Anaemia in Indians aged 10–19 years: Prevalence, burden and associated factors at national and regional levels

Abstract Anaemia control programmes in India are hampered by a lack of representative evidence on anaemia prevalence, burden and associated factors for adolescents. The aim of this study was to: (1) describe the national and subnational prevalence, severity and burden of anaemia among Indian adolescents; (2) examine factors associated with anaemia at national and regional levels. Data (n = 14,673 individuals aged 10–19 years) were from India's Comprehensive National Nutrition Survey (CNNS, 2016–2018). CNNS used a multistage, stratified, probability proportion to size cluster sampling design. Prevalence was estimated using globally comparable age‐ and sex‐specific cutoffs, using survey weights for biomarker sample collection. Burden analysis used prevalence estimates and projected population from 2011 Census data. Multivariable logistic regression models were used to analyse factors (diet, micronutrient deficiencies, haemoglobinopathies, sociodemographic factors, environment) associated with anaemia. Anaemia was present in 40% of girls and 18% of boys, equivalent to 72 million adolescents in 2018, and varied by region (girls 29%–46%; boys 11%–28%) and state (girls 7%–62%; boys 4%–32%). Iron deficiency (ferritin < 15 μg/L) was the strongest predictor of anaemia (odds ratio [OR]: 4.68, 95% confidence interval [CI]: [3.21,6.83]), followed by haemoglobinopathies (HbA2 > 3.5% or any HbS) (OR: 2.81, 95% CI: [1.66,4.74]), vitamin A deficiency (serum retinol <20 ng/ml) (OR: 1.86, 95% CI: [1.23,2.80]) and zinc deficiency (serum zinc < 70 μg/L) (OR: 1.32, 95% CI: [1.02,1.72]). Regional models show heterogeneity in the strength of association between factors and anaemia by region. Adolescent anaemia control programmes in India should continue to address iron deficiency, strengthen strategies to identify haemoglobinopathies and other micronutrient deficiencies, and further explore geographic variation in associated factors.


| INTRODUCTION
Globally, one in four individuals aged 10-24 years (~430 million) suffer from anaemia, with the highest prevalence found in low-and middle-income countries (Azzopardi et al., 2019). As the years between childhood and adulthood represent a sensitive period for developmental, physiological and behavioural changes, anaemia in this formative phase of life can reduce work capacity, impair neurocognitive and pubertal development and increase susceptibility to infections (Y. Balarajan et al., 2011;Haas & Brownlie, 2001;Scott et al., 2018). For adolescent girls also entering pregnancy (~21 million cases annually; Darroch et al., 2016), the consequences of anaemia extend to maternal and neonatal mortality as well as poor birth outcomes (Y. Balarajan et al., 2011;Black et al., 2013;Figueiredo et al., 2018;Haider et al., 2013).
India is home to 253 million adolescents 10-19 years of age, among the largest cohorts globally. Limited nationally representative nutrition survey data exist for this age group. National Family Health There are nutritional and nonnutritional causes of anaemia: micronutrient deficiencies anind genetic blood disorders, including haemoglobinopathies, inflammation, infectious diseases and other physiological conditions such as menstruation and pregnancy (Y. Balarajan et al., 2011;Chaparro & Suchdev, 2019;GBD 2017 Disease andInjury Incidence andPrevalence Collaborators, 2018;Karakochuk et al., 2015;Nguyen et al., 2015;World Health Organization [WHO], 2017). In India, a recent analysis characterized types of anaemia among children and adolescents aged 1-19 years .
Among anaemic adolescents in this study, 21.3% had iron deficiency only (iron deficiency anaemia), 25.6% had folate or vitamin B12 deficiency without iron deficiency (folate or vitamin B12 deficiency anaemia), 18.2% had iron deficiency plus folate or vitamin B12 deficiency (dimorphic anaemia), 31.4% had no iron or folate or vitamin B12 deficiency (anaemia of other causes) and 3.4% had anaemia of inflammation. Anaemia is also associated with several individual-and household-level characteristics such as education, age at marriage and wealth (Y. S. Balarajan et al., 2013;Chakrabarti et al., 2018;Nguyen et al., 2018;Prieto-Patron et al., 2018). Existing studies on factors associated with anaemia in India mostly focus on women of reproductive age and children (Y. S. Balarajan et al., 2013;Nguyen et al., 2018;Varghese & Stein, 2019;, although there are some small studies on adolescents (Ahankari et al., 2017;Bharati et al., 2009;Mukherjee, 2016;Thomas et al., 2015;Toteja et al., 2006). Only one of these studies (Ahankari et al., 2017) assessed Using data from CNNS, this paper (1) describes the national and regional prevalence, severity and burden of anaemia in adolescents aged 10-19 years in India; and (2) examines factors associated with anaemia in this population, at both national and regional levels, to inform public health solutions to reduce adolescent anaemia in India.

Key messages
• The Comprehensive National Nutrition Survey (CNNS) 2016-2018 is the only nationally representative survey to measure haemoglobin, haemoglobinopathies, biomarkers of micronutrient deficiencies, diet and social factors in Indian adolescents aged 10-19 years.
• Our prevalence estimates translate to 72 million adolescents being anaemic, with Uttar Pradesh having twice as many anaemic adolescents as any other state.
• Iron deficiency, haemoglobinopathies, vitamin A deficiency and zinc deficiency were associated with an increased risk of anaemia; these associations may vary subnationally. factors (e.g., education level) and household factors (e.g., sanitation). A subsample of 14,669 adolescents with a valid measure of haemoglobin (Hb) level and survey weights were included in the prevalence and burden analyses. Biological samples to assess micronutrient status were collected from two out of three adolescents, selected using systematic random sampling (MoHFW et al., 2019). Accounting for missing variables on biomarkers, the multivariable models included 6156 adolescents overall, of which 3058 boys and 3098 girls ( Figure 1).

| Biomarker measurement
The following anaemia-related parameters were analysed: Hb, Hb

| Outcome
Anaemia was the primary dependent variable, defined as altitudeadjusted Hb < 11.5 g/dl for boys and girls 10-11 years, Hb < 12.0 g/dl for boys 12-14 years and girls 12-19 years and Hb < 13.0 for boys 15-19 years following standard WHO cutoffs (WHO, 2011a) (Supporting Information: Table S1). We used haemoglobin cutoffs currently recommended by the WHO as these cutoffs allow for international and interstudy comparability, although we note a recent call for a reexamination of the cutoffs in the Indian population (Sachdev et al., 2021). For descriptive analyses, we also report the prevalence of mild (boys and girls 10-11 years: 11.0-11.4 g/dl; boys 12-14 years and girls 12-19 years: 11.0-11.9 g/dl; boys 15-19 years: 11.0-12.9 g/dl), moderate (8.0-10.9 g/dl) and severe (<8.0 g/dl) anaemia, again following standard WHO cutoffs (WHO, 2011a).

| Predictors
We developed a conceptual framework based on existing frame- literature on risk factors and pathophysiology of anaemia (Chaparro & Suchdev, 2019), as well as data availability in the CNNS ( Figure 2).
Broadly, we classified factors (predictors) as either proximate or distal.
Proximate factors included dietary factors, micronutrient deficiencies and the presence of genes for haemoglobinopathies. Diet was assessed using a food frequency questionnaire, which asked adolescents how many days in a typical week they consumed various food groups.
We constructed an 'animal source foods' indicator from eggs, fish, chicken or meat, coded as 1 if any of these foods were consumed at least one day per week and 0, if otherwise. Consumption of supplements or tablets included consumption of IFA supplements in the last week, multivitamin supplements in the last month and deworming tablets in the last 6 months. Micronutrient deficiencies were constructed as binary variables following sex-and age-specific WHO cutoffs (Benoist, 2008;Namaste et al., 2017;WHO, 2011aWHO, , 2011b: iron deficiency as ferritin < 15 ng/ml; folate deficiency as erythrocyte folate < 151 ng/ml; vitamin B12 deficiency as serum B12 < 203 pg/ml; vitamin A deficiency as serum retinol < 20 μg/dl; vitamin D deficiency as serum 25(OH)D < 20 ng/ml; zinc deficiency as serum zinc < 66 g/dl to <74 μg/dl depending on sex, gender and fasting status (Supporting Information: Table S1). To account for the effects of inflammation on nutritional status biomarkers, individuals with CRP > 5 mg/L (3.9%) were excluded from the analyses ). An indicator of haemoglobinopathies was set equal to 1 if the adolescent had either thalassaemia trait (HbA2 > 3.5%) or sickle cell β thalassaemia (any HbS) and 0 if both tests were negative.
Distal factors included age (10-14 or 15-19 years), area of residence (rural/urban), wealth index, caste (scheduled caste/tribe, other backward classes (i.e., disadvantaged groups in India) and others), religion (Hindu and non-Hindu), schooling status, parent's education (literate or illiterate), environmental factors (improved toilet, water and soap for handwashing), exposure to sources of information that may be educational with respect to health, such as mass media (newspaper/ radio/TV), access to school meals) and geographical region of residence.
Principal component analysis was used to construct the asset-based household wealth index, following Demographic and Health Survey guidelines (Rustein & Johnson, 2004), which was further categorized into quintiles, where the highest quintile represented the richest and lowest quintile the poorest.

| Statistical analysis
Descriptive analyses were conducted to report the characteristics of the study sample and provide regional estimates for anaemia (classified by types and severity). Maps were used to visualize state-wise variability in anaemia prevalence and burden separately for boys and girls. Burden of anaemia was calculated as the product of the anaemia prevalence from CNNS data and the total eligible projected adolescents for each state in 2018 from Population Projections for India and States 2011-2036 (National Commission on Population, MoHFW, 2020), which estimated population on the basis of the Census 2011 age-sex data. National and regional multivariable logistic regression models were used to examine associations between anaemia and its associated factors among adolescent boys and girls. Model fit was assessed using the Hosmer-Lemshow goodness-of-fit test and Akaike and Bayesian information criterion (Vrieze, 2012). In the regional multivariable regression models, sex-specific analyses were not conducted owing to limitations in sample size. To check for the robustness of the main results, we applied structural equation modelling with a maximum likelihood for estimation to account for missing values (Allison, 2003).
All analyses accounted for the multistage cluster sampling design and F I G U R E 2 Conceptual framework for factors associated with anaemia in adolescents. Boxes with a solid outline are factors included in the regression analysis and boxes with a grey dotted outline are factors not included. Data on food security were collected in the Comprehensive National Nutrition Survey, but were not publicly available at the time of writing. WASH, water sanitation and hygiene.
survey weights specific to biomarker data. All analyses were conducted using Stata v.17.0.

| Sample characteristics
Animal source foods were consumed at least once per week by 46.3% of adolescents and consumption was slightly more frequent in boys compared with girls (Table 1). Only 8.9% of adolescents consumed IFA supplements in the last week and 25.8% consumed deworming tablets in the last 6 months. Micronutrient deficiencies were common in both sexes, with iron and vitamin D deficiencies higher in girls and B12, folate and zinc deficiencies slightly higher in boys. Overall, folate deficiency (35.7%) was the most common and vitamin A deficiency was the least common (12.3%) micronutrient deficiency. Among distal factors, 74.3% of the sample lived in rural areas, 81.8% were Hindu, 32.7% were scheduled caste or tribe, and 77.3% were currently in school. Less than half of adolescents had parents who were both literate (42.4%) and around half had access to improved sanitation (55.8%) or materials for handwashing (50.9%). Sixty percent of adolescents were exposed to mass media, such as television, radio, newspaper or magazines less than once per week. One in four had access to a mid-day meal in school.
3.2 | Prevalence of adolescent anaemia at national, regional and state levels The prevalence of anaemia among adolescents was 28.5% overall, corresponding to more than 72 million adolescents ( Figure 3, panel a).
Anaemia was higher in girls (39.6%, 48.7 million) than boys (17.6%, 23.7 million). In terms of severity, 17.6% of adolescents had mild anaemia, 10.0% had moderate anaemia and 0.9% had severe anaemia. Anaemia was highest among girls aged 15-19 years (47.5%) and lowest among adolescent boys aged 10-14 years (17.1%). Anaemia prevalence varied widely across regions, ranging from 29.0% in the South to 45.8% in the East for girls; and from 10.8% in the South to 28.4% in the Northeast for boys ( Figure 3,
T A B L E 2 Odds of anaemia by proximate and distal factors in Indian adolescents aged 10-19 years, results from the national-level multivariable logistic regression models

| Comparison with other studies
Other studies have reported on the strong association between iron deficiency and anaemia in Indian adolescent boys and girls, but none to our knowledge have done so at national and regional levels. Our findings are consistent with recent studies on school-aged children and adolescents from Nepal (Ford et al., 2020a), Kuwait (Shaban et al., 2020) and Bangladesh (Ahmed et al., 2000), and communitybased studies on adolescent girls in India (Patel et al., 2017;Srivastava et al., 2016;Thomas et al., 2015). Similar associations between anaemia and iron deficiency are well documented in women of reproductive age and children Ford et al., 2020b;George et al., 2012;Petry et al., 2019;.
Although IFA has been recommended and guidelines for weekly IFA supplementation exist, coverage of IFA supplements among adolescents was low (9%), suggesting supply, access and/or adherence issues. Consumption of IFA supplements in the last week was not a significant predictor of anaemia in the overall national model or in any region, except for among boys where IFA consumption predicted higher odds of anaemia. We speculate that this may simply reflect better adherence among anaemic boys compared with nonanaemic boys.
Our findings on associations between vitamin A deficiency and anaemia are aligned with previous studies on adolescents in Nepal (Ford et al., 2020a) and Bangladesh (Ahmed et al., 2000). Vitamin A deficiency is known to be associated with reduced iron binding capacity and transferrin saturation. Additionally, vitamin A also modulates iron homeostasis by regulating hepcidin synthesis and plays a critical role in immune modulation, with vitamin A deficiency increasing susceptibility to anaemia of infection (da Cunha et al., 2014;Semba & Bloem, 2002). Similar findings on the relationship between zinc status and anaemia have been reported from studies on school-aged children and adolescents in New Zealand (Houghton et al., 2016) and Turkey (Atasoy & Bugdayci, 2018).
Folate deficiency was inversely associated with anaemia in our study. Previous cross-sectional studies have also reported similar findings (Arsenault et al., 2009;Caicedo et al., 2010;Morris et al., 2007;Rogers et al., 2003;Saraya et al., 1973). Mechanistically, this inverse association has been attributed to competition between iron and folate. The intestinal transporter protein, essential for normal iron and folate absorption and homeostasis (PCFT/HCP1), has a higher affinity for folate (Arsenault et al., 2009;Qiu et al., 2006;Shayeghi et al., 2005). This can lead to a competitive reduction of haem--iron absorption, resulting in lower haemoglobin synthesis (Arsenault et al., 2009). In contrast, an analysis of nationally representative data from 10 surveys Merrill et al., 2017) studying factors associated with anaemia in children and women of reproductive age showed no association between anaemia and folate deficiency, a finding additionally supported by a systematic review on the haematologic effects of folate deficiency (Metz, 2008). Therefore, we suggest caution when interpreting our finding on folate deficiency being associated with reduced odds of anaemia.  The presence of thalassaemia trait (HbE) or sickle-cell βthalassaemia was associated with higher odds of anaemia. Genetic haemoglobin disorders can be homozygous (manifesting in the disease) or heterozygous (trait) and can lead to defective formation of haemoglobin, thereby increasing the risk of anaemia. In India, previous studies have investigated the prevalence of thalassaemia traits and sickle cell disease (Bhukhanvala et al., 2012;Madan et al., 2010;Mohanty et al., 2008Mohanty et al., , 2014; however, these studies did not examine the association of haemoglobinopathies with anaemia in adolescents. Thalassaemia traits have been found to be associated with anaemia in a nationally representative sample of Malawian children (McGann et al., 2018), cohorts of rural children in Karnataka (Pasricha et al., 2010), India and young children and women from Cambodia (George et al., 2012;Karakochuk et al., 2015). This finding underscores the importance of screening programmes in schools for the assessment of thalassaemia traits, supplemented by necessary counselling. India currently has detailed guidance from the National Health Mission on population-level screening programmes for the detection of carriers of β-thalassaemia HbS and HbE (MoHFW, 2016;Patra et al., 2015). The policy on prevention and control of haemoglobinopathies also recommends community education and awareness generation, though the status of implementation for these programmes remains unclear.
We found few associations between nonnutritional or distal factors and anaemia other than being older and female being associated with higher odds of anaemia compared with being younger and male. Wealth, parental literacy, sanitation, hygiene, mass media exposure and receiving free school meals were not associated with anaemia. Previous studies that have found associations between nonnutritional factors and anaemia have not controlled for micronutrient status (Ahankari et al., 2017;Nguyen et al., 2018). It may be that inclusion of additional proximate factors masks associations between distal factors and anaemia. To test this hypothesis, we ran additional exploratory models (Supporting Information: Table S2). In bivariate analysis, odds of anaemia were lower among adolescents in school (compared with those out of school), from richer households, when parents were literate, when adolescents had access to soap and water for handwashing, when

| Strengths and limitations
Our study provides a current description of the prevalence, burden and associated factors of anaemia in a nationally representative sample of Indian adolescents. In addition to providing data on an understudied age group-NFHS includes 15-19-year-old but not 10-14-year-old adolescents-CNNS uses gold standard methods for estimation of anaemia and other micronutrient deficiencies. Most previous field surveys, including NFHS, have used HemoCue 201+ for estimating haemoglobin concentrations from capillary blood samples, which is known to overestimate anaemia prevalence in hot and humid environments (Whitehead et al., 2019). Inaccuracies with the HemoCue method have also been attributed to differential dilution pressure due to milking, skin temperature and depth of needle penetration (Boghani et al., 2017;Gwetu et al., 2013). CNNS used an automated haematology analyser, which provides higher precision and accuracy and is based on the WHO-recommended cyanohaemoglobin method of estimation MoHFW et al., 2019). The different methods of Hb assessment between NFHS and CNNS are responsible for the different prevalence estimates.
Our study is not without methodological limitations. First, CNNS is cross-sectional in nature and hence precludes inference of any causal relationship between anaemia and its associated factors.
Second, exclusion of individuals with elevated CRP may introduce bias, but as AGP was not available and as most CRP values were clustered at low CRP levels, it was a challenge to address the effects of inflammation on acute-phase proteins using a regression-based approach . However, we conducted a sensitivity analysis to compare the results of regression models with and without individuals with elevated CRP and found no differences (Supporting Information: Table S2). Third, dietary data only included the frequency of food group consumption; a detailed dietary assessment is necessary to study the potential contribution of dietary constituents to adolescent anaemia. Fourth, our regional analysis should be treated as exploratory and interpreted with caution due to the limited sample size, particularly in the Central (n = 537) and West (n = 442) regions. This issue may underlie unexpected significant associations such as the negative association between vitamin B12 deficiency and anaemia in the West. Ideally, we would have been able to conduct a state-level regression analysis given that many decisions are made at the state level in India, but the sample with biomarker data was far too small for such an analysis.

| Policy and programme context and implications
India is among the few countries in South Asia and the world to have a comprehensive programme to address anaemia in the adolescent While there is a growing consensus that nutrition-specific strategies alone will not end anaemia, our findings highlight the need for nutrition-specific strategies that focus on iron but also address deficiencies in other micronutrients such as vitamin A and zinc. The IFA supplementation strategy has supply and compliance barriers (Ramakrishnan et al., 2012;Sethi et al., 2017), which still need to be

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.