Prospective data from the Women’s Health Initiative on depressive symptoms,stress,and inflammation

Abstract

This study examined the longitudinal association of depressive symptoms and stressful life events with inflammation in the Women’s Health Initiative. Women aged 50 years and older (N = 7477) completed questionnaires assessing depressive symptoms and stressful life events at baseline and 15 years later. Serum measures of C-reactive protein were collected at both assessments. In bivariate analyses, C-reactive protein predicted 15-year depressive symptoms and stressful life events (ps < .03) and baseline depressive symptoms and stressful life events predicted later C-reactive protein (ps < .03). These longitudinal relationships were not maintained in multivariate adjusted analyses. Combined with previous research, this suggests the relationship between depression, stressful life events and inflammation attenuates with time.

Keywords

depression inflammation psychological distress stress women’s health

Inflammation, an immune response to pathogens or injury, is theorized to be one pathway leading to depression (Dantzer et al., 2008). While acute inflammation is helpful in response to pathogens or injury, chronic inflammation has been associated with greater symptoms of depression (Dowlati et al., 2010; Howren et al., 2009) and may be one way through which depression increases the risk for poor health such as cardiovascular disease (Musselman et al., 1998). Research has supported the theory that depression and depressive symptoms are related to increases in chronic inflammation, although the relationship may be bidirectional with depression leading to increased inflammation and vice versa (Dowlati et al., 2010; Howren et al., 2009). One meta-analysis of 24 studies showed that a diagnosis of major depression in healthy adults was associated with increases in tumor necrosis factor (TNF)-alpha and interleukin (IL)-6, markers of general and chronic inflammation, but unrelated to other markers of inflammation IL-1beta, IL-4, IL-8, IL-10, and interferon (IFN)-gamma (Dowlati et al., 2010). Another meta-analysis of 62 articles investigated the relationship of depression with C-reactive protein (CRP) and IL-6 (Howren et al., 2009). This meta-analysis found that CRP and IL-6 were related to greater depressive symptoms in healthy populations, and diagnosed depression was related to increased inflammatory markers. However, the association was less pronounced for depressive symptoms than for a diagnosis of depression.

A few studies have investigated the longitudinal relationship between depressive symptoms and chronic inflammation and found significant relationships. One study of 263 healthy older adults (Stewart et al., 2009) showed that baseline depressive symptoms predicted 6-year CRP; however, the reverse relationship was not observed—baseline CRP did not predict depressive symptoms at 6 years. Another study of 3609 adults found depressive symptoms predicted CRP 2 years later, with physical activity mediating this relationship (Hamer et al., 2009). In a study of 1781 pre- and peri-menopausal women, CRP was found to predict depressive symptoms 1 year later, although the effect was small (Matthews et al., 2010). A review of eight longitudinal studies found only small effects (r = .03) for the longitudinal relationship between depression and chronic inflammation over an average of 5 years and found a publication bias suggesting studies with positive findings were more likely to be published (Valkanova et al., 2013). When the authors corrected for this bias, the effects were reduced but still significant. Overall, these studies suggest a positive, bidirectional relationship between chronic inflammation and depressive symptoms, but the effects tend to be small and follow-ups are typically within 5 years.

The Women’s Health Initiative (WHI) Long Life Study (LLS) provides an opportunity to examine the longitudinal relationship of chronic inflammation with depressive symptoms and stressful life events (SLEs) in older women to examine whether chronic inflammation could be the mechanism between depressive symptoms and risk of disease. The WHI LLS measured CRP at baseline and approximately 15 years later allowing an examination of the relationship between depressive symptoms and chronic inflammation with a longer follow-up period than was typical in previous studies. Consistent with prior cross-sectional and longitudinal studies summarized above (Dowlati et al., 2010; Howren et al., 2009; Valkanova et al., 2013), we hypothesized a bidirectional relationship between inflammatory markers and depressive symptoms or SLEs 15 years later with chronic inflammation predicting later depressive symptoms and depressive symptoms predicting later inflammation. This study has a longer time between assessments than previous studies. As depression can be related to stress in that stress may be both a cause of depression (Monroe et al., 2007) and result of depression (Wu and Andersen, 2010), we also examined the relationship of stress (measured via SLEs) with chronic inflammation.

Method

Participants and procedures

The WHI studies began in 1993 and had two major components: the Clinical Trials and the Observational Study (The Women’s Health Initiative Study Group, 1998). Over 40 clinical centers across the United States accrued a diverse study population of postmenopausal women aged 50–79 years (N = 161,808). The first phase of the program ended in 2005 and all women whom were still alive and under active surveillance had the option to re-consent for participation in two Extension Study (ES) phases (2005–2010, n = 115,400 who consented to ES1, and 2010–2015, n = 93,500 who consented to ES2). During ES2, women who had biomarkers such as CRP collected at baseline and had measures of depressive symptoms and SLEs available were eligible to participate in the LLS visits, which occurred between 2011 and 2013. In 2011–2012, women also completed an additional measure of depressive symptoms and SLEs. On average, the follow-up assessments occurred 15.37 years (1.11 standard deviation (SD)) after baseline assessments. In total, 8017 women provided blood samples at baseline and follow-up, which were then assayed for CRP. We excluded women who had CRP values over 15 mg/L at either baseline or the follow-up as these levels could indicate acute infection, and our primary interest was chronic inflammation (McIntyre et al., 1997; Stewart et al., 2009). Our final sample included 7477 women.

Measures

Depressive symptoms

The six-item version of the Center for Epidemiological Studies Depression scale–short form (CESD-SF; Tuunainen et al., 2001) was completed at baseline and follow-up. Participants rated each item on a 0 (rarely) to 3 (most or all the time) scale. One item, enjoying life, is reverse-scored. A score of 5 or more is indicative of elevated depressive symptoms. For the sum scores, the Cronbach’s alpha at baseline was .66. We utilized item response theory (IRT), specifically the graded response model, to calculate interval-level scores from the six items, and higher scores indicate more depressive symptoms. The IRT analyses estimated the level of depressive symptoms measured by each item, as well as the accuracy, and then used this to calculate interval-level scores. Scores were calculated so the mean was 0 and the SD was 1.

SLEs

At baseline and follow-up, women indicated whether 11 SLEs (e.g. ill spouse, relationship break-up, financial problems) occurred in the year prior to measurement. A count of SLEs was then calculated. Due to skew, a natural log transform was applied to the data.

CRP

Blood was drawn at baseline and follow-up to measure CRP. Within 2 hours of blood draws, blood samples were processed including centrifuge, freezing, and shipping to the central laboratory at the Fred Hutchinson Cancer Research Center in Seattle, Washington. Final processing of all samples occurred within 36 hours of blood draw. CRP assays were run in batches 6–12 months after draw by the University of Minnesota Advanced Research and Diagnostic Laboratory. Due to skew, we used a natural log transform on the CRP values.

Covariates

Baseline questions asked women for their age, highest education level completed, marital status, race/ethnicity, smoking behavior (never, previous, current (Willemse et al., 2004), alcohol use (servings per week), beta-blocker use (Ohtsuka et al., 2001), anti-inflammatory medication use, hormone therapy use (Kirschbaum et al., 1999), body mass index (Festa et al., 2001), physical activity (Ford, 2002), and diet (Giugliano et al., 2006). Physical activity was calculated using a questionnaire that asked participants about intensity and time spent in light, moderate, and vigorous activity and walking (Nguyen et al., 2013; Seguin et al., 2012). Metabolic expenditure (MET) per week was then calculated. Diet was based on a questionnaire, the Food Frequency Questionnaire, and used the Healthy Eating Index 2005 scoring (Guenther et al., 2008).

Analysis plan

Path analyses using a cross-lagged panel design were used to examine the longitudinal relationship of depression and SLEs with inflammation. This design tests directional paths from baseline variables (depressive symptoms/SLEs, inflammation) with all follow-up variables (inflammation, depressive symptoms/SLEs) using multiple regression. We used full information maximum likelihood to estimate the model parameters in Amos 22. All analyses controlled for age, education, marital status, ethnicity, body mass index, anti-inflammatory medication use (including nonsteroidal anti-inflammatory drugs (NSAIDs)), beta-blocker use, smoking, alcohol use, physical activity, and diet. A principal components analysis was used to reduce the number of covariates to two component scores but still control for confounding variables (age, education, smoking behavior, alcohol use, body mass index, physical activity, and diet). Marital status, race/ethnicity, and medication use were entered separately as covariates. One path analysis comparing depressive symptoms with CRP was constructed, and a second path analysis was constructed comparing SLEs with CRP. The paths from baseline depressive symptoms and SLEs to follow-up CRP tested whether baseline depressive symptoms or SLEs were related to subsequent CRP. The paths from baseline CRP to follow-up depressive symptoms and SLEs tested whether baseline inflammation was related to subsequent depressive symptoms or stressful events. To evaluate model fit, we used the root mean square error of approximation (RMSEA; Browne and Ceduk, 1993) and comparative fit index (CFI). The values less than .08 indicate good fit for the RMSEA and the values above .94 for the CFI indicate good fit. We also report the chi-square test of perfect fit below; however, for samples over 200, this test can be significant and indicate poor fit due to sample size alone and not due to actual poor fit (Hu and Bentler, 1999).

Results

Descriptive statistics are reported in Table 1. Consistent with the overall WHI, women were an average of 63.47 years old at baseline (SD 6.91 years). Due to the sampling procedure of the LLS, only women who were Caucasian, Hispanic, or African American were included in the sample. Slightly more than half the sample were married (58.6%) and had baseline incomes above US$35,000 (61.2%), the median income in the United States in 1993 when WHI started, but a substantial portion of the sample was not married (41.4%) or had incomes below the median (38.8%). At baseline, 14.3 percent of women had elevated depressive symptoms indicative of possible depression. At the follow-up, 14.4 percent of women had elevated depressive symptoms.

Table 1.

Descriptive statistics of the WHI women in this study (N = 7477).

Variable	Mean (SD) or % (n)
Age at baseline	63.47 (6.91)
Ethnicity
Caucasian	53.8% (4021)
African American	30.2% (2256)
Hispanic	16.0% (1200)
Married or long-term relationship	58.6% (4381)
Income
Unknown	2.8% (209)
<US$35,000/year	41.6% (3107)
>US$35,000/year	61.2% (4579)
Education
Unknown	0.5% (39)
Less than bachelor’s	60.0% (4487)
Bachelor’s or higher	39.5% (2951)
BMI	28.61 (5.59)
Clinical Trial participant	77.0% (5761)
Hormone Replacement Trial, control	31.7% (2373)
Hormone Replacement Trial, intervention	31.7% (2370)
Smoking at baseline	6.0% (451)
Alcohol use at baseline	65.6% (4908)
Baseline physical activity, METs per week	12.12 (13.46)
Baseline HEI score	66.74 (10.73)
	Mean (SD)
Baseline values on variables
Depressive symptoms, IRT scores, mean 0, SD 1	−.003 (.753)
Number of stressful life events, possible range from 0 to 11	1.80 (1.47)
CRP values	3.33 (3.04)
Follow-up values on variables
Depressive symptoms, IRT scores, mean 0, SD 1	−.008 (.732)
Number of stressful life events, possible range from 0 to 11	1.47 (1.31)
CRP values	2.54 (2.54)

WHI: Women’s Health Initiative; SD: standard deviation; BMI: body mass index; MET: metabolic expenditure; HEI: Healthy Eating Index; IRT: item response theory; CRP: C-reactive protein.

Initial bivariate correlations showed some cross-sectional and longitudinal relationships between depressive symptoms, SLEs, and inflammation. Baseline CRP was associated with baseline depressive symptoms (r = .045, p < .001) and baseline SLEs (r = .074, p < .001), but follow-up CRP was not related to follow-up depressive symptoms (r = −.012, p = .355) or follow-up SLEs (r = .023, p = .092). Baseline depressive symptoms (r = .027, p = .028) and SLEs (r = .053, p < .001) were associated with CRP at follow-up. Baseline CRP was associated with depressive symptoms (r = .029, p = .021) and SLEs (r = .048, p < .001) at the follow-up.

The relationships seen in the preliminary bivariate analyses were not maintained in the controlled path analyses. The results for the path analyses are reported in Figure 1. For depressive symptoms, model fit was acceptable (χ² = 7.643, p = .006; CFI = .999; RMSEA = .018). CRP and depressive symptoms were still significantly related at baseline in the multivariate model (p < .001). Both depressive symptoms at baseline and CRP at baseline predicted their respective values at the 15-year follow-up (both ps < .001). However, the level of depressive symptoms at baseline was not significantly associated with CRP at the 15-year follow-up (p = .966). CRP at baseline was not significantly associated with depressive symptoms at the 15-year follow-up (p = .984).

Figure 1.

Results of cross-lagged path analyses comparing depressive symptoms and number of stressful life events (SLEs) to inflammation (C-reactive protein (CRP)). Covariates in both models were age, highest education level completed, marital status, race/ethnicity, smoking behavior, alcohol use, beta-blocker use, anti-inflammatory medication use, body mass index, physical activity and diet.

For SLEs, the model fits well (χ² = .065, p = .799; CFI = 1.000; RMSEA < .001). CRP and SLEs were still related at baseline (p < .001). SLEs at baseline significantly predicted SLEs 15 years later (p < .001), and CRP at baseline significantly predicted CRP 15 years later (p < .001). However, SLEs at baseline did not significantly predict CRP 15 years later (p = .233), and CRP at baseline did not significantly predict SLEs 15 years later (p = .933).

As a final step, we calculated post hoc power achieved using a standard multiple regression model and reran stratified analyses by ethnicity as this could have moderated the relationship of inflammation with depression. We achieved 45.4 percent power for baseline depressive symptoms predicting follow-up CRP and for baseline CRP predicting follow-up depressive symptoms. For SLEs predicting follow-up CRP, we achieved 94.7 percent power. The amount of variance accounted for by the baseline predictors of interest was very small (less than .1%–.3% of the total variance). The results from the path analyses did not change when analyses were stratified by ethnicity.

Discussion

This study examined the cross-sectional and prospective associations of depressive symptoms, SLEs and inflammation (CRP) in a large cohort of older women at study entry and 15 years later. CRP was significantly, positively related to depressive symptoms and SLEs at baseline. Although longitudinal bivariate relationships were significant between depressive symptoms and SLEs with inflammation, more controlled path analyses showed no longitudinal relationship between baseline depressive symptoms and SLEs with inflammation 15 years later. Conversely, baseline inflammation was not related to depressive symptoms and SLEs at the 15-year follow-up in controlled analyses. Effect sizes (<1% of the variance) were consistent with previous studies (Valkanova et al., 2013). Given the adequate power, observed effect size, and consistency of our baseline results with previous cross-sectional research, this study suggests that the longitudinal relationship between depressive symptoms and inflammation is not maintained at 15 years.

Although the study results initially may seem at odds with previous research, the methodology of previous research should be considered. Previous studies utilized shorter follow-up assessment times, with an average of 5 years (Valkanova et al., 2013). It is likely that depressive symptoms and stressful events are related to inflammation at least up to 5 but not 15 years later. It is unlikely that these disparate findings from previous research were due to age as the majority of previous studies included middle-aged to older adults as in our sample. However, due to the prolonged follow-up in this study, the additional effect of age at the follow-up may have led to a restricted range of depressive symptoms as distress tends to decrease with age (Stone et al., 2010). The levels of CRP were comparable between our study (2.54–3.33 mg/L) and previous studies (2.2–3.69 mg/L) so a restricted range on inflammation is unlikely to explain the null results. However, gender may also explain the different findings as the relationship between depression and inflammation, particularly for CRP, may be stronger in men (Morris et al., 2011; Vetter et al., 2013). Two studies with strong results utilized mixed gender samples (Hamer et al., 2009; Stewart et al., 2009), and the results from the Study of Women’s Health Across the Nation (SWAN) while still significant from inflammation to depression, had only a trend toward significance from depression to inflammation (Matthews et al., 2010). In addition to the long follow-up interval, the exclusive focus on women might explain our longitudinal null findings. Overall, these results suggest that the relationship between depressive symptoms and SLEs with inflammation may not exist 15 years later for middle-aged and older women.

Although this study had several strengths including a large sample with a long follow-up interval, some limitations are noted. First, a self-report measure of depressive symptoms rather than a diagnosis of depression was used. However, our methodology is consistent with many prior studies on this topic and a substantial minority had elevated depressive symptoms at both times. Also, numerous studies have shown that depression is a continuous, not categorical, construct (Hankin et al., 2005; Ruscio and Ruscio, 2000). Second, only one marker of inflammation, CRP, was available and only at two times. Although stress has been linked to many other immune markers such as natural killer cell lysis (Segerstrom and Miller, 2004) and more severe forms of stress (childhood adversity, poverty) have been linked to CRP (Hansel et al., 2010), few have examined SLEs and CRP. It remains unclear whether relationships would have emerged at 15 years between depressive symptoms and number of SLEs and additional inflammatory markers.

The study results have implications for future research. Given the literature on gender differences in the association of depression and inflammation and our findings with a relatively large sample, biomarkers of depression and stress may differ substantially by gender and future research should explore which biomarkers might be most useful to measure in women. Also, causal or risk relationships between psychosocial factors and inflammation (at least as measured by CRP) would be expected to diminish with time and distance from the precipitating event or onset of symptoms. Previous studies have supported the relationship of biomarkers with depression within shorter time periods such as 5 or 6 years and, combined with our results, suggest that the mechanism for disease risk, such as cardiovascular disease, from depressive symptoms could be inflammation within 5 years but not 15 years.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The WHI program was funded by the National Heart, Lung, and Blood Institute; National Institutes of Health; US Department of Health and Human Services through contracts HHSN268201100046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, HHSN268201100004C, and HHSN271201100004C. Dr Jones was funded by a fellowship from the National Institute on Aging (T32 AG027677).

References

Browne

Ceduk

(1993) Alternate ways of assessing model fit. In: Bollen

Long

(eds) Testing Structural Equation Models. London: SAGE, pp. 136–162.

Dantzer

O’Connor

Freund

. (2008) From inflammation to sickness and depression: When the immune system subjugates the brain. Nature Reviews: Neuroscience 9(1): 46–56. DOI: 10.1038/nrn2297.

Dowlati

Herrmann

Swardfager

. (2010) A meta-analysis of cytokines in major depression. Biological Psychiatry 67(5): 446–457. DOI: 10.1016/j.biopsych.2009.09.033.

Festa

D’Agostino

Jr Williams

. (2001) The relation of body fat mass and distribution to markers of chronic inflammation. International Journal of Obesity and Related Metabolic Disorders 25(10): 1407–1415. DOI: 10.1038/sj.ijo.0801792.

Ford

(2002) Does exercise reduce inflammation? Physical activity and C-reactive protein among US adults. Epidemiology 13(5): 561–568. DOI: 10.1097/01.Ede.0000023965.92535.C0.

Giugliano

Ceriello

Esposito

(2006) The effects of diet on inflammation—Emphasis on the metabolic syndrome. Journal of the American College of Cardiology 48(4): 677–685. DOI: 10.1016/j.jacc.2006.03.052.

Guenther

Reedy

Krebs-Smith

(2008) Development of the Healthy Eating Index-2005. Journal of the American Dietetic Association 108(11): 1896–1901. DOI: 10.1016/j.jada.2008.08.016.

Hamer

Molloy

de Oliveira

. (2009) Persistent depressive symptomatology and inflammation: To what extent do health behaviours and weight control mediate this relationship? Brain, Behavior, and Immunity 23(4): 413–418. DOI: 10.1016/j.bbi.2009.01.005.

Hankin

Fraley

Lahey

. (2005) Is depression best viewed as a continuum or discrete category? A taxometric analysis of childhood and adolescent depression in a population-based sample. Journal of Abnormal Psychology 114(1): 96–110. DOI: 10.1037/0021-843X.114.1.96.

10.

Hansel

Hong

Camara

. (2010) Inflammation as a psychophysiological biomarker in chronic psychosocial stress. Neuroscience and Biobehavioral Reviews 35(1): 115–121. DOI: 10.1016/j.neubiorev.2009.12.012.

11.

Howren

Lamkin

Suls

(2009) Associations of depression with C-reactive protein, IL-1, and IL-6: A meta-analysis. Psychosomatic Medicine 71: 171–186.

12.

Bentler

(1999) Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal 6(1): 1–55.

13.

Kirschbaum

Kudielka

Gaab

. (1999) Impact of gender, menstrual cycle phase, and oral contraceptives on the activity of the hypothalamus-pituitary-adrenal axis. Psychosomatic Medicine 61(2): 154–162.

14.

McIntyre

Harper

Macdougall

. (1997) Serum C-reactive protein as a marker for infection and inflammation in regular dialysis patients. Clinical Nephrology 48(6): 371–374.

15.

Matthews

Schott

Bromberger

. (2010) Are there bi-directional associations between depressive symptoms and C-reactive protein in mid-life women? Brain, Behavior, and Immunity 24(1): 96–101. DOI: 10.1016/j.bbi.2009.08.005.

16.

Monroe

Slavich

Torres

. (2007) Major life events and major chronic difficulties are differentially associated with history of major depressive episodes. Journal of Abnormal Psychology 116(1): 116–124. DOI: 10.1037/0021-843x.116.1.116.

17.

Morris

Zhao

Ahmed

. (2011) Association between depression and inflammation—Differences by race and sex: The META-Health study. Psychosomatic Medicine 73(6): 462–468. DOI: 10.1097/PSY.0b013e318222379c.

18.

Musselman

Evans

Nemeroff

(1998) The relationship of depression to cardiovascular disease: Epidemiology, biology, and treatment. Archives of General Psychiatry 55(7): 580–592.

19.

Nguyen

Herting

Kohen

. (2013) Recreational physical activity in postmenopausal women is stable over 8 years of follow-up. Journal of Physical Activity & Health 10(5): 656–668.

20.

Ohtsuka

Hamada

Hiasa

. (2001) Effect of beta-blockers on circulating levels of inflammatory and anti-inflammatory cytokines in patients with dilated cardiomyopathy. Journal of the American College of Cardiology 37(2): 412–417.

21.

Ruscio

(2000) Informing the continuity controversy: A taxometric analysis of depression. Journal of Abnormal Psychology 109(3): 473–487.

22.

Segerstrom

Miller

(2004) Psychological stress and the human immune system: A meta-analytic study of 30 years of inquiry. Psychological Bulletin 130(4): 601–630. DOI: 10.1037/0033-2909.130.4.601.

23.

Seguin

Lamonte

Tinker

. (2012) Sedentary behavior and physical function decline in older women: Findings from the women’s health initiative. Journal of Aging Research 2012: 271589 (10 pp.). DOI: 10.1155/2012/271589.

24.

Stewart

Rand

Muldoon

. (2009) A prospective evaluation of the directionality of the depression-inflammation relationship. Brain, Behavior, and Immunity 23(7): 936–944. DOI: 10.1016/j.bbi.2009.04.011.

25.

Stone

Schwartz

Broderick

. (2010) A snapshot of the age distribution of psychological well-being in the United States. Proceedings of the National Academy of Sciences of the United States of America 107(22): 9985–9990. DOI: 10.1073/pnas.1003744107.

26.

The Women’s Health Initiative Study Group (1998) Design of the Women’s Health Initiative clinical trial and observational study. Controlled Clinical Trials 19(1): 61–109.

27.

Tuunainen

Langer

Klauber

. (2001) Short version of the CES-D (Burnam screen) for depression in reference to the structured psychiatric interview. Psychiatry Research 103(2–3): 261–270.

28.

Valkanova

Ebmeier

Allan

(2013) CRP, IL-6 and depression: A systematic review and meta-analysis of longitudinal studies. Journal of Affective Disorders 150(3): 736–744. DOI: 10.1016/j.jad.2013.06.004.

29.

Vetter

Wadden

Vinnard

. Power-Up Research Group (2013) Gender differences in the relationship between symptoms of depression and high-sensitivity CRP. International Journal of Obesity 37(Suppl. 1): S38–S43. DOI: 10.1038/ijo.2013.95.

30.

Willemse

Postma

Timens

. (2004) The impact of smoking cessation on respiratory symptoms, lung function, airway hyperresponsiveness and inflammation. European Respiratory Journal 23(3): 464–476.

31.

Andersen

(2010) Stress generation over the course of breast cancer survivorship. Journal of Behavioral Medicine 33(3): 250–257. DOI: 10.1007/s10865-010-9255-y.