Untangling weight from height: A generalised body mass index (gBMI) approach to evaluating child and adolescent nutritional status in India

Methods

This observational study based on data from multiple cross-sectional national surveys was undertaken by the department of Biostatistics and Epidemiology, St John’s Research Institute, Bengaluru, India between May 2024 and August 2025. Ethical approval for the study was obtained from the Institutional Ethics Committee of St. John’s Medical College, Bengaluru, India, which is registered with the regulatory authority. Multiple rounds of National Family Health Surveys (NFHS)^22-24 and Comprehensive National Nutritional Survey (CNNS)²⁵ data were utilised to develop the gBMI, which was validated against a cross-sectional data set collected from apparently healthy children aged 5–16 years attending urban schools in Bengaluru, Karnataka, India, between December 2011 and October 2013. All assessments were conducted at St. John’s Research Institute and Medical College Hospital with ethical approval by the Institutional Ethics Committee, parental consent, and child assent (for those over 10 years).

Extraction of healthy subset

A subset of healthy children from above mentioned national surveys was selected using selection criteria adopted from WHO Multicentre Growth Reference Study (MGRS)²⁶ to ensure that study participants had optimal growth potential under ideal conditions. The description of detailed selection criteria can be found in Majumder et al 2024²⁷ but summary diagram is depicted in Figure 1.²⁸

Close

Fig. 1. Steps of selection of analytical sample [anaemia is defined as per World Health Organization (WHO) criteria²⁸].

Export to PPT

Applying these to NFHS-3, 4, 5, and CNNS identified 13,245 children under five years (1,821; 4,531; 4,918; 1,934, respectively). From CNNS, 3,583 children aged 5–9 years and 3,077 aged 10–19 years were included. About 85% belonged to the top wealth quintile. Anthropometric Z-score distributions were comparable across surveys.²⁹

Results

After excluding the upper and lower 5% of calculated BMI values, the final analytical ‘healthy child’ sample included 12,466 under 5-year children (CNNS: 1,824; NFHS-3: 1,785; NFHS-4: 4,412; NFHS-5: 4,445) and 6,487 children aged 5–19 years (CNNS: 3,583 aged 5–9; 3,077 aged 10–19). The dataset comprised 9,991 boys and 8,962 girls: 1,293 boys and 1,279 girls below one year; 5,269 boys and 4,625 girls aged 1 to below five years; 1,881 boys and 1,626 girls aged 5–9 years; and 3,351 boys and 2,979 girls aged 10–19 years. Detailed age- and gender-specific frequency distributions are provided in Supplementary Figure 1.

The estimated exponent parameter (${\alpha }_t)$ for gBMI was calculated separately for each age (in months) for boys and girls using the analytical sample. Age-dependent exponent parameters (${\alpha }_t)$ for Indian children aged 0–19 years are presented in Supplementary Table I, and the trend over age is illustrated in Figure 2. The estimated exponents (${\alpha }_t)$ ranged from 1.3 to 2.9 for both boys and girls. At early ages (0–59 months), both sexes had similar exponent values, which increased simultaneously from 1.3 to 2.6 up to age 6 years. A rapid increase in the exponent parameters was observed between age 8 to 14 years. In this period, exponent estimates ranged from 2.6 to 2.9. For boys, the highest exponent value (2.9) was recorded at age 11, while girls reached their maximum (2.7) at age 10. After this periods, the exponents (${\alpha }_t)$ gradually declined, reaching a value of 2 at age 19 ( Fig. 2). Overall, the exponent values for boys were generally higher than those for girls. Estimated Pearson’s correlation coefficients (with 95% confidence intervals) for the residuals from the height models (m) were calculated for both boys and girls, and their associations are shown in scatter plots ( Fig. 3). The estimated correlations were close to zero for both sexes.

Close

Fig. 2. The figure depicts estimated exponent (αt of height) parameter, ‘‘ of gBMI for boys and girls at each age in months.

Export to PPT

Close

Fig. 3. Association of height with the relative adiposity [i.e., residuals of equation (1)] from the healthy sample of Indian 0-19 yr (A) boys and (B) girls depicted almost null. This confirms that gBMI is free from influence of linear growth of height.

Export to PPT

GAMLSS estimates are presented in Supplementary Table II. Figure 4 displays the gBMI-for-age Z-score (gBAZ) reference centile curves for boys and girls aged 0–19 years. The model demonstrates a satisfactory fit to the data, as indicated by the goodness-of-fit metrics shown in Supplementary Figure 2.

Table summarises the prevalence of acute undernutrition and overnutrition among children and adolescents in different age groups (<5 yr, 5–9 yr, 10–14 yr, and 15–19 yr) using three references: WHO growth standard (WHZ for <5yr and BAZ for >5yr children)³¹ India-specific growth references (WHZ for <5yr and BAZ for >5yr children),³² and the gBAZ (this analysis, for all children). The NFHS 5 data was used for this evaluation in <5yr children and data from the CNNS was used for >5yr children.

For acute undernutrition, as measured by wasting, the WHO standard consistently produced higher prevalence rates across all age groups compared to both the India specific and gBMI references. Among children under five years, the prevalence of wasting according to the WHO weight-for-height standard was 19.2%, 10.9%, according to India-specific weight-for-height growth reference, and 10.7% by the current gBMI reference. Similarly, in the 10–14-year age group, the prevalence of thinness was 22.9% by the WHO BAZ, markedly higher than the estimates by the India-specific reference (5.7%) and the current gBAZ (3.4%) reference (Table).

For overweight (WAZ>2) or obesity (BAZ>2), the WHO standard yielded the lowest prevalence estimates across all age groups. For example, in the 10–14-year age group, the prevalence was 1.3% by WHO standard compared to 1.6% by India-specific growth reference²⁷ and 3.9% by the current gBMI reference. Among adolescents aged 15–19 years, the prevalence by the gBMI reference was significantly higher (8.7%) compared to the WHO BMI standard (0.05%) and the India-specific BMI reference (0.79%) (Table).

Figure 5 shows the association of BAZ and gBAZ with body fat mass (kg) across different Z-score categories within validation data set. The boxplots (Figure 5 for example) demonstrate that, for both indices, FM increased with higher Z-score categories, indicating a positive association. However, in higher Z-score ranges, particularly above 2, the gBAZ showed higher median fat mass values than BAZ, suggesting an improved sensitivity in capturing adiposity. This observation is further supported by the correlation analysis; where gBAZ exhibited a slightly stronger correlation with fat mass (r=0.70) than BAZ (r=0.65). These findings indicate that the gBAZ may serve as a more robust index than the conventional BAZ in reflecting actual fat mass among children and adolescents.

Close

Fig. 4. Centiles of reference growth of gBAZ in Indian 0-19y boys (A) and girls (B). The range between the 2.5th and 97.5th percentiles represents normal or healthy growth. The curves compare the changes in gBMI over age for boy and girl across different levels of possible range of gBAZ in a reference population.

Export to PPT

Close

Fig. 5. Distribution of body fat mass (kg) across different Z-score categories of body mass index (BMI) and generlised body mass index (gBMI). The bold middle lines compare median, height of rectangles compare interquartile range and the whiskers compare entire range.

Export to PPT

Discussion

The present study addresses critical limitations in conventional weight-for-height measures that are used to assess nutritional status among children and adolescents, with reference to acute undernutrition and obesity. This is consistent with previous findings that BMI, although commonly used for adults, does not differentiate between fat mass and fat free mass. Moreover, the dependence of BMI on height is particularly problematic during childhood growth phases and puberty.^1,9,11,12

In response to these challenges, the present analysis has proposed a novel, population-specific metric, the generalised body mass index (gBMI), which adapts the power-type index concept originally described by Benn (1971).¹⁴ By estimating an age- and sex-specific exponent parameter that modifies the weight-to-height relationship, the gBMI isolates weight to be independent of height, thus providing an age-appropriate and more accurate indicator of acute undernutrition and obesity. The exponent values were found to vary dynamically with age and sex, reflecting physiological growth changes, particularly during puberty when rapid changes in body composition occur.^33,34 This dynamic adjustment overcomes the key limitation of BMI, that is, its fixed exponent of 2, which inadequately captures complex growth patterns in developing children.³¹

The study’s robust methodology, utilising large, nationally representative datasets from NFHS and CNNS surveys, ensured the generation of reliable and generalisable reference standards for Indian children and adolescents. The careful extraction of healthy children based on WHO criteria minimised confounding factors, increasing the metric’s validity for optimal growth assessment. Furthermore, the validation using an independent cross-sectional dataset from urban Bengaluru, which included direct measures of body fat through DXA, reinforced the index’s practical clinical applicability.

The gBMI exponent ranged from 1.3 in infancy to peaks of 2.9 in pubertal boys and 2.7 in girls, but thereafter stabilising to 2 by late adolescence, highlighting the physiological rationale for an adaptable exponent. Importantly, correlation analyses confirmed that the gBMI residuals were independent of height, affirming the success of the methodological approach in disentangling weight from linear growth. Several studies have reported that the exponent of height in the BMI formula varies, particularly in children under 5 years and during puberty.^15-20,32 Recent studies have also estimated age-specific height exponents for allometric indices among children across multiple countries including data from India and have proposed pooled or region-specific estimates.^7,21,32 However, none have established a consistent or universal exponent-for-age across different populations using representative, healthy samples of children and adolescents. Therefore, the gBMI formula should be population-specific, or further comprehensive studies are needed to determine a uniform exponent for age applicable across populations.

When comparing prevalence estimates of acute undernutrition and overweight in Indian children, the gBAZ yielded lower estimates of wasting than the WHO WHZ or BAZ standards but aligned more closely with India-specific WHZ or BAZ references.²⁷ Moreover, gBAZ detected a substantially higher prevalence of overnutrition in adolescents compared to the WHO and Indian-specific references,^27,30 highlighting its increased sensitivity to adiposity. Despite the validation dataset having a low prevalence of overweight and obese children, the relatively stronger correlation of gBAZ with measured body fat mass compared to conventional BMI-for-age Z-scores underscores the enhanced capability of gBMI to reflect adiposity more accurately, which is crucial for evaluating targeted nutritional interventions. This attribute confirms the theoretical advantage of using a population- and age-specific power index to assess nutritional status rather than relying on fixed power indices like BMI.

However, this study has several limitations. Estimates derived from cross-sectional health datasets across multiple surveys may be subject to bias. The proposed reference for gBMI-for-age should be validated using longitudinal cohort data before it is considered for policy implementation. Additionally, the validation component is limited by the supplementary dataset, which contains insufficient numbers of overweight or obese adolescents to robustly assess the efficacy of gBMI as a measure of excess adiposity compared to BMI.

In conclusion, the findings from this analysis reveal that the gBMI provides a meaningful improvement over existing indices by offering an adaptable, height-independent, and adiposity-correlated measure for assessing acute undernutrition and overweight/obesity in Indian children and adolescents. The use of gBMI can refine nutritional surveillance, enabling better identification of at-risk populations and more effective public health interventions specific to the diverse physiological growth trajectories in this population. Future research should explore the longitudinal predictive validity of gBMI in relation to health outcomes and its applicability in Indian as well as other populations. Incorporating gBMI into clinical and public health practice warrants evaluation for feasibility and impact on malnutrition monitoring and obesity prevention programmes.