Materials and methods

We established a cohort comprising all Danish women and men born in 1935–89. This cohort was linked to national health registries by means of the unique identification number ascribed to all Danish inhabitants to obtain information about reproductive histories and RA hospitalisations. Although the risk factors considered were related to pregnancy, we also analysed their possible impact in men because fathers probably share major socioeconomic characteristics with mothers but do not share the biological factors related to pregnancy. Thus, fathers served as a control for unmeasured lifestyle factors.

Reproductive histories

For both women and men, information about live births was obtained from the Civil Registration System, a continuously updated database that has recorded all demographic events in Denmark since 1 April 1968. As described elsewhere,27 this database provides a virtually complete record of live births for women born in 1935 or later. For each parent, children were defined as pregnancies that resulted in at least one live born child. For women born in or after 1955, we also obtained detailed information about induced abortions (International Classification of Diseases (ICD-8) 640, 641, 642; ICD-10 O04, O05, O06) from the National Registry of Induced Abortions and the Danish National Patient Registry (period 1974–2004). For ectopic pregnancies (ICD-8 631; ICD-10 O00), spontaneous abortions (ICD-8 643; ICD-10 O03), missed abortions (ICD-8 63461, 63462, 63463, 63469, 6451; ICD-10 O021), hydatidiform moles (ICD-8 63429, 6450; ICD-10 O01), hyperemesis (ICD-8 63809, 63899, 76249; ICD-10 O21), gestational hypertension (ICD-8 63700, 63702; ICD-10 O12, O13, O16), and pre-eclampsia (including eclampsia) (ICD-8 63703, 63704, 63709, 63719, 63799, 63999, 76219, 76229, 76239, 76299; ICD-10 O14, O15), we used data from the Danish National Patient Registry (period 1977–2004). Data about stillbirths were obtained from the Medical Birth Registry and the Danish National Patient Registry (period 1973–2004).

RA outcomes

Information about first inpatient hospital contacts for RA (ICD-8 71219, 71229, 71239, 71259; ICD-10 M05 and M06) was obtained from the Danish National Patient Registry (period 1977–2004).

Stratification of person-years and RA outcomes

Women and men were analysed separately. Each cohort member contributed person-years at risk from her or his 15th birthday or 1 January 1977, whichever came later, to the date of first RA hospitalisation, death, emigration, disappearance or 1 January 2005, whichever came first. By the end of the study (31 December 2004), cohort members born in 1935 or later were 15–69 years old and women born in 1955 or later were 15–49 years old. Specifically, person-years and RA outcomes were stratified in a time-dependent manner according to age, birth cohort, calendar period (all in 1-year age groups), marital status (unmarried, married, separated/divorced or widowed) and each of the following live birth-related variables: number of children (0, 1, 2, 3, 4+), age at birth of first child (<20, 20–24, 25–29, 30+ years) and time since birth of the most recent child (0–<2, 2–4, 5–9, 10–14, 15–19, 20+ years). For women born in 1955 or later, we also stratified person-years and RA outcomes in a time-dependent manner by the number of spontaneous abortions (0, 1+), missed abortions (0, 1+), stillbirths (0, 1+), any of the three preceding idiopathic pregnancy losses (0, 1, 2+), induced abortions (0, 1, 2+), ectopic pregnancies (0, 1+), hydatidiform moles (0, 1+) and according to the number of pregnancies complicated by hyperemesis (0, 1+) or pregnancy-related hypertension, including diagnoses of gestational hypertension (0, 1+) or pre-eclampsia/eclampsia (0, 1+) and time since most recent pregnancy-related hypertensive disease (0–<2, 2–4, 5–9, 10–14, 15–19, 20+ years). This time-dependent handling of exposures and outcomes ensured that only pregnancies that preceded the first recorded RA hospitalisation were considered.

Poisson regression analysis

The statistical analysis of the resulting table of stratum-specific RA incidence rates was carried out as a log-linear Poisson regression analysis, yielding rate ratios (RRs) of first inpatient hospitalisation rates for RA and 95% confidence intervals (CIs). In RR calculations we adjusted for marital status as well as for age, birth cohort and calendar period using cubic splines restricted to be linear in the tails.28 Adjustment for other variables was performed using the categorisations described above. In trend tests, categorised quantitative variables were treated as continuous variables by replacing each category by the person-years-weighted median of the original variable in that category. Trend tests were carried out only after acceptance (p⩾0.05) of the corresponding model reduction from a categorical to a linear description of the association between RA and the variable in question. Two-sided tests were applied; p values <0.05 and 95% CIs excluding unity were considered statistically significant. To further address findings in women whose pregnancies were complicated by hyperemesis, gestational hypertension and pre-eclampsia, a set of parallel analyses were made for men who fathered children born after pregnancies complicated by these diseases.

Results

Women

Among 2 140 056 women born in 1935–89, 7017 patients were hospitalised with RA during 43.5 million person-years of follow-up between 1977 and 2004. Median age at first hospitalisation for RA was 45 years (range 15–69). Compared with nulliparous women, parous women were at no unusual risk of RA (RR 0.96; 95% CI 0.89 to 1.04). However, compared with one-child mothers, women with two or more children had a significantly reduced risk (RR 0.84; 95% CI 0.78 to 0.90 for two-children mothers; RR 0.83; 95% CI 0.77 to 0.91 for three-children mothers) after adjusting for age at first childbirth (table 1). Women who had their first child before age 20 years were at higher risk than women who were 20–24 years old at their first childbirth (RR 1.20; 95% CI 1.12 to 1.29), and the risk declined further with increasing age at first childbirth (p for trend <0.001).

We also studied the risk of RA for a subset of 1 387 186 women born in 1955–89 for whom data on adverse pregnancy outcomes and pregnancy-related diseases were available. This analysis was based on 1648 cases of RA occurring during 23.6 million person-years of follow-up (tables 2 and 3). The median age at first RA hospitalisation was 30 years (range 15–49). After adjusting for number of children and age at birth of first child, RRs did not differ significantly between women with and without histories of spontaneous abortions, missed abortions, stillbirths, induced abortions, ectopic pregnancies or hydatidiform moles (table 2).

As shown in table 3, however, women with histories of pregnancies complicated by hyperemesis (RR 1.70; 95% CI 1.06 to 2.54), gestational hypertension (RR 1.49; 95% CI 1.06 to 2.02) or pre-eclampsia (RR 1.42; 95% CI 1.08 to 1.84) had a significantly increased risk of RA compared with women without such histories. Indeed, the risk of RA increased significantly with increasing number of pregnancies complicated by gestational hypertension or pre-eclampsia (p for trend = 0.003). After taking the number of such complicated pregnancies into account, the risk of RA did not differ with time since the most recent complicated pregnancy.

Men

In the cohort of 2 243 840 men born in 1935–89, 3041 cases of RA occurred during 45.3 million person-years of follow-up. The median age at first hospitalisation for RA was 46 years (range 15–69). Overall, the risk of RA was unrelated to whether a man had children or not (table 1), but the risk of RA was higher among men who fathered their first child at a young age (p for trend <0.001). Men were at a transiently reduced risk of RA in the first 2 years following the birth of the most recent child (RR 0.72; 95% CI 0.57 to 0.91) compared with the period 10–14 years after. Men fathering children following pregnancies complicated by hyperemesis (RR 1.10; 95% CI 0.55 to 1.93; n = 10 RA cases), gestational hypertension (RR 0.92; 95% CI 0.63 to 1.30; n = 30) or pre-eclampsia (RR 0.86; 95% CI 0.59 to 1.20; n = 32) were at no unusual risk of RA compared with other men.

Discussion

With 7017 hospitalised cases of RA among women and 3041 cases among men, our study is by far the largest to investigate possible associations between reproductive factors and risk of RA. Also, our study is the first to carry out parallel analyses of the impact of children in women and men, an approach which helps to distinguish between biological and social causes of the observed associations. Salient observations include an increased RA risk in one-child mothers, in women and men who became parents at a young age, and in women with pregnancies complicated by hyperemesis, gestational hypertension or pre-eclampsia.

Controversy surrounds the possible impact of parity. In some studies, women without children were found to be at increased risk of RA8 9 17 20 while others found no association.7 10 11 13 Our study clearly supports the view that nulliparity is not a marker of increased RA risk. However, we found that one-child mothers were at a significantly higher risk of RA than women with two or more children. Theoretically, women with one child only might have experienced fertility or pregnancy problems more frequently than women with two or more children. If so, the finding that one-child mothers are at an increased risk of RA is compatible with studies showing increased frequencies of subfertility before the onset of RA.6 12 14 17 21

In contrast to most previous studies, we found a significant inverse association between age at birth of first child and subsequent risk of RA among both women and men.7 9 11 13 17 Specifically, there was a significant 3% and 4% decrease in the risk of RA with each 1-year increase in age at first childbirth among women and men, respectively. The similar finding for men suggests that this association may not be related to pregnancy. Instead, socioeconomic factors29 or lifestyle factors (such as tobacco smoking) that prevail in lower socioeconomic strata of society may be involved.30 We were able to adjust for socioeconomic differences that follow marital status, but we had no information about tobacco consumption in our study cohort. Thus, a possible confounding influence of tobacco consumption, which has a negative effect on fertility31 while conferring an increased risk of RA,15 cannot be excluded.

We found no significant association between pregnancy terminations—whether by induced abortions or idiopathic pregnancy losses—and subsequent risk of RA. These observations are in agreement with negative findings in most previous studies that examined pregnancy losses as potential risk factors for RA.6 9 12 13 19

To the best of our knowledge, our study is the first to examine the possible role of hyperemesis, gestational hypertension and pre-eclampsia in relation to RA risk. We observed significant increases in the risk of RA for all three conditions and, among women with pregnancies complicated by gestational hypertension or pre-eclampsia, the risk was proportional to the number of such complicated pregnancies.

Nausea and vomiting are common in the first trimester of pregnancy and probably attributable to the changing hormonal milieu of pregnancy.32 However, only the few cases severe enough to require hospitalisation with a diagnosis of hyperemesis would be registered in our records. Hospitalisation for hypertensive complications of pregnancy, including pre-eclampsia, usually occurs in the third trimester.33 We are not aware of a common link between hyperemesis and hypertensive pregnancy complications that would explain why both would be related to subsequent RA risk. There may be both medical and social causes leading to hospitalisation. Generally, pregnancy complications have been reported to be more frequent in women from lower socioeconomic strata,34 35 so the observed excess risk associated with hyperemesis and hypertensive pregnancy complications might in part reflect the influence of uncontrolled socioeconomic factors. However, we did not see a similar significant relationship among men whose children’s mothers were hospitalised for hyperemesis, gestational hypertension or pre-eclampsia, as might be expected for a socioeconomic explanation.

When considering possible biological mechanisms for the increased risk of RA following hyperemesis and hypertensive pregnancies, it should be recalled that immunological abnormalities may occur even decades before a clinical diagnosis of RA.36 37 38 It is conceivable that subclinical immunological activity in women destined to develop RA might somehow disturb these women’s immune adaptation to pregnancy in ways that result in hyperemesis, gestational hypertension or pre-eclampsia. Moreover, women with already established rheumatic disease are also more likely to develop pre-eclampsia than non-rheumatic women.39

Another mechanism to be considered is fetal microchimerism. Fetal cells may persist in maternal circulation for decades but are usually cleared soon after delivery.40 41 The concept that circulating immune competent fetal cells may give rise to autoimmune diseases has gained increasing attention in recent years.22 Of relevance in the present context, studies have shown that women with pre-eclampsia have considerably higher concentrations of circulating fetal cells.26 Thus, our finding of an increased risk of RA following pre-eclamptic pregnancies might be viewed as favouring a microchimeric mechanism. However, we saw no clear relationship between parity and RA risk, whereas one might theoretically expect that the risk of RA would be lowest in nulliparous women and increase directly with the number of children. Theoretically, the impact of fetal microchimeric cells on the risk of RA might depend on the mother’s ability to clear the fetal cells, a feature which—for as yet unclear reasons—might differ between women with hyperemesis, gestational hypertension, or pre-eclampsia and women who do not experience such pregnancy complications.

Despite its large size, its cohort design and its population-based nature, our study has a number of limitations that need attention. Overall, the cohort comprised women and men in the age span 15–69 years but, for the analyses of pregnancy complications, we only had data for women aged 15–49 years. Thus, our findings may not necessarily apply to RA with onset in older age groups. Also, to identify RA outcomes in our study, we relied on first inpatient hospital contacts for RA among our cohort members because the validity of such inpatient records has been found to be >70% in the Danish National Patient Registry.42 However, it should be noted that, for some patients with RA, an inpatient hospital contact may occur rather late in the course of the disease. For the 10 058 cohort members who developed RA in our study, we do not know when the initial immunological changes took place that eventually led to the diagnosis of RA and to admission to hospital. For the entire cohort, the median age at first RA hospitalisation was 45 years for women, which is well beyond the age when most women have completed their reproduction. However, for the subset of women born after 1955 (ie, the cohort for which we had information about pregnancy losses and pregnancy complications), the median age at first RA hospitalisation was 30 years. We cannot exclude the possibility that, for some of these women, the disease process leading to RA hospitalisation might have preceded the recorded reproductive events. Furthermore, an unknown proportion of patients with RA were not identified owing to milder disease that did not require hospitalisation. Our findings may not therefore necessarily apply to patients with mild RA.

We believe that routine records of live births and stillbirths are close to 100% valid and complete in the Danish health registers, but formal validation of other reproductive variables has been undertaken only for pre-eclampsia.42 43 It is therefore noteworthy that our significant findings of an increased risk of RA associated with hyperemesis, gestational hypertension and pre-eclampsia occurred despite the possible influence of non-specific data errors that would tend to favour null findings.

Prevalent cases of RA could potentially have influenced the risk estimations. For example, women diagnosed with RA before the start of the Danish National Patient Registry in 1977 might have had their disease recorded for the first time in connection with one of the studied reproductive events at some point after 1977. However, such bias is unlikely to have produced the observed significant association with hyperemesis, gestational hypertension and pre-eclampsia because the risk of RA in the first 2 years after the most recent complicated pregnancy was negligible.

Finally, we had only a limited number of variables available to control for potential confounding by socioeconomic factors. We adjusted for marital status, but other confounders might have impacted on the observed risk associations between reproductive factors and the risk of RA. For instance, to the extent that tobacco smoking is linked to the pregnancy factors under study, RRs might be confounded. However, the observed increased risk of RA in women with a history of pre-eclampsia cannot plausibly be explained by unmeasured tobacco consumption because tobacco smoking, an established risk factor for RA,15 protects against pre-eclampsia.44 To partially account for the lack of ability to control for socioeconomic factors, we carried out all analyses of the impact of live births in both women and men. For instance, by finding a statistically significant association between young age at first childbirth and the risk of RA in both men and women, it seems likely that the underlying mechanism among women is not a physiological effect of pregnancy. However, a possible influence of socioeconomic factors and other unmeasured confounders cannot be excluded.

Three main results emerged from this study. Women who had given birth to only one child and women and men who had their first child at a young age were at an increased risk of RA, and women with a history of pregnancies complicated by hyperemesis, gestational hypertension or pre-eclampsia were at an increased risk of RA of 42–70%. Socioeconomic factors may be responsible, at least in part, for the link with young age at first childbirth. If they represent genuine new clues to the aetiology of RA, the novel associations with pregnancy complications might be explained by a deterioration in immune adaptability to pregnancy in women predisposed to RA or by increased fetal microchimerism. Confirmatory studies are clearly warranted.