Yoga-based lifestyle treatment and composite treatment goals in Type 2 Diabetes in a rural South Indian setup- a retrospective study

This multicentre retrospective study examined the effects of adjunct yoga-treatment in achieving composite cardiovascular goals for type 2 diabetes (T2D), set forth by the American Diabetes Association (ADA) in rural Indian settings. Records were extracted for 146 T2D patients, aged ≥20–70 years, and treated under the “Apollo Total Health Programme” for rural diabetes management, for the period April 2016 to November 2016. The study cohort comprised of two treatment groups (n = 73 each); non-yoga group (standard of care) and yoga group (adjunct yoga-treatment). Propensity score matching was applied between the study groups to define the cohort. Composite cardiovascular scores were based on the combination of individual ADA goals; A1c < 7%, blood pressure (BP) < 140/90 mmHg, stringent BP (<130/80 mmHg) and lipid, LDL-C < 100 mg/dl [risk factor for atherosclerotic cardiovascular disease]. Logistic regression was used to compare between the two treatment groups. Compared to standard of care, adjunct yoga-treatment was found to significantly facilitate the attainment of ADA composite score by 8-fold; A1c, ~2-fold; LDL-C, ~2-fold; BP < 140/90 mmHg and <130/80 mmHg by ~8-and ~6-fold respectively. This study provides the first evidence for significant efficacy of adjunct yoga-treatment for the attainment of favourable treatment goals for T2D in rural Indian settings. Clinical Trial Registration Number: CTRI/2020/02/0232790


Results
Baseline Characteristics. Records of a total of 146 T2D patients were retrieved. The mean age of the study cohort was 55.61 ± 10.90 years, the majority were female 64.40% (n = 94) , and 76.03% (n = 111) belonged to low socioeconomic status. The recruited study cohort had an average duration of diabetes of ~6 years. According to the ADA criteria, 54.79% (n = 80) of the total study cohort was found to be above recommended A1c targets (≥7.0%, 53 mmol/ml), 81.51% (n = 119) above BP1 targets (≥130/80 mmHg), 47.26% (n = 69), above BP2 targets (≥140/90 mmHg) and 47.94% (n = 70) above lipid targets (LDL ≥ 100 mg/dl) 19 . Overall, at baseline, 93.83% (n = 137) and 85.62% (n = 125) of the study cohort was found to be above combined ADA composite scores 1 and 2, respectively. The cohort was also observed to have generalized obesity with a mean BMI of 26.69 ± 4.58 Kg/ m 2. Importantly, 85.62% (n = 125) of the study cohort was found to be overweight/obese according to the Asian cut off for BMI (≥23 kg/m 2 ) 20 . At baseline, subjects in the yoga group had significantly lower DBP levels than the non-yoga group (yoga, 80.66 ± 9.30 mmHg vs. 84.52 ± 10.12 mmHg; P = 0.032) but were similar to the control group with respect to other parameters ( Table 1).
With respect to the status of A1C goals of ADA (<7%), 46.58% (n = 34) subjects were found to meet criteria for in the yoga group at baseline, however, the percentage increased to 54.79% (n = 40) at follow-up (Table 2). In the non-yoga group, the percentage of subjects with A1c criteria decreased from 43.84% (n = 32) to 36.99% (n = 27) ( Table 2). The difference in the distribution between the patients meeting ADA criteria for A1c was statistically significant between yoga and non-yoga treatment groups at the follow-up (P = 0.046). When analysed by multiple logistic regression, modelled by covariates, age, sex, duration of diabetes, socioeconomic status, baseline A1c values, yoga treatment was found to be significantly associated with the ~2-fold (OR = 2.44, 95% CI = 1.19-5.00, P = 0.015) higher chances of attainment of favourable A1c cut off (<7%) as compared to standard of care ( Table 2).
The percentage of subjects who met the ADA-criteria with respect to favourable LDL-C, < 100 mg/dl, increased from 52.05% (n = 38) to 54.79% (n = 40) in the yoga group (Table 2). However, in the non-yoga group, there was a decrease from 52.05% (n = 38) to 38.36% (n = 28) in the number of subjects who met the LDL-C criteria ( Table 2). The distribution of patients with favourable LDL-C values was not significant between yoga and non-yoga groups at the follow-up (Table 2). However, when analysed by logistic regression, adjusted for covariates and, baseline lipid status, yoga treatment was found to be significantly associated with the ~2-fold (OR = 2.22, 95% CI = 1.06-4.68, P = 0.035) increased chances for the attainment of favourable LDL-C outcome (<100 mg/dl) as compared to standard of care alone (Table 2).
We assessed the BP outcomes with old and revised favourable cut-offs recommended by ADA (Table 3). When analysed with old cut-off (<130/80 mm Hg), we could observe a pronounced increase in the percentage of subjects meeting the favourable BP outcome from 21.92% (n = 16) to 34.25% (n = 25) in the yoga group (Table 2). On the contrary, in the non-yoga group, the number of T2D patients who met BP criteria of <130/80 mm Hg decreased from 15.07% (n = 11) to 8.22% (n = 6) ( Table 2). When analysed by logistic regression, yoga treatment was found to be associated with ~6.4-fold (OR = 6.37, 95% CI = 2.24-18.08, P = 0.001) increase the chances of favourable BP cut-offs (<130/80) at follow-up. When analysed with revised new BP cut-off (<140/90 mm Hg), we could observe a pronounced increase in the percentage of subjects meeting the favourable BP outcome from 60.27% (n = 44) to 84.93% (n = 62) in the yoga group (Table 2), yoga treatment was also found to be associated with 8.28-fold (95% CI, 3.52-19.48, P < 0.0001) increased chances for the revised favourable BP cut-offs,. In the non-yoga group, the number of T2D patients who met BP criteria decreased from 45.21% (n = 33) to 39.73% (n = 29) ( Table 2).
We also analysed the status of cardiovascular control for the subgroup of study cohort with uncontrolled diabetes (A1c ≥ 8.0%), n = 44. We could observe 63.16% success towards attainment of lipid goal (LDL < 100 mg/dl) and 26.32% for BP targets (130/80 mmHg) 89.47% for BP target (140/90) by 6-months of yoga treatment (data not shown). However, the controls exhibited deterioration with respect to these goals (data not shown).
Effect of yoga treatment as compared to standard of care was demonstrated with respect to the attainment of favourable BMI cut-off (<23 Kg/m 2 ) for Asians ( www.nature.com/scientificreports www.nature.com/scientificreports/ was found to be associated with 62-fold (OR = 61.73, 95% CI = 3.19-1193) increased chances of attainment of the favourable BMI cut-off over a period of 6 months ( Table 2).
Outcomes in continuous measures. Over the study period of around 6-months, the yoga-group exhibited significant within-group beneficial mean changes and percent changes in A1c, −0.50%, (−5.03%); FBS,  Table 3). With respect to triglyceride (TG), we could observe an unexpected increase in the mean TG levels in the yoga group, 11.74 ± 3.72 mg/dl (31.94%) ( Table 3). We observed pronounced worsening of the metabolic variables in the non-yoga group (Table 3). We could observe a deteriorating trend in the mean difference of these variables from baseline in the non-yoga group (Table 3). Significant within-group differences were also observed in the non-yoga group for FBS, 11.14 mg/dl (18.31%); BMI, 0.76 Kg/m 2 (4.06%); HDL-c, −2.51 mg/dl (−4.72%), and TG, 67.70 mg/dl (59.01%). Between-study group differences between yoga group and non-yoga group very significant with respect to all the studied parameters (Table 3).

Discussion
Type 2 diabetes is associated with vascular complications and enhanced risk of cardiovascular events. Therefore, ADA has suggested a multifactorial targeted approach towards efficient management of T2D [glycemic, lipid, and blood pressure] 5,7 . Use of such composite endpoints in clinical studies also help in the better understanding of the net effect of an intervention or a therapy rather than individual endpoints 21,22 . Further, studies targeting composite endpoints have been reported to have higher statistical efficiency as compared to those with individual endpoints 22 . We hereby highlight the grim status of diabetes management with respect to the attainment of the ADA laid primary treatment goals, in rural Indian settings. At baseline of the study, 54.79% was found to be above ADA laid A1c targets (≥7.0%), 47.26% was found to be above lipid targets (LDL-C > 100 mg/dl) and 81.51% were found to be above BP cut-offs of>130/80 mmHg, and 47.26% were found to be above the revised BP cut-offs (<140/90 mmHg) 5 . Overall, only 6.16% (n = 9) of the total study cohort was found to be meeting the composite   www.nature.com/scientificreports www.nature.com/scientificreports/ score of all the three treatment targets at the baseline. However, a prior report by Menon et al. 15 , indicated only 1-3% of Indian T2D population achieving the combined treatment goals of ADA in an urban clinical setup. The pathophysiological link between obesity and T2D was also evident in the study cohort, wherein, 62.33% (n = 91) of the cohort sample was found to be obese. As majority of the cohort sample (~80%), belonged to low socioeconomic status 23 , this poor status of cardiovascular risk control could be attributed to lack of pharmacologic management, governed by poor awareness and socioeconomic status in the rural Indian settings.
Lifestyle management plays an essential role in the efficient control of diabetes status 6,24 . Yoga, as a lifestyle intervention has been reported to lead to beneficial health outcomes related to cardiovascular and metabolic disorders including T2D 25,26 . Based on its high reported receptivity and cost-effectiveness, yoga holds a strong potential as a lifestyle management skill in Indian scenario 3 . This is the first report wherein the efficacy of yoga treatment was assessed in aiding the cardiovascular fitness with respect to achievement of ADA laid primary treatment goals of T2D in rural Indian settings. Our findings reflect a magnitude of success of 10 Table 3. Distribution of continuous variables between yoga and non-yoga treatment groups at baseline and follow up. Data are represented as mean ± SE (95% confidence interval) P-value of < 0.05 was set as significant between baseline and follow-up on within group and between group comparisons; CI, confidence interval, A1c; P value-Within study groups, Test statistics t-test; P value-Between study groups test statistic-ANCOVA. (2020) 10:6402 | https://doi.org/10.1038/s41598-020-63133-1 www.nature.com/scientificreports www.nature.com/scientificreports/ a success rate of 19% with respect to ADA composite endpoint over a period of 12 months on patients with uncontrolled T2D 27 . Further, compared to usual care, 6-months of yoga treatment was also found to be associated with 8-fold higher (OR = 8.22, 95% CI = 2.02-33.49) success towards the attainment of ADA composite scores (with revised BP cut-off of 140/90 mmHg) in rural Indian settings. With respect to individual composite goals, 6 months of yoga treatment was found to have a higher likelihood of attainment of A1c goal by ~2-fold (OR = 2.44, 95% CI = 1.19-5.00); LDL-C by ~2-fold (OR = 2.22, 95% CI = 1.06-4.68); blood pressure<140/90 mmHg by ~8-fold [OR = 8.28, 95% CI, 3.52-19.48)] in rural T2D population compared to standard of care. The observed 2-fold higher potential of yoga-treatment as compared to standard of care towards attainment of LDL-c targets in the rural Indian T2DM cohort deserves clinical attention. Control of dyslipidemia in Indian T2DM patients has been reported to be very poor; with almost half of them not reaching their LDL-C goal 28 . These findings are important as Coronary Artery Disease (CAD) mortality remains high in the Indian patients with T2DM. Similarly, the observed 8-fold increased impact of yoga on BP control as compared to standard of care is an important clinical outcome. Hypertension is a prevalent co-morbidity in T2D patients associated with an increased risk of cardiovascular events and mortality 29,30 . This coexistence has been reported to enhance the risks of nephropathy and retinopathy 31,32 .
As previously reported by Ikramuddin et al., intense lifestyle intervention could aid in 31% success towards attainment of glycemic, 70% for lipid and 70% for BP targets (130/80 mmHg) over a period of 12 months in a mixed ethnic population of uncontrolled T2D (HbA1c ≥ 8.0%) 27 . Interestingly, in the present subgroup of patients with uncontrolled T2D, HbA1c ≥ 8.0% yoga treatment of 6 months duration was found to be effective with 26.32% success in attainment of glycaemic control, 63.16% for lipid, and 26.32% for BP goals (130/80 mmHg). These findings indicate that if assessed for long-term effects, yoga treatment could match the magnitude of the potential of intense lifestyle-interventions as described for overweight/obese uncontrolled T2D patients 27. Glycemic control is the primary target of diabetes management strategies towards prevention of the devastating complications such as blindness, kidney failure and amputations 7,9 . We could observe significant absolute decrease in mean A1c% by 0.5 in the yoga group over a period of 6 months. The magnitude of reduction by 0.5% in A1c holds strong clinical significance, based on the reported epidemiological association between 1% reduction in the A1c value with 14% reduction in myocardial infarction (MI), 21% reduction in diabetes-related mortality and 37% reduction in microvascular complications 33 . Further significant and beneficial yoga treatment-induced mean percentage reductions in SBP (4.33%); DBP (5.66%); and TC (1.65%) could also be observed in the present study. These findings support earlier reported beneficial cardiovascular effects of yoga 25,26 .
Weight loss remains a major challenge in diabetes due to the complex interplay between metabolic, neuroendocrine and psychological factors 34 . In a prior study intensive lifestyle intervention of 1 year was reported to achieve an average 8.6% weight loss along with significant reduction of A1C, and CVD risk factors, with sustained effects up to 4 years 24 . In the present study, we could observe a 4.18% reduction in weight over a period of 6 months under yoga-treatment. Our results highlight the equivalent potential of yoga to intense weight-loss intended lifestyle interventions in overweight/obese T2D patients. A favourable and differential effect of 61-fold (OR = 61.73, 95% CI = 3.19-1193) was also observed for BMI outcome of <23 kg/m 2 in the yoga group against standard of care. The observed significant impact of yoga-treatment on BMI outcome in diabetes bears strong clinical relevance as weight management is an important component of efficient diabetes care 7 . Further, weight loss through lifestyle changes remains the first-line therapy for T2DM 34 . Available observational evidence suggest various clinical benefits of weight loss in diabetes including improvement in glycemic control, reduced risk of cardiovascular events alongwith improvements in quality of life, mobility, and physical function 35,36 . However, we could not assess the sustenance of weight-related effects of yoga in this short-duration study. Based on the proposed self-regulation modality of yoga, wherein yoga could lead to repatterning of hedonic neurocircuitries, we speculate sustained weight-loss effects with long-term yoga-treatment 37 .
The study is limited by its observational nature. The difference-in-means method used to establish the equivalence between the study groups could be limited in its capacity to control the confounding by baseline variables. To this end, we conducted propensity score matching with "nearest neighbour" method for matching of the treatment groups for key covariates. Further, logistic regression was also done to adjust for the effect of the covariates to assess the study outcomes. The observed poor outcome in the standard of care group deserves attention of physicians and clinicians working in the rural sector of India. The poor outcome could possibly be attributed to poor adherence to medication and prescribed physical activity in the rural T2D population suggesting that there is a need to explore strategies to facilitate adherence with the patients/caregivers 17 . Since Indian patients were found to be receptive to yoga, yoga-based treatment could be a pragmatic solution for effective diabetes management. Based on the epidemic proportions of T2DM in India, there is an urgent need to conduct a large, prospective, long-term study of the efficacy of yoga on attaining all of the ADA goals in the rapidly increasing T2DM population. Early initiation of yoga treatment to target adequate diabetes care has the potential to prevent the devastating complications including not just the microvascular but also loss of time from work and quality of life.

Methods
Cohort identification. The study was part of an ongoing service activity defined as the "Total Health Programme" (THP) India's first integrated rural healthcare service delivery network, initiated by the Apollo Group (https://www.apollohospitals.com/corporate/initiatives/csr-at-apollo/total-health-programme). THP aimed at patients for various diseases across the adopted villages for effective disease management in collaboration with Swami Vivekananda Yoga Anusandhana Samsathana (S-VYASA) (http://svyasa.edu.in/). During the month of April/May 2016, 324 adults diagnosed with T2D, aged ≥20 years to 70 years, from 12 nearby villages of the Chittoor district, Andhra Pradesh, India, were originally referred to the Apollo health scheme for diabetes management. T2D was defined as per the American Diabetes Association criteria 19 . When the records were screened, out of 277 initially referred patients, only 150 were found to complete regular supervised yoga www.nature.com/scientificreports www.nature.com/scientificreports/ treatment, and amongst them 73 only had sufficient laboratory data until November 2016. Figure 1 details the steps involved in cohort selection. Medical records of these 73 patients were retrieved for the study. Patients undergoing insulin treatment, pregnant or breastfeeding, or who had severe vascular, hepatic, renal diseases or cancer were excluded from the study. Patients with atherosclerotic cardiovascular disease (ASCVD) 7 were also excluded. Records of an equal number of T2D patients were retrieved as a non-yoga group who opted for only standard of care treatment at the rural Apollo clinics during the same period of time. Difference-in-means of the age, the proportion of sex and blood A1c levels were matched between yoga and non-yoga groups before finalizing the selection of the cohort sub-groups (n = 73 each). This was followed by matching the sub-groups through propensity score matching. Thus, we defined a cohort of 146 T2D patients, diagnosed with T2D as per ADA criteria 19 . Both yoga and non-yoga groups were followed from the (index date; date of first check up with prescribed treatment) until the end of 6 months. Parameters of interest were included at the time of admission/index date (baseline) and at an average of 6 months of follow-up. Written consent was obtained from the study subjects and the study was approved by the Institutional Ethics Committee of Swami Vivekananda Yoga Anusandhana Samsathana, Bengaluru, India. Informed consent was obtained from all the study subjects. All methods were performed in accordance with the relevant guidelines and regulations. The study was registered with ClinicalTrials. gov (NCT01212133); registration number: (CTRI/2020/02/0232790).
Measures. The duration of yoga-treatment was approximately six months. Primary parameters of interest were the follow-up status of revised ADA laid treatment goals of diabetes;A1C < 7.0% (<53 mmol/mol), and BP cut-offs (<140/90 mmHg). Additionally based on ADA definition of risk factors of Atherosclerotic Cardiovascular Disease (ASCVD) 7 , the treatment goals also included <100 mg/dl of LDL-C, and stringent BP goals (<130/80 mmHg). Composite score was defined based on the meeting of all the target goals. Secondary outcomes were continuous measures of A1c, FBS (fasting blood glucose), PPBS (postprandial blood glucose), LDL-Cholesterol (LDL-C), SBP, DBP, weight, total cholesterol (TC), triglyceride (TG) diastolic blood pressure (DBP), and body mass index (BMI). Patient demographic and anthropometric information including age, sex, www.nature.com/scientificreports www.nature.com/scientificreports/ socioeconomic status, duration of diabetes, medication, weight, blood pressure, height was also extracted. BMI was calculated as weight in kilograms divided by the square of height in meters. Asian cut-off for BMI (≥25 kg/m 2 ) was used to define obesity 20 . Intervention. The administration of yoga was carried out at Apollo rural and satellite clinics (https://www. apollohospitals.com/corporate/initiatives/csr-at-apollo/total-health-programme) for T2D patients from nearby villages. Non-yoga group, the T2D patients received the standard of care for diabetes as per ADA guidelines from a physician-coordinated team 5 . The patients were also referred for diabetes self-management education and support for strengthening and empowering their diabetes knowledge and self-care behaviors as per ADA guidelines. The yoga treatment given to the patients was derived from a validated integrated yoga module developed by Angadi et al. 38 . The treatment protocol included daily supervised administration of yoga sessions for one hour. The yoga module was comprised of loosening practices, asanas, pranayama, relaxation techniques, and meditation; (detailed protocol has been appended as a supplement table no. 1). Only certified yoga therapists were involved in the administration of the yoga-treatment. Both the yoga and non-yoga treatment groups were followed from the date of admission into the clinics, till November 2016.

Statistical analyses.
Missing data were minimal. Continuous variables were tested for normality with the Shapiro-Wilk test. We used descriptive statistics with mean and 95% confidence intervals [CIs]), and standard, or percentages (numbers) for representation of T2D patient's baseline characteristics. Categorical variables were described using frequencies. Socioeconomic status was determined by using Kuppuswamy's scale 22 . Outcome measures were compared using Analysis of covariance (ANCOVA) to adjust for baseline measures and to provide an unbiased estimate of the mean group differences. A General linear model (GLM) for multivariate analysis was developed with covariates of baseline values of the outcome variables, age, medication and duration of diabetes. P-value of < 0.05 was set as significant and < 0.0001 was set as highly significant. Statistical analysis was performed using SPSS version 21.0, Microsoft Excel-2013 and R studio version 1.1.423. For comparisons within treatment groups from baseline to follow-up, a Wilcoxon signed rank test was performed. Propensity scores were calculated for each subject based on primary baseline covariates known to be associated with diabetes treatment and/or the study outcomes, including age, sex, socioeconomic status, disease duration, medication, and biochemical parameters using the "nearest neighbour" method (Appendix, supplementary material). Logistic regression was then used to identify predictors of successful achievement of the favourable ADA and BMI outcomes. Models of the relationships were created with independent variable including age, sex, duration of diabetes, yoga treatment vs. non-yoga treatment, baseline values of variables of biological relevance.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.