PATIENTS AND METHODS

Patients

The present patient cohort has been extensively described in previous reports.^9,11,12 The patients were gathered together for a case-control study of the association between hormonal factors and RA.^11,12 In total, 138 consecutive incident female patients with RA visiting the outpatient rheumatology clinic of the Leiden University Medical Centre for the first time between 1982 and 1986 with symptoms of recent onset (0–5 years) and aged 20–50 years at first visit were included in the study. Of the 138 patients invited to participate, three initially refused to cooperate and three were lost during the first years as they moved out of the area. One hundred and thirty two women were prospectively followed up for 12 years. Data were collected at the start of the study, yearly in the first six years and, finally, at the follow up after 12 years. Because the patients had their final assessment done over a two year period, not all patients were assessed at exactly 12 years of follow up. The median disease duration at the last follow up visit was 12 years (range 10–14). This final assessment will be referred to as the 12 year follow up assessment.

Hormonal factors

At the start of the study detailed information on OC use was collected, including dates and duration of use in monthly periods. A life calendar with major life events when contraceptive use might begin (for example, engagement, marriage, childbirth) and a tray with examples of OC packages available in the Netherlands from 1960 to that date were used to help women recall their use of contraceptives. The records of their general practitioners were checked if a patient could not give complete information. Pregnancies before the onset of RA were recorded. At all assessments the number of months of OC use and the number of pregnancies during the previous period were recorded.

Disease activity

At baseline the number of swollen joints, the Ritchie articular index,¹³ and the Westgren ESR scores were obtained as parameters of disease activity. The disease activity score (a pooled index of the ESR, the number of swollen joints, and the Ritchie articular index) was calculated as a combined index of disease activity.¹⁴

Radiology

Radiographs of the hands and feet were taken at study start and after 12 years of follow up and scored using the Sharp score modification van der Heijde (SHS).¹⁵ The SHS separately identifies joint space narrowing (JSN) and erosions in the metacarpophalangeal, proximal interphalangeal, and metatarsophalangeal joints and two interphalangeal joints of the big toes. The joint of the thumb has only an erosion score. Furthermore, six areas in each wrist are assessed for JSN and six others for erosions. The narrowing in both hands and feet is scored from 0 to 4 for each joint. The maximum erosion score in the hands is 5, in the feet 10. The maximum score for the SHS thus is 448: 168 for JSN, 280 for erosions.

After the 12 years' follow up, radiographs of shoulders, elbows, hips, knees, ankles, and subtalar joints were scored according to the Larsen score (LS) for large joints.¹⁶ A score of 1 was given in cases of slight JSN, but not in case of soft tissue swelling and/or osteoporosis only. Larsen scores of 2–5 were given in case of erosive disease; higher scores indicating more damage. In weight bearing joints erosions are not obligatory for Larsen grade 2. The presence of a total joint replacement or an arthrodesis was graded with the maximum score of 5. The total LS is calculated by summing all the scores of the separate joints and ranges between 0 and 60.

Physical functioning

The Health Assessment Questionnaire (HAQ) is used as the parameter for physical functioning and was adapted for the Dutch population.¹⁷ The questionnaire consists of eight categories and the responses are scored on a four point scale. The questionnaire has a final column in which respondents can indicate the use of any aid or device. The use of one of these is scored by at least a 2. The highest score of each category is taken as the score for that category. The final score of the questionnaire is the averaged score of all the categories, and ranges between 0 and 3.

Analysis

To see whether there were baseline differences between patients with or without OC use or with or without pregnancies, the baseline variables were compared between these groups of patients. Differences were tested for significance using Student's t test or Mann-Whitney test, where appropriate.

Outcome at 12 years was also compared between these subgroups. In addition, the effects on outcome of pregnancies and OC use before RA onset, and during RA were evaluated.

The effect of OC use and pregnancies on outcome was further studied by separating the sample into subgroups of patients based on the tertiles of total months of OC use and on the tertiles of the number of pregnancies. In all cases we used non-parametric test statistics.

Finally, Spearman rank correlations and ordinal polytomous logistic regression were used to analyse the association between outcome variables and several hormonal parameters.

RESULTS

Table 1 shows the demographic and baseline variables. There were no statistically significant differences between the groups except for the age of disease onset between patients with or without OC use. Furthermore, there was also no difference in baseline variables, including OC use and pregnancies, between the women studied and those lost to follow up.

The median duration of OC use before the onset of RA was 60 months (range 0–248) and during the course of the RA the median duration of OC use was 34 months (0–144). The median of the total duration of OC use was 94 months (0–392). The median number of pregnancies was 2 (0–8) before the onset of RA and 0 (0–4) during the course of RA. The median of the total of pregnancies was 2 (0–8).

Table 2 shows the outcome variables for radiographic joint damage and disability in the four subgroups of patients. The patients who either used OC or had been pregnant had less radiographic joint damage of the small and large joints, but this difference did not reach statistical significance. Disability as measured by the HAQ was less in the group with OC use and pregnancies; in the group of women who had been pregnant before the onset of RA the difference in HAQ was the largest (0.38).

By further study possible differences in outcome between groups, the medians of the SHS, Larsen, and HAQ, were calculated for each tertile of cumulative OC use and of cumulative pregnancies (table 3). Again the patients within the highest tertile of pregnancies or months of OC use had less radiological damage and less disability. Patients who had been pregnant three or more times had significantly less disability than those who had not been pregnant.

The Spearman rank correlations of the outcome parameters with the exposure to pregnancy or cumulative OC use were calculated. Only the association of OC use before the onset of RA with the LS of large joints was statistically significant. After checking for age, rheumatoid factor, and SHS at onset, the association became non-significant.

When polytomous logistic regression analysis on the tertiles of the SHS distribution was used, the (pseudo) R² of the combined influence of OC use and pregnancies was only 0.026. Neither number of pregnancies (odds ratio (OR)=0.96, p=0.70) nor months of OC use (OR=0.99, p=0.11) were significantly related to SHS tertile.

DISCUSSION

This study shows that OC use and pregnancy only minimally influence the long term outcome in RA as assessed by joint damage and disability. There is, however, a trend towards less joint damage and disability in patients with long term OC use and multiple pregnancies.

The study shows that the median HAQ score in women with a total of two pregnancies during life is 0.63 lower than the score in women with no child or one. A difference between HAQ scores of more than 0.2 is considered clinically relevant.¹⁸ The trend towards less radiographic joint damage and less disability in women with long OC use is noticeable because the women in the OC use group were significantly older than the group without OC use. Disability in RA has been shown to be greater in older patients, so the trend towards less disability in the older patients with long OC use is even more remarkable. Disease modifying drug (DMARD) use influences outcome and was shown to have a beneficial effect on radiographic damage and function. All women were similarly treated by the less aggressive standard of two decades ago. Women who wanted to get pregnant often discontinued drug treatment. Therefore the trend towards less radiographic damage and disability is not likely to be due to DMARD use. The trend towards a milder disease in the group of patients with a high exposure to female sex hormones, however, did not show statistical significance, which might be owing to a lack of power. This is not surprising considering the large ranges of the outcome variables (table 2 and 3). Three hundred and thirty female patients divided into two subgroups would be needed to reach statistical significance, given the data of the present study and the trend found. In the univariate analysis the associations were not strong. Moreover, in a multivariate analysis the combined influence of OC use and pregnancies explained no more than at most 4% of the variance of the 12 year outcome as measured by the SHS, LS, or HAQ.

Pregnancy was first reported to induce remission in RA in 1935.¹⁹ Since then, many studies have reported the ameliorating effect of pregnancy on disease activity. As a possible explanation for this phenomenon it has been suggested that levels of endogenous female hormones influence the disease. This view was further strengthened by the finding that OC use had a protective role in the incidence of RA and also had a restraining influence on the course of the disease.^7,11,20 In one study on hormone treatment in RA some improvement was reported in disease activity when oestrogen was given.²¹ In view of this evidence it was plausible that the long term outcome of RA might be influenced by female sex hormones. Our study, however, shows that hormonal parameters contribute only minimally to the explanation of long term radiological damage and disability.

This is the first prospective long term outcome study to investigate the association between hormonal factors and joint damage. One other study reported the long term outcome in RA. In 1966 Oka described a retrospective case control study in a group of 100 consecutive patients who had been pregnant after the onset of RA.⁶ These women were compared with an equal number of women who had not been pregnant since developing RA. In that study, which comes with the usual limitations of a retrospective study, no significant differences were found in functional ability or disease activity. This is in accordance with the present study.

In conclusion, this study shows that OC use and pregnancy do not significantly influence outcome in RA. There is, however, a trend for patients with multiple pregnancies and long term OC use to have less radiographic joint damage and a better functional level. This information is important in counselling young women with RA.