Estimates of major non-communicable disease risk factors for India, 2010 & 2015: A summary of evidence

The 2011 United Nations Summit provided a global platform for non-communicable diseases (NCDs), and national governments were urged to ‘customize the implementation’ of their commitments1. A global monitoring framework (GMF) with voluntary targets was adopted to monitor global progress on NCDs2. In May 2013, the 66^th World Health Assembly adopted the comprehensive GMF for the prevention and control of NCDs1. The GMF included a set of indicators capable of application across regions and country settings to monitor trends and assess progress made in the implementation of national strategies and plans on NCDs3.

India became the first country globally to define its National Monitoring Framework with country-specific targets and indicators4. In the national advisory group consultation in June 2013, India identified 10 targets and 21 indicators to be achieved by 2025. The targets and their indicators address morbidity and mortality (2 indicators), risk factors (11 indicators) and national systems response (8 indicators). Among the key changes were the inclusion of an additional target of 50 per cent relative reduction in proportion of households using solid fuels as primary source of cooking and focus on the current use rather than the harmful use of alcohol4. All targets for risk factors are relative reduction as compared to 2010 values. The progress on the achievements to the targets was to be reviewed every five years (2015, 2020) till the 2025 deadline. As India lacks baseline values for 2010 to monitor the progress of its achievements, it is imperative to generate the 2010 baseline values while establishing mechanisms for measuring future national targets.

A review of the global status report on NCDs by the WHO showed that there were some important differences in the way the indicators were framed, which prevented India from adopting those as its baseline values5. These include the use of tobacco smoking in its report as opposed to India’s focus on tobacco use (including smokeless tobacco) and focus on current alcohol use as opposed to harmful use of alcohol in the WHO report and non-inclusion of the solid-fuel use in the report4,5. This study was initiated to assess the burden of NCDs and their risk factors, which included preparing national estimates for the prevalence of major NCD risk factors for 2010 and 2015 using available data from surveys and published literature. This study presents the result of this systematic exercise focusing only on those risk factors of NCDs for which targets have been established at the national level.

Material & Methods

Studies and reports published on Indian populations from 2008 to 2017 on the prevalence of one or more of the following risk factors such as raised blood pressure, raised body mass index, physical inactivity, alcohol consumption, tobacco consumption and solid cooking fuel use. The list of risk factors and their definitions based on the WHO recommendations is shown in Table I. Raised blood sugar was removed from the list of risk factors to be studied to avoid duplication of work as diabetes was included in the list of diseases for which estimates were being prepared. Sodium intake was also dropped as initial review showed that there would not be enough studies to generate national-level estimates for 2010 and 2015.

Search strategy: PubMed, Google Scholar and Cochrane database were searched using the search strategy combining the terms, ‘NCD risk factors’, ‘risk factor profile’, ‘lifestyle risk factors’, ‘epidemiology’. For each risk factor separate search strategies were developed (Supplementary Table I). Grey literature focusing on large-scale national survey reports was also identified. For all these multi-State surveys, each State was taken as a unit of study, if the information was available at State level. Two investigators screened all the abstracts of the articles and the reports, and then applied the eligibility criteria. They then independently extracted data from the articles using a data extraction form. Data extracted from each included article were publication date, study design, study duration and dates, participant data and results of the study. Critical appraisal was conducted using Joanna-Briggs (JBI) tool for prevalence studies, but no study was excluded based on poor quality6. Each of the papers was reviewed by at least two study investigators and a third investigator mediated any disagreements relating to eligibility, outcome definition or data.

Inclusion & exclusion criteria: All observational/cross-sectional/prevalence studies in English language between 2008 and 2017, at national or sub-national level conducted in India, covering any age group between 18 and 69 yr were included in the study. Community based studies covering either or both sexes and prevalence studies using identified case definitions by WHO were included. General review or policy papers or case-control studies or case reports and studies for which abstract were not available, studies involving special groups (occupation), subgroup or institution-based studies and studies which used case definitions other than that specified by WHO were excluded.

Statistical analysis: To generate a national estimate, India was divided into eight strata - age (18-44 and 45-69 yr); gender (male and female) and place of residence (urban and rural). This was based on the preliminary analysis of data which showed that for most variables, these were major determinants of the risk factor levels. Risk factor data for each of the eight strata was abstracted. Data from studies that did not match with the included age group was extracted to the closest age stratum. For each risk factor, meta-analysis of the data in each stratum was conducted to determine the combined estimate. An established meta-analytic method was then adopted for pooling prevalence data using Meta-XL version 5.3, a plugin software package for Microsoft Excel (EpiGear International Pty Ltd., The Netherlands) and calculated pooled estimates separately for 2010 and 2015 using a random effects model (http://epigear.com/index_files/metaxl.html). The strata-specific estimates were combined to arrive at national estimate using population weight of each stratum. Population weights for 2010 and 2015 derived from the 2011 Census of India Report and Population Projections for India and States 2001-2026 report7 by Census of India, respectively, were used (Supplementary Table II).

Even though the evaluation of the credibility of results from a meta-analysis has become important, not much guidance is available for rating credibility of population estimates generated by meta-analyses. There are two major approaches which deal with quality of evidence or confidence on results of the evidence synthesis process – GRADE and CINeMa8,9. Both these approaches use the overlapping criteria of inconsistency/heterogeneity, indirectness, imprecision, reporting bias and study bias or limitations. Indirectness pertains to study happening in a group being used to make recommendations for another group and is not relevant to our purpose. Therefore the credibility of the estimate was assessed using four parameters: average JBI Score for the quality of studies; representativeness based on the number of zones (north, south, east, west, north-east and central) contributing to each of the estimates; relative precision by dividing the width of confidence interval (CI) with the estimate and consistency based on the difference in estimates across two time periods for subgroups of age, geographic location and gender. Details are provided in Supplementary Table III.

Results

The results of the literature search and decisions at different phases as per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for each risk factor are shown in Table II. A total of 84 studies contributed to the estimates for 2010 and 45 studies contributed to 2015 and these are listed in Supplementary Table IVA and B, respectively. One study could contribute data to more than one risk factor. Overall, more studies contributed to estimates for overweight/obesity (81) and tobacco (54) as compared to estimates for alcohol (23) and physical inactivity (35). The data on solid fuel use were mainly from large national-level surveys as opposed to physical activity, which was not covered in any national-level survey.

Based on the above-defined criteria the credibility of the estimates generated for 2010 and 2015 was assessed. Overall, the estimates scored between moderate and high levels of credibility. They were least credible for overweight and obesity estimates and best for use of solid cooking fuel. There were no major differences in credibility scores for 2010 and 2015 (Table III).

The estimates for each of the risk factors for 2010 and 2015 are shown in Table IV. The estimate of current tobacco consumption showed an 18.5 per cent relative decline from 27 per cent (21.4-32.6%) in 2010 to 22 per cent (17.3-26.6%) in 2015. The absolute decline during this period was higher for men (42.6-35.4%) as compared to women (10.7-7.7%) and more for urban (30.3-15%) as compared to rural (35.7-32.6%). This resulted in narrowing of gender differential and widening of rural-urban differentials. The estimates for the use of solid fuel as a source of cooking showed a modest decline from 61.5 per cent (50.2-72.5) to 55.7 per cent (41.6-69.8) between 2010 and 2015 with a larger reduction occurring in urban areas (from 29 to 17%) as compared to rural areas (from 78 to 73%).

The estimates for current alcohol use remained around 15-16 per cent during both the time points of 2010 and 2015 with no major rural/urban or gender differentiates during this period. The estimates also showed higher consumption of alcohol among men as compared to women and similar prevalence in rural-urban areas. The estimates for prevalence of physical inactivity showed a 15.6 per cent relative decrease from 44.2 per cent (37.8-50.6%) in 2010 to 37.3 per cent (32.7-41.9%) in 2015. The absolute decline was much higher for women (54.1-42%) who, nonetheless, continued to have higher levels of physical inactivity as compared to men who did not show much change during this period. Physical inactivity estimates showed an increase in rural residents (35.4-49.9%), whereas a significant decline was noted among the urban residents (54.5-30.7%).

The prevalence of raised blood pressure showed a small but insignificant increase from 20.2 per cent (18.4-22.1%) in 2010 to 21.6 per cent (19.3-23.9%) in 2015. While the estimates for men showed a significant increase (from 14.1 to 21.8%), women showed a small but non-significant decrease (from 26.6 to 21.3%). Both rural (from 20.8 to 23.1%) and urban areas (from 28.0 to 33.9%) showed an increase in the prevalence of raised blood pressure. The estimates for overweight showed a small insignificant increase; 20.5 per cent (12.6-28.4%) in 2010 and 21.7 per cent (18.0-25.4) in 2015. The estimate for men showed an increase (from 16.2 to 20.3%) while for women it showed a small decline (from 25 to 23%). A decline was noted among rural residents (27-18.8%) but an increase was seen among urban residents (from 12.6 to 34.8%). The estimates for obesity remained similar for all strata as well as at the national level during both the time periods.

Discussion

This data synthesis exercise for estimating 2010 baseline values for NCD-related indicators for whom national targets have been set along with the estimates for 2015, was conducted at the behest of the Ministry of Health and Family Welfare (MoHFW), Government of India. The strengths of this study are use of stratification to address population diversity, reasonable precision of the estimates and a transparent and standard approach to estimation and moderate-to-high levels of credibility of the estimates. While unlikely for the estimates for 2010, the estimates for 2015 could undergo a change as more studies are subsequently published. The weaknesses of the present investigation are lower precision of estimates at strata level and lack of national representativeness for some risk factors. Despite large number of individuals contributing to the estimates, the precision of estimates is only modest as meta-analysis considers the study as a unit. The precision is also affected by the number of strata used for estimation, with lesser number of strata resulting in more studies per strata and higher precision. However, given the fact that different risk factors showed different variability across the three parameters (age, residence and sex), our choice of preferring validity over precision is reasonable.

Before these estimates are accepted, they should be critically evaluated. As national-level studies conducted during this period have already been included in the process of generating these estimates, these cannot themselves be used for validation. We propose two ways to assess the reasonableness of our estimates. The first is to make comparisons of the change across strata over time for a risk factor and see if these are consistent with what is known about that risk factor and its control efforts in the country. The other way is to compare these estimates with those that have been generated by other global agencies and collaborations. This was done for each of the risk factors separately and ss described below.

For alcohol, the present study reports that there may be a little change in the country as well as for gender and place of residence against a target of 10 per cent reduction by 2025. This is consistent with what is known about alcohol control strategy in India. India does not have a national alcohol control policy. While some States have alcohol prohibition policies, their implementation is quite erratic. Different reviews of India’s alcohol control policy have identified its weaknesses and areas needing strengthening10,11. There are two key indicators which WHO uses for alcohol consumption, one is on per capita consumption of liquor and the second one is on harmful use of alcohol. The Global Burden of Disease has estimated that there was a 38 per cent increase in per capita consumption of alcohol in India from 1990 to 201712. It is well documented that one of the characteristics of alcohol consumption in India is heavy episodic binge drinking rather than regular alcohol use. Thus, this increase in consumption need not necessarily mean increase in prevalence. Drafting a comprehensive national alcohol policy and working on its implementation is required if India has to bring down its alcohol consumption.

Current use of tobacco showed an 18.5 per cent decline between 2010 and 2015, against a target of 30 per cent decline by 2025. This is supported by the fact that India has shown a strong political will in its approach towards tobacco control and introduced most of the measures described under the Framework Convention of Tobacco Control. India has clearly outlined its tobacco control strategy and introduced necessary legislation to support this strategy. This is consistent with 17 per cent decline in tobacco use between the two global adult tobacco surveys (2009-2010 and 2016-2017), which themselves have contributed to the current estimates13. A comparison of our estimate with other estimates is complicated by the use of smoked tobacco as a global indicator. The global tobacco epidemic report 2019 estimate for current tobacco use for India for 2017 at 27.8 per cent (20.8-34.3) is higher than our estimates 22.0 per cent (17.3-26.6%), though the CIs overlap. Our estimates for men seem to be almost same at 42.7 per cent (32.8-52.3), but differ for women with global report estimating it at 12.9 per cent (8.9-16.3) and our estimates being much lower at 7.7 per cent (4.6-10.9%)14. Further efforts to strengthen tobacco control need to address smokeless tobacco and tobacco use among women15.

Our estimate of 61.5 per cent of the households using solid fuel in 2010 is quite close to 61 per cent estimated by Bonjour et al16. The estimate of 55.5 per cent for 2017 by the India State-Level Disease Burden group matches quite well with 55.7 per cent for 2015 in India17. This is about 10 per cent lower than the target of 50 per cent relative reduction set by the government to India. Our data also show that rural India continues to have high solid fuel use. Keeping this in mind, India launched a national initiative named Prime Minister Ujjwala Yojana in 2016 to improve access of rural households to clean fuel (https://www.pmujjwalayojana.com/about.html). While it has made significant progress, it is facing significant implementation challenges that are being addressed, which makes it likely that the progress will be accelerated18.

Our estimates for physical inactivity showed a 15.6 per cent decline between 2010 and 2015 at national level which is steeper than the target set with women and urban areas showing significant improvement and worsening in rural areas. Guthold et al19 estimated physical inactivity levels for 2016 as 34·0 per cent (22.3-47.7), which is quite close to our estimate of 37.3 (32.7-41.9) for 2015. There has been some government-led initiative to promote physical activity in India, largely restricted to urban India, but the most decline in physical inactivity is being driven by increased awareness among the urban people. Anecdotal evidence is available to support lifestyle changes due to increased access to technology in rural areas resulting in decreased physical activity. Our estimates of levels of overweight and obesity for 2010 and 2015 are higher than those reported in the WHO Global status report 201420. What is similar in both the estimates is that the overweight/obesity levels are relatively stagnant, which is in keeping with the target of halting the rise of overweight and obesity. A more vigorous effort addressing environmental, structural and information determinants is required to promote physical activity in India with special focus on women and rural areas. A recent review concluded that ‘physical activity (PA) promotion seems to be lacking in the policy perspective and currently physical inactivity as a risk factor is not considered seriously’21.

Our estimates for 2010 for raised blood pressure (20.2%) are slightly lower than the WHO estimate for the same year (22.9%) as well as for 2015 (21.6 vs. 25.4%)20. Our estimate is also lower than that of Anchala et al22 whose meta-estimate was 29.8 per cent (95% CI: 26.7-33.0) for studies conducted between 2011 and 2013, but their national estimate was based only on regional stratification. Both the estimates show that there has been a slight increase or little change as compared to the target of a 30 per cent reduction in raised blood pressure. India is currently focusing on increasing diagnosis and improving access to treatment of hypertension with the launch and expansion of the India Hypertension Control Initiative23. However, the focus on prevention is weak with no specific major intervention aimed at salt reduction at population level. Major barriers indicated by the stakeholders for salt reduction in India were social and cultural beliefs, a large unorganized food retail sector and the lack of proper implementation of even existing food policies24. These need to be addressed effectively.

In conclusion, this study provides with credible estimates for 2010 and 2015 for monitoring progress on the targets set for NCD prevention and control in India. These estimates enable us to clearly identify areas that India is doing well and areas that it should focus on in future, if it has to achieve the targets it has set for itself. The need for India to establish a sustainable surveillance mechanism for primary collection of data through a nationwide survey has already been repeatedly made25. Furthermore, India completed a nationwide NCD monitoring survey in 2017-18. Reading of the recent nationwide NCD monitoring survey results (2017-18) in conjunction with these estimates can provide further useful insights on the trend of risk factors to improvise our strategies26. Overall, it is suggested that these mechanisms need to be institutionalized and linkages to policy and programmes must be strengthened.