Optimism, pessimism and health biomarkers in older couples

Risk of cardiovascular conditions (CVD) dramatically increases with age (Marengoni et al., 2011). These health conditions strongly impair the quality of life and functioning of older individuals, putting them at an increased risk of early mortality. The effects on the quality and length of life (Lloyd-Sherlock et al., 2019), together with the economic burden these conditions create, suggests the need to identify potential risk and protective factors of such conditions. Though cardiovascular problems are multi-determined, optimism and pessimism—two ways of characterizing individuals’ outcome expectations—have been suggested to influence the onset and severity of cardiovascular problems (Boehm & Kubzansky, 2012; Rozanski et al., 2019). In brief, optimists expect things to go their way and pessimists do not (Kubzansky et al., 2004). One of the suggested mechanisms linking optimism/pessimism to cardiovascular health have been cardiovascular-related biomarkers (e.g., C-reactive protein [CRP] and high-density lipoprotein [HDL])(Heath & Grant, 2020; Lagrand et al., 1999). However, few consistent relationships between optimism/pessimism and these biomarkers have emerged.

Further, emerging evidence suggests that the psychological characteristics of a person’s spouse can influence their health (Slatcher, 2010). Couples are necessarily interdependent (Chopik & Lucas, 2019; Hoppmann & Gerstorf, 2014; Jackson et al., 2015; Mejía & Gonzalez, 2017). Many factors bound couples together such as similarity in traits and experiences, but most relevant to the current study, couples engage in similar health behavior and have similar risk factors for poor health (Chopik & Lucas, 2019; Mejía & Gonzalez, 2017). For example, when one person makes a positive health change in their life, their partner is quick to follow (Jackson et al., 2015). Altogether, because health risks and health behavior are coordinated within couples, identifiable psychological characteristics in one person may predict indicators of cardiovascular functioning in their partner (Hoppmann & Gerstorf, 2014). Examining the effect of spousal optimism and pessimism on biomarkers is a crucial step in helping to understand the mechanisms linking psychological characteristics to health across the lifespan. In the current study, we examined associations between optimism, pessimism, and cardiovascular biomarkers in a large sample of older couples followed over an eight-year period.

Optimism, Pessimism, and Links with Health across the Lifespan

Although researchers initially conceptualized a unidimensional construct labeled optimism, accumulating evidence suggest two independent constructs of optimism and pessimism (Kubzansky et al., 2004). The independence of optimism and pessimism is more pronounced in older samples (Benyamini, 2005; Shrira et al., 2011), is affirmed in studies examining the heritability of each (Plomin et al., 1992), and has been established in psychometric and discriminant validity studies (Creed et al., 2002; Marshall et al., 1992). Such extensive work highlights the importance of understanding the unique effects of pessimism and optimism on health. Indeed, highly optimistic people are likely to act in healthier ways compared to their less optimistic counterparts (e.g., adhere to healthier diets, exercise, avoid smoking; Giltay et al., 2007). In addition, higher pessimism and lower optimism may increase negative reactions to ecological stressors, that can expose individuals to higher levels of stress-related hypothalamic-pituitary-adrenal axis activation, leading to aggravated cytokine production inflammation (Malek et al., 2015). This overproduction of inflammation and stress responses likely creates cardiovascular problems over time through allostatic processes (McEwen, 2005). However, importantly, strain-inflammation associations are inconclusive, with one recent study even showing inverse associations across time (Gough & Godde, 2018).

Only few studies have shown prospective associations between future expectations (i.e., optimism and pessimism) and cardiovascular problems among middle aged and older adults (Kim et al., 2014; Kim, James, et al., 2019; Matthews et al., 2004; Tindle et al., 2009). However, a handful of studies tested associations between future expectations and biomarkers that can function as precursors of cardiovascular problems. To indicate cardiovascular risk, these studies focused on inflammatory markers such as CRP and lipid indices, such as HDL. Most of these studies have used a cross-sectional design to test between-person associations, examining whether optimists and pessimists differ in levels of cardiovascular biomarkers. Critically reviewing these studies suggests that the associations found appear somewhat weak and inconsistent. For example, in a large cross-sectional study of midlife adults, higher optimism was linked to lower cardiovascular risk, as indicated by levels of HDL and triglycerides, but not by low-density lipoprotein (LDL) or total cholesterol levels (Boehm et al., 2013). In addition, in a large sample of midlife and older adults, pessimism—and to a lesser extent optimism—was linked to some inflammatory indices, such as interleukin-6 and CRP but not others (Roy et al., 2010). After adjusting for related constructs (i.e., depression and body-mass-index [BMI]), even these associations became mostly non-significant (Roy et al., 2010). One prospective study found that baseline optimism and pessimism were linked to overall levels of some (e.g., interleukin-6) but not all inflammatory markers (e.g., CRP). The associations were attenuated when additional factors were controlled for (e.g., age, depression, health habits). In addition, baseline optimism was not associated with changes in these markers over time (Ikeda et al., 2011).

To our knowledge, no study has used a prospective design to test within-person predictive processes linking relative deviations in individual’s future expectations with relative changes in health biomarkers. Support for this direction comes from a recent study of older adults, in which changes in optimism were linked to subsequent changes in chronic illness and perceived health (Chopik et al., 2018). In this study we test a similar model for biological functioning.

Spousal Optimism and Health across the Lifespan

According to the interpersonal perspective on risk and resilience (Smith et al., 2013; Yang et al., 2016), both an individual’s own optimism and pessimism and their partner’s optimism and pessimism should contribute to cardiovascular functioning. Numerous studies have linked health indicators across spouses, suggesting that a couples’ shared environment may have effects on health and well-being. Living together can expose couples to similar environmental risks and stressors, and stress experienced by one person may have an effect on their spouse (Kiecolt-Glaser et al., 2019; Repetti et al., 2009; Slatcher, 2010; Wofford et al., 2019). Troubled marriages can also be a source of stress, and through this route, affect health (Kiecolt-Glaser, 2017; Robles, 2014).

Changes in life circumstances in older adulthood, such as retirement, may result in placing an increased importance on the marital relationships (Luong et al., 2011). Moreover, older adults usually enjoy close relationships and put a greater emphasis on maintaining well-being in these relationships (Luong et al., 2011). This is likely to accentuate the reciprocal effects of optimism/pessimism on health between partners. For example, having a spouse high in optimism is associated with better subjective health, physical functioning, and fewer chronic illnesses for individuals—over and above an individual’s own levels of optimism (Chopik et al., 2018; Kim, Chopik, & Smith, 2014).

More optimistic or less pessimistic partners may more strongly support each other in the adoption of healthy habits (e.g., exercising) and the adherence to medical treatments. Partners were found to affect each other’s levels of physical activity, and this, in turn, mediates the relationship between life satisfaction (a correlate of optimism) and mortality (Stavrova, 2019). Conversely, pessimists might discourage their partners from maintain a healthy lifestyle, as they are less likely to believe their actions will have an impact on health. In one study, pessimism was associated with worse health habits, such as smoking (Kubzansky et al., 2004), and these habits can create an unhealthy living environment for both members of a couple through behavioral contagion or second-hand smoking. Partners’ pessimism can also impinge on the relational climate, increasing stress in the relationship (Srivastava et al., 2006). Optimism, on the other hand, is a resource, and optimistic couples are characterized by greater perceived support, better conflict resolution, and more relationship satisfaction (Assad et al., 2007). The optimism of one spouse, might reduce stress, thus positively affecting health through reducing allostatic processes (McEwen, 2005). But is being married to an optimist or a pessimist associated with a healthier biological profile?

The Current Study

Our primary aim was to test associations between wives and husbands’ optimism and pessimism and key biomarkers related to cardiovascular functioning—CRP and HDL—in a large sample of older couples. We examined how these dyadic processes were associated over three waves, each separated by four years (i.e., 8-years total). The longitudinal design enables us to distinguish between-and-within partners’ variability of future expectation and their associations with biomarkers. At the within-level we tested how deviations in partners’ future expectations from one wave to the next (i.e., relative to each partner own mean level) contribute to subsequent changes in their biomarkers levels. We tested these association both within a given wave and across one wave. At the between-level, we tested how partners’ average level of future expectation (i.e., aggregated across the three waves) is associated with their average level of biomarkers (i.e., aggregated across the three waves).

At the within-level, for concurrent individual effects (i.e., actor effects), we expected that, for individuals, increases in optimism and decreases in pessimism, relative to their mean level, would be associated with relative increases in HDL and decreases in CRP within the same assessment wave (HO1a). At the within-level, for concurrent partner effects, we expected that increases in individual optimism and decreases in pessimism, relative to their own mean level, would be associated with relative increases in HDL and decreases in CRP for their partners within the same assessment wave (HO1p).

We also tested whether within-person deviation in one person’s optimism and pessimism, (i.e., relative to this person mean level) preceded deviations in their (HO2a) and their partner’s (HO2p) biomarkers across one wave (i.e., lagged-effects). In other words, these lagged effects examine whether within-person deviations in partners’ future expectations in one wave (i.e., from waves 1 to 2), predict deviations in partners’ health biomarkers at the next wave. The rationale to expect these cross-wave within-person effects is that changes in partners’ future expectations may have pervasive and enduring effects on partners’ stress levels or health behaviors that can set into motion changes in CRP and HDL.

At the between-level, we expected that participants higher in optimism or lower in pessimism, relative to the rest of the sample, would also be higher in HDL and lower in CRP (HO3a). We also expected that participants higher in optimism or lower in pessimism, relative to the rest of the sample, would have partners who were higher in HDL and lower in CRP (HO3p, see Figure 1 Panel 3 for between-level actor-partner effects.

Figure 1.

The study’s conceptual model. Black arrows represent actor effects and dashed arrows represent partners effects. t= a given wave; t+1= next wave. Panel 2 - HO2 model is adjusted for autoregression associations and concurrent associations with partners’ future expectations. All models are adjusted for partners’

Method

Study Population

The sample included 3,243 married couples from the Health and Retirement Study (HRS), a longitudinal panel study of individuals 50 years and older followed every two years; psychological and biomarker data were collected every four years. To be included in this study, both partners had to provide biological data and reports on optimism and pessimism in at least one of the three waves [Wave_1,2006 (couples n=2024), Wave_2,2010 (couples n=2037) and Wave_3,2014 (couples n=2093)]. At Wave 1, wives’ average age was 63.24 (SD = 8.65) and husbands average age was 63.15 (SD=8.67). The average relationship length was 34.30 (SD=14.41). Because only two waves of the 2008 cohort were available¹, they were not considered here as we would be unable to model within-person associations for that cohort.

Measures

Biological Measures

Cardiovascular biomarkers were obtained using blood-based samples collected during at-home assessments. To capture cardiovascular functioning, we used measures of inflammatory and metabolic functioning which were collected in the HRS (for more details see Crimmins & Kim, 2017). Blood samples were obtained by finger-prick with a sterile lancet to release several drops of blood onto filler paper. Blood samples were assayed for CRP at the University of Vermont using a standard ELIZA assay. HDL was assayed at Biosafe Laboratories in 2006, and at 2010 and 2014 at the University of Washington. To standardize concentration values across different assays, HRS produced an adjusted value using the distribution obtained in a similarly aged nationally representative sample with conventional assays, the National Health and Nutrition Examination Survey (for more details, see Crimmins et al., 2014).

C-Reactive Protein (CRP) is an indicator of inflammation associated with acute and chronic conditions ranging from injury to chronic inflammatory diseases. Chronically high levels of CRP are associated with CVD and hypertension (Heath & Grant, 2020; Lagrand et al., 1999). Briefly, CRP may lead to an inflammation of the atherosclerosis by inducing adhesion molecule expression in the endothelial cells. To make use of all available data, biomarker values that were larger than three standard deviations above the mean were replaced with values corresponding to three standard deviations above the mean (i.e., Winsorized; Reifman & Keyton, 2010). A total of 85 (85/3614= 2.3%) values for husbands and 87 for wives (87/3744= 2.4%) were higher than 3 SD above the mean. After we recoded these values at 3 SD above the mean (raw value= 18.16 for husbands; raw value=18.42 for wives), skewness => 2. We performed a natural log transformation to adjust for right-skew. This resulted in CRP scores that were sufficiently normal to render parametric statistics valid (Garson, 2012).

High-density lipoprotein (HDL) is considered “good cholesterol,” with higher levels of HDL associated with decreased risk of cardiovascular diseases (Rader & Hovingh, 2014). CRP and HDL are both considered indicators of allostatic load (Thomson et al., 2019). A total of 51 (51/3271= 1.56%) values for husbands and 54 for wives (54/3471= 1.56%) were higher than 3 SDs above the mean. After we recoded these values at 3 SDs above the mean (raw value= 82.59 for husbands; raw value=101.65 for wives), the distributions were sufficiently normal to render parametric statistics valid. See Table 1 for descriptive statistics.

Table 1.

Descriptive Statistics for the Study Constructs

	Mean	SD	Min	Max	Skewness	Kurtosis
1. Pes-W	2.32	1.03	1.00	6.00	.66	−.22
2. Pes -H	2.56	1.01	1.00	6.00	.42	−.44
3. Opt- W	4.59	.94	1.00	6.00	−.65	.32
4. Opt - H	4.55	.90	1.00	6.00	−.66	.63
5. HDL- W	58.53	13.94	21.42	101.65	.64	.33
6. HDL- H	49.18	10.98	13.64	82.59	.68	.49
7. CRP- W	3.56	4.65	.05	18.42	1.95	3.72
8. CRP- H	2.94	3.73	.02	18.41	2.57	6.87
9. LnCRP-W	.75	1.15	−2.94	2.91	−.32	−.17
10. LnCRP-H	.48	1.12	−.400	2.91	−.04	−.07
11. Age- W	61.49	8.62	30.00	86.00	.00	−.14
12. Age- H	66.75	7.76	31.00	94.00	.07	−.28
13. BMI- W	28.11	6.06	10.58	67.35	1.21	2.94
14. BMI- H	28.17	4.62	15.45	61.08	1.06	2.63
15. Cond- W	1.92	1.32	0.00	8.00	.60	.27
16. Cond- H	2.09	1.30	0.00	7.00	.49	−.06
17. Dep- W	1.18	1.49	0.00	8.00	1.82	3.31
18. Dep- H	.92	1.29	0.00	8.00	2.12	4.94
19. ADL- W	.19	.55	0.00	5.00	4.13	20.98
20. ADL- H	.21	.52	0.00	4.33	3.42	13.37

Note. W=wife; H=husband; Pes= raw scores on the pessimism scale of the Life Orientation Test – Revised (LOT-R); Opt= raw scores on the optimism scale of The Life Orientation Test – Revised; HDL = High-density lipoprotein; CRP= C-reactive protein; Cond=Number of self- reported health conditions; Dep = scores on the Center for Epidemiologic Studies Depression Scale; ADL= Activities of Daily Living; Age = age at Wave 1 (2006); BMI = body Mass Index at Wave 1. LnCRP= natural log-transformed C-reactive protein values.

^*p<.05.

^**p<.01.

^***p < .001

Optimism and Pessimism

Participants completed The Life Orientation Test – Revised (LOT-R; Scheier et al., 1994). The LOT-R consists of three positively worded items measuring optimism and three negatively worded items measuring pessimism. Respondents rated their agreement with the statements on a scale ranging from 1(strongly disagree) to 6(strongly agree). Items were averaged to create the two subscales. Cronbach’s alpha ranged across the three waves from .78 to .85 for the optimism scale and .76 to .79 for the pessimism scale.

Covariates

We included several covariates due to their putative associations with either future expectations or the cardiovascular biomarkers. Individuals reported on their age, marital length, household income, and medications. Participants indicated whether a doctor had ever diagnosed them with various health conditions (e.g., diabetes, cardiovascular diseases). Depression during the past week was measured at each wave with the 8-item short form of the Center for Epidemiologic Studies Depression Scale (CESD; (Radloff, 1977) given that future expectations is the hallmark of depression (Zetsche et al., 2019). We also controlled for participants’ Activities of Daily Living (ADLs). Participants were asked whether they had difficulties in bathing, eating, dressing, walking across the room, and getting in and out of bed. Responses were coded as 1 if participants experienced difficulties, and 0 if not, and the number of positive answers were summed to create an overall ADL score. Weight and height were measured, from which BMI (weight/height²) was calculated.

Data Analyses

To test hypotheses HO1–3, we used two-level path analyses with assessments (i.e., waves) nested within persons (i.e., wives and husbands) and regressed each biomarker on individuals’ and partners’ future expectations at both the within-person (i.e., partners’ optimism and pessimism as time-varying predictors) and between-person levels simultaneously (i.e., average levels of partners’ optimism and pessimism across the three time-points). This multi-level approach allowed us to partition the variance into within-person and between-person components. We used random intercepts and fixed slopes for these analyses.

At the within-person level (HO1), we tested whether changes in one person’s optimism and pessimism predicted concurrent changes in their partners’ biomarkers (i.e., are increases/decreases in individual/partner optimism associated with increases/decreases in biomarkers?). To account for the interdependence between partners’ health biomarkers within each wave, we predicted each biomarker (i.e., HDL, CRP) simultaneously for both partners to allow for co-variation between the wife and husband residual within each dyad at each wave.

To test HO1 we used the following generic equations at Level-1 (i.e., within-person effects):

Level 1:

$${\mathit{\text{Biomarker-wife}}\mathit{\left(}\mathit{\text{same-wave}}\mathit{\right)}}_{ijw}={\beta }_{0jw}+{\beta }_{1jw}{\mathit{\text{ same-wave wife optimism}}}_{1ij}+{\beta }_{2jw}{\mathit{\text{ same-wave wife pessimism}}}_{2ij}+{\beta }_{3jw}{\mathit{\text{ same-wave husband optimism}}}_{3ij}+{\beta }_{4jw}{\mathit{\text{ same-wave husband pessimism}}}_{4ij}+{\beta }_{\text{5}jw}{\mathit{\text{ wave of reporting}}}_{5ji}+r_{\text{ijw}}$$ (1)

$${\mathit{\text{Biomarker-husband}}\mathit{\left(}\mathit{\text{same-wave}}\mathit{\right)}}_{ijh}={\beta }_{0jh}+{\beta }_{1jh}{\mathit{\text{ same-wave wife optimism}}}_{1ij}+{\beta }_{2jh}{\mathit{\text{ same-wave wife pessimism}}}_{2ij}+{\beta }_{3jh}{\mathit{\text{ same-wave husband optimism}}}_{3ij}+{\beta }_{4jh}{\mathit{\text{ same-wave husband pessimism}}}_{4ij}+{\beta }_{5jh}{\mathit{\text{wave of reporting }}}_{5ij}+r_{\text{ijh}}$$ (2)

For example, Equation 1 specifies that a wife’s biomarker at time i in dyad j (Biomarker-wife (same-wave)_ijw) is a function of the wife’s intercept specific to dyad j (β_0jw), a wife’s slope specific to dyad j representing the effect of within-person variation in optimism (β_1jw), a wife slope specific to dyad j representing the effect of within-person variation in pessimism (β_2jw), a husband slope specific to dyad j representing the effect of within-subject variation in optimism on his wife’s biomarker (β_3jh), a husband slope specific to dyad j representing the effect of within-person variation in pessimism on his wife’s biomarker (β_4jh), a wife slope representing the passage of time (i.e., across waves 1–3; β_5jw), and a residual specific to time i for dyad j (r_ijw). Equation 2 is the equivalent for husbands.

To test HO2 that changes in future expectations precede changes in partners’ biomarkers, we conducted lagged analyses in which future expectations at wave t predicted change in health biomarkers from wave t to t + 1, while controlling for autoregressive effects of health biomarkers and concurrent effects of partners’ future expectations (see Bolger et al., 2003). We allowed covariation between the lagged variables of partners’ future expectations and the biomarker. We person-mean centered the lagged variables (Enders & Tofighi, 2007), and added these variables to Level-1 Equations 1 and 2.

The between-person HO3 was tested at Level-2 simultaneously with HO1–2 at Level-1. At the between-person level, we tested how wives’ and husbands’ average levels of biomarkers were associated with average future expectations (i.e., is being an optimist or pessimist on average associated with biomarkers, on average?). We ran the models together for husbands and wives to allow for co-variation between the wife and husband residuals at each dyad.

We used the following generic equations at Level-2 (i.e., between-person effects):

Level 2:

$${\beta }_{0jw}={\gamma }_{00\text{w}}+{\gamma }_{01\text{w}}\mathit{\text{ wife optimism }}+{\gamma }_{02\text{w}}\mathit{\text{ wife pessimism }}+{\gamma }_{03\text{w}}\mathit{\text{ husband optimism }}+{\gamma }_{04\text{w}}\mathit{\text{ husband pessimism}}\text{+ }{u}_{0\text{jw}}$$ (3)

$${\beta }_{0jh}={\gamma }_{00\text{h}}+{\gamma }_{01\text{h}}\mathit{\text{ wife optimism }}+{\gamma }_{02\text{h}}\mathit{\text{ wife pessimism }}+{\gamma }_{03\text{h}}\mathit{\text{ husband optimism }}+{\gamma }_{04\text{h}}\mathit{\text{husband pessimism+ }}{u}_{0\text{jh}}$$ (4)

Level-2 Equations specify between-person variation in the coefficients of the Level-1 equations. Thus, for example for a wife, Equation 3 specifies that between-person variation in a wife’s biomarkers is a function of an intercept (γ_00w), an effect of between-person differences in wife optimism (γ_01w), an effect of between-person differences in wife pessimism (γ_02w), an effect of between-person differences in husband optimism (γ_03w), an effect of between-person differences in husband pessimism (γ_04w), and a residual component specific to each dyad (u_0jh). Equation 4 is the equivalent for husbands. Although the Level-1 notation allows for between-person variation in β_1–5jw/h slopes for wife and husband, there is in fact no actual variation, merely a constant value for each. Equations 1–4 are the equivalent for husbands.

We ran analyses with and without controlling for depression, health condition, ADLs, and BMI, which were all associated with one or more of the study’s constructs (see Table 2). Analyses were conducted in Mplus 7.4, using full information maximum likelihood (FIML) that can handle missing data and a robust estimator (MLR) to handle non-normal distribution of the data (Chopik et al., 2020; Muthén & Muthén, 2012).²

Table 2.

Intercorrelations Between the Study Constructs

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18
1. Pes-W
2. Pes -H	.39**
3. Opt- W	−.45**	−.18**
4. Opt - H	−.14**	−.38**	.17**
5. HDL- W	−.08**	−.08**	.01	.01
6. HDL- H	−.08**	−.09**	.04	.04	.24**
7. CRP- Wa	.07**	.08**	−.01	−.07*	−.16**	−.03
8. CRP- Ha	.12**	.17**	−.05	−.12**	−.03	−.08*	−.01
9. Age- W	.02	−.01	−.01	.03	−.03	−.03	−.01	−.01
10. Age- H	.03	.01	−.02	.03	−.02	−.02	−.03/.00	−.01	.79***
11. BMI- W	.14**	.16**	−.03	.02	−.25**	−.12*	−.05**	−.04*	−.09**	−.08**
12. BMI- H	.04	.10**	−.01	−.02	−.07*	−.20**	−.04*	−.05**	−.12*	−.15*	.25***
13. Cond- W	.23*	.13*	−.14**	−.01	−.17**	−.09**	.03	−.04*	.27***	.22**	.30***	.23**
14. Cond- H	.13*	.19**	−.08**	−.12**	−.08**	−.15**	−.03	.001	.25**	.30***	.12**	.07*	.21**
15. Dep- W	.40***	.22**	−.28**	−.03	−.08**	−.10**	.02	.001	−.04*	−.04*	.18**	.04*	.38***	07**
16. Dep- H	.19**	.37**	−.08**	−.30***	−.08**	−.10**	−.02	−.01	−.09*	−.09*	.14**	.07*	.10**	.26***	.28***
17. ADL- W	.16**	.14**	−.08*	−.02	−.05	−.07*	.02	−.01	.11*	.09**	.19**	.01	.32***	.09**	.37**	.15**
18. ADL- H	.14**	.18**	−.06*	−.10**	−.06*	−.06*	−.02	−.01	.09*	.14**	.07*	.12**	.12**	.27**	. 18*	.38**	.15*

Note. Associations are between average level of constructs, aggregated across the three waves. W=wife; H=husband; Pes= raw scores on the pessimism scale of the Life Orientation Test – Revised (LOT-R); Opt= raw scores on the optimism scale of The Life Orientation Test – Revised ; HDL = High-density lipoprotein; CRP= C-reactive protein; Cond=Number of self- reported health conditions; Dep = scores on the Center for Epidemiologic Studies Depression Scale; ADL= Activities of Daily Living; Age = age at Wave 1 (2006); BMI = body Mass Index at Wave 1.

^aCorrelations are with natural log-transformed C-reactive protein values

^*p<.05.

^**p<.01.

^***p < .001

Results

Preliminary Analyses

Table 1 presents descriptive statistics for the study’s variable and covariates. Table 2 presents the correlations among the study variables. Across all waves, individuals’ and partners’ future expectations were significantly correlated. Also, a person’s pessimism levels were negatively related to their own and their partner’s HDL, and positively to CRP levels. Husbands’ optimism negatively correlated with their own and their wives’ CRP levels.

The associations between partner’s future expectations and depression were all significant with one exception: the association between husbands’ optimism and wives’ depression. An individual’s HDL levels were negatively associated with their own and their partner’s depression (rs ranged .08–.10). An individual’s CRP levels were associated with their own depression and wives’ CRP also associated with husbands’ depression.

Paired sample t-tests showed that husbands reported higher pessimism (t(1428)= 7.98, p < .001; Cohen’s d =.22) than wives. Wives had significantly higher HDL and CRP levels than their husbands (t(1332)=−.21.93, p < .001, d = .77; t(1359) = −6.01, p < .001, d=.22; respectively).

Hypotheses Testing³

Within-person Concurrent Effects of Optimism and Pessimism on Partners’ Biomarkers (HO1)

We first tested how changes in partners’ future expectations (i.e., optimism and pessimism) associated with concurrent changes in partners’ biomarkers (i.e., HDL, CRP), at the same time point (i.e., at the same-wave t). Results are presented in the top panel of Table 3. As can be seen all effect were negligible in size and non-significant.

Table 3.

Multilevel Models to Predict Health Biomarkers

	HDL				CRP
Within-dyad predictions	Model 1a		Model 1b		Model 2a		Model 2b
Concurrent effects (HO1)	Wife	Husband	Wife	Husband	Wife	Husband	Wife	Husband
Pes-W (same wave)	.02(.02)	.001(.02)	.03(.02)	.03(.02)	.01(.03)	−.01(.02)	−.01(.03)	−.001(.02)
Pes -H (same wave)	−.02(.02)	.02(.02)	−.02(.02)	−.003(.02)	.02(.02)	−.02 (.02)	−.01(.02)	−.01(.03)
Opt-W (same wave)	.03(.02)	.02(.03)	.02(.03)	.01(.02)	.001(.03)	−.001(.03)	.02(.03)	−.01(.02)
Opt -H (same wave)	.001(.02)	−.001(.03)	.01(.02)	−.01(.02)	−.03(.03)	.01(.02)	−.02(.03)	.02(.02)
Lagged effects (HO2)
Pes-W (prior wave)	.001(.02)	.001(.03)	.02(.02)	.01(.03)	−.01(.02)	.02(.02)	.00(.02)	.00(.02)
Pes -H (prior wave)	−.001 (.02)	−.01(.03)	−.01 (.02)	−.02(.03)	−.01(.02)	.02(.02)	−.03(.03)	.01(.02)
Opt -W (prior wave)	−.02(.03)	−.03(.04)	−.02(.03)	−.02 (.03)	.003(.04)	−.01(.02)	−.02(.03)	−.01(.02)
Opt -H (prior wave)	.02(.02)	−.01(.03)	.02(.02)	−.01(.03)	−.02(.02)	−.01(.02)	.00 (.02)	−.001(.02)
Biomarker (prior wave)	−.23(.03)***	−.33(.03)**	−.35(.02)***	−.33(.02)**	−.16(.02)***	−.20(.03)***	−.22(.02)***	−.25(.03)***
Covariates
Dep- W(same wave)			−.01(.02)	−.01(.02)			−.01(.03)	.01(.02)
Dep- H(same wave)			.02(.02)	−.01(.02)			.001(.03)	−.01 (.03)
Cond-W(same wave)			.03(.05)	.00(.03)			−.04(.03)	.06(.03)*
Cond-H(same wave)			−.09(.05)	.01(.03)			−.01(.03)	−.04(.03)
Adl-W(same wave)			.01(.02)	−.01(.02)			.02(.03)	.03(.03)
Adl-H(same wave)			−.02(.02)	−.01(.02)			−.04(.02)	−.01(.02)
Time (wave number)	−.08(.02)***	.07(.02)**	−.09(.03)***	.07(.03)*	−.22(.02)***	−.12(.03)***	−.21(.03)***	−.18(.03)***
Between-dyads differences (HO3)
Pes-W(average level)	−.13(.04)***	−.08(.05)	−.05(.04)	.09(.10)	.07(.04)*	.08(.04)*	.01(.03)	.05(.04)
Pes-H (average level)	−.09(.04)*	−.11(.05)*	−.03(.04)	−.11(.09)	.07(.04)	.15(.04)***	.01(.03)	.11(.04)**
Opt-W(average level)	−.02(.04)	.001(.05)	−.03(.04)	−.11(.10)	.03(.04)	.01(.04)	.05(.03)	−.001(.04)
Opt-H(average level	.03(.04)	.001(.04)	.05(.04)	−.05(.09)	−.04(.03)	−.06(.03)	−.09(.03)**	−.05(.03)
Covariates
Dep-W(average level)			−.05(.04)	−.07(.04)			.02(.04)	−.06(.04)
Dep-H(average level)			−.01(.04)	−.06(.05)			−.03(.03)	−.01(.05)
Cond-W(average level)			−.10(.04)*	−.05(.05)			.08(.04)*	.08(.04)*
Cond-H(average level)			.03(.05)	−.13(.04)***			−.001(.03)	.07(.03)*
ADL_W(same wave)			.04(.04)	−.03(.04)			.05(.03)	−.05(.03)
ADL_H(same wave)			−.05(.04)	.02(.04)			.06(.03)	.07(.03)*
BMI-W (baseline)			−.31(.04)***	-			.42(.03)***	-
BMI-H(baseline)			-	−.19(.04)***			-	.28(.03)**
R squared for within	.008*	.05	.12***	.12***	.08***	.05***	.10***	.10***
R squared for between			.13***	.14***	.02***	.05***	.24***	.17***

Note. All values are standardized betas. Standard errors are in parentheses. Models a tests the direct effects of partners’ future expectation on health biomarkers without covariates. Models b tests the direct effects of partners’ future expectation on health biomarkers with covariates. HO= Hypothesis; W=wife; H=husband; Pes= raw scores on the pessimism scale of the Life Orientation Test – Revised (LOT-R); Opt= raw scores on the optimism scale of The Life Orientation Test – Revised; HDL = High-density lipoprotein; CRP= C-reactive protein; Cond=Number of self- reported health conditions; Dep = scores on the Center for Epidemiologic Studies Depression Scale; ADL= Activities of Daily Living; BMI = body Mass Index at Wave 1. Coefficients are for log transformed CRP values.

^*p<.05.

^**p<.01.

^***p < .001

Within-person Lagged Effects of Optimism and Pessimism on Partners’ Biomarkers (HO2)

To further explore the sequential order of change in partners’ future optimism and pessimism and health biomarkers, we conducted lagged analyses with optimism and pessimism at wave t-1 predicting changes in partners’ biomarkers from wave t-1 to the next wave t. Results are summarized in the middle panel of Table 3. All effects were negligible in size and non-significant.

Between-person Effects of Optimism and Pessimism on Partners’ Biomarkers (HO3)

At the between-person level, we tested if average levels of partners’ optimism and pessimism were associated with average levels of biomarkers. Results for the between-person effects are summarized in the bottom panel of Table 3. Husbands’ pessimism was positively associated with own CRP levels, and husbands’ optimism was negatively associated with their wives’ CRP.

Discussion

In the current study we examined whether relative changes in husbands and wives (i.e., within-person effects) contribute to changes in their health biomarkers, within the same wave (HO1) and across waves (HO2). We also tested whether average levels of future expectations were associated with average levels of biomarkers (i.e., between-person effects) (HO3).

Results for within-person effects were all non-significant, both within (HO1) and across waves (HO2). To our knowledge, this is the first study to test within-person effects of future expectations on biomarkers. The non-significant effects within-people across waves suggest that subtle relative changes in future expectations do not translate into subsequent changes in health biomarkers. This is in line with non-significant effects of future expectations on epigenetic processes and cellular aging (Kim, Fong, et al., 2019). But why would future expectations be associated with physical health but not these health biomarkers? It is possible that future expectations exert their influence on physical health via other pathways, such as attenuated stress response and its related physiological mechanisms and particularly the HPA, or the frequency of physical activity (Huffman et al., 2016; Lai et al., 2005). Nevertheless, biomarkers should still be the mechanisms linking future expectations to CVD. These mechanisms should be examined in future studies.

The between-person (HO3) associations were largely non-significant. However, we found an association between husbands’ pessimism and their own CRP, and an association between husbands’ optimism and wives’ CRP. Nonetheless, both effects were small in magnitude (Table 3, Models b). Finding rather weak associations between average biomarkers and future expectations for individuals (i.e., actor effects) (HO3a) is consistent with most prior studies (Ikeda et al., 2011; Roy et al., 2010). As an exception, Boehm (2013) found positive associations between optimism and HDL, albeit among participants with a wider age range. It may be that the influence of future expectations on health biomarkers declines when increase in age-related morbidity, overrides the influence of psychological characteristics on biomarkers (Beard et al., 2016). In fact, our findings suggest that partners’ physical health conditions are the strongest predictor of between-person differences in biomarkers and not their future expectation or mental health. The non-significant associations between average HDL and average own or partner future expectations may be because HDL is mainly a product of metabolic cascades that are strongly governed by genetics (Vitali et al., 2017) and health behaviors (Vincent et al., 2019). Therefore, the direct contribution of future expectations may be rather small, especially in older age when metabolic problems increase (Yu et al., 2019).

The one between-person association that appeared robust was between husbands’ pessimism and their own CRP levels. This might suggest that men’s pessimism is associated with physiological or behavioral processes that weaken the immune system. At the physiological level, men’s pessimism might be linked to strong activation of stress systems, which, in turn, through a cascade of chemical reactions, influences the functioning of the immune system (Malek et al., 2015). At the behavioral level, higher pessimism might discourage men from engaging in a healthy lifestyle, such as regular exercise, which consequently may increase CRP levels (Kasapis & Thompson, 2005).

Our data do not explain why we did not obtain a similar association for wives (i.e., actor effect; HO3a). Other factors, such as the importance of appearance, may be a strong motivator to choose a healthy diet and physical activity, whereas this might not have as much effect for men. Gender differences should be further examined. In addition, the reason why pessimism—and not optimism—was related to husbands’ CRP is unclear, but pessimism might be linked to other health-undermining conditions that could explain this effect, such as loneliness.

Finally, we found an inverse association between husbands’ optimism and wives’ CRP, albeit only after we controlled for covariates. Perhaps husbands’ low optimism is expressed in ways that undermine their wives’ immunity. Low-optimistic husbands may be more worried on average, which may increase their wives’ stress-response system, decreasing immune functioning. Husbands’ low optimism may also undermine their wives’ motivation to maintain a healthy lifestyle. Non-significant associations between wives’ future expectations and husbands’ CRP suggests that wives’ expectations are not robustly related to their partners’ biomarkers. The difference might lie in the sources husbands and wives turn to for support: women tend to disclose their thoughts and worries to female friends, whereas men tend mostly to disclose to their wives (Ermer & Proulx, 2020). As a result, the personal characteristics of husbands might have a stronger influence on wives’ health than vice versa.

A few limitations are worth noting. First, we used the two biological markers of CVD that were available in the HRS. However, future studies should replicate these null results with other biological markers of inflammation (e.g., cytokines [interleukin (IL)-1, IL-6, IL-8] (Zakynthinos & Pappa, 2009), and metabolic activity (e.g., low-density lipoprotein cholesterol)(Albert & Tang, 2018). Second, we did not find any prospective effects between future expectations and biomarkers, possibly because of the long gaps between waves (four years)(Kuiper & Ryan, 2018). Future research should examine these effects over shorter intervals. Also, both HDL and CRP tend to change in concert with the onset of medical conditions and health habits. Although we controlled for these factors, it is plausible that a single sample of health biomarkers does not reliably reflect a person’s baseline level of these markers, and additional measures across days or weeks are needed.

Despite these limitations, our data suggest that, among older adults, future expectations are not a major player in inter-or intrapersonal variability in couples’ biomarkers. As it is important to target risk factors, it is also important to clarify misconceptions and eliminate non-risk factors in order to promote empirically-driven prevention programs. For example, it is plausible that the previously found links between future expectations and CVD are mediated by other physiological systems, which are more closely related to the stress processes, specifically the autonomic and endocrine systems(Lai et al., 2005) and CVD (Gordan et al., 2015). Future studies may also clarify whether the previously found associations between future expectation and CVD may reflect the contribution of other psychological factors, and not future expectations. (Sutin et al., 2010).

Conclusion

The current study examined between-and within-person effects of future expectations on health biomarkers. We found that future expectations—both of individuals and their spouses—were largely unrelated to biomarkers, with one exception of husbands’ high pessimism and low optimism being related to their own and partners’ CRP levels, respectively. Our results provide some guidance for future research to examine factors that might link future expectations to health—whether it be reports of perceived health or physiological indicators of illness. Studies following couples across time should also partition the variance of between-and within individual effects, like we did here. Altogether, our results help refine our understanding about the psychological antecedents of physical health and contextualize these effects in a dyadic setting.