Mind–body practices have been shown to have therapeutic effects in a wide variety of physical and psychological conditions, such as stress, depression, and anxiety (D'Silva et al., 2012; Maric et al., 2021). Despite growing research interest in the psychophysiology of mind–body practices (Brandmeyer et al., 2019; Riley and Park, 2015), there has been little consensus on the choice of mechanistic or therapeutic outcome variables, and there are rarely direct comparisons on the effects of different mind–body modalities (e.g., meditation vs. yoga) that are popular among US adults (Dossett et al., 2020). Therefore, we conducted a randomized controlled trial (RCT) to comprehensively compare meditation and yoga against an active control condition of stress education, with a large battery of outcome measures including self-report psychological questionnaires, stress biomarkers and gene expression measures.

A large number of self-report questionnaires have been used in mind–body research (Crosswell and Lockwood, 2020). The Perceived Stress Scale (PSS) is one of the most widely used self-report measures of stress (Cohen et al., 1983). Additional measures on depression (Zigmond and Snaith, 1983), anxiety (Spielberger et al., 1971), burnout and resilience (Wagnild and Young, 1993) are also frequently utilized. Therefore, this study included a large battery of psychological measures to comprehensively evaluate their reliability in detecting psychological symptom changes across different mind–body modalities.

“Allostatic load” is a concept regarding the accumulative “wear and tear” of chronic stress on physiological systems (McEwen and Stellar, 1993; Prior, 2021). While the human body can adapt to mild levels of stress, repeated and prolonged environmental challenges can eventually lead to inability to maintain a normal homeostatic balance due to chronic alterations in physiological systems (McEwen and Stellar, 1993; Prior, 2021). Long-term adverse effects of maladaptive response to chronic stress (Dhabhar, 2014) include inflammation and immunosuppression (Barrett et al., 2021), cardiovascular dysfunction and disease (Kivimäki and Steptoe, 2018), accumulation of abdominal fat (Aschbacher et al., 2014), loss of bone minerals (Kelly et al., 2019), decreased neurogenesis (Du Preez et al., 2021), increased neuronal cell death and associated atrophy in the limbic system (Schoenfeld et al., 2017). Many stress biomarkers have been shown to respond to mind–body practices, for example, prior mind–body studies have shown changes with interleukin-6 (IL-6) levels, despite having mixed findings (Bower and Irwin, 2016; Moreno, 2024). Cortisol is another biomarker that is frequently investigated in mind–body research which also has inconclusive findings in the literature (O'Leary et al., 2016). Mindfulness-based interventions have been shown to reduce blood pressure (Intarakamhang et al., 2020; Solano López, 2018). Mind–body practices have also shown promises in weight management (López-Alarcón et al., 2025). Due to the multi-facet effects of mind–body practices and the potential limitation with a particular biomarker, in this study we utilized a composite “allostatic load index” to capture biomarkers from several physiological systems including the neuroendocrine system, the immune system, the cardiovascular system, and the metabolic system (Logan and Barksdale, 2008; Seeman et al., 2010). Using such a composite measure to reflect stress-related dysregulation across several physiological systems increases measure reliability.

Inflammatory genetic transcription factors are also well-researched biomarkers of chronic stress (Natoli et al., 2011). These transcription factors regulate inflammatory and anti-inflammatory gene expression through modulating the production of cytokines, chemokines, and other inflammatory molecules (Cohen et al., 2012). Notable transcription factors include nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) (Biancalana et al., 2021) anti-inflammatory and pro-inflammatory pathways, as well as tumor necrosis factor receptor 2 (TNFR2; Medler and Wajant, 2019), and IL-6 JAK STAT3 signaling (Johnson et al., 2018) that are indicative of inflammatory and immune responses. Prior mind–body research demonstrated reduction of the NF-κB pro-inflammatory transcription control pathway (Bhasin et al., 2013; Dutcher et al., 2022; Oyler et al., 2023) as well as the TNFR2 pathway (Bhasin et al., 2013). Therefore, this study also measured several genomic biomarkers from blood samples and evaluated their relationships with self-report measures.

This study utilized a three-arm RCT design to compare the effects of meditation vs. yoga vs. an active control of stress education, to identify potential common and unique psychophysiological effects of meditation and yoga. The study was conducted with chronically stressed but otherwise healthy adults to focus on stress-related effects and mechanisms while minimizing the influences from pre-existing disease conditions.

This 3-arm RCT compared meditation, yoga, and an active control among chronically stressed but otherwise healthy adults with a large battery of outcome measures. The three arms did not show significant differences in any of the outcome measures. Most psychological measures demonstrated significant post-intervention improvements within all study arms, but physiological measures did not show significant improvements within any study arms. Several psychological questionnaires showed strong correlations with physiological measures, such as PSS-10, PWBS, CAMS-R and STAI-SS, but did not survive multiple comparison.

Self-report measures are often criticized for subjectivity, and there have been reports of biological differences in the absence of psychological differences between mind–body interventions and controls, and vice versa (Joss et al., 2024a; Joss et al., 2024b). However, there are also limitations with physiological measures because they can be affected by diurnal fluctuations and many life style factors such as food and smoking (Garde et al., 2009). For example, prior research suggests cortisol response in laboratory-induced acute stress can more accurately reflect intervention effects than accumulative cortisol (Rogerson et al., 2024). Because all the physiological measures used in this study were global accumulative measures, which could be influenced by many factors in real life, the allostatic measures used in this study may not be sensitive enough to detect intervention-related changes. Prior studies on the effects of mind–body interventions on gene expression usually utilized a hypothesis-free gene set enrichment analysis (GESA) approach (Bhasin et al., 2018; Bhasin et al., 2013; Kuo et al., 2015), whereas this study focused on specific gene sets such as the BioCarta NF-kB Pathway based on our prior research (Bhasin et al., 2018; Bhasin et al., 2013; Dusek et al., 2008a; Dusek et al., 2008b; Kuo et al., 2015). It is possible that this hypothesis-driven approach might have been too restricted in its ability to detect potentially important changes in genetic expression related to mind–body interventions. Therefore, the mostly null findings with physiological outcome measures in this study could have been related to the research methodology of measuring global accumulative biomarkers and using a hypothesis-driven genomic analysis approach.

Multiple review articles have demonstrated overall mixed findings regarding the effects of mind–body programs on stress biomarkers such as cortisol (O'Leary et al., 2016) and IL-6 (Bower and Irwin, 2016; Moreno, 2024), despite some reports of post-intervention reduction (Walsh et al., 2016). For example, one recent study reported post-intervention elevation of IL-6 among older adults (Lindsay et al., 2022), which was interpreted as indicative of adaptive boost of immune competence for fighting off infections among older adults (Lindsay et al., 2022; Moreno, 2024). A recent meta-analysis of 105 meditation studies with total N = 3,826 found overall small effect size on immune functioning including IL-6 levels (Oyler et al., 2023). Although our study showed increased IL-6 within the stress education and meditation treatment arms, the elevated values were still within the normal range for healthy adults (Said et al., 2021). The physiological or biopsychosocial mechanisms for IL-6 level fluctuations within the normal range among healthy adults are still not well understood (Picotte et al., 2009). Our study also found elevated HDL in the yoga treatment arm at the follow-up timepoint, which is consistent with findings on the association of yoga and HDL in a recent meta-analysis of 53 studies with a total sample size of N = 13,191 (Ghazvineh et al., 2022). Our findings add to the literature of mixed findings on the effects of mind–body interventions for stress biomarkers for consideration in future research.

A major limitation of this study is the use of a relatively healthy study sample. This study only enrolled chronically stressed but otherwise healthy adults with no mental illness. Any prospective participants with DSM axis-I disorders were excluded. The purpose of choosing the healthy adult study population was to investigate the effects of mind–body interventions while avoiding confounding factors associated with any particular medical or mental health disorders. This kind of healthy adult sample is representative of a large number of healthy adult participants of mind–body programs in the general public, therefore this study design has generalizability for the public. Nevertheless, in this study, the healthy adult sample might have led to floor effects that limited the detectability of intervention-related physiological changes with existing technology within the short timeframe of the study, whereas study populations with psychopathology could have larger effects that are more detectable (Wielgosz et al., 2019). For example, a recent systematic review of 44 meta-analyses of mindfulness-based interventions (k = 336 RCTs, N = 30,483 participants) suggests strongest effects were observed from disease populations, whereas findings from healthy adult populations were inconsistent with overall small effect sizes when compared with active control conditions (Goldberg et al., 2022).

As discussed above, the stringent eligibility criteria of this study excluded any prospective participants that met diagnosis criteria for any DSM axis-I disorders; future studies may consider loosening the eligibility criteria to include patients with mild or moderate levels of Major Depressive Disorder and Generalized Anxiety Disorder, which also represent a large number of participants of mind–body programs in the general public. Prospective participants on any psychiatric medications were excluded from this study out of concerns of affecting genomic measures, but such exclusion criteria may not be necessary for future clinical trials focused on testing the clinical efficacy of mind–body programs. The “chronically stressed” criterion was based on self-report, which could have susceptibility to inaccuracies and biases, because self-reported perceived stress is not necessarily equivalent to physiological stress levels, for example, someone could be experiencing chronic stress but lacks the awareness to realize they have been stressed thus under-report the “perceived stress” levels, whereas someone could be hyper-sensitive to the fluctuation in stress levels without actually having high levels of stress biomarkers. Nevertheless, although biomarker-based eligibility criteria might have been more objective, there is currently no consensus on appropriate stress biomarkers with feasible clinical application, and it is time consuming and resource demanding to utilize the composite allostatic load index with 14 biomarkers for screening purposes.

Furthermore, the stress education active control arm (Hoge et al., 2013) might have had unexpectedly strong stress reduction elements, such as using audio books as daily home practice and raising awareness on diet and exercise. Such an active control condition has shown similar effects in reducing stress, psychological symptoms and blood pressure compared to mind–body interventions (Dusek et al., 2008a; Dusek et al., 2008b; Joss et al., 2024a; Joss et al., 2024b; Joss et al., 2024c). Most importantly, recent meta-analyses have demonstrated that mind–body interventions often had no superiority over active control conditions in prior studies (Goldberg et al., 2022), therefore the lack of between-group effects in this study is consistent with the existing literature.

There are several other limitations with this study. (1) Given the above-mentioned limitations with the active control arm, this study could have benefited from including a passive control arm, such as a waitlist or “do nothing” control arm, which might have been able to provide information on longitudinal psychological and physiological changes without any interventions. (2) The null findings in this study could have been related to the limitations with the instruments and assays used for screening and assessments. Newer technologies, such as AI-based facial biomarkers (Al-Jebrni et al., 2020; Wang et al., 2025), is more objective than self-report measures and potentially more specific and sensitive than traditional physiological measures. Therefore, future studies should consider utilizing these AI based technologies. (3) Lack of study sample demographic diversity is another limitation, which compromised the generalizability of study findings. Despite all the limitations, this study can still provide methodological insights for future mind–body research.