Previous studies have evaluated the correlation between rheumatoid arthritis (RA) risk and pregnancy history, with conflicting results. Fetal cells acquired during pregnancy provide a potential explanation for modulation of RA risk by pregnancy. The present study was undertaken to examine the effect of parity on RA risk.
We examined parity and RA risk using results from a population-based prospective study in Seattle, Washington and the surrounding area and compared women who were recently diagnosed as having RA (n = 310) with controls (n = 1,418). We also evaluated the distribution of parity in cases according to HLA genotype.
We found a significant reduction of RA risk associated with parity (relative risk [RR] 0.61 [95% confidence interval 0.43–0.86], P = 0.005). RA risk reduction in parous women was strongest among those who were younger. Most striking was that RA risk reduction correlated with the time that had elapsed since the last time a woman had given birth. RA risk was lowest among women whose last birth occurred 1–5 years previously (RR 0.29), with risk reduction lessening progressively as the time since the last birth increased (for those 5–15 years since last birth, RR 0.51; for those >15 years, RR 0.76), compared with nulliparous women (P for trend = 0.007). No correlation was observed between RA risk and either age at the time a woman first gave birth or a woman's total number of births. Among cases with the highest genetic risk of RA (i.e., those with 2 copies of RA-associated HLA alleles), a significant underrepresentation of parous women versus nulliparous women was observed (P = 0.02).
In the present study, there was a significantly lower risk of RA in parous women that was strongly correlated with the time elapsed since a woman had last given birth. While the explanation for our findings is not known, HLA-disparate fetal microchimerism can persist many years after a birth and could confer temporary protection against RA.
It has been suspected that pregnancy and childbirth may influence the risk of developing rheumatoid arthritis (RA), but epidemiologic findings have been inconsistent. Both case–control studies and cohort studies (1–11) have evaluated gravidity and/or parity as etiologic factors, with some showing a reduced risk of RA associated with pregnancy or childbirth and others showing no significant association (Table 1). Elevations in levels of sex hormones during pregnancy revert very rapidly after parturition, and it would be difficult to explain a protective effect of parity on this basis. It has been recently determined that fetomaternal cell trafficking during pregnancy results in the long-term persistence of small numbers of cells in mothers and children, referred to as fetal and maternal microchimerism, respectively (12, 13). Fetal microchimerism, which persists in the mother, offers a potential explanation for an effect of parity on RA risk in women.
Table 1. Studies of gravidity and/or parity and risk of RA*
Author, year (ref.)
ACR = American College of Rheumatology (formerly, the American Rheumatism Association); ICD = International Classification of Diseases; RF = rheumatoid factor.
Study showed a statistically significant association between parity and/or gravidity and risk of rheumatoid arthritis (RA).
Danish national registry, ages 15–69 at first inpatient hospitalization for RA
In the current study, we analyzed pregnancy history and RA risk using detailed data from a prospective case–control study of newly diagnosed RA in women in Seattle and the surrounding area. We reasoned that a lesser effect of parity might be observed among women who were older and that the contradictory results of prior studies might be explained by considering both a woman's age and the time that elapsed since previous pregnancies. We also investigated whether parity might be especially beneficial for women who are at greater risk of RA because they carry 2 copies of RA-associated HLA risk alleles. Thus, the underlying hypothesis leading to the current study was that pregnancy could provide protection against RA, but that this effect, like that of a vaccine, diminishes over time.
PATIENTS AND METHODS
The current study derives from a prospective population-based case–control study of newly diagnosed RA in women living in King County, WA, or receiving medical care at Group Health Cooperative, a large, Seattle-based prepaid health plan. That study recruited women (ages 18–64 years) using a surveillance system involving rheumatologists, family physicians, and internists to identify RA cases diagnosed from November 1986 through February 1991 and was designed to evaluate risk of RA in relation to prior or current oral contraceptive use. Each potential case was evaluated in person by a board-certified rheumatologist, who obtained standardized information on clinical history and conducted a standardized joint examination. A rheumatoid factor test was performed at the University of Washington Clinical Immunology Laboratory, and a review of outpatient medical records was completed for each participant. Of potential cases identified, 93% agreed to the study examination and personal interview, and of these, 87% were found to be eligible. The initial study identified 349 women whose disease met the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1958 criteria for definite or probable RA (14). The current study was limited to 319 of these women whose disease met the ACR 1987 revised criteria for RA (15) following a review of the rheumatologist's physical examination, rheumatoid factor test result, and medical record abstracts.
Control subjects in King County were identified using the Mitofsky-Waksberg method of random digit telephone dialing (16). Screening for eligibility was successfully completed for 96% of the households contacted. Group Health Cooperative control subjects were selected using age-stratified random sampling from the plan's master enrollment file. Participation rates among controls were 78% both in King County and at Group Health Cooperative, and 1,457 of 1,541 willing participants met the eligibility criteria.
Each subject was interviewed in person about events prior to a specified reference date that was defined for cases as the date of the first physician visit for symptoms ultimately diagnosed as RA. The first visit to a physician for joint symptoms was utilized as the reference date for RA onset because a diagnosis of RA is sometimes delayed and, conversely, patient recall of first joint symptoms may precede RA onset. Each control subject was assigned a reference date that was selected at random from the distribution of reference dates among the cases. Detailed information regarding demographic characteristics and reproductive history was obtained by trained interviewers. Four cases and 38 control subjects who were pregnant at the reference date and 5 cases and 1 control with missing relevant data points were excluded from the present study, yielding a total of 310 cases and 1,418 controls for analysis.
HLA genotyping of RA cases.
DNA was extracted either from heparinized whole blood or from peripheral blood mononuclear cells isolated using Ficoll-Hypaque gradient centrifugation. For a few cases, DNA was extracted from a mouth swish sample or from serum. DNA-based typing was used to identify specific HLA–DRB1 alleles, as described previously (17).
The primary outcome measure for analysis was disease status (RA case versus control). Logistic regression models were used to estimate the association between parity and disease status, while adjusting for confounding factors. The resulting odds ratio estimates were interpreted as relative risks (RRs) because of the rarity of RA in the population from which cases and controls were drawn, and 95% confidence intervals (95% CIs) were also calculated. Gravidity was defined as the number of pregnancies prior to the reference date, and parity was defined as the number of those pregnancies that lasted at least 20 weeks and resulted in a live birth or in stillbirth.
Factors examined as potential confounders included age at reference date, race, education level, marital status, body mass index, smoking status, oral contraceptive use (ever or never), history of spontaneous or induced abortion, and menopausal status. A factor was included as a confounder if there was a difference of ≥10% in any of the estimated coefficients of interest between the multivariable model including the factor and the model without it. The following characteristics were considered predictors of RA risk: number of children, age at the time of the first birth, and time that had elapsed since the last birth, with adjustment for other confounders as needed. Two-sided P values from regression models were derived from the Wald test. The distribution of the RA-associated HLA alleles (shared epitope) according to the number of copies was compared across parity status within RA cases using the chi-square test for trend. Analyses were performed using SAS software, version 8 (SAS Institute).
Characteristics of the cases and control participants are summarized in Table 2. The groups were similar with respect to age; the median age of both cases and controls was 43 years. Compared with controls, women diagnosed as having RA were less likely to be white, to have been married, and to have attended college.
Table 2. Characteristics of the RA cases and controls*
Cases (n = 310)
Controls (n = 1,418)
Values are the number (%). RA = rheumatoid arthritis.
Married or living as married
Divorced or separated
Highest level of education
College or graduate school
Parity of any number, compared with women who had never been pregnant, was significantly associated with a lower risk of RA (adjusted RR 0.61 [95% CI 0.43–0.86], P = 0.005) (Table 3). Gravidity without parity conferred no advantage as RA risk among women who were gravid but nulliparous did not differ significantly from that among nulligravid women. The associations of gravidity with case status were similarly negative when gravidity was subcategorized according to history of induced or spontaneous abortion (data not shown). Thus, all further analyses compared parous subjects with the combined group of nulliparous subjects, regardless of gravidity.
Table 3. Adjusted RRs of RA by gravidity and parity and by parity within age subgroups*
Values are the number (%). RA = rheumatoid arthritis; 95% CI = 95% confidence interval.
For all ages, relative risks (RRs) were adjusted for age, race (white versus nonwhite), and oral contraceptive use (ever versus never); within age subgroups, RRs were adjusted for race and education level (college or more versus less than college).
We next investigated whether the reduction in RA risk among parous versus nulliparous women differed according to whether disease onset occurred in reproductive or postreproductive years. RA risk was essentially halved for parous women who were younger than 45 years of age, whereas there was no significant association of parity in women 45 years and older (Table 3). The difference was not simply chronological with advancing age; a test for a linear trend in the association of parity and RA risk over age resulted in a P value of 0.12. There was limited variability in parity status among older women, as <10% of women 45 years and older were nulliparous.
Of special interest in the current study was the evaluation of RA risk in women according to the amount of time that had elapsed since a woman had last given birth. We found that RA risk was lowest among women whose last birth had been 1–5 years prior, with risk reduction progressively lessening with increasing time since the last birth. For women who had last given birth 1–5 years previously, the proportion of cases was 11 of 120 subjects (9%) (RR 0.29, P < 0.001); for women who had last given birth 5–15 years previously, the proportion of cases was 49 of 345 subjects (14%) (RR 0.51, P < 0.001); and for women who had last given birth >15 years previously, the proportion of cases was 140 of 805 subjects (17%) (RR 0.76, P = 0.18), relative to nulliparous women, in whom the proportion of cases was 97 of 406 subjects (24%) (Figure 1). A test for trend across these levels resulted in a P value of 0.007 in an adjusted model.
Although some studies have shown an increased risk of RA postpartum (18), in our study population overall, there was no significant increased or decreased risk in the year following delivery (Figure 1). Using the postpartum year as a reference and restricting analysis to parous women resulted in a similar pattern of attenuating RR with increasing years since the last pregnancy as follows: for 1–5 years since the last pregnancy, RR 0.28 (P = 0.006); for 5–15 years, RR 0.45 (P = 0.03); and for >15 years, RR 0.59 (P = 0.23).
While subject age and time since last childbirth were correlated, the association of RA risk with the time that had elapsed since the last birth persisted after adjusting for age. Moreover, the pattern among women younger than 35 years of age was similar to the pattern seen in the entire cohort; the association of parity and risk of RA was strongest among those who had last given birth 1–5 years prior (RR 0.27 [95% CI 0.11–0.67], P = 0.005) and weaker among those who had last given birth >5 years previously (RR 0.59 [95% CI 0.30–1.17], P = 0.13). In contrast, RA risk was not significantly associated with age at the time of the first birth or with the total number of births (Table 4).
Table 4. Adjusted RRs of RA by parity characteristics in parous subjects*
Although samples were not available from children of the study subjects, we were able to examine the hypothesis that parity is especially beneficial for women who are at greatest risk of RA, by conducting HLA genotyping of all RA cases and analyzing the results according to parity. Results of HLA genotyping were evaluated for women with RA according to the number of copies (whether 0, 1, or 2) of an RA-associated HLA allele. A similar sequence of the HLA–DRβ1 molecule is thought to underlie HLA-associated RA risk, referred to as the “shared epitope” (19), with 2 copies of the shared epitope conferring the greatest risk of RA.
Consistent with the hypothesis, among women with RA who had 2 copies of the shared epitope, fewer were parous (18%) than nulliparous (29%) (Table 5). The converse was observed for women with RA who had 0 copies of the shared epitope, of whom 36% were parous and 27% were nulliparous (P for trend = 0.02 for the association of shared epitope copy number and parity). Moreover, mirroring the overall study findings, the highest-risk HLA genotype was most underrepresented in parous women who had last given birth 1–5 years prior (0% had 2 copies of the shared epitope), with progressive attenuation with increasing time since the last birth (14% of those who had last given birth 5–15 years previously, and 21% of those who had last given birth >15 years previously, had 2 copies of the shared epitope) (Table 5).
Table 5. Distribution of the RA-associated HLA SE among RA cases, by copy number, parity, and time elapsed since the patient last gave birth*
Values are the number (%). RA = rheumatoid arthritis.
A test for trend in association between the number of copies of the shared epitope (SE) and parity resulted in a P value of 0.02.
A beneficial role of pregnancy in RA was initially described more than 70 years ago, when Hench reported that most women with RA who become pregnant experience an improvement of their disease during pregnancy (20). Many years later, studies first addressed the question of whether a woman's prior pregnancy history affected her risk of developing RA. Three studies reported in 1989 and 1990 showed a significant difference in RA risk according to a woman's pregnancy history. Results were expressed as an increased risk among nulliparous women but could also have been expressed as reduced risk in parous women. These retrospective case–control studies were carried out in The Netherlands, the UK, and the US (1, 3, 4). A later study in Sweden also showed a significant association (8). However, in other retrospective and prospective studies, no significant association has been demonstrated (2, 5–7, 9–11). In addition to differences in the study populations and the analyses that were conducted, in some studies the disease of the participants could not be conclusively defined as having met ACR criteria, and in other studies investigators were not able to obtain information regarding the date of RA onset, meaning that pregnancies could have occurred before or after RA and that incident RA cases might not have been identified. Analyses were conducted according to parity in some studies and according to gravidity in others.
In the current study, we found that parous women overall, compared with nulliparous women, were ∼40% less likely to have been diagnosed as having RA. The inverse association with prior pregnancy was observed only for parity; gravidity without parity did not impact RA risk. We reasoned that age for adult women can be divided into reproductive and postreproductive years and is not simply chronological as in males, and that the contradictory results of prior studies might be explained by considering a woman's age and the time that had elapsed since prior pregnancies. We found that the lower risk of RA among parous women was most evident in younger women (<45 years of age), with no significant reduction of RA risk in women older than 45. More striking was the finding that the time since the woman's last childbirth strongly and significantly correlated with RA risk, even after controlling for the mother's age. RA risk was lowest among women who had last given birth 1–5 years prior, with risk reduction progressively lessening as the time since the last birth increased. In contrast, no correlation with a woman's age at the time she first gave birth or with total number of births was observed. Moreover, these relationships of time and risk held up even among the subset of women who were <35 years old.
It is difficult to envision a mechanism by which hormonal changes could explain an effect of prior pregnancies on RA risk many years later. However, it has recently become known that the trafficking of fetal cells to the mother during pregnancy results in long-term persistence of a small number of fetal cells in the mother (as well as maternal cells in her offspring). Both adverse and beneficial effects of fetal microchimerism are currently under investigation (12, 13, 21). While the explanation for our study results is not known, persistent fetal microchimerism could potentially explain the protective effect of parity on RA risk. The absence of association with gravidity alone could also be accounted for on this basis, because the types of fetal cells transferred in early versus late gestation have been found to differ (22, 23) and/or the duration of exposure differs for gravidity without parity.
There are a number of known and putative risk factors for RA, including smoking and obesity (24), as well as prior use of oral contraceptives, breastfeeding, and irregular menstrual periods (10). HLA class II alleles encoding a similar sequence of the DRβ1 chain, referred to as the shared epitope, are known to increase the risk of RA (19). As was recently reported, if a woman acquires microchimerism that carries the shared epitope, whether fetal, maternal, or of iatrogenic origin, it might be expected to convey an increased risk of RA (25). The results of the current study suggest the converse may also occur; i.e., we hypothesize that acquisition of fetal microchimerism with RA-protective HLA alleles could reduce the risk of RA. Protection against RA has been proposed for HLA molecules that carry the sequence DERAA instead of the shared epitope at the same location of the DRβ1 chain. Interestingly, a recent study showed that the DERAA sequence was significantly underrepresented as a noninherited maternal HLA antigen among RA patients (26), an observation that also lends support to the hypothesis that microchimerism with RA-protective HLA alleles (maternal or fetal) could impact RA risk.
In conclusion, in the current study, we observed a lower risk of RA among parous women and found a significant correlation between time elapsed since a woman's last childbirth and reduction in RA risk. While any disease-protective mechanism underlying our findings is unknown, an attractive explanation is a protective effect of HLA-disparate fetal microchimerism. The additional finding that parous women were underrepresented among those who had 2 copies of the RA-associated shared epitope, and thus were at higher genetic risk of RA, is also consistent with this explanation. The diminishing strength of protection against RA over time could result from changes in maternal immunity as the mother ages, but could also occur due to aging of the fetal microchimerism as time progresses. RA is a common disease that causes substantial morbidity and disability, and the possibility that pregnancy might act like a vaccine in conferring temporary protection against RA merits further investigation.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Guthrie had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Dugowson, Voigt, Koepsell, Nelson.
Acquisition of data. Dugowson, Voigt, Koepsell, Nelson.
Analysis and interpretation of data. Guthrie, Dugowson, Voigt, Koepsell, Nelson.