Life events, caregiving, and risk of autoimmune rheumatic diseases in the Women’s Health Initiative Observational Study

Abstract

Background:

Growing evidence suggests psychosocial stressors may increase risk of developing autoimmune disease. We examined stressful life events and caregiving in relation to incident Rheumatoid Arthritis (RA) and Systemic Lupus Erythematosus (SLE) in the Women’s Health Initiative Observational Study cohort.

Methods:

The sample of post-menopausal women included 211 incident RA or SLE cases reported within 3 years after enrollment, confirmed by use of disease modifying anti-rheumatic drugs (i.e., Probable RA/SLE), and 76,648 non-cases. Baseline questionnaires asked about life events in the past year, caregiving, and social support. We used Cox regression models to calculate Hazard Ratios (HR) and 95% Confidence Intervals (CI), adjusting for age, race/ethnicity, occupational class, education, pack-years smoking and BMI.

Results:

Incident RA/SLE was associated with having 3 or more life events (e.g., age-adjusted HR 1.70; 95%CI 1.14, 2.53; p_trend=0.0026). Elevated HRs were noted for physical (2.48; 1.02, 6.04) and verbal (1.34; 0.89, 2.02) abuse (p_trend=0.0614), 2 or more interpersonal events (1.23, 95%CI 0.87, 1.73; p_trend=0.2403), financial stress (1.22; 95%CI 0.90, 1.64), and caregiving 3 or more days per week (1.25; 95%CI 0.87, 1.81; p_trend=0.2571). Results were similar excluding women with baseline symptoms of depression or moderate to severe joint pain in the absence of diagnosed arthritis.

Discussion:

Our findings support the idea that diverse stressors may increase risk of developing probable RA or SLE in post-menopausal women, supporting the need for further studies in autoimmune rheumatic diseases including childhood adverse events, life event trajectories, and modifying psychosocial and socioeconomic factors.

INTRODUCTION

Autoimmune rheumatic diseases, including the two most common systemic autoimmune diseases, rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE), affect as many as 1.6 million adults in the U.S. (1). Sometimes co-occurring, RA and SLE share risk factors including female sex and family history of RA or SLE, while RA is increased in older women and SLE predominately affects reproductive age women (2–5). Characterized by complex etiologies involving environmental and genetic factors (6–8), only a few modifiable risk factors have been identified besides smoking.

A broad literature supports the idea that stress may play a role in worsening RA and SLE symptoms and outcomes, and growing evidence suggests stress may trigger a variety of autoimmune diseases (9, 10). Recent studies show that history of trauma or post-traumatic stress disorder is associated with risk of developing RA or SLE (11–18). Given the known effects of stress on the immune system, such as immunosuppression and inflammation (19–21), these findings support a broader hypothesis that other stressors may increase risk of RA/SLE. Stressful life events can include acute experiences such as death of a partner or a serious accident, abuse, interpersonal conflicts such as marital separation, or financial stress, with some experiences reflecting traumatic stressors, patterns of longer-term experiences and chronic stress, or compounding trajectories of stressors across the lifespan (22). Research on stressful life events and RA/SLE is limited; one large case-control study in Sweden reported stressful life events in the 5 years prior to diagnosis were associated with an increased odds of developing RA (23), while a smaller study of SLE in southern Sweden showed no associations with life events in the past year (24).

Individuals experience external stressors within a broad socioeconomic and psychological, and social context. The Reserve Capacity model posits that lower socioeconomic status (SES) modifies the impact of stressors, undermining tangible, interpersonal and intrapersonal resources for coping, such as social support, amplifying over time the potential for adverse effects on health (25, 26). Past research in the Women’s Health Initiative (WHI) found that having more negative life events was related to lower education, non-white race, lower social support, and adverse health behaviors (e.g., smoking, BMI)(27, 28). While the latter may have direct physiological impacts, the others may reflect differential reserve and resources, which may modify the impact of life events on health, including autoimmunity, inflammation, and disease (Figure 1). Race/ethnicity and age may further contextualize this relationship. In the WHI Extension Study (mean age 77 years), younger women and those of Black or African American (versus White) race/ethnicity reported higher resiliency (29).

Figure 1.

Life-course and recent stressors, reserve capacity, and pathways contributing to the development of autoimmune disease

In the WHI Observational Study (OS) cohort, we previously noted associations of SES-related covariates with risk of RA (i.e., non-professional occupation) and SLE (i.e., lower education)(30). Here we investigated whether risk of RA/SLE in the first three years of follow-up was associated with recent life events and caregiving, another potential stressor associated with depressive symptoms in the WHI (31). We hypothesized that having more stressful life events and caregiving might contribute to risk of developing RA or SLE. We also examined specific types of life events and in secondary analyses explored potential differences in associations with RA/SLE by age and indicators of Reserve Capacity (occupational class, social support).

MATERIALS AND METHODS

Study sample

The WHI-OS cohort includes 93,676 women (enrolled 1994–98) from 40 clinical centers across the United States (32). At baseline and year 3 of follow-up, women were asked if a doctor ever told them they had systemic lupus erythematosus (SLE), or arthritis, and if so, what type, i.e., rheumatoid arthritis (not rheumatism) or other/don’t know. For the current study, eligible participants were those with complete data on RA/SLE status [exclude missing, N=2,429 (2.6%)], medication use [exclude missing, N=13,871 (14.8%)], the life events scale and major covariates [exclude missing, N=842 (0.9%)]. The analysis sample excluded prevalent RA or SLE cases who used disease modifying anti-rheumatic drugs (DMARDs; N=815, 0.9%) and potential cases with either RA or SLE without DMARDS or DMARDS without RA/SLE at baseline or follow-up (N=2,533; 2.7%). Probable cases were identified based on a new self-reported diagnosis during the first three years of follow-up, confirmed by DMARD use at year 3 (a highly specific method for case ascertainment in the absence of medical records review or physician validation)(33). Derivation of the analysis sample is shown Supplemental Figure 1, and characteristics of the study sample relative to the complete OS Cohort are shown in Supplemental Table 1. The final analysis sample of 74,897 women included 211 cases of probable incident RA (N=176), SLE (N=27), or both (N=8) and 74,686.

The institutional review boards of the participating institutions approved protocols and consent forms, which were signed by the women at enrollment.

Stress and covariates

Baseline questionnaires included a standardized life events scale, asking about 11 items representing external stressors in the past year, i.e., “hard things that sometimes happen to people” (34). We used the total count of events reported (0, 1, 2, 3, 4+ for descriptive frequencies, and collapsed categories to reduce the impact of small cell size in modeling and secondary analyses). We also considered different types: interpersonal (6 items: e.g., spouse was deceased, close friend or family member deceased or serious illness, divorce or breakup, close friend/family member divorce, close friend/family member lost job or retired, major conflict with children or grandchildren), financial stress (1 item: major money problems), and abuse (2 items: physically or verbally abused by a family member or close friend). Physical abuse was infrequently reported in the absence of verbal abuse (27), and for multivariable modeling we grouped these as three categories (none, verbal abuse only, and physical abuse ± verbal abuse). Other items included death of a pet, and major accidents, disasters, muggings, unwanted sexual experiences, robberies, or similar events.

Baseline questionnaires also asked about recent caregiving, a potential source of chronic stress: questions asked whether a woman was regularly providing care for an ill relative or friend, and how many days per week in the past 4 weeks, which we grouped into 3 categories (none, up to 2 days, 3 or more days per week)(35). Social support was assessed through the general social support index (including emotional, tangible, affection, and positive social interactions), which we dichotomized as lower (≤median) and higher (>median)(36). Depressive symptoms in the past 2 weeks were assessed by the modified CESD-6 scale, dichotomized using a cut-point of ≥0.06 (37). Joint pain or stiffness in the past 4 weeks was rated as none, mild, moderate, severe. Other covariate data included age, self-reported race/ethnicity, education, occupational class, smoking pack-years, and body mass index (BMI)

Analyses

We modeled risk of developing RA/SLE using Cox proportional hazards regression to calculate Hazard Ratios (HR) and 95% Confidence Intervals (CI). Initial models adjusted for age, while fully adjusted models also included race/ethnicity, occupational class, education, pack-years of smoking and BMI. No evidence of confounding was seen in fully adjusted models, so age-adjusted results are reported in the text. Proportionality was assessed by adding a term for the interaction between the exposure and log-transformed survival time (p-values: 0.6455 total life events, 0.1999 interpersonal events, 0.0466 financial stress, 0.6681 for abuse, and 0.6730 caregiving. Graphs of the survival function versus log-transformed survival time were also examined, showing the departure from proportionality for financial stress was based on a small number of cases at one extreme. Trends tests for life event and Interpersonal event counts were based on linear terms in the models.

In sensitivity analyses, we excluded women with symptoms, which could indicate pre-clinical or undiagnosed RA (i.e., moderate to severe joint pain without doctor diagnosed arthritis or missing data on arthritis: 5343 non-cases, 35 cases). We also ran models excluding women with depressive symptoms (8,859 non-cases, 44 cases), which may be a cause or effect of chronic inflammation, autoimmunity, and response to pain, and may also result from earlier life events and past trauma (38–41).

In secondary analyses, we explored potential modifiers of the relationship of RA/SLE risk with life events, including age, social support, and occupational class. Interaction p-values were derived by comparing models including both variables with and without cross-product term; the test statistic, based on the deviance method, was tested against a chi-square distribution (degrees of freedom=difference in degrees of freedom between the two models) with a p<0.10 for statistical significance, allowing a higher Type 1 error rate to accommodate the lower power for testing interactions.

Results

Table 1 shows across categories of the number of life events reported (0, 1, 2, 3, and 4+) that women who reported more life events or stressors were slightly younger and had lower social support. They were more likely to be Black or Hispanic, have a lower educational attainment and occupational class, greater BMI and pack-years of smoking, and were more likely to report depressive symptoms or moderate to severe joint pain or stiffness. The most common stressor was death of a friend/family member (range 48% to 82% across categories 1 to 4+), followed by financial stress (11% to 76%), major conflict (8% to 64%), verbal abuse only (not including physical, 3% to 37%), major accident (3% to 26%), divorce/breakup (1% to 14%), spouse/partner died (1% to 10%), and physical abuse (with or without verbal abuse, <1% to 9%). Frequent caregiving (3 or more times per week) was reported by 11% and 13% of those with 0 or 1 life events, to 20% of those with 4+ events.

Table 1.

Characteristics of study participants by number of stressful life events at baseline

	Number of events reported
	0 (n=16836)		1 (n=23948)		2 (n=17452)		3 (n=9527)		≥ 4 (n=7134)
	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Age	64.3	(7.2)	64.1	(7.3)	63.3	(7.3)	62.6	(7.2)	61.6	(7.2)
Social support	37.6	(7.2)	37.0	(7.3)	36.0	(7.6)	34.9	(7.9)	32.8	(8.3)
	N	%	N	%	N	%	N	%	N	%
Race/ethnicity
White	14863	88.3	20966	87.5	14849	85.1	7905	83.0	5477	76.8
Black	752	4.5	1378	5.8	1327	7.6	838	8.8	819	11.5
Hispanic	426	2.6	563	2.4	502	2.9	358	3.8	441	6.2
American Indian	43	0.3	66	0.3	59	0.3	40	0.4	61	0.9
Asian/Pacific Islander	571	3.4	704	2.9	482	2.8	233	2.4	203	2.8
Unknown	181	1.1	271	1.1	233	1.3	153	1.6	133	1.9
Education
< HS Graduate	565	3.4	806	3.4	697	4.0	441	4.7	431	6.1
High school/GED	2665	15.9	3761	15.8	2711	15.6	1470	15.6	1075	15.2
School after high school	5478	32.8	8258	34.7	6431	37.1	3695	39.1	2963	41.9
≥College degree	8007	47.9	10962	46.1	7491	43.2	3834	40.6	2598	36.8
Occupation
Managerial/Professional	7507	46.4	10639	46.2	7496	44.7	3951	43.2	2714	40.0
Technical/Sales/Admin	4386	27.1	6470	28.1	4858	28.9	2735	29.9	2062	30.4
Service/Labor	2335	14.4	3503	15.2	2762	16.5	1674	18.3	1460	21.5
Homemaker only	1951	12.1	2439	10.6	1672	10.0	789	8.6	556	8.2
Body-mass index (kg/m2)
<25	7950	47.8	10520	44.4	7030	40.8	3638	38.6	2335	33.1
25 - <30	5574	33.5	8058	34.0	5951	34.5	3252	34.5	2408	34.1
≥30	3106	18.7	5112	21.6	4258	24.7	2536	26.9	2310	32.8
Pack years of smoking
Never-smoker	8918	54.6	12310	53.1	8791	52.1	4689	51.0	3488	50.8
<5	2320	14.2	3440	14.9	2500	14.8	1436	15.6	1092	15.9
5 - <20	2288	14.0	3327	14.4	2451	14.5	1328	14.4	974	14.2
≥20	2807	17.2	4087	17.6	3140	18.6	1743	19.0	1310	19.1
Specific life events1
Friend/family member died	---	NA	11560	48.4	11201	64.3	7069	74.3	5868	82.4
Financial stress	---	NA	2534	10.6	5391	30.9	4839	50.9	5402	75.9
Major conflict	---	NA	1961	8.2	3744	21.5	3727	39.2	4538	63.7
Abuse – verbal only	---	NA	826	3.4	1812	10.4	1945	20.4	2658	37.3
Major accident	---	NA	637	2.7	1234	7.1	1175	12.3	1859	26.1
Divorce of break-up	---	NA	117	0.7	327	1.9	440	4.6	1006	14.1
Spouse/partner died	---	NA	339	1.4	628	3.6	548	5.8	694	9.7
Physical (± verbal abuse)2	---	NA	16	0.1	90	0.5	162	1.7	635	8.9
Caregiving
None or infrequent	11196	66.8	14788	62.0	10164	58.6	5110	54.0	3540	49.9
Up to 2 times/week	3659	21.8	5880	24.7	4518	26.0	2764	29.2	2102	29.6
3 or more times/week	1902	11.4	3166	13.3	2674	15.4	1593	16.8	1450	20.4
Depression symptoms3	577	3.5	1527	6.5	1910	11.2	1501	16.1	1991	28.6
Joint pain or stiffness
Symptom did not occur	5915	35.2	7635	32.0	4816	27.7	2373	25.0	1423	20.0
Mild	7817	46.6	11260	47.1	8288	47.6	4361	45.9	3137	44.2
Moderate	2592	15.4	4040	16.9	3352	19.3	2112	22.3	1813	25.5
Severe	459	2.7	960	4.0	940	5.4	646	6.8	729	10.3

¹Common or severe events listed, including include financial stressor (i.e., major money problems); spouse or partner died, divorce or break-up; major conflict (i.e., with children or grandchildren); major accident (or disaster). Other included close friend/family member divorced, close friend/family member lost job or retired; pet died.

²Physical abuse occurred in the absence of reported verbal abuse for 8, 33, 45, and 72 women across the 4 event categories.

³Depression symptoms if CES-D score ≥ 0.06.

Table 2 shows frequencies of life events in the past year, caregiving days per week, and symptoms of depression and joint pain in RA/SLE cases and non-cases. Across all life events evaluated, 12.8% of cases reported 4 or more and 18.5% reported 3 events. Across the different types of events: 10% reported at least 3 interpersonal life events and 24.2% reported 2 events; 28.4% reported financial stress. Only 2.4% reported physical abuse, while 12.3% reported verbal abuse only. Caregiving 3 or more days per week was reported by 17.5% of cases. At baseline, 17.3% had symptoms of depression, and 8.1% reported moderate to severe joint pain in the absence of diagnosed arthritis. Similar frequencies were seen in RA cases, while frequencies in SLE cases are shown in Supplemental Table 2. Covariate frequencies are shown in Supplemental Table 3.

Table 2.

Incident RA/SLE by stressful life events, caregiving, depressive symptoms, moderate to severe joint pain or stiffness, and doctor diagnosed arthritis other than RA.

	Non-cases (n=74686)		RA/SLE (n=211)1		RA (n=184)
	N	%	N	%	N	%
Life events (all)
0	16797	22.5	39	18.5	32	17.4
1	23884	32.0	64	30.3	59	32.1
2	17410	23.3	42	19.9	34	18.5
3	9488	12.7	39	18.5	36	19.6
≥ 4	7107	9.5	27	12.8	23	12.5
Interpersonal
0	22895	30.7	60	28.4	47	25.5
1	29581	39.6	79	37.4	73	39.7
2	15581	20.9	51	24.2	46	25.0
≥ 3	6629	8.9	21	10.0	18	9.8
Financial stress
No	56448	75.7	151	71.6	133	72.3
Yes	18106	24.3	60	28.4	51	27.7
Missing	132
Abuse
None	66731	89.3	180	85.3	161	87.5
Verbal only	7215	9.7	26	12.3	20	10.9
Physical +/− verbal	740	1.0	5	2.4	3	1.6
Caregiving (days/week)
None/infrequent	44678	60.1	120	56.9	103	56.0
Up to 2 times/week	18869	25.4	54	25.6	49	26.6
3 or more times/week	10748	14.5	37	17.5	32	17.4
Missing	391		0
Depression (CES-D/DIS ≥ 0.06)
No	65827	89.8	167	82.7	148	84.6
Yes	7471	10.2	35	17.3	27	15.4
Missing	1388		9
Joint pain/stiffness
No symptoms	22137	29.7	25	11.8	23	12.5
Mild	34776	46.7	87	41.2	76	41.3
Moderate	13838	18.6	71	33.6	62	33.7
Severe	3706	5.0	28	13.3	23	12.5
Missing	229
Doctor diagnosed arthritis2	31978	42.8	107	50.7	92	50.0
Moderate-Severe joint pain/stiffness  + Diagnosed arthritis   + No diagnosed arthritis  Missing arthritis	12430 3808 1306	16.7 5.1	64 17 18	30.3 8.1	55 14 16	29.9 7.6

¹Total sample includes 8 cases of RA with SLE and 27 cases of SLE-only

²Doctor diagnosed arthritis other than RA, other non-specified or unknown type

In proportional hazard regression models (Table 3), RA/SLE risk was associated with a greater number of recent life events or stressors (e.g., age-adjusted HR=1.70; 95%CI 1.14, 2.53 for 3 or more versus none; p-trend 0.0026), and with physical abuse (e.g., HR=2.48; 95%CI 1.02, 6.03). Elevated HRs were seen for increased interpersonal events (HR 1.23, 95%CI 0.87, 1.73), financial stress (HR 1.22; 95%CI 0.90, 1.64), and more frequent caregiving (>2 days/week versus none; HR 1.25; 95%CI 0.87, 1.81), though confidence limits did not exclude the null and trend tests were not statistically significant. Estimates did not appear to be confounded by demographic or behavioral factors (race/ethnicity, education, occupation, pack-years of smoking and BMI), as results were similar in fully adjusted models. In sensitivity analyses (Supplemental Table 4), associations of RA/SLE with the number of life events and physical abuse persisted after excluding women with moderate or severe joint pain or stiffness not due to diagnosed arthritis or depressive symptoms. The trend for more frequent caregiving became statistically significant after excluding women with moderate to severe joint pain in the absence of diagnosed arthritis (p_trend=0.0424) but was attenuated excluding those with depressive symptoms.

Table 3.

Associations of stressful life events or stressors with incident RA or SLE adjusting for age and covariates’

Stressful life events1	Number of non-cases	Number of RA/SLE cases	HR (95% CI) 2 Age-adjusted	HR (95% CI) 2 Fully adjusted
All life events/stressors
0	16797	39	1.0	1.0
1 to 2	41294	106	1.11 (0.77, 1.60)	1.16 (0.78, 1.73)
≥ 3	16595	66	1.70 (1.14, 2.53)	1.75 (1.14, 2.69)
p-trend			0.0026	0.0039
Interpersonal
0	22895	60	1.0	1.0
1	29581	79	1.02 (0.73, 1.43)	1.17 (0.82, 1.67)
≥ 2	22210	72	1.23 (0.87, 1.73)	1.25 (0.86, 1.81)
p- trend			0.2403	0.2465
Financial stress
No	56448	151	1.0	1.0
Yes	18106	60	1.22 (0.90, 1.64)	1.15 (0.83, 1.59)
p- trend			0.2078	0.4020
Abuse
None	66731	180	1.0	1.0
Verbal only	7215	26	1.34 (0.89, 2.02)	1.36 (0.89, 2.10)
Physical (± verbal)	740	5	2.48 (1.02, 6.03)	2.50 (1.02, 6.14)
p- trend			0.0614	0.0524
Caregiving
None	44678	120	1.0	1.0
≤ 2 times/week	18869	54	1.06 (0.77, 1.46)	1.04 (0.74, 1.47)
> 2 times/week	10748	37	1.25 (0.87, 1.81)	1.31 (0.89, 1.92)
p- trend			0.2571	0.2115

¹Categories collapsed to reduce the variance in multivariable models.

²Hazard ratios and 95% confidence intervals estimated from Cox proportional hazards regression models, adjusted for age, or fully adjusted for age, race/ethnicity, education, occupation, pack-years of smoking and BMI.

In secondary analyses exploring potential modifiers (Figure 2 and Supplemental Table 5), several associations were more apparent in women with lower occupational status (i.e., those with non-professional/managerial jobs or homemakers) or social support, or ages 50–64 years. Abuse showed strong, statistically significant associations with RA/SLE in women with lower occupational status, lower social support, and ages 50–64 years (ORs 1.60 to 1.79), though interactions were not statistically significant. The association of financial stress with RA/SLE was also more apparent in women with a lower occupational status (p_interaction =0.0147) and in women aged 50–64 years (p_interaction=0.0814). By contrast, the association of RA/SLE with interpersonal events was more apparent in women aged 65+ years (p_interaction=0.0742).

Figure 2.

Life events associated with RA/SLE, stratified by social and demographic buffering factors Age-adjusted Hazard Ratios and 95% Confidence Intervals.

Discussion

Results of this prospective study support the hypothesis that stressful life events may play a role in the development of RA and SLE in post-menopausal women. We found that women reporting a greater number of events in the past year, including death of a partner or close friend, conflict with children or grandchildren, divorce, abuse (physical or verbal), financial problems, and major accidents or disasters, had a 70% increased risk of being diagnosed with probable RA or SLE within three years of enrollment. These results extend prior literature showing associations with diverse, contemporary psychosocial stressors in a population of middle-aged and older women.

Our prospective findings for probable RA/SLE are consistent with a retrospective study of 2,774 RA cases in Sweden showed that having 3 or more life events in the past 5 years was associated with RA in women and with anti-citrullinated peptide antibody (ACPA)-negative cases (ORs 1.3–1.4) , though their study sample was younger (Median age 55 years, range 18–70) and clinically validated (23). Looking at individual types of events, they saw stronger (OR≥1.4), statistically significant associations for interpersonal conflict (with a spouse or children or at work), increased or decreased responsibility at work, unemployment, change in residence, and divorce. In a retrospective analysis of 85 SLE cases, the strongest associations (though none of them statistically significant) included severe conflict or being deeply offended by someone, along with death of a child (24). In our overall sample, RA/SLE was not associated with increased interpersonal events (a group that included events such as divorce, death of someone close, conflict with children, a family member/friend loss of a job or retirement), or with financial problems. Focused research is needed to address the impacts of stress from interpersonal conflict and loss.

While relatively uncommon, we found that physical abuse (with or without verbal abuse) in the past year was associated with risk of RA/SLE. Partner abuse is an important, understudied traumatic stressor in women, and these findings add to a growing literature focused on post-traumatic stress disorder (PTSD) and trauma as risk factors for RA and SLE (10–12, 18, 42). Previous research in the WHI has shown that abuse is related to depressive symptoms and overall mortality risk (43, 44). Though focused on events in the past year, the question specifies abuse in a close relationship, which may also indicate an extended pattern of traumatic stress (45). We did not look at associations specifically in SLE cases, but note that 25.7% reported physical or verbal abuse, compared to 12.5% of those with RA and 10.7% of non-cases. Given the small number of exposed cases, these findings warrant cautious interpretation.

Our results suggest that overall caregiving was not associated with developing RA/SLE. We saw an elevated HR for more frequent caregiving (3 or more days per week), but the CI included the null. Notably, these results were sensitive to excluding women with baseline symptoms of joint pain (increased HR) or depression (attenuated HR). Caregiving is a common experience in mid-life and older women caring for parents or partners and is typically viewed as a source of chronic stress. But research on caregiving often fails to capture the context or heterogeneity of caregiving, and any potential positive effects on health and wellbeing (46–49). Conversely, caregiving may be accompanied by life events such as illness or death of a spouse or relationship conflict. In the current study sample, women with more life events also reported more frequent. Caregiver stress may be better captured in future studies using methods such as a latent class analysis (50).

We explored the use of markers of Reserve Capacity (i.e., occupational class and social support) to contextualize the environment in which stressors occur, and age, as potential effect modifiers on the relationship of life events with RA/SLE. In stratified models, most differences were in the expected direction, with greater risk in women with fewer resources, i.e., lower occupational status and lower social support. Most did not reflect significant interactions (i.e., p<0.10), except for financial stress, which was associated with increased RA/SLE in women with lower occupational status and in women ages <65 years. Having more interpersonal events was also associated with RA/SLE in those who were 65 and older, but not in women <65. These differences could be related to exposure heterogeneity across subgroups, or variable response depending on life stages and co-occurring events. Cautious interpretation is warranted given the exploratory nature of these analyses.

Psychological stress or distress is experienced when demands outweigh psychological, social, and materials resources, with impacts on health through diverse pathways, including psychopathology. We did not adjust for depressive symptoms, which may be a marker for physiologic effects on the pathway to disease. Baseline depressive symptoms in the past 4 weeks were more common in women who later developed RA/SLE, however observed associations with life events persisted in women without symptoms. We did not examine diagnosed depression or anti-depressant use, nor did we evaluate potential mediation or modification by depression, which warrants further consideration as it offers a potential opportunity in clinical settings to identify and support individuals at risk.

Stress can affect health through different biological mechanisms, depending on the type or timing of the stressor, and other historical and concurrent exposures that modify neuroendocrine responses. Chronic and acute stress due to experiences viewed as threatening, unpredictable or uncontrollable may lead to immune dysfunction and dysregulation, resulting in a variety of long and short-term changes, including increased susceptibility to infection, reduced healing, and inflammation (51, 52). Autoantibodies in RA (and other autoimmune diseases) can arise well in advance of disease onset as part of a larger constellation of factors, including systemic inflammation, preceding clinically apparent disease (32, 53–55). Stress effects on the immune system could act at an earlier stage on the causal pathway leading to autoantibody production, for example, in the mucosal origins hypothesis for RA, or at a later stage contributing to clinical pathology leading to diagnosis (56). Time to diagnosis following initial symptoms varies and it is likely that some cases in our study had undiagnosed disease at baseline. Early symptoms of disease may include joint-specific inflammation; however, results were unchanged in models excluding those reporting moderate or severe joint pain at baseline not due to other or unknown forms of doctor-diagnosed arthritis.

This study has limitations. Although the sample was of sufficient size to detect modest associations with frequently reported life events, the low incidence of probable RA/SLE and short follow-up time limited our ability to examine less common individual exposures (e.g., physical abuse, or experiencing an accident or disaster or another major traumatic event), or conduct analyses limited to SLE. The WHI-OS cohort is a volunteer sample of women who were ineligible for the clinical trials for various reasons; most were white, while Black and Hispanic women were more likely to be excluded from our study sample due to missing data, further limiting generalizability. The small number of non-white women in the study sample precluded analyses of racial/ethnic disparities in RA/SLE in relation to life events (57).

Self-report of RA/SLE is known to be non-specific, so we limited our analyses to probable clinical cases based on their use of disease-specific medication. During our study period, the paradigm of early initiation of DMARDs following diagnosis was emerging, but not widespread, especially in older patients. Of those enrolled in Medicare Part D, only 24% of RA patients used DMARDs in 1996, rising to only 41% by 2003 (58, 59). Thus, our cases may represent those with more aggressive disease or otherwise better access to and uptake of DMARDs. We lacked data on disease phenotype, including anti-CCP antibodies at RA diagnosis. Recent findings in male RA patients showed that PTSD symptoms were related to serum cytokine levels in those with anti-CCP antibodies (60), while a large retrospective study showed no associations of life events with anti-CCP negative RA in men, but few differences in associations with RA by anti-CCP status among women (23). Further research is warranted on the role of stress and RA/SLE in larger preclinical samples, including those with anti-CCP antibodies and individuals with a family history of RA/SLE.

Stress exposure assessment can be challenging in studies of autoimmune diseases, which themselves can be stressful or cause physical and psychological changes that may result in recall bias or reverse causality (39). Events were measured only at one time point, and no data were available on past adult life events, traumas, and childhood adverse experiences, which may proliferate and form trajectories impacting health in older women (61, 62). Results for physical abuse, in particular, are likely to reflect a longer trajectory of abuse (physical and verbal) with cumulative effects on health. Repeated stress measures, also including perceived stress and psychological sequelae of traumatic stress, over a longer follow-up period, are needed to identify whether specific events versus sustained patterns of trauma or chronic stress confer increased risk of RA/SLE, especially given the potential latency of effects in the development and progression of disease.

We saw limited evidence of confounding by BMI and smoking, race/ethnicity, or socioeconomic factors (occupational status and education), but cannot rule out the possibility of unmeasured confounders.

In sum, the results of this prospective analysis highlight the possible role of stress due to recent life events as proximal risk factors for RA or SLE in post-menopausal women, and support the need to consider a diverse range of stressors and contextual factors in future studies. If replicated, our findings also suggest opportunities to identify individuals who may be at higher risk for developing RA or SLE.

SIGNIFICANCE AND INNOVATIONS.

• In a well-characterized cohort of post-menopausal women, those reporting 3 or more major life events in the past year at baseline had a 70% increased risk of developing rheumatoid arthritis (RA) or systemic lupus erythematosus (SLE) in the subsequent three years.

• Associations were not confounded by sociodemographic and lifestyle factors and were robust to sensitivity analyses excluding women with baseline depressive symptoms or unexplained moderate to severe joint pain.

• Our findings for diverse stressors, ranging from interpersonal factors to financial stress to abuse, add to a growing literature on the role of psychosocial stressors in the development RA and SLE.