Association of trauma and posttraumatic stress disorder with incident systemic lupus erythematosus (SLE) in a longitudinal cohort of women

Abstract

Objective

To conduct the first longitudinal study examining whether trauma exposure and PTSD are associated with increased risk of incident SLE in a civilian cohort.

Methods

We examined the association of trauma exposure and PTSD symptoms with SLE incidence over 24 years of follow-up in a U.S. longitudinal cohort of women (N=54,763). Incident SLE with ≥4 American College of Rheumatology criteria was ascertained by self-report and confirmed by medical record review. PTSD and trauma exposure were assessed with the Short Screening Scale for DSM-IV PTSD and the Brief Trauma Questionnaire. Women were categorized as having: no trauma, trauma and no PTSD symptoms, subclinical PTSD (1-3 symptoms), or probable PTSD (4-7 symptoms). We examined whether longitudinally assessed health risk factors (e.g., smoking, body mass index (BMI), oral contraceptive (OC) use) accounted for increased SLE risk among women with versus without trauma exposure and PTSD.

Results

During follow-up, 73 cases of SLE occurred. Compared to women with no trauma, probable PTSD was associated with increased SLE risk (HR_{4-7 symptoms} =2.94, 95% CI=1.19-7.26, p<0.05). Subclinical PTSD was associated with increased SLE risk, though this did not reach statistical significance (HR_{1-3 symptoms} =1.83, 95% CI=0.74-4.56, p=0.19). Smoking, BMI and OC use slightly attenuated associations (e.g., probable PTSD adjusted HR=2.62, 95% CI=1.09-6.48, p<0.05). Trauma exposure, regardless of PTSD symptoms, was strongly associated with incident SLE (HR=2.87, 95% CI=1.31, 6.28, p<0.01).

Conclusions

This study contributes to growing evidence that psychosocial trauma and associated stress responses may lead to autoimmune disease.

Psychosocial stress exposure may alter immune function, and exposure to severe stressors, as well as high levels of distress after stressor exposure, have been implicated in autoimmune disease pathogenesis^1-3 and associated with subsequent development of autoimmune disease^4,5. Posttraumatic stress disorder (PTSD) is the sentinel stress-related disorder and indicates severe psychological distress occurring in response to a traumatic stressor. Epidemiologic research has suggested that PTSD may increase risk for autoimmune disease^6,7, including rheumatoid arthritis,^7-10 autoimmune thyroiditis^9,10, inflammatory bowel disease, multiple sclerosis⁹ and psoriasis¹⁰. Thus, PTSD may also be associated with increased risk for systemic lupus erythematosus (SLE).

SLE is an autoimmune disorder associated with renal failure¹¹, myocardial infarction^12,13, fatal infection^14,15, and premature mortality (standardized mortality ratio=2.4)¹⁴ which has an incidence 3 to 13 times higher among women than men¹⁶. High prevalence of anxiety and psychological distress are well documented among individuals with SLE^17-19, and stress and emotional distress are often implicated by SLE patients as triggers of their disease flares²⁰. However, evidence regarding whether traumatic experiences, stress or PTSD increase SLE risk is sparse as just one study comprised of predominantly male military veterans has specifically examined SLE risk association with PTSD. In a study using computer medical records of individuals enrolled in the Department of Veterans Affairs health care system, war veterans with a PTSD diagnosis had significantly higher risk of subsequent SLE diagnosis (adjusted relative risk, RR=1.85)⁹ than those without a diagnosis of a psychiatric disorders over a median of 4 years of follow-up. Moreover, the association of PTSD with SLE (as well as other autoimmune diseases) was stronger than those of other psychiatric disorders. This sample was 88% male, and the median time elapsed between psychiatric diagnosis and immune disorder diagnosis was just over 7 months (220 days), raising concerns about potential confounding or reverse causation due to undetected autoimmune disease.

No longitudinal studies of SLE risk in association with PTSD or traumatic events among civilians have been conducted. Additionally, no studies have examined whether trauma exposure alone, irrespective of psychological sequelae, is associated with increased risk of SLE. Finally, lifestyle factors that are more common in persons with PTSD have been identified as possible risk factors for increased systemic inflammation and autoimmune disease, including smoking, obesity, and oral contraceptive use^16,21-26. These behavioral factors may partially account for the relationship between trauma, PTSD, and autoimmune disease, yet their role has not been examined.

In the present study, we tested the hypothesis that trauma and PTSD are associated with increased risk of incident SLE in a large longitudinal cohort of civilian women. We further examined whether higher prevalence of health risk behaviors, namely, smoking, sedentary lifestyle, obesity, alcohol use and oral contraceptive use, might account for possible increased risk of SLE in women with versus without PTSD and trauma exposure. Finally, as depression has been associated with SLE¹⁸ and is frequently co-morbid with PTSD²⁷, we ascertained the association of PTSD with SLE independently of depression.

Methods

Participants

The Nurses' Health Study II is an ongoing cohort of 116,430 female nurses initially enrolled in 1989 and followed with biennial questionnaires. The present study included follow-up through 2013. This study included women who returned a supplementary 2008 questionnaire on trauma exposure and PTSD symptoms (N=54,763). This questionnaire was sent to a subsample of participants (N=60,804, response rate=90.1%). To retain participation in the ongoing longitudinal cohort, only women who have already returned their biennial questionnaire are sent supplementary questionnaires. Women missing data on trauma or PTSD symptoms (N=3,930) were excluded. This study was approved by the Institutional Review Board of Brigham and Women's Hospital. Return of the questionnaire via US mail constitutes implied consent.

Case ascertainment

Methods for SLE case identification and validation according to ACR 1997 Classification Criteria have been reported.^28,29 Nurses were asked to report all new physician-diagnosed SLE on each questionnaire. Women who self-reported new cases were then asked to complete the connective tissue disease screening questionnaire (CSQ)²⁸, to provide the name and address of the health care provider who had diagnosed SLE, and to sign a medical records release. For all women who scored positive for symptoms of SLE on the CSQ, we attempted to obtain medical records from the time of diagnosis. These records were independently reviewed by two board-certified rheumatologists for all ACR criteria for SLE and other features consistent with SLE.

We excluded participants who reported an existing diagnosis of SLE at baseline, and censored those who self-reported connective tissue disease in follow-up without SLE confirmation by medical record review (N=531). The sensitivity and specificity of this two-stage screening procedure have been shown to be high.³⁰

PTSD and trauma ascertainment

Trauma exposure and PTSD symptoms were assessed on a supplementary 2008 questionnaire. The 16-item Brief Trauma Questionnaire queried lifetime exposure to 15 types of traumatic events (e.g., serious car accident, sexual assault) and an additional item queried any traumatic event not covered in the other questions³¹. Respondents were asked to identify which trauma was their worst or most distressing; they were then asked their age at this worst trauma as well as their age at their first trauma. PTSD symptoms were assessed in relation to their worst trauma with the 7-item Short Screening Scale for DSM-IV PTSD³². Four or more symptoms on this scale have been associated with PTSD diagnosis (sensitivity=80%, specificity=97%, positive predictive value=71%, negative predictive value=98%)³².

For each year of follow-up, participants were characterized as: trauma unexposed, trauma exposed/no PTSD symptoms, trauma/1-3 PTSD symptoms (subclinical PTSD), or trauma/4-7 PTSD symptoms (probable PTSD). Trauma and PTSD status were time-updated over the follow-up period, with date of onset of trauma and PTSD symptoms determined as follows. A respondent was considered trauma unexposed for each year of follow-up before her age at first trauma. For each year of follow-up after her age at first trauma, she was considered trauma exposed/no PTSD symptoms. For each year of follow-up after her age at worst trauma, she was characterized as having no symptoms, 1-3 symptoms, or 4-7 PTSD symptoms based on her responses to the PTSD screen. If a woman reported only one trauma, her age at first and age at worst trauma were considered the same. For example, a woman who reported experiencing her first trauma at age 40, her worst trauma at age 50 and 6 PTSD symptoms in relation to this worst trauma would be coded as: trauma unexposed before age 40, trauma exposed with no PTSD symptoms after age 40, and trauma exposed with 4-7 PTSD symptoms after age 50. To mitigate concerns about reverse causality, women who reported illness as their worst trauma were excluded from analyses of PTSD and SLE to avoid including women whose SLE could have induced PTSD symptoms (N=2129, N=18 SLE cases).

For analyses of the association of any lifetime trauma with SLE risk (irrespective of PTSD symptoms), we coded trauma exposure according to a woman's age at her first trauma. For each year of follow-up, a woman was considered trauma exposed if she was older than her reported age at first trauma and was considered trauma unexposed if she was younger than her age at first trauma or reported never experiencing a traumatic event. Women whose first trauma was illness were excluded (N=408, N=4 SLE cases).

Health risk behaviors and demographic covariates

We selected covariates that have been related to either PTSD or SLE or both^{16,21-23,25,26,29,33}. Covariates were time-updated, such that for each year of follow-up the most recent report was used (Supplemental Figure). Oral contraceptive use, current smoking, and weight were queried on each biennial questionnaire, 1989-2013. Body mass index (BMI) was calculated in kg/m² based on weight reported on the biennial questionnaires and self-reported height as reported on the baseline questionnaire (1989). Self-reported weight was highly reliable (r=0.97) in a validation study³⁴. Physical activity was queried in 7 waves (1989-2013) and was characterized as 0-9 or 10+ metabolic equivalent (mets)/week. Alcohol use was queried in 1989, 1995, 2005 and 2009. Lifetime history of depression was assessed in 2001. Indicators of socioeconomic status included U.S. Census tract median household income obtained from geocoded home addresses and maximum of parents' education when respondents were infants, reported in 2005 as: high school or less, some college, or college or more. Self-reported race was coded as white or non-white. As race has been strongly associated with SLE risk¹⁶, we included race in all models.

Statistical analyses

We examined prevalence of all covariates by trauma/PTSD status at baseline in 1989.

To ascertain the association of trauma exposure and PTSD with incident SLE, we calculated hazard ratios for each level of trauma/PTSD symptoms using Cox proportional hazards regressions with age in months as the measure of time (time metameter), with trauma unexposed as the referent. We additionally examined the association of PTSD symptoms coded as a continuous variable (0 to 7) with SLE incidence.

We evaluated census tract median income and parents' education in infancy as possible demographic covariates; as we found they were not associated with SLE and did not alter the association of trauma or PTSD with SLE, we did not include them in models. We separately examined the association of each health risk factor, e.g., smoking, sedentary behavior, obesity, alcohol use, and oral contraceptive use, with SLE incidence using Cox proportional hazards models. To ascertain the extent to which higher prevalence of health risk factors in women with PTSD could account for any observed elevated risk of SLE, we calculated hazard ratios for SLE for each level of trauma/PTSD symptoms in models including as covariates all health factors associated with SLE, using Cox models. We did not conduct formal mediation analyses, as the health risk factors we examined could be either confounders, to the extent they were present prior to trauma or PTSD onset, or mediators, to the extent they increased following onset of trauma or PTSD. Given the rarity of SLE, and that these health risk factors typically initiate prior to the age of most women at our study enrollment, we did not have sufficient power to distinguish confounding from mediation. Thus, these analyses address the question, “Does potentially higher prevalence of health risk factors in women with trauma/PTSD account for any observed elevated risk of SLE?” without distinguishing whether the factors are confounders or mediators of the PTSD-SLE relation.

To examine whether trauma exposure per se confers increased risk of SLE, regardless of whether women subsequently developed PTSD, we calculated hazard of incident SLE in relation to any lifetime trauma exposure using Cox models with age in months as the measure of time, with trauma unexposed as the referent.

Although we established the date of onset of PTSD, because PTSD assessment was in 2008, we conducted careful tests of the possibility of reverse causality in the relationship between PTSD and SLE. That is, we assessed the likelihood that pre-diagnosis SLE symptoms might either elicit PTSD symptoms or increase likelihood of reporting PTSD symptoms, in three sensitivity analyses. First, we conducted analyses excluding any SLE cases occurring within 2 years of PTSD onset/trauma from analyses. Next, we examined the association of PTSD status at cohort enrollment in 1989 (i.e., without updating PTSD status over time) in relation to incident SLE over follow-up, to ensure that PTSD likely occurred well before SLE onset, given that women who reported SLE at cohort enrollment were excluded. Finally, we tested whether SLE incidence was associated with increased subsequent risk of PTSD; for this analysis we excluded women with PTSD symptom onset prior to developing SLE or prior to cohort enrollment and conducted Cox models with age as the time measure, adjusted for race. In these models, women were considered SLE-unexposed before their SLE diagnosis or if they did not report SLE and SLE-exposed in the year of their confirmed SLE diagnosis and for subsequent years. To ascertain whether PTSD was associated with SLE independently of depression, we conducted analyses excluding women with depression prior to PTSD onset. Thus, we conducted sensitivity analyses including only women who: 1) reported no depression prior to 2001, when lifetime history of depression was assessed, and 2) had their worst trauma prior to 2001 or did not experience a traumatic event.

Results

Compared to women with no trauma exposure, women with 4-7 PTSD symptoms at baseline were of similar age, were more likely to have ever used oral contraceptives and to have ever smoked, and were less likely to have a healthy BMI (Table 1). Over the 24-year follow-up period, 73 women developed SLE. In time-updated models, smoking, oral contraceptive use, and BMI were each associated with increased risk of developing SLE, while alcohol use and sedentary behavior were not (Table 2). Nearly all women with SLE in our sample had seen an ACR member rheumatologist and were anti-nuclear antibody (ANA) positive (Table 3).

Table 1. Baseline age-standardized characteristics by trauma and PTSD status Nurses' Health Study II, 1989, N=50,242.

	No trauma	Trauma no PTSD	PTSD 1-3 symptoms	PTSD 4-7 symptoms
N	14,885	19,579	9,514	6,264
Age in 1989, years, mean (SD)*	34.1 (4.8)	34.8 (4.6)	35.2 (4.5)	35.2 (4.4)
Non-white, %	5	5	6	6
Census tract median income
<40,000, %	11	13	13	15
Parents' education
High school or less, %	50	49	49	48
Oral contraceptive use
Never, %	19	16	14	14
Past, %	67	73	74	76
Current, %	14	12	11	10
Smoking status
Never, %	72	66	64	60
Past, %	19	22	24	26
Current, %	10	12	12	14
Alcohol consumption >= 5g/day, %	20	20	21	20
BMI, mean (SD)	23.6 (4.6)	23.9 (4.8)	24.0 (4.9)	24.2 (5.2)
Exercise
0-9 mets/wk, %	38	37	38	37
10+ mets/wk, %	62	62	62	63
Lifetime history of depression (2001)	5	8	8	20

Note: Values are standardized to the age distribution of the study population. Values of polytomous variables may not sum to 100% due to rounding.

^*Not age adjusted.

Table 2. Association of health risk factors with SLE incidence among women followed in the NHSII ^†.

		Person-years	SLE cases	HR (95% CI)
Model 1	Body mass index (kg/m2)
	18.5 to <25	596,488	34	1.0 [Reference]
	25 to <30	295,029	17	1.11 (0.61, 2.00)
	30+	237,617	19	1.71 (0.95, 3.04)
Model 2	Smoking status
	Never	764,548	37	1.0 [Reference]
	Past	297,275	26	1.93 (1.16, 3.21)
	Current	93,212	10	2.06 (1.01, 4.18)
Model 3	Oral contraceptive use
	Never	150,051	3	1.0 [Reference]
	Ever	1,004,984	70	3.62 (1.14,11.52)
Model 4	Exercise (metabolic equivalents/wk)
	0-9	427,431	33	1.0 [Reference]
	10+	691,039	39	0.74 (0.46, 1.17)
Model 5	Alcohol consumption
	None	283,434	18	1.0 [Reference]
	>0, <5g/day	599,448	38	1.08 (0.61, 1.90)
	=>5g/day	272,153	17	1.15 (0.59, 2.26)

^†All models adjusted for race. Each model contains only one health risk factor. “Missing” not shown.

Table 3. Characteristics of incident SLE cases, NHSII, 1989-2011, N=73 cases.

Characteristic	N	%
Race, White	68	93.2
Anti-nuclear antibody (ANA) positive	72	98.6
Anti-double stranded DNA antibody (anti-dsDNA) positive	39	53.4
Arthritis, yes	53	72.6
Hematologic involvement	47	63.4
Renal involvement	7	9.6
Seen by an ACR member rheumatologist	65	89.0

In models adjusted only for race and age, high PTSD symptoms (4-7 symptoms) were associated with increased SLE risk. This association was somewhat attenuated in models further adjusted for smoking, oral contraceptive use, and BMI (Figure, panel A). One to three PTSD symptoms were associated with elevated risk of SLE, however this did not reach statistical significance. Continuous PTSD symptoms were associated with increased risk of incident SLE in models adjusted for race and age with borderline statistical significance (per symptom, HR=1.11, 95% CI=0.99, 1.24). Any versus no trauma exposure was strongly associated with increased SLE risk (Figure, panel B). This association was slightly attenuated after accounting for smoking, BMI, and oral contraceptive use.

Figure.

Hazard ratio for SLE in association with exposure to trauma and symptoms of posttraumatic stress disorder (PTSD).

Figure note: Part A. HR_{Trauma, no PTSD}=1.96, 95% CI=0.82, 4.66; HR_{1-3 PTSD symptoms}=1.83, 95% CI=0.74, 4.56; HR_{4-7 PTSD symptoms}=2.94, 95% CI=1.19, 7.26. Adjusted model: HR_{Trauma, no PTSD}=1.85, 95% CI=0.77, 4.40; HR_{1-3 PTSD symptoms}=1.68, 95% CI=0.68, 4.19; HR_{4-7 PTSD symptoms}=2.62, 95% CI=1.09, 6.48.

Part B. HR_Trauma=2.83, 95% CI=1.29, 6.21. Adjusted model: HR_Trauma=2.61, 95% CI=1.19, 5.73 All hazard ratios were calculated with Cox proportional hazards models with age in months as the time measure. All models are adjusted for race.

In sensitivity analyses excluding the first two years of follow-up, results were similar to those in the main analyses (race- and age-adjusted models: HR_{PTSD 4-7}= 3.26, 95% CI=1.32, 8.05; HR_{trauma exposure}=2.73, 95% CI=1.25, 6.00). In models examining risk of incident SLE in association with trauma and PTSD status at baseline in 1989, there was a strong association of trauma and PTSD with SLE incidence (race- and age-adjusted models: HR_{trauma, no PTSD}=2.41, 95% CI=1.21, 4.81; HR_{PTSD 1-3}=2.13, 95% CI=0.81, 5.63; HR_{PTSD 4-7}=3.83, 95% CI=1.65, 8.91). In sensitivity analyses we did not find evidence that SLE increased risk of subsequently developing PTSD (HR_SLE=0.92, 95% CI=0.29, 2.93).

In analyses restricted to women without a diagnosis of depression prior to 2001 and with their worst event occurring prior to 2001 (N=42,677, N SLE cases=22), the association of trauma and PTSD symptoms with SLE risk was slightly stronger than in the main analyses, though confidence intervals were wide and estimates did not reach statistical significance, perhaps because there were relatively few cases (HR_{trauma, no PTSD}=3.30, 95% CI=0.72, 15.12; HR_{PTSD 1-3}=2.87, 95% CI=0.58, 14.3; HR_{PTSD 4-7}=4.06, 95% CI=0.74, 22.35).

Conclusion

In this large longitudinal study of civilian women, trauma exposure and PTSD were strongly associated with increased risk of incident SLE. We found a nearly three-fold elevated risk of incident SLE among women with probable PTSD and more than two-fold higher risk of incident SLE among women who had experienced any traumatic event compared with trauma-unexposed women.

These associations do not appear to result from either confounding or reverse causality in the association, whereby SLE caused trauma exposure or PTSD. We excluded trauma caused by illness, and findings were consistent when trauma and PTSD exposure were time lagged and when trauma and PTSD were characterized at baseline, substantially prior to most SLE diagnoses. Results were also largely similar when excluding women who had received a diagnosis of depression. Moreover, we found that SLE diagnosis itself did not predict risk of subsequently developing PTSD. These results strengthen evidence that the experience of trauma and PTSD may increase risk of subsequent SLE.

We examined whether several health risk factors accounted for the association between trauma or PTSD and SLE and found that an elevated risk of SLE in women with trauma or PTSD remained after accounting for these factors. These results recommend taking a closer look at biological changes subsequent to trauma and PTSD as potentially important mechanisms by which trauma and PTSD increase risk for SLE. To date, much of the research on biological changes associated with trauma and PTSD has been conducted examining whether higher levels of inflammation are evident^35-43 and considering alterations to the functioning of the hypothalamus-pituitary-adrenal (HPA)-axis^44-46. As both HPA-axis functioning and dysregulated inflammatory processes and have been implicated in SLE, they are promising candidates for underlying mechanisms linking trauma-PTSD with subsequent SLE¹. PTSD has been associated with dysregulation of the HPA-axis^44-46. HPA-axis function differs in persons with versus without SLE⁴⁷, although it is unknown whether these abnormalities precede SLE onset.

Animal models have demonstrated a link between PTSD and increased systemic inflammation. For example, in rats exposed to predator-induced PTSD, inflammatory micro-RNA were up-regulated in the brain, adrenal glands, and blood⁴⁸. In humans, numerous studies have found associations of PTSD and trauma exposure with elevated circulating inflammatory molecules, including tumor necrosis factor-alpha (TNF-α), C-reactive protein (CRP), and interleukin-6 (IL-6)^35-43. Additionally, persons with versus without PTSD have a more hyperactive immune response and higher circulating immunoglobulin-M levels^10,49 It is also possible that additional health risk factors that we did not measure may be relevant. For example, sleep disturbance has been associated with PTSD⁵⁰ and is prevalent among patients with SLE⁵¹. In a mouse model of SLE, sleep deprivation induced earlier disease onset⁵².

Our findings are subject to several limitations. Our sample was predominantly white; therefore, this association should be studied in women of other races. Because of the timing of our PTSD assessment, lifetime exposure to trauma and PTSD symptoms were reported after many SLE cases had occurred (N=65 of 73). Moreover, our trauma and PTSD assessment was in 2008, thus, women may have experienced trauma and PTSD between 2008 and the end of follow-up in 2013. Misclassification of trauma and PTSD-exposed women as unexposed is likely to have biased results towards the null hypothesis. Although we conducted multiple analyses to assess if our associations might be a result of SLE leading to trauma and PTSD, and did not find evidence to support this, nonetheless, our study design cannot definitively rule out this possibility. Although our sample was large, the number of incident SLE cases was moderate, limiting statistical power. Moreover, the women who enrolled in the NHS2 cohort were nurses and likely more interested in health-protective behaviors than women in the general population, and this difference may have biased our results. Our study has a number of important strengths. It is the first longitudinal study to examine this question among a healthy civilian cohort with incident SLE validated by medical record review. Participants were not selected on the basis of trauma or PTSD status, thus results are more generalizable than clinic-based studies.

Our study contributes to growing evidence that psychosocial trauma and associated stress responses lead to autoimmune disease. Identification of the biological pathways by which psychosocial trauma may increase risk for autoimmune disease is crucial and may provide greater insight into disease etiology, as well as strategies for prevention. In addition, identification of mechanisms by which trauma and PTSD are associated with increased risk of SLE may indicate mechanisms for the association of PTSD and trauma exposure with other chronic diseases. Future studies are needed to determine whether treatment for PTSD affects these pathways, and whether lifestyle interventions can reduce the risk of autoimmune disease subsequent to trauma or PTSD.

Supplementary Material

Supp FigS1

Supplemental Figure: Timeline of data collection, Nurses' Health Study II