Bereavement is Associated with Reduced Systemic Inflammation: C-Reactive Protein Before and After Widowhood

Abstract

Background:

Bereavement is associated with poorer health and early mortality. Increased systemic inflammation is one pathophysiological pathway thought to explain this health risk. However, few studies have examined systemic inflammation before and after widowhood.

Purpose:

The current study examined the associations between inflammation and widowhood status before and after bereavement in a sample of married adults who became widowed between assessments in the English Longitudinal Study of Ageing.

Methods:

We examined levels and change over time in systemic inflammation, as assessed by C-reactive protein (CRP), among participants who became bereaved (n = 199). We then compared these results to a sample of participants whose spouse remained living, selected using a propensity score matching algorithm (n = 199).

Results:

Contrary to expectations, widowed participants’ CRP decreased following bereavement, d = −0.29, p < .001. Change in CRP was not associated with pre-loss depression levels, caregiving status, marital quality, number of chronic diseases, prescribed medications, body mass index, age, or sex. Compared to continuously married participants, widowed participants’ evidenced a significantly greater decrease in CRP after their spouse’s death, β = −0.14, p < .001.

Conclusions:

Widowed adults’ systemic inflammation decreased significantly following bereavement, both as a group and compared to people who remained married. We discuss possible explanations for this counterintuitive finding, including the measure of inflammation used in the study and the timing of the study measurements.

1. Introduction

Being involved in high-quality social relationships is associated with better health and longevity (1–2). Unfortunately, the opposite is also true: the loss of close social relationships through divorce or bereavement is associated with increased risk of poor health and early death (3–7). The death of a spouse can be particularly detrimental (5–8)—people who become widowed are at a 40% increased risk for early death, even after adjusting for relevant covariates (6). Despite existing evidence for poorer health following bereavement, mechanisms underlying these associations are less well studied.

One pathophysiological pathway that may explain the association between widowhood and poorer health is inflammation (8–10). The social signal theory of depression suggests that the social threat and stressors experienced during bereavement promote a pro-inflammatory phenotype, which subsequently increases depressive symptoms and social withdrawal (11). Although a number of studies have examined the association of widowhood and systemic inflammation (8–10,12), the majority have examined levels of inflammation in the acute period following bereavement. Such studies are valuable, but have two notable limitations. First, these studies have not examined pre-widowhood levels of inflammation. Bereaved adults may evidence higher levels of systemic inflammation prior to widowhood (i.e., a pre-widowhood effect), which would have implications for how to address health risk associated with increased systemic inflammation. Indeed, prior studies have found pre-widowhood effects for depression, functional status, and quality of life (13–14). Second, past studies have typically examined acute effects on systemic inflammation in the first year following bereavement. Because systemic inflammation impacts health over time (15), it is important to examine whether higher levels of inflammation following a spouse’s death are maintained longer term, and might explain increased risk of poorer health. Systemic inflammation levels in the years following widowhood are of particular interest because the psychological and physical health impacts of bereavement are thought to dissipate over time (16–17), but it is unknown if acute effects on systemic inflammation subside as well.

The association between bereavement and inflammation following widowhood may vary according to psychosocial characteristics. For example, people who have higher depressive symptoms (18), complicated grief (10), or higher genetic risk (19) are at greater risk of increased inflammation after bereavement. In contrast, people who are caregivers for a sick spouse may experience some degree of relief after their partner dies (20), which could attenuate increases in inflammation (21), perhaps due to the end of an effort-reward imbalance (22) associated with the stress of spousal caregiving (23).

To our knowledge no studies of widowhood have accounted for (or compared) pre-widowhood levels of systemic inflammation or examined change in inflammation over several years between bereaved individuals and their continuously married counterparts. The current study used longitudinal data from the English Longitudinal Study of Ageing (ELSA) to examine systemic inflammation before and after bereavement. We hypothesized that widowed participants would have greater increases in inflammation following bereavement as compared to their non-bereaved counterparts. In addition, we hypothesized that widowed participants would evidence higher levels of inflammation before bereavement and increased inflammation after bereavement relative to participants who remained married. We also conducted exploratory analyses to determine whether individual differences (e.g., depression) moderated change in inflammation and if survivorship bias may have affected the results.

2. Material and Method

2.1. Participants and Study Design

ELSA collected a representative sample of adults in England over the age of 50 (24). As described by ELSA documentation (24), the sample was drawn from households that had previously responded to the Health Survey for England between 1998 and 2011 and were enrolled for assessment in ELSA. Detailed information about eligibility criteria have been previously published (25). Participants were enrolled by the ELSA study and provided informed consent. ELSA currently has eight waves of data collected every two years from 1998 to 2017 (24). These assessments were supplemented by home visits by a nurse every other wave (Wave 2, 4, 6, and 8), during which blood samples were collected and assayed. For the present study, we included all participants who participated in the blood draws and were married at an initial wave (Wave 2, 4, or 6), then widowed by the time of the next wave (Wave 4, 6, or 8 respectively). This resulted in a final sample of 199 bereaved participants assessed pre-and post-widowhood (4 years apart). For all participants, the wave prior to their partner’s death was used as the pre-widowhood assessment, and the wave following their partner’s death was used as the post-widowhood assessment. Participants (n = 2,289) who were married at an initial wave (Wave 2, 4 or 6) and remained married at the next wave (Wave 4, 6, or 8) were used to create a comparison sample of still-married adults matched to the widowed sample using a propensity score algorithm (n = 199). Sample characteristics are reported in Table 1.

Table 1.

Clinical and Demographic Characteristics of the Study Sample

	Widowed n = 199	Matched sample n = 199	Continuously married n = 2,289
Sex	72.4%	72.4%	52.5%
Caregiving status*	24.6%	3.5%	4.5%
Smoking status	6.6%	7.7%	10.5%
Age	70.7 ± 8.2	70.6 ± 7.8	65.0 ± 7.8
Depressive symptoms	1.2 ± 1.7	1.1 ± 1.5	1.0 ± 1.6
Marital quality*	3.3 ± 0.6	3.4 ± 0.5	3.4 ± 0.5
Chronic diseases	0.7 ± 0.7	0.7 ± 0.6	0.6 ± 0.6
Medications	0.6 ± 0.9	0.6 ± 0.8	0.4 ± 0.7
BMI	28.5 ± 5.6	27.9 ± 5.1	27.8 ± 4.5
Baseline lgCRP	1.22 ± 0.4	1.25 ± 0.4	1.19 ± 0.4
Follow up lgCRP*	1.13 ± 0.4	1.25 ± 0.4	1.18 ± 0.4

Note: Data are percentages and means ± standard deviations from all available data unless otherwise noted. Sex is percentage women, marital status is percentage married, smoking status is percentage providing care for their spouse, smoking status is percentage smoking. Continuously married participants (n = 2,289) values were used by selecting a random subsample of eligible occasions. BMI = body mass index.

^*indicates factors for which the widowed group and the propensity score-matched group differed significantly (p < .05).

2.2. Measures

2.2.1. Marital Status.

Participants self-reported their marital status at each ELSA assessment.

2.2.2. Systemic Inflammation.

Blood samples were collected in participants’ homes by trained nurses. Samples were analyzed using the N Latex C-reactive protein (CRP) mono Immunoassay on the Behring Nephelometer II Analyzer (26) to produce CRP levels (mg/l). CRP values above 10 mg/l were recoded as missing, as such levels likely represent acute infection (27). CRP values were then log-transformed, matching standard practice (28). The mean log transformed CRP values 1.24 at the pre-assessment and 1.19 at the post-assessment. There were 22 (5.5%) and 34 (8.5%) cases of CRP scores greater above 10 mg/l for the pre-and post-assessments, respectively. The number of excluded CRP scores did not vary significantly by widowed status.

2.2.3. Depressive Symptoms.

The Center for Epidemiologic Studies Depression Scale (29) assessed participants’ self-reported symptoms of depression using eight items. Higher scores reflected more depressive symptoms and emotional disturbance.

2.2.4. Marital Quality.

Marital quality was assessed using six items (30) measuring participants’ perception of marital support and strain. Higher scores indexed better quality.

2.2.5. Chronic Diseases.

Participants reported whether they had four major categories of chronic diseases—cardiovascular disease (e.g., stroke, myocardial infarction), diabetes, lung disease, or cancer. A count of these variables (range 0–4) assessed number of diseases.

2.2.6. Prescribed medications.

Participants reported whether they were currently taking medications for the chronic diseases they reported at each ELSA assessment, e.g., whether they were taking medication for high blood pressure, diabetes, etc. The number of prescribed medications was a count of how many medications the participants endorsed taking (range 0–4).

2.2.7. Body Mass Index (BMI).

Participants’ weight and height were assessed by a trained nurse in participants’ homes and BMI was calculated using the standard formula.

2.2.8. Smoking.

Participants self-reported whether they were currently smoking cigarettes.

2.2.9. Caregiving status.

Participants reported whether they were their spouse’s caregiver.

2.3. Data Analysis

We first examined change in pre-to post-widowhood CRP for participants who became widowed (n = 199), as well as whether change in CRP was associated with age, sex, pre-loss depression levels, marital quality, smoking status, caregiving status. We next compared the widowed participants to participants that remained married during the matching waves (n = 2,289). As would be expected, the widowed sample (70.6 years old, SD_age = 8.2; 72.9% women) was significantly older (d = 0.95, p < .001) and made up of more women (d = 0.24, p < .001) than the continuously married group (63.5 years old, SD_age = 7.7; 47.5% women). To account for these differences, we used the propensity score matching algorithm (31) integrated within SPSS version 26. The sample was selected using a 1:1 ratio within each possible study period by propensity to become widowed using logistic regression based on the predictors of age and sex (match tolerance = 0.01, selected without replacement). Widowed participants who were not matched within study period were then matched using the full sample (match tolerance = 0.05). This method eliminated differences in age and sex between the widowed and still-married groups (ds < .02, ps > .85). Using the propensity-matched sample, we then tested whether widowhood status predicted pre-widowhood CRP levels and change in CRP from pre-to post-widowhood¹. Age, sex, depressive symptoms, marital quality, smoking, chronic diseases, prescribed medications, BMI and caregiving status were included as covariates in all models, and were also tested as moderators in exploratory analyses. Full information maximum likelihood estimation in MPLUS version 8.3 (32) was used to account for missing data. This method incorporates all available data and produces estimates that outperform other missing-data treatments when data are missing at random (33).

3. Results

3.1. Widowed Sample

Change in CRP within the widowed sample was tested using a paired-sample t-test. In contrast to our hypothesis, widowed participants evidenced a significant decrease in CRP from pre-to post-widowhood, Cohen’s d = −0.29, 95% CI [−0.14, −0.44], p < .001, a reduction of 0.46 mg/L. Sex, age, depressive symptoms, marital quality, smoking, chronic diseases, prescribed medications, BMI, and caregiving status were not associated with change in CRP (Table 2). Of note, there was a small to moderate non-significant (p = .074) negative association between body mass and change in CRP (no other associations had a p < .10).

Table 2.

The Association of C-reactive Protein with Study Variables in Bereaved Sample Outcome: Post-widowhood CRP

N = 199	B	95% CI	P
Pre-widowhood CRP	0.75	[0.65, 0.84]	< .001
Age	0.07	[−0.04, 0.17]	.234
Sex	0.00	[−0.11, 0.11]	.990
BMI	−0.11	[−0.24, 0.01]	.074
Smoking status	0.09	[−0.04, 0.22]	.169
Chronic diseases	0.04	[−0.10, 0.18]	.591
Prescribed medications	−0.11	[−0.25, 0.03]	.119
Caregiving status	0.00	[−0.11, 0.11]	.998
Depressive symptoms	0.03	[−0.08, 0.13]	.649
Marital quality	0.04	[−0.08, 0.16]	.494

Note: CRP = C-reactive protein ; BMI = body mass index.

3.2. Comparisons to Non-Widowed Sample

In covariate-adjusted models, there was no significant difference in CRP level at the pre-widowhood assessment between people who became widowed compared to those who remained married, β = −0.06, 95% CI [−0.15, 0.04], p = .256. From the pre-to post-widowhood assessment, however, people who became widowed showed a significant reduction in CRP compared to people who remained married, β = −0.14, 95% CI [−0.22, −0.06], p < .001.

3.3. Exploratory Analyses

3.3.1. Moderation.

Sex, age, depressive symptoms, marital quality, smoking, BMI, chronic diseases, prescribed medications, and caregiving status did not moderate the association between widowhood and change in CRP level, ps > .28.

3.3.2. Survivorship Bias.

It was possible that there were people who became bereaved but did not survive until the next study wave. These participants would have been excluded from our study sample. ELSA included proxy reports collected after participants died that asked whether the deceased participant had a surviving spouse at the time of death. Participants who were married at an initial assessment but were reported as not having a surviving spouse at the time of death by their proxy likely had become widowed and then had died in the period after becoming widowed and the next study wave. Using proxy reports, we identified 23 ELSA participants who completed a nurse visit while married, then became widowed but did not survive to the next study wave. This sample was small in size, but significantly more male (33.8% more male, d = 0.36, p < .001) and older (8.8 years older, d = 0.99, p < .001) than people who became widowed and survived to the next study wave. The deceased participants’ pre-widowhood CRP levels were significantly higher than the widowed participants who survived to the next study wave when controlling for age and sex, β = 0.16, 95% CI [0.02, 0.43], p = .029.

4. Discussion

The current study examined associations between systemic inflammation and widowhood using a sample drawn from a representative longitudinal cohort of English older adults (N = 398). In contrast to our predictions, widowed participants showed a significant decrease in inflammation from pre-to post-widowhood, both within-participants and when compared to participants whose spouse remained alive. In addition, we did not find any evidence to support pre-widowhood effects (13–14) on inflammation—participants who became widowed had comparable levels of inflammation levels prior to widowhood compared to those who did not experience subsequent bereavement. It is worth noting, however, that the few participants who experienced widowhood but died before the post-assessment had elevated CRP. This is in line with current knowledge linking inflammation and widowhood to mortality.

What might explain our counterintuitive results? It is possible that the results reflect meaningful reductions in systemic inflammation years after bereavement. For example, prior work has found that caregivers and family members of people with high levels of chronic disease burden evidence improvements in emotional distress after the death of a care recipient (34–35). Following a period of risk for poorer health in the immediate aftermath of spousal loss, it is possible that decreased distress due to relief from the challenges associated with a spouse approaching death might translate to decreased inflammation. Future research should assess these possibilities.

Results may also be due to study artifacts. We propose three possibilities related to: (1) the measure of inflammation used in this study, (2) unmeasured variation in responses to bereavement, and (3) bias related to the timing of the assessment periods. First, it is possible that the measure of inflammation used in this study, CRP, is affected differently by bereavement than other markers of inflammation. A recent systematic review (8) found that inflammation increased following bereavement measured across a variety inflammatory biomarkers, including lymphocyte production, neutrophil activity, and some inflammatory cytokine levels—e.g., interleukin 6, interleukin 1β, and tumor necrosis factor α. However, neither study assessing CRP in this review evidenced a significant association (36–37).

Second, it is possible that increased inflammation only occurs in people for whom bereavement is particularly stressful, such that our primary analyses missed meaningful variation in inflammatory response. Responses to bereavement are highly heterogeneous and many people are resilient to this stressor (38). Indeed, prior research has suggested that it is the people most affected by their bereavement, such as those with complicated grief (10), or who use maladaptive coping strategies, such as emotional suppression (39), that evidence poorer health after bereavement. Although we tested for moderation by depressive symptoms, the ELSA study did not include measures of complicated grief or coping strategies following bereavement. Future research should test whether psychological responses specific to the death of a spouse moderate the association between widowhood and inflammation.

Third, it is possible that the timing of the ELSA waves impacted the study results. For example, the length of time between waves could have resulted in survivorship bias. Prior work has found that the health effects of widowhood, including mortality (6), are most evident in the 6 months following the death of a spouse and dissipate over time (6, 16–17). Because ELSA assessments of inflammation were four years apart, our analysis may have missed the period during which bereaved adults evidenced higher inflammation, or excluded participants who died in the period between bereavement and the next ELSA wave. This possibility is supported by our exploratory results showing that people who became widowed but did not survive to the next assessment had higher pre-widowhood inflammation (and presumably poorer health) compared to those who survived and were included in our primary analyses. It was also not known when in that four-year period the death of a spouse occurred, which could obfuscate important differences in the time course of inflammation levels following widowhood.

Our study is not without limitations. Other measures of inflammation may have been more sensitive to widowhood, but such measures (e.g., leukocyte counts, ESR, IL1β, IL6, IL8, or TNFα) were not available across all ELSA waves used in this study; thus, we could not examine within-person changes in these measures. We also did not have a measure of complicated grief, and the timing of the ELSA waves and widowhood (relative to the assessment periods) may have limited our ability to detect significant effects or heterogeneity across time (e.g., acute versus longer-term effects). In addition, although propensity scoring algorithms are helpful in identifying useful comparison groups (31), causality cannot be established when comparing observational data from such groups. Finally, although we adjusted our models for a variety of relevant covariates, it is not possible to control for all possible alternative explanations. For example, we could not adjust for differences in diet, since this factor was not measured consistently across waves in the ELSA dataset. In addition, some covariates were measured using self-report (e.g., chronic disease count, prescribed medications) that might be more accurate if reported by a medical provider.

4.1. Conclusion

In contrast to our expectations, the current study found prospective decreases in systemic inflammation in a sample of bereaved adults, relative to their continuously married counterparts. Although preliminary, these results suggest that widowed adults who survive past the initial period of risk following bereavement may have decreased systemic inflammation in the years following their spouse’s death. Our findings highlight the importance of examining the association between bereavement and inflammation over longer time frames—as well as accounting for health both pre-and post-widowhood—to determine the time course of inflammatory response to the loss of a spouse.

Figure 1.

Mean CRP levels in the sample by widowed status. CRP values have been log transformed.

* = p < .05

Highlights.

• Increased systemic inflammation may explain risk for poor health after bereavement

• Few studies examine inflammation before and years after widowhood

• Contrary to expectations, systemic inflammation decreased following bereavement

• This decrease remained when comparing to people whose spouse remained alive

• More studies examining inflammation after widowhood over longer periods are needed