Transitions to Family Caregiving and Latent Variables of Systemic Inflammation Over Time

Abstract

Circulating levels of inflammatory biomarkers may be influenced by chronic psychological stressors such as those experienced by family caregivers. However, previous studies have found mostly small and inconsistent differences between caregivers and control samples on individual measures of systemic inflammation. Latent variables of inflammation were extracted from six biomarkers collected from two blood samples over 9 years apart for 502 participants in a national cohort study. One-half of these participants transitioned into a sustained family caregiving role between the blood samples. Two latent factors, termed “up-regulation” and “inhibitory feedback,” were identified, and the transition to family caregiving was associated with a lower increase over time on the inhibitory feedback factor indexed by interleukin (IL)-2 and IL-10. No caregiving effect was found on the up-regulation factor indexed primarily by IL-6 and C-reactive protein. These findings illustrate the advantages of using latent variable models to study inflammation in response to caregiving stress.

Keywords

family caregiving inflammatory biomarkers inflammation inhibitory feedback latent variables

Introduction

Chronic stress has been implicated as a risk factor for many medical conditions, possibly through compromised immune system functioning and increased systemic inflammation (Hänsel et al., 2010; Kiecolt-Glaser et al., 2003; Liu et al., 2017; Wirtz & von Känel, 2017). Inflammation may be a common pathway linking psychosocial influences to many chronic health problems (Friedman & Shorey, 2019), and biomarkers of inflammation have been linked to loneliness (Nersesian et al., 2018), depression (Smith et al., 2018), and other forms of psychosocial stress (Kiecolt-Glaser et al., 2010). Through the activation of the hypothalamic–pituitary–adrenal axis and subsequent up-regulation of glucocorticoids and catecholamines, chronic stress may affect both pro-inflammatory and anti-inflammatory pathways (Elenkov & Chrousos, 1999; Tian et al., 2014). Job stress, childhood adversity, and family caregiving have all been studied as naturally occurring, chronic stressors that might affect inflammation and subsequent health outcomes (Hänsel et al., 2010; Liu et al., 2017).

Several previous studies have found elevations in inflammatory biomarker levels for family caregivers compared to non-caregiving comparison groups (e.g., Kiecolt-Glaser et al., 2003; Lutgendorf et al., 1999; von Känel et al., 2012). However, recent systematic reviews of these and other studies have characterized the findings as being mixed and inconsistent (Allen et al., 2017; Potier et al., 2018). A recent meta-analysis of 44 distinct effects from 20 published studies found the overall caregiver versus control effect size to be quite small (d = 0.14) and of questionable clinical significance (Roth et al., 2019). The results were also remarkably inconsistent across the studies. Interleukin (IL)-6, for example, is a well-studied pro-inflammatory cytokine that was found in an early study to be significantly elevated in 18 family caregivers compared to 15 non-caregiving controls (Lutgendorf et al., 1999). However, at least eight subsequent studies have found no significant differences between caregivers and non-caregiving controls on this specific inflammatory biomarker (Roth et al., 2019).

Studies of associations between family caregiving and biomarkers of inflammation are often hampered by important limitations. First, the large majority of these studies have involved comparisons of relatively small convenience samples. The few larger, population-based studies typically do not find significant differences between caregivers and non-caregiving controls on inflammatory biomarker measures (Kang & Marks, 2014; Kim & Ferraro, 2014). Second, most previous studies of caregiving–inflammation associations have been cross-sectional comparisons of persons who have already been caregivers for many months or years compared to variably-assembled non-caregiving comparison groups. Although some analyses of longitudinal changes in inflammation over time for caregivers have been reported (e.g., Kiecolt-Glaser et al., 2003; von Känel et al., 2012), these studies did not collect measures of inflammation from the caregivers prior to their transition into the caregiving role. Third, the psychometric properties of many observed measures of circulating inflammatory biomarkers can be problematic. The frequency distributions are often highly skewed and may contain outliers. Even after transformations and outlier removal, substantial random measurement error may persist in many measures.

Factor analytic methods (Brinkley et al., 2012; Graham-Engeland et al., 2018) and latent variable modeling methods (Bandeen-Roche et al., 2009) have been applied previously to multiple correlated measures of inflammation in efforts to extract composite measures with better psychometric properties and to better understand inflammatory regulation. However, these measurement modeling methods have not been previously applied, to our knowledge, in research on biomarker associations with family caregiving or other forms of chronic, psychological stress. Applying such methods to multiple inflammatory biomarker measures collected from participants both before and after they took on significant family caregiving responsibilities should help control measure-specific random measurement error and further clarify the impact that caregiving stress might have on systemic inflammation.

The Caregiving Transitions Study (CTS) is a national population-based study of persons who have transitioned into a family caregiving role while participating in another large longitudinal national cohort study, the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study (Roth, Haley, Rhodes, et al., 2020). All participants in REGARDS were assessed for caregiving status at enrollment and had two blood samples taken an average of 9.3 years apart. Participants in the REGARDS study who transitioned from being non-caregivers at the first blood sample to family caregivers at some point prior to the second blood sample were potentially eligible to be enrolled in the CTS. For each enrolled caregiver who met additional eligibility criteria, an individually matched non-caregiving control participant was also enrolled in the CTS. Biomarkers extracted from the blood samples include high density C-reactive protein (CRP), tumor necrosis factor alpha receptor 1 (TNFR1), D-dimer, IL-2, IL-6, and IL-10.

Previous analyses of CTS data indicate that the transition to family caregiving was associated with significant increases in depressive symptoms, and perceived stress, and greater age-related declines in health-related quality of life compared to the longitudinal changes of the non-caregiving controls (Haley et al., 2020). For the biomarker data, previous analyses of individual biomarker measures indicated that both caregivers and controls showed age-related increases over the 9.3-year period on all of these measures except for CRP (Roth, Haley, Sheehan, et al., 2020). The caregivers showed a small but significantly greater increase over time on TNFR1 compared to the controls when the analysis was restricted to matched pairs with complete data only, but that difference was only 0.14 standard deviation units (d). Caregivers showed smaller increases over time than controls on IL-2, but that effect only approached statistical significance, and caregivers and controls did not otherwise differ significantly on any of the other four individual inflammatory biomarkers studied.

In this manuscript, we report analyses of latent variables of inflammatory biomarker change from the CTS data that address the psychometric and measurement error limitations outlined above. That is, we re-examine the CTS inflammation data by applying a latent variable measurement approach to the multiple, inter-correlated biomarkers of inflammation collected. After identifying a satisfactory latent variable measurement model, we tested measurement invariance, and then we examined the effects of the transition to caregiving on changes on those latent variables over time after adjusting for key demographic variables (age, sex, race, and body mass index). We hypothesized that the latent inflammation variables would be sensitive to changes over time in caregiving status such that those who transitioned into family caregiving roles between blood samples would show greater changes on the latent variables than the matched non-caregiving controls.

Methods

Overview of Study Design

The REGARDS study is a longitudinal cohort study that enrolled 30,239 participants in 2003–2007 from across the continental United States who were 45 or more years of age at the time of enrollment. By design, only participants with a self-reported race of African American or White were enrolled. Persons with a previous diagnosis of cancer requiring chemotherapy/radiation and those on a waiting list for or already residing in a nursing home were excluded. The REGARDS study included a computer-assisted telephone interview (CATI) at enrollment and subsequent follow-up telephone interviews every 6 months to assess possible stroke events and other major changes in health status. Two in-home assessments were also conducted that were separated by an average of 9.3 years. Blood samples were collected during both in-home visits. Body mass index (BMI) was assessed, and current prescription medication use was also collected during both in-home assessments.

The CTS is an observational study nested within the larger REGARDS study that enrolled REGARDS participants who became family caregivers at some point between the first and second REGARDS in-home assessments. As described in more detail, caregivers were those who indicated that they were “currently providing care on an on-going basis to a family member, friend, or neighbor with a chronic illness or a disability.” For each enrolled caregiver, an individually matched non-caregiving control participant was also enrolled in the CTS. Biomarker assessments were performed on the frozen blood samples from the REGARDS in-home assessments for the CTS participants.

Full descriptions of the design, participant eligibility criteria, enrollment procedures, and assessments have been published previously for both the REGARDS study (Howard et al., 2005) and the CTS (Roth, Haley, Rhodes, et al., 2020; Haley et al., 2020). Both studies were approved by the Institutional Review Board of the University of Alabama at Birmingham. Written informed consent was obtained prior to the first blood sample.

Participants

The present analyses are of data that were collected from 251 persons who became family caregivers between the two REGARDS in-home assessments and 251 matched controls who reported no caregiving responsibilities throughout their participation in the REGARDS study. As described in detail previously (Roth, Haley, Rhodes, et al., 2020), the controls were matched individually with the caregivers on seven factors: age (± 5 years), sex, race, education level, marital status, self-rated health, and self-reported history of serious cardiovascular disease from the REGARDS baseline CATI. Overall, the CTS sample was 65% female, 36% African American, and 76% married or cohabiting with a partner.

Procedure

As part of the REGARDS baseline CATI, each participant was asked about their demographic information, health history, and other health risk factors. This included a question about caregiving status. Specifically, all REGARDS participants were asked: “Are you currently providing care on an on-going basis to a family member with a chronic illness or disability? This includes any kind of help such as watching your family member, dressing or bathing this person, arranging care, or providing transportation?” Those who answered “no” were designated as non-caregivers at REGARDS baseline. In a later caregiving screening CATI conducted an average of 11.8 years after the baseline CATI, updated caregiving status information was collected. Specifically, during this caregiving screening CATI, participants were asked: “Are you currently providing care on an on-going basis to a family member, friend, or neighbor with a chronic illness or a disability? This would include any kind of regular help with basic activities such as dressing, bathing, grooming this person, managing bills, arranging for medical care, watching or supervising this person, or providing transportation.” Participants who answered “yes” to this question from the caregiving screening CATI and “no” to the similar question from the REGARDS baseline CATI were further screened for eligibility to be enrolled as incident caregivers in the CTS. Those who answered “no” to the caregiving status questions at both interviews were further screened to serve as matched non-caregiving controls.

Participants who transitioned into a caregiving role prior to the caregiving screening CATI were further asked (1) their relationship with the person receiving that care; (2) when they began providing care because of that person’s disability of health problem (approximate month and year); (3) whether that person has “Alzheimer’s disease, another form of dementia, or serious memory problems”; (4) how many hours of care they provided per week because of that person’s disability or health problem; and (5) how much of a mental or emotional strain it was to provide that care (no strain, some strain, or a lot of strain). Only caregivers who reported a month and year when they began providing care that was at least 3 months before they completed the second REGARDS in-home assessment were eligible to be enrolled in the CTS. Caregivers also had to report providing at least 5 hours of care per week in order to be eligible. Subsequent in-depth interviews confirmed that all care recipients needed assistance with at least one activity of daily living (ADL) or an instrumental activity of daily living (IADL) and that most caregivers provided substantial ADL and IADL care (Haley et al., 2020; Sheehan et al., 2021).

For controls, in addition to matching with a caregiver on the seven factors listed above, each enrolled control participant reported no significant family caregiving responsibilities throughout his or her period of participation in the REGARDS study. In addition, non-caregiving controls who were matched to spouse caregivers had to be married, and non-caregiving controls who were matched to an adult child caregiver had to have at least one living parent.

Biomarker Assay Methods

Fasting morning blood samples were collected by trained phlebotomists from REGARDS participants at their homes during both of the REGARDS in-home assessments. The time lag between the 1^st and 2^nd REGARDS in-home assessments ranged from 7.6 years to 12.4 years (M = 9.3 years) for the participants included in the present analyses. Blood samples were centrifuged and shipped overnight on ice to the Laboratory for Clinical Biochemistry Research at the University of Vermont, where they were re-centrifuged and stored at −80°C. Standardized collection, shipping, and processing methods were used and have been described in more detail elsewhere (Gillett et al., 2014; Howard et al., 2005).

Biomarker Measures of Systemic Inflammation

Six circulating biomarkers of inflammation (CRP, D-dimer, TNFR1, IL-2, IL-6, and IL-10) were assayed for the present analyses. These biomarkers were selected based on their use in previous studies of inflammation in caregivers (Allen et al., 2017; Potier et al., 2018; Roth et al., 2019) and on whether valid measures could be obtained from frozen blood samples that were several years old. All six biomarkers assess systemic inflammation in the human body (Hänsel et al., 2010). CRP and IL-6 are classic measures of systemic inflammation, and are the most commonly studied measures of inflammation in the caregiving literature (Roth et al., 2019). TNFR1 is a more recent addition to chronic inflammation measurement, and has often been the inflammatory biomarker most associated with adverse health outcomes in older adults (Gross et al., 2019; Varadhan et al., 2014). D-dimer is generated from clotting-related processes and has been utilized as a surrogate marker of inflammation in many population studies of older adults (Cohen et al., 2003; Walston et al., 2002). IL-2 and IL-10 are thought to represent feedback mechanisms or anti-inflammatory influences on immune system regulation (Banchereau et al., 2012; Boyman & Sprent, 2012; Couper et al., 2008).

High sensitivity C-reactive protein (CRP) was measured using a BNII nephelometer (high sensitivity CRP; Dade Behring Inc.). The inter-assay coefficients of variation (CVs) were 3–6% with a detection level of 0.16 μg/mL. D-dimer was assessed using an immunoturbidimetric assay (Liatest D-DI; Diagnostica Stago, catalog# 00,515) on a Sta-R analyzer (Diagnostica Stago). The lower limit of detection range of the assay was 0.01–20 μg/mL, and inter-assay CVs were 1.48%. Tumor necrosis factor alpha receptor 1 (TNFR1) was measured using an R&D systems Elisa assay (catalog # DRT100). Detectable range 78–5000 pg/mL with an inter-assay CV of 3.4%. IL-2, IL-10, and IL-6 were measured with a Meso Scale Discovery (MSD) Pro-inflammatory panel, catalog # K15049G. MSD assays were read using a MESO QuickPlex SQ 120. Detectable ranges are as follows: IL-2: 0.07–2860 pg/mL, inter-assay CV: 18.12%; IL-10: 0.02–674 pg/mL, inter-assay CV: 10.6%, IL-6: 0.05–1500 pg/mL, and inter-assay CV: 5.2%. For IL-2, 25% of levels from the first assessment and 16% of the levels from the second assessment were below the detectable range, and 0.034 was inserted as the value for these observations, which represents the midpoint between 0 and the lowest detectable score of 0.068.

Statistical Analyses

Frequency distributions were examined for all biomarkers from both blood samples and were observed to be highly positively skewed. Consistent with previous analyses of these data and similar biomarker data from the Cardiovascular Health Study (Jenny et al., 2012), a log(base 2) transformation was applied to each biomarker at each assessment. After the log transformation, each biomarker was further examined for possible outliers using Tukey’s (1977) “outer fences” method that is based on the interquartile range (IQR). Specifically, all values that were more than 3*IQR above the 75^th percentile or more than 3*IQR below the 25^th percentile were designated as extreme outliers and recoded as missing in the primary analyses. No outliers were detected after log transformation using this method for CRP, TNFR1, or D-dimer. For IL-2, IL-6, and IL-10, values above 0.95, 8.23, and 2.22 pg/mL, respectively, were identified as extreme positive outliers. Collectively, 28 of the 2919 values (0.96%) across the IL-2, IL-6, and IL-10 assessments were identified as outliers.

Latent variables were extracted from the log-transformed biomarker measures and structural equation models were estimated using Mplus 8.4 (Muthén & Muthén, 1998-2017). All Mplus analyses used full information maximum likelihood estimation to reproduce the observed variances and covariances of the log-transformed biomarker measures as well as possible given the constraints of the model. In addition, a robust variance correction was used to account for the caregiver-control matched sampling design. Specifically, the TYPE=COMPLEX option, MLR estimator, and the CLUSTER option in Mplus were used to compute standard errors that took into account the non-independence of the observations due to matching.

The analyses proceeded sequentially through the following steps. First, consistent with previous findings (Bandeen-Roche et al., 2009), we examined the fit of a 2-factor model for biomarkers from both blood samples separately with index items (factor loadings fixed to 1.0 on one factor and 0 on the other factor) for IL-6 on the first latent variable and for IL-10 on the second latent variable (Figure 1). Non-significant factor loadings were trimmed (i.e., fixed to zero) in subsequent measurement models to achieve greater parsimony. Second, the evaluation of the 2-factor model was extended to include both waves of data simultaneously. Correlated residuals between the first and second blood samples were specified for all six of the biomarker measures in these analyses.

Figure 1.

Original unconstrained 2-factor measurement model.

Third, because the ultimate goal of these analyses was to compare groups (i.e., caregivers vs. controls as well as groups defined by key demographic variables) on changes in the latent constructs over time, measurement invariance analyses were conducted to ensure that the meaning of the constructs was consistent across comparison groups and across time. Such comparisons on latent variables can be compromised by non-invariance unless appropriate steps are taken (Brown, 2015; Putnick & Bornstein, 2016). The measurement invariance testing procedures and results are presented in detail in supplemental materials 1.

After the measurement modeling and invariance testing were completed, a final measurement model with partial invariance was adopted. Unstandardized factor loadings for the biomarkers on their respective latent factors were constrained to be equal over time. Longitudinal structural or causal models were then constructed and tested in the context of this final measurement model. The first structural model examined whether the transition to family caregiving that occurred between the first and second in-home assessments affected either latent variable at the second assessment more than what was observed for the non-caregiving controls. The second model included caregivers only and examined whether caregiving relationship (spouse vs. non-spouse), intensity (number of hours of care per week), duration (in years prior to the second blood sample), type (dementia vs. non-dementia caregiving), or caregiving strain (none, some, or a lot) affected the latent biomarker measures. For both models, the respective latent variables from the first assessment served as a covariate of the latent variable from the second assessment along with participant race, sex, age, and BMI. Estimates in standard deviation units (i.e., Y-standardization or STDY estimates in Mplus) were tested for statistical significance and interpreted accordingly. Overall model fit was examined using the root mean square error of approximation (RMSEA), the comparative fit index (CFI), and the standardized root mean square residual (SRMR) fit statistics.

Several sensitivity analyses were conducted to examine whether analytic modifications would have any meaningful impact on the results. These sensitivity analyses included (1) recoding Tukey outliers to the respective outlier cut point scores instead of recoding those scores to missing, (2) removing the temporal constraints on the equality of the unstandardized factor loadings over time, and (3) adding observed indicators for use of statin medications and antidepressants at each in-home assessment as additional covariates in the structural models. The substantive findings did not change across any of these sensitivity analyses, so only the primary analyses with outliers recoded as missing; constrained factor loadings over time; and participant race, sex, age, and BMI as the only covariates are further presented here. More information on the results of the sensitivity analyses is available from the authors upon request.

Results

Demographic and other descriptive information for the participants in these analyses are provided in Table 1. By design, caregivers and controls were individually matched on sex and race and closely comparable on education, age at both REGARDS in-home assessment, and self-rated health from the REGARDS enrollment interview. Caregivers and controls did not differ on the time interval between the two in-home assessments. Participants who transitioned into caregiving were found to have slightly but significantly higher BMI scores at both assessments (ps < .05).

Table 1.

Descriptive Information for Incident Caregivers and Matched Non-Caregiving Controls.

Variable	Caregivers (N = 251)	Controls (N = 251)
Sex, female N (%)	163 (65)	163 (65)
Race, African American N (%)	90 (36)	90 (36)
Education, N (%)
Less than high school graduate	9 (4)	10 (4)
High school graduate	58 (23)	58 (23)
Some college	78 (31)	69 (27)
College graduate and above	106 (42)	114 (45)
Annual household income, N (%)
< $20,000	19 (8)	19 (8)
$20,000–$34,000	49 (20)	48 (19)
$35,000–$74,000	105 (42)	97 (39)
> $75,000	57 (23)	68 (27)
Refused	21 (8)	19 (8)
Self-rated health at 1st REGARDS interview
Excellent	61 (24)	54 (22)
Very good	81 (32)	88 (35)
Good	84 (33)	88 (35)
Fair	22 (9)	21 (8)
Poor	3 (1)	0 (0)
Age at 1st in-home assessment, M (SD)	60.2 (7.7)	59.8 (7.3)
Age at 2nd in-home assessment, M (SD)	69.5 (7.9)	69.1 (7.5)
BMI at 1st in-home assessment, M (SD)	30.5 (6.4)	28.6 (6.2)
BMI at 2nd in-home assessment, M (SD)	30.4 (6.5)	29.0 (6.5)
Years elapsed between 1st and 2nd REGARDS in-home assessments, M (SD)	9.3 (0.9)	9.2 (0.8)
Duration of caregiving in years before 2nd in-home assessment, M (SD)	3.4 (2.4)	---
Hours of care per week, M (SD)	35.7 (27.8)	---
Caregiving relationship, spouse N (%)	128 (51)	---
Caring for a person with dementia, N (%)	117 (47)	---
Caregiving strain, N (%)
None	59 (24)	---
Some	135 (54)	---
A lot	56 (22)	---

For the caregivers, they averaged 3.4 (SD = 2.4) years of caregiving prior to the second REGARDS in-home assessment and 35.7 (SD = 27.8) hours of care per week. Just over half (n = 128, 51%) were caring for a spouse or co-residing partner, and 117 (47%) reported caring for a person with Alzheimer’s disease or another form of dementia.

Descriptive statistics for and Pearson product–moment correlations among the log-transformed biomarker measurements from both blood samples are available in supplemental materials 2. Test–retest correlations across the two assessments ranged from 0.80 for TNFR1 to 0.41 for IL-2. Data were missing on all six biomarkers for 19 participants at baseline and two participants at the second assessment. No participants were missing data on all six biomarkers at both time points. The useable sample sizes for model testing, therefore, at baseline, the second assessment, and both time points were n = 483, n = 500, and n = 502, respectively.

Latent Variable Measurement Models

The originally proposed measurement model (Figure 1) fit very well for the six biomarkers at baseline (χ² = 6.42, df = 4, p = 0.170, RMSEA = 0.035, CFI = 0.990, SRMR = 0.015), but had three standardized loadings that were small and not statistically significant (i.e., f1 → IL-2, f2 → CRP, and f2 → D-dimer). These non-significant loadings were dropped, and the model was re-estimated, resulting in a more parsimonious model with excellent fit (χ² = 7.42, df = 7, p = 0.387, RMSEA = 0.011, CFI = 0.998, SRMR = 0.019). The same measurement model was also fit to the biomarker data at the second assessment and was also found to have excellent fit (χ² = 12.15, df = 7, p = 0.096, RMSEA = 0.038, CFI = 0.980, SRMR = 0.023).

The first latent variable (f1) was labeled “up regulation” based on the same label applied previously to a latent variable with similar indicators (Bandeen-Roche et al., 2009). This latent variable had strong standardized factor loadings for IL-6 and CRP, with more moderate loadings for D-dimer and TNFR1. The second latent variable (f2) was labeled “inhibitory feedback” because of its sizable standardized factor loadings for the two inhibitory or anti-inflammatory biomarkers, IL-2 and IL-10. The two latent factors were modestly correlated with each at both assessment waves.

Given the strong fit and the consistency of this measurement model across time, it formed the basis for subsequent invariance testing. As described in more detail in supplemental materials 1, we examined both metric and scalar invariance across key demographic variables (sex, race, age, BMI, and caregiving status) and across time (blood sample 1 vs. blood sample 2). Significant scalar non-invariance was identified for sex, race, and age for TNFR1, which was then accounted for by adjusting for these demographic effects on that specific indicator in modified or partial measurement invariance models. The small split-loadings for TNFR1 on the inhibitory feedback latent variable were further reduced to insignificant levels after taking this invariance into account, and these loadings were, therefore, fixed to zero in the final partial invariance measurement model.

The standardized factor loadings for a combined measurement model of the biomarker data from both blood samples are provided in Figure 2. The final partial invariance measurement model that included the predictors for TNFR1 and constrained the unstandardized factor loadings for each indicator on its underlying latent factor to be equal over time was found to provide excellent fit to the observed data (χ² = 172.37, df = 78, p < 0.001, RMSEA = 0.049, CFI = 0.950, and SRMR = 0.077). It was subsequently adopted as the measurement model to be used in the structural models.

Figure 2.

Completely standardized (STDYX) factor loading estimates for the biomarker measurement model at the first/second assessments. P < 0.001 for all loadings.

Structural Model of Caregiving Transition Effects on Latent Inflammation Variables

The results from the structural model that tested the effects of the transition to family caregiving on changes in the latent variables of inflammation over time are summarized in Figure 3. Sex, race, age, and BMI were specified as covariates of all four latent factors, and, as mentioned above, sex, race, and age were specified as predictors of TNFR1 at both waves to account for the scalar non-invariance for that biomarker. The model depicted in Figure 3 provided excellent fit to the observed data (χ² = 213.98, df = 108, p < 0.001, RMSEA = 0.044, CFI = 0.950, SRMR = 0.079). Because the latent variables from the first assessment (f1 and f2) were predictors of those same latent variables at the second assessment (labeled as f3 and f4, respectively), additional predictive effects on f3 or f4 represent predictions of changes on those latent constructs over the 9.3 year period.

Figure 3.

Final structural model. Completely standardized (STDYX) estimates are shown except for caregiver versus control paths, which are Y-standardized (STDY). Sex, race, age, and BMI are also predictors of all four latent factors, and sex, race, and age are predictors of TNFR1 at both waves, but these paths are not illustrated. Unstandardized factor loadings were constrained to be equal across waves. Only one correlated residual is illustrated, but all six were estimated.

For the “up-regulation” latent variable (f3), BMI was a significant predictor (b = 0.047; SE = 0.009; 95% confidence interval (CI): 0.030, 0.064; p < 0.001) and the effect for age approached statistical significance (b = 0.012; SE = 0.007; 95% CI: −0.001, 0.025; p = 0.067). The results for caregiving status indicate that caregivers and controls did not differ significantly on changes to the “up-regulation” latent variable over time (b = 0.084, SE = 0.078, 95% CI: −0.068, 0.237; p = 0.279). For the “inhibitory feedback” latent variable (f4), significant predictive effects were found for participant age (b = 0.027, SE = 0.010; 95% CI: 0.007, 0.047; p = 0.008), race (b = −0.340, SE = 0.151, 95% CI: −0.636, −0.043; p = 0.025), and caregiving status (b = −0.277; SE = 0.132; 95% CI: −0.536, −0.018; p = 0.036). Thus, after adjusting for the other covariates, inhibitory feedback increased by about 0.03 SDs for each year of age, decreased by 0.34 SDs for African Americans compared to Whites, and decreased by 0.28 SDs for those who transitioned to family caregiving between assessments compared to non-caregiving controls.

No additional statistically significant effects were found in the structural equation model for the caregivers only. That is, among the incident caregivers only, no effects were found on the longitudinal changes for the up-regulation or inhibitory feedback latent variables for caregiving relationship, hours of care per week, duration of care, dementia versus non-dementia caregiving, or caregiving strain (all ps > 0.10).

Discussion

The present paper builds on previous work that has examined inflammatory biomarker changes as a function of the transition to family caregiving. We applied a latent variable measurement model and found that the covariances among six circulating inflammatory biomarker measures could be well-accounted for by two underlying latent variables. Consistent with previous work (Bandeen-Roche et al., 2009), we identified an “up-regulation” latent variable that was primarily identified by strong factor loadings for IL-6 and CRP, with more modest factor loadings for D-dimer and TNFR1. Our two-factor model also included an “inhibitory feedback” latent variable that was identified by significant loadings for IL-2 and IL-10.

The transition to family caregiving was not found to be associated with longitudinal changes on the up-regulation latent variable compared to changes observed for the non-caregiving controls. This null finding for the caregiving transition on the up-regulation latent variable is consistent with the frequently reported lack of differences when caregivers and non-caregivers have been previously compared on IL-6 and CRP in cross-sectional analyses (Roth et al., 2019). Conversely, differential changes were found between those who transitioned to caregiving and non-caregiving controls on the inhibitory feedback latent variable that was indexed by IL-2 and IL-10. Significant predictive effects on this latent variable were also found for age and race. Longitudinal increases on this factor were reduced for persons who transitioned into a family caregiving role compared to non-caregiving controls, for younger compared to older participants, and for African American participants compared to White participants.

The potential health implications of these findings for the caregiving transition on the inhibitory feedback latent variable are intriguing and deserve further investigation. Both IL-2 and IL-10 have anti-inflammatory properties (Banchereau et al., 2012). IL-2 is essential for the development and the survival of immunosuppressive Foxp3⁺ regulatory T-cells (Chinen et al., 2016), and IL-10 is a cytokine that inhibits the activity of Th1 cells, macrophages, and dendritic cells, all of which are needed for appropriate immune responses (Couper et al., 2008). The suppression of an inhibitory feedback mechanism in persons who transitioned into a family caregiving role may suggest that the stress of caregiving led to a chronic low-grade inflammation, and this dampened anti-inflammatory effect might be a mechanism by which caregivers are at greater risk for some inflammation-induced chronic diseases, such as incident cardiovascular disease (Capistrant et al., 2012; Mortensen et al., 2018; Wirtz & von Känel, 2017). Conversely, it is noteworthy that the participants who transitioned into family caregiving did not show the typical age-related increases on this inhibitory feedback latent variable, and that African Americans also had lower increases over time on this latent variable. Both caregivers (Mehri et al., 2021; Roth et al., 2015) and older African Americans (Roth et al., 2016; Wing et al., 1985; Yao & Robert, 2011) show lower mortality rates and enhanced survival compared to their respective comparison groups. Perhaps this dampened inhibitory feedback mechanism provides a common pathway to explain these somewhat disparate and paradoxical mortality findings.

Although the latent variable analyses did uncover a potentially interesting effect for the transition to caregiving on an inhibitory feedback mechanism for inflammation, analyses of caregiving-specific predictors did not reveal any additional significant effects. That is, neither caregiving intensity (e.g., hours of care, dementia caregiving, and strain) nor caregiving duration were found to significantly predict individual differences in changes on either latent variable in our analyses. We also did not detect any differences between spouse and non-spouse caregivers. Previous analyses of CTS data have demonstrated expected differences between these caregiving subtypes on self-reported measures of well-being and caregiving burden (Haley et al., 2020; Sheehan et al., 2021), but those self-reported differences in stress and well-being did not further translate into significant differences in inflammation in the present analyses.

Previous analyses of these biomarker data from the CTS using individual linear regression models for each biomarker separately found small and usually marginal (ps < 0.10) covariate-adjusted effects for the caregiving transition on TNFR1 and IL-2, with caregivers showing greater increases on TNFR1 but lesser increases on IL-2 over time compared to the non-caregiving controls (Roth, Haley, Sheehan, et al., 2020). The small caregiving effect observed previously for TNFR1 was not extended to a difference on the up-regulation latent variable in the present analyses. However, the transition to family caregiving was associated with a reduced age-related increase on the inhibitory feedback latent variable in the present analyses. Descriptive data from our previous paper showed that the caregivers had flat longitudinal trajectories and showed very little change across the 9.3-year interval on IL-2 and IL-10, whereas the non-caregiving controls showed moderate age-related increases over time on both biomarkers (Roth, Haley, Sheehan, et al., 2020). The caregiving effect on the inhibitory feedback factor in the present analyses, therefore, is consistent with these differential longitudinal trajectories and with the marginal effect for caregiving on IL-2 that were previously reported.

The standardized effect size of 0.28 standard deviation units between caregivers and controls on the change observed for the inhibitory latent variable is generally larger than caregiving effects that have been previously reported for caregiving on individual inflammatory biomarkers from earlier studies (Roth et al., 2019) and from the earlier regression analyses on these individual biomarker measures from the CTS (Roth, Haley, Sheehan, et al., 2020). The improved reliability of latent variables in comparison to the reliabilities for individual observed variables might account for some of these effect size differences. In general, the present analyses illustrate the potential advantages of using latent variable approaches to remove measure-specific random measurement error and increase effect size estimates when analyzing effects on multiple, correlated physiological variables.

Future studies will be important to further examine biomarker differences and investigate possible mechanisms and health implications. For example, Kim and Yoon (2020) found that sleep quality can be an important compensatory pathway that protects against inflammation, and sleep quality can also be impacted by caregiving stress (Gao et al., 2019). Social factors such as marital quality (Wong & Shobo, 2017) and other roles that might either elicit stress or provide resources, such as parenting and employment (Barnett, 2015) are also potentially involved in affecting the health trajectories of caregiving. Future studies including such contextual factors alongside our latent variable approaches for assessing inflammation should be further illuminating.

The present study has many strengths including the national population-based sample of caregivers, the careful confirmation of sufficient and sustained caregiving activities for the caregivers, and the case-by-case matching methods used to select control participants (Roth, Haley, Rhodes, et al., 2020). The CTS appears to be unique in being the only prospective study of caregiving that has biomarker data from both before and after participants transitioned into the family caregiving role. There are limitations to the study as well. The CTS has carefully matched incident caregiver and non-caregiving control samples, but some remaining differences between caregivers and controls undoubtedly exist. While we were able to control for a few measured differences, including BMI and certain medication classes, influences for other, unmeasured potential confounders are certainly possible.

Another limitation is the analysis of only six biomarkers of inflammation, and different latent variable solutions might have been found if more or different inflammatory biomarkers were analyzed. The inhibitory feedback latent variable was indexed by only two observed biomarkers (IL-2 and IL-10) and is essentially based on the rather modest correlation of 0.28 between those two biomarkers at both assessments. Future research is necessary to confirm whether an inhibitory mechanism is responsible for this relatively modest correlation between these two biomarkers. Additional anti-inflammatory biomarkers have been identified (Banchereau et al., 2012) and might be included in more comprehensive measurement models in future studies.

In conclusion, the investigation of systemic inflammation and other physical health effects associated with the long-term exposure to psychological stress continues to be a complex and multifactorial endeavor. Innovations in measurement modeling for multiple correlated physiological indicators offer much potential for gaining new insights in this important area. We believe these methods should continue to be used in relatively large, population-based samples of family caregivers and matched samples of non-caregiving controls. Our findings suggest that caregiving is not associated with significant increases in the classic inflammatory response as captured by such indicators as IL-6 or CRP, but extended caregiving may dampen an inhibitory response mechanism indexed by circulating IL-2 and IL-10 levels. With the rapid increase in the number of family caregivers that is occurring around the world, it will be important to continue to advance sound research with appropriate methodological sophistication and rigor so that we can better understand and anticipate needs and challenges that confront older adults with disabilities and the family members who care for them.

Supplemental Material

sj-pdf-1-roa-10.1177_01640275221084729 - Supplemental Material for Transitions to Family Caregiving and Latent Variables of Systemic Inflammation Over Time

Supplemental Material, sj-pdf-1-roa-10.1177_01640275221084729 for Transitions to Family Caregiving and Latent Variables of Systemic Inflammation Over Time by David L. Roth, John P. Bentley, Debora Kamin Mukaz, William E. Haley, Jeremy D. Walston, and Karen Bandeen-Roche in Research on Aging

Footnotes

Acknowledgments

The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at .

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: This work was supported by a cooperative agreement [U01 NS041588] co-funded by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), National Institutes of Health, Department of Health and Human Services. The Caregiving Transitions Study was further supported by an investigator-initiated grant [RF1 AG050609] from the NIA. Additional support for the analyses reported here was provided by the Johns Hopkins University Claude D. Pepper Older Americans Independence Center funded by the NIA (P30 AG021334). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or NIA. Representatives of the NINDS were involved in the review of the manuscript but were not directly involved in the collection, management, analysis, or interpretation of the data.

ORCID iDs

David L. Roth

John P. Bentley

Supplemental Material

Supplemental material for this article is available online.

Author Biographies

David L. Roth , PhD, is a Professor and the Director of the Center on Aging and Health at Johns Hopkins University. His 30+ years of research experience in applied gerontology include interests in family caregiving and health aging. He is the 2021 recipient of the M. Powell Lawton award from the Gerontological Society of America.

John P. Bentley , PhD, is a Professor and the Director of the Center for Pharmaceutical Marketing and Management at the University of Mississippi. He is a pharmacist, biostatistician, and health services researcher. His interests include patient-reported outcomes; medication use, misuse, and outcomes; latent variable analysis; and statistical mediation and moderation.

Debora Kamin Mukaz , M.S., PhD, is a postdoctoral associate at the Laboratory for Clinical Biochemistry Research at the University of Vermont. Her research focuses on the epidemiology of cardiovascular and cardiometabolic conditions, and on the effects of COVID-19 in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study.

William E. Haley , PhD, is a Distinguished University Professor at the University of South Florida School of Aging Studies. His research focuses on the psychological, social, and health effects of family caregiving, stress process theories of caregiving, racial/ethnic diversity in caregiving, benefits of caregiving, and intervention research.

Jeremy D. Walston , M.D., is the Raymond and Anna Lublin Professor of Geriatric Medicine and Gerontology and the Director of the Claude D. Pepper Older Americans Independence Center at Johns Hopkins University. He has expertise and research interests in frailty, resilience, inflammation, and the biology of healthy aging.

Karen Bandeen-Roche , PhD, is the Hurley Dorrier Professor of Biostatistics and the Chair of the Department of Biostatistics at the Johns Hopkins Bloomberg School of Public Health. She is known for her leadership as a biostatistician and for her research on aging and aging-related frailty.

References

Allen

A. P.

Curran

E. A.

Duggan

Cryan

J. F.

Chorcorian

A. N.

Dinan

T. G.

Clarke

(2017). A systematic review of the psychobiological burden of informal caregiving for patients with dementia: Focus on cognitive and biological markers of chronic stress. Neuroscience and Biobehavioral Reviews, 73, 123-164. https://doi.org/10.1016/j.neubiorev.2016.12.006.

Banchereau

Pascual

O’Garra

(2012). From IL-2 to IL-37: the expanding spectrum of anti-inflammatory cytokines. Nature Immunology, 13(10), 925-931. https://doi.org/10.1038/ni.2406.

Bandeen-Roche

Walston

J. D.

Huang

Semba

R. D.

Ferrucci

(2009). Measuring systemic inflammatory regulation in older adults: evidence and utility. Rejuvenation Research, 12(6), 403-410. https://doi.org/10.1089/rej.2009.0883.

Barnett

A. E.

(2015). Adult child caregiver health trajectories and the impact of multiple roles over time. Research on Aging, 37(3), 227-252. https://doi.org/10.1177/0164027514527834.

Boyman

Sprent

(2012). The role of interleukin-2 during homeostasis and activation of the immune system. Nature Reviews Immunology, 12(3), 180-190. https://doi.org/10.1038/nri3156.

Brinkley

T. E.

Hsu

F. C.

Beavers

K. M.

Church

T. S.

Goodpaster

B. H.

Stafford

R. S.

Pahor

Kritchevsky

S. B.

Nicklas

B. J.

(2012). Total and abdominal adiposity are associated with inflammation in older adults using a factor analysis approach. Journal of Gerontology: Medical Sciences, 67(10), 1099-1106. https://doi.org/10.1093/gerona/gls077.

Brown

T. A.

(2015). Confirmatory factor analysis for applied research (2nd.). Guilford publications.

Capistrant

B. D.

Moon

J. R.

Berkman

L. F.

Glymour

M. M.

(2012). Current and long-term spousal caregiving and onset of cardiovascular disease. Journal of Epidemiology and Community Health, 66(10), 951-956. https://doi:10.1136/jech-2011-200040.

Chinen

Kannan

A. K.

LevineFan

A. G. X.

Klein

Zheng

Rudensky

A. Y.

(2016). An essential role for the IL-2 receptor in T_reg cell function. National Immunology, 17(11), 1322-1333. https://doi.org/10.1038/ni.3540.

10.

Cohen

H. J.

Harris

Pieper

C. F.

(2003). Coagulation and activation of inflammatory pathways in the development of functional decline and mortality in the elderly. American Journal of Medicine, 114(3), 180-187. https://doi.org/10.1016/s0002-9343(02)01484-5.

11.

Couper

K. N.

Blount

D. G.

Riley

E. M.

(2008). IL-10: the master regulator of immunity to infection. Journal of Immunology, 180(9), 5771-5777. https://doi.org/10.1038/ni.3540.

12.

Elenkov

I. J.

Chrousos

G. P.

(1999). Stress hormones, Th1/Th2 patterns, pro/anti-inflammatory cytokines and susceptibility to disease. Trends in Endocrinology and Metabolism, 10(9), 359-368. https://doi.org/10.1016/S1043-2760(99)00188-5.

13.

Friedman

Shorey

(2019). Inflammation in multimorbidity and disability: an integrative review. Health Psychology, 38(9), 791-801. http://dx/doi.org/10.1037/hea0000749.

14.

Gao

Chapagain

N. Y.

Scullin

M. K.

(2019). Sleep duration and sleep quality in caregivers of patients with dementia: a systematic review and meta-analysis. JAMA Network Open, 2(8), e199891. http://doi.org/10.1001/jamanetworkopen.2019.9891.

15.

Gillett

S. R.

Boyle

R. H.

Zakai

N. A.

McClure

L. A.

Jenny

N. S.

Cushman

(2014). Validating laboratory results in a national observational cohort study without field centers: the Reasons for geographic and racial differences in stroke cohort. Clinical Biochemistry, 47(16-17), 243-246. https://doi.org/10.1016/j.clinbiochem.2014.08.003.

16.

Graham-Engeland

J. E.

Sin

N. L.

Smyth

J. M.

Jones

D. R.

Knight

E. L.

Sliwinski

M. J.

Engeland

C. G.

(2018). Negative and positive affect as predictors of inflammation: Timing matters. Brain, Behavior, and Immunity, 74, 222–230. https://dx.doi.org/10.1016/j.bbi.2018.09.011.

17.

Gross

A. L.

Walker

K. A.

Moghekar

A. R.

Pettigrew

Soldan

Albert

M. S.

Walston

J. D.

(2019). Plasma markers of inflammation linked to clinical progression and decline during preclinical AD. Frontiers in Aging Neuroscience, 6(11), 229. https://doi.org/10.3389/fnagi.2019.00229.

18.

Haley

W. E.

Roth

D. L.

Sheehan

O. C.

Rhodes

J. D.

Huang

Blinka

M. D.

Howard

V. J.

(2020). Effects of transitions to family caregiving on well-being: a longitudinal population-based study. Journal of the American Geriatrics Society, 68(12), 2839-2846. https://doi.org/10.1111/jgs.16778.

19.

Hänsel

Hong

Camara

R. J.

von Känel

(2010). Inflammation as a psychophysiological biomarker in chronic psychosocial stress. Neuroscience and Biobehavioral Reviews, 35(1), 115-121. https://doi.org/10.1016/j.neubiorev.2009.12.012.

20.

Howard

V. J.

Cushman

Pulley

Gomez

C. R.

R. C.

Prineas

R. J.

Graham

Moy

C. S.

Howard

(2005). The Reasons for geographic and racial differences in stroke study: objectives and design. Neuroepidemiology, 25(3), 135-143. https://doi.org/10.1159/000086678.

21.

Jenny

N. S.

French

Arnold

A. M.

Strotmeyer

E. S.

Cushman

Chaves

P. H. M.

Ding

Fried

L. P.

Kritchevsky

S. B.

Rifkin

D. E.

Sarnak

M. J.

Newman

A. B.

(2012). Long-term assessment of inflammation and healthy aging in late life: the cardiovascular health study all stars. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 67(9), 970-976. http://doi.org/10.1093/gerona/glr261.

22.

Kang

Marks

N. F.

(2014). Filial caregiving is associated with greater neuroendocrine dysfunction: evidence from the 2005 national survey of midlife in the U.S. SAGE Open Medicine, 2. https://doi.org/10.1177/2050312113520152.

23.

Kiecolt-Glaser

J. K.

Gouin

J. P.

Hantsoo

(2010). Close relationships, inflammation, and health. Neuroscience and Biobehavioral Reviews, 35(1), 33-38. http://doi.org/10.1016/j.neubiorev.2009.09.003.

24.

Kiecolt-Glaser

J. K.

Preacher

K. J.

MacCallum

R. C.

Atkinson

Malarkey

W. B.

Glaser

(2003). Chronic stress and age-related increases in the proinflammatory cytokine IL-6. Proceedings of the National Academy of Sciences, 100(15), 9090-9095. http://doi.org/10.1073/pnas.1531903100.

25.

Kim

Ferraro

K. F.

(2014). Do productive activities reduce inflammation in later life? Multiple roles, frequency of activities, and C-reactive protein. The Gerontologist, 54(5), 830-839. http://doi.org/10.1093/geront/gnt090.

26.

Kim

Yoon

(2020). Volunteering, subjective sleep quality, and chronic inflammation: a 5-year follow-up of the national social life, health, and aging project. Research on Aging, 42(9-10), 291-299. https://doi.org/10.1177/0164027520922624.

27.

Liu

Y. Z.

Wang

Y. X.

Jiang

C. L.

(2017). Inflammation: the common pathway of stress-related diseases. Frontiers in Human Neuroscience, 11, 316. http://doi.org/10.3389/fnhum.2017.00316.

28.

Lutgendorf

S. K.

Garand

Buckwalter

K. C.

Reimer

T. T.

Hong

S. Y.

Lubaroff

D. M.

(1999). Life stress, mood disturbance, and elevated interleukin-6 in healthy older women. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 54(9), M434-M439. http://doi.org/10.1093/gerona/54.9.M434.

29.

Mehri

Kinney

Brown

Rajabi Rostami

(2021). Informal caregiving and all-cause mortality: a meta-analysis of longitudinal population-based studies. Journal of Applied Gerontology, 40(7), 781-791. http://doi.org/10.1177/0733464819893603.

30.

Mortensen

Dich

Lange

Ramlau-Hansen

C.H.

Head

Kivimaki

Hulvej Rod

(2018). Weekly hours of informal caregiving and paid work, and the risk of cardiovascular disease. The European Journal of Public Health, 28(4), 743-747. http://doi.org/10.1093/eurpub/ckx227.

31.

Muthén

L. K.

Muthén

B. O.

(1998-2017) Mplus. User’s guide (8). Muthén & Muthén.

32.

Nersesian

P. V.

Han

H. R.

Yenokyan

Blumenthal

R. S.

Nolan

M. T.

Hladek

M. D.

Szanton

S. L.

(2018). Loneliness in middle age and biomarkers of systemic inflammation: findings from midlife in the United States. Social Science & Medicine, 209, 174-181. http://doi.org/10.1016/j.socscimed.2018.04.007.

33.

Potier

Degryse

J. M.

de Saint-Hubert

(2018). Impact of caregiving for older people and pro-inflammatory biomarkers among caregivers: a systematic review. Aging Clinical and Experimental Research, 30(2), 119-132. https://doi.org/10.1007/s40520-017-0765-0.

34.

Putnick

D. L.

Bornstein

M. H.

(2016). Measurement invariance conventions and reporting: the state of the art and future directions for psychological research. Developmental Review, 41, 71-90. https://doi.org/10.1016/j.dr.2016.06.004.

35.

Roth

D. L.

Fredman

Haley

W. E.

(2015). Informal caregiving and its impact on health: a reappraisal from population-based studies. The Gerontologist, 55(2), 309-319. https://doi.org/10.1093/geront/gnu177.

36.

Roth

D. L.

Haley

W. E.

Rhodes

D. J.

Sheehan

O. C.

Huang

Blinka

M. D.

Howard

V. J.

(2020). Transitions to family caregiving: enrolling incident caregivers and matched non-caregiving controls from a population-based study. Aging Clinical and Experimental Research, 32(9), 1829-1838. https://doi.org/10.1007/s40520-019-01370-9.

37.

Roth

D. L.

Haley

W. E.

Sheehan

O. C.

Huang

Rhodes

J. D.

Durda

Cushman

(2020). The transition to family caregiving and its effect on biomarkers of inflammation. Proceedings of the National Academy of Sciences, 117(28), 16258-16263. https://doi.org/10.1073/pnas.2000792117.

38.

Roth

D. L.

Sheehan

O. C.

Haley

W. E.

Jenny

N. S.

Cushman

Walston

J. D.

(2019). Is family caregiving associated with inflammation or compromised immunity? A meta-analysis. The Gerontologist, 59(5), e521-e534. https://doi.org/10.1093/geront/gnz015.

39.

Roth

D. L.

Skarupski

K. A.

Crews

D. C.

Howard

V. J.

Locher

J. L.

(2016). Distinct age and self-rated health crossover mortality effects for African Americans: evidence from a national cohort study. Social Sciences & Medicine, 156, 12–20. https://doi.org/10.1016/j.socscimed.2016.03.019.

40.

Sheehan

O. C.

Haley

W. E.

Howard

V. J.

Huang

Rhodes

J. D.

Roth

D. L.

(2021). Stress, burden, and well-being in dementia and non-dementia caregivers: insights from the Caregiving Transitions Study. The Gerontologist, 61(5), 670-679. https://doi.org/10.1093/geront/gnaa108.

41.

Smith

K. J.

Ollis

Schmitz

(2018). The association between C-reactive protein, interleukin-6 and depression among older adults in the community: a systematic review and meta-analysis. Experimental Gerontology, 102, 109-132. https://doi.org/10.1016/j.exger.2017.12.005.

42.

Tian

Hou

Yuan

T. F.

(2014). A possible change process of inflammatory cytokines in the prolonged chronic stress and its ultimate implications for health. The Scientific World Journal, 780616. https://doi.org/10.1155/2014/780616.

43.

Tukey

J. W.

(1977). Exploratory data analysis. Addison-Wesley.

44.

Varadhan

Yao

Matteini

Beamer

B. A.

Xue

Q. L.

Yang

Walston

(2014). Simple biologically informed inflammatory index of two serum cytokines predicts 10 year all-cause mortality in older adults. Journal of Gerontology: Medical Sciences, 69(2), 165-173. https://doi.org/10.1093/gerona/glt023.

45.

von Känel

Mills

P. J.

Mausbach

B. T.

Dimsdale

J. E.

Patterson

T. L.

Ziegler Grant

(2012). Effect of alzheimer caregiving on circulating levels of C-reactive protein and other biomarkers relevant to cardiovascular disease risk: a longitudinal study. The Gerontologist, 58(4), 354-365. https://doi.org/10.1159/000334219.

46.

Walston

McBurnie

M. A.

Newman

Tracy

R. P.

Kop

W. J.

Hirsch

C. H.

Gottdiener

Fried

L. P.

(2002). Frailty and activation of the inflammation and coagulation systems with and without clinical comorbidities: results from the cardiovascular health study. Archives of Internal Medicine, 162(20), 2333-2341. https://doi.org/10.1001/archinte.162.20.2333.

47.

Wing

Manton

K. G.

StallardHames

E. C. G.

Tryoler

H. A.

(1985). The black/white mortality crossover: investigation in a community-based study. Journal of Gerontology, 40(1), 78-84. https://doi.org/10.1093/geronj/40.1.78.

48.

Wirtz

P. H.

von Känel

(2017). Psychological stress, inflammation, and coronary heart disease. Current Cardiology Reports, 19(11), 111. https://doi.org/10.1007/s11886-017-0919-x.

49.

Wong

J. D.

Shobo

(2017). Types of family caregiving and daily experiences in midlife and late adulthood: the moderating influences of marital status and age. Research on Aging, 39(6), 719-740. https://doi.org/10.1177/0164027516681050.

50.

Yao

Robert

S. A.

(2011). Examining the racial crossover in mortality between African American and White older adults: a multilevel survival analysis of race, individual socioeconomic status, and neighborhood socioeconomic context. Journal of Aging Research, 132073. https://doi.org/10.4061/2011/132073.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.21 MB