Results

Two hundred and eighty one cases and 507 referents participated in the study, resulting in response rates of 67% and 59%, respectively. The mean age was 62 years for the cases and 58 years for the referents, and 64% and 51%, respectively, were women. Seventy nine per cent of the cases were seropositive. The mean duration of the disease in the cases at the time of entry into the study was 8 years, as was the mean time span between the dummy date of onset and entry into the study for referents. For cases the mean interval between year of onset of symptoms and year of diagnosis was less than four years.

The response rate was low and therefore the possible influence of non-response bias was evaluated. These results are presented in detail in the study on occupational determinants of importance for RA.22 In summary, the comparison revealed that non-responding women were older than the women who responded, and that all non-responders were more likely to be single and to earn less than the corresponding group of responders.

Men who were current smokers had a significantly increased occurrence of RA (exposed cases/referents 40/62, odds ratio (OR) 2.2, 95% confidence interval (CI) 1.1 to 4.2, with adjustment for age and socioeconomic status). Pack years of smoking were calculated by multiplying the mean number of packs per day by the total number of years of smoking. Those with no previous use of any form of tobacco were regarded as the reference group. The results are presented in table 1. In men an increase in risk was found with increasing number of pack years. Slightly increased risks were at first seen for non-smoking men passively exposed to smoking fumes and for men reporting previous use of tobacco for buccal application (Swedish snuff), but the risks did not remain when adjustment was made for socioeconomic status.

A non-significant increased risk of RA was seen in both men and women who consumed at least 750 ml of alcohol per drinking session at the age of 25 compared with total abstainers. With adjustment also for smoking and socioeconomic status, the ORs decreased towards unity (data not shown).

When men and women were analysed together with adjustment for age and sex, atopic allergy was less common amongst cases than among referents (exposed cases/referents 32/81, OR 0.6, 95% CI 0.4 to 1.0) and this difference was almost significant. The results were similar after adjustment for smoking habits. When the cause of the allergy was examined, somewhat decreased ORs were seen for allergic reactions to pollen (exposed cases/referents 13/39, OR 0.6, 95% CI 0.3 to 1.2) and to house dust (exposed cases/referents 3/8, OR 0.5, 95% CI 0.1 to 2.0). The results for men and women are shown in tables 2 and3.

Results regarding other comorbidity, previous infections and medication, as well as results concerning educational level, residential exposures, some other external factors and psychosocial stress are also shown in tables 2 and 3. Only one of the women who reported use of insulin had previously been treated with corticosteroids. Vaccinations did not influence the risk of RA (data not shown). Most of the women who reported a joint injury at least 20 years before the diagnosis of RA had their injury before the age of 20. Previous operations and other kinds of trauma were not associated with RA (data not shown). No increase in risk was seen from exposure to pets (data not shown). Apart from the non-significant increase in the risk for RA in women exposed to solvents, no associations between RA and leisure time exposures to chemicals were found (data not shown).

The results on reproductive factors are shown in table 4. “Normal” age of menarche and menopause was defined from the age distributions seen in the female referents, and those with onset of menarche at age 13–15 and of menopause at age 48–53 were considered as the reference group. When the onset of menopause was studied, those with a history of gynaecological operations (12 cases and 20 referents) were excluded from the analyses. In consideration of the length of the fertile period, those with a fertile period of 34–40 years were regarded as the reference group. There was no difference between cases and referents with regard to interrupted pregnancies (data not shown).

The logistic regression analyses did not produce any major changes in risk estimates to indicate confounding in men or women. Apart from some occupational exposures, the analyses included exposure to farm animals and private wells for men, and oral contraceptive use and some psychosocial factors for women.

Discussion

The response rate was lower than expected, and analyses were therefore made to evaluate the possible presence of a non-response bias. When age, socioeconomic status, and income were studied, there were no indications of a systematic non-response bias concerning socioeconomic factors since the observed differences between responders and non-responders seemed to be of similar magnitude in both cases and referents.

Subjects were unaware of the specific aim of the study and of their case-referent status to avoid any possible recall bias and most of the hypotheses tested are unknown to the public. As a dummy date of onset was used for referents, cases and referents were required to recall exposures over an equally long period of time. The awareness of the negative impact of tobacco and alcohol use in general may have resulted in an underestimation of the subjects' true consumption, but such a tendency of underreporting is probably equally likely in both cases and referents.

The use of prevalent cases of RA from the University Hospital in Linköping may have resulted in a selection bias. It is possible that those affected by more severe disease had died before the time of inclusion in the study, while those with less severe disease may have been referred to other health care units. Assuming an association between the analysed factors and disease severity, it is possible that those who had died represent cases exposed to risk factors and unexposed to preventive factors, while the opposite could be assumed for the less severe cases. However, even if an association between certain factors and disease severity is assumed, a possible counterbalancing loss of extreme cases can be expected to minimise the selection bias.

Several kinds of exposures, which probably have different mechanisms of action, have been studied, and it is reasonable to assume different latency periods for these. Varying latency requirements have therefore been tested and used when relevant.

There was an increased risk for RA in both current and previous smokers, and in men there was a dose-response relationship which further strengthens the credibility of a true cause-effect relationship. These results are in agreement with previous findings.12-14 ,23 The tendency for higher ORs in seropositive cases is also in agreement with previous findings of an association between smoking and RA severity.23 ,24 Smoking may also cause an increased production of RF in non-rheumatic subjects25 and the RF concentration in patients with RA has been shown to be linearly related to the number of years of smoking.23 Furthermore, the association between smoking and RA, especially seropositive RA, seems to be stronger in men than in women.13 ,23

Education has been regarded as a socioeconomic marker and a low educational level has previously been found to be associated with an increase in severity of RA.26 However, in a recent study no association was seen between RA and a low educational level.13 In the present study an inverse relationship between educational level and RA prevailed after adjustments for smoking habits and occupation. It is possible that a higher educational level is associated with a healthier lifestyle or a different pattern of healthcare use.

A recent study27 has shown that coffee consumption is associated with both RF seropositivity and development of RA. This is interesting because the amount of coffee consumed is likely to vary with other lifestyle factors previously found to be associated with RA such as smoking. However, questions concerning coffee consumption were not included in the present study.

Previous studies of the coexistence of atopic manifestations in patients with RA have shown an inverse relationship between the two conditions,6-8 supporting the theory of a functional dichotomy between Th1 and Th2 type responses.3 Assuming a lifelong commitment to either Th1 or Th2 type responses, a history of allergic manifestations in cases and referents was compared. The results from the present study support an inverse relationship between RA and atopic allergy, while the results concerning the presence of asthma and eczema were divergent between the sexes. It is possible that subjects have reported other than atopic conditions and Th1 mediated contact dermatitis might have been reported as atopic eczema. The question concerning atopic allergy is probably more reliable and is supported by the decreased risk of allergic reactions to pollen and house dust.

An association between Th1 dominated conditions is supported by the relationship between RA and insulin treatment in women as diabetes mellitus type 1 is considered to be a Th1 condition.28Autoimmune thyroiditis is also regarded as a Th1 condition29 and has previously been found to be associated with RA.5 However, those reporting thyroid conditions in this study may also have had other thyroid diseases.

Various infections have been suggested to contribute to the development of RA, and the association with antibiotics in men seen in this study may be regarded as an indication of previous bacterial diseases. In women there was also an unexpected negative relationship between RA and rubella, previously considered to be a possible risk factor for RA.30

The increased risks associated with the use of a private well and problems with mould indoors are new findings. Private wells are found in the countryside in Sweden, and the risk might be conveyed by contamination of chemicals used in nearby cultivations. Wells can also contain various organic degradation products, minerals, and bacteria.

The increased risk of RA associated with customer exposure to hair dyes and/or bleach prevailed after adjustment for work as a hairdresser. Hairdressers have previously been found to have an increased risk for RA17 and for lymphoma,31 a condition associated with RA32 and possibly also associated with customer use of hair dyes.33

The significantly decreased risk seen in women who have ever breast fed a child is not in accordance with the suggested association between RA and hyperprolactinaemia.2 ,10 However, such an inverse relationship has also been reported previously.34 Although previous studies have not associated treatment with non-contraceptive sex hormones with any major change in risk of RA,35 our study indicates a possible beneficial effect from their use.

In conclusion, the findings support the hypothesis of a functional dichotomy between RA and atopy, which is characterised as a Th2 dominated disease, while there is a clinical correlation between RA and other Th1 dominated conditions such as diabetes mellitus. Most of the results relating to the role of reproductive factors and hormonal treatment are in agreement with existing theories on the influence of hormonal factors on the development of RA. Furthermore, the negative impact of lifestyle factors such as smoking is confirmed while higher education seems to have a protective effect. Most of the results on external factors such as physical and psychological trauma and residential exposures are hypothetical as there are no obvious explanations for these findings. It is likely that the risk factors and beneficial factors reported in this study act together with other aetiological or protective factors to facilitate or to prevent the end point of RA. Further studies are warranted to confirm this.