Mindfulness-Based Stress Reduction reduces pro-inflammatory gene regulation but not systemic inflammation among older adults: A randomized controlled trial

Abstract

Objective:

Aging is associated with increased pro-inflammatory gene expression and systemic inflammation, and psychosocial stress may accelerate these changes. Mindfulness interventions show promise for reducing psychosocial stress and extending healthspan. Inflammatory pathways may play a role. In a sample of lonely older adults, we tested whether mindfulness training reduces proinflammatory gene expression and protein markers of systemic inflammation.

Methods:

Lonely older adults (65–85 years; N=190) were randomly assigned to an 8-week Mindfulness-Based Stress Reduction (MBSR) or matched Health Enhancement Program (HEP). Blood was drawn pre- and post-intervention and at 3-month follow-up. In peripheral blood mononuclear cells (PBMCs), RNA profiling was used to assess transcriptional regulation by pro-inflammatory NF-kB as well as β-adrenergic CREB, antiviral IRF, and glucocorticoid receptor (GR) transcription factors. Plasma was assayed for proinflammatory markers IL-6 and CRP. Analyses tested time (pre, post, follow-up) by condition (MBSR versus HEP) effects.

Results:

MBSR reduced NF-kB (d=.17, p=.028) but did not alter CREB (d=.10, p=.20), IRF (d=.13, p=.086), or GR activity (d=.14, p=.063) relative to HEP over time. Contrary to predictions, there were no time × condition effects of MBSR compared to HEP on reducing circulating IL-6 or CRP.

Conclusions:

In lonely older adults, MBSR reduced cellular pro-inflammatory gene regulation in ways that would predict reduced disease risk. However, no similar effect was observed for circulating protein markers of inflammation. These results provide specificity about how mindfulness interventions may impact distinct inflammatory markers among aging adults in ways that may have important implications for healthspan.

Trial Registration:

Clinical Trials identifier NCT02888600

Although life expectancy has increased dramatically in the past 100 years, human healthspan has not kept pace (1), with the majority of US adults living with one or more chronic diseases of aging (2). This disparity presents important questions about how to slow the process of biological aging, in turn reducing risk for chronic disease and extending healthspan—years lived without disability (3). A multitude of interrelated physiological processes deteriorate with age, contributing to a systemic pro-inflammatory state that predicts morbidity and disability (4). Importantly, psychosocial factors like loneliness predict inflammation and thus may accelerate aging (5). Interventions that target psychosocial factors related to this process of ‘inflammaging’ (6) may reduce chronic disease risk and extend healthspan (1). Mindfulness training interventions show promise for reducing loneliness (7,8) and improving health-relevant outcomes (9). Here we test whether an 8-week mindfulness (vs. health promotion) intervention can improve inflammatory processes among lonely but otherwise healthy older adults.

Loneliness and other chronic psychosocial threats are thought to activate intracellular signaling pathways that contribute to a chronic inflammatory state. In particular, loneliness has been associated with upregulated cellular expression of proinflammatory genes transcribed by nuclear factor kappa B (NF-κB). Loneliness also appears to contribute to a broader conserved transcriptional response to adversity (CTRA) characterized by upregulated expression of β-adrenergic signaling genes transcribed by cAMP response element-binding protein (CREB) and downregulated expression of antiviral genes regulated by interferon regulatory factors (IRF) and anti-inflammatory glucocorticoid receptor (GR) genes (10,11). Loneliness has also been linked with plasma protein markers of systemic inflammation, such as IL-6 and CRP (12-14). Although both associate with chronic inflammatory disease risk and biological aging, cellular RNA profiles and plasma protein markers of inflammation represent distinct aspects of inflammatory biology and have distinct impacts on disease. Cellular RNA primes circulating immune cells to respond to damaged tissue, and these circulating cells actively traffic into tumors, atherosclerotic plaques, etc. to produce a local response. Although this response is adaptive in the context of disease, chronic activation of inflammatory genes in response to psychosocial threat can contribute to chronic disease (15). In contrast, although pro-inflammatory cytokines are released by circulating immune cells in response to localized damage, the plasma inflammatory proteins measured here are primarily produced by adipose, liver, and spleen tissues and passively diffuse throughout the vascular space. As such, while circulating inflammatory proteins are markers of disease risk, cellular RNA profiles are more directly linked with disease activity.

Mindfulness meditation interventions, which train people to monitor their present-moment sensory experiences through an open and accepting perspective (16), may decrease psychosocial stress and thus reduce inflammatory biology. By welcoming all emotional experiences—even distressing ones—to arise and pass without elaboration, mindfulness practice is thought to moderate psychosocial stress response pathways in ways that reduce inflammation and chronic disease risk (17). Indeed, mindfulness interventions have been shown to reduce feelings of loneliness (7,8) and mitigate physiological reactivity to social threat (18). Mindfulness interventions have also been shown to reduce pro-inflammatory gene expression in high-stress samples: stressed employees (19), medical residents (20), and breast cancer survivors (21) (for reviews, see (15,22-24)). And, although not all evidence is consistent, several studies have found that mindfulness interventions reduce plasma inflammatory markers in high-risk samples (25,26), including stressed (27), older (28,29), and overweight (28,30) adults, those with higher baseline inflammation (31,32), and breast cancer patients (21,33,34). Notably, in our previous pilot work, an 8-week Mindfulness-Based Stress Reduction (MBSR) intervention (vs. waitlist control) reduced pro-inflammatory NF-κB gene expression and yielded a trend toward lower CRP in a small sample of lonely older adults (8). Although expression of inflammatory genes is known to trigger the acute release of inflammatory cytokines into circulation, peripheral inflammation is primarily mediated by other factors (e.g., adiposity (35); diet (36); sleep disturbance (37)). As such, changes in cellular gene expression and changes in plasma inflammatory markers represent two distinct pathways through which mind-body interventions may influence health.

Lonely older adults are an ideal population to test hypotheses in geroscience. Can biological aging be slowed or reversed? Aging itself is a primary risk factor for systemic inflammation and chronic disease, and loneliness is thought to exacerbate these age-related outcomes (38). Intervening to target this psychosocial risk factor in a healthy but at-risk population may provide insight as to whether psychological interventions can slow processes underlying biological aging (3). We previously showed in the present sample of lonely older adults that MBSR improves innate immunocompetence marked by increased release of IL-6 from immune cells stimulated with endotoxin (39) while preserving the sensitivity of immune cells to the anti-inflammatory effects of glucocorticoids (40). Whether this pattern of effects extends to allow for efficient recovery from an acute inflammatory response back to initial levels of inflammation—thus helping to protect against chronic systemic inflammation—is unknown. However, interventions that allow older adults to maintain optimal immune responding while protecting against chronic systemic inflammation and inflammatory disease may extend healthspan (1).

The primary aims of this efficacy trial are to evaluate the effects of MBSR (vs. a Health Enhancement Program [HEP]) on biomarkers of inflammatory disease risk among lonely older adults. Proinflammatory gene expression (indexed by transcriptional indicators of pro-inflammatory NF-κB activity) and circulating protein markers of inflammation (IL-6 and CRP) were assessed at pre-intervention, post-intervention, and 3-month follow-up in a sample of lonely older adults (N=190). Following the results of our pilot trial (8), MBSR was expected to reduce NF-κB transcription factor activity and plasma concentrations of IL-6 and CRP relative to HEP. Secondary analyses explored whether MBSR vs. HEP reduced β-adrenergic CREB and increased anti-viral IRF and anti-inflammatory GR transcription factor activity that characterize the CTRA (11). Associations between changes in plasma protein markers, proinflammatory gene expression, and loneliness and moderation by additional risk factors were explored in supplementary analyses.

Materials and Methods

Participants

Healthy, lonely older adults (N=190; 65-85 years) were enrolled and randomized in this two-arm parallel trial (see Table 1 for baseline characteristics and Figure S1, Supplemental Digital Content, for CONSORT flow chart). Participants were recruited from the greater Pittsburgh area through local senior organizations and research registries, outreach events at local organizations and senior housing, and newspaper, radio, bus, email, and mailed advertisements. Eligible participants were English speaking adults between the ages of 65 and 93 years; were not diagnosed with or being treated for a current health problem or chronic disease known to affect inflammatory biology; were not using medication affecting cardiovascular or immune system function, except blood pressure medications in the final two intervention cohorts (n=54; a criterion changed to increase enrollment); reported moderate to high levels of perceived loneliness (>=4 on the Short Form UCLA-R (41)); had no current substance abuse problem; were not diagnosed with or being treated for severe mental illness; spent no more than 90 minutes/week in regular mind-body practice (e.g., guided meditation, yoga, tai chi); had no significant cognitive impairment; and had no problems with attending study assessments.

Table 1.

Baseline characteristics of randomized participants.

	Full Sample	MBSR	HEP	Condition
Characteristic	(N=190)	(N=93)	(N=97)	Difference
Age in years	69.77 (0.31)	69.96 (0.45)	69.59 (0.43)	F(1,188)=0.35, p=.55
Sex				χ2(1)=0.53, p=.47
Female	149 (78.42%)	75 (80.65%)	74 (76.29%)
Male	41 (21.58%)	18 (19.35%)	23 (23.71%)
Race				χ2(3)=0.27, p=.97
American Indian/Alaska Native	0 (0.00%)	0 (0.00%)	0 (0.00%)
Asian	2 (1.05%)	1 (1.08%)	1 (1.03%)
Black/African American	22 (11.58%)	11 (11.83%)	11 (11.34%)
White/Caucasian	161 (84.74%)	78 (83.87%)	83 (85.57%)
Bi- or Multi-Racial	5 (2.63%)	3 (3.23%)	2 (2.06%)
Ethnicity					χ2(1)=2.58, p=.11
Hispanic or Latino	6 (3.16%)	1 (1.08%)	5 (5.15%)
Not Hispanic or Latino	184 (96.84%)	92 (98.92%)	92 (94.85%)
Education Level				χ2(8)=8.96, p=.35
No High School Diploma	3 (1.58%)	1 (1.08%)	2 (2.06%)
GED	2 (1.05%)	0 (0.00%)	2 (2.06%)
High School Diploma	9 (4.74%)	2 (2.15%)	7 (7.22%)
Technical Training	6 (3.16%)	2 (2.15%)	4 (4.12%)
Some College, no degree	23 (12.11%)	11 (11.83%)	12 (12.37%)
Associate Degree	6 (3.16%)	4 (4.30%)	2 (2.06%)
Bachelor’s Degree	57 (30.0%)	34 (36.56%)	23 (23.71%)
Master’s Degree	71 (37.37%)	33 (35.48%)	38 (39.18%)
MD, PhD, JD, PharmD	13 (6.84%)	6 (6.45%)	7 (7.22%)
Marital Status				χ2(4)=0.83, p=.93
Married/Living with partner	82 (43.16%)	39 (41.94%)	43 (44.33%)
Widowed	27 (14.21%)	14 (15.05%)	13 (13.40%)
Separated	5 (2.63%)	2 (2.15%)	3 (3.09%)
Divorced	43 (22.63%)	23 (24.73%)	20 (20.62%)
Single	33 (17.37%)	15 (16.13%)	18 (18.56%)
BMI a	27.94 (0.44)	27.41 (0.62)	28.44 (0.61)	F(1,188)=1.40, p=.24
Pre-intervention raw CRP (μg/mL)b	2.25 (0.25)	1.94 (0.36)	2.54 (0.35)	χ2(1)=1.45, p=.229
Pre-intervention raw IL-6 (pg/mL)b	3.22 (0.19)	2.71 (0.27)	3.69 (0.27)	χ2(1)=6.38, p=.012
UCLA Loneliness c	42.39 (0.70)	43.09 (1.00)	41.72 (0.98)	F(1,187)=0.95, p=.33
Current mind-body practice d	21 (11.60%)	11 (12.79%)	10 (10.53%)	χ2(1)=0.23, p=.64

Note: Data reported as means and (SEs) or (%). ^aFor N=34, height used in BMI calculations estimated from averages for females and males. ^bN=187 (MBSR N=92; HEP N=95). ^cUCLA Loneliness data available for N=188 at baseline (MBSR N=92; HEP N=96). ^dInformation about pre-intervention mind-body practice (regular practice <90 minutes per week) available for N=181 at baseline (MBSR N=86; HEP N=95).

This trial was funded through NIH project R01AT008685, which aimed to test whether MBSR reduces loneliness and biomarkers of inflammatory disease risk in lonely older adults (pro-inflammatory gene expression, CRP, and IL-6). The trial was pre-registered with Clinical Trials identifier NCT02888600, with IL-6 and CRP registered as primary outcomes and pro-inflammatory gene expression and loneliness as secondary outcomes. Primary loneliness outcomes are reported by Dutcher et al. (42). Previous trial outcomes associated with NIH project F32AT009508 have been reported separately, showing that MBSR (vs. HEP) buffers increases in glucocorticoid receptor resistance (40) and increases LPS-stimulated IL-6 release, an index of innate immunocompetence (39). Study procedures were approved by the Carnegie Mellon University and University of Pittsburgh IRBs and all participants provided written informed consent. Study data were collected between October 2016 and February 2020 and are available by request. Trial recruitment concluded when recruitment goals had been reached. Intervention instructors were blind to outcome measures and laboratory technicians were blind to condition.

MBSR has shown a broad range of effect sizes for improving inflammatory pathways. Our pilot work comparing MBSR to no treatment in a lonely older adult sample showed medium effects on CRP (d=.58) (8) and proinflammatory NF-kB gene expression (ds=.68-.77), and MBSR has also shown large effects on reducing neurogenic inflammation compared to an active HEP intervention (d=2.8) (43). A priori power calculations conservatively estimated a small-medium effect (d=.3) of MBSR compared to HEP on primary trial outcomes. Assuming a d=.3 effect size and a pre-post correlation of r=.8, we planned to enroll N=188 participants to achieve acceptable power (86%).

Procedure

The study was run in eight intervention cohorts. Participants were pre-screened for eligibility by phone. At a pre-intervention study appointment, eligible participants completed questionnaire assessments. Participants attended a separate afternoon appointment for a non-fasting blood draw in the Behavioral Immunology Lab at University of Pittsburgh (PI: ALM). 30mL blood was drawn via venipuncture for later assessment of gene expression and IL-6 and CRP (see Measures). They then completed three days of ambulatory assessments and were randomly assigned to condition at the first intervention session. Participants were allocated 1:1 to MBSR or HEP interventions using a computerized random number generator, with procedures implemented separately for each of eight intervention cohorts. Randomization was blocked by age (≤75 vs. >75 years) and baseline depressive symptoms (Beck Depression Inventory-II(44) score ≤13 [minimal depression] vs. >13 [mild, moderate, or severe depression]) to ensure balance across groups. Allocation sequence was concealed; only author AGCW had access to the sequence and otherwise was not involved in data collection. At post-intervention, participants completed an afternoon blood draw, questionnaire assessment, and ambulatory assessment as at pre-intervention. Afternoon blood and questionnaire assessments were again collected at three-month follow-up. All blood draws were conducted between 1pm and 8pm, with 85% of draws between 3pm and 7pm. Blood draw times were standardized within participants (mean difference in blood draw time across three assessments = 8 minutes, SD = 52 minutes). Participants were debriefed and compensated up to $475, including stepped bonus payments for high adherence. Adherence was monitored throughout the study period and non-adherent participants were contacted to encourage engagement.

Materials

Intervention Programs

Participants were randomly assigned to MBSR or a structurally matched HEP (45). Both 8-week interventions are standardized curriculum-based group programs. In this study, the structure was adapted so that interventions consisted of 8 weekly 2-hour group sessions (rather than 2.5-hour sessions), a day-long retreat during the sixth week, and approximately 45 minutes of home practice assignments six days per week. Content covered in each class is summarized below. During the 3-month follow-up period, participants were encouraged to continue mindfulness (MBSR) or health (HEP) practices; they received a list of community resources for continued support, weekly practice reminder emails, and continued access to meditation audios (MBSR only).

The MBSR program was based upon the 2014 MBSR Curriculum Guidelines (Center for Mindfulness in Medicine, Health Care, and Society, University of Massachusetts Medical School). MBSR includes guided mindfulness meditations intended to foster awareness of present-moment experiences and an open, accepting, and nonjudgmental perspective. Guidance and group discussions also encourage nonjudgmental awareness in everyday life, including when experiencing stress or other challenging emotions. Group discussions involve exploration of habitual reactions to stress, and the cultivation of skills such as pausing before responding. The daylong retreat after Week 6 involved silent guided meditation practice and reflection. Foundationally, the course centers on self-care and group support, including support around meeting the challenge of integrating meditation practice into daily life. Home practice recordings guide participants through body awareness, mindful movement, and seated meditation.

The HEP program, which was originally developed to enhance health and to match the structure of MBSR without training mindfulness, followed guidelines outlined in MacCoon et al. (2012) (45). HEP utilizes behavioral health principles to counteract the effects of stress. Participants engage in group discussions and activities to promote experiential learning of strength, aerobic, flexibility, and balance exercises, nutritional concepts such as inflammatory and anti-inflammatory properties of food, and stress management through creative expression, particularly music. The daylong retreat after Week 6 involved conversation, group meal preparation, functional movement exercises, and creative expression. Home practice assignments guide participants through age-appropriate physical fitness, nutrition and meal preparation, and music engagement exercises.

MBSR classes were taught by one of two certified MBSR instructors. The senior teacher, who taught cohorts 1 and 4-8, had been teaching MBSR for 14 years and had a personal meditation practice of 25 years, and the teacher for cohorts 2-3 had been teaching MBSR for 8 years and had a personal meditation practice of 13 years. The senior teacher also provided peer support and monitoring every two weeks for cohorts 2-3. Both teachers completed all MBSR teacher trainings required by the U Mass Center for Mindfulness in Medicine, Health Care, and Society. HEP classes were taught by one of two Registered and Licensed Dietitian Nutritionists. The teacher for cohorts 1-5 was a senior research nutritionist and medical writer, and the teacher for cohorts 6-8 was a nutrition educator and exercise coach. MBSR and HEP instructors followed a written template of activities and discussion topics for each session. All intervention sessions were monitored in real time by trained research staff who attended for the purpose of monitoring intervention fidelity and addressing questions about the research procedures.

Measures

Proinflammatory Gene Expression.

Blood samples (8mL) were collected via venipuncture at pre-intervention, post-intervention, and 3-month follow-up in BD Vacutainer^® CPT^™ tubes, a fully-closed system containing sodium heparin anticoagulant, Ficoll^™ Hypaque^™ liquid density solution, and an inert gel barrier. Samples were centrifuged for 20 minutes at 1455 x g, allowing for separation and isolation of peripheral blood mononuclear cells (PBMCs) within the blood collection tube. PBMCs were spun at 10300 x g and cell pellets were resuspended in 1.5mL Invitrogen^™ RNAlater^™ Stabilization Solution and frozen at −80°C.

PBMCs were shipped on dry ice to the UCLA Social Genomics Core Laboratory for genome-wide transcriptional profiling. Total RNA was extracted from PBMCs (RNeasy; Qiagen, Valencia CA), tested for suitable mass (Nanodrop ND1000) and integrity (Agilent Bioanalyzer), and assayed by RNA sequencing following the manufacturers’ standard protocols (19). Then, analysis of differential gene expression and promoter-based bioinformatics was conducted as described in Analyses below. Analyses included 177 participants who had data available for at least one of the three assessments (Figure S1, Supplemental Digital Content).

Circulating IL-6 and CRP

were assessed from plasma samples collected at pre-intervention, post-intervention, and 3-month follow-up. Blood samples (6mL) collected in sodium citrate tubes were centrifuged for 10 minutes at 1500 x g and plasma was aliquoted for storage at −80°C until assay.

IL-6 was assayed in batch in the Behavioral Immunology Lab using the Ella Automated Immunoassay Platform (Protein Simple; Simple Plex Human IL-6 [2^nd gen]; catalog # SPCKB-PS-003028, lot # 10386). All available timepoints for each participant were assayed on the same plate. The standard range of this assay is 0.28-2652 pg/ml, and inter- and intra-assay coefficients of variability were 8.73% and 2.34%, respectively. Raw IL-6 concentrations were log-transformed and outliers (n=4 values) were winsorized to the upper boundary of three standard deviations for primary analysis. IL-6 samples were available for at least one of the three assessments from 187 participants and all data were included in primary analyses (Figure S1, Supplemental Digital Content).

Additional plasma samples were shipped on dry ice to the Laboratory for Clinical Biochemistry Research (LCBR) at the University of Vermont for CRP assay. CRP levels were determined using a particle-enhanced immunonephelometric assay with the Siemens BNII Nephelometer from Dade Behring (Newark, DE). The standard range of this assay is 0.156–500 μg/mL and the inter-assay coefficient of variability was 2.51%. Raw CRP concentrations were log-transformed and 1 outlying value was winsorized to the upper boundary of three standard deviations for primary analysis. CRP samples were available for at least one of the three assessments from 187 participants and all data were included in primary analyses (Figure S1).

Intervention and Home Practice Adherence:

Class attendance was recorded via sign-in sheet, with adherence calculated as total number of sessions out of nine. Self-reported home practice was assessed via daily Qualtrics links; daily practice duration was averaged across the eight-week intervention. Home practice during follow-up was encouraged with weekly reminder emails and was reported retrospectively during monthly phone calls. Specifically, participants reported the frequency and duration of formal and informal practice each week in the past month; self-reported weekly practice durations were averaged across the follow-up period to create indexes of formal and informal practice.

Loneliness

was assessed at pre-intervention, post-intervention, and 3-month follow-up using the 20-item UCLA Loneliness Scale (46) (N=188). The summed loneliness score had acceptable reliability (Cronbach’s α=.93) and higher scores reflect higher loneliness. Dutcher et al. (42) report on intervention-related changes in loneliness.

Analyses

Analyses were conducted using Stata 17 software (StataCorp, College Station, Texas) and SAS (Cary, NC). Preliminary analyses tested for condition differences in demographics and other baseline characteristics using chi-square (for categorical variables) and ANOVA tests (for continuous variables). ANOVAs also tested for condition differences in intervention adherence and home practice. Primary analyses include all available data and do not adjust for covariates. Covariate-adjusted models are reported in the Supplemental Digital Content.

Gene expression analyses were conducted in two steps. The first step identified genes that were differentially expressed as a function of study condition. Gene-specific transcript abundance values (gene transcripts per million mapped reads) were log_2-transformed prior to analysis. Repeated measures ANOVAs estimated the magnitude of change in gene expression over time (average of post-intervention and follow-up – pre-intervention; repeated measure) as a function of study condition (MBSR vs. HEP). Conditional models reported in Supplementary Analyses additionally adjusted for measured levels of mRNAs indicating prevalence of major leukocyte subsets, a mechanism of change in RNA which affects gene expression profiles (10). The second step quantified the prevalence of transcription factor-binding motifs (TFBMs) in core promoter sequences of each differentially expressed gene. Genes showing >50% differential change over time across conditions served as input into TELiS promoter-based bioinformatics analysis of transcription factor activity (47). This analysis tested for condition differences in upregulated and downregulated genes in the prevalence of the following TFBMs: the pro-inflammatory transcription factor, NF-κB (assessed by the TRANSFAC position-specific weight matrix V$CREL_01), the β-adrenergic signaling pathway CREB (V$CREB_Q3), the anti-viral type 1 Interferon Response Factor (IRF) family (V$IRF_Q6_01), and the anti-inflammatory GR pathway (V$GR_Q6_01).

IL-6 and CRP analyses used mixed linear models (MLMs) to test for time (pre-intervention, post-intervention, or 3-month follow-up) × condition (MBSR vs. HEP) differences on log-transformed IL-6 and CRP using the Stata Mixed procedure. Planned comparisons testing for MBSR vs. HEP differences from pre- to post-intervention and pre-intervention to 3-month follow-up were calculated within these MLMs. Hypothesis tests were 2-sided and a priori significance was set at α=.05. All available data were included in analyses, including pre-intervention data for 15 participants who dropped out; however, IL-6 and CRP data were not available at any time point for 3 participants, leaving a sample of 187. MLMs model all available data and provide unbiased estimates to account for data missing at random (MAR). Cohort was identified as a correlate of missingness (IL-6: t(568)=4.03, p<.0005; CRP: t(568)=3.93, p<.0005), with more blood samples missing in earlier cohorts; cohort was included as a covariate in Supplementary Analyses. MLMs capture both within- and between-individual variability. Time was modeled as a random effect across individuals with an independent covariance structure, with pre-intervention values used as the first repeated measure to test for time × condition interactions. Condition was modeled as a fixed effect using maximum likelihood estimation.

Exploratory analyses testing associations between plasma inflammatory markers, NF-κB expression, and loneliness are described and reported in the Supplemental Digital Content. Moderation of primary results by risk factors age and BMI are also reported in the supplementary materials.

Within-group Cohen’s d effect sizes were calculated by dividing the pre-post (or pre-follow-up) mean difference by the pooled standard deviation. Between-group effect sizes were calculated by dividing the difference between pre-post (or pre-follow-up) mean change in each condition by the pooled standard deviation of change.

Results

Preliminary Analyses

Participants were 69.77 years of age on average (SD=4.27; range: 65-85 years) and most were female (78%), white (85%), non-Hispanic (97%), and college-educated (74%). Intervention groups did not differ by age, sex, race, ethnicity, education, marital status, pre-intervention BMI, pre-intervention loneliness, or experience with mind-body practices (Table 1). There were also no baseline group differences in raw CRP (M_difference=0.61 μg/mL, SE=0.50, 95% CI: [−0.38, 1.59], χ²(1)=1.45, p=.23) or log CRP (M_difference=0.10, SE=0.17, 95% CI: [−0.23, 0.43], χ²(1)=0.36, p=.55), but raw baseline IL-6 concentrations were higher among HEP participants (M=3.69 pg/mL, SE=0.27) compared to MBSR participants (M=2.71 pg/mL, SE=0.27) (M_difference=0.97, SE=0.39, 95% CI: [0.22, 1.73], χ²(1)=6.38, p=.012), as was baseline log IL-6 (M_difference=0.19, SE=0.09, 95% CI: [0.01, 0.37], χ²(1)=4.38, p=.036) (Table 1). MLMs used in primary analyses account for baseline differences.

Supplementary Table S1 describes intervention adherence outcomes. Participants completed an average of 8 out of 9 intervention sessions (8 weekly sessions plus retreat day), with 63% of participants attending all 9 sessions. There were no group differences in intervention adherence, but HEP participants were more likely to attend the day-long retreat (87% vs. 74%; p=.031). Participants logged an average of 35.70 (SE=1.26) minutes of home practice per day. HEP participants practiced significantly more than MBSR participants (39.90 [95% CI: 36.42-43.38] vs. 31.31 [95% CI: 27.76-34.87] minutes per day; p=.001) during the 8-week intervention. Thus, home practice and retreat attendance were included as covariates in analyses reported in the Supplemental Digital Content. Home practice (formal + informal practice) during the three-month follow-up did not differ across groups, although HEP participants reported significantly more formal practice than MBSR participants (59.31 [95% CI: 52.43-66.20] vs. 17.15 [95% CI: 9.98-24.32] minutes each week; p<.001). Fifteen participants dropped out of the study, with no group differences in dropout rate (MBSR: 9.68% vs. HEP: 6.19%; p=.37). No study-related adverse events were reported.

Primary Analyses

Immune cell gene regulation

In analyses of PBMC RNA profiles, MBSR was expected to reduce bioinformatic indications of activity in pro-inflammatory NF-κB signaling. Analyses revealed 57 gene transcripts that differed by >50% between groups in the magnitude of change in average expression level from pre-intervention to post-intervention and 3-month follow-up (39 downregulated and 18 upregulated in MBSR relative to HEP). Consistent with predictions, promoter-based bioinformatics analysis of genes showing differential change over time in MBSR relative to HEP indicated reduced activity of pro-inflammatory transcription factor NF-κB (0.65-fold relative prevalence of NF-κB binding motifs in promoters of genes up-regulated in MBSR compared to HEP; −0.62 log₂-motif ratio ± SE=0.28, p=0.038; Cohen’s d=.17) (Figure 1). In supplemental analyses adjusted for leukocyte subset prevalence, this effect of MBSR vs. HEP on NF-κB activity was lessened (0.68-fold; −0.55 log2-motif ratio ± SE=0.30, p=0.064; Supplementary Material).

Figure 1.

MBSR vs. HEP effects on gene expression activity. Error bars represent standard errors of the mean log2 ratio for each transcription factor.

Second, MBSR was expected to alter gene transcription activity related to CTRA, including reducing β-adrenergic signaling (CREB) pathways and increasing activity of anti-viral (IRF) and anti-inflammatory GR pathways following MBSR relative to HEP. TELiS bioinformatics analysis of genes showing differential change following MBSR vs. HEP indicated (1) no changes in activity of β-adrenergic transcription factor CREB (0.89-fold; −0.18 log₂-motif ratio ± SE=0.14, p=0.20; d=.10), (2) a trend towards increased anti-viral IRF transcription factor activity (1.38-fold; 0.46 log₂-motif ratio ± SE=0.27, p=0.086; d=.13), and, (3) unexpectedly, a trend toward reduced anti-inflammatory GR transcription factor activity following MBSR relative to HEP (0.86-fold; −0.22 log₂-motif ratio ± SE=0.12, p=0.063; d=.14). Supplemental analyses adjusting for leukocyte subset prevalence showed significant effects of MBSR vs. HEP on reduced CREB, increased IRF, and reduced GR transcription factor activity (Supplementary Material).

Circulating plasma inflammatory markers

MBSR was expected to decrease plasma IL-6 and CRP compared to HEP from pre-intervention to post-intervention and 3-month follow-up. MLMs tested for time (pre, post, follow-up) × condition (MBSR, HEP) interactions on IL-6 and CRP.

Contrary to predictions, IL-6 analyses showed no time × condition effects across pre, post, and follow-up timepoints (χ²(2)=2.18, p=.34) and no time × condition effects from pre- to post-intervention (b=−0.08 ± SE=0.08, χ²(1)=0.74, p=.39, d=.06) or 3-month follow-up (b=−0.12 ± SE=0.08, χ²(1)=2.16, p=.14, d=.11) (Table 2).

Table 2.

Raw and log-transformed CRP and IL-6 at pre-intervention, post-intervention, and 3-month follow-up in MBSR and HEP.

	MBSR		HEP		Pre-Post	Time x Condition
Characteristic	(N=92)		(N=95)		Cohen’s d	Difference (Log)
CRP	Raw (μg/mL)	Log	Raw (μg/mL)	Log
Pre-intervention	1.94 (0.36)	0.08 (0.12)	2.54 (0.35)	0.18 (0.12)	--	--
Post-intervention	2.12 (0.39)	0.22 (0.12)	2.75 (0.37)	0.23 (0.12)	.05	χ2(1)=0.42, p=.52
3-month follow-up	2.55 (0.41)	0.31 (0.13)	2.24 (0.40)	0.13 (0.12)	.15	χ2(1)=4.18, p=.041
IL-6	Raw (pg/mL)	Log	Raw (pg/mL)	Log
Pre-intervention	2.71 (0.27)	0.83 (0.07)	3.69 (0.27)	1.02 (0.06)	--	--
Post-intervention	2.86 (0.30)	0.88 (0.07)	3.50 (0.29)	1.01 (0.07)	.06	χ2(1)=0.74, p=.39
3-month follow-up	3.03 (0.32)	0.95 (0.07)	3.73 (0.31)	1.02 (0.07)	.11	χ2(1)=2.16, p=.14

Note: Data reported as means and (SEs). Cohen’s d estimates between-group differences in log change from pre-intervention to post-intervention and 3-month follow-up.

Similarly, CRP analyses showed no time × condition effects across pre, post, and follow-up timepoints (χ²(2)=4.36, p=.11) and no time × condition effects from pre- to post-intervention (b=−0.09 ± SE=0.13, χ²(1)=0.42, p=.52, d=.05). There was a significant time × condition effect from pre-intervention to 3-month follow-up (b=−0.28 ± SE=0.14, χ²(1)=4.18, p=.041, d=.15), but, opposite to predictions, this effect was driven by increases in CRP in the MBSR group (M_change=0.23 ± SE=0.10; χ²(1)=5.53, p=.019, d=.26) relative to HEP (M_change=−0.05, SE=0.09; χ²(1)=0.26, p=.61, d=.04) (Table 2). Covariate-adjusted models described in the Supplemental Digital Content showed the same pattern of effects for IL-6 and CRP.

Discussion

In a large sample of lonely older adults, mindfulness meditation training altered pro-inflammatory gene expression—but did not reduce plasma inflammatory markers—relative to a matched intervention focused on enhancing health behaviors. Specifically, mindfulness training showed downregulated pro-inflammatory NF-κB gene expression relative to HEP. No other immune cell gene regulation pathways involved in the CTRA showed significant changes in unadjusted analyses, although after adjusting for changes in immune cell abundance, there was some indication that MBSR can alter CTRA pathways. No group differences were observed on circulating IL-6, and findings indicate a relative increase in plasma CRP levels three months after the intervention among individuals who received mindfulness training. This is the largest study to date testing mindfulness intervention effects on intra- and extra-cellular markers of inflammation. It suggests specificity in how mindfulness interventions influence inflammatory biology, providing evidence that mindfulness training does not consistently reduce markers of systemic inflammation, but can alter pro-inflammatory gene regulatory pathways in circulating immune cells. These changes in immune cell regulation may have implications for biological aging, with potential to reduce risk or progression of inflammatory diseases (11).

Findings are consistent with prior work showing downregulating effects of mindfulness interventions on pro-inflammatory NF-κB in smaller samples of lonely older adults (8), stressed customer service workers (19), and breast cancer survivors (21). Reduced NF-κB activity has also been observed in samples of older lonely or distressed adults following other mind-body (tai chi (48,49); yogic meditation (50)) and cognitive behavioral (49,51) interventions. Here, relative to HEP, MBSR showed significant downregulating effects on NF-κB activity in unconditional analyses. Effects were smaller after adjusting for changes in leukocyte subset prevalence. Psychosocial stress and accompanying sympathetic nervous system activity is known to alter relative numbers of leukocyte subsets in peripheral circulation, particularly increasing monocyte abundance. Together, these findings suggest that mindfulness training-related decreases in NF-κB activity may be partially driven by changes in cell subtype numbers, possibly reflecting decreases in monocytes numbers, which are the primary cell subtype responsible for NF-κB transcription.

Mindfulness and mind-body interventions have also been shown to alter other pathways involved in the conserved transcriptional response to adversity (CTRA). Primary unconditional models showed no significant effects of mindfulness intervention on β-adrenergic CREB, anti-viral IRF, or anti-inflammatory GR transcriptional pathways. Supplemental analyses that adjusted for intervention-related changes in circulating numbers of leukocyte subsets—which eliminate effects of changes in monocyte abundance—showed changes in gene transcription activity that are relatively consistent with previous work. Previous trials have shown downregulating effects of mind-body interventions on CREB activity (20,49,51,52), indicating alterations in sympathetic nervous system signaling. Here, CREB activity was similarly downregulated following MBSR relative to HEP in models adjusted for leukocyte subsets. Mind-body interventions have shown less reliable effects on anti-viral gene expression, showing upregulating effects on IRF activity in some studies (21,50) and no effects in others (19,49). Here, IRF activity was downregulated following MBSR (vs. HEP) in adjusted models. Mind-body interventions have tended to upregulate the expression of genes with GR response elements compared with both inactive and health education control groups (21,49,52). Inconsistent with these previous trials, in this sample, MBSR vs. HEP tended to downregulate GR activity. It is possible that interventions focused on nutrition, exercise, and sleep hygiene also exert effects on glucocorticoid signaling. Since downregulated GR activity following mindfulness vs. health enhancement has not been observed before (and these effects reached the threshold for significance only in leukocyte-adjusted analyses), this pattern warrants replication before novel interpretations are considered. Measurement of circulating cortisol in conjunction with GR gene expression is needed to understand whether mindfulness training alters cortisol outflow in ways that adjust HPA-axis feedback loops. We previously reported in this sample that MBSR (vs. HEP) maintains functional immune cell sensitivity to glucocorticoids (40), but whether either intervention altered circulating cortisol is unknown. It’s possible that mindfulness training alters glucocorticoid activity primarily in the context of acute stress (18); effects on basal cortisol levels are rarely observed (53). Together with previous evidence, these findings suggest that mindfulness training may alter CTRA in immune cells of people exposed to chronic psychosocial stress (11), although these effects are only apparent after adjusting for sympathetically-driven changes in inflammatory monocyte production.

Mindfulness training did not reduce circulating protein markers of inflammation, and instead MBSR participants showed a relative increase in CRP at three-month follow-up. A growing body of literature shows mixed effects of mindfulness interventions on plasma inflammatory biomarkers (9,22). Although some studies have found support for the idea that mindfulness interventions decrease markers of inflammation in samples at risk for chronic inflammatory disease (26,54), larger efficacy trials have sometimes failed to replicate these effects (e.g., (55,56); for a review, see (25)). In the present study, we aimed to recruit an at-risk sample of lonely older adults to replicate our previous pilot findings in this population showing marginal effects of MBSR vs. waitlist control on reducing CRP (8). However, this study’s exclusion criteria may have winnowed our sample to an unusually healthy, potentially ‘biologically elite’ sample of lonely older adults who may already have some protection against the health effects of loneliness (40). Supplemental analyses exploring higher risk subgroups show no indication that MBSR reduces IL-6 or CRP among older or overweight lonely adults. Moreover, the HEP comparison intervention focused on nutrition, exercise, and sleep hygiene, positive health behaviors that relate to lower levels of systemic inflammation (35-37), perhaps by targeting inflammatory proteins in adipose and other tissues. As such, these health behaviors may buffer ‘inflammaging’ more reliably than an intervention focused primarily on psychological skills. Indeed, greater home practice duration correlated with greater decreases in both IL-6 (r=−.25, p=.017) and CRP (r=−.33, p=.001) in the HEP group only, providing some evidence that adopting positive health behaviors is beneficial for mitigating systemic inflammation among healthy lonely older adults. Across groups, exploratory analyses also showed associations between changes in loneliness and IL-6, supporting the hypothesis that improvements in psychosocial distress (whether through learning mindfulness skills, engaging in a supportive group environment, or other nonspecific intervention effects) may be an independent pathway to improve systemic inflammation.

This study provided a unique opportunity to test whether changes in cellular transcriptional measures of inflammatory biology parallel changes in plasma protein inflammatory markers, or whether those molecularly distinct aspects of inflammatory biology are functionally distinct as well. Few trials have measured both gene expression and protein marker outcomes to test this pathway. Notably, similar to the present results, Bower et al. (2015) found effects of mindfulness training (vs. no treatment) on NF-κB, CREB, and IRF expression, but mindfulness training did not significantly impact plasma IL-6, CRP, or TNF (21). Similarly, our pilot trial of lonely older adults showed downregulating effects of mindfulness training vs. no treatment on NF-κB activity, but no effects on IL-6 and marginal decreases in CRP (8). Here, mindfulness vs. health enhancement intervention downregulated NF-κB, but did not reduce circulating IL-6 or CRP. Moreover, changes in intracellular gene expression did not correspond with changes in circulating markers of inflammation, suggesting that intervention-induced changes in NF-κB in circulating immune cells do not associate with changes in plasma protein biomarkers. This is not altogether unexpected because plasma inflammatory proteins are not primarily produced by circulating immune cells, but rather by other tissues such as adipose, liver, and spleen. As such, the present results imply that mindfulness training may have distinct effects on distinct tissue systems, with more pronounced impacts on the biology of circulating leukocytes.

Inflammatory processes play a central role in biological aging and in truncating healthspan. The present results join our previously reported trial findings showing that mindfulness training may cultivate a change in perspective that diminishes psychosocial stress in ways that alter select immune processes related to inflammatory disease risk. Mindfulness training initiated changes in intracellular expression of pro-inflammatory NF-κB genes, enhanced the inflammatory response to immune challenge (39), and preserved the sensitivity of immune cells to glucocorticoids (40). However, mindfulness training did not reduce circulating plasma markers of inflammation in this study, and it showed mixed effects on CTRA gene expression pathways relative to HEP. Collectively, these mindfulness training effects have potential implications for altered biological responding to psychosocial threat, physical recovery processes, and inflammatory disease risk.

This trial has several strengths and limitations. It is the largest trial to date testing mindfulness intervention changes in multiple immune processes related to biological aging. Intervention and protocol adherence was excellent, allowing for a reliable test of hypotheses. Mindfulness intervention effects were observed above and beyond effects of a stringent active comparison intervention, suggesting that the development of mindfulness skills is uniquely capable of getting ‘under the skin’ to impact markers of health. However, although this design is useful for testing active mechanisms beyond social support and other nonspecific treatment effects, a no treatment control group would be necessary to test how MBSR effects compare to the standard process of inflammaging over the course of five to six months. Instead, this trial compared MBSR to an intervention that boosted positive behaviors known to support healthy aging, so benefits of either program relative to typical aging remain unknown. Second, several sample characteristics have potential to limit generalizability of results. A minority of participants had exposure to mind-body practices, which could conceivably boost or attenuate effects. And although we recruited adults ages 65-93 years and age did not consistently moderate effects, the majority of the sample was younger than 70 years; it may be more difficult for middle-old or oldest-old adults to access or engage with mindfulness training. Medical exclusions also resulted in a healthy older adult sample; results may not generalize to higher-risk older adults. However, results are largely consistent with mindfulness trial findings in a range of populations. Still, an important question is whether mindfulness interventions can extend healthspan in populations on an accelerated trajectory of declining health. Finally, the measurement of long-term health outcomes is an important next step for understanding whether mindfulness interventions have potential to extend healthspan.

Overall, the present findings suggest that mindfulness training may benefit lonely older adults by altering immune processes underlying biological aging. The trial provides specificity about these immune processes, suggesting that mindfulness interventions alter intracellular pro-inflammatory NF-κB gene regulation pathways but not extracellular markers of systemic inflammation. Compared to an intervention focused on positive health behaviors, mindfulness training downregulated pro-inflammatory gene expression, which may help offset the effects of chronic psychosocial stress on inflammatory disease risk.

Acronyms:

CREB

cAMP response element-binding protein

CRP

C-Reactive Protein

GR

glucocorticoid receptor

HEP

Health Enhancement Program

IL-6

Interleukin-6

IRF

interferon regulatory factors

MBSR

Mindfulness-Based Stress Reduction

MLM

mixed linear model

NF-κB

nuclear factor kappa B

PBMCs

peripheral blood mononuclear cells

RNA

ribonucleic acid

TELiS

Transcription Element Listening System

TFBMs

transcription factor-binding motifs

Data availability:

Data that support the findings reported in this paper are available from the corresponding author upon reasonable request.