A longitudinal study of age-based change in blood pressure reactivity and negative affect reactivity to natural stressors

Abstract

Objective:

Aging is theoretically accompanied by emotional gains, but physiological self-regulatory losses. Emotional and physiological regulation can be operationalized as the extent of increase in negative affect and blood pressure upon experiencing a stressor (i.e., reactivity). The direction of age-based changes in negative affect reactivity to stressors is uncertain. Additionally, evidence for age-based increases in blood pressure reactivity to stressors is based largely on age-based differences observed in cross-sectional and lab-based studies. The present study is the first to examine long-term longitudinal changes in stress-related reactivity for both blood pressure and negative affect in the natural environment.

Methods:

375 healthy adults aged 50-70 years completed six days of hourly ambulatory blood pressure assessment and electronic diary reports of social conflict and task demand and control. 255 participants repeated three days of assessment in a six-year follow-up. With reactivity operationalized as the change in an outcome in association with momentary social conflict, task strain, or task demand (i.e., a model derived slope parameter), multilevel models were used to assess aging-based change in blood pressure and negative affect reactivity over the course of the six-year follow-up.

Results:

Aging is associated with increased diastolic blood pressure reactivity to social conflict and task demand (β_{social_conflict}= 0.48, p=.007; β_{task_demand} =0.19, p=.005), increases in negative affect reactivity to social conflict and task strain (β_{social_conflict}=0.10, p<.001; β_{task_strain}=0.08, p=.016), and increases in systolic blood pressure reactivity to task-based stress(β_{task_strain}=1.29, p=.007; β_{task_demand}=.23 p= .032).

Conclusion:

Findings suggest age-based increases in affective and cardiovascular reactivity to natural stressors.

INTRODUCTION

Aging is a multidimensional process. In the cardiovascular domain, increased sympathetic activation in arterial wall muscles and increased arterial stiffness lead to higher clinical blood pressure (BP) with age (1). Structural and functional changes in the arteries and vascular system also lead to slower recovery from BP spikes (1). Nevertheless, these biological declines are accompanied by theorized socioemotional gains (2). Socioemotional theories of aging suggest individuals benefit from improved emotion-regulation strategies and changing motivations across adulthood (2,3). Socioemotional Selectivity Theory suggests that with age, adults view time as scarcer, which motivates focus on positive experiences and social ties (4). Older adults report lower negative affect, more satisfying relationships, more positive attentional biases, and greater engagement in stressor-avoidance techniques than younger adults (3,5–8). Less clear is whether and how biological and psychosocial aging are reflected in concurrent cardiovascular and negative affect reactivity to momentary stressors.

Physiological changes with age are hypothesized to increase BP reactivity to stressors. Indeed, a meta-analysis of age-related differences in cardiovascular reactivity concluded that systolic blood pressure (SBP) reactivity to laboratory stressors was higher among older adults than younger adults (9). While there was some evidence for increased diastolic blood pressure (DBP) reactivity, this did not exceed a conventional threshold for significance (9). A study of middle-aged and older adults found that decreased baroreflex sensitivity and increased carotid artery stiffness are associated with greater mean arterial BP reactivity to stress—potential mechanisms for age-differences in reactivity (10). Changes in BP reactivity are important to assess because increasing evidence links high BP reactivity to advanced subclinical cardiovascular disease (CVD), and increased risk of hypertension, stroke, and CVD-related death (11–13).

Features of the current literature limit the conclusions we can draw about aging and cardiovascular reactivity to everyday stressors. First, most studies of BP reactivity use lab-based stressors. While laboratory findings elucidate mechanisms of cardiovascular response to stress, it is unclear whether age-differences in BP reactivity extend to everyday stressors (9). Second, studies addressing age-effects in BP reactivity almost exclusively use cross-sectional samples, and most compare older and younger adults, excluding middle-aged (9). The only longitudinal study of BP reactivity found that older adults experienced larger increases in SBP (but not DBP) reactivity to laboratory stress tasks over ~10 months (14). These results may not generalize to everyday stressor experiences, as cross-sectional evidence with ecological momentary stressors suggest ambulatory DBP (but not SBP) reactivity is greater in older adults (15). Additionally, though older adults experienced greater increases in BP reactivity than younger adults over ~10 months (14), cardiovascular aging unfolds on a larger time-scale (e.g., half decades, decades) (16). Empirical work must confirm whether longitudinal age change follows patterns of cross-sectional age differences (17), and whether laboratory-based reactivity generalizes to momentary stressor experiences.

While mean negative affect (NA) is expected to decline with age, theory and evidence are mixed regarding age-effects on NA reactivity to stressors. Some evidence suggests older adults show higher affect reactivity (18–20); others find lower reactivity (15,21,22); and yet others find no age differences (23,24). Strength and Vulnerability Integration theory suggests that accumulated self-regulation experience provides older adults with the resources to reduce exposure to negative social ties, and pre-emptively or attentionally distance themselves from stressors (3). However, age-based benefits disappear in the face of stressors that induce sustained emotional arousal and hinder appraisal or avoidance strategies (3). Additionally, affect reactivity may differ by type of stressor experienced (23,25–27). For instance, older adults have demonstrated lower reactivity to interpersonal stressors than younger adults but equivalent reactivity to overloads at work or home (26). The nature of the stressor and the timing of assessments may lead to disparate findings on whether age is associated with increases or decreases in affective stressor reactivity (3,17,28). The above studies have measured NA reactivity by differing methods, including daily diary, ecological momentary assessment (EMA), and responses to laboratory-based stressors. The present study uses two types of common everyday stressors—social conflict and task strain—to examine age-based changes in momentary NA reactivity. EMA of social conflict and task strain aids comparability to both experimental and daily diary work because these two stressor types are thematically similar to the interpersonal and cognitive challenges employed in laboratory reactivity studies, and they also represent some of the most important daily stressors found in daily diary studies. Temporally, EMA offers immediate assessment, like lab studies, but with naturalistic stressors as in daily diary designs.

Only three studies have examined the congruence of BP and affective responses to stress across age, though with different methods and mixed results. In a cross-sectional, laboratory-based study, Luong and Charles (22) found that older adults had both lower affective and cardiovascular (heart rate and BP) reactivity to laboratory-based social conflict compared to younger adults. Older adults’ reactivity advantages were primarily due to differences in appraisal and coping strategy use (22). In a longitudinal, laboratory-based study, Uchino et al., (14) examined age differences (sample aged 30-70 years) in ~10 month change in NA and BP reactivity averaged between a speech and mental arithmetic lab-task. They found that compared with their younger counterparts, older adults experienced larger decreases in NA reactivity over 10 months, with concurrent increases in SBP reactivity and parasympathetic withdrawal. In another study, Uchino et al., (15) used EMA to test cross-sectional age differences among middle aged adults (M(SD)=52.80(10.00)) in BP and NA reactivity to “an everyday hassle or problem.” Older adults experienced lower NA reactivity but higher DBP reactivity (15). The present study aims to address whether age-based changes in ambulatory blood pressure (ABP) reactivity and NA reactivity to momentary stressors follow similar patterns when assessed longitudinally, over time intervals longer than those previously assessed in this literature.

Using data from adults aged 50-70 followed for six years, the present study is the first to test long-term longitudinal changes in simultaneous cardiovascular and affect reactivity to ambulatory stressors. The present study’s use of EMA and ABP monitoring extends foundational laboratory research to allow for the study of BP reactivity to everyday stressful experiences. The focus on middle aged adults also fills a gap in age-based reactivity literature that often compares older adults to younger adults. Midlife is a key developmental period when intervention and prevention for cardiovascular health may have greater impact than later in life or after the onset of disease, particularly for slow accumulating disease processes like atherosclerosis (16). With advancing age, we expect emotion regulation benefits will result in decreased NA reactivity to stressors, while physiological vulnerabilities will result in increased stressor-related BP reactivity. We assess two types of stressors—task strain and social conflict—to explore whether patterns of reactivity change differ across two stressor types, each of which has been shown to be associated with elevated subclinical atherosclerosis (13,29).

METHODS

To study the associations between cardiovascular reactivity, negative affect reactivity and subclinical CVD, the present study analyzes data from the Pittsburgh Healthy Heart Project (PHHP), a prospective epidemiological study of bio-behavioral predictors of subclinical CVD progression. Below, we briefly describe information relevant to the present study. For detailed information about sampling criteria, recruitment, and study procedures, see Kamarck et al. (30,31).

Participants

Two waves of the PHHP, consisting of N=375 community-dwelling, middle-aged adults (aged 50-70 years at baseline), were included in analyses. Participants were 86% White, 52% male, and 51% held a bachelor’s or higher degree (see Table 1). The study recruited adults with no hypertension or history of long-term medical disorders including CVD, and no pharmacological treatment for hypertension or hypercholesterolemia in the past year at baseline. At the 6-year follow-up, 30% of the sample reported current or past use of antihypertensive medication.

Table 1.

Participant characteristics of the Pittsburgh Healthy Heart Study

	Baseline (N=375)	Follow-up (N=237)
	Mean ± SD or N (%)	Mean ± SD or N (%)
Age:	60.48 ± 4.78	66.31 ± 4.65
Sex: Women	189 (50.40)	123 (51.90)
Race: Black	58 (15.50)	31 (13.10)
White	312 (83.20)	200 (84.40)
Other	5 (1.30)	6 (2.60)
Education: HS or Less	89 (23.70)	52 (21.90)
Some College	98 (26.10)	55 (23.20)
College	95 (25.30)	66 (27.80)
Master’s +	93 (24.80)	64 (27.00)

Note: No significant differences between baseline and follow-up in percentage of women (t-test), racial demographic (chi-square test), or education (chi-square test) at α = .05.

Procedures

The baseline study (N_baseline=375; N_follow-up=237), consisted of 10 visits over the course of a 5-month period, completed September 1998-October 2000. An initial visit included informed consent, medical screening, and demographic and health behavior information collection. Most relevant to the present study, approximately 1 month after the initial visit, participants were trained to use an automated ABP monitor (Accutracker DX; SunTech Medical, Raleigh, NC) and a self-report electronic diary (Palm Pilot Professional; Palm, Inc, Santa Clara, CA). This was followed by a 3-day ambulatory monitoring period and four visits associated with a second 3-day ambulatory monitoring period four months later (32). A six-year follow-up consisted of six visits, completed from February 2005-July 2006, including one medical history visit and three visits related to a third 3-day period of ambulatory monitoring. Participants were paid $200 for the baseline procedures and $350 for the follow-up. This study was conducted in compliance with the University of Pittsburgh Institutional Review Board.

Ambulatory Blood Pressure

At baseline, participants initially completed 3 days of ABP monitoring, with the Accutracker DX collecting data every 45 minutes during waking hours. Approximately 4 months later, participants were retrained and completed an additional 3 days of monitoring (31).

After each ABP assessment, an electronic diary (Palm Pilot Professional) administered a 45-item self-report questionnaire adapted from the Diary of Ambulatory Behavioral States (DABS) (32). Questionnaire ratings were timestamped (using software from Invivodata, Inc, Scotts Valley, CA) to monitor compliance. Data from each EMA assessment was considered matched to the ABP assessment if the questionnaire was completed within 30 minutes of the ABP assessment; the average time lag between ABP and EMA was small: at baseline, mean (SD)= 5.4 (3.25) minutes and at follow-up, mean (SD)= 2.32(2.19). As part of the DABS, participants were asked about BP-relevant behaviors, including temperature comfort, speech, and posture during BP assessment; physical activity within 10 minutes prior to assessment; and food, caffeine, alcohol, and drug consumption within 30 minutes of assessment. Participants provided a mean of 101 (range: 19-148) observations with both ABP and electronic diary data across the two baseline monitoring periods.

At the 6-year follow-up, participants completed an additional 3 days of ambulatory monitoring, using the same ABP unit and a similar self-report electronic diary (Palm Z22). The protocol was identical to baseline, with the addition of overnight ABP collected on two of the three monitoring days. Both daytime ABP and electronic diary data were available for 235 participants. Participants provided a mean of 45 (range: 5-76) observations with both ABP and electronic diary data at the 6-year follow-up.

Task Demand and Task Strain

Each DABS assessment included a three-item Task Demand scale (i.e., “required working hard?,” “required working fast?,” and “juggling several tasks at once?”) and a two-item Decisional Control scale (i.e., “could change activity if you chose to?,” and “choice in scheduling this activity?”). These items were derived from larger scales in the Job Content Questionnaire (34) but as part of the DABS, they were designed to reference all activities “during the past 10 minutes” both in and out of the workplace. Participants responded to the items using a visual analog scale (VAS) anchored at NO=1 to YES=11 at baseline and NO=1 to YES=10 at follow-up. A cursor was located at the center of each VAS denoting moderate agreement to the item in each case. To make the scales from the two time points equivalent, we subtracted 1 from all baseline scores and multiplied by .9, and we also subtracted 1 from scores at follow-up, resulting in a comparable range of 0=NO to 9=YES for both periods. Values from respective items were averaged to produce scale scores for Task Demand and Decisional Control in each case.

For each wave of data collection, observations where scale scores were greater than the sample median on Task Demand and less than or equal to the sample median on Decisional Control for the given wave were marked as high Task Strain (1=task strain, 0=not task strain). The use of this dichotomous variable was based on historical precedent for studies of job strain (35), demonstrated associations with ABP (36), and a statistical interest in only the particular quadrant of the demand/control interaction (37). While we traditionally assess Task Strain, there is also evidence that Task Demand on its own is a stimulus for stress-related cardiovascular reactivity (38,39). Thus, Task Strain as a binary variable and Task Demand as a continuous variable were examined as alternative indicators of task-based stress.

Social Conflict

In each electronic diary assessment at baseline and follow-up, participants were asked about their most recent social interaction. If they indicated the social interaction occurred within the past 10 minutes, their data was included in the present analyses. Participants then answered three follow-up questions (“someone treated you badly?”, “someone interfered with your efforts?”, and “someone in conflict with you?”) using the visual analog scales, anchored at NO and YES, described above. As above, scores ranging from 0 to 9 were derived for each item and were averaged across items, yielding a social conflict score for each observation.

Negative Affect

During each electronic diary assessment at baseline and follow-up, participants were also administered questions about their current mood. Negative Affect was assessed using 3 items (i.e., “How feeling? Sad?”, “Frustrated/Angry?”, “Nervous/Stressed?”) to which participants responded using the visual analogue scale, anchored at NO and YES, described above. As above, corrected item scores ranged from 0 to 9, and were averaged across items for each observation as a measure of momentary negative affect. Chronbach alpha= .81, indicating that the scale is reliable; test-retest reliability across 6-years = .44, as might be expected because this NA scale was not intended as a trait measure, but rather is sensitive to momentary and daily situational effects. When assessing sensitivity to within-person, across wave change, R_kf=.64 indicates reliability of the negative affect measure across both waves, and R_c=.57 indicates adequate reliability of change in negative affect across waves (40). Note that recent arguments suggest standard cutoffs should be relaxed for within-person reliability, especially in the case of EMA designs when there are fewer items for reduced participant burden (41)

Analysis

Computing BP Reactivity

BP reactivity was operationalized as the change in ambulatory SBP or DBP that accompanies the presence of high task strain (as indicated by either the binary strain or continuous demand measures) or high social conflict. The effect of task strain or social conflict on BP is modeled with multilevel models (42) to account for the nested data structure (i.e., moments within data collection waves within individuals). For example, in the case of task strain, ABP for moment i, wave j, and person k was modeled as:

$${\text{Blood}\_\text{Pressure}}_{\text{ijk}}=\left({\beta }_{0\text{j}}+{\text{u}}_{0\text{k}}+{\text{v}}_{0\text{jk}}\right)+\left({\beta }_{1\text{j}}+{\text{u}}_{1\text{k}}\right){\text{Task}\_\text{Strain}}_{\text{ijk}}+{\beta }_{2\text{j}}{\text{Mean}\_\text{Task}\_\text{Strain}}_{\text{jk}}+{\beta }_{3\text{j}}{\text{BaselineAge}}_{\text{k}}+{\beta }_{4\text{j}}{\text{BaselineAge}}_{\text{k}}\ast {\text{Task}\_\text{Strain}}_{\text{ijk}}+{\text{e}}_{\text{ijk}}$$

Following Sliwinski et al. (20), we grand-mean centered the task strain and social conflict values to maintain a consistent meaning of social conflict/task strain values across individuals and waves (baseline and follow-up). With this centering, the coefficients are interpreted as follows (42): β₀₀ is the fixed effect sample average for baseline BP (intercept); β₀₁ is the fixed effect sample average change in BP intercept at follow-up. Note that a dummy variable for wave (0=baseline, 1=follow-up) was the means for achieving wave-specific parameters (e.g., β₀₀ and β₀₁), so calculating change was part of the equation. u_0k is a random effect allowing for between-person differences in intercept, and v_0jk is a random effect allowing within-person, between-wave differences in intercept (i.e., individual differences in 6 year changes in ABP or NA); β₁₀ is the fixed effect sample average baseline ABP reactivity slope, indicating the extent to which BP changes in response to higher levels of task strain or social conflict at the momentary level; and u_1k is a random effect allowing for between-person differences in momentary reactivity. Of greatest interest for the present study, β₁₁ is the change in momentary reactivity from baseline to the follow-up period.

β₂₀ is the effect of between-person differences in wave-level stress (task strain or social conflict) on blood pressure at baseline, and β₂₁ is the change in this effect from baseline to follow-up. Adjusting for the wave-level effects allows us to distinguish momentary reactivity from blood pressure differences that may accompany chronic stress exposure. We also controlled for the unique effect of baseline age on BP and NA at each wave (baseline:β₃₀ and follow-up:β₃₁), and examined whether cross-sectional age moderated momentary reactivity at baseline (β₄₀) or changes in momentary reactivity from baseline to follow-up (β₄₁).

Effects of momentary confounds on ABP were controlled for at level 1 as fixed effects: dummy variables for whether the participant was standing, lying down, talking, the temperature was too cold or too hot at the time of assessment, whether the participant had engaged in light, moderate, or heavy physical activity, and whether the participant had eaten a meal or snack, or had consumed caffeine, alcohol, or drugs in the 30 minutes prior to ABP monitoring.

Negative affect reactivity was modeled equivalently, but without adjusting for momentary ABP confounds.

A spatial power covariance function was used to model auto-correlated errors over continuous time (i.e., time of ABP and electronic diary assessment). Random effects are assumed to be multivariate normally distributed with variances ${\sigma }_{u_{0k}}^2,{\sigma }_{u_{1k}}^2,\mathrm{}{\sigma }_{v_{0jk}}^2$ and covariance ${\sigma }_{u_{0k}{u}_{1k}}^{}$. Multilevel models were fit using SAS PROC MIXED with REML estimation and incomplete data treated as missing at random (43)SAS code available).

Sensitivity Analyses

Two sensitivity analyses were performed to check assumptions made in the above models. First, we used the same multilevel model described above to test whether there were significant differences in BP reactivity or NA reactivity between the two baseline monitoring periods: the 3-day initial monitoring period and the 4 month 3-day monitoring period to justify our assumption that both periods could be combined. Second, we tested the original BP models with antihypertensive medication use as a covariate and as a moderator of stress-related reactivity. This allowed us to check whether the use of antihypertensive medications by 30% of the sample at the 6-year follow-up altered our findings.

RESULTS

Cross-sectional Associations

Cross-Sectional Descriptive Statistics

Table 2 includes descriptive statistics and cross-sectional correlations for variables of interest at baseline and follow-up, using person-level average values. Mean ambulatory SBP and DBP were highly correlated at each wave, though slightly less so at follow-up (r_baseline=.67, p<.001; r_follow-up=.55, p<.001). Both ambulatory SBP and DBP were slightly positively associated with average momentary task strain at baseline, but not at follow-up (SBP: r_baseline=.11, p=.031; r_follow-up=.08, p=.20; DBP: r_baseline=.13, p=.010; r_follow-up=.10, p=.13), while only SBP was related to task demand at baseline (SBP: r_baseline =.10, p=.048) (29). Average ambulatory SBP and DBP were not related to average momentary social conflict at baseline (r_baseline=.04, p=.40; DBP: r_baseline=.06, p=.22), while SBP was slightly positively related to social conflict at follow-up (SBP: r_follow-up=.12, p=.069; DBP: r_follow-up=.06, p=.36). These descriptive correlations suggest some associations between mean ABP and stressor exposure between persons.

Table 2.

Descriptive Statistics and Pearson Correlations of Key Study Variables, using person-level average values.

	Baseline (N=375)		Follow-up (N=237)		Correlations
	Mean	(SD)	Mean	(SD)	SBP	DBP	Negative Affect	Social Conflict	Task Strain	Task Demand	Age
SBP	128.74	(12.26)	137.41*	(15.44)		.67*	.07	.04	.11*	.10*	.08
DBP	78.71	(7.20)	77.11*	(7.81)	.55*		.12*	.06	.13*	.10	−.13*
Negative Affect	2.16	(1.12)	0.89*	(0.93)	.08	.06		.70*	.53*	.52*	−.23*
Social Conflict	1.45	(1.00)	0.38*	(0.48)	.12	.06	.53*		.49*	.39*	−.21*
Task Strain	0.26	(0.25)	0.19*	(0.24)	.08	.10	.42*	.46*		.70*	−.20*
Task Demand	2.96	(1.25)	1.72*	(1.27)	.13	.08	.40*	.40*	.82*		−.22*
Baseline Age	60.01	(4.78)	60.31	(4.65)	.001	−.25*	−.17*	−.15*	−.22*	−.18*

Note:

^*denotes significant at α<.05 (in “follow-up mean” column, this indicates significant difference from baseline). Top triangle of correlation matrix is from the baseline data, bottom triangle is from the follow-up. Note: SBP=Systolic Blood Pressure; DBP=Diastolic Blood Pressure.

Mean momentary NA was very strongly associated with social conflict at baseline and strongly at follow-up (r_baseline=.70, p<.001; r_follow-up=.53, p<.001). Similarly, mean NA was positively associated with task strain, with a stronger association at baseline than follow-up (r_baseline=.53, p<.001; r_follow-up=.42, p<.001). These descriptive correlations suggest NA associations with stressor exposure between persons.

Cross-sectional age was not associated with SBP (r_baseline=.08, p=.11; r_follow-up=.001, p=.98), but it was negatively associated with both average ambulatory DBP (r_baseline=−.13, p=.015; r_follow-up=−.25, p<.001) and average momentary NA (r_baseline=−.23, p< .001; r_follow-up=−.17, p=.009). Baseline age was also negatively associated with both social conflict (r_baseline=−.21, p<.001; r_follow-up=−.15, p=.019) and task strain (r_baseline=−.20, p< .001; r_follow-up=−.22, p=.001).

Cross-Sectional Age Differences: Results from Multilevel Model

As suggested in the descriptive statistics above, results from the multilevel models found that cross-sectionally, older adults experienced lower levels of DBP and NA at baseline (and decreasing levels of DBP than younger participants (see Table 3 and 4: β_30;31). Cross-sectional (baseline) age was not associated with BP or NA reactivity to social conflict at baseline, or change in these reactivity slopes at follow-up (Table 3: β_40;41, all p>.05). Similarly, cross-sectional age was not associated with BP reactivity to task strain or task demand, or change in these reactivity slopes at follow-up (Table 3 columns 1, 2, 4, and 5: β_40;41, all p>.05). However, older adults at baseline had slightly lower NA reactivity to task demand at baseline, and slightly greater increases in NA reactivity to task strain and task demand (Table 4: Task Strain: β₄₁=0.04, p<.001; Task Demand: β₄₁= 0.01, p<.001). In sum, cross-sectionally, older adults experienced lower DBP and NA levels, no differences in BP reactivity to stress, but slightly greater increases in NA reactivity to task demand over time.

Table 3.

Multilevel models of 6-year change in systolic blood pressure (SBP) reactivity, diastolic blood pressure (DBP) reactivity, and negative affect reactivity to momentary social conflict.

	SBP		DBP		Negative Affect
	Est.	(SE)	Est.	(SE)	Est.	(SE)
Fixed Effects
Intercept at Baseline, β00	121.60*	(1.30)	75.40*	(0.80)	2.07*	(0.04)
Change in Intercept β01	11.31*	(1.83)	−1.18	(0.89)	−0.21*	(0.11)
Baseline SC (Reactivity) β10	0.43*	(0.13)	0.08	(0.08)	0.33*	(0.01)
Change in SC Reactivity β11	0.21	(0.25)	0.48*	(0.18)	0.43*	(0.02)
Baseline Wave Mean of SC β20	0.42	(0.68)	−0.09	(0.35)	0.47*	(0.04)
Change in Wave Mean of SC β21	2.59	(1.78)	−0.10	(0.92)	0.57*	(0.11)
Cross-Sectional Age at Baseline β30	0.26	(0.15)	−0.19*	(0.08)	−0.02*	(0.01)
Change in Cross-Sectional Age Effect β31	−0.20	(0.21)	−0.33*	(0.11)	0.02	(0.01)
Cross-Sectional Age x Baseline SC Reactivity β40	0.04	(0.03)	0.02	(0.02)	0.003	(0.002)
Cross-Sectional Age x Change in SC Reactivity β41	0.04	(0.05)	−0.05	(0.04)	0.007	(0.004)
Random Effects
Level 3: σ2intercept	68.74*	(11.85)	31.63*	(3.79)	0.23*	(0.05)
σintercept, reactivity slope	−0.75	(1.32)	−0.32	(0.55)	0.01	(0.01)
σ2reactivity slope	0.88*	(0.30)	0.51*	(0.14)	0.03*	(0.003)
Level 2: σ2intercept	96.35*	(9.76)	18.11*	(2.13)	0.34*	(0.04)
Level 1: σ2residual	171.12*	(1.86)	74.17*	(0.85)	0.60*	(0.01)
σ2residual: spatial power	0.91*	(0.01)	0.88*	(0.01)	0.94*	(0.004)
Model Fit (AIC)	240843.4		214938.2		82534.9

Note: SC=Social Conflict. AIC=Akaike Information Criteria. N_baseline=375 and N_follow-up=237. Covariates for BP models include dummy variables for whether the participant was standing, sitting, lying down, and talking at the time of the BP reading, whether the temperature was too hot or too cold, whether the participant had engaged in light, moderate, or heavy physical activity in the 10 minutes prior to the reading, and whether the participant had consumed a meal, snacks, caffeine, alcohol, or drugs in the half hour before the reading.

Table 4.

6-year change in systolic blood pressure (SBP) reactivity, diastolic blood pressure (DBP) reactivity, and negative affect reactivity to momentary strain.

	Task Strain						Task Demand
	Systolic BP		DBP		Negative Affect		SBP		Diastolic BP		Negative Affect
	Est.	(SE)	Est.	(SE)	Est.	(SE)	Est.	(SE)	Est.	(SE)	Est.	(SE)
Fixed Effects
Intercept at Baseline, β00	120.97*	(1.08)	73.94*	(0.65)	2.13*	(0.05)	120.77*	(1.09)	73.95*	(0.65)	2.05*	0.05
Change in Intercept β01	9.24*	(0.93)	−1.30*	(0.43)	−1.12*	(0.06)	10.36*	(1.14)	−0.98	(0.55)	−0.79*	0.08
Baseline TS (Reactivity) β10	1.15*	(0.27)	1.00*	(0.17)	0.40*	(0.02)	0.43*	(0.07)	0.50*	(0.05)	0.11*	0.01
Change in TS Reactivity β11	1.29*	(0.48)	0.54	(0.32)	0.08*	(0.03)	0.23*	(0.11)	0.19*	(0.07)	0.01	0.01
Baseline Wave Mean of TS β20	3.32	(2.70)	0.33	(1.37)	1.97*	(0.19)	0.34	(0.55)	−0.44	(0.28)	0.34*	0.04
Change in Wave Mean of TS β21	3.65	(3.59)	−0.54	(0.32)	−0.96*	(0.29)	0.31	(0.79)	0.24	(0.38)	−0.15*	0.05
Cross-Sectional Age at Baseline β30	0.24	(0.15)	−0.19*	(0.08)	−0.03*	(0.01)	0.24	(0.15)	−0.19*	(0.08)	−0.03*	0.01
Change in Cross-Sectional Age Effect β31	−0.20	(0.20)	−0.34*	(0.10)	−0.02	(0.01)	0.05	(0.19)	−0.36*	(0.10)	−0.02	0.01
Cross-Sectional Age * Baseline TS Reactivity β40	0.05	(0.05)	0.04	(0.03)	−0.01	(0.01)	0.01	(0.02)	0.01	(0.01)	−0.004*	0.001
Cross-Sectional Age * Change in TS Reactivity β41	0.14	(0.10)	−0.001	(0.07)	0.04*	(0.01)	0.01	(0.02)	−0.02	(0.01)	0.01*	0.001
Random Effects
Level 3: σ2intercept	72.76*	(11.87)	31.05*	(3.67)	0.26*	(0.06)	74.00*	(11.98)	31.42*	(3.73)	0.28*	0.06
σintercept, reactivity slope	−3.72	(2.96)	−1.96	(1.03)	0.03	(0.02)	0.29	(0.79)	−0.52	(0.28)	0.02*	0.004
σ2reactivity slope	6.54*	(1.53)	2.06*	(0.57)	0.13*	(0.02)	0.77*	(0.12)	0.27*	(0.05)	0.01*	0.001
Level 2: σ2intercept	99.64*	(9.52)	19.68*	(2.06)	0.48*	(0.05)	99.32*	(9.51)	19.71*	(2.09)	0.45*	0.04
Level 1: σ2residual	170.39*	(1.34)	71.28*	(0.59)	0.58*	(0.005)	169.85*	(1.34)	70.89*	(0.59)	0.58*	0.005
σ2residual: spatial power	0.91*	(0.005)	0.87*	(0.01)	0.94*	(0.003)	0.91*	(0.005)	0.87*	(0.01)	0.95*	0.003
Model Fit (AIC)	397881.8		354265.5		134085.8		397715.0		353966.5		133230.5

Notes: TS=Task Stress, AIC=Akaike Information Criteria. N_baseline=375 and N_follow-up=237. Covariates for BP models include dummy variables for whether the participant was standing, sitting, or lying down, talking at the time of the BP reading, whether the temperature was too hot or too cold, whether the participant had engaged in light, moderate, or heavy physical activity in the 10 minutes prior to reading, and whether they had eaten a meal, snacks, caffeine, alcohol, or drugs in the half hour before the reading

Changes in Physiological and Affect Reactivity with Age

Table 3 displays six year changes in SBP and DBP reactivity and NA reactivity to momentary social conflict, and Table 5 displays model derived estimates of rest and reactivity changes from baseline to follow-up. Baseline SBP reactivity to social conflict was 0.43 (p<.001; i.e., ambulatory SBP increased by 0.43 mmHG with a one unit increase in social conflict), and it did not significantly change from baseline to follow-up (β₁₁=0.21, p=.41). There was no significant DBP reactivity to social conflict on average at baseline (β₁₀=0.08, p=.32), but DBP reactivity increased by 0.48 from baseline to follow-up (p=.007; i.e., at follow-up, ambulatory DBP increased by 0.56 mmHG with a one unit increase in social conflict). NA increased by .33 units (p<.001) with a one-unit increase in social conflict at baseline, and NA reactivity to social conflict increased from baseline to follow-up by .43 units per one unit increase in social conflict (β₁₁=0.10, p<.001). In sum, DBP and NA reactivity to social conflict increased over six years, while SBP reactivity to social conflict did not change.

Table 5.

Model derived stress-based reactivity and rest (low stress) values of Systolic Blood Pressure, Diastolic Blood Pressure, and Negative Affect.

		Social Conflict	Task Strain	Task Demand
Systolic Blood Pressure	Rest at Baseline	121.17	120.97	120.34
	Rest at Follow-up	132.26	130.21	130.47
	Difference in Rest Values from Baseline to Follow-up	11.10	9.24	10.13
	Reactivity at Baseline	0.87	1.15	0.87
	Reactivity at Follow-up	1.29	2.44	1.33
	Difference in Reactivity from Baseline to Follow-up	0.42	1.29	0.46
Diastolic Blood Pressure	Rest at Baseline	75.31	73.94	73.46
	Rest at Follow-up	73.66	72.64	72.28
	Difference in Rest Values from Baseline to Follow-up	−1.66	−1.30	−1.17
	Reactivity at Baseline	0.17	1.00	0.99
	Reactivity at Follow-up	1.13	1.54	1.38
	Difference in Reactivity from Baseline to Follow-up	0.96	0.54	0.39
Negative Affect	Rest at Baseline	1.74	2.13	1.94
	Rest at Follow-up	0.95	1.02	1.13
	Difference in Rest Values from Baseline to Follow-up	−0.80	−1.12	−0.80
	Reactivity at Baseline	0.66	0.40	0.23
	Reactivity at Follow-up	1.80	0.48	0.25
	Difference in Reactivity from Baseline to Follow-up	1.14	0.08	0.02

Note: “Rest” values are estimated at 1 unit below the grand mean for social conflict and task demand and at 0 for task strain, reactivity is the difference in outcome associated with an increase in social conflict or task demand from 1 unit below the mean to 1 unit above the mean (columns 1 and 3) and from absence (0) to presence (1) of task strain (column 2)

Table 4 presents similar data for task strain and task demand. Baseline SBP reactivity to task strain was 1.15 (i.e., ambulatory SBP increases by 1.15 mmHg in the presence of task strain, p< .001), and this increased from baseline to follow-up (β₁₁=1.29, p=.007). Similarly, baseline SBP reactivity to task demand was 0.43 (p< .001), and this increased at follow-up by .23 (p=.032 i.e., at follow-up, ambulatory SBP increases by 0.66 mmHg per one-unit increase in task demand). Baseline DBP reactivity to task strain was 1.00 (p< .001), and increased with marginal significance at follow-up (β₁₁=0.54, p=.096). Baseline DBP reactivity to task demand was 0.50 (p< .001), and increased longitudinally (β₁₁=0.19, p=.005). NA reactivity to task strain was 0.40 (p< .001) at baseline, and it increased from baseline to follow-up (β₁₁=0.08, p=.016). NA reactivity to task demand was 0.11 (p<.001) at baseline and increased with marginal significance longitudinally (β₁₁=0.01, p=.117). In sum, NA reactivity, SBP reactivity, and DBP reactivity to task-based stress generally increased over six years.

Sensitivity Analyses

We found no significant differences in BP reactivity or NA reactivity between the baseline and the 4 month follow-up data (all ps>.05, data not shown), which justified our decision to combine them as one “baseline” measure. Analyses testing whether antihypertensive medication use affected results found that while those who take BP meds at follow-up had higher mean BP at baseline, reactivity and change in reactivity was not affected by BP medication use (data not shown).

DISCUSSION

The present study was the first to longitudinally examine long-term age-based changes in both BP reactivity and NA reactivity to stressors in daily life. Using intensive longitudinal data collected 6-years apart, we found that advancing age during midlife to young-old age is related to increased NA reactivity (replicated across social conflict and task strain), increased DBP reactivity (replicated across social conflict and task demand), and increased SBP reactivity to task-based stress. These data suggest an increased vulnerability to momentary stressors with age.

This study extended previous work in this area that has been primarily lab-based and has involved primarily cross-sectional comparisons. With two common and impactful natural stressors—social conflict and task strain—the findings show that increases in BP reactivity with age extend to natural experiences. Longitudinal DBP reactivity increases with age, replicating Uchino et al.’s (15) cross-sectional findings, and suggesting age-based findings are not due to cohort or selection effects that may be reflected in group comparisons of older and younger adults. Unlike Uchino et al.’s (14) short-term longitudinal, lab-based study, cross-sectional age was not associated with BP reactivity or change in BP reactivity in this sample. Lack of cross-sectional replication may have been due to the more restricted age-range in the present study’s sample compared to the Uchino et al. (14) sample. Additional statistical controls for average stressor experiences across each person and wave further ensured that period effects did not bias reactivity estimates. Methodological advancements in the present study add evidence that the patterns of stressor reactivity change over time and are similar across diastolic blood pressure and NA.

Selection, Optimization, and Compensation theory posits that both gains and losses accompany development, with the ratio progressively favoring losses with advancing age (2). Indeed, the present findings show that despite mean level affect improvement with age, affective and BP reactivity increase during momentary stress. Further, cross-sectional age in the present study was related to greater NA reactivity to task-based stress, corroborating findings between age-differences and age-change. That NA reactivity increase with age is seemingly contrary to current psychological theories of aging which suggest that adults respond less negatively as they age (44). Previous findings do offer some clues to understand this seeming contradiction. Namely, in contexts in which avoidance techniques are less applicable (e.g., when interpersonal problems turn into arguments), NA reactivity benefits for older adults are not apparent (3,25). The present findings may reflect such contexts. Demand and control perception also changes with age (45). Older adults may have higher thresholds for what constitutes strain after employing attentional and avoidant strategies. The scenarios that characterize task strain could potentially reflect stressors that are less avoidable with age, that is, stressors to which older adults are more vulnerable (3). Because operational definitions of stress in diary reports involve an appraisal of event as stressful, these events may be more salient across age than laboratory stressors (20).

The increase in BP reactivity with age is consistent with prevailing cardiovascular aging theories that increased arterial stiffness, systemic vascular compliance, and impaired baroreflexes lead to greater BP fluctuations in late midlife and older age(46). However, in the midlife to young old sample, the present study found that aging is associated with increases in DBP across stressor types but only increased SBP reactivity to task-based stress. This is contrary to previous lab-based findings, which report stronger differences in SBP, but is consistent with Uchino et al.’s cross-sectional examination of reactivity to momentary hassles/problems (15). Several possible reasons arise for this discrepancy. First, it is possible that older adults’ coping strategies are different when facing stressors in natural contexts than they are when facing artificial stressors in the lab (9). Different types of coping orientations have been associated with different patterns of cardiovascular response, with beta-adrenergic activity (cardiac responses) associated with active coping efforts and alpha-adrenergic activity (vascular responses) with passive coping (47,48). Aging processes may be associated with the use of more passive coping efforts, especially in cases of social conflict, due to loss of available resources for active coping as well as improved emotion-regulation abilities from life experience (3). If participants favor more passive coping as they age, it is possible that the stressors they face may thereby elicit stronger vascular responses, accounting for the observed increase in DBP reactivity. Similarly, there is evidence that cardiac and vascular dimensions of cardiovascular reactivity can vary as a function of variations in the provoking stimulus (49). It is possible that the types of stressors examined in this study have more vascular relevance than active coping stressors that frequently characterize laboratory studies. Corroborating current findings with other stressor types will help future research understand age-relevant changes in BP reactivity and their underlying mechanisms.

Several limitations of the present study can be addressed in future work. In the analytical approach, the present study separately examined reactivity to two types of stressors to confirm similar patterns of effects across stressor types. This approach could be further expanded to address other types of stressors. However, approaches that highlight stressor reactivity as a personal attribute would suggest aggregating across these and more types of stressors to obtain a more reliable between-person variable (50). Nevertheless, the use of multiple items for each scale, across multiple occasions, helps generalize the application of each reactivity score, and furthers understanding of the limits of reactivity as a between-person indicator. Related to such scales, there may be shared method variance in the self-reported stressors and NA, all items of which are measured on the same visual analog scale. Study design limitations also included the sampling design, with two waves of measures of participants 50-70 at baseline. Due to modeling constraints with the two waves of data, we did not allow for random effects of within-person, between-wave reactivity slopes. While the present study tested for fixed effects in reactivity change over time, future work is needed to account for individual differences in such effect and whether those individual differences may be important for predicting future health and well-being. Further work is also needed to examine reactivity change patterns in old-old and oldest-old populations.

Future work is also need to determine whether the similar patterns of change in NA and diastolic blood pressure reactivity to stress reflect underlying age-based vulnerabilities, stressor experience or psychosocial resource change, or are not due to the same underlying processes at all. Direct measurement of peripheral resistance and cardiac output changes in the natural environment would help clarify some of the mechanisms which may be causing age-based changes in BP reactivity, but would require more sophisticated ambulatory assessments (51). Relatedly, direct measurement of coping techniques (e.g., re-appraisal, distraction) and cognitive appraisals (e.g., degree of threat) may clarify emotion-regulation strategies that result in differing NA responses. Beyond appraisal and coping, several person, environmental, or person-by-environment interactions are relevant to understanding the present results. Though the present study could not assess this, it is also possible that threatening characteristics of the stressors, themselves, may change with age (52). For example, in long-established relationships, daily social tension might be perceived as less threatening to how others view the individual, but more threatening to the individual’s health. Future work must also assess whether there are qualitative differences in stressors faced over time, despite similar quantitative ratings. For example, task strain after retirement may differ in content and consequences than task strain before retirement. While stressor exposure generally decreases with age, there is evidence that the qualitative components (such as stressor types) changes over time (53).

The present study provided the first long-term longitudinal evidence of convergence between increases in BP and NA reactivity with age. While age may be associated with reductions in NA, these benefits do not translate into improvements in NA or DBP reactivity in the context of everyday stressors. Future work will be needed to uncover the mechanisms accounting for these discrepancies and differences in DBP and SBP reactivity to everyday stressors. Future work is also needed to determine whether these increases in NA and BP reactivity at the momentary level are similarly reflected over longer time-scales such as hours or days, as post-event recall of stressful experiences after longer time might be less negative for older adults (17). Indeed, age-related benefits in coping strategies that extend over longer time spans may moderate downstream health consequences of increases in momentary stress-related cardiovascular and negative affect reactivity that appear to be associated with aging.

Abbreviations:

BP

Blood Pressure

SBP

Systolic Blood Pressure

DBP

Diastolic Blood Pressure

CVD

Cardiovascular Disease

EMA

Ecological Momentary Assessment

ABP

Ambulatory Blood Pressure

PHHP

Pittsburgh Healthy Heart Project

NA

Negative Affect