Offspring-sex modifies the association between early-pregnancy adiposity and 2-year-old total physical activity – The Glowing Study

Background: Rodent models suggest that in utero exposure to under and overnutrition programs offspring physical activity (PA) behaviors. Such nexus has not been established in humans. This study evaluated the association of early pregnancy maternal adiposity with offspring PA at age 2 years (2-yo-PA) taking into consideration prenatal and postnatal factors. Methods: Women (n=153) were enrolled early in pregnancy (<10 weeks). At enrollment, maternal adiposity [air displacement plethysmography, fat mass index (FMI, kg/m2)] and PA (accelerometers, activity counts) were measured, and age, race, and education self-reported. Gestational weight gain was measured at the research facility. Offspring birthweight and sex were self-reported. At age 2 years, parental feeding practices (child feeding questionnaire) were assessed, whereas anthropometrics (length and weight) and physical activity (accelerometers) were objectively measured. Offspring body mass index z-scores were calculated. Generalized linear regression analysis modeled the association of maternal FMI and 2-yo-PA [average activity counts (AC)4/day]. Results: There was an interaction between maternal FMI and offspring sex in association with 2-yo-PA (β= −1.03, p= 0.030). Specifically, 2-yo-PA was lower in girls compared to boys when maternal FMI was ≥7 kg/m2. Maternal PA early in pregnancy positively associated with 2-yo-PA (β= 0.21, p= 0.005). In addition, children born to women with college education tended to be more active compared to children born to women without college education (β= 3.46, p= 0.059). Conclusions: Sexual dimorphism was observed in the associations of maternal adiposity with 2-yo-PA, with girls being less active compared to boys only when maternal FMI was ≥7 kg/m2.


INTRODUCTION
The discovery of the developmental origins of health and disease, which emerged from studies of famine-exposed communities at the turn of the twentieth century (1,2), has critical relevance in contemporary societies. In the United States, at least 50% of children are born every year to women with overweight [Body Mass Index (BMI) 25.0-29.9 kg/m 2 ] or obesity (BMI ≥ 30 kg/m 2 ) (3). Evidence supports that in utero exposure to maternal overweight/obesity increases cardiometabolic risk in the offspring, starting as early as birth and infancy. Maternal adiposity early in pregnancy, for example, is directly related to offspring adiposity at birth (4) and adiposity accrual in girls over the rst two years of life (5). Indeed, by age two years, the obesity risk in children born to mothers with obesity is twice that of children born to women with normal weight (6). What is unknown is whether maternal obesity per se programs progeny behaviors pertinent to energy balance, such as physical activity behavior, thereby in uencing children's obesity risks..
A study in rodents reported that prenatal exposure to undernutrition (i.e., 50% of habitual calorie intake) and overnutrition (i.e., a high-fat [23%] diet) modi ed the disposition of progeny to engage in physical activity later in life (7). Speci cally, compared to controls, exposed offspring exhibited markedly reduced degree of wheel-running activity. Similarly, using a mouse model, Johnson et al. (8) evaluated the effects of a high-fat diet during pregnancy on offspring physical activity. The authors reported offspring showed reduced voluntary physical activity when in the home-cage setting. In addition, compared to males, female offspring were more likely to sleep during dark cycles which are regarded as the time when rodents reach peak activity. These ndings suggest that prenatal overnutrition may in uence offspring's attitudes and behaviors related to physical activity in a sex-dependent manner.
In humans, only one study has assessed the effects of gestational weight gain (GWG) on objectively measured physical activity in children (n = 113) (9). Physical activity was measured with accelerometers, and information on maternal pre-pregnancy weight status and GWG collected using recall questionnaires.
At age 4 years, average activity counts per day decreased with increasing GWG in boys. In girls, the association between physical activity and GWG was modi ed by maternal overweight/obesity (n = 22) and followed an inverted U shape. These ndings are consistent with animal models showing in utero overnutrition may program offspring's physical activity patterns. The study is limited, however, by the small number of women with overweight and obesity as well as its cross-sectional design, which relied on maternal data recall.
Methodological di culties likely account for the paucity of studies in humans assessing whether maternal overweight/obesity programs offspring physical activity behaviors. Separating the effects of prenatal and postnatal in uences requires rigorous and costly studies with repeated measures over extended time periods. To address these research limitations, we conducted objective measurements of physical activity by accelerometry in 2-year-old participants of the longitudinal observational Growing Life, Optimizing Wellness study (GLOWING, NCT01131117). The GLOWING study's primary goal is to assess the impact of maternal health before and throughout pregnancy on offspring obesity risk. On the basis of the limited evidence from murine models and human studies, we hypothesized that total physical activity (i.e., average daily activity counts) in 2-year-old offspring is inversely and sexdependently associated with early-pregnancy maternal adiposity. Speci cally, we hypothesized that physical activity declines with increasing maternal adiposity in 2-year-old boys and girls, but girls are signi cantly less active than boys.

Subjects
Mother -offspring pairs (n = 153) were participants enrolled in the GLOWING study between 2011 and 2014 at the Arkansas Children's Nutrition Center. The GLOWING study is an ongoing longitudinal observational study evaluating the impact of maternal health prior to and during pregnancy on offspring growth and obesity risk. Inclusion criteria were: BMI of 18.5-35 kg/m 2 at enrollment, second parity, singleton pregnancy, ≥ 21 years old, and conception without assisted fertility treatments. Only offspring who were born full term, healthy and had no ongoing medical conditions at 2 years of age were eligible to participate in the present analysis. Exclusion criteria at enrollment and/or during pregnancy were: maternal preexisting or ongoing medical conditions including gestational diabetes, complications during pregnancy, medications known to in uence fetal growth, maternal active smoking, alcohol consumption in any amount, and being an athlete (de ned as being engaged in a professional sports activity). The institutional review board at the University of Arkansas for Medical Sciences approved the study protocol. All parents gave written informed consent.

Measurements
Maternal measurements during pregnancy Anthropometrics and gestational weight gain Maternal height was measured to the nearest 0.1 cm at enrollment using a standard wall-mounted stadiometer (Tanita Corp., Tokyo, Japan). Body weight was measured at enrollment, 12-week gestation, and every 6 weeks thereafter using a tared scale (Perspective Enterprises, Portage, MI, USA) to the nearest 0.1 kg. Gestational weight gain was computed from the rst measured weight to gestation week 36.
Adherence to the Institute of Medicine (IOM) gestational weight gain guidelines was evaluated by adjusting the guidelines to re ect the last measure at gestation week 36 (10).
Maternal physical activity early in pregnancy Average daily step counts and total activity counts were measured with the Actical accelerometer (Philips Respironics Co. Inc., Bend, Oregon, USA) at enrollment. The Actical is a small (2.8 x 2.7 x 1.0 cm; weight: 17g) accelerometer measuring omnidirectional gross motor activity. The device was placed on the participants' ankles on the non-dominant side. Each participant wore the Actical for 2-5 days (2 days: n = 3; 3 days: n = 10; 4 days: n = 82; 5 days: n = 58). The monitor was programmed to record movement activity beginning at 11:59 PM on a given day. Participants were instructed to maintain usual activities while wearing the monitor. Total activity counts per day early in pregnancy were obtained using the visual identi cation method (refer to the 'Offspring physical activity at age 2 years' section for more details).

Demographic characteristics
Mothers self-reported race (i.e., American Indian or Alaska Native, Asian, Black or African American, Hispanic or Latino, Native Hawaiian or other Paci c Islander, or White), age, and date of last menstrual period. At the post-natal two-week research study visit, participants reported infant's birth weight, infant's race, and sex. Weight for gestational age at birth z-scores were computed from the maternal reported infant's birth weight using the International Fetal and Newborn Growth Consortium for the 21st Century (INTERGROWTH -21) standards (11).
Parental feeding practices at age 2 years Parental feeding practices were measured using the Child Feeding Questionnaire (CFQ) (12). This scale targets parents of children ages 2 to 11 years and is comprised of 7 subscales. Four of them evaluate parental beliefs toward their child's obesity proneness, and 3 evaluate parental control attitudes and practices regarding child feeding. The latter 3 were used in this study and include: restriction (limit access to certain foods), monitoring (keep track of what the child eats), and pressure to eat (pressure the child to eat more food, particularly during meals).
Offspring measurements at age 2 years Anthropometric measurements at age 2 years Offspring weight was measured to the nearest 0.01 kg using tared scale (SECA 727, SECA, Ontario, CA) and length was measured to the nearest 0.1 cm by using a length board (Easy Glide Bearing Infantometer, Perspective Enterprises, Portage, MI). BMI z-scores (BMI-Z) were computed based on the World Health Organization (WHO) Child Growth Standards for children ages 0 to 5 years (13).

Physical activity at age 2 years
Children wore Actical accelerometer at the ankle for 2 to 9 days (2 days: n = 4; 3 days: n = 7; 4 days: n = 5; ≥5 days: n = 137). The Actical data were downloaded using the Actical software. Data were summarised as 60-second epochs. Data in 60 second epochs were processed using a custom semi-automated algorithm to identify waking wear time (14). We adapted this algorithm (originally developed for uniaxial or triaxial accelerometer worn on the hip) with a re nement sample (n = 10), where rules were trialled and evaluated graphically, relative to visual identi cation. Several thresholds were trialled, and then the one informally judged to be the best performing (considering not only error amount, but the type of failures and how often they were likely to occur) was selected. Brie y, the algorithm attempts to identify in-bed and non-wear periods by searching for prolonged periods (≥ 180 minutes) of low activity (rolling averages of 30 minutes either side of each minute < 50cpm). The algorithm then searches for sustained periods of higher intensity activity (rolling average 30 minutes either side of ≥ 50 counts per minute for ≥ 10 minutes). A single non-wear rule was applied to all of the data. All minutes in continuous periods of ≥ 120 minutes of zero counts per minute, allowing for < 3 minutes with counts 1 to 50 counts per minute, were classed as non-wear. To screen out low-movement periods when the device was removed after data collection, any in-bed wear ≥ 24 hours after the last valid day was treated as invalid data for in-bed wear. Daily total activity counts and counts per minute were determined for waking wear periods as well as for rest periods. Participants were included in analyses if they had at least one day of 10 + hours of wear.
Data were screened for abnormal values, and outliers were visually examined by two researchers to reach a consensus decision.

Statistical analysis
Data measures in the interval scale are summarized as mean ± SD whereas data measures in the ordinal or nominal scale are summarized as percentages and counts. Categorical variables between groups were compared using the Chi-square or Fisher exact tests. A generalized linear model was used to model the association of 2-year-old physical activity (response variable) with early pregnancy FMI (primary term of interest) and with other prenatal and post-natal variables. Prenatal variables included maternal characteristics at enrollment [i.e., age, education (i.e., college vs. not college education) and physical activity (i.e., average daily activity counts)], as well as GWG (i.e., adequate, inadequate, and excessive). Postnatal variables were offspring sex, offspring race, offspring birthweight z-score, BMI-Z at age 2 years, and parental feeding practices (restriction, pressure, and monitoring).
The interaction between early pregnancy FMI and covariates were assessed. The base model with the study's primary terms of interest produced a signi cant early pregnancy FMI x offspring sex interaction (β = -1.14, p = 0.020). No interactions were observed between maternal adiposity and any other of the considered variables. Our nal model was created using principles of "Purposeful Selection" advocated by Hosmer et al. (15). Brie y, we evaluated several subsets among the important confounding variables in addition to the terms in the base model. The best subset of these terms was identi ed by observation of minimum model AIC statistic. The normality of the residuals was assessed by SAS Proc Univariate with the Shapiro-Wilk normality test. The test was non-signi cant which is consistent with no departure from normality in the residuals. Statistical analyses were conducted with SAS® 9.4 (Cary, NC, U.S.A).

RESULTS
Study participants (Table 1)  Data presented as mean ± SD, counts and percentages. BMI = body mass index. Unless speci ed otherwise the sample size for each variable is n = 153 Women were predominantly Caucasian (89%), and ≈ 30 years old at the time of enrollment. Seventy-three percent (n = 112) had college education and 54% (n = 83) excessive weight. Forty-six percent, 36%, and 18% of women had adequate, excessive, and inadequate GWG, respectively. The average gestational age at birth was ≈ 39 weeks and 58% of children were boys.
Maternal physical activity (i.e., average activity counts per day x 10 4 ) measured early in pregnancy positively associated with offspring total physical activity at 2 years of age (β = 0.20, p = 0.014). There was no association between 2-year-olds total physical activity and maternal age, maternal adiposity, GWG, birthweight z-score, and parental feeding practices (i.e., restriction, pressure, and monitoring).
Generalized linear regression analysis: base and nal models evaluating associations of 2-year-old offspring total physical activity (activity counts 10 4 ) with prenatal and postnatal variables of interest (Table 3) Non-college degree SE = standard error, CL = con dence limit, FMI = fat mass index In the nal model, maternal FMI, offspring sex, the interaction of maternal FMI with offspring sex, maternal physical activity, and maternal education were retained (Table 4). Compared to boys, the average activity counts per day of girls decreased with increasing maternal adiposity (Figure ). The beta estimate of the nal model for the maternal FMI x offspring sex interaction (β = -1.03, p = 0.031) changed by 9.6% from that obtained in the base model (β = -1.14, p = 0.020). Similarly, the beta estimate of the nal model for early pregnancy physical activity (total activity counts x 10 4 ) did not change compared to bivariate associations (bivariate association: β = 0.21, p = 0.014 vs. nal model: β = 0.21, p = 0.005). For every 10,000 activity counts accrued early in pregnancy, 2-year-old offspring daily activity counts increased by 2,100. After accounting for the interaction of maternal adiposity x offspring sex and maternal physical activity early in pregnancy, the beta estimate for maternal education decreased by 30% compared to bivariate associations (bivariate association: β = 5.0, p = 0.012 vs. nal model: β = 3.46, p = 0.059). Estimates (β) are presented on the scale of the response variable. CI = con dence interval, χ2 = chi square. Adjusted associations: estimates adjusted for early pregnancy maternal physical activity and maternal education.
The association between 2-year-old offspring physical activity with sex was compared at different levels of maternal adiposity (FMI = 4, 5, 6, 7, 8, 9, and 10 kg/m 2 ) ( Table 4). In both adjusted and unadjusted models, offspring total physical activity at age 2 years was signi cantly lower in girls compared to boys when early pregnancy maternal FMI was ≥ 7 kg/m 2 (Table 5).  Fig. 2).
For girls, the best tted model included early pregnancy adiposity (β = -0.66, p = 0.037) and maternal education (β = 4.69, p = 0.052) ( Fig. 2A). Maternal education caused a 20% change in the β estimate of early pregnancy adiposity (from β = -0.82 to β = − 0.66) therefore it was retained in the nal model. For boys (Fig. 2B), early pregnancy physical activity was the strongest predictive variable of offspring physical activity. Speci cally, for every 10,000 activity counts early in pregnancy, 2-year-old boys' daily activity counts increased by 2,400.

DISCUSSION
This study aimed to examine the relationship between early pregnancy maternal adiposity and objectively assessed 2-year-old offspring physical activity, taking into account prenatal and postnatal factors that may impact this outcome. We speci cally hypothesized that, offspring physical activity is inversely associated with early pregnancy maternal adiposity in both sexes, but girls are less active than boys. Our data showed an interaction between offspring sex and early pregnancy maternal adiposity in relation to 2-year-olds physical activity. Speci cally, girls were less active compared to boys in mothers with early pregnancy FMI ≥ 7.0 kg/m 2 . In addition, maternal physical activity early in pregnancy and maternal education directly associated with total physical activity of 2-year-old offspring. When analyses were strati ed by sex, early pregnancy maternal adiposity and maternal education were the strongest variables associated with girls' physical activity. On the other hand, early pregnancy maternal physical activity was the strongest predictor of boy's physical activity.
Little research has been conducted to study maternal obesity programming of offspring physical activity behaviors. Murine models have shed some light indicating that in utero exposure to undernutrition and overnutrition impacts adult offspring physical activity patterns in a sex-dependent manner, but results from these studies are contrasting. Cunha et al. (7)  Taken together, these rodent models suggest that adult physical activity behaviors are programed during prenatal development. Differences in rodent models, and age of testing may explain the variation in results between studies. What remains unknow is how physical activity behaviors in younger offspring relate to maternal obesity in pregnancy. Our data show that the association of 2-year-old offspring total physical activity with early pregnancy maternal adiposity is modi ed by offspring sex. Girls were less active than boys when maternal adiposity early in pregnancy was ≥ 7 kg/m 2 . Sexual dimorphism in physical activity has been reported in children and adolescents (4 to 18 years) with girls spending more time in sedentary activities compared to boys (16-18). It is less clear whether physical activity in infancy and toddlerhood (2 to 3 years of age) varies by sex. Hager et al. (19) evaluated physical activity in toddlers from low-income families (n = 191) from the Baltimore, Maryland region in the United States.
Both, mother and offspring wore Actical accelerometers. In this study, 73% of children were < 2 years of age, mothers were predominantly Black (68%), had overweight/obesity (72%), and completed high school education only (72%). Male sex, non-Black race, and maternal physical activity were all directly associated with offspring moderate to vigorous physical activity.
While the primary outcome of the aforementioned study and ours are different, the results of Hager et al.
are pertinent to our study. Sex differences in infants'/toddlers' physical activity levels were observed in a population primarily comprised of Black women with excessive weight. Other studies objectively measuring physical activity levels of infants/toddlers do not report differences in physical activity between boys and girls. For instance, Hnatiuk et al. (20) characterized physical activity levels and patterns of 18 to 19 month-old Australian children (n = 295) from the Melbourne region. Physical activity was measured using hip-worn GT1M (uniaxial) Actigraph accelerometers. In this study, total daily physical activity, and minutes in moderate to vigorous physical activity did not differ between boys vs. girls (delta = 3.6 minutes/day), although, boys were more active than girls in the morning hours. This study did not report on maternal weight status and 75% of women had post-secondary education.
Bivariate associations from our study ( Table 2) showed that female offspring and Black maternal race negatively associated with offspring total physical activity. However, when maternal adiposity was considered, sex-differences in offspring born to women with FMI < 7 kg/m 2 were attenuated and became evident when this FMI cutoff was exceeded. Contrary to Hager's study, our participants were predominantly educated Caucasians (89%), so it is possible that a lack of statistical power prevented us from conclusively linking maternal Black race to offspring physical activity in our nal model. Whether the association with maternal race is driven by social disparities or biology, cannot be established with the current study design. Our data suggests the possibility that prenatal exposure to maternal overweight/obesity programs early life offspring physical activity behaviors in a sex dependent manner.
However, more studies are needed to con rm this nding. We have previously reported that dietary palmitate (i.e., fat) oxidation is lower in 2-year-old girls than in boys, and that early pregnancy maternal adiposity directly associates with girls' but not boys' adiposity accrual over the rst two years of life (5,21). It is unknown if sexual dimorphism in offspring physical activity behaviors is contributing to these ndings.
Wasenius et al. (9) found a negative association between self-reported maternal gestational weight gain (kg) and objectively measured physical activity (accelerometry) in 3.6-year-old children (β = -3.2, p = 0.049). When analyses were strati ed by sex, they found an interaction between self-reported prepregnancy BMI status and gestational weight gain in association with offspring total physical activity. In girls but not boys, physical activity decreased with increasing gestational weight but only in offspring born to mothers with overweight or obesity (i.e., BMI ≥ 25 kg/m 2 ). On the other hand, in women with normal weight, gestational weight gain directly associated with offspring physical activity. In our study, gestational weight gain (absolute [kg] or status [inadequate, adequate, excessive], did not associate nor interacted with maternal adiposity (or BMI) in relation to 2-year-old offspring physical activity. A limitation of the former study was that questionnaires were used to estimate gestational weight gain which could have increased recall bias.
Our study had some limitations including a sample comprised of mostly Caucasian mothers (89%) and living in Arkansas, which may limit the generalizability of the current ndings to other racial or ethnic groups. However, the study also had numerous strengths including longitudinal collection of data, measurement of maternal anthropometrics directly in clinic, consideration of parenting styles, and most importantly direct objective measures of maternal and offspring physical activity using accelerometry.

Figure 1
Regression plot showing the interaction between early pregnancy maternal adiposity (X axis) and offspring sex in association with 2-year-olds mean activity counts per day (Y axis).

Figure 2
Page 19/19 Fitted plot showing the association of 2-year-old offspring daily activity counts with early pregnancy maternal fat mass index (2A), and early pregnancy maternal daily activity counts (2B).