Yoga-based lifestyle intervention for healthy ageing in older adults: a two-armed, waitlist randomized controlled trial with multiple primary outcomes

Abstract

Yoga-based clinical research has shown considerable promise in varied ageing-related health outcomes in older adults. However, robust frameworks have yet to be used in intervention research to endorse yoga as a healthy ageing intervention to test the multidimensional construct of healthy ageing. This was an assessor-masked, randomized controlled trial conducted among 258 sedentary, community-dwelling older adults aged 60–80 years, randomly allocated to 26-week yoga-based intervention (YBI) (n = 132) or waitlist control (WLC) (n = 126). The effectiveness of YBI was assessed through two separate global statistical tests, generalized estimating equations and rank sum-based test, against a comprehensive healthy aging panel comprised of ten markers representing the domains of physiological and metabolic, cognitive, physical capability, psychological, and social well-being. The secondary outcomes were individual primary marker scores, Klotho, inflammatory markers, and auxiliary blood markers. We could establish the healthy aging effect of the 26-week YBI over WLC using two models of global statistical test (GEE, β = 0.29; 95% CI = 0.20 to 0.38, p < 0.001), and rank sum-based test (β = 0.28, 95% CI = 0.19 to 0.36, p < 0.001). There were also significant improvements in direction of benefit at individual levels of all the aging markers. Exploratory evaluation with adopted indices from contemporary clinical trials also validated the potential of YBI for healthy aging; HATICE adapted composite score (mean difference =  − 0.18; 95% CI =  − 0.26 to − 0.09, p < 0.001) and healthy ageing index (mean difference =  − 0.33; 95% CI =  − 0.63 to − 0.02, p = 0.03). The global effect of YBI across multiple ageing-related outcomes provides a proof of concept for further large-scale validation. The findings hold a great translational value given the accelerated pace of population aging across the globe. Trial registration: CTRI/2021/02/031373.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11357-024-01149-5.

Introduction

As per the latest estimates, India has been estimated to have overtaken China as the world’s most populous nation [1], with projections of a nearly doubled proportion of older adults by the year 2050 [2]. Driven by enhanced life expectancy and fertility drops, rapid population ageing has become a significant cause of global demographic shifts [3, 4]. Such rising trends of the population aging call for swift implementation of preventive strategies to enhance the healthy survivorship of the older adults. It is well reflected in the strategized announcement by the World Health Organization of the period “2021–2030” as the decade of healthy aging [5].

The inevitable gradual deterioration in health accompanied by the ageing process manifests across various dimensions of physical and mental health, wherein a sedentary lifestyle acts as a catalyst to enhance the pace of the decline [6]. Physical activity has been reported to be an effective means to alleviate this ageing-associated decline [7, 8]. Therefore, lifestyle promotion of physical activity has been a critical focus of beneficial ageing policies to prevent disability and fast deterioration in health in older adults [9]. Due to its gentle approach, there is an increasing global interest towards considering yoga as a physical activity regimen in the older population [10, 11]. As per the international exercise guidelines, we also find mention of yoga, under the recommendations for optimal ageing and maintenance of functional capacities in older adults with respect to their balance and gait-promoting effects [12]. As an intervention, yoga has often been directed at several physiological systems, many of which intersect with the concept of healthy ageing phenotype (HAP) [13–15]. Overall, the effects of yoga-based interventions are focused on physical capability, cognition, and well-being as supported by meta-analyses [13, 14]. Recently, yoga has been summarized to affect frailty markers associated with clinically meaningful outcomes in older adult populations [14]. However, domains of metabolic and physiological health have received scant attention in yoga-based research on older adults [14].

Taken together, the quality of scientific evidence on the health benefits of yoga in older adults has improved in terms of extensive and better study design with small to moderate effect sizes, in line with other forms of exercise programs [14, 15]. However, a strong and comprehensive framework of evidence remains to be established to endorse yoga as an intervention for healthy ageing. This limitation is driven by the lack of standard markers for healthy ageing [15] and use of single primary markers against the multidimensional construct of healthy ageing [16, 17].

The purpose of the study is to explore the effectiveness of regular practice of yoga under the umbrella term of “healthy ageing” across several domains of health [16]. Achieving statistically significant impact on all the individual outcomes would be too stringent; hence, we envisaged a two-criteria approach to resolve the complex attribute of healthy aging (1) by adopting an array of markers representing the five domains of health (physiological and metabolic health, physical capability, cognitive function, psychological and social well-being) and representing the healthy ageing phenotype (2) with the use of a comprehensive statistical models to capture overall direction and impact of yoga-based lifestyle intervention. Further, the efficacy of the intervention was also attested with improvements on a separate set of mechanistic biomarkers of longevity (Klotho) and inflammation (C-reactive protein and Tumor necrosis factor-receptor II), to boost confidence in the trial’s findings.

Methods

Study design

The study tested the holistic responsiveness of healthy ageing indicators to yoga-based lifestyle intervention (YBI) vs. waitlist control in older adults. The design and method of the trial entitled “Yoga for the healthy ageing phenotype (yHAP)” have been described previously [18]. Briefly, yHAP was a 26-week, monocentric, randomized, two-armed, assessor-masked, waitlist-controlled trial. Given the emerging second wave of COVID-19, the protocol was amended with a detailed account of the changes provided in (Appendix 1 and 2). The present study adhered to the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines [19]. The ethics committee approval was taken from the institutional ethical committee. Written informed consent was obtained before the start of the study, and all participants were informed about the randomization.

Participants

Participants were sedentary older adults aged 60–80 years and representing both genders. They were further included if they were interested and willing to participate in study, able to perform moderately strenuous yoga asana, and/or diagnosed with lifestyle diseases like diabetes mellitus, obesity, and hypertension.

The recruitment took place between February 2021 and August 2021, from the residential communities of Bangalore City, India, through the distribution of pamphlets and door-to-door visits in the intervention’s vicinity facility (Appendix 3; detailed community-wise distribution has been appended in supplementary Table 1). Exclusion criteria were (i) recent engagement in any regular structured exercise or yoga, (ii) history of other chronic (coronary heart disease, arthritis, chronic kidney disease) or neurological ailments, (iii) concurrent participation in another research study, and (iv) any recent surgery limiting the practice of yoga. Given the rise in COVID-19 infections, an additional exclusion of recently infected subjects was also amended in the trial (Supplementary Table 2).

Randomization and masking

An external statistician randomized the participants in a 1:1 ratio using a permuted block size of 20, stratified by community centers with a computer-generated random sequence allocator. The assignment codes were put in sealed, opaque envelopes and were concealed from the researcher involved in the enrollment. The codes were then opened by non-study personnel at the site. Outcome assessors were also masked to the group allocations. However, participants, intervention staff, and other site staff could not be blinded due to the nature of the yoga-based intervention.

Intervention

The YBI intervention aligned with the integrative yoga approach consisting of a combination of postures and breathing techniques interspersed with short bouts of relaxation (Supplementary Table 3). These easy-to-perform yogic practices were chosen based on the high likelihood of older adults adhering to the program [20]. The frequency and duration of the procedures aligned with the previously published reports on older adults [20–23]. All the participants of the YBI group followed the intervention for 26 weeks. For the initial 12 weeks, contact classes of yoga practice were conducted by certified yoga teachers for 60 min and 5 days/per week. Following the same, participants were advised to do daily home-based practice for the next 12 weeks; this was done to integrate the intervention into their daily routine settings. During this period, there were also follow-up contact classes, conducted at 15 days interval, wherein the yoga instructors documented the self-reported attendance for the home-based practice by the participants, and also delivered yoga session. Dietary advice was also based on the ancient yoga-based concepts aligned with the research-based evidence for healthy nutrition in older adults (Supplementary Table 4).

Personal time with the trainer was given separately throughout the intervention phase to all the participants to achieve fidelity and make them understand the practices more clearly. All the participants received monthly phone calls to assess any personal changes in health. Logbooks were maintained during the individual sessions, and participants were asked to record attendance. The yoga trainer considered the self-records during each contact session to obtain the daily attendance. The research team personally called participants who missed more than two consecutive sessions to understand why they were absent and avoid any unfavorable impacts of the intervention. Throughout the intervention phase, participants could meet with trainers discreetly to help them comprehend the techniques. In the program’s second half, participants were asked to maintain an attendance register, which the yoga instructor reviewed during the contact sessions, held at 15-day intervals. If the participants missed the contact sessions, their attendance was telephonically examined by the project staff. Adherence was based on the number of sessions attended. Adverse events were defined as undesirable symptoms or responses during a yoga session, and the answers were recorded by the yoga instructor on an individual basis post-yoga session.

Participants in the control group did not receive any intervention besides their usual care and were instructed to continue their daily activities (without engaging in regular structured exercise).

Outcomes

Primary outcome

All outcomes were assessed at baseline, and post intervention (at end of 26 weeks). The primary outcome was a differential overall change in the markers of the healthy aging phenotype between the YBI and WLC groups, assessed using two different methods of global statistics, the generalized estimating equations (GEE) and rank-sum tests. Domain-wise healthy aging markers-physiologic and metabolic health (glycated hemoglobin (HbA1c), systolic blood pressure (SBP), forced expiratory volume (FEV1), low-density lipoprotein-cholesterol (LDL-C)), cognitive health (digit symbol substitution test (DSST), and ratio of trail making test parts A and B (TMT A/B)), physical capability (gait speed (GS), hand grip strength (HGS)), psychological well-being (WHO Quality of Life Instrument-Short Form (WHOQol-BREF)), and social well-being (University of California, Los Angeles Loneliness Scale (UCLA)) were carefully selected from the panel of aging markers [16].

Secondary outcomes

The secondary outcomes were individual scores of the marker of healthy aging phenotype, along with Klotho levels, body mass index (BMI), complete blood count, blood urea, and creatinine composite inflammatory score (TNF-RII, CRP).

Exploratory outcomes

We also explored the effects of intervention with two adopted composite scores: the healthy aging index (HAI), an estimate of physiologic aging [24], and a composite of BMI, SBP, and LDL-c adopted from the “Healthy Ageing Through Internet Counselling in the Elderly (HATICE) trial” [25]. To examine whether treatment groups differ in mean differences in the two samples, we used Hotelling’s T2 test, which is a multivariate equivalent of Student’s t test.

Assessments

Age was calculated from birth as per the baseline evaluation date. Marital status was categorized into four groups (single/married/divorced/widowed). Education was assessed in years. Employment status is categorized as either currently employed or unemployed. Smoking status was categorized into three groups (never, former, current). Nutritional status was assessed using the Mini Nutritional Assessment questionnaire (MNA) [26]. BMI was calculated as weight (kg)/height (m²). Height was measured using a stadiometer, and body weight was measured using a digital scale. Digital BP monitor (Omron model HEM-7120, Omron Healthcare Co. Ltd.) was used for the measurement of blood pressure at the end of the 5 min of rest as the average of two measurements 60 s apart. A portable spirometer (Schiller spirometer (SP-1A)) was used to test the lung functions of the participants. Forceful exhalation for 1 s after a maximum inhalation will be set as forced expiratory volume in 1 s (FEV1). The grip strength of each hand was measured using a hydraulic hand dynamometer (Baseline® measurements; Fabrication Enterprises Inc, Elmsford, NY, USA). The protocol employed an average of three strength tests as the resultant score for each hand. Gait speed was assessed by a self-paced 15-foot walk test [27]. For cognition, we used the digit symbol substitution test (DSST) using correct number–symbol matches in 90 s [28] and the B/A ratio of the trail-making test (TMT) [28]. An individual’s perceived physical health, psychological health, social relations, and environment-related QOL were assessed using the WHOQOL-BREF questionnaire (World Health Organization Quality of Life) [29]. An assessment of social isolation and loneliness was conducted using the UCLA loneliness scale version 3, which consists of 20 items [30]. frailty scores (non-frail and at risk of frailty) were assigned to participants using a 10-item instrument, Frailty Index for Elders [31]. Diabetes was defined by either self-report or HbA1c >  = 6.5% [31]; hypertension was defined as on medication or an average of two blood pressure (BP) readings where systolic BP ≥ 140 mmHg and/or a diastolic BP ≥ 90 mmHg. Dyslipidemia was calculated with criteria (LDL-C ≥ 130 mg/dL, TC ≥ 200 mg/dL, TG ≥ 150 mg/dL, or low levels of HDL-C < 40) [32].

Venipuncture was done post a minimum of 8-h fasting, and blood samples were collected in EDTA tubes. According to the manufacturer’s guidelines, glycated hemoglobin (HbA1c) was measured using the Beckman Coulter Chemistry Analyzer AU480 system. An automatic chemistry analyzer (Mindray BS390, Mindray Bio-Medical Electronics Co., Ltd., Shenzhen, China) was used to measure glucose, lipid, urea, and creatinine under quality control standards. Inflammatory markers (TNF-RII and CRP) were assessed using Luminex-based immunoassay (Luminex Corporation, Austin, TX). Anti-ageing marker, Klotho, serum levels were measured by enzyme-linked immunosorbent assay using a soluble α-Klotho assay kit (Human Klotho DuoSet ELISA, R&D Systems, Minneapolis, USA). These markers’ baseline means and SDs were combined to calculate z-scores. The z-scores were then averaged to calculate composite z-scores separately for baseline and post-intervention data.

The HAI was scored using markers of cardiovascular, lung, cognitive, metabolic, and kidney function (SBP, FBS, DSST, creatinine, and FEV1) [24]. Components were scored from 0 to 2 from most to least healthy using clinical cut points for fasting blood sugar (0: < 100 mg/dL, 1: 100–126 mg/dL, 2: >  = 126 mg/dL) as recommended by the American Diabetes Association. There are cut points for SBP (0: < 126 mm/hg; 1: 126–142 mm/hg; 2: >  = 142 mm/hg) and DSST (0: ≥ 42 points, 1: 30–42 points, and 2: < 30 points) as per the study reported by O’Connell et al. [24]. For creatinine and FEV1, tertile-based cutoffs were derived from the study participants’ baseline dataset, creatinine (0: >  = 88.4 mg/dL; 1: 70.7–88.3 mg/dL; 2: < 70.7 mg/dL) and FEV1 (male 0: < 0.87 L/sec; 1: 0.88–1.61 L/s; 2: > 1.61 L/s and female 0: < 0.99 L/s; 1: 1.0–1.72 L/s; 2: >  = 1.73 L/s). The five components were then summed to give a score of 0–10 for each participant at each time point.

Statistical analysis

Sample size estimation

The sample size estimation has been presented in detail in the study protocol [18]. It aligned with the assumptions of the global statistical tests. We preferred to choose a sample size of n1 − n2 = 100, in line with the adequate sample size proposed for 10 multiple outcomes, under the simulations reported for sample size calculations with computation of the global treatment effect under various possible settings. Further, assuming an attrition rate of 25% over 6 months, we derived a sample size of n = 250 for the present study.

Data were analyzed with IBM SPSS version 27 (IBM Corp., Armonk, NY) and R Studio (version 2023.06.01). Continuous variables were tested for normality with the Shapiro–Wilk test. We used descriptive statistics with mean (standard deviation) or number (percentages) for representation of baseline characteristics. Between group differences for the baseline demographic characteristics were compared using the χ² test of independence for categorical variables and the Mann–Whitney U test for continuous variables. There was only 14.7% loss of data at follow-up, with no compromise in the power of the study [18]; hence, imputation of data was not deemed necessary, and complete case analysis was done removing the dropouts from analysis. Statistical significance was accepted in 2-sided tests at p < 0.05 and that with p < 0.01 as highly significant. We referred to the Cohen’s guidelines for the classification of effect sizes based on coefficient β = 0.10–0.29 as small, 0.30–0.49 as medium, and 0.50 as large [33]. Adjusted mean differences (AMDs) were presented and 95% confidence intervals (CI) were reported. Assessment of influence of YBI on frailty was done using logistic regression, with frailty status as binomial variable (at risk of frailty vs. non-frail, with the latter used as a reference category).

Subgroup analyses were conducted to examine modification of the intervention effect by stratification of age, gender, and comorbidities. Forest plots depicted interaction effects and reported as P_interaction. Interaction terms were constructed by multiplying the intervention and the grouping variables.

Global statistical tests (GSTs)

Primarily, the GST was conducted using generalized estimating equations (GEEs) to take into account the correlation among the multiple markers of the healthy aging panel. The analyses of GEE were performed using R Studio version 4.3.1. “glmtoolbox” package was used to perform the GEE analysis with cluster correlated data and “metan” package for the correlation plot. We used the population-averaged or marginal model of the generalized estimating equations (GEE) to find the overall effect of the regressors on the outcome variable with different scales. Further, the primary study parameters were converted to z-scores before conducting all analyses to place them on the same metric. The z-scores of the variables were calculated using the baseline means and the standard deviations of the combined values. The correlation among the dependent variables was considered by defining the best correlation structure in the GEE model. The final model was developed by refitting the regression coefficients to correct the correlation between the variables to obtain summarized parameter estimate of the β coefficient.

The global effect of the yoga intervention was also assessed using the rank-transformed analysis of covariance based on the O’Brien’s rank-sum test [34]. Outcome-specific ranks were generated separately for baseline and intervention values of each of the ten outcome markers. This was followed by the calculation of overall rank sums for each subject by summing their outcome-specific ranks. The differences between groups were then evaluated by running a general linear model with follow-up values based overall rank sums as the dependent variable, adjusted for the study covariates and the baseline values of the overall rank-sums [33]. Under sensitivity analysis, we performed complete case analysis to check if results of primary outcome were affected by missingness in data, and we also performed multivariate ANCOVA, another method of global test, to derive the Hotelling’s T² statistic, to report the overall mean differences across groups.

Results

Baseline characteristics of participants

The flow of participants is presented in Fig. 1. Among the 484 screened individuals, 162 (33.47%) did not meet the eligibility criteria, 39 (8.06%) refused to participate, and 25 (5.16%) did not come for baseline assessment and an overall 258 (53.32%) individuals were randomized into yoga-based intervention (n = 132) and waitlist control (n = 126) within 15 days (about 2 weeks) of the baseline assessment. Women representation was lower in eligible, non-participants compared to the study participants (Supplementary Table 5). At baseline, participants were between the ages of 60–80 years old (median: 66 years (IQR = 63–70)), and were predominantly women (55.1%), with 8 years of education (median: 8 (IQR = 5–10)) and 98.42% are married. Most of the participants were never smokers (93.79%) but had high prevalence of dyslipidemia (95.34%), hypertension (41.47%), and diabetes (67.82%). Demographic profiles of yoga and waitlist groups (Table 1) were evenly distributed between groups (p > 0.05) except for generalized obesity status, with higher proportion in YBI compared to WLC groups (p = 0.007).

Fig. 1.

Trial profile based on the consolidated standards of reporting trials guidelines for transparent reporting of trials

Table 1.

Participant demographics by group

Demographic variable	Total (n = 258)	Yoga (n = 132)	Waitlist control (n = 126)	p-value
Age in years, median (IQR)	66 (63–70)	66 (63–69)	66.5 (63–70)	0.53
Women, n (%)	142 (55.04)	71 (53.78)	71 (56.34)	0.68
Years of education, median (IQR)	8 (5–10)	8 (5–10)	8 (5–10)	0.84
Currently employed, n (%)	126 (48.84)	59 (44.69)	67 (53.17)	0.17
Marital status, n (%)
Single	4 (1.55)	2 (1.51)	2 (1.58)	0.88
Currently married	216 (83.72)	109 (82.57)	107 (84.92)
Divorced	7 (2.71)	3 (2.27)	4 (3.17)
Widow/widower	31 (12.01)	18 (13.63)	13 (10.32)
Living alone	16 (6.20)	7 (5.30)	9 (7.14)	0.54
Smoking, n (%)
Never	242 (93.79)	127 (96.21)	115 (91.26)	0.28
Former	7 (2.71)	2 (1.51)	5 (3.97)
Current	9 (3.48)	3 (2.27)	6 (4.76)
Lifestyle diseases, n (%)
Hypertension	107 (41.47)	58 (43.93)	49 (38.88)	0.48
Diabetes mellitus	175 (67.82)	86 (65.15)	89 (70.63)	0.35
Generalized obesity	81 (31.39)	51 (38.63)	30 (23.81)	0.01*
Dyslipidemia	246 (95.34)	126 (95.45)	120 (95.24)	0.93
Medication, n (%)
Anti-hypertensive	101 (39.14)	58 (43.93)	43 (34.13)	0.15
Blood glucose lowering	88 (34.11)	50 (37.87)	38 (30.16)	0.26
Lipid altering	62 (24.03)	38 (28.78)	24 (19.04)	0.44
MNA scores, median (IQR)	12 (11–13)	12 (11–13)	12 (11–13)	0.18

Baseline characteristics were summarized using percentages for categorical variables and median in interquartile range for continuous variables; chi-square tests were performed for categorical variables, and Mann–Whitney U test for continuous variables. MNA Mini Nutritional Assessment. Hypertension (either SBP >  = 140 or DBP >  = 90); diabetes mellitus (glycated hemoglobin >  = 6.5); generalized obesity (BMI ≥ 25 kg/m²). Dyslipidemia (LDL-C ≥ 130 mg/dL, or TC ≥ 200 mg/dL or TG ≥ 150 mg/dL or low levels of HDL-C < 40). *p < 0.05

Intervention compliance and adherence

After an average intervention period of 30 weeks, data on primary outcomes were available for 85.3% (n = 220) participants (YBI-119, WLC-101). Change of residence, non-availability for follow-up, and ill health were the most common reasons for dropping out (n = 38) (Fig. 1). The overall attrition rate was 13%. There was an overall rate of adherence of 90.80% based on class attendance The rate of adherence to the intervention was found to be high in the initial 12 weeks of the program (92.47%) compared to the later 12 weeks (90.42%).

Primary outcome

Testing of multiple outcomes has been considered a challenge in clinical trials; hence, we deemed the use of global statistical to present a proof of concept in the study. Using the GEE, we found a statistically significant and overall beneficial impact of the YBI vs. control for the healthy aging panel of multiple markers, β = 0.29 (95% CI = 0.20 to 0.38, p < 0.001). The GEE analysis also indicated significant associations between healthy aging model and other covariates such as hypertension and obesity HAP (Supplementary Table 7 and Supplementary Fig. 1). Additionally, we found a significantly substantial overall beneficial impact of the intervention, β = 0.28 (95% CI = 0.19 to 0.36, p < 0.001), obtained from the rank-sum test adjusted for baseline values of covariates.

Secondary outcomes

When dissected for the individual markers of HAP (Table 2), yoga intervention was found to have the most considerable statistically significant effect on physical domain of WHOQoL-BREF (β-coefficient = 0.48; 95% CI = 0.38 to 0.59; p < 0.001) followed by DSST (β-coefficient = 0.41; 95% CI = 0.33 to 0.49; p < 0.001), FEV1 (β-coefficient = 0.27; 95% CI = 0.16 to 0.38; p < 0.001), SBP (β-coefficient =  − 0.25; 95% CI =  − 0.33 to − 0.16; p < 0.001), hand grip strength (β-coefficient = 0.20; 95% CI = 0.14 to 0.26; p < 0.001), gait speed (β-coefficient = 0.20; 95% CI = 0.08 to 0.32; p = 0.001), other domains of WHOQoL-BREF like psychological health (β-coefficient = 0.19; 95% CI = 0.09 to 0.29; p < 0.001), social relationships (β-coefficient = 0.18; 95% CI = 0.10 to 0.27; p < 0.001), glycated hemoglobin, HbA1c (β-coefficient =  − 0.17; 95% CI =  − 0.27 to − 0.08; p < 0.001), TMT B/A (β-coefficient =  − 0.14; 95% CI =  − 0.25 to − 0.04; p = 0.007), environmental health (β-coefficient = 0.13; 95% CI = 0.05 to 0.20; p < 0.001), UCLA loneliness score (β-coefficient = 0.13; 95% CI = 0.03 to 0.23; p = 0.009), and LDL-C (β-coefficient =  − 0.08; 95% CI =  − 0.15 to − 0.01; p = 0.035). Additional secondary markers have also been tested (Table 3) and there were significant differences with respect to Klotho z-scores (β-coefficient = 0.06; 95% CI = 0.02 to 0.09; p = 0.007). Yoga intervention was also found to be protective against frailty (odds ratio = 0.35, 95% CI = 0.19 to 0.56; p < 0.05). Significant difference was not found in composite scores of inflammatory markers (β-coefficient =  − 0.04; 95% CI =  − 0.16 to 0.09; p = 0.14).

Table 2.

Effect of yoga intervention on the primary outcome, secondary outcomes and adapted composite scores

Variable	Baseline		Follow-up		Effect size estimates
	Yoga (n = 119)	Waitlist control (n = 101)	Yoga (n = 119)	Waitlist control (n = 101)	Adjusted mean difference (95% CI)	β-coefficient (95% CI)
Global statistical tests
Generalized estimating equations	0.03 (0.34)	− 0.06 (0.37)	0.12 (0.29)	− 0.16 (0.37)	0.20 (0.14 to 0.26)	0.29 (0.20 to 0.38)**
Rank-sum test	1325.57 (264.20)	1242.89 (274.28)	1389.45 (231.56)	1175.69 (269.10)	147.89 (102.15 to 193.63)	0.28 (0.19 to 0.36)**
Secondary markers
Physiological and metabolic health
Glycated hemoglobin (%)	6.91 (0.76)	6.84 (0.79)	6.70 (0.69)	6.90 (0.89)	− 0.29 (− 0.45 to − 0.13)	− 0.17 (− 0.27 to − 0.08)**
LDL-c (mg/dL)	111.21 (27.62)	109.39 (28.72)	111.01 (24.32)	112.98 (28.61)	− 3.93 (− 7.58 to − 0.27)	− 0.08 (− 0.15 to − 0.01)*
Systolic blood pressure (mm/Hg)	136.65 (15.97)	134.66 (19.72)	126.97 (9.85)	132.78 (13.76)	− 6.43 (− 8.68 to − 4.19)	− 0.25 (− 0.33 to − 0.16)**
FEV1 (L/s)	1.31 (0.59)	1.34 (0.64)	1.83 (0.48)	1.53 (0.55)	0.30 (0.17 to 0.42)	0.27 (0.16 to 0.38)**
Cognitive function
DSST score	34.52 (4.61)	34.16 (4.41)	37.81 (4.96)	33.17 (4.32)	4.26 (3.39 to 5.12)	0.41 (0.33 to 0.49)**
TMT-B/A (ratio)	2.64 (0.29)	2.62 (0.26)	2.59 (0.29)	2.64 (0.25)	− 0.07 (− 0.13 to − 0.02)	− 0.14 (− 0.25 to − 0.04)*
Physical capability
Hand grip strength (kg)	21.73 (6.56)	20.48 (6.79)	24.42 (6.21)	20.45 (6.87)	2.71 (1.91 to 3.49)	0.20 (0.14 to 0.26)**
Gait speed (m/s)	0.80 (0.19)	0.85 (0.18)	0.89 (0.19)	0.83 (0.21)	0.09 (0.03 to 0.14)	0.20 (0.08 to 0.32)**
Psychological well-being
Physical health	59.31 (3.53)	59.61 (3.54)	62.82 (3.51)	58.27 (4.58)	4.55 (3.56 to 5.54)	0.48 (0.38 to 0.59)**
Psychological health	51.68 (5.75)	51.27 (5.96)	56.11 (6.27)	53.12 (5.68)	2.34 (1.21 to 3.48)	0.19 (0.09 to 0.29)**
Social relationships	42.09 (9.89)	41.66 (10.13)	49.75 (9.15)	46.51 (9.27)	3.45 (1.87 to 5.04)	0.18 (0.10 to 0.27)**
Environmental health	59.23 (7.51)	56.49 (8.02)	62.80 (7.00)	58.73 (7.68)	1.93 (0.78 to 3.08)	0.13 (0.05 to 0.20)**
Social well-being
Loneliness score	46.74 (6.82)	43.60 (6.67)	45.43 (6.08)	41.68 (6.36)	1.78 (0.46 to 3.09)	0.13 (0.03 to 0.23)*
Adopted composite scores
HATICE adopted composite	0.07 (0.60)	− 0.10 (0.63)	− 0.05 (0.58)	− 0.02 (0.64)	− 0.18 (− 0.26 to -0.09)	− 0.16 (− 0.23 to − 0.09)**
Healthy aging index (HAI)	5.39 (1.60)	5.12 (1.51)	5.77 (1.47)	5.97 (1.71)	− 0.33 (− 0.63 to -0.02)	− 0.09 (− 0.19 to 0.00)*
Additional HAI markers
Serum creatinine (mg/dL)	1.09 (0.15)	1.08 (0.13)	1.01 (0.13)	1.02 (0.14)	− 0.01 (− 0.05 to 0.03)	− 0.03 (− 0.13 to 0.08)
Fasting blood glucose (mg/dL)	144.19 (52.38)	125.94 (26.29)	148.35 (52.24)	130.11 (26.05)	0.12 (0.03 to 0.22)	0.01 (− 0.01 to 0.34)
Additional HATICE markers
BMI (kg/m2)	23.68 (6.36)	21.67 (4.48)	23.90 (5.71)	21.81 (3.99)	0.34 (− 0.01 to 0.69)	0.03 (− 0.00 to 0.07)

Baseline and post-intervention data are presented as mean (standard deviation); group differences in the primary outcomes were analyzed by global statistical tests (GSTs); generalized estimating equations (GEE) and rank-sum tests (RST). Data of GST’s are presented as mean (standard deviation) of composite z-scores for GEE and composite ranked score for RST. Effect sizes are reported as AMD and beta coefficient (β), analyzed with the univariate analysis. Models are adjusted for baseline values of variables and other study covariates (age, gender, education, employment status, marital status, smoking status, comorbidities, and medications). Healthy aging index (HAI) scores are computed as sum of cutoff-based values of five healthy aging markers (fasting blood glucose, creatinine, DSST, SBP, and FEV1). HATICE trial adopted scores are composite of z-scores of BMI, SBP, and LDL-c. LDL-c low-density lipoprotein-cholesterol, FEV1 forced expiratory volume in 1 s, DSST digit symbol substitution task, TMT B/A trail making task; *p < 0.05, **p < 0.001

Table 3.

Effect of yoga intervention on additional secondary markers at baseline and follow-up across groups

Variable	Baseline		Follow-up		Effect size estimates
	Yoga (n = 119)	Waitlist control (n = 101)	Yoga (n = 119)	Waitlist control (n = 101)	Adjusted mean difference (95% CI)		β-coefficient (95% CI)
Systemic integrity
S-Klotho z-scores	− 0.10 (0.88)	0.12 (1.11)	− 0.05 (0.92)	0.07 (1.08)		0.12 (0.04 to 0.19)	0.06 (0.02 to 0.09) *
Inflammatory markers
Composite Z-score (TNF-RII, CRP)	0.02 (1.00)	0.03 (0.96)	− 0.12 (0.95)	− 0.05 (0.98)		− 0.08 (− 0.33 to 0.18)	− 0.04 (− 0.16 to 0.09)
Auxiliary blood biochemistry
Blood urea (mg/dL)	20.87 (5.24)	23.76 (16.68)	23.54 (4.38)	24.52 (5.32)		− 0.19 (− 1.36 to 0.99)	− 0.02 (− 0.14 to 0.09)
Total WBC (count/µL)	7645 (1631)	7964 (1352)	7585 (1486)	8031 (1617)		− 155.79 (− 371.76 to 60.18)	− 0.05 (− 0.12 to 0.02)
Mean corpuscular volume (fl)	88.33 (4.78)	85.82 (4.68)	86.86 (4.97)	86.95 (4.69)		− 2.36 (− 3.26 to − 1.46)	− 0.20 (− 0.29 to − 0.12) **
Mean corpuscular hemoglobin (pg)	32.44 (10.18)	28.57 (8.34)	33.61 (9.53)	29.78 (8.45)		− 0.07 (− 0.99 to 0.85)	− 0.01 (− 0.06 to 0.05)
MCHC (g/dL)	33.83 (2.73)	35.36 (7.75)	34.53 (3.45)	36.21 (8.04)		− 0.39 (− 1.31 to 0.53)	0.03 (− 0.10 to 0.04)
Neutrophils (count/µL)	65.45 (5.56)	65.79 (5.23)	66.66 (5.52)	67.09 (5.78)		0.26 (− 0.78 to 1.29)	0.02 (− 0.07 to 0.11)
Lymphocytes (/µL)	26.75 (5.42)	26.29 (6.55)	28.45 (5.41)	27.68 (5.74)		0.18 (-0.98 to 1.35)	0.02 (− 0.08 to 0.12)
Monocytes (%)	6.35 (1.30)	5.43 (2.45)	6.48 (1.13)	5.66 (2.29)		0.00 (-0.18 to 0.18)	0.01 (− 0.05 to 0.05)
Eosinophils (%)	2.55 (1.01)	2.62 (1.12)	2.54 (0.60)	2.74 (0.73)		− 0.16 (− 0.27 to − 0.05)	− 0.12 (− 0.20 to − 0.04)
Basophils (%)	0.28 (0.24)	0.28 (0.23)	0.34 (0.24)	0.36 (0.25)		− 0.03 (− 0.08 to 0.03)	− 0.06 (− 0.17 to 0.06)
RBC count (million cells/µL)	4.47 (0.44)	4.47 (0.41)	4.74 (0.70)	4.61 (0.41)		0.15 (0.00 to 0.29)	0.13 (0.00 to 0.26)
Platelet count (lakhs/mm3)	2.27 (0.62)	2.35 (0.63)	2.61 (0.57)	2.59 (0.58)		0.08 (− 0.01 to 0.18)	0.07 (− 0.00 to 0.15)
Hemoglobin (g/dL)	12.80 (1.48)	12.92 (1.22)	12.23 (1.18)	12.20 (1.19)		0.06 (− 0.16 to 0.29)	0.03 (− 0.07 to 0.12)

Baseline and post-intervention data are presented as mean (standard deviation); group differences are analyzed by general linear model adjusting for baseline values of variables and other study covariates (age, gender, education, employment status, marital status, smoking status, comorbidities, and medications). Effect size is represented as the beta coefficient (β) and adjusted mean difference (AMD) along with confidence interval (CI), BMI body mass index. *p < 0.05, **p < 0.001

Subgroup analysis

Subgroup analyses were performed according to age, gender, and comorbidities (Fig. 2), though no significant interactions with the study groups’ trends of improvement were found in age group <  = 66 years (β-coefficient = 0.29; 95% CI = 0.18 to 0.41, p < 0.001), in females (β-coefficient = 0.32; 95% CI = 0.20 to 0.42, p < 0.001), and with presence of diabetes (β-coefficient = 0.31; 95% CI = 0.22 to 0.40, p < 0.001) and absence of hypertension (β-coefficient = 0. 0.32; 95% CI = 0.20 to 0.45, p < 0.001).

Fig. 2.

Subgroup analysis of the primary outcome

Exploratory analyses

An overall global benefit of the yoga-based intervention on the multiple domains of healthy aging was also established with two separate analyses, HATICE adopted composite (mean difference =  − 0.18; 95% CI =  − 0.26 to − 0.09, p < 0.001) and HAI (mean difference =  − 0.33; 95% CI =  − 0.63 to − 0.02, p = 0.03) as compared to waitlist control (Table 2). In relation to sensitivity of the global test for primary outcome analysis, we also conducted multivariate Hotelling’s multivariate test which yielded significant results (T² = 1.40, p < 0.001).

Adverse effects

A few participants presented with mild pain in shoulder and lumbar region (n = 12) during yoga sessions. The complaints were intervened by corrective possess (lumbar stretch for back pain) and alternate nostril breathing at the end of the session.

Discussion

We present findings from a randomized clinical trial on the overall benefit of a 26-week yoga intervention on healthy ageing phenotype. Using the GEE model of the global test, there is a summarized statistical effect of β = 0.29, 95% CI = 0.20 to 0.38 for a 26-week YBI over WLC, taking into account the correlations between the outcomes. The findings were also confirmed using rank-sum-based global test, wherein a similar magnitude of the effectiveness of the intervention could be established β = 0.28, 95% CI = 0.19 to 0.36. The observed effect sizes of β, 0.28 and 0.29, accord with small intervention effect [33] that could be explained by short duration of the intervention and a sample size estimate based on exploratory hypotheses, which could have also compromised the power of the trial. Moreover, we have adjusted the apparent correlations among the markers in the global statistical test; the residual unadjusted correlations could indicate a synergism in the overall influence of the intervention. There were unexpected associations observed between GEE model for healthy aging and presence of hypertension and obesity, explained by the plausible beneficial effects of certain drugs such as recently reported effect of antihypertensive drug, rilmenidine, on health/lifespan in Caenorhabditis elegans [35]. However, this hypothesis definitely needs further validation.

To the best of our knowledge, this is the first study to use a global test strategy to measure the holistic effectiveness of a lifestyle-based intervention. The global tests used in the study provide a critical assessment of treatment efficacy and chosen markers of HAP are potentially applicable in community-based intervention studies, expected to change with age, capable of predicting ageing-related phenotypes, and amenable to modification by lifestyle interventions [16, 24, 25] since there is no criterion reference for assessing healthy ageing, and no single outcome could be deemed sufficient to capture the concept of healthy ageing. Overall, the trial’s findings support that targeting the complex multidimensional construct of healthy ageing [16] is feasible and findings of this trial provide a viable platform for future comparisons of similar interventional research.

The beneficial influence of our interventional research on healthy ageing phenotype aligns with observational data from a national community-based sample in India in 2017–2018, wherein daily yoga practice was associated with 1.3 odds of successful ageing (adjusted OR = 1.34, 95% CI = 1.11 to 1.61) among older adults (≥ 65 years) [36]. However, the conceptual difference between the terms “healthy ageing (HA)” and “successful ageing (SA)” limits the direct comparison of the effect sizes; HA primarily captures the essence of physical and cognitive functional preservation [37] but without the requirement of disease avoidance, which otherwise forms the base for SA.

As recommended earlier [38], we also ran secondary tests of individual outcomes to support the trial’s primary hypothesis. In terms of statistical significance, there were significant improvements in HbA1c status (β =  − 0.17; 95% CI =  − 0.27 to − 0.08), DSST (β-coefficient = 0.41; 95% CI = 0.33 to 0.49), FEV1 (β-coefficient = 0.27; 95% CI = 0.16 to 0.38), and SBP (β-coefficient =  − 0.25; 95% CI =  − 0.33 to − 0.16). We also reviewed the significance of the observed changes regarding available minimum clinically significant differences (MCIDs). The observed reduction in HbA1c was at par with the reported MCIDs [39] of 0.3–0.5%. Similarly, there was an improvement by approximately four symbols in DSST (AMD = 4.26, 95% CI = 3.39 to 5.12) which were in the range of the recently proposed minimal clinically significant difference (MCID) of 3–5 symbols (though the MCID has been set for older adults who have fallen [40]). We also observed moderate improvements in FEV1, the lung function marker (AMD = 0.30; 95% CI = 0.17 to 0.42) aligning with the reported benefit of yoga-based respiratory training on pulmonary function, maximum expiratory and inspiratory pressures (PEmax and PImax, respectively) in the older adults as written by Santaella et al. [41]. Though lower than the proposed MCID [41], the mean differences in hand grip strength (AMD = 2.71, 95% CI = 1.91 to 3.49) are in line with the reported standardized mean differences (SMD) for exercise training interventions (SMD = 0.28, and SMD = 4.12) [42, 43]. Concurrent improvements in variables like grip strength and SBP indicate the need to conduct larger trials to establish the strength training potential of yoga for older adults. These results become more intriguing given the relevance of the reported lower hypotensive responsivity of the older adults to strength training exercises [43]. Our results also support the previously reported influence of yoga on improvement of quality of life [44] and subjective well-being [45] in older adults.

Further, we confirmed the potency of the intervention by adopting the indices used in contemporary clinical trials to assess healthy ageing. The healthy ageing index is an epidemiologically established predictor of mortality and comorbidities [24]. Compared with the results of the recently reported secondary analysis of intentional weight loss intervention on HAI [46], our study’s mean difference (AMD =  − 0.33; 95% CI =  − 0.63 to − 0.02) is smaller in magnitude by 0.2 scores (0.5 vs 0.3, respectively). The significant improvements in HAI could be attributed to improvements in cognition (DSST), blood pressure (SBP), and pulmonary (FEV1) markers.

In healthy ageing through internet counselling in the elderly (HATICE), a multinational, randomized controlled trial has indicated modest improvement of cardiovascular risk scores (mean difference =  − 0·05, 95% CI − 0·08 to − 0·01) composed of the standardized composite score (z-score) of systolic blood pressure, LDL cholesterol, and body mass index (BMI) in an older population by coach-supported internet-based self-management [25]. Using the same composite index, we observed a higher mean difference of − 0.18 (95% CI =  − 0.26 to − 0.09) over 26 weeks that could be due to the inflation attributed by non-active control group. The observed trends of improvements in the health metrics support the potential of yoga to positively modulate the trajectory of ageing rather than just attenuating the health decline.

There was a paralleled upregulation in the Klotho levels, which has been considered a system integrator with pleiotropic functionality related to cognition, vascular function, physical strength, and metabolism with an overarching holistic modality underpinning the hallmarks of ageing [47]. The significant increase in an exerkine [48], Klotho paralleled with trends of decline in the inflammatory status [46], supports a mechanical ageing framework. These findings prompt us to investigate further the mediation effects of YBI in future studies powered for causative analysis [49].

Strengths and limitations

Adopting the United Nations Decade of Healthy Ageing resolution, this study provides a significant thrust to clinical research testing the effectiveness of lifestyle-based interventions for measurable metrics of healthy ageing. The primary strength of the study was the use of a comprehensive global statistical test to present the overarching effectiveness of the intervention on the health of older adults. Including subjects with lifestyle diseases is a valuable addition to the existing literature on older adults, given the inclusion of mainly healthy subjects in the previously reported ageing clinical trials, wherein these morbidities often are part of the exclusion criteria. The study is limited by the short term of the intervention, which determines the ascertainment of the sustenance of the effects. On a distinct note of short intervention, the observed high impact of yoga over 6 months is a significant aid to bring rapid beneficial changes in the health of older adults. We propose a translational model with short intermittent intervention bouts to improve health that could be implemented in older adults, which could also overcome the non-adherence issues of lifestyle interventions. Conceived as a lifestyle intervention, the sessions in the YBI group included varied aspects of yogic philosophy, including the practice of physical postures, regulated breathing, and meditation, along with dietetic advice as an integral component. We observed an increase in BMI in both the YBI and WLC groups, where diet could be a confounding variable. Since we had only given dietary advice to the YBI group as part of the intervention, and neither of the groups was monitored and assessed for nutritional practices, it could be a limitation in the study.

The clinical relevance of the findings also remains underscored, given the uncertainty in the definition of healthy ageing. Additionally, an inactive control condition could have inflated the effect size in this study. However, it was deemed appropriate to include the same to actuate the effects of yoga realistically.

Conclusion

The global effect of YBI across multiple ageing-related outcomes provides a proof of concept for its use as a public health intervention to promote healthy aging in older adults. The trial’s findings support that the complex multidimensional construct of healthy ageing could be captured successfully in intervention research. Given the accelerated pace of population aging across the globe and limited interventional research on healthy aging, finding holds a great translational value and highlights the need for public health intervention with a “never too late” attitude. The findings also need to be extended to institutionalized older adults, and to those living in old age homes, and with multimorbid conditions.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

All authors contributed to the study conception and design. Data collection and analysis were performed by Atmakur Snigdha and Amrutha Jose. The first draft of the manuscript was written by Vijaya Majumdar and Atmakur Snigdha and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

GMR Airports Infrastructure Limited.

Data availability

Data will be available on reasonable request for research proposals. Corresponding author will provide the deidentified participant data at vijaya.majumdar@svyasa.edu.in.

Declarations

Ethical approval and consent to participate

Ethics approval was obtained from the Institutional ethical committee (Swami Vivekananda Yoga Anusandhana Samsthana) and signed informed consent is obtained from participants.

Consent for publication

Not applicable.

Conflict of interest

The authors declare no competing interests.