Grandparental and Parental Obesity Influences on Childhood Overweight: Implications for Primary Care Practice

Study Sample

To examine patterns of BMI among children, their parents, and their grandparents, we used data from the Panel Study of Income Dynamics (PSID), a 35-year panel survey of a nationally representative sample of nearly 8000 US families in 2003, and the Child Development Supplement (CDS), a nationally representative study of children of PSID families.¹⁰ The PSID has a genealogical design whereby descendants of sample members are themselves interviewed when they begin to live independently. As a result, grandparents of the CDS children (ie, members of the original PSID cohort) are interviewed even when they do not live in the same household as the grandchildren who are part of the CDS cohort.

Data Collection

Self-reported weight and height information was provided by parents and grandparents in 2003. For children, weight and height information was collected using in-home assessments: a tape measure was used to assess height and a scale was used to measure weight in 2002/2003, when the CDS children were aged 5 to 19 years. Height and weight data were then used to calculate BMI for each respondent, using height and weight data converted to metric measures by the investigators for the BMI calculation.

For children, age- and sex-specific growth curves provided by the Centers for Disease Control and Prevention (CDC)¹¹ were used to define BMI categories. By convention, BMI ≥95th percentile was defined as overweight, BMI ≥85th percentile but <95th percentile was defined as at risk for overweight, <5th percentile was defined as underweight, and otherwise BMI was normal weight. Similarly by CDC convention, for adults BMI ≥30 was defined as obese, BMI 25 to 29.9 was defined as overweight, BMI <18.5 was defined as underweight, and otherwise BMI was normal weight.

Data Analysis

To account for the complex stratified sampling of the PSID and permit inference to the national population, all analyses were sample weighted. Only weighted data are presented, except where otherwise noted.

To compare the PSID sample with the broadly accepted standard national measures of BMI from the National Health and Nutrition Examination Survey (NHANES), we assessed the frequency distribution of children across BMI categories in the PSID measured in 2002/2003 alongside the distribution from NHANES data measured in 1999 to 2002.²

To examine the association of child BMI with parental and grandparental BMI, we generated composite measures of adult BMI as follows. For parental BMI, we set the BMI category (underweight, normal weight, overweight, obese) based on mother's BMI and, if missing, used father's BMI. For grandparental BMI, we set the BMI category based on maternal grandmother's BMI and, if missing, used (in order) maternal grandfather's BMI, paternal grandmother's BMI, and paternal grandfather's BMI. This selection rubric was based on analyses that suggest maternal obesity is more strongly determinant of obesity patterns than paternal obesity.^6,7

With these categories, we used linear regression models to estimate the proportion of children who were overweight in families with different combinations of parental and grandparental BMI categories. We used dummy variables for the different parental and grandparental groups, including missing. We included the association with missing data because it may correspond to clinical situations in which grandparental data are not available because grandparents are deceased or when grandparental weight status is otherwise unknown. We constructed similar models for all children who were either at risk for overweight or overweight (ie, BMI ≥85th percentile for sex-specific BMI). We do not report associations with parental and grandparental underweight because of possible correlation with adults’ chronic illness that could not be confirmed in the dataset and because of small sample size for this group that makes estimates unstable.

This study was approved by the University of Michigan Institutional Review Board. Three of the coauthors (KM, RFS, FS) coordinated data collection for the PSID and CDS during the study period.

Study Sample

The sample included 2591 children aged 5 to 19 years. Overall, the prevalence of different BMI categories in the 2002/2003 PSID was quite similar to BMI prevalence data from the 1999 to 2002 NHANES (Table 1). Although rates of overweight seemed slightly higher in the PSID than in the NHANES, this is consistent with the fact that the prevalence of overweight has increased rapidly among children in recent years and the PSID data were measured at a later date.^1,12

We had parental BMI data for 2437 children (94.1% of sample) and grandparental BMI data for 1573 children (60.7% of sample). The distribution of children in the sample by their parental and grandparental weight status (including missing) appears in Table 2. Among children with known parental and grandparental BMI information, it was most common for children to have normal-weight parents and normal-weight grandparents (12.83% overall), followed by normal-weight parents and overweight grandparents, overweight parents and overweight grandparents, and overweight parents and normal-weight grandparents.

Risk of Childhood Obesity with Parental and Grandparental Obesity

Among children with known parental and grandparental BMI, the proportion of children overweight for age and sex was 7.9% among individuals with normal-weight parents and normal-weight grandparents (Table 3, top left-hand corner). This proportion of children who were overweight was less than half of the proportion who were overweight in the sample (18.6%; Table 1).

Parental and grandparental overweight and obese status had independent associations with childhood overweight. Among children with normal-weight grandparents (Table 3, left-hand column), parental overweight and obese status had a stepwise, statistically significant positive association with children's overweight prevalence. Similarly, among children with normal-weight parents (Table 3, top row), grandparental obesity was associated with a significantly increased prevalence of children's overweight, compared with families with normal-weight grandparents.

Furthermore, for children with overweight grandparents and with obese grandparents, the presence of parental obesity (but not overweight) was associated with significantly increased prevalence of children's overweight compared with children with normal-weight parents (Table 3). Among children with obese parents, regardless of grandparental weight status, overweight prevalence was almost twice the prevalence rate in the overall sample.

For children whose grandparental BMI data were missing, their probability of overweight corresponded strongly with their parental BMI group. The exception was for children whose parental BMI was normal, but if their grandparental BMI was missing they were significantly more likely to be overweight than if the grandparental BMI was measured as normal (P < .0001).

Comment

This is the first national study to confirm the association of children's overweight with parental obesity observed in smaller samples^6–8 and is also the first analysis of a national sample to report an association of children's overweight with grandparental weight status. The positive associations of children's overweight with parental and grandparental obesity in this national sample underscore the continuing clinical and public health importance of familial patterns of obesity in an era when environmental explanations for obesity are receiving considerable research and policy attention.^3–5,13

Our findings also highlight for family physicians and other primary care practitioners that asking about parental and grandparental patterns of weight can inform the risk of overweight for children. This is important because the majority of children are of normal weight, and it can be difficult for health care providers to anticipate which children are at greatest risk of progressing to overweight later in childhood. These data, although cross-sectional in nature, indicate the relative proportions of a broad age-group of children who may be overweight at some point in childhood, based on their parents’ and grandparents’ weight characteristics. In the past, many primary care providers may have informally assessed this likelihood of overweight for children in their practice based on parental weight alone. The added insight from these findings is that, for children with normal-weight parents, grandparental obesity is a risk factor for childhood overweight.

Familial effects probably have dual origins in genetic and behavioral mechanisms common to parents and their children. More than 500 genes have been putatively linked to obesity,¹⁴ and behavior related to dietary preferences and to patterns of physical and sedentary activity are established early in individuals’ lifetimes.¹³ Many experts have argued that prevention of obesity during childhood is essential to success of efforts to combat the national obesity epidemic.^13,15–17 Yet, clinicians and policymakers today have few evidence-based effective tools for obesity prevention among children.^13,18 The strong and independent associations of parental and grandparental obesity with children's overweight in this study lend further imperative to the development of innovative intervention strategies that target young children in their families.

The association of grandparental obesity with childhood overweight, especially in families with normal parental weight, raises additional questions about the intrafamilial mechanisms of obesity. Importantly, although the PSID/CDS includes information about grandparents co-residing with their grandchildren at points in time, it does not include information about duration of co-residence or the grandparents’ roles in child care that we believe would be important for an analysis of obesogenic environments. In addition, we were not able to adjust for children's exposure to their grandparents’ dietary or physical activity habits, which may have directly influenced children's behaviors. In addition, it is possible that grandparents’ behaviors related to their own obesity may have influenced their grandchildren through the children's parents, even if those parents were themselves normal weight. It is also plausible that genetic predisposition to obesity may be variably expressed across generations. More research is warranted to explore these multigenerational effects, especially in clinical situations where children with normal-weight parents may be at increased risk for overweight in the setting of grandparental obesity.

Our findings must be interpreted with important caveats. A chief limitation of this dataset is that adult BMI status was calculated from respondents’ self-reported height and weight. Estimates of the prevalence of obesity based on self-report may be affected by bias toward lower BMI from underreporting of weight influenced by social desirability.^19,20 Of note, self-report adult BMI data from the PSID closely match other self-report data available from concurrent waves of the National Health Interview Survey.²¹ Another limitation is that we assessed intrafamilial BMI associations at one point in time; future work will permit assessments of these relationships over longer periods.

Last, we had BMI data for grandparents for approximately 60% of the children's sample, which largely reflects survival status of the grandparents rather than item nonresponse on the height and weight questions. Importantly, children with normal-weight parents were at increased risk if their grandparental BMI data were missing, raising the possibility that deceased grandparents had been obese and perhaps died prematurely related to health consequences of obesity such as coronary heart disease and diabetes. Information that clinicians can collect from families regarding deceased grandparents’ weight status may well be more complete and informative than data available for this analysis.