Concordance of weight status between mothers and children: a secondary analysis of the Pakistan Demographic and health survey VII

Abstract

Background

Familial concordance of weight status is an emerging field of study that may guide the development of interventions that operate beyond the individual and within the family context. There is a dearth of published data for concordance of weight status within Pakistani households.

Methods

We assessed the associations between weight status of mothers and their children in a nationally representative sample of households in Pakistan using Demographic and Health Survey data from 2017–18. Our analysis included 3465 mother–child dyads, restricting to children under-five years of age with body mass index (BMI) information on their mothers. We used linear regression models to assess the associations between maternal BMI category (underweight, normal weight, overweight, obese) and child’s weight-for-height z-score (WHZ), accounting for socio-demographic characteristics of mothers and children. We assessed these relationships in all children under-five and also stratified by age of children (younger than 2 years and 2 to 5 years).

Results

In all children under-five and in children 2 to 5 years, maternal BMI was positively associated with child’s WHZ. For all children under-five, children of normal weight, overweight, and obese women had WHZ scores that were 0.21 [95% CI (confidence interval): 0.04, 0.37], 0.43 [95% CI: 0.25, 0.62], and 0.51 [95% CI: 0.30, 0.71] units higher than children of underweight women, respectively. For children ages 2 to 5, children of normal weight, overweight, and obese women had WHZ scores that were 0.26 [95% CI: 0.08, 0.44), 0.50 [95% CI: 0.30, 0.71), and 0.61 [95% CI: 0.37, 0.84] units higher than children of underweight women, respectively. There was no association between maternal BMI and child WHZ for children under-two.

Conclusions

The findings indicate that the weight status of mother’s is positively associated with that of their children, particularly after age 2. These associations further strengthen the call for research regarding interventions and policies aimed at healthy weight promotion among mothers and their children collectively, rather than focusing on individuals in isolation.

Background

There is a growing body of literature focusing on concordance of cardiometabolic disease risk factors and outcomes within families and households, suggesting that noncommunicable disease (NCD) prevention programs may be more effective if the family – rather than the individual – was targeted for intervention. Studies show increased likelihood of adults living with someone with a chronic condition, such as obesity, diabetes, hypertension, hyperlipidemia, and other cardiometabolic conditions, having that same condition. This includes concordance between parent–child [1, 2], sibling [3, 4], spousal dyads [5, 6], and between any co-residing members of households [7, 8]. These studies investigate the genetic predisposition to disease [7], as can be seen in strictly familial relationships, and also the effect of the shared environment [9] and health behavior influences [10, 11], such as in co-residing studies.

Intergenerational concordance within households – and specifically that between parents and children – may reflect a composite of genetic, epigenetic, and shared environmental factors. The magnitude of intergenerational concordance of NCD risk within households in Pakistan is largely unknown. Studies focusing on parent–child concordance have been primarily done in high-income countries, such as in the United Kingdom and Australia [12–17]. A systematic review and meta-analysis looking at high-, middle-, and low-income countries establishes the parent–child concordance of obesity as well, but this study did not include Pakistan and did not analyze continuous child WHZ [18]. There is some additional evidence from low- and middle-income countries (LMICs) regarding concordance of biomarkers such as hemoglobin A1c and C-reactive protein [2]. Specifically in South Asia, studies show statistically significant associations between child under-five obesity and maternal weight, but do not further stratify these relationships by child age [19–21].

Within Pakistan, there is particular importance of the parent–child dyad. Pakistan has one of the highest average household sizes and percentage of multigenerational households in the world, suggesting extended sharing of environment between generations. Moreover, Pakistan is following a global trend of increasing life expectancy, signaling a large population of individuals at higher risk of developing NCDs that the Pakistani healthcare system will have to manage in the future [22]. Additionally, children in Pakistan are becoming increasingly vulnerable to different forms of malnutrition: in the 2018 Pakistan National Nutrition Survey (PNNS), wasting was noted in 17.7% of children, the highest in Pakistan’s history, and prevalence of overweight children under-five almost doubled from 5% in 2011 to 9.5% in 2018. Similarly, women ages 15–49 experienced an increase in overweight and obesity increasing from 28% to 37.8% from 2011 to 2018 [23]. This parallel increase in excess weight across a wide range of Pakistani age demographics may have important implications for cardiometabolic disease at the population level. Thus, there is a need to study Pakistani household concordance of NCD risk factors to further understand the extent, if present, of household concordance in Pakistan, especially with an additional focus on parent–child concordance to advocate for interventions that involve the full household, and not only prevention for adults.

In the present study, we analyzed the concordance of maternal and child weight status, a critical risk factor for NCDs. We further investigate the dynamics of these associations across different age groups of children under 5.

Methods

Data sources and participants

Pakistan’s Demographic and Health Survey, Round VII (PDHS-7) survey was conducted in 2017–18 by the National Institute of Population Studies and funded by the United States Agency for International Development (USAID). PDHS-7 was the 7th round of a nationally representative survey, with information about household, demographic, and maternal and child health indicators. The survey design and sample size calculations were formulated to provide reliable information at the national level and subnational levels, as well as for urban and rural areas separately, with data collected from the four provinces of Punjab, Sindh, Khyber Pakhtunkhwa, and Balochistan, the regions of Azad Jammu and Kashmir and Gilgit Baltistan, the Islamabad Capital Territory, and the Federally Administered Tribal Areas. The sampling design used a two-stage stratified approach and, because of non-proportional sample allocation, the sample weights were generated by the PDHS team. Participants were recruited and data was collected from November 2017 to April 2018.

The respondents included all ever-married women aged 15–49 in all selected households. In one-third of the selected households, height and weight were directly measured for ever-married women 15–49 and children under-five years of age. For this study, we included all children under-five who had WHZ information and BMI information on their mothers, leading to a sampling-weight adjusted sample size of 3465. Figure 1 depicts the sampling scheme and relevant response rates for PDHS-7 [24].

Fig. 1.

Flow diagram of household and participant selection for the 2017-18 Pakistan Demographic and Health Survey. ^*If an EB was greater than 300 households, then the EB was segmented (determined by the field team before household listing) and only one segment was selected to form the cluster (probability of selection was proportional to the household size). ^**No replacements or changes of pre-selected households were allowed during the implementing stage. ^***Survey successfully administered in 561 out of 580 EBs, 19 clusters dropped due to security issues during the fieldwork

Study measures

Child weight status: Weight-for-height z-score (WHZ)

Child weight and height measurements were taken during a single visit. Weight measurements were taken using Seca scales (model no. Seca 878U) and height measurements using a Shorr Board (recumbent if child’s age was less than 24 months and standing if older). Anthropometry data in PDHS-7 was collected by two female enumerators in each field team (total of 44 enumerators) who jointly took measurements. All enumerators were trained beforehand to standardize procedures for anthropometry, including hands-on training to measure ten children twice to assess accuracy and precision of measurements and further training for those enumerators who were out of range more three or more times [24].

The dependent variable was weight-for-height z-score (WHZ) for children under-five per the 2006 Child Growth Standards released by the World Health Organization, reliably shown to describe both overweight (WHZ > 2) and wasting (WHZ < -2) [25]. Every one unit in z-score represents one standard deviation from the mean weight-for-height from a study of more than eight thousand children recruited from Brazil, Ghana, India, Norway, Oman, and the United States of America. For example, a WHZ of -2 indicates that the child’s weight is two standard deviations below the mean for their height and sex [26]. Notably, children contributing data to the development of the growth standards were exclusively or predominantly breastfed for at least the first four months, introduced to complementary foods by six months of age, and partially breastfed for at least 12 months. Consequently, growth standards were derived based on breastfed children [27–29].

Maternal weight status: Body Mass Index (BMI)

The primary exposure of interest for this study was maternal BMI. BMI was measured in the same way as child WHZ, with SECA 878U scales and Shorr Boards (standing height measurement). Maternal BMI was coded as obese (30 or greater), overweight (25 to less than 30), healthy weight (18.5 to less than 25), and underweight (less than 18.5), which was the reference category ([ref]).

Covariates

Several other covariates were also considered in the analysis. A covariate was included in the statistical models if it had evidence of correlation with child weight outcome and if there was sufficient variation in the covariate for it to add information to the model, e.g. if a factor has 99% ‘yes’ response rate and 1% ‘no’ rate, then its use was avoided because it would not meaningfully explain variation in the outcome. Through this process, the following nine covariates were selected for analysis: urban/ rural specific wealth index [20, 25, 30–32], sex of child [25, 30], mother’s employment status [30, 31, 33], type of place of residence [32, 34–36], mother’s youngest child under-five [32, 37, 38], mother’s oldest child under-five [32, 37, 38], child age in months, child breastfeeding status [39–41], and maternal age in years [33, 37, 38].

Household wealth status was measured as quintiles of assets as provided by the PDHS-7 survey. Wealth index quintile categories were labeled as richest, rich, middle, poor, poorest[ref]. Child sex was coded as male or female[ref]. Type of place of residence was coded as urban or rural[ref]. Child breastfeeding status was coded as never breastfed, ever breastfed but not currently breastfeeding, and still breastfeeding[ref]. Mother’s employment status, youngest child under-five and oldest child under-five were all coded as no or yes[ref].

Analysis

A dyadic dataset of mothers and their children was created by merging anthropometric and demographic characteristics of mothers and children per DHS guidance for analysis. Mother’s characteristics were treated as exposures for the child [42]. Furthermore, in studying cardiometabolic risk concordance between younger children and their mothers, it is important to consider the impact of children being born small for gestational age (SGA) can have on anthropometric concordance. This is particularly important consideration in Pakistan, where the prevalence of SGA births is > 45% [43]. As approximately 85% of children born SGA have experienced sufficient catch up growth by age 2 [44–46], stratifying analysis by this age point in relation to concordance which, to our knowledge, has not been done before, could yield important findings regarding the ‘unmasking’ of parent–child concordance. Thus, we stratified children by two years of age in our analysis and modeling.

All analysis and modeling accounted for the complex survey design [42]. All data restructuring and analysis was done using IBM SPSS Statistics v29.0.0.0 with Complex Samples v29. The code used for restructuring and analysis can be found via the following link: 10.5281/zenodo.7794384.

Descriptive statistics

Reported statistics (sampling weight-adjusted) included background characteristics of the participants which include average age of mother (in years) and children (in months), mean BMI of mothers, and mean child WHZ with 95% confidence intervals. Other reported adjusted population descriptive statistics were percentage of dyads that were urban vs rural; percentage of male and female children; percentage of mothers currently employed; percentage of children who were never breastfed, previously breastfed but not currently breastfeeding, or currently breastfeeding; percentage of children being the youngest child; percentage of children being the oldest under-five; and the percentage of children in each wealth index quintile. The mean child WHZ with 95% confidence interval and sampling weight-adjusted sample size was also reported for each factor level.

Modeling associations between maternal and child weight status

We estimated linear regression models to investigate concordance between maternal BMI and child WHZ, measured continuously. Three linear models were generated: 1) all children under-five years, 2) all children under-two years, 3) all children ages 2 to 5, with the models adjusted for the 9 covariates. We reported the R² value and the Wald F for each model and reported the parameter estimates with 95% confidence interval and t-test value for each categorical level. For each model, the statistically significant variables were remodeled including only these variables and an additional term for each two-variable combination to assess for interaction between the statistically significant variables. Cases with missing values for any variables were treated as invalid. Associations for all models were considered statistically significant when p < 0.05.

Further considerations for analysis

We assessed multicollinearity between the independent variables using the variance inflation factor (VIF) calculation. We used a conservative cutoff of 3, as an indicator to further investigate relationships [47]. Using this cutoff, no multicollinearity was observed among the independent variables. To account for multiple comparisons and counter the increased risk of Type I error, the Holm-Bonferroni method was used to adjust the P-value of 0.05 accordingly [48, 49]. Regarding outliers, the DHS data is screened before dissemination for plausibility of Z-scores. For children who had WHZ that were below -5 standard deviations or above + 5 standard deviations were flagged and their WHZ was not reported. Similarly, women who had BMI below 12 or above 60 were also flagged and did not have their BMI reported [42]. Further consideration of outliers was not made because of these adjustments by the PDHS data processing team.

Results

For the analysis, out of 4671 sample unweighted cases, 4130 had complete variable information (with 541 missing) and were included, which led to a sampling-weight adjusted sample size of 3465. With cases that had information on maternal BMI and child WHZ, only two did not have complete information on the other covariates.

Summary statistics

Of the overall sampling weight-adjusted sample of 3465 children (Fig. 1), 1406 (40.6%) children were less than 2 years old, and 2059 (59.4%) children were ages 2 to 5. The mean age of mothers was 29.0 years (95% CI [28.6, 29.4]) and children under-five years was 28.8 months (95% CI [28.0, 29.5]). The mean maternal BMI was 24.6 (95% CI [24.2, 25.1]). The mean child WHZ for the entire sample was -0.30 (95% CI [-0.36, -0.24]), mean child WHZ for children under-two was -0.41 (95% CI [-0.50, -0.31]), and mean child WHZ for children ages 2 to 5 was -0.23 (95% CI [-0.29, -0.16]). The greatest mean child WHZ was in children who had never been breastfed at 0.07 (95% CI [-0.22, 0.36]), and the category with the lowest mean child WHZ was children with mothers who were underweight at -0.66 (95% CI [-0.80, -0.51]). Table 1 summarizes maternal, child, and household characteristics of the analyzed maternal-child dyads and the mean child WHZ for each factor level is summarized in Table 2.

Table 1.

Percentage of sample in each factor level and mean values of continuous variables (n = 3465), with 95% confidence interval from PDHS-7 dataset used in analysis

Variable	All	Children under 2	Children 2 to 5 years
Maternal Characteristics
		Percent (95% CI)
Maternal BMI Category
30 +	15.7 (13.1 – 18.8)	13.1 (10.4 – 16.4)	17.5 (14.5 – 20.9)
25 to < 30	27.6 (24.8 – 30.6)	27.9 (24.5 – 31.7)	27.4 (24.3 – 30.7)
18.5 to < 25	46.2 (42.9 – 49.6)	46.7 (42.6 – 50.8)	45.9 (42.1 – 49.8)
< 18.5	10.4 (8.5 – 12.7)	12.3 (9.9 – 15.1)	9.2 (7.2 – 11.7)
Mother currently employed
No	86.7 (83.9 – 89.0)	34.6 (30.1 – 39.4)	31.4 (27.6 – 35.5)
Yes	13.3 (11.0 – 16.1)	65.4 (60.6 – 69.9)	68.6 (64.5 – 72.4)
		Mean (95% CI)
Maternal BMI	24.6 (24.2, 25.1)	24.3 (23.8, 24.8)	24.9 (24.4, 25.4)
Maternal Age (yrs.)	29.0 (28.6, 29.4)	27.4 (26.9, 27.9)	30.3 (29.6, 30.5)
Child Characteristics
		Percent (95% CI)
Youngest child
No	34.3 (32.8 – 35.9)	5.5 (4.2 – 7.2)	54.0 (51.3 – 56.7)
Yes	65.7 (64.1 – 67.2)	94.5 (92.8 – 95.8)	46.0 (43.3 – 48.7)
Oldest child under-five
No	32.5 (30.9 – 34.1)	56.4 (52.9 – 59.8)	16.2 (14.3 – 18.2)
Yes	67.5 (65.9 – 69.1)	43.6 (40.2 – 47.1)	83.8 (81.8 – 85.7)
Sex of child
Male	51.1 (49.1 – 53.1)	51.2 (47.3 – 55.0)	51.0 (48.7 – 53.3)
Female	48.9 (46.9 – 50.9)	48.8 (45.0 – 52.7)	49.0 (46.7 – 51.3)
Breastfeeding status
Never breastfed	4.4 (3.3 – 5.8)	4.1 (2.9 – 5.8)	4.6 (3.2 – 6.4)
Previously breastfed but not currently breastfeeding	63.4 (61.3 – 65.5)	24.9 (22.1 – 28.0)	90.1 (88.0 – 91.9)
Currently breastfeeding	32.2 (30.2 – 34.2)	70.9 (67.7 – 74.0)	5.3 (4.2 – 6.7)
		Mean (95% CI)
Child WHZ	-0.30 (-0.36, -0.24)	-0.41 (-0.50, -0.31)	-0.23 (-0.29, -0.16)
Child Age (mos.)	28.8 (28.0, 29.5)	11.0 (10.5, 11.5)	40.9 (40.4, 41.5)
Household Characteristics
		Percent (95% CI)
Area
Urban	32.7 (29.1 – 36.6)	34.6 (30.1 – 39.4)	31.4 (27.6 – 35.5)
Rural	67.3 (63.4 – 70.9)	65.4 (60.6 – 69.9)	68.6 (65.4 – 72.4)
Wealth index (urban/ rural specific)
Richest	18.3 (14.7 – 22.5)	19.1 (14.7 – 24.5)	17.7 (14.1 – 21.9)
Richer	20.3 (17.0 – 24.0)	21.9 (18.0 – 26.3)	19.1 (15.7 – 23.1)
Middle	20.1 (17.4 – 23.0)	20.2 (17.0 – 23.9)	19.9 (17.0 – 23.2)
Poor	21.4 (18.4 – 24.8)	20.2 (16.8 – 24.1)	22.3 (18.8 – 26.1)
Poorest	20.0 (16.5 – 24.0)	18.6 (15.1 – 22.5)	21.0 (16.8 – 25.8)

Table 2.

Mean Child WHZ and sampling weight-adjusted sample size (Adj. N) of analyzed sample, stratified by level of selected factors

Variable	Adj. N	Mean child WHZ (95% CI)
All	3465	-0.30 (-0.36, -0.24)
Maternal BMI Category
30 +	545	-0.02 (-0.16, 0.11)
25 to < 30	956	-0.14 (-0.25, -0.03)
18.5 to < 25	1602	-0.41 (-0.51, -0.31)
< 18.5	361	-0.66 (-0.80, -0.51)
Mother currently employed
No	3003	-0.29 (-0.36, -0.23)
Yes	462	-0.35 (-0.52, -0.19)
Youngest child
No	1189	-0.21 (-0.29, -0.13)
Yes	2276	-0.35 (-0.42, -0.28)
Oldest child under-five
No	1126	-0.40 (-0.51, -0.28)
Yes	2339	-0.26 (-0.32, -0.20)
Sex of child
Male	1770	-0.32 (-0.39, -0.24)
Female	1695	-0.29 (-0.36, -0.21)
Area
Urban	1134	-0.31 (-0.40, -0.21)
Rural	2331	-0.30 (-0.37, -0.22)
Breastfeeding status
Never breastfed	150	0.07 (-0.22, 0.36)
Previously breastfed but not currently breastfeeding	2201	-0.25 (-0.31, -0.18)
Currently breastfeeding	1111	-0.46 (-0.56, -0.36)
Wealth index (urban/ rural specific)
Richest	633	0.04 (-0.09, 0.17)
Richer	702	-0.17 (-0.29, -0.06)
Middle	695	-0.36 (-0.50, -0.22)
Poor	743	-0.52 (-0.64, -0.40)
Poorest	693	-0.45 (-0.58, -0.31)

Associations between mother and child weight status

Maternal BMI was positively associated with child WHZ for all children among children under-five. Children of normal weight, overweight, and obese women had WHZ scores that were 0.21 [95% CI: 0.04, 0.37], 0.43 [95% CI: 0.25, 0.62], and 0.51 [95% CI: 0.30, 0.71] units higher than children of underweight women, respectively. Maternal BMI was also positively associated with child WHZ among children ages 2 to 5, where children of normal weight, overweight, and obese women had WHZ scores that were 0.26 [95% CI: 0.08, 0.44), 0.50 [95% CI: 0.30, 0.71), and 0.61 [95% CI: 0.37, 0.84] units higher than children of underweight women, respectively.

For all children under-five, household wealth index was positively associated with child WHZ (Richest: 0.34 [95% CI: 0.16, 0.51]; Rich: 0.18 [95% CI: 0.00, 0.35]; Middle: 0.00 [95% CI: -0.18, 0.18]; Poor: -0.13 [95% CI: -0.31, 0.05]; Poorest [ref.]). Household wealth was similarly positively associated with child WHZ for children under 2 (Richest: 0.42 [95% CI: 0.12, 0.72]; Rich: 0.20 [95% CI: -0.09, 0.49]; Middle -0.32 [95% CI: -0.59, -0.05]; Poor: -0.15 [95% CI: -0.46, 0.17]; Poorest: [ref]) and for children aged two to five (Richest: 0.29 [95% CI: 0.09, 0.49]; Rich: 0.16 [95% CI: -0.03, 0.34]; Middle 0.22 [95% CI: 0.02, 0.42]; Poor -0.11 [95% CI: -0.28, 0.06]; Poorest: [ref.]).

There were no statistically significant interaction effects for any linear model. Table 3 summarizes information for each model.

Table 3.

Linear regression results of the association between maternal BMI and child WHZ, with selected covariates

Predictor	Analysis I – all children under-five			Analysis II—children under-two			Analysis III—children 2 to 5 years
(Category)	Coefficient (95% CI)	t-value	P	Coefficient (95% CI)	t-value	P	Coefficient (95% CI)	t-value	P
Mother BMI category			(< 0.001)			0.181			(< 0.001)
30 or greater	0.51(0.30, 0.71)	4.930	< .001	0.29(-0.06, 0.64)	1.625	0.105	0.61(0.37, 0.84)	5.018	< .001
25 to < 30	0.43(0.25, 0.62)	4.613	< .001	0.25(-0.06, 0.56)	1.597	0.111	0.50(0.30, 0.71)	4.826	< .001
18.5 to < 25	0.21(0.04, 0.37)	2.502	0.13	0.06(-0.22, 0.33)	0.402	0.688	0.26(0.08, 0.44)	2.873	0.004
less than 18.5	*			*			*
Mother currently working			0.369			0.845			0.183
No	-0.08(-0.25, 0.09)	-0.899	0.369	0.03(-0.22, 0.27)	0.196	0.845	-0.13(-0.32, 0.06)	-1.333	0.183
Yes	*			*			*
Youngest child			0.062			0.021			0.153
No	0.12(-0.01, 0.24)	1.874	0.062	0.37(0.06, 0.68)	2.324	0.021	0.10(-0.04, 0.24)	1.431	0.153
Yes	*			*			*
Oldest child under-five			0.570			0.637			0.678
No	-0.04(-0.18, 0.10)	-0.569	0.57	-0.05(-0.24, 0.15)	-0.473	0.637	-0.04(-0.24, 0.16)	-0.416	0.678
Yes	*			*			*
Sex of child			0.403			0.163			0.925
Male	-0.04(-0.14, 0.06)	-0.838	0.403	-0.13(-0.31, 0.05)	-1.398	0.163	0.01(-0.10, 0.11)	0.094	0.925
Female	*			*			*
Area			0.447			0.827			0.422
Urban	-0.04(-0.16, 0.07)	-0.761	0.447	-0.02(-0.20, 0.16)	-0.219	0.827	-0.05(-0.19, 0.08)	-0.804	0.422
Rural	*
Breastfeeding status			0.063			0.477			0.237
Never breastfed	0.35(0.04, 0.65)	2.255	0.025	0.3(-0.21, 0.81)	1.149	0.251	0.30(-0.17, 0.78)	1.254	0.211
Previously breastfed, not currently breastfeeding	0.15(-0.02, 0.32)	1.704	0.089	-0.02(-0.25, 0.20)	-0.188	0.851	0.09(-0.31, 0.49)	0.424	0.672
Still breastfeeding	*			*			*
Wealth index (urban/ rural specific)			(< 0.001)			(< 0.001)			(< 0.001)
Richest	0.34(0.16, 0.51)	3.702	< .001	0.42(0.12, 0.72)	2.771	0.006	0.29(0.09, 0.49)	2.852	0.005
Rich	0.18(0.00, 0.35)	1.986	0.048	0.20(-0.09, 0.49)	1.387	0.166	0.16(-0.03, 0.34)	1.679	0.094
Middle	0.00(-0.18, 0.18)	0.015	0.988	-0.32(-0.59, -0.05)	-2.339	0.02	0.22(0.02, 0.42)	2.124	0.034
Poor	-0.13(-0.31, 0.05)	-1.432	0.153	-0.15(-0.46, 0.17)	-0.925	0.355	-0.11(-0.28, 0.06)	-1.242	0.215
Poorest	*			*			*
Mother age (yrs)	0.00(-0.01, 0.01)	-0.159	0.874	0.00(-0.02, 0.02)	0.171	0.865	0.00(-0.01, 0.01)	-0.616	0.538
Child age (mos)	0.00(-0.01, 0.00)	-0.549	0.583	0.01(-0.01, 0.02)	1.269	0.205	-0.01(-0.02, 0.00)	-2.428	0.016
R^2	0.053			0.062			0.066
Wald F	7.182			3.345			7.43

Overall variable p-values in parenthesis denote statistically significant associations (Holm-Bonferroni corrected)

^*denotes reference category

Discussion

We found that maternal BMI was positively associated with child WHZ in children under 5. This positive association was present for children ages 2 to 5, even after accounting for several socio-demographic factors, but not for children under-two. The findings suggest that there is intergenerational concordance in weight status in Pakistani mothers and young children which is unmasked with increased age.

The findings of parent–child concordance of BMI-WHZ for all children under-five are consistent with previous literature which specifically explore parent–child concordance of NCD risk [2, 12–18] and of association studies in South Asia which have observed a positive association between maternal and children under-five weight status [19–21]. However, previously described phenomenon of the double burden of malnutrition at the household level (DBMHL) suggests co-existence of undernourished children and overweight mothers in the same household, but this phenomenon is primarily described for child stunting and, in some cases, underweight [50–58]. Although a recent study by Biswas et al. [59], in which they analyzed the co-existence of overweight mothers with all types of undernourished children under-five (stunted, wasted, and underweight) in South and Southeast Asia from 2006 to 2017 data, described positive relationships between maternal weight status and all three types of child undernutrition, with the greatest co-existence in Pakistan. However, this study did not analyze the relationship between non-overweight or obese mothers and their children, nor did it split the children under-five into two age groups [59]. Another study by Hossain et al. showed a positive association between maternal overweight status and children ages 24 to 59 months overweight status, but this study did not include children from 0 to 24 months due to most classification systems defining child overweight status after 24 months [21]. In our analysis, using WHZ for the child’s weight status provided more flexibility in analyzing a wider age group.

Furthermore, double burden of malnutrition (DBM) at the country or population level is well-established in previous literature [60–63]. DBM prevalence has been increasing in Pakistan as shown by PNNS 2018 results, with wasting noted in 17.7% of children, the highest in Pakistan’s history, prevalence of overweight children under-five almost doubling from 5% in 2011 to 9.5% in 2018, and women ages 15–49 experiencing an increase in overweight and obesity increasing from 28% to 37.8% from 2011 to 2018. Both, DBMHL and DBM, can co-exist with maternal-child BMI-WHZ concordance, as the increasing burden of DBM, with an increasing percentage of overweight children, could be leading to the more recent findings of positives association of maternal-child weight status. Co-existence of these phenomena within the same society highlight the complexity of interventions and approaches required to address malnutrition among LMICs, as different regions and households are experiencing various challenges.

To our knowledge, there are no studies which specifically investigate the relationship between maternal and child concordance under-five stratified by 2 years of age, as we analyzed in this study. These findings, which show a positive association between maternal BMI and child WHZ only in the 2 to 5 age group, potentially indicate that concordance between child WHZ and maternal BMI is ‘unmasked’ after age 2, when the majority of early childhood catch-up growth has occurred [44, 45]. As in, children who may be born SGA may not show concordance of WHZ with maternal BMI during the first two years of life, but as they experience their expected catch up growth and have spent more time in a shared environment with their mothers, their weight status starts to become concordant with their mothers, as seen by our analysis which shows the difference in association between children older than and younger than two. This suggests that children under-two who do not initially appear to have WHZ concordant to their mother’s BMI may develop this positive relationship later in life.

Furthermore, a targeted review on maternal and child overweight and obesity in LMICs shows that junk food, such as potato chips, sponge cakes, sugary biscuits, and sugary drinks are a significant part of child diet by the first two years of life [64], potentially yielding an avenue of intervention and an explanation for the development of mother–child concordance after two years of age. This relationship could also be affected by other dietary aspects such as breastfeeding status and weaning, although in our analysis, breastfeeding history did not show a statistically significant association with weight status. However, it is important to note that the population surveyed showed little variability in breastfeeding, with only 4.4% of the population having been never breastfeed. Furthermore, limitations of the survey prevented us from investigating duration and exclusivity of breastfeeding, which can play a role in the strength of association between breastfeeding and weight status. Prior studies have shown that longer duration of breastfeeding decreases excess weight gain in children [39], and that exclusively breastfed children through 6 months tend to be more likely to grow within the normal limits set by the 2006 WHO Child Growth Standards while also being protective against overweight, obesity, and type 2 diabetes through childhood and adulthood [40, 41, 65]. Overall, this may indicate an avenue for early interventions for weight-related risks in children, where healthier nutritional sources during infancy may lead to a decreased sharing of NCD risk found between very young children and their mothers as children age. This approach is well established within public health, focusing on building public policy that focuses on health in domains beyond simply clinical settings, such as the commitment to health promotion and disease prevention societal domains laid out by experts within declarations such as the Ottawa Charter for Health Promotion and the 2030 Agenda for Sustainable Development Goals (SDGs) [66–68].

Within Pakistan specifically, although there have been attempts to establish a national policy to implement health promotion strategies for prevention and reduction of NCDs, there is room for improvement and further work. For example, the National Action Plan for Non-Communicable Disease Prevention, Control and Health Promotion was introduced in 2003 as a means to develop a robust NCD surveillance system, with a focus on policy measures and communications focusing on diet, activity, and accessibility of healthier foods [69]; however, due to government changes, the policies and plans were not implemented. Similarly, in 2009, a public–private partnership dubbed the National Commission for Prevention of NCDs was proposed by the Pakistan Ministry of Health, but implementation was halted due to legal concerns [70]. Consequentially, despite calls from experts in Pakistan [71], there remains little in terms of surveillance systems in Pakistan to track the progress of NCDs, the country’s progress in achieving NCD-related SDGs, or even understanding the role of household/ family intervention to mitigate risks of a rapidly growing burden [72, 73].

Regarding other covariates which were analyzed, we showed wealth index as a statistically significant positive factor across all regression analyses. This association existing in Pakistan is well established in the literature [74–79]. Increasing wealth quintile has been suggested to correlate with access to nutrient rich foods and increased sedentary behaviors [20, 80–83].

Considering strengths of this study, the analysis was done on a nationally representative dataset which involved a complex sampling process to obtain the best possible demographic snapshot of Pakistan. The analysis also included a significant number of covariates to unveil potential confounders and used rigid levels of correction and validation regarding significance and multi-collinearity. Limitations of the study included the use of cross-sectional data, whereas a cohort study could provide more in-depth data on trends in specific households.

Conclusions

We found that maternal-child concordance of weight status among children ages 2 to 5, but not among children under 2. Future work should involve more detailed data collection of health behaviors and environmental factors relevant to obesity and cardiometabolic disease early in life within the household. Such information may elucidate pathways for NCD risk concordance between mothers and children, and possibly all family members, to develop evidence or family-wide health promotion interventions for healthy weight across the life course.

Abbreviations

adj

Adjusted

BMI

Body Mass Index

CI

Confidence interval

DBM

Double burden of malnutrition

DBMHL

Double burden of malnutrition at the household level

LMICs

Low- and middle- income countries

mos.

Months

NCD

Noncommunicable disease

PDHS-7

Pakistan’s Demographic and Health Survey, Round VII

PNNS

Pakistan National Nutrition Survey

SGA

Small for Gestational Age

USAID

United States Agency for International Development

ref.

Reference

VIF

Variance Inflation Factor

WHO

World Health Organization

WHZ

Weight-for-height z-score

yrs

Years

Authors’ contributions

The concept was drafted by F.A. and R.I. F.A. and S.A.P. contributed to the analysis plan. M.K.A., R.I., and S.A.P. advised on paper conceptualization. F.A. was the primary writer of the article with comprehensive feedback from M.K.A., S.A.P., and R.I. All authors read and approved the final manuscript.

Funding

No grants or other funding sources were provided for the analysis and writing of this manuscript. Regarding survey design and data collection, financial support was provided by the United States Agency for International Development.

Availability of data and materials

The datasets generated and/or analyzed during the current study are available with DHS program. The data can be requested at: https://dhsprogram.com/data/available-datasets.cfm

Declarations

Ethics approval and consent to participate

The PDHS-7 survey was reviewed and approved by the National Bioethics Committee of the Pakistan Health Research Council and the Inner City Fund/ ICF Institutional Review Board. Adult participants were consented both, for themselves and for their children (24). Prior to obtaining access to and performing analysis on the PDHS-7 data, the researchers of the present study submitted a concept note, gained data access, and obtained permission from USAID. The data were de-identified and anonymized by the DHS program prior to being shared, therefore not considered human subjects research by an author-affiliated institution (Emory University, Atlanta, GA, USA) and not requiring further ethical clearance.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.