Do breast-feeding and other reproductive factors influence future risk of rheumatoid arthritis?: Results from the Nurses' Health Study
To explore the contribution of female hormonal factors occurring prior to the onset of rheumatoid arthritis (RA), such as age at menarche, parity, age at first birth, breast-feeding, use of oral contraceptives (OCs), irregular menstrual cycles, and postmenopausal hormone (PMH) use, to the subsequent development of RA in a large female cohort.
We studied female reproductive and hormonal risk factors for RA in a cohort of 121,700 women enrolled in the longitudinal Nurses' Health Study. The diagnosis of incident RA (between 1976 and 2002) in 674 women was confirmed by a connective tissue disease screening questionnaire and blinded medical record review for American College of Rheumatology criteria. Sixty percent of the patients with RA were rheumatoid factor positive. The relationship between potential risk factors, including age, age at menarche, parity, age at first birth, total lifetime history of breast-feeding, use of OCs, and irregular menstrual cycles and the multivariate-adjusted risk of RA was estimated using Cox proportional hazards models.
Using a multivariate model that adjusted for age, body mass index, smoking, parity, and other hormonal factors, we observed a strong trend for decreasing risk of RA with increasing duration of breast-feeding (P for trend = 0.001). For women who breast-fed (compared with parous women who did not breast-feed), the risk ratios (RRs) and 95% confidence intervals (95% CIs) were as follows: breast-feeding for ≤3 total months, RR 1.0 (95% confidence interval [95% CI] 0.8-1.2); for 4–11 total months, RR 0.9 (95% CI 0.7–1.1); for 12–23 total months, RR 0.8 (95% CI 0.6–1.0); and for ≥24 total months, RR 0.5 (95% CI 0.3–0.8). Very irregular menstrual cycles were associated with an increased risk of RA (RR 1.4, 95% CI 1.0–2.0). Age at menarche ≤10 years was associated with an increased risk of seropositive RA (RR 1.6, 95% CI 1.1–2.4) but not significantly associated with risk of RA. Parity, total number of children, age at first birth, and OC use were not associated with an increased risk of RA in this cohort.
In this large cohort, breast-feeding for >12 months was inversely related to the development of RA. This apparent effect was dose-dependent, with a significant trend toward lower risk with longer duration of breast-feeding. Irregular menstrual cycles and earlier age at menarche increased the risk of RA. Other reproductive hormonal factors were not associated with RA risk.
There is epidemiologic, clinical, and experimental evidence implicating hormones in the incidence and clinical expression of rheumatoid arthritis (RA). RA is 2–4 times more likely to develop in women than in men (1, 2), perhaps due to a protective effect of androgens in men (3–5). However, results of epidemiologic studies of various reproductive factors are inconsistent. In women, RA frequently develops at times when sex steroid hormone levels are changing, such as during the postpartum and perimenopausal periods (6–9). A case–control study among women in whom RA developed during the postpartum months after their first pregnancy demonstrated an increased risk of RA among breast-feeding women compared with population-based postpartum controls (10), but 2 other case–control studies showed no significant risk of RA associated with breast-feeding (11, 12). In contrast, a protective effect of breast-feeding was suggested in a population-based cohort study in Norway, in which total time of lactation was associated with decreased mortality from RA, with an approximate dose-response relationship (13), and in a cohort study of elderly-onset RA (14).
In most studies (11, 15–22), use of oral contraceptives (OCs) is protective. However, several studies have shown no association between OCs and the risk of RA (23–27). Reports about the effect of other reproductive factors (e.g., age at menarche and age at first birth) on the development of RA have also been inconclusive (14, 27, 28). Several studies have shown a peak incidence of RA in perimenopausal (6) or early postmenopausal women (29), suggesting an influence of perimenopausal hormonal changes; however, results of studies of the risk associated with use of exogenous hormones are inconsistent (19, 24, 27, 30–32).
Thus, we sought to explore the contribution of female hormonal factors occurring prior to the onset of RA, such as age at menarche, parity, age at first birth, breast-feeding, OC use, irregular menstrual cycles, and postmenopausal hormone use and the subsequent development of RA in a large female cohort.
PATIENTS AND METHODS
The Nurses' Health Study (NHS) is a prospective cohort of 121,700 female nurses, ages 30–55 years in 1976, when the study began. Information is collected from the subjects via biennial questionnaires regarding diseases, lifestyle, and health practices. Overall, we have maintained >90% followup of the original cohort (33). All aspects of this study were approved by the Partners HealthCare Institutional Review Board.
Identification of RA.
From 1976 until 1982, participants self-reported a diagnosis of RA or other connective tissue diseases (CTDs) including systemic lupus erythematosus, mixed connective tissue disorder, scleroderma, polymyositis, dermatomyositis, or Sjögren's syndrome, in a write-in section of the questionnaire. Since 1982, participants have been specifically asked whether RA was diagnosed by a physician. For the purposes of this study, we contacted 11,674 women reporting RA and 1,681 women reporting other CTDs on any of the biennial questionnaires from 1976 until 2002, to request permission to review their medical records and to request that they complete a connective tissue disease screening questionnaire (CSQ), which includes 6 questions pertaining to symptoms of RA (34).
In total, 9,031 (77%) of the 11,674 subjects with self-reported RA responded to the mailings (n = 5). After excluding subjects who denied having the diagnosis of RA (n = 1,974), those in whom RA was diagnosed before 1976 (n = 466), those who denied permission for record review (n = 855), and those with negative results for RA symptoms on the CSQ (n = 3,042), we requested medical records from 2,797 women with RA or CTD self-reports and positive CSQ results, and obtained 2,674 records with adequate information (96%). Two rheumatologists trained in chart abstraction independently conducted a medical record review, examining the charts for the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) diagnostic criteria for RA (35). Subjects for whom 4 of the 7 diagnostic criteria were documented in the medical record were considered to have definite RA.
We confirmed 789 cases of definite RA diagnosed between 1976 and 2002, for a case confirmation rate of 29% of the medical records reviewed and 7% of the original self-reports. This is nearly identical to the self-reported RA case confirmation rate of 6% reported for the Iowa Women's Health Study, another large, prospective, female cohort study (36).
Population for analysis.
For all analyses, we excluded prevalent cases of RA as of June 1976, women who reported RA but in whom the diagnosis was not confirmed by medical record review, and nonresponders to the supplementary questionnaire on RA. We also excluded women who reported any cancer (except nonmelanomatous skin cancer) at baseline or during followup, because a cancer diagnosis and treatment can affect behaviors such as hormone use, smoking, or breast-feeding. Thus, the final group studied included 104,642 women who were followed up from 1976 until 2002; 674 cases of incident RA were identified among those women.
Information on reproductive factors and potential confounding variables.
All exposure information was self-reported on the mailed questionnaires, which were administered every 2 years since 1976. Data on age at menarche were collected in 1976 and categorized as ≤10 years, 11 years, 12 years, 13 years, and ≥14 years. Information on age at first birth was collected in 1976 and was categorized as <20 years, 20–24 years, 25–29 years, and >29 years. Parity, categorized as 0, 1, 2, 3, or ≥4 children, was assessed from 1976 through 1984 (after which the cohort was ages 38–63 years, and childbirth was uncommon).
Information on the use of OCs was collected from 1976 through 1982 and categorized as never, current, or past use (after which the cohort was ages 36–61 years, and current use was rare). In this cohort, there was little current use of OCs (0.8% of total person-years of exposure time was attributed to current use); therefore, OC use was categorized as never or ever. Duration of OC use was categorized as never, <5 years, or ≥5 years. Time since last use of OCs was categorized as never, <5 years, or ≥5 years. Information about total lifetime breast-feeding history, which was collected in 1986, was categorized as ≤3 months, 4–11 months, 12–23 months, and ≥24 total months.
Irregular menstrual cycles, defined as “the regularity of your natural menstrual periods between the ages 20–35 years when you were neither pregnant nor taking oral contraceptives,” were assessed in 1982, and responses were categorized as regular, usually regular, usually irregular, or very irregular. Postmenopausal hormone use was categorized as current, past, or never, based on reports from each biennial questionnaire during the followup period, with an additional category for premenopausal women. Age was categorized as <50 years, 50–54 years, 55–59 years, 60–64 years, and ≥65 years. Information on smoking was collected every 2 years and categorized as never, past, current 1–14 cigarettes/day, or current ≥15 cigarettes/day. Body mass index (BMI) was computed for each 2-year interval, using the most recent weight in kilograms divided by height in meters squared as reported in 1976 (when the mean [±SD] age of the cohort was 42.9 ± 7.2 years) and categorized as <22 kg/m2, 22–24.9 kg/m2, 25–29.9 kg/m2, 30–34.9 kg/m2, and ≥35 kg/m2.
We used Cox proportional hazards models to study the association between RA (diagnosed between age 30 years and age 79 years) and reproductive factors, adjusting for potential confounding factors, age, smoking, and BMI (37). In this type of analysis, data for each person include person-time of followup (from the date of receipt of the first questionnaire to the end of the followup period), the date of diagnosis of RA as defined in the medical record, and the date of death or loss to followup (whichever came first). Information from each 2-year questionnaire was used to analyze the risk of RA in the next 2-year analytic cycle.
The following variables were included in the Cox models as time-varying covariates and were updated at each 2-year analysis cycle: age, smoking, BMI, parity, OC use, and postmenopausal hormone use. For variables for which the question was initially asked every 2 years and then taken off the questionnaire (e.g., parity or OC use), we updated the information until the last assessment, then carried forward the last reported information throughout subsequent analytic cycles. In addition, the following variables were assessed once and then carried forward through followup: age at menarche, age at first birth, breast-feeding, and irregular menstrual cycles. Age-adjusted and multivariate analyses of age at first birth and breast-feeding history were limited to parous women.
There is a possibility that the women in whom RA developed between 1976 and 1986 may have recalled their breast-feeding history differently from those in whom RA did not develop. Therefore, the prospective association between breast-feeding and risk of RA in women ages 40–79 years was studied in 338 patients in whom RA was diagnosed after 1986, the year in which breast-feeding history was reported on the NHS questionnaire. Similarly, the prospective association between irregular menstrual cycles and risk of RA in women ages 36–79 years was studied in 494 patients in whom RA was diagnosed after 1982, the year in which menstrual cycle patterns were reported. Relative risk (RR) was calculated as the ratio of incidence rates in exposed subjects compared with unexposed subjects.
Because the focus of this study was reproductive factors, we examined characteristics of the population categorized according to the number of live-born children. Women who had more children were slightly more likely to have breast-fed for longer durations and were more likely to have received OCs than were women who had fewer children (Table 1). Otherwise, the distribution of other potential confounders was similar across categories of parity. Of 674 RA patients with a mean (±SD) age at diagnosis of 56 ± 9 years, 404 (60%) were seropositive, 203 (30%) had radiographic changes characteristic of RA, and 323 (48%) reported any steroid use on questionnaires since 1990 (Table 2). Women with postmenopausal-onset RA had similar characteristics. Most of the RA cases (82%) had been diagnosed by a member of the ACR. In 1976, when subjects were ages 30–55 years, >95% had already had their last pregnancy. The mean (±SD) time between the last pregnancy and the date on which RA was diagnosed was 25 ± 9 years.
Table 1. Age-adjusted characteristics of subjects at baseline in 1976, according to parity*
|Age, mean years||42.7||41.4||40.7||42.3||44.7|
|Body mass index, mean kg/m2||23.4||23.3||23.1||23.4||23.9|
|Smoking, %|| || || || || |
|OC use, %|| || || || || |
|Mean duration of OC use, years||3.7||3.8||4.2||3.9||3.8|
|Breast-feeding, months, %|| || || || || |
| Data missing||0||19||17||16||16|
|PMH use|| || || || || |
| Data missing||2||1||1||1||1|
Table 2. Characteristics of the patients with RA*
|Age at diagnosis, mean ± SD years||56 ± 9||60 ± 7|
|Rheumatoid factor positive†||404 (60)||258 (59)|
|Radiographic changes‡||203 (30)||125 (28)|
|Steroid-treated§||323 (48)||220 (50)|
|RA diagnosed by ACR member||555 (82)||363 (84)|
The risk of RA appeared to be increased in middle-aged women (i.e., those ages 50–64 years) (Table 3). Compared with women younger than age 50 years, the relative risk of RA in women ages 50–54 years was 1.7 (95% CI 1.4–2.1) and that in women ages 55–59 years was 1.5 (95% CI 1.2–1.9); women ages 60–64 years had a 50% elevation in risk. Rheumatoid factor–positive RA had a similar association with age.
Table 3. Age and relative risk of RA, Nurses' Health Study, 1976–2002*
|<50||189||897,338||1.0 (referent)||110||895,898||1.0 (referent)|
|50–54||161||457,508||1.7 (1.4–2.1)||106||456,078||1.9 (1.5–2.5)|
|55–59||137||424,111||1.5 (1.2–1.9)||84||422,796||1.6 (1.1–2.2)|
|60–64||95||311,606||1.5 (1.1–1.9)||52||310,887||1.4 (1.0–1.9)|
|≥65||92||343,583||1.3 (1.0–1.6)||52||345,106||1.2 (0.8–1.7)|
In age-adjusted models among parous women, total lifetime breast-feeding of 12–23 months was associated with a 30% reduction in the risk of RA (RR 0.7, 95% CI 0.6–1.0), and breast-feeding for at least 24 months was associated with a 50% reduction in risk (RR 0.5, 95% CI 0.3–0.7, P for trend = 0.001) (Table 4). After adjustment for potential confounders, including number of children, we found a significant reduction in the risk of RA for women who breast-fed at least 24 months total (RR 0.5, 95% CI 0.3–0.8, P for trend = 0.001) compared with parous women who did not breast-feed. Limiting the analysis to 404 RF-positive RA cases demonstrated a similar reduction in the risk of RA for breast-feeding ≥24 months (RR 0.6, 95% CI 0.3–1.1) among parous women (Table 5). In a multivariate model that included only parous women, we tested for potential interaction between number of children (categorized as ≤2 versus ≥3) and breast-feeding (0–12 months versus ≥13 months), but the P value for the interaction was 0.7. Because breast-feeding was assessed in 1986, we repeated the analysis, beginning followup in 1986 (n = 338 incident RA cases), and results were similar (Table 6) to the those from the combined retrospective/prospective analysis.
Table 4. Female hormonal factors and relative risk of RA, Nurses' Health Study 1976–2002*
|Age at menarche, years|| || || || |
| ≤10||46||141,867||1.3 (0.9–1.8)||1.3 (0.9–1.8)|
| 11||104||396,304||1.0 (0.8–1.3)||1.0 (0.8–1.3)|
| 12||167||643,876||1.0 (referent)||1.0 (referent)|
| 13||204||747,701||1.1 (0.9–1.4)||1.1 (0.9–1.3)|
| ≥14||146||485,100||1.1 (0.9–1.4)||1.1 (0.9–1.4)|
|Age at first birth, years§|| || || || |
| <20||60||231,478||1.0 (referent)||1.0 (referent)|
| 20–24||291||1,018,459||1.1 (0.8–1.4)||1.1 (0.8–1.4)|
| 25–29||191||706,845||1.0 (0.7–1.3)||1.0 (0.7–1.4)|
| >29||60||279,623||0.8 (0.6–1.2)||0.8 (0.5–1.1)|
|Parity|| || || || |
| 0||60||165,031||1.0 (referent)||1.0 (referent)|
| 1||44||181,054||0.7 (0.5–1.0)||0.8 (0.5–1.2)|
| 2||174||675,096||0.7 (0.5–1.0)||0.8 (0.5–1.1)|
| 3||170||655,544||0.7 (0.5–1.0)||0.7 (0.5–1.1)|
| ≥4||217||725,770||0.8 (0.6–1.1)||0.8 (0.5–1.2)|
|Parous¶|| || || || |
| Nulliparous||60||165,031||1.3 (1.0–1.8)||1.3 (0.9–1.9)|
| Parous||605||2,237,464||1.0 (referent)||1.0 (referent)|
|Breast-feeding§|| || || || |
| None||222||695,882||1.0 (referent)||1.0 (referent)|
| ≤3 months||144||427,650||1.0 (0.9–1.3)||1.0 (0.8–1.2)|
| 4–11 months||110||413,971||0.8 (0.7–1.1)||0.9 (0.7–1.1)|
| 12–23 months||54||234,377||0.7 (0.6–1.0)||0.8 (0.6–1.0)|
| ≥24 months||19||131,484||0.5 (0.3–0.7)#||0.5 (0.3–0.8)#|
|Menstrual periods|| || || || |
| Very regular||340||1,146,045||1.0 (referent)||1.0 (referent)|
| Usually regular||127||418,417||1.0 (0.8–1.2)||1.0 (0.8–1.2)|
| Usually irregular||67||200,389||1.1 (0.9–1.5)||1.1 (0.8–1.4)|
| Very irregular||34||74,948||1.5 (1.1–2.2)||1.4 (1.0–2.0)|
|OC use|| || || || |
| Never||354||1,255,476||1.0 (referent)||1.0 (referent)|
| Ever||304||1,058,766||1.1 (0.9–1.3)||1.1 (0.9–1.3)|
|Duration of OC use|| || || || |
| Never||354||1,229,156||1.0 (referent)||1.0 (referent)|
| <5 years||186||682,020||1.0 (0.9–1.2)||1.1 (0.9–1.3)|
| ≥5 years||99||348,321||1.0 (0.8–1.3)||1.0 (0.8–1.3)|
|PMH use**|| || || || |
| Never||153||459,330||1.0 (referent)||1.0 (referent)|
| Past||167||439,345||1.2 (0.9–1.4)||1.3 (1.0–1.6)|
| Current||86||250,876||1.1 (0.8–1.4)||1.0 (0.8–1.3)|
Table 5. Female hormonal factors and relative risk of seropositive RA among 404 RA patients, Nurses' Health Study 1976–2002*
|Age at menarche|| || || || |
| ≤10 years||33||141,880||1.6 (1.1–2.4)||1.6 (1.1–2.4)|
| 11 years||66||396,330||1.1 (0.8–1.5)||1.2 (0.8–1.6)|
| 12 years||95||643,938||1.0 (referent)||1.0 (referent)|
| 13 years||120||747,770||1.1 (0.8–1.4)||1.1 (0.8–1.4)|
| ≥14 years||85||485,161||1.2 (0.9–1.6)||1.1 (0.9–1.5)|
|Parity‡|| || || || |
| Nulliparous||33||165,055||1.2 (0.9–1.7)||1.0 (0.6–1.6)|
| Parous||367||2,237,672||1.0 (referent)||1.0 (referent)|
|Breast-feeding§|| || || || |
| None||128||695,966||1.0 (referent)||1.0 (referent)|
| ≤3 months||86||427,703||1.1 (0.8–1.4)||1.1 (0.8–1.3)|
| 4–11 months||66||414,009||0.9 (0.6–1.2)||0.9 (0.7–1.2)|
| 12–23 months||38||234,394||0.9 (0.6–1.3)||1.0 (0.6–1.4)|
| ≥24 months||13||131,488||0.6 (0.3–1.0)¶||0.6 (0.3–1.1)|
|Menstrual periods|| || || || |
| Very regular||201||1,146,170||1.0 (referent)||1.0 (referent)|
| Usually regular||77||418,459||1.0 (0.8–1.4)||1.0 (0.8–1.4)|
| Usually irregular||43||200,413||1.2 (0.9–1.7)||1.2 (0.9–1.7)|
| Very irregular||18||74,960||1.4 (0.8–2.2)||1.3 (0.8–2.1)|
|Oral contraceptive use|| || || || |
| Never||219||1,229,271||1.0 (referent)||1.0 (referent)|
| Ever||175||1,060,506||1.0 (0.8–1.2)||1.0 (0.8–1.2)|
Table 6. Prospective association between breast-feeding and risk of RA and between irregular menstrual cycles and risk of RA*
|Breast-feeding†|| || || || |
| None||141||415,999||1.0 (referent)||1.0 (referent)|
| ≤3 months||86||256,393||1.0 (0.8–1.3)||0.9 (0.7–1.2)|
| 4–11 months||58||247,977||0.7 (0.5–0.9)||0.7 (0.5–1.0)|
| 12–23 months||38||141,135||0.8 (0.6–1.1)||0.8 (0.6–1.2)|
| ≥24 months||13||78,950||0.5 (0.3–0.9)‡||0.5 (0.3–0.9)§|
|Menstrual periods¶|| || || || |
| Very regular||272||878,434||1.0 (referent)||1.0 (referent)|
| Usually regular||101||321,128||1.0 (0.8–1.2)||1.0 (0.8–1.3)|
| Usually irregular||48||154,112||1.0 (0.7–1.3)||1.0 (0.7–1.3)|
| Very irregular||22||57,762||1.2 (0.8–1.8)||1.2 (0.8–1.8)|
Very irregular menstrual cycles at ages 20–35 years were associated with an increased risk of RA (n = 674 RA cases) in both age-adjusted (RR 1.5, 95% CI 1.1–2.2) and multivariate models (RR 1.4, 95% CI 1.0–2.0). Because this variable was assessed in 1982, we also repeated this analysis, beginning followup in 1982 (n = 494 incident RA cases), and the results were not significant (Table 6). In addition, OC use (defined as never and ever as a time-varying covariate) was not associated with the risk of RA; neither duration of OC use nor time since last use of OCs was associated with the risk of RA. To examine whether OC use that occurred closer to the time of diagnosis might be associated with the risk of RA, we reexamined the data using time-varying information on OC use from 1976 to 1982 and limited the followup to 323 incident RA cases diagnosed from 1976 to 1986 (when women were ages 30–65 years) (data not shown). The results were similar to those from the primary analysis. Early age at menarche was associated with a modest increased risk of RA in both age-adjusted (RR 1.3, 95% CI 0.9–1.8) and multivariate (RR 1.3, 95% CI 0.9–1.8) models (Table 4) but had a stronger association with seropositive RA (RR 1.6, 95% CI 1.1–2.4) (Table 5). Other reproductive factors, including number of children and age at first birth, were not associated with the risk of RA in this cohort.
Cox proportional hazards models limited to 440 women with RA diagnosed when they were postmenopausal demonstrated no significant association between current use of postmenopausal hormone use and risk of RA. The relative risk of RA associated with current use of hormones by postmenopausal women was 1.0 (95% CI 0.8–1.3), and the relative risk associated with past use was 1.3 (95% CI 1.0–1.6).
Breast-feeding was associated with a slightly lower risk for the development of RA among parous women in this large cohort study. Long duration of total lifetime breast-feeding suggested a dose-response effect with RA risk. Very irregular menstrual cycles between the age of 20 years and 35 years were associated with a modest increase in risk of RA. Earlier age at menarche was associated with an increased risk of developing seropositive RA.
Epidemiologic data suggest that in women, conditions associated with excess female hormones such as estrogen and/or progesterone may be protective against the development of RA. Women are at decreased risk of developing RA during pregnancy, when estradiol and progesterone levels are high. The 12-month postpartum period, particularly the first 3 months (when hormone levels fall dramatically), appears to represent a period of increased risk (9). Nonetheless, the association between exogenous estrogen use and RA is unclear. Use of OCs decreases risk in most studies (11, 15–22) but not in others (23–27). Our results do not support an association between OCs and risk of RA in this cohort of women who were followed up between the ages of 30 years and 79 years. In our analysis, neither ever use of OCs nor duration of use was associated with RA, even during the years closest to the time of use; however, we lacked detailed information on dose and dates of OC use prior to the start of the cohort study and had low power to detect associations with current use. Thus, we were unable to examine whether OC use in the 1960s, when the dose of estrogen was high, was protective, as demonstrated by a recent study (15).
In most (19, 38–43) but not all (44) previous studies, nulliparity was a risk factor for developing RA. However, we did not observe a significant association in this cohort after adjusting for breast-feeding and other reproductive factors, nor did we observe any interaction between number of children and breast-feeding. It is possible that the protective effect of parity in other studies is attributable to unmeasured confounding by breast-feeding among parous women.
Results of epidemiologic studies suggest that the incidence of RA increases with age (1, 45), with a peak incidence around the time of menopause (6, 29). If the perimenopausal risk of RA is increased due to falling estrogen levels, then estrogen therapy should reduce the risk, although results of studies are mixed (14, 19, 24, 27, 29–32). We confirmed that the risk of RA increases with age and demonstrated a peak risk at the time of menopause (age 50–54 years). However, exogenous estrogen therapy among postmenopausal women did not reduce the risk of RA in this cohort.
We identified a new risk factor for RA, irregular menstrual cycles, which, to our knowledge, has not previously been studied. It is possible that this is a chance finding among exploration of many reproductive risk factors. Alternatively, in premenopausal women, irregular menses may be a marker of polycystic ovary syndrome, a condition associated with excess androgen. Interestingly, androgens have been proposed to protect against RA (46), based on limited evidence—an increased incidence of RA in women compared with men, reduced levels of androgens in RA patients (3, 4), and the beneficial action of testosterone on disease activity in established RA (47, 48). A recent cohort study demonstrated an increased risk of elderly-onset RA in women with polycystic ovary syndrome (14). This area clearly needs additional study.
We demonstrated an inverse relationship between breast-feeding and the risk of RA. The biologic mechanism for this is unclear. We hypothesized that breast-feeding would increase the risk of RA, based on studies that demonstrated enhancement of the Th1 response by prolactin, the primary hormone that is elevated during breast-feeding. However, our findings demonstrated the opposite. Although there is evidence that prolactin is a potent immunostimulatory molecule (49–55), other studies suggest immunosuppressive effects (56–58). Hormones other than prolactin could be involved in the reduction in risk associated with breast-feeding. Patients with RA have dysregulation of the hypothalamic–pituitary axis (HPA) (59) and inappropriately normal cortisol levels despite high levels of inflammation mediators that normally stimulate the HPA axis and cortisol production (60, 61). A recent study in postmenopausal women demonstrated elevation of cortisol levels in women who had previously breast-fed for 12 months or longer (62), suggesting a potential mechanism for reduction in risk for RA via elevated cortisol levels. During pregnancy, progesterone is antiinflammatory (63–67), and lymphocytes develop progesterone receptors at a rate that continues to increase throughout breast-feeding (66). Breast-feeding might reduce RA risk via prolonged antiinflammatory effects from progesterone.
Our data suggest a prolonged inverse relationship with breast-feeding, because the mean time since the last pregnancy among women with RA was 25 years. Therefore, protopathic bias, in which the inverse relationship might be partially explained by changes in behavior due to early symptoms of RA, would be very unlikely. Nonetheless, the decreased risk of RA associated with breast-feeding could be attributable to unmeasured confounding.
Case–control studies of the association between breast-feeding and the risk of RA have had conflicting results. One study in which women with postpartum-onset RA were compared with healthy postpartum controls demonstrated a significantly higher prevalence of breast-feeding in the patients with RA (10). However, a genetic analysis from this study (68) suggested that the reported increase may have been attributable to confounding by HLA–DR4 status. In contrast, another case–control study of mothers with RA compared with control mothers demonstrated no increased frequency of breast-feeding in those with RA, although there was an association of breast-feeding with more severe disease (11). Two prospective studies of breast-feeding and RA had findings similar to ours (13, 14). Again, further study of this issue is warranted.
Strengths of our study include the large number of incident cases of RA, the repeated assessment of exposures, prospective assessment of most exposures, and the long followup period. Limitations include the assessment of breast-feeding history and menstrual cycle regularity several years into followup, which raises the possibility of recall bias. Although our analyses of breast-feeding history and menstrual cycle regularity combined retrospective and prospective data, when we eliminated RA cases diagnosed before the report of these 2 exposures, the results were similar, suggesting that recall bias was not a problem. There was an attenuation of the observed effects of breast-feeding and menstrual cycle irregularity in analyses limited to seropositive RA, which is contrary to our prior expectations, perhaps due to unmeasured confounders among women at risk for seropositive RA or due to the smaller number of cases in the analysis.
The validation of self-reported RA through medical record review rather than by physical examination is a potential weakness of the study. However, 82% of the RA cases were diagnosed by ACR members, which adds support to the validity of the diagnoses. There is potential for misclassifying RA cases as non-cases when relying solely on medical record documentation. Therefore, those women who self-reported RA or other CTDs in whom the diagnosis of RA was not confirmed by medical record review were excluded from the analyses. Additional limitations include the use of self-reported exposure data and the possibility of unmeasured confounding (e.g., hormone levels). Finally, although participants are not a random sample of women living in the US, it is unlikely that the basic biologic relationships among these women will differ from those among women in general.
In conclusion, we observed that breast-feeding was inversely associated with the risk of RA, with a strong trend for decreasing risk of RA with increasing duration of breast-feeding, and that early age at menarche is positively associated with seropositive RA. In addition, we identified a novel risk factor, irregular menstrual cycles, that increased the risk of subsequent RA. The risk of RA increases at ages 50–54 years. Postmenopausal hormone use, however, does not protect against RA. One interpretation of the data is that the inability to breast-feed and irregular menses are markers for a deranged hypothalamic axis in subjects prone to the development of RA. These findings suggest avenues for further research into the hormonal mechanisms involved in RA, because the complex relationships between RA and reproductive hormones clearly warrant further study (3, 46).
We thank the dedicated nurses in the NHS, who have now participated in the study for more than 26 years. We also thank Christine Jones for her efforts to obtain medical record information, Gideon Aweh for his programming expertise, and Karen Costenbader for reviewing the manuscript.