Prevalence of metabolic syndrome among adult population in India: A systematic review and meta-analysis

Objective This review was done to determine the prevalence of metabolic syndrome (MS) among adult general population in India. We also wanted to find the gender, setting, and region-wide distribution of MS in India. Methods We conducted systematic searches in various databases including Medline, ScienceDirect, Cochrane library and Google Scholar from inception until August 2019. We included studies conducted in India reporting the prevalence of MS among adults aged 18 years or more. We used the Newcastle Ottawa scale to assess the quality of included studies. We carried out a meta-analysis with random-effects model and reported pooled prevalence with 95% confidence intervals (CIs). We used the Funnel plot to assess publication biases. Results In total, we analysed 113 data from 111 studies with 133,926 participants. Majority of the included studies (76 out of 111) had low risk of bias. We found significant heterogeneity among the included studies (p<0.001). We also found a symmetrical funnel plot indicating an absence of publication bias. The prevalence of MS among adult population in India was 30% (95%CI: 28%-33%). There was a steady increase in the burden across the age groups from 13% (18–29 years group) to 50% (50–59 years). We also found that people living in urban areas (32%; 95%CI: 29%-36%) had higher prevalence when compared to tribal (28%; 95%CI: 21%-36%) or rural adults (22%; 95%CI: 20%-25%). Gender distribution of MS showed that the females had higher prevalence (35%; 95%CI: 31%-38%) when compared to males 26% (95%CI: 22%-29%). Conclusion Almost one in three adults in India suffer from MS. Females, people living in urban areas and in northeast region had higher prevalence of MS. Development and implementation of policies and protocols for the screening of MS would enable us in early diagnosis and treatment with special focus towards the vulnerable and high-risk groups.

Introduction Disease patterns around the globe are undergoing rapid structural changes over the last three decades, with a sudden increase in the burden of Non-Communicable Diseases (NCDs) and a decreasing trend of communicable diseases [1]. The Global Burden of Disease (GBD) study, brings to light this phenomenon of epidemiological transition in India, with a 62.7% of the total mortality in 2016 contributed by the NCDs. Key elements contributing to the development of these NCDs have been identified and are studied together under the heading of Metabolic Syndrome (MS) [2].
Metabolic syndrome is a constellation of interconnected physiological, biochemical, clinical and metabolic factors that directly increases the risk of cardiovascular diseases, type 2 diabetes mellitus (DM) and all-cause mortality. It is constituted by abdominal obesity, insulin resistance, hypertension, and hyperlipidemia [3]. Various diagnostic criteria have been proposed for quantifying MS. But the most widely used ones are from the International Diabetic Federation (IDF) and the National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III) [4,5].
Metabolic syndrome increases the risk of developing type 2 DM and cardiovascular diseases over the next 5 to 10 years by five and two-fold respectively [3]. Furthermore the patients with MS have, an average four-fold increased risk of developing stroke & myocardial infarction and a two-fold risk of dying from a similar event compared with those without MS, regardless of previous history of cardiovascular events [6]. Identifying the individuals with (or) at risk of developing MS would help to inform the probabilities of worse outcomes and thus an urgent need for a promotive, preventive or curative action.
A recent study from the United States reported the prevalence of MS to be around 22.9% [7]. Various population-based studies were conducted in India too, to quantify the same and the results ranged from 10 to 30 percentage [8]. A study conducted in the eleven large urban cities of India during 2006-2010 reported the prevalence of MS as high as 35% [9]. Owing to the behavioural habits, urban population, seem to be the most vulnerable group for developing MS [10]. Prevalence of MS, thus, seems to vary greatly from region to region and from ethnicity to ethnicity.
Even though evidences are being generated on MS from different parts of the country, there is no nationwide pooled estimate to comment on the burden of MS in the Indian subcontinent, which could drive policy action [11]. With almost one third of the total population living in the urban areas, and with 40% of the population aged between 30 and 70 years, understanding the overall prevalence of MS in India becomes essential for predicting the future burden of type 2 DM and cardiovascular diseases [12].
While the NCD care in the country is segregated under the broad headings of Diabetes, and Cardiovascular diseases, MS informs us the need to look at the picture as a whole. Results from this study would, thus, help guide national policy making process to give due importance to prevention and treatment of MS in the community and clinical settings. This could thus be an essential key to keep a check on the ever-increasing burden of the NCD epidemic in the country. Hence, this review was conducted to generate a pooled prevalence estimate for MS in India.

Design and registration
We conducted a systematic review and meta-analysis of cross-sectional studies. We have obtained approval from the JIPMER Scientific Advisory Committee (JSAC). We have also obtained exemption from Institutional Ethics Committee (IEC). We have registered our protocol on the international prospective register of systematic review (PROSPERO). PROSPERO Registration Number is CRD42019147277. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist for reporting systematic reviews incorporating meta-analyses for reporting our review.

Eligibility criteria
Type of studies. We included studies conducted in India reporting the prevalence of MS for the current review. There was no restriction related to study design, communities (rural/ urban) or age groups. We included the studies irrespective of the setting in which the study was conducted (community or facility based or workplace based). Studies reported as full text were included while unpublished data were excluded from the review.
Type of participants. We included studies conducted among adults aged 18 years or more. We excluded the studies conducted among specific diseased population.
Type of outcome measure. Studies reporting the prevalence of MS and diagnosed it as per the NCEP ATP III or IDF guidelines were included to obtain the pooled prevalence estimate for MS in India.

Search strategy
We conducted extensive electronic search in the following databases: Medline, ScienceDirect, Cochrane library and Google Scholar. Combination of medical subject heading (MeSH) and free text terms were used for carrying out literature search. The detailed search strategy used to search the Medline database has been reported in S1 File. Similar strategies were used in Cochrane Library, ScienceDirect and Google scholar for literature search of published studies. Search was conducted in all the databases from inception of the database to August 2019 with English language restriction for publication. We also checked the reference list of primary studies obtained through electronic search and relevant articles were included in the review and analysis.

Selection of studies
Two independent investigators (YK and TR) independently performed the literature search and screened the title, abstract and keywords of all the studies identified for possible inclusion in the review. Full text article was obtained for those studies that are found to be relevant. Further screening of abstract and full text of the retrieved articles was done independently by two investigators (YK and TR) to select the studies which satisfy the eligibility criteria of the current review. Any disagreements during the entire selection process between two authors were resolved either through consensus or consultation with third investigator (SR). Quality of the overall review process was monitored by the third investigator (SR).

Data extraction and management
Primary investigator (YK) extracted the relevant study characteristics for the review from the included studies using the Cochrane Public Health group Data extraction and Assessment Template. Following data were extracted: General information. Author, Study title, Publication year. Methods section. Study design, study period, study setting (community/workplace/facility), community (urban/rural/tribal), geographical region, state, sample size, sampling technique, diagnostic criteria, outcome assessment and statistical tests employed.
Outcome section. Mean age, non-response rate and their characteristics, prevalence of MS.
Primary investigator (YK) transferred the obtained data into the software STATA version 14. Data entry was double checked for correct entry by another investigator (SR) through comparison of data presented in review and included study reports.

Risk of bias assessment in included studies
Two independent authors (SR and SM) assessed the quality of all the included studies using the Newcastle-Ottawa (NO) scale adapted for cross sectional studies [13]. Two criteria (selection and outcome) were used to assess the risk of bias. Following domains were used for assessing the risk of bias under selection criteria: representativeness of the sample, justification of sample size, rate of non-respondents and their characteristics and use of validated measurement tool. Under Outcome criteria, outcome assessment and statistical test employed were used to assess the risk of bias. Each of these outcomes was rated as high (1 point) or low (0 points) based on the quality of evidence and availability of information. Studies scoring more than or equal to 3 points were considered to have high risk of bias.

Statistical analysis
Meta-analysis was performed with the selected studies using STATA 14.2 (StataCorp, College Station, TX, USA). For each of the studies, standard error was calculated using the reported prevalence of MS and total sample size. "Metaprop" function was used for performing analysis of the current review [14]. To minimize the effect of extremely small or large prevalence on the overall estimate, Freeman Tukey double arc-sine transformation was done to stabilise the variance [14]. Final pooling of data was done with random effects model and reported as proportion with 95% confidence interval.
Evidence of between-study variance due to heterogeneity was assessed through following methods: Chi square test for heterogeneity and I 2 statistics to quantify the inconsistency. I 2 less than 25% was mild, 25-75% was moderate and more than 75% was considered as substantial heterogeneity [15]. Sub-group analysis was performed based on age group, gender, geographical region (North/South/West/East/Northeast/Central), study setting (community/workplace/ facility), diagnostic criteria, waist circumference measurement, year of publication, quality of studies, representativeness of the sample and urban/rural/tribal communities. Potential covariates for meta-regression were study setting, region, diagnostic criteria, representativeness of the sample, year of publication, mean age and quality of evidence. We performed multivariable meta-regression analysis by including the study level factors with p value less than 0.2 in the univariate model. We performed sensitivity analysis to assess the robustness of the results by excluding the studies conducted exclusively on elderly (�60 years), studies having high risk of bias or not conducted on a representative sample, and studies conducted before 2007.
Study specific prevalence estimates and pooled estimates were graphically represented through forest plot for both combined and subgroup analysis. Publication bias was assessed and graphically represented by funnel plot; asymmetry of the plot was tested using Egger's test and p value less than 0.10 was considered as statistically significant publication bias [16].

Study selection
We conducted a systematic search to find studies that report the prevalence of MS from the dates of database inception until August 2019. During the first stage of screening (title, abstract and keywords), we retrieved 310 studies from the following databases: Medline, ScienceDirect, Cochrane library and Google Scholar. After removal of duplicates, we reviewed 249 articles' full texts against our eligibility criteria for the possible inclusion in the review. We reviewed the bibliographies of the retrieved articles and found four more relevant studies. Finally, we analysed 111 studies with 133,926 participants satisfying the inclusion criteria (Fig 1)  .

Methodological quality of the included studies
We performed assessments of risk of bias for all the included studies using NO scale. First, selection bias domains were assessed and it was found that 64 out of 111 studies (58% of studies) had high risk of bias related to representativeness of the sample for our review, 80 (72%) studies did not report proper justification for sample size, 69 (62%) studies did not report nonresponse rate or their characteristics. All the studies reported use of validated measurement tool. Under the outcome domain, only one study did not independently assess the outcome; nine studies did not report the statistical methods used in the study. Almost one-third i.e. 35 (31.5%) of the included studies had high risk of bias as per NO scale.

Burden of metabolic syndrome in India
Pooled estimate was calculated after adjusting for population size weights and graphically depicted in Fig 2. The overall pooled prevalence of MS among adult population in India was 30% (95%CI: 28%-33%). State wise analysis of MS showed that the maximum prevalence of MS was reported in Madhya Pradesh (50%) followed by New Delhi (43%), Odisha (43%) and Telangana (42%). Least pooled prevalence of MS was found in Jammu & Kashmir (15%) followed by Haryana (18%) and Punjab (21%). There was a significant heterogeneity found among the studies included in our review (I 2 = 99.1%; p<0.001). Additional subgroup analysis was performed to explore the source of heterogeneity and obtain separate estimates based on age group, gender, study setting, geographical area and regions, year of publication, representativeness of sample, waist circumference measurement, quality of studies and diagnostic criteria ( Table 2).
Age wise distribution of metabolic syndrome in India. The pooled prevalence of MS differed significantly across the age groups (p<0.001). There was a steady increase in the burden across the age groups from 13% (18-29 years group) to 50% (50-59 years). There was significant heterogeneity among the studies reporting prevalence across all the age groups with I 2 >90% and p<0.001.
Gender distribution of metabolic syndrome in India. In total, 82 studies have reported prevalence of MS among female adult population while 76 studies have reported for males. The pooled prevalence of MS among adult females in India was 35% (95%CI: 31%-38%); while for males, the pooled prevalence was 26% (95%CI: 22%-29%). There was significant heterogeneity among the studies reporting prevalence in females and males (I 2 = 98%, p<0.001).
Northeast India. Only 4 studies were conducted in Northeast India to determine the prevalence of MS. The pooled prevalence of MS in Northeast India was 35% (95%CI: 25%-46%). There was a significant heterogeneity found among the studies conducted in South India (I 2 = 96.9%; p<0.001) (S6 Fig). There was no sufficient number of studies to see the urban-rural and gender-based differences in the burden of MS.

Additional subgroup analysis and sensitivity analysis
Subgroup analysis based on diagnostic criteria revealed a statistically significant difference in the pooled estimate between NCEP ATP-III and IDF criteria (29% vs 34%; p = 0.05). However, we did not see any statistically significant difference in the burden of MS based on year of publication, difference in measurement of waist circumference, quality of studies or representativeness of sample.
We performed sensitivity analysis by excluding the studies conducted exclusively on elderly (�60 years), studies having high risk of bias or not conducted on a representative sample, and studies conducted before 2007. This is to provide robust and latest estimate on the burden of MS among adult population in India. In total, 47 out of the 111 studies satisfied these criteria and included in the sensitivity analysis. The pooled prevalence of MS obtained in sensitivity analysis was still almost similar to the overall findings (29%; 95%CI: 25%-33%) indicating that the estimate is robust to the changes in age group, year, and quality of studies.

Meta-regression
First, we conducted univariate meta-regression with individual study level factors such as study setting, geographical region, diagnostic criteria, year of publication, mean age, representativeness of the sample and quality of studies (Table 3). We found significant association with mean age and study setting. However, we performed multivariable meta-regression with factors having p value less than 0.2 such as study setting, mean age, and diagnostic criteria used. The adjusted model was able to explain 24.5% of the between-study variability and the model was statistically significant (p = 0.004).

Publication bias
Egger's test was performed for the assessment of publication bias. There were no small study effects with non-significant coefficient value (Co-efficient: 0.78; 95%CI: -0.95 to 2.52; p = 0.372) which shows lack of evidence of publication bias. Graphical representation of the test of publication bias was depicted through funnel plot in Fig 3. Funnel plot also showed symmetric plot indicating the absence of publication bias.

Discussion
We have conducted this review to obtain a comprehensive estimate of burden of MS among adult population in India. We have also captured the gender distribution, urban-rural-tribal differences and geographical region wise estimates to find any significant difference in the estimates of MS. In total, we analysed data from 111 studies with 133,926 participants. Majority of the studies were conducted in Southern states followed by Northern states. Majority (83) were from community-based studies. Out of these, 49 have reported separate estimates for urban region, 34 for rural and 4 for tribal region. Majority of the included studies (76 out of 111) had low risk of bias.
The prevalence of MS among adult population in India was 30% (95%CI: 28%-33%). Similar reviews published from other low middle income countries in Middle East, South East Asia and Latin American region also reported that almost one third of general population have MS  [128][129][130]. This recent upsurge in the prevalence of MS among low middle income countries including India might be directly linked with rapid economic development and urbanization in the country. This rapid industrialization can influence drastic changes in lifestyle patterns and nutrition [131][132][133].
We also found that people living in urban areas (32%; 95%CI: 29%-36%) were found to have higher prevalence of MS when compared to people living in tribal (28%; 95%CI: 21%-36%) or rural areas (22%; 95%CI: 20%-25%). These findings were in line with the studies conducted in neighbouring low middle-income countries [129]. Unhealthy lifestyles, better socio-economic status, decreased physical activity, stress, excessive salt and red meat consumption might be the contributory factors influencing the higher prevalence of MS in urban areas [134].
Age wise distribution showed that there is a steady rise in the burden of MS with increase in age groups and it differed significantly (p<0.001). Gender distribution of MS showed that the females had higher prevalence (35%; 95%CI: 31%-38%) when compared to males 26% (95%CI: 22%-29%). Similar reviews in Eastern Mediterranean [135], South East Asian [136], and Western pacific [137] regions abide by our findings. Major reason for this finding could be the gender specific risk factors such as menopause, polycystic ovarian syndrome and use of hormonal contraceptives among women [138]. Other probable reasons could be the excess risk women carries in terms of elevated body weight, increased waist girth, and low high density lipoprotein cholesterol when compared to men [139][140][141][142].
Highest prevalence of MS was reported in Madhya Pradesh (50%) followed by New Delhi (43%), Orissa (43%) and Telangana (42%). Stratified analysis across the geographical regions showed that people living in Northeast India (35%) have highest prevalence of MS followed by Eastern India (33%), regions with unique lifestyle and culture [142]. This highlights the need of understanding the influence of sociocultural, ethno-geographical factors in determining the risk of MS. This high prevalence among the north eastern states could also be due to lesser number of studies (n = 4) from the region. Hence, the current review may not be representative of the entire north eastern population. Further studies should be carried out in this region to have conclusive evidence on the burden of MS.
The major strength of the study is that we have tried to provide the first comprehensive review on burden of MS among adult general population in India. We also provided estimates based on gender, study setting, community and geographical regions. Test for publication bias have found that there was no significant bias in the current review. However, our review has certain limitations. Summarizing and concluding the burden of MS as 30% among adults in India with demographic and socio-economic differences is difficult because of the inherent heterogeneity. We have tried to overcome this limitation by conducting subgroup analysis based on study setting, geographical regions and provided individual prevalence estimates. The chi square test for heterogeneity also revealed significant variability across the national and region wise included studies. Hence, we tried to explain the between-study variability using meta-regression and found the potential sources of heterogeneity.
In spite of these limitations, current review provides important baseline information on the burden of MS among adult population in India. Our review has showed a higher prevalence among women, people living in urban areas and in specific geographical pockets. The findings of our review highlight that MS is a major public health problem in India. It is necessary for the government to allocate adequate resources and establish appropriate cost-effective interventions to tackle the burden of MS. Development and implementation of policies and protocols for the screening (focusing on population level screening strategies) would enable us in early diagnosis and treatment. Special focus should be given towards the vulnerable and highrisk groups. A wholesome comprehensive health care approach encompassing various levels of health care is necessary to achieve better management of MS. This crucial step would not only enable us to reduce the mortality imposed by its individual components but also to curtail the out-of-pocket expenditure incurred by these conditions. Further comprehensive meta-analysis on population-based studies is necessary to find the factors responsible for MS which will help the policy makers, especially in the low and low middle-income regions to devise region specific interventions.