Results

Completion rates for all nine surveys during the study period ranged from 75% to 97% with the average being 85%. Between 1980 and 1996, the proportion of group A subjects in the entire responders to questionnaire ranged from 3.2% to 5.1% (mean 3.9%) and group B subjects from 4.4% to 7.1% (mean 5.8%). Almost all group A (99%–100%, mean 99.8%) and B subjects (95%–100%, mean 98.0%) underwent an examination either at the centre or at home.

There were 16 incident cases (8 men and 8 women) during the study period. As table 1 shows, the crude incidence rates for the total population were 0.35 per 1000 person years (PY) between 1965 and 1975, 0.22 between 1975 and 1985, and 0.09 between 1985 and 1996, for men 0.27, 0.18, and 0.20, and for women 0.43, 0.26, and 0.00. Age and sex adjusted incidence rates during those three time periods for the total population were 0.39, 0.24, and 0.08 per 1000 PY. Age adjusted incidence rates for men were 0.29, 0.20, and 0.16, and for women 0.48, 0.28, and 0.00, respectively. The adjusted incidence significantly decreased during the study period for the total population (p<0.025) and women ( p<0.01), whereas no significant change could be seen for men (p>0.5) (fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Changes of adjusted incidence.

Four cases (2 men and 2 women) out of 16 incident cases, diagnosed at least once as having definite RA, were apparently found to be completely remitted during the follow up examinations. Among these four cases, one always showed seronegativity, while in the other three, seropositivity fluctuated. However, they showed seronegativity at least once during the observation period, whereas the remaining 12 cases were always seropositive, suggesting a strong correlation between seronegativity and a benign course of RA. These four remitted cases were also non-erosive at the time of diagnosis. Later, one (woman) showed erosive change, 21 years after the diagnosis, without meeting the criteria for definite RA, and the other (man) became erosive 17 years after the diagnosis, fulfilling the criteria. The remaining two remitted cases (one man and one woman) continued to be non-erosive during the follow up period without fulfilling the criteria.

Table 2 shows the point prevalence rates of RA in nine consecutive surveys during the study period. The crude prevalence rates for both men and women combined varied from 2.0 to 4.7 per 1000 persons, for men from 0.0 to 2.4, and for women from 2.0 to 7.0. Age and sex adjusted prevalence rates for both men and women combined varied from 1.7 to 5.4, while age adjusted prevalence rates for men were from 0.0 to 2.5, and for women from 2.3 to 8.2. The adjusted prevalence rate tended to decrease for both men and women combined (p<0.1), and showed a significant decline for women (p<0.025), whereas for men no significant change was observed (p>0.5) (fig3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Changes of adjusted prevalence.

Additional information on RA patients was provided from a public health nurse, national health insurance invoices, general practitioners, and hospitals. However, no additional cases were found in the study population from these sources.

Discussion

During a 31 year study period of the Kamitonda population, there was a decrease in the incidence of RA in the whole population, which was confined to women. A decrease of prevalence was also evident for the whole population, which was also restricted to women, reflecting the decrease of incidence in women.

Until now, four studies have reported the incidence of RA in Japan. Each was community based and performed from the late 1950s to 1970s. Two were successively performed on atomic bomb survivors in Hiroshima and Nagasaki, and showed the crude incidence of definite RA to be 0.4 per 1000 PY for all ages in the first 1958–1964 survey and 0.9 per 1000 PY in the second 1965–1966 survey.5 ,6 The third, conducted by our study group in the Tojo district, Tondabayashi, Osaka from 1962 to 1966, estimated the crude incidence of definite RA in inhabitants of all ages to be 0.6 per 1000 PY.7 The fourth was the early results of the present Kamitonda study from 1965 to 1980,8 in which the incidence of both probable and definite RA combined were described. Hence, subjects with probable RA at the first survey in 1965 who later met criteria for definite RA were not regarded as incident cases in our previous report, being different from the present report. As seen in this study, the crude incidence of definite RA in Kamitonda between 1965 and 1975 was 0.35 per 1000 PY, being within an expected range in Japan at that time. The incidence of RA in Kamitonda is probably representative of the whole population of Japan. In addition, the figures are similar to those obtained by recent studies in white populations. They were 0.4 in the UK1and 0.37 in Finland,14 although the survey methods were quite different, the former being based on a prospective case notification and the latter on records review.

Secular changes in the incidence of RA have been suggested in several reports. Of these, Linos et al,15 Dugowson et al,16 and Hochberg17 recognised a declining trend only in women, whereas Chan et al 18 concluded from a three year observation that there was no secular change in incidence. On the other hand, Kaipiainen-Seppanen et al 19observed an approximately 15% decline in the incidence of RA from 1980 to 1990, affecting seronegative RA nearly exclusively, particularly in men. These studies were mostly based on records from hospitals and general practitioners, or members of a specified group, with the inevitable selection bias ensuing. In only one report, by Jacobssonet al,4 incidence and its secular decline in women was shown by a community based approach, but it was from a group of Pima Indians, a particular population having unusually high seropositivity and a much higher incidence rate than found in white populations.

Even in a community based study like ours, the problem of selection bias may be encountered during the long term follow up period. In this study, selection bias may have been introduced by modification of the survey method, being obliged to do so for a variety of reasons. In the earlier period from 1965 to 1975, every effort was made to examine all inhabitants, either at the centre or at home, regardless of rheumatic complaints, because the study was being performed not only for RA but also for other rheumatic diseases. In the later period from 1980 to 1996, examination at the centre or at home was restricted to those who had rheumatic complaints. We believe, nevertheless, that this modification did not influence our results. Firstly, there were no incident cases who seemed to have gone undetected in the preceding surveys. Secondly, the decline of incidence was only observed in women, while the incidence of men remained much the same throughout the study period. Thirdly, a decline of incidence for women was also noted from the second to the third period. Fourthly, the information provided from several sources did not add any new cases who were not found by our surveys. The size and low mobility of the study population enabled us to gather adequate long term information on the health status of the inhabitants.

As is always the case of a longitudinal population based study of RA, subjects with a benign form of RA whose symptoms were transient and relapsing could have been missed.20 As our evaluation of inhabitants was made at survey intervals of 3–5 years, patients with RA in whom onset and remission occurred between two consecutive surveys may have been ignored. However, we consider this possibility to be very low, because we did not encounter any person with such history throughout all of the surveys. Furthermore, we did not find any relapsed case of RA who may have escaped detection previously. The completion rates in our surveys were high and subjects with a past history of arthritis or rheumatic complaints possibly related to RA were all interviewed by doctors. Nevertheless, the possibility of missing mild cases could not be completely neglected, but would be minimised by the prolonged observation of the same population.

In a longitudinal observation of incident cases, four of 16 cases later showed complete remission. One of them fulfilled the criteria again in 1992, 17 years after the initial diagnosis of RA . We did not consider this person as a new case in 1992. Whether we had counted this case as new or not, does not influence the nature of our results, as this relapsed case was male. We further tested how the results would change, assuming that one more woman with RA had been found during 1985–1996. However, the trend of declining incidence of total population (χ² =4.20–4.42) and of female RA (χ²=4.78–5.12) still remained significant with the statistical power of p<0.05, regardless of the age of the postulated woman.

The decline of incidence of RA in Kamitonda may be explained by a change of some environmental factors preferentially occurring more obviously in women. Linos et al,15 recognising the declining incidence of female RA, suggested that the use of oral contraceptives was responsible, whereas Jacobsson et al 4 argued against the supposition that use of oral contraceptives could explain the declining incidence, as the use of such pills was constant in the Pima Indian population during the survey period. It should be noted that, in Japan, oral contraceptives have not been available on the free market, but only when a doctor recognised their necessity and prescribed them. Therefore, the use of these drugs has invariably been very low in our country, at approximately 1% of total contraception.21 The very low rate of current or ever oral contraceptive users in population was also confirmed in our previous report.22 Accordingly, the use of oral contraceptives cannot be a responsible factor for the observed phenomenon in our country. The most marked changes occurring in Japanese women in the past years are seen in social factors, of which we can enumerate the following; an increase of economic independence, a higher educational level, and a marked decrease in the birth rate, thus facilitating employment and social activities. These phenomena are apparently in accordance with the hypothesis by James who stated that a higher androgen concentration in women responding to an augmented role in the society could prevent the onset of RA.23

The rather unusual finding that the cumulative incidence was almost the same in men and women might also argue that the onset of RA is more dependent on external life conditions in women than in men, as suggested above. On the other hand, three of 16 incident cases (19%) had an occurrence of RA in first degree relatives. Interestingly, all three incident cases with familial occurrence were in men. According to Kwoh et al,24 in cases with multiple familial occurrence of RA, the female to male ratio for probands was 1:1 versus 3:1 in cases with single familial occurrence. Therefore we speculate that although the number of patients is very small, RA in men might have been seen more often in genetically disposed persons, thus revealing the constant incident rate of RA in men. As a result, the occurrence of RA may in the future become nearly equal between sexes.

It is worthy to note, however, that our report concerns the significant declining tendency of incidence of RA in Japanese women, as obtained from community based follow up surveys covering a period of more than 30 years. Nevertheless, as it is based on an obviously small number of incident cases, the extent of our conclusions is limited, as seen by the fairly broad confidence interval of the incident density estimate.