are confined to women [5]. As age‐ and sex‐specific estimates are used widely for the planning of current and future health‐care provision, we conducted a two‐stage population‐based survey to provide contemporary estimates of the prevalence of RA for seven age–sex bands (four in women and three in men).

This prevalence study was conducted in the area in which the Norfolk Arthritis Register (NOAR) is set. NOAR is a primary‐care‐based inception cohort of adults with inflammatory polyarthritis (of which RA is a subset) with onset since 1990 [12]. The study reported in this paper presents a number of additional opportunities for understanding the occurrence of RA. First, it offers the opportunity to establish the proportion of previously identified incident cases of RA that is identifiable in a subsequent cross‐sectional prevalence study. This will provide valuable information on remission rates—information which is essential for modelling the burden of disease. Secondly, it offers an opportunity to assess the completeness of notification to NOAR (by identifying cases of RA which should have been known to NOAR) and adjusting our incidence estimates accordingly. It also offers a further chance to evaluate the 1987 ACR classification criteria for RA [2] in a population.

It is generally believed that the occurrence of RA does not vary amongst people of Northern European origin (whether living in Europe or North America) [13]. However, there is evidence within the UK that the prognosis of RA may be influenced by social deprivation. [14, 15]. We therefore explored the effect of area of residence on the estimated prevalence (which is influenced by disease duration as well as incidence) of RA within Norfolk.

Methods

We conducted a two‐stage population‐based survey. The first stage was a screening questionnaire and the second stage included a clinical examination of positive responders to the screening questionnaire. The study was approved by the Norwich Local Research Ethics Committee.

Sampling

Eleven computerized general practices (GenP) reflecting urban, rural and coastal populations within the former Norwich Health Authority participated in the study. The total population covered by these practices was approximately 60 000. According to National Census data, <1% of the population of Norfolk belong to ethnic minorities.

The registers of adult patients from each GenP were divided into four age bands (16–44, 45–64, 65–74 and 75+ yr) for each gender. Data from NOAR show that the incidence of RA in men aged 16–44 yr is very low [2.9 per 100 000; 95% confidence interval (CI) 0.8, 7.6] [12]. Therefore, men aged 16–44 yr were not included in the study because of the large sample size that would have been required to provide robust estimates of the prevalence of RA in this group. The sample size of each of the remaining seven age–sex bands (Table 1) was based on being able to show that a prevalence of RA outside the range of 0.5–2.0 times that reported by Lawrence [1] was significantly different from Lawrence's data, and allowed for a 25% non‐response rate. The sample size drawn from each practice was weighted according to the size of the practice. Within each GenP, the subjects in each of the seven age–sex bands were sorted by date of birth. We then selected every nth subject, where n is the number of people in that age–sex band divided by the sample size in that age–sex band required from that GenP. GPs were shown the list of selected subjects and could request that individuals should be excluded if they felt, for example, that the survey would cause severe psychological distress or the individual was terminally ill. Subjects excluded in this way were replaced by the next person on the list (sorted by date of birth).

Fig. 1. 

Trend in the age‐specific prevalence of RA (annual consulting rate per 10 000) for adult males and females from the last two morbidity surveys by the Royal College of General Practitioners (1981–82 and 1991–92) [10, 11].

Fig. 1. 

Trend in the age‐specific prevalence of RA (annual consulting rate per 10 000) for adult males and females from the last two morbidity surveys by the Royal College of General Practitioners (1981–82 and 1991–92) [10, 11].

Table 1. 

Sample structure, response rates and age‐ and sex‐specific prevalences of RA using various assumptions

 Female age groups (yr)
 

 

 

 
Male age groups (yr)
 

 

 

 

 
16–44
 
45–64
 
65–74
 
75+
 
16–44
 
45–64
 
65–74
 
75+
 
Stratum sample 2799 869 439 414 – 1279 724 526 
Dead/not at address 283 31 13 – 78 21 22 
True sample size 2516 838 430 401 – 1201 703 504 
Response rate (%) 79.0 89.5 89.8 79.1 – 80.4 86.2 81.5 
Number of positive 173 183 145 134 – 170 127 93 
   responders         
Proportion assessed 0.71 0.87 0.86 0.70 – 0.76 0.87 0.84 
Number with RA 14 11 12 – 11 
Minimum RA prevalencea (%) 0.12 1.67 2.56 2.99 0.02e 0.58 1.14 2.18 
   (95% CI) (0.03, 0.35) (0.91, 2.80) (1.28, 4.58) (1.55, 5.23)  (0.23, 1.20) (0.49, 2.24) (1.09, 3.90) 
RA prevalenceb (%) 0.15 1.87 2.84 3.77  0.72 1.32 2.70 
   (95% CI) (0.03, 0.44) (1.02, 3.13) (1.42, 5.07) (1.95, 6.59)  (0.29, 1.49) (0.57, 2.60) (1.35, 4.84) 
RA prevalencec (%) 0.21 2.15 3.31 5.36  0.94 1.51 3.08 
   (95% CI) (0.07, 0.66) (1.28, 3.61) (1.85, 5.92) (3.10, 9.27)  (0.45, 1.98) (0.76, 3.00) (1.73, 5.51) 
 Female age groups (yr)
 

 

 

 
Male age groups (yr)
 

 

 

 

 
16–44
 
45–64
 
65–74
 
75+
 
16–44
 
45–64
 
65–74
 
75+
 
Stratum sample 2799 869 439 414 – 1279 724 526 
Dead/not at address 283 31 13 – 78 21 22 
True sample size 2516 838 430 401 – 1201 703 504 
Response rate (%) 79.0 89.5 89.8 79.1 – 80.4 86.2 81.5 
Number of positive 173 183 145 134 – 170 127 93 
   responders         
Proportion assessed 0.71 0.87 0.86 0.70 – 0.76 0.87 0.84 
Number with RA 14 11 12 – 11 
Minimum RA prevalencea (%) 0.12 1.67 2.56 2.99 0.02e 0.58 1.14 2.18 
   (95% CI) (0.03, 0.35) (0.91, 2.80) (1.28, 4.58) (1.55, 5.23)  (0.23, 1.20) (0.49, 2.24) (1.09, 3.90) 
RA prevalenceb (%) 0.15 1.87 2.84 3.77  0.72 1.32 2.70 
   (95% CI) (0.03, 0.44) (1.02, 3.13) (1.42, 5.07) (1.95, 6.59)  (0.29, 1.49) (0.57, 2.60) (1.35, 4.84) 
RA prevalencec (%) 0.21 2.15 3.31 5.36  0.94 1.51 3.08 
   (95% CI) (0.07, 0.66) (1.28, 3.61) (1.85, 5.92) (3.10, 9.27)  (0.45, 1.98) (0.76, 3.00) (1.73, 5.51) 

aPrevalence of RA assuming that none of the non‐responders to the screening questionnaire and examination had RA. This gives a minimum prevalence but seems the most robust assumption.

bPrevalence of RA assuming that non‐responders have same rate of RA as responders.

cPrevalence of RA assuming that those who declined examination had the same rate of RA as those who agreed to be examined.

eMales aged 16–44 yr were not included in the survey. This prevalence figure was calculated by assuming that the female:male ratio of RA in the 16–44 yr age group is the same as that observed in NOAR for the incidence of RA in the same age group (i.e. 6.45:1).

The Townsend score

The Townsend score is a composite index of material deprivation based on the following four variables from the 1991 National Census: percentage of unemployed males; percentage of households which do not own a car; percentage of households which are not owner‐occupied; and percentage of overcrowded households [16]. The enumeration district is the smallest geographical unit to which Townsend scores can be assigned and has a few hundred residents. Townsend scores for the whole of England and Wales have been divided into quintiles. Each subject in this prevalence study was assigned to one of these quintiles on the basis of the Townsend score of his/her enumeration district (determined from the postcode).

Screening

The GenPs were studied sequentially. The selected subjects from each GenP were sent a validated screening questionnaire [17], a covering letter signed by their GP, and a prepaid envelope. The screening questionnaire included questions on demography, joint complaints and the British version of the Stanford Health Assessment Questionnaire (HAQ) [18, 19]. Two weeks after the date of the first mailing, non‐responders were sent a reminder postcard. Two weeks after the second mailing the remaining non‐responders were sent a second letter, a screening questionnaire and a prepaid envelope.

Positive responders were individuals who reported ever having swelling of two or more joints (excluding the ankles) lasting for 4 or more weeks, or who had ever been told by a doctor that they had RA. They were invited to attend a clinic at the GenP surgery for examination by a research nurse. Subjects who were unable to attend the clinic were offered a home visit. The research nurse interviewed the subject and completed a second, more detailed questionnaire. The subject's joints were examined for tenderness, soft tissue swelling and deformity. A joint was said to be ‘deformed’ if the subject could not adopt the anatomical position or had lost one‐third or more of the normal range of movements. Subjects were also examined for subcutaneous nodules. On the basis of this examination, a modified version of the 1987 ACR criteria was applied (Table 2). A similar modification [20], which required access to the patient's records, had been validated. This modification (Table 2) was made for use in a population survey without access to patient records. It was validated against a rheumatologist's diagnosis based on the study documentation. Those who satisfied two or more of the modified ACR criteria were asked to give a blood sample for rheumatoid factor analysis (unless there was documented evidence of previous seropositivity from our laboratory) and to have X‐rays of the hands and feet (unless X‐rays taken within the last 2 yr were available). The questionnaires, blood samples and X‐rays were returned to Manchester for data entry and analysis. X‐rays were read for erosions by two observers using Larsen's method [21]. A score of two or more in any joint indicated that the subject had erosions. A third observer arbitrated if there was disagreement about the presence of erosions.

In addition, 251 individuals who were negative responders were invited to attend for clinical assessment. They were selected at random with the aim of examining approximately 21 from each GenP (three from each age–sex band).

Table 2. 

The American College of Rheumatology Revised Classification Criteria for Rheumatoid Arthritis [2] (as applied in this study)

1. Morning stiffness Morning stiffness in and around the joints lasting for at least 1 h before maximal 
    improvement (ever) 
2. Arthritis of 3 or more joint areas Swelling and tenderness, or deformity (reduction in the range of movement by at least 
    one‐third or inability to adopt the anatomical position) of at least three joint areas (out of 
    14 possible areas: right or left proximal interphalangeal (PIP), metacarpophalangeal 
    (MCP), wrist, elbow, knee, ankle, metatarsophalangeal) 
3. Arthritis of hand joints Swelling and tenderness, or deformity (as defined above) of at least one hand joint area: 
    wrist, PIP or MCP 
4. Symmetrical arthritis More than 50% of the joint areas involved (as defined in criterion 2) have bilateral 
    changes 
5. Rheumatoid nodules Subcutaneous nodules over bony prominences or extensor surfaces, or in juxta‐articular 
    regions 
6. Serum rheumatoid factor Demonstration of rheumatoid factor by latex agglutination at a titre >1 in 40 
7. Radiographic changes Erosions detected by two independent observers on posterior–anterior hand or foot X‐raysa 
1. Morning stiffness Morning stiffness in and around the joints lasting for at least 1 h before maximal 
    improvement (ever) 
2. Arthritis of 3 or more joint areas Swelling and tenderness, or deformity (reduction in the range of movement by at least 
    one‐third or inability to adopt the anatomical position) of at least three joint areas (out of 
    14 possible areas: right or left proximal interphalangeal (PIP), metacarpophalangeal 
    (MCP), wrist, elbow, knee, ankle, metatarsophalangeal) 
3. Arthritis of hand joints Swelling and tenderness, or deformity (as defined above) of at least one hand joint area: 
    wrist, PIP or MCP 
4. Symmetrical arthritis More than 50% of the joint areas involved (as defined in criterion 2) have bilateral 
    changes 
5. Rheumatoid nodules Subcutaneous nodules over bony prominences or extensor surfaces, or in juxta‐articular 
    regions 
6. Serum rheumatoid factor Demonstration of rheumatoid factor by latex agglutination at a titre >1 in 40 
7. Radiographic changes Erosions detected by two independent observers on posterior–anterior hand or foot X‐raysa 

aAlthough X‐rays of both the hands and feet were read, no patients satisfied the ACR criteria because they had erosions in the feet but not the hands.

Statistics

Prevalence in each stratum was calculated by dividing the number of cases of RA by the true size of the stratum sample (the number of questionnaires mailed out minus the number of individuals found to have died or moved away). The 95% confidence intervals (CI) were based on the Poisson distribution of the number of cases. The association between material deprivation (Townsend score) and response was assessed using the ‘svylogit’ (i.e. logistic regression for survey data) command in STATA version 6.0 [22]. This method of analysis takes account of between‐stratum differences in the sampling fraction by using probability weights.

Results

Age‐ and sex‐specific prevalence of RA

On average, each GenP asked for 20 individuals to be excluded and replaced. Half of the substitutions were because the individual was known to have died. Most of the remaining substitutions were due to dementia, severe mental illness or terminal illness. Questionnaires were then mailed to 7050 people (Fig. 2). Four hundred and fifty‐seven subjects were assumed to have died or moved away, either because the questionnaire was returned by the Post Office or the subject was not listed in the electoral register. Replies were received from 5424 individuals (an overall adjusted response rate of 82%). Response rates were higher in those aged 45–74 yr than in those older or younger than this. Response rates were slightly lower in men than in women of the same age group (Table 1). Overall, a successful match of postcode to enumeration district could be made for 6502 (98.6%) subjects. Response rates were highest from the most affluent areas (Table 3).

The 1025 positive responders were invited to attend for a clinical assessment and 818 (80%) agreed. Sixty‐six subjects satisfied at least four of the seven ACR criteria when applied as listed in Table 2. There was no association between the prevalence of RA and the Townsend score (P=0.514) (Table 3). The age‐ and sex‐specific prevalences of RA (together with their 95% CI) are shown in Table 1. The study did not generate estimates for men aged 16–44 yr. If we assume that the female:male ratio of RA in the 16‐ to 44‐yr‐old age group is the same as that observed in the NOAR incidence estimates [12] (i.e. 6.45:1), then the prevalence of RA in men aged 16–44 yr would be 0.02%. Extrapolating these data to the population of the UK yields an estimate of the overall prevalence of RA in adults of 0.81% (1.16% for women and 0.44% for men, a female:male ratio of 2.7:1). On the basis of these figures, there were around 386 600 people in the UK with adult‐onset RA in the year 2000.

The 1987 ACR criteria for RA do not have any exclusions. Four individuals who satisfied the ACR criteria in this survey were known to have other inflammatory conditions which may have accounted for the findings: one case each of scleroderma, Wegener's granulomatosis, ankylosing spondylitis and juvenile‐onset chronic arthritis. In addition, many people in the older age groups had been given a clinical diagnosis of osteoarthritis. None of these people was excluded from our analysis. In addition to the modified ACR criteria in list format (Table 2) we also applied the ACR criteria in their original list and classification tree format [2]. Substantially fewer cases were identified using the list (4/7 criteria) format and somewhat fewer by the classification tree format (Table 4).

Fig. 2. 

Study subjects.

Fig. 2. 

Study subjects.

Table 3. 

Crude and age‐ and sex‐adjusted response rates by Townsend quintile (England and Wales) and association between Townsend quintile and RA status (adjusted for age group and sex)

  Response rate
 

 
Association between deprivation and RA
 

 

 

 
Townsend quintile % of Norfolk   Number of    
(England and Wales)
 
sample
 
Crude rate
 
Adjusted ratea
 
RA cases
 
Adjusted OR
 
95% CI
 
P (Z)c
 
1 (most affluent) 23.0 85.5% 85.4% 16 1.00 – – 
28.3 82.9% 82.8% 18 1.13 0.55, 2.33 0.733 
26.5 82.1% 82.2% 15 0.88 0.42, 1.86 0.738 
4 and 5 (most deprived)b 22.3 78.1% 78.2% 17 1.50 0.72, 3.09 0.277 
Total 100 82.2% – 66 – – 0.514 
  Response rate
 

 
Association between deprivation and RA
 

 

 

 
Townsend quintile % of Norfolk   Number of    
(England and Wales)
 
sample
 
Crude rate
 
Adjusted ratea
 
RA cases
 
Adjusted OR
 
95% CI
 
P (Z)c
 
1 (most affluent) 23.0 85.5% 85.4% 16 1.00 – – 
28.3 82.9% 82.8% 18 1.13 0.55, 2.33 0.733 
26.5 82.1% 82.2% 15 0.88 0.42, 1.86 0.738 
4 and 5 (most deprived)b 22.3 78.1% 78.2% 17 1.50 0.72, 3.09 0.277 
Total 100 82.2% – 66 – – 0.514 

aAdjusted for age and sex.

bTownsend quintiles 4 and 5 have been combined as there were only three RA cases in quintile 5.

cAdjusted Wald test.

Table 4. 

Number of cases of RA identified by the original ACR criteria and our modified criteria in list and tree format

  Original criteria
 

 
Age group Modified criteria   
(yr)
 
List
 
List
 
Classification tree
 
Females    
   16–44 
   45–64 14 12 
   65–74 11 11 
   75+ 12 
Males    
   45–64 
   65–74 
   75+ 11 
Total 66 35 57 
  Original criteria
 

 
Age group Modified criteria   
(yr)
 
List
 
List
 
Classification tree
 
Females    
   16–44 
   45–64 14 12 
   65–74 11 11 
   75+ 12 
Males    
   45–64 
   65–74 
   75+ 11 
Total 66 35 57 

Sensitivity analysis

We used the number of questionnaires mailed minus the number of people assumed to have died or moved away as the true denominator in our prevalence estimates. In effect, this method assumes that none of the non‐responders to the screening questionnaire or to the examination phase had RA. These estimates, therefore, give the ‘minimum’ RA prevalence in the population surveyed (Table 1). A number of other assumptions could have been used. For example, we could have assumed that all the non‐responders had RA (which would have given a maximum prevalence), but this seems intrinsically extremely unlikely. A more plausible assumption is that the proportion who had RA was the same among non‐responders as among responders to the screening questionnaire. If this was the case, one would expect the same prevalence of RA amongst those who responded to the second and third mailings (who would have been labelled as non‐responders if the reminders had not been sent) as in those who responded to the first mailing. In fact the proportion with RA fell progressively with each mailing: the odds ratio (OR) of screening positive in the second vs the first mailing was 0.79 (95% CI 0.67, 0.95) and the OR of screening positive in the third vs the first mailing was 0.52 (95% CI 0.43, 0.63). This favours the assumption that all remaining non‐responders did not have RA (although it is likely that some cases remain undetected this way). Alternatively, we may assume that non‐responders to the screening questionnaire did not have RA, but that those who declined examination had the same prevalence of RA as those who were examined (Table 1). One way to look at this would be to compare the HAQ scores from the screening questionnaire for those positive responders who agreed to an examination and those who did not agree. If the prevalence of RA were the same in those who declined examination as in those who agreed, one would expect the HAQ scores in the two groups to be similar. In fact, those who were examined had higher HAQ scores than those who declined examination. Sixty‐two per cent of those who were examined and 49% of those who declined examination had an HAQ score >0. Those who had an HAQ score between 0.1 and 2.0 were significantly more likely to agree to an examination than those with an HAQ score of 0, whereas those with an HAQ score >2.0 were equally likely to attend as those with an HAQ score of 0 (Table 5). Again, this favours the assumption that those who were not examined did not have RA (although some very disabled cases of RA will have been missed).

Table 5. 

Relationship between HAQ score and agreeing to be examined

HAQ score (from Odds ratio of agreeing   
screening questionnaire)
 
to be examined
 
95% CI
 
P
 
– – 
0.1–0.5 1.54 1.04, 2.29 0.033 
0.6–1.0 2.46 1.41, 4.27 0.001 
1.1–1.5 2.23 1.14, 4.37 0.019 
1.6–2.0 2.02 1.05, 3.87 0.035 
2.1–2.5 1.08 0.54, 2.15 0.823 
2.6–3.0 1.01 0.44, 2.32 0.984 
HAQ score (from Odds ratio of agreeing   
screening questionnaire)
 
to be examined
 
95% CI
 
P
 
– – 
0.1–0.5 1.54 1.04, 2.29 0.033 
0.6–1.0 2.46 1.41, 4.27 0.001 
1.1–1.5 2.23 1.14, 4.37 0.019 
1.6–2.0 2.02 1.05, 3.87 0.035 
2.1–2.5 1.08 0.54, 2.15 0.823 
2.6–3.0 1.01 0.44, 2.32 0.984 

Negative responders

Two hundred and twelve (84%) of the negative responders who were invited for examination agreed to attend (median number from each practice=20, range 16–30). They included 120 women and the median age was 67.5 yr (range 20–94 yr). None of these people satisfied the ACR criteria. Thus, the screening questionnaire had 100% sensitivity in detecting RA, with a lower 95% CI of 98.3% based on a binomial distribution.

RA in remission

The names selected for inclusion in this prevalence survey were matched against the register of people who had ever been notified to NOAR. Patients are notified to NOAR from primary or secondary care if they have soft tissue swelling of two or more joints lasting for 4 or more weeks with an onset since 1 January 1990 [12]. Twenty‐four matches were found, 13 of whom had ever satisfied the 1987 ACR criteria (as part of the NOAR assessments) when applied cumulatively (i.e. if a single criterion was satisfied it was carried forward as positive to all subsequent assessments). All 13 were positive responders. However, only five of the 11 patients who attended for an examination satisfied the ACR criteria at the time of the assessment. All of the remaining six were in remission on treatment, with few or no joint abnormalities on examination. There were no cases who had previously satisfied the ACR criteria (as part of the NOAR assessments) who had gone into spontaneous remission.

Discussion

These are the first age‐ and sex‐specific estimates of the prevalence of RA in the UK since the seminal paper by Lawrence published 40 yr ago [1]. Lawrence combined the results of two population surveys conducted in Leigh, Lancashire (an urban area) and Wensleydale, Yorkshire (a rural area). The total sample size was 2590 adults and 2234 (86%) were examined. He found 24 cases who satisfied the recently published 1958 ARA criteria for definite RA [23]. He took as the denominator those individuals who attended for an examination (thus assuming the same rate of RA in those who were and those who were not examined) (Table 6). However, Lawrence commented that ‘those who refuse to co‐operate are generally persons without rheumatic symptoms’. Thus, it would be more appropriate to calculate the prevalence figures from his study using the total sample as the denominator (Table 6), and this would make the results more comparable with our study. Although the classification criteria used were different from those used in our study, it appears that the prevalence of RA has altered differently in men and women in the last 50 yr (Table 6). However, because of the small number of cases identified (in particular in Lawrence's study), we had very low statistical power to make either within‐ or between‐stratum comparisons between the two surveys. Thus, it is unsurprising that logistic regression analyses (adjusted for age group) and using the study identifier as a dummy variable failed to show that the overall change in prevalence was statistically significant, for either women (OR=0.80, 95% CI 0.46, 1.40, P=0.430) or men (OR=1.28, 95% CI 0.45, 3.65, P=0.640). Nevertheless, whereas the prevalence of RA has fallen in all age groups except those aged over 75 yr for women, it has risen in the two age groups that can be compared for men. The only other cross‐sectional survey of the prevalence of RA in the UK using similar methods was conducted in 1990 and was confined to women aged 45–64 yr [5]. The authors reported a prevalence of 0.87% using the original ACR criteria, which is significantly lower than the prevalence in that stratum (2.14%) reported by Lawrence (P=0.037).

Our findings are compatible with a number of previous studies that have shown a fall in the incidence of RA in women since the 1960s [4, 24]. This fall has been attributed to a protective effect of the oral contraceptive pill (OCP) [25] or some other factor related to the use of the OCP. A NOAR case–control study of incident RA also showed this protective effect [26]. It is possible that use of the OCP reached saturation levels in the late 1980s and that is why the prevalence of RA in women has now stabilized. Women currently aged over 75 yr are too old to have taken the OCP and this may be why the prevalence of RA has not changed in this age group.

There have only been a few studies of the prevalence of RA in mainland Europe since the introduction of the 1987 ACR criteria. All report a lower prevalence than identified in the present study. In 1994 the prevalence of RA in Oslo, based on a register of attendance at a hospital rheumatology clinic and satisfying the 1987 ACR criteria, was 0.67% for women and 0.19% for men aged 20–79 yr [27]. A Swedish study based on a two‐stage population screening survey similar to that used in our study (but confined to adults aged 20–75 yr) estimated a total population prevalence for this age group of 0.51% [28]. A two‐stage population screening survey in Brittany (the first phase being conducted by telephone) found a prevalence of 0.74% in women and 0.26% in men [29]. RA appears to be less frequent in southern than northern Europe. Rates reported from north‐west Greece are 0.45% for women and 0.19% for men (aged 16+ yr) [30], those in northern Italy are 0.51% for women and 0.13% for men [31], and those in Yugoslavia are 0.29% for women and 0.09% for men [32]. Some of these differences may be explained by study design or by different age–sex structures of local populations.

On the other hand, our figures are somewhat lower than the most recent estimates from Rochester, Minnesota [8] (1.40% for women and 0.74% for men). These US estimates were based on individuals who had attended for medical care and who had ever satisfied the ACR criteria prior to the ‘prevalence date’ (31 December 1985). Thus, these US results provide an estimate of cumulative prevalence, whereas our estimates provide a point prevalence of people with clinical and/or radiological evidence of current or past RA. The inclusion of joint deformity in our interpretation of the ACR criteria (Table 2) enabled us to capture individuals who were in remission but had some accumulated damage from their RA [20]. However, we will have missed individuals who were in spontaneous or treatment‐induced remission and had no sequelae of previous disease activity. Matching our prevalence study against the NOAR register showed that six out of 13 patients who had previously satisfied the ACR criteria (as part of the NOAR assessments) were on disease‐modifying treatment and failed to satisfy the criteria on the day of the prevalence survey, but none had entered spontaneous remission. All subjects in NOAR have less than a 10‐yr history of arthritis. In order to adjust the prevalence estimates for cases of RA which are in remission, we also need to know what proportion of RA cases of more than 10 yr duration might have been missed. The proportion is likely to be lower in this group due to cumulative joint deformity and less effective treatment strategies in the past. Amongst the 735 subjects assessed who failed to satisfy the ACR criteria on the day of the examination and who were not on NOAR, a further eight subjects were on a disease‐modifying drug (methotrexate, sulphasalazine, an antimalarial or azathioprine). One had lupus and another ankylosing spondylitis, but the remaining six had a consultant diagnosis of RA or undifferentiated inflammatory polyarthritis. It is likely that some of these individuals would have satisfied the ACR criteria at some point in the past. We were not, therefore, able to adjust accurately our estimates of RA prevalence for cases missed due to treatment‐induced remission. Nevertheless, it is important to consider these individuals in any RA prevalence estimates that are being used for health service planning, as these people represent a cost to the health service. We did not include them in the present study as they did not satisfy our criteria for a prevalent case and, while these excluded individuals were a cost to the health service, some of the subjects who were included in our estimates had never sought medical help for their arthritis. With improving treatment of RA, it is likely that there will be an increasing number of people who have previously been diagnosed as having RA but, as a result of successful treatment, have no evidence of the disease on a prevalence date.

In conclusion, we believe that these estimates are robust and generalizable, and provide the best currently available figures on the burden of RA in the UK population.

Correspondence to: D. Symmons.

Table 6. 

A comparison of the estimates of the prevalence of RA in Lawrence's paper [1] and the present study

 Lawrence's study
 

 

 

 

 
Norfolk study
 

 

 

 

 
Age group (yr)
 
Total sample (a)
 
Number examined (b)*
 
Number of cases
 
Prevalence using ba as denominator (%)
 
Prevalence using a as denominator (%)
 
Sample size
 
Number of cases
 
Prevalence (%)
 
Comparison of rates using ba as denominator compared with Norfolk 2000 estimates
 
Comparison of rates using a as denominator compared with Norfolk 2000 estimates
 
Females           
   16–44 640 568 0.18 0.16 2516 0.12 −32% −24% 
   45–64 468 406 10 2.46 2.14 838 14 1.67 −32% −22% 
   65–74 167 139 4.32 3.59 430 10 2.56 −41% −29% 
   75+ 72 61 3.28 2.78 401 11 2.99 −0.1% +8% 
Males           
   45–64 435 378 0.53 0.46 1201 0.58 +10% +27% 
   65+ 165 145 1.38 1.21 1207 19 1.57 +14% +30% 
 Lawrence's study
 

 

 

 

 
Norfolk study
 

 

 

 

 
Age group (yr)
 
Total sample (a)
 
Number examined (b)*
 
Number of cases
 
Prevalence using ba as denominator (%)
 
Prevalence using a as denominator (%)
 
Sample size
 
Number of cases
 
Prevalence (%)
 
Comparison of rates using ba as denominator compared with Norfolk 2000 estimates
 
Comparison of rates using a as denominator compared with Norfolk 2000 estimates
 
Females           
   16–44 640 568 0.18 0.16 2516 0.12 −32% −24% 
   45–64 468 406 10 2.46 2.14 838 14 1.67 −32% −22% 
   65–74 167 139 4.32 3.59 430 10 2.56 −41% −29% 
   75+ 72 61 3.28 2.78 401 11 2.99 −0.1% +8% 
Males           
   45–64 435 378 0.53 0.46 1201 0.58 +10% +27% 
   65+ 165 145 1.38 1.21 1207 19 1.57 +14% +30% 

aMethod used in original paper.

We are grateful to Lynn Massingham and Teresa Kempson for their work in examining the subjects and co‐ordinating the project. We thank Drs Marwan Bukhari and Beverley Harrison for their assistance in reading the X‐rays. We thank the GP practices for access to their patient lists and the use of their premises for clinics. The study was funded from the core support by the Arthritis Research Campaign, UK, to the ARC Epidemiology Unit.

References

1
Lawrence JS. The prevalence of rheumatoid arthritis.
Ann Rheum Dis
 
1961
;
20
:
11
–7.
2
Arnett FC, Edworthy SM, Bloch DA et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis.
Arthritis Rheum
 
1988
;
31
:
315
–24.
3
Hochberg MC. Changes in the incidence and prevalence of rheumatoid arthritis in England and Wales 1970–1982.
Semin Arthritis Rheum
 
1990
;
19
:
294
–302.
4
Silman AJ. Trends in the incidence and severity of rheumatoid arthritis.
J Rheumatol
 
1992
;
19(Suppl. 32)
:
71
–3.
5
Spector TD, Hart DJ, Powell RJ. Prevalence of rheumatoid arthritis and rheumatoid factor: evidence for a secular decline.
Ann Rheum Dis
 
1993
;
52
:
254
–7.
6
Jacobbson LTH, Hanson RI, Knowler WC et al. Decreasing incidence and prevalence of rheumatoid arthritis in Pima Indians over a 25 year period.
Arthritis Rheum
 
1994
;
37
:
1158
–65.
7
Kaipiainen‐Seppanen O, Aho K, Laakso M. Shift in the incidence of rheumatoid arthritis toward elderly patients in Finland during 1975–90.
Ann Rheum Dis
 
1996
;
55
:
608
–11.
8
Gabriel SE, Crowson CS, O'Fallon WM. The epidemiology of rheumatoid arthritis in Rochester, Minnesota 1955–85.
Arthritis Rheum
 
1999
;
42
:
415
–20.
9
Shichikawa K, Inoue K, Hirota S et al. Changes in the incidence and prevalence of rheumatoid arthritis in Kamitonda, Wakayama, Japan, 1965–1996.
Ann Rheum Dis
 
1999
;
58
:
751
–6.
10
McCormick A, Fleming D, Charlton J.
Morbidity statistics from general practice
 . Fourth national study 1991–1992. London: HMSO,
1995
.
11
Royal College of General Practitioners, OPCS, Department of Health and Social Security.
Morbidity statistics from general practice
 . Third national study 1981–1982. London: HMSO,
1986
.
12
Symmons DPM, Barrett EM, Bankhead CR, Scott DGI, Silman AJ. The incidence of rheumatoid arthritis in the United Kingdom: results from the Norfolk Arthritis Register.
Br J Rheumatol
 
1994
;
33
:
735
–9.
13
Silman AJ, Hochberg MC.
Epidemiology of the rheumatic diseases
 . Oxford: Oxford Medical Publications,
1993
:
7
–68.
14
McEntegert A, Morrison E, Capell HA et al. Effect of social deprivation on disease severity and outcome in patients with rheumatoid arthritis.
Ann Rheum Dis
 
1997
;
56
:
410
–3.
15
Maiden N, Capell HA, Madhok R, Hampson R, Thomson EA. Does social disadvantage contribute to the excess mortality in rheumatoid arthritis patients?
Ann Rheum Dis
 
1999
;
58
:
525
–9.
16
Townsend P, Phillimore P, Beattie A.
Health and deprivation
 . London: Croon Helm,
1988
.
17
MacGregor AJ, Riste LR, Hazes JMW, Silman AJ. Low prevalence of rheumatoid arthritis in Black Caribbeans compared with Whites in inner city Manchester.
Ann Rheum Dis
 
1994
;
53
:
293
–7.
18
Fries JF, Spitz PW, Kraines RG. Measurement of patient outcome in arthritis.
Arthritis Rheum
 
1980
;
23
:
137
–45.
19
Kirwan JR, Reeback JS. Stanford Health Assessment Questionnaire modified to assess disability in British patients with rheumatoid arthritis.
Br J Rheumatol
 
1986
;
25
:
206
–9.
20
MacGregor AJ, Bamber S, Silman AJ. A comparison of the performance of different methods of disease classification for rheumatoid arthritis. Results of an analysis from a nationwide twin study.
J Rheumatol
 
1994
;
21
:
1420
–6.
21
Larsen A, Dale K, Eek M. Radiographic evaluation of rheumatoid arthritis and related conditions by standard reference films.
Acta Radiol
 
1977
;
18
:
481
–9.
22
Statacorp Stata statistical software: Release 6.0. College Station (TX): Stata Corporation,
1999
.
23
Ropes MW, Bennett GA, Cobb S, Jacox R, Jessar RA. 1958 revision of diagnostic criteria for RA.
Bull Rheum Dis
 
1958
;
9
:
175
–6.
24
Dugowson CE, Koepsell TD, Voigt LF, Bley L, Nelson JL, Daling JR. Rheumatoid arthritis in women: Incidence rates in women in a Group Health Cooperative, Seattle, Washington, 1987–9.
Arthritis Rheum
 
1991
;
34
:
1502
–7.
25
Symmons D, Harrison B. Early inflammatory polyarthritis: results from the Norfolk Arthritis Register with a review of the literature. I. Risk factors for the development of inflammatory polyarthritis and rheumatoid arthritis.
Rheumatology
 
2000
;
39
:
835
–43.
26
Brennan P, Bankhead C, Silman A, Symmons D. Oral contraceptives and rheumatoid arthritis: results from a primary care based incident case–control study.
Semin Arthritis Rheum
 
1997
;
26
:
817
–23.
27
Kvien T, Glennås A, Knudsrød OG, Smedstad LM, Mowinckel P, Førre O. The prevalence and severity of rheumatoid arthritis in Oslo.
Scand J Rheumatol
 
1997
;
26
:
412
–8.
28
Simonsson M, Bergman S, Jacobsson LTH, Petersson IF, Svensson B. The prevalence of rheumatoid arthritis in Sweden.
Scand J Rheumatol
 
1992
;
28
:
340
–3.
29
Saraux A, Guedes C, Allain J et al. Prevalence of rheumatoid arthritis and spondyloarthropathy in Brittany, France.
J Rheumatol
 
1999
;
26
:
2622
–7.
30
Drosos AA, Alamanos I, Voulgari V et al. Epidemiology of adult rheumatoid arthritis in north west Greece 1987–1995.
J Rheumatol
 
1997
;
24
:
2129
–33.
31
Cimmino MA, Parisi M, Moggiana G, Mela GS, Accardo S. The prevalence of rheumatoid arthritis in Italy: the Chiavari study.
Ann Rheum Dis
 
1998
;
57
:
315
–8.
32
Stojanovic R, Vlajinac H, Palic‐Obradovic D, Janosevic S, Adanja B. Prevalence of rheumatoid arthritis in Belgrade, Yugoslavia.
Br J Rheumatol
 
1998
;
37
:
729
–32.

Comments

0 Comments