National surveys and regional cross-sectional studies do not provide a robust indication of secular change and its quantum, especially in those reaping the benefits of development, for example, the middle socio-economic strata in urban settings. Information of this nature would provide a reasonable yardstick for improvement goals in childhood anthropometry, especially if novel nutrition interventions are being considered. Further, such data would identify the potential burden of over-nutrition in children born to relatively undernourished parents; a nutrition transition pattern that could explain the emerging epidemic of non-communicable diseases (NCDs) in the country [7].

A robust estimate could be obtained through an intergenerational cohort design. This would offer the advantages of partially controlling for genetic variations while comparing anthropometry of children with their parents at similar ages. To the best of our knowledge, no such data are available in the Indian context to inform policy. This communication reports the intergenerational change in anthropometric indices of children below ten years of age and their predictors in the New Delhi Birth Cohort (NDBC).

Statistical analysis: Data analysis was performed using SPSS 20.0. The intergenerational change in anthropometry was computed using two types of standardized scores (Z-scores); the World Health Organization (WHO) reference and internally within the cohort. For the former, F2 and F1 generation anthropometry was converted to WHO length/height-for-age, weight-for-age and BMI-for-age Z-scores at the date of measurement, using an SPSS macro for age-groups 0-5 years [11] and 5-19 years [12]. For the internal cohort Z-scores, the longitudinal height, weight and BMI measurements for the F1 generation were modelled into growth charts using Cole’s LMS method [13,14]. These charts were used to compute age- and sex-specific Z-scores for F2 children at the date of measurement. To compare intergenerational anthropometry at similar ages, the F1 Z-score was interpolated at the exact age of the F2 measurement, using neighbouring F1 measurements. The interpolation was acceptable if the F1 observation was within 6 months for ages up to 1 year; within 1 year for ages 1 to 2 years; within 1.5 years for ages 2 to 3 years; and within 2 years thereafter. To estimate the comparison in absolute units, the interpolated F1 Z-score at the exact date of F2 measurement was back-transformed using the L, M and S values for the cohort Z-scores [15].

The intergenerational change in size was computed as F2-F1. This intergenerational model had a 3-level data structure comprising the F0 generation having one or more F1 children (F1-F1 siblings) and the F1 generation having one or more F2 children (F2-F2 siblings). Thus, a mixed model [16] approach was used to allow for this structure. There were 13 sibships among the F1 generation. We fitted models to allow for this third level of variation, but there were too few sibships to disturb the intergenerational effect size estimates and their standard errors; so subsequently we ignored this variance component.

The change was quantified in two age groups: 0-5 years and 5-10 years, adjusting for all combinations of the sex of the child and parent. This intergenerational change was further adjusted for socio-demographic characteristics. These included maternal (F0) schooling, wealth and water supply, sanitation and hygiene (WASH) score at F1 birth; and education of the F1 parent (cohort subject), wealth and WASH score at F2 measurement. We used the 1st principal component score [17] to derive the wealth and WASH scores at F1 birth and F2 measurement from the individual variables (listed in the footnotes of the relevant tables). These covariates were used in multivariate models as standardized units (mean 0), so that uncentered variables do not cause shifts in computed intergenerational changes.

Among the 337 F1 participants there were 13 pairs of F1-F1 siblings and among the 432 F2 children, 245 were single children, 89 were paired siblings and 3 were three siblings (Fig. 1). We thus analyzed 432 F2-F1 pairs, which included 178 (41.2%) father-son, 138 (31.9%) father-daughter, 71 (16.4%) mother-son and 45 (10.4%) mother-daughter comparisons. Among them, 116 (72 boys) were aged 0-5 years and 316 (177 boys) were 5-10 years old, with mean (SD) ages of 3.5 (1.2) and 7.9 (1.3) years, respectively. The mean (SD) height-for-age, weight-for-age and BMI-for-age (WHO Z-scores) were -0.97 (1.39), -0.68 (1.19) and -0.08 (1.37), respectively for younger children and -0.42 (1.22), -0.31 (1.55) and -0.12 (1.49), respectively for older children.

Fig. 1 The three level structure of the intergenerational data set.

The socio-demographic characteristics at the time of the parent’s (F1 generation) birth and at the time of the child’s (F2 generation) measurement are summarized in Web Table 1. Comparatively, parents (F1) had a poorer socio-demographic profile; one-third were residing in a flat or bungalow, the mean (SD) members sharing a room were 3.6 (1.7), four-fifths had shared toilet facilities, 60% had a common water supply and only one-fourth of their F0 mothers had completed 10 or more years of education. In contrast, at the time of the child’s measurement, almost all of them were living in either flat or independent house, mean (SD) members sharing a room were 1.9 (0.9), most had separate toilet facilities (96%) and water supply (81%), and 62% of their F1 mothers had 15 or more years of education (graduate).

TABLE I Intergenerational Changes in Anthropometry

Using the WHO definition [18], in comparison to their parents, under-five children had a substantially lower prevalence of stunting (44.7% vs 18.4%; n=114) and underweight (25.9% vs 12.5%; n=112). However, wasting remained unchanged (4.6% vs 5.5%; n=109).

TABLE II Predictors of Intergenerational Change in Anthropometric Indices (Multivariate Model)

Table II depicts the multivariate associations of intergenerational change. Parents’ (F1) anthropometry had the most consistent and strongest associations (P <0.001), with effect sizes ranging from 0.47 to 0.65 SD decrease per SD. Thus children of shorter, lighter and thinner parents had gained more in height, weight and BMI, respectively. The age of the parent (F1) at child birth and the age of the child (F2) at measurement were significant (P<0.001) positive predictors for F2 height-for-age and weight-for-age but not for BMI-for-age. Amongst the socio-demographics characteristics, poorer WASH status at the F1 parents’ birth predicted greater increase in all three F2 indices. In contrast, higher parental (F1) education predicted greater F2 gain in weight and BMI, and had no association with F2 height gain. Table III quantifies the adjusted intergenerational changes across sub-groups of various predictors.

TABLE III Adjusted Intergenerational Changes Across Different Subgroups of Predictors (Multivariate Model)

The scatterplot in Fig. 3 illustrates the transition in BMI categories among parents and children at the same age. The cut-offs were slightly modified from WHO recommendations [19] to ensure sufficient numbers in each category: Thin <-1.5 Z, Normal -1.5 to 0.5 Z, and Overweig`ht >0.5 Z. In comparison to 127 children, only 29 parents were overweight at the same age. While a higher proportion of children of overweight parents were overweight, the greatest increase in BMI categories occurred in thin parents. A quarter of their children were overweight while half were normal. In contrast, among all children born to overweight parents, only 41% were overweight and 59% were normal.

These categories were defined using WHO BMI Z-score as follows: Thin <–1.5, Normal –1.5 to 0.5, and Overweight >0.5. Fig. 3 Scatterplot of transition in Body Mass Index (WHO Z scores) in parents and their children at corresponding ages.

In this intergenerational study, we documented that children below ten years of age were considerably taller (~1 SD) and heavier (0.8 to1.5 SD), while only 5-10 years old were broader (~1 SD), than their parents at corresponding ages. Independent predictors of greater gains in children included parents with lower anthropometric indices and poorer WASH facilities; higher age of parents at child birth and of children at measurement (height and weight); and more parental education (weight and BMI).

Important limitations of this study include a somewhat small sample size based on a proportion of the currently available cohort. Among the parents, there was a considerably higher representation of fathers because outmigration after marriage was common in female participants (F1). The main strengths of our study include the comparison of children and their parents at corresponding ages from carefully collected prospective data for the latter, an urban LMIC setting of relatively rapid socio-economic development, and appropriate multi-level modelling with available confounder adjustment.

In the absence of targeted food or nutrient supplementation, it is reasonable to ascribe these intergenerational gains to improvements in general living conditions. However, from a policy perspective, filtering out important predictors is desirable. Poorer water supply and sanitation facilities of the parents (F1) at birth predicted greater F2 gains. A Cochrane systematic review of cluster randomized trials, intervening for only 9-12 months also suggests that WASH interventions confer a small benefit on linear growth (~0.1 SD) in under-five children [23]. The current national impetus on "Swach Bharat Abhiyaan" is therefore timely and appropriate. In conformity with earlier experiences [24,25], higher parental literacy was a positive predictor, thereby re-emphasizing the importance of improving education. Higher ages of parents at child birth and of children at measurement were important positive predictors of intergenerational gains in height and weight. A quadratic relation of childhood anthropometry with age of child birth has been documented in pooled analyses of cohorts from low- and middle- income countries (LMICs) [26]; our sample was probably constituted by the linear component of this association. However, both these variables are also proxies for exposure duration, thereby suggesting that sustained improvement in living conditions resulted in greater benefit.

Poorer parental anthropometry had the strongest and most consistent predictive ability for greater intergenerational gains. Potential explanations for this include statistical regression to the mean, narrowing of socio-economic inequalities and greater biological response among deprived strata. This observation augurs well for attempting equity for secular increases in the height of populations. However, excessive BMI gain in children of thin parents, if primarily due to increased adiposity, could be providing the backdrop for the current escalation of cardio-metabolic risk factors in Indian children and adolescents [7,27,28]. This hypothesis is in concordance with our earlier observation of increased risk of diabetes mellitus in adults who were relatively thin as children but continued to become obese relative to themselves [9].

Under-five children were considerably taller and heavier but only older subjects were additionally broader than their parents. This provides more direct evidence of the earlier postulate, based on cross-sectional comparisons [29], that children "grow up" (get taller) before "growing out" (get broader). This observation is also consistent with the increased prevalence of obesity in older children from India [28] and other countries [29-31]. The underlying mechanisms for this phenomenon are unclear to us and merit further exploration.

Our data suggest that sustained improvement in general living conditions leads to considerable increases in height, weight and BMI within one generation. The current governance focus on inclusive development is therefore apt, especially if the benefits percolate preferentially to the underprivileged. Isolated vertical interventions (for example, nutrient supplementation) should only be entertained if there is convincing evidence of substantial benefit above that expected from developmental transition. Findings from the NFHS-4 survey of notable reductions in stunting and underweight with nearly stagnant or even increased wasting prevalence in some states should not fuel exaggerated concerns and action to screen for and treat for severe acute malnutrition; as this phenomenon may occur in a population undergoing development-related anthropometric transition. Vigilance may be required to address the potential of greater cardio-metabolic risk in families showing large intergenerational increases in BMI.

In conclusion, over one generation in an urban middle-class population, whose general living conditions had improved, under-five children have become considerably taller and heavier and 5-10 year old children have additionally become broader, than their parents at corresponding ages. Child populations probably "grow up" before "growing out".

Contributors: SS, ARA, HPS, CHDF and SKB: conceptualized the study. SS, ARA, CO and HPS: analyzed the data. SS: drafted the initial manuscript. All authors contributed to the critical revision of the article.

Funding: Indian Council of Medical Research for supporting Ms. Sikha Sinha through the Senior Research Fellowship Scheme. The original cohort studies were supported by the National Center for Health Statistics, USA and the Indian Council of Medical Research, India. Later phases have been supported by Department of Biotechnology, India; British Heart Foundation, UK; Wellcome Trust, UK; Medical Research Council; UK; Department for International Development, UK; and Bill and Melinda Gates Foundation, USA.

Competing interests: None stated.