The CNNS was conducted from February 2016 to October 2018 by the Ministry of Health and Family Welfare, Government of India, collaborating with United Nations Children's Fund (UNICEF) and the Population Council. The methodological details of the survey are available in detail in the CNNS report (12). Briefly, the CNNS collected data from three target age groups- 0–4 years, 5–9 years, and 10–19 years in all states of India. A multistage, stratified, probability proportional to size (PPS) survey design covering rural and urban households was used. Villages in the rural area and municipal wards in the urban area formed the primary sampling units (PSU). Census Enumeration Blocks (CEB) were the secondary sampling units in urban areas. The PSUs were selected by PPS, and households were selected randomly. Only one child or adolescent was chosen from each age group in each household. Adolescents with major physical deformity, cognitive disabilities, chronic illness (e.g., tuberculosis, cancer, liver, and renal disease), acute febrile or infectious illness, acute injury, ongoing fever, and pregnant adolescents were excluded. The present secondary data analysis needed no institutional ethical approval. However, Ethical approvals for the CNNS were obtained from the Population Council's Institutional Review Board, New York, and the Institutional Ethical Committee of Post Graduate Institute of Medical Education and Research, Chandigarh, India (PGI/IEC/2015/1508).

The CNNS used a household and an individual questionnaire for all participants using the Computer-Assisted Personal Interviewing (CAPI) method in the principal languages of the state and/or English after obtaining informed consent from the caregiver and assent from the adolescent (12). For everyone, data were collected on age, gender, location/region, religion, social class, current educational status, mother's education, wealth index, and region of the country as per the standard definitions (Supplementary Table 1). Height and weight were collected using a Stadiometer and digital weighing scale, respectively. During the CNNS survey, anthropometry measurements were internally monitored, and a three tier system of data quality assurance mechanisms were implemented.

Individual-level data for 10–19 years collected from the CNNS was accessed for this secondary analysis. The age group 10–19 years included children who had completed 120 months till 228 months. The age group was classified as early adolescents (120–179 months) and late adolescents (180–228 months). World Health Organization (WHO) Anthro Plus software was used to derive the BMI-for-age z-score (BAZ) and Height-for-age z-score (HAZ) (13). WHO standard classifications were followed to define the outcome variables. Stunting was defined as HAZ < −2 standard deviation (SD), Thinness as BAZ < −2 SD, Overweight as BAZ > +1 SD, and Obesity as BAZ > +2SD (13). The proportion of adolescent children with stunting, thinness, overweight, and obesity, along with 95% confidence intervals (95% CI) was estimated at the national level and for each state and union territory. The estimated burden of stunting, thinness, overweight, and obesity for individual states and union territories are plotted on an Indian map using R studio.

Data was collected from 35,831 adolescents during the CNNS survey, of which 31,941 adolescents had BAZ scores, and 32,045 had HAZ score and were included in the final analysis (Figure 1). The collected data included children aged 10 years and above (120 months onwards) till 19 years (239 months). As the z-scores were available in WHO references till the age of 19 years (228 months) only, the data of children aged 229–239 months was excluded for analysis. STATA version 13 (StataCorp LLC, College Station, TX, USA) was used for all statistical analyses. The association between socio-demographic characteristics and individual anthropometric indicators (Stunting, Thinness, and overweight) was assessed through a Univariate logistic regression analysis. The socio-demographic variables include age, gender, religion, residence, social class, maternal education, wealth index, and region of India. A multivariable logistic regression model examined the association between socio-demographic characteristics and individual anthropometric indicators. Variables with a p-value < 0.25 on univariate analysis, variables of contextual and clinical importance were included in the final model (14). A p-value < 0.05 was considered statistically significant.

Stunting (HAZ < −2 SD) was seen in 8,877 Indian adolescents with a burden of 27.4% (95% CI 26.4, 28.4%). The burden of thinness was 24.4% (95% CI 23.5, 25.4%) as seen in 7,896 adolescents. The double burden of stunting and thinness in the same individual was seen in 8.6% (95% CI 8.3, 8.9%) adolescents, while the prevalence of co-existing stunting and overweight was seen in 0.8% (0.7, 0.9%). The burden of overweight and obesity among Indian adolescents was 4.8% (95% CI 4.5, 5.1%) and 1.1% (95% CI 0.9, 1.3%), respectively (Table 1).

A higher prevalence of stunting was seen in north-eastern states—Assam (41.3, 95% CI 37.5, 45.2), Meghalaya (45.9, 95% CI 40.4, 51.6), Nagaland (41.8, 95% CI 36.8, 47.0) and Jharkhand (34.1, 95% CI 29.6, 38.9) compared to the national average (27.4 95% CI 26.4, 28.4). Seven states have thinness more than the national average (24.4, 95% CI 23.5, 25.4), two of which are from the Southern region—Telangana (29.1, 95% CI 25.7, 32.7), and Karnataka (27.3, 95% CI 23.6, 31.4). Overweight (India 4.8, 95% CI 4.5, 5.1) was seen to be more prevalent in Tamil Nadu (12.4, 95% CI 10.1, 15.1), Delhi (12.0, 95% CI 9.6, 14.9%), Goa (14.3, 95% CI 12.0, 16.9), Kerala (9.5, 95% CI 7.3, 12.3), Sikkim (9.3, 95% CI 6.9, 12.4), Arunachal Pradesh (11.1, 95% CI 8.9, 13.8), and Tripura (9.3, 95% CI 7.5, 11.5). States with a prevalence of overweight lower than the national average include Madhya Pradesh (1.7, 95% CI 1.1, 2.5), Uttar Pradesh (2.1, 95% CI 1.5, 3.0), Bihar (1.8, 95% CI 1.1, 3.0), and Jharkhand (1.9, 95% CI 1.2, 3.1). Obesity was found to be higher than the national average (1.1%, 95% CI 0.9, 1.3) in states like Goa (5.0%, 95% CI 3.7, 6.5), Delhi (3.3%, 95% CI 2.1, 5.0), Tamilnadu (2.6%, 95% CI 1.7, 3.8), Punjab (2.6%, 95% CI 1.8, 4.0), and Manipur (2.6%, 95% CI 1.7, 3.9) (Figure 2).

Adolescents in the age group of 15–19 years (AOR 1.23, 95% CI 1.11, 1.36), female gender (AOR 1.20, 95% CI 1.08, 1.33), Muslim religion (AOR 1.20, 95% CI 1.02, 1.42), scheduled caste (AOR 1.36, 95% CI 1.16, 1.59) and scheduled tribes (AOR 1.57, 95% CI 1.25, 1.98) were at increased odds of stunted growth compared to their counterparts (Table 2). Similarly, adolescents never attending the school (AOR 1.33, 95% CI 1.09, 1.63), belonging to the lowest wealth index (AOR 2.54, 95% CI 2.05, 3.14) were at increased odds of being stunted. Adolescents from all other parts of India were at increased odds of stunting compared to adolescents from the northern region of the Country (Table 2).

Adolescents from other backward classes (AOR 1.26, 95% CI 1.08, 1.47), lowest wealth index families (AOR 1.66, 95% CI 1.33, 2.07), western India (AOR 1.21, 95% CI 1.02, 1.44) were at increased odds of thinness compared to their counterparts at a statistically significant level. The age group of 15–19 years (AOR 0.68, 95% CI 0.61, 0.76) compared to 10–14 years age group, Female gender (AOR 0.57, 95% CI 0.51, 0.63) compared to males were at decreased odds of thinness. Adolescents from eastern India (AOR 0.79, 95% CI 0.66, 0.93) and north-eastern India (AOR 0.59, 95% CI 0.43, 0.88) were at decreased odds of thinness compared to adolescents from northern India (Table 3; Supplementary Table 2).

Adolescents in the age group of 10–14 years (AOR 1.26, 95% CI 1.06, 1.49), residence in urban areas (AOR 1.43, 95% CI 1.19, 1.71), others (general) category of social class (AOR 1.30, 95% CI 1.07, 1.57), mothers' education more than 12 years (AOR 1.66, 95% CI 1.33, 2.06) highest wealth index families (AOR 2.11, 95% CI 1.59, 2.79) were at increased odds of overweight compared to their counterparts. Similarly, adolescents residing in north-eastern India (AOR 1.87, 95% CI 1.28, 2.74), eastern India (AOR 1.49, 95% CI 1.10, 2.01), and southern India (AOR 1.85, 95% CI 1.48, 2.30) were at increased odds of overweight compared to adolescents residing in north India (Table 4).

In the present study, the prevalence of stunting was higher in late adolescence (30% in late adolescence vs. 25.6% in early adolescence). According to the Global School-Based survey conducted in 57 Low-Middle-Income countries (LMICs) on children aged 12–15 years, 14.6% of adolescents in India were stunted (15). Our analysis has shown higher estimates of stunting, which could be due to the inclusion of 15–19-year-old subjects, where the prevalence of stunting was 4% higher than 10–14-year adolescents. Among Indian adolescents in the 15–19 years age group, Bhargava et al. reported that stunting was 29.1% (95% CI 28.6–29.6) in NFHS-3 (2005-06), which increased to 34.1% (95% CI 33.9–34.4) in NFHS-4 (2015-16) (16). Our estimates from CNNS data on stunting are lower than both NFHS-3 and NFHS-4. The lower estimates could be because NFHS represents only the late adolescence (15–19 years), while CNNS includes both early and late adolescence (10–19 years).

The burden of stunting among girls was higher than boys, corroborating with previous estimates by Bhargava et al. on NFHS data (16). Indian boys and girls had similar growth faltering compared to the WHO median until 14 years, and after this age, girls grew slower, leading to a higher burden of stunting. This was demonstrated from the review of National nutrition monitoring bureau (NNMB) data from their surveys in 1975-79 and 2012-13 (17). Stunting among Indian adolescents is lower compared to adolescents of other countries in Asia [48% in rural Bangladesh (18) and 23.7% in Malaysia (19)] and other African countries [12.2% in Ethiopian adolescent girls (20) and 9.7% in Somalian refugee camps (6)]. No significant difference between gender at any age was reported in contrast to the present findings (21). The risk of stunting increases with the increase of age. This is attributable to different chronic factors such as food security, education, gender disparities, women's health (22). Furthermore, repeated infections and environmental stress apart from poor nutrition collectively from intrauterine stages through adolescence culminate in stunting (22). The odds of being stunted were high among North-eastern India, lowest wealth index families, scheduled tribes, scheduled caste, never attending the school, and Muslim religion consistent with predictors from the secondary analysis on NFHS data (16). Increased odds of stunting in North-eastern India compared to northern India may be attributable to the lower growth potential of the hilly north-eastern region. At the same time, the other factors are all related to vulnerability, affordability, and deprivation.

In the current study, thinness was higher in boys (29.9%) than girls (18.9%). In NFHS-3, the reported prevalence of thinness was 22.3% (boys) and 9.9% (girls), which subsequently decreased to 16.5 and 9% in NFHS-4 in boys and girls, respectively. In many neighboring countries, thinness among boys is higher than girls, viz Indonesia (11 vs. 5%), Bangladesh (19.6 vs. 15.4%), and Nepal (37.8 vs. 26.2%) (23–25). This pattern of higher prevalence of thinness among boys is similar to aggregates based on 57 LMICs (6.7% in boys and 4.5% in girls) (15). It might be attributed to the increased gain in height among boys compared to girls leading to higher levels of thinness.

In the present study, the prevalence of thinness was 27.6% in early adolescence and 20.9% in late adolescence, which was similar to findings by Kumar P et al. observed in adolescents of Bihar and Uttar Pradesh (26). This could be due to increased growth spurt in the early adolescent stage compared to the late adolescent stage (20). This might also be due to higher calorie intake deficiencies among early adolescents than late adolescents (27). In the current study, apart from being a boy or belonging to early adolescence, the odds of being thin were high among those from—other backward classes, lowest wealth index families, rural residence, north-eastern India, or whose mothers had no schooling, all of which were associated with increased risk of undernutrition.

Our analysis has shown a marginally higher prevalence of overweight (BAZ > 1SD) among boys (4.9%) compared to girls (4.7%), with no statistically significant difference. The prevalence rates of overweight reported in NFHS-4 were slightly higher than CNNS data, and there was a marginal increase between NFHS-3 (3.0% in boys and 4.3% in girls) and NFHS-4 (6.2% in boys and 4.9% in girls) (16). The burden of overweight is lower in India compared to international estimates. Systematic analysis of 1,769 studies has shown that in developing countries, the prevalence of overweight among children and adolescents has increased from 8.1 to 12.9% in boys and 8.4 to 13.4% in girls between 1980 and 2013 (28). Late adolescents (4.1%) had a lower burden of overweight than early adolescents (5.2%) and had statistically significant lower odds of being overweight. Adolescents residing in urban areas and those of the highest wealth index families were at increased odds of being overweight compared to their counterparts. The odds of being overweight were higher among adolescents from the south and west India compared to those residing in Northern India.

The literature about coexisting stunting with thinness or obesity is sparsely available amongst adolescents in India. The UDAYA study conducted in Uttar Pradesh and Bihar states reports that the co-existence of stunting with thinness was seen in 11 percent of boys and 8 percent of early adolescent girls (26).

The strength of the present analysis is the use of CNNS data, a large, quality-controlled, recent, and representative national sample. We have used BAZ scores, an appropriate indicator for reporting nutritional status among adolescents unlike BMI which is appropriate for adults and used in NFHS surveys. We have not analyzed and triangulated other forms of malnutrition, including the micronutrient deficiencies and co-existence of different forms of malnutrition which is a limitation of this analysis. Another limitation of this analysis is that the socio-demographic predictors of malnutrition among adolescents needs to be triangulated with clinical characteristics and dietary intakes. Since adolescent malnutrition especially stunting and thinness, are high, correcting these in this age group could be the game-changer in breaking the intergenerational cycle of malnutrition. With the identified burden of overweight and obesity, interventions focusing on the early lifestyle changes in this age group must be taken up through already existing national health programs. Identification of state-specific determinants of malnutrition coupled with effective interventional strategies can benefit in a targeted reduction in the burden of malnutrition. Growth monitoring and nutritional surveillance like quarterly anthropometric assessment can contribute to the real-time correction of malnutrition among adolescents. School health programs and POSHAN Abhiyan (Prime Minister's Overarching Scheme for Holistic Nourishment) must be linked with the periodic anthropometric assessment of this age group.