RheuniatnlomY

Rheumatol Int (1992) 12:33-38

INTERNATIONAL Clinical and Experimental investigations

9 Springer-Verlag 1992

Prevalence of radiographic osteoarthritis in two elderly European populations E. Bagge 1, A. Bjelle 1, H.A. Valkenburg z, and A. Svanborg 3 1 Department of Rheumatology, University of G6teborg, Sahlgren Hospital, S-413 45 G6teborg, Sweden 2 Department of Epidemiology, Erasmus University, Rotterdam, and Faculteit der Geneeskunde, Van der Borenstraat 7, NL-1081 BT Amsterdam, The Netherlands 3 Geriatric Section, University of Illinois at Chicago, Department of Medicine, Box 6998, Chicago, Illinois 60680, USA, and Department of Geriatric and Long-Term Care Medicine, University of G6teborg, G6teborg, Sweden Received April 22, 1991/Accepted September 27, 1991

Summary. Was compared the prevalence of radiographic osteoarthritis (ROA) in two elderly populations, one in G6teborg (Sweden) and one in Zoetermeer (The Netherlands). The comparison was preceded by an inter-observed analysis of 150 films, revealing lower percentage agreements and kappa values when analysing the films on a five-point scale compared to the analysis in a dichotomy of "abnormal" and "normal" films. An intraobserver analysis of 50 films revealed higher percentage agreement and kappa values of well over 0.75, suggesting that radiographs could be read by one observer when different populations are being compared. The prevalence of hand ROA was higher in the G6teborg population, while the prevalence of knee ROA showed no significant differences. No significant increase with age of the cohorts was found for hand ROA in either population and no increase with age for knee ROA was found in the Swedish population. In the Zoetermeer population, knee ROA increased with age in both sexes. Key words: Inter- and intra-observer variation - Prevalence of ROA - Age-related trends

Introduction Radiographic examination of the joints for signs of osteoarthritis (OA) is still one hallmark establishing the diagnosis [1-4]. As long as clinical signs and symptoms insufficiently predict the presence of radiographic osteoarthritis (ROA) in populations [4-9], the use of radiographs will be mandatory for studying determinants of clinical and radiological (progression of) OA. Different prevalence rates of ROA in different populations can be due to differences in criteria applied when evaluating the films, or to observer variance. In an epidemiological study of 70-, 75- and 79-year-old cohorts in G6teborg, Sweden, considerably lower rates of hand and knee ROA were observed than had been reported in other studies [10]. Furthermore, the rates of ROA of the knees

tended to decrease with the age of the cohort, suggesting a cohort effect. As the prevalence rates of ROA in some cross-sectional population surveys were reported to decrease with increasing age [11], a cross-country study between Sweden and The Netherlands was designed. The aim of the present study was to compare prevalence rates of hand and knee OA at different ages in two elderly populations with a generally similar socio-economic structure, having first completed a detailed methodological comparison of inter- and intra-observer variations.

Study population and methods G6teborg, Sweden. The longitudinal study "70-year-old people in G6teborg" was initiated in 1971/1972, with systematic sampling of 30% of the 70-year-old population in G6teborg (cohort I, n= 1148; participation rat 85%). Five years later, the survivors of chort I were followed up (n = 743; participation rate 92%) and at the same time a new sample of 70-year-olds was obtained by the same procedure (cohort II, n = 1281; participation rate 81%). The survivors of cohort II at age 75 (n = 768; participation rate 88%) and of the first sample at age 79 (n = 538; participation rate 82%) were re-examined in 1980/1981. In 1981/1982 a third cohort of 70-year-olds (cohort IIL n = 806; participation rate 77%) was systematically sampled and examined. Subsamples from the three cohorts were studied for joint disorders [10, 11] and selected for radiographic examination of hand and knee joints (Table I). Incompleteness of examinations was mainly due to logistical difficulties. Comparison of respondents and non-respondents of the general examination revealed no significant differences of marital status, income or registration by the local temperance boards [13, 14]. The subsamples did not differ significantly from samples with regard to these parameters, physical activity, absence of definable disease or frequency of joint complaints. Zoetermeer, The Netherlands. Between April 1975 and April 1978, 3091 adult men and 3493 women, aged 20 years and older, were examined for rheumatic complaints and conditions in the EPOZstudy (Epidemiological Preventive Organisation Zoetermeer) [9, 12]. These 6584 men and women constituted approximately 80% of the adult population of two districts of the city of Zoetermeer, one of urban and the other of rural descent. In the survey centre, radiographs were taken of all of hands and from 45 years onwards (1359

34 Table 1. Radiographed subsamples of two elderly populations

Table 2. Inter-observer variation in the prevalence of all grades of radiographic osteoarthritis (grade 0-4) in two elderly populations

G6teborg Age

Women

70 75 79 Total Zoetermeer Median age

(range)

70.0 75.0 79.9

(67-72) (73 -77) (78-92)

Total

Men

Total

Joint

Zoetermeer (100 radiographs)

Grteborg (50 radiographs)

Agreement [%]

Kappa

Agreement [%1

Kappa

DIP PIP IP I MCP II-V MCP I CMC I

71.1 68.5 72.3 60.0 63.3 62.6

0.57 0.54 0.59 0.40 0.47 0.45

82.7 67.3 75.0 67.3 70.9 54.9

0.67 0.53 0.61 0.49 0.53 0.36

65 73 43

63 59 31

128 132 74

181

153

334

Women

Men

Total

35 30 56

17 24 31

52 54 87

Total

66.9

0.52

69.5

0.55

121

72

193

Knee

63.6

0.51

76.9

0.64

men and 1598 women) radiographs of the knees were also obtained. The radiographs were read by two observers for the first half of the study and by one observer for the second half, independent of other findings [11]. HAV was one of the readers throughout the study. For the present study, the subjects were selected from the total available group, in that they clustered around the ages 70, 75 and 80 and that a total of approximately 200 people were available.

Radiographic techniques. Hand and knees were radiographed in postero-anterior projections. All radiographs of the knees were taken in the standing position, in Grteborg with parallel feet and in Zoetermeer with the feet slightly externally rotated. In the G6teborg study, Agfa-Gevaert radiochrome films were used. In the Zoetermeer study, Kodak No-screen films were used for the hands and Kodak Definix films for the knees.

Table 3. Intra-observer variation in the prevalence of all grades of radiographic osteoarthritis (grade 0-4) in G6teborg radiographs read by one observer within a four-month interval Joint

Agreement [%]

Kappa

DIP PIP IP I MCP II-V MCP I CMC I

86.0 87.0 90.0 92.0 93.0 86.0

0.70 0.79 0.81 0.78 0.87 0.80

Total

89.0

0.82

Knee

84.0

0.74

Radiographic examination. In the previous Grteborg study the ROA grade was based on the presence of joint space narrowing, bone cysts and sclerosis, while osteophytes alone were not used as a criterion for ROA [10, 14]. In the present study, the radiographs obtained in 1980/1981 and in 1981/1982 were re-examined according to the criteria described by Kellgren and Lawrence [3] and read by one observer (EB). ROA was defined as grade 2 (definite osteophytes and possible joint space narrowing) or higher. Hand joints on each side were recorded separately in the following groups: DIP, PIP, IP I, MCP II-V, MCP I, CMC I. The intracarpal and radiocarpal joints were difficult to interpret in some of the Agfa Gevaert films and therefore excluded from the analyses. Radiographs of each knee were recorded separately. However, in the present study ROA of the left and right side were taken the ROA grade, thus representing the most affected side. The present report was restricted to a comparison of prevalence rates of grade 2 + OA in the two populations. For the inter-observer study, two sets of films were randomly selected from Grteborg and Zoetermeer. In the first stage, 50 radiographs were read by both EB and HAV after mutual discussion and agreement on the criteria to be applied for ROA. In the second stage, the readings of 100 Zoetermeer radiographs were compared. An intra-observer study within a 4-month interval was performed by EB on 50 randomly selected radiographs of the Grteborg population. The mortality in the Grteborg population of 70, 75 and 79-yearolds radiographed in 1980/1981 and 1981/1982 was checked from the death registers up until June 1990. In the Zoetermeer population information was available on whether the respondents died throughout the 14 years of follow up since the start of the study.

Statistical analyses. The chi-square test was used for comparison ' between the two populations with respect to different joints simul-

taneously. Proportions were compared by Fisher's exact test, twotailed. Comparisons between groups were performed by elimination of the influence of sex and age using Mantel's test, two-tailed [15]. The inter- and intra-observer agreements were analysed as observed percentage agreement and kappa statistics [16]. Calculations were performed both over the whole range of grades (5-point scale) and on a dichotomy between "abnormal" (grade 2 + and grade 3 +) or "positive" films and "normal" (grade 0-1 or 0-2) or "negative" films.

Results Inter- and intra-observer studies F o r the i n t e r - o b s e r v e r analysis o f separate joints, the whole range o f grades ( 0 - 4 ) was i n c l u d e d (Table 2). The lowest a g r e e m e n t b e t w e e n the two observers was f o u n d for the M C P I I - V , M C P I a n d C M C I joints. However, w h e n the analysis was d o n e s e p a r a t i n g grades 0 - 1 f r o m grades 2 - 4 a n d grades 0 - 2 f r o m grades 3 - 4 , the interobserver analysis o f all h a n d j o i n t s together r e n d e r e d a k a p p a value of 0.67 a n d 0.65 respectively in the Zoetermeet films, a n d 0.74 a n d 0.69 respectively in the G 6 t e b o r g films. It is obvious, therefore, that classifying the films as " p o s i t i v e " a n d " n e g a t i v e " increased the strength o f the associations. As expected, the i n t r a - o b s e r v e r analysis (Table 3) showed a c o n s i d e r a b l y better a g r e e m e n t t h a n the inter-

35 Table 4. a Differences in prevalence rates (%) for radiographic osteoarthritis of hand and knee joints in 70-year-olds of two populations DIP

PIP

IP I

MCP II-V

MCP I

Women Grade 2 + G6teborg Zoetermeer

86.1 65.7

55.3 40.0

67.7 40.0

52.3 42.9

73.3 45.7

Differences

20.4 *

15.3

27.7 **

Men Grade 2 + G6teborg Zoetermeer

74.6 47.1

46.0 29.4

77.8 35.3

63.5 23.5

80.9 29.4

Differences

27.5 *

16.6

42.5 **

40.0 **

51.5 ***

9.4

26.6 **

CMC I

Total

Knee

28.1 28.6

93.8 74.3

55.3 45.7

-- 0.5

19.5 **

9.6

26.8 17.7

92.1 76.5

52.4 11.8

9.1

15.7

40.6 **

9 P