PHYSICAL ACTIVITY MEASURED EARLY IN PREGNANCY AND BIRTHWEIGHT- A STUDY FROM THE ARCH COHORT By Diana Haggerty A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Epidemiology-Master of Science 2017 ABSTRACT PHYSICAL ACTIVITY MEASURED EARLY IN PREGNANCY AND BIRTHWEIGHT- A STUDY FROM THE ARCH COHORT By Diana Haggerty Excessive birthweight is a growing concern and challenge in current obstetrics, and has been associated with perinatal complications, such as shoulder dystocia, and long-term health complications such as type 2 diabetes. (1-3) Variation of birthweight measures used to describe excessive birthweight contributes to the lack of consensus in the literature. In addition, these measures do not allow for the study of the effect of Physical Activity (PA) at non-traditional points in the distribution. The effect of PA may not be found at mean birthweight estimates, but an effect may be present above the 70th percentile of birthweight z-scores standardized for infant sex and gestational age.(4) We hypothesized that women who performed 150 minutes or more of PA per week on average would have lower odds of delivering a neonate whose birthweight z-score was above the 70th percentile of the sample™s birthweight z-score compared to the neonates of women who were sedentary. We also sought to replicate findings that birthweight z-score varies by PA category above the 70th percentile of birthweight z-score using a sample in which PA behavior data was collected prospectively.(4) We used data from the Archives for Research on Childhood Health cohort. We calculated birthweight z-score and trichotomized PA. We used multivariable logistic regression to model the association between PA measured early in pregnancy and birthweight z-score above the 70th percentile. This study did not support the hypothesis that the odds of delivering a neonate above the 70th percentile would be lower for women who performed 150 minutes or more of PA per week compared to the neonates women who were sedentary.iii This thesis is dedicated to the families who participated in ARCH. Without you, this work would not exist.iv ACKNOWLEDGMENTS I am grateful to the Archive for Research in Child Health and principal investigator Nigel Paneth MD MPH for the use of these data and to Michigan State University's Biomedical Research Informatics Core for making them available. I would like to acknowledge my thesis committee members, including Dr. James Pivarnik (chair), Dr. Nigel Paneth, and Dr. Joseph Gardiner. v TABLE OF CONTENTS LIST OF TABLES––––––––––––––––––––––––––––––––––––––..––––................–.. vi LIST OF FIGURES––––––––––––––––––––––...–––––––––––––.––––––––............... vii KEY TO ABBREVIATIONS––––––––––––––––––––.––..––––..............––––––––––.–..– viii INTRODUCTION––––––––––––––––––––––––––––––––––..............–––––––––––– 1 Mechanism of High Birthweight–––––––––––––––––––––––––––––––.............–. 2 Risk Factors that May Confound or Modify the Relationship between PA and Birthweight–––.. 3 Physical Activity and Pregnancy––––––––––––––––––––––––.–––––.............––... 3 Physical Activity and Birthweight–––––––––––––––––––––––––.–––.............–––. 4 Gaps in Knowledge.......................................................................................................................... 6 Objective......................................................................................................................................... 6 METHODS––––––––––––––––––––––––––––––––––––––––––––..–..............–––. 7 Conceptual Model.......................................................................................................................... 7 Inclusion......................................................................................................................................... 8 Exposure.......................................................................................................................................... 8 Outcome......................................................................................................................................... 9 Covariates....................................................................................................................................... 9 Statistical Analysis......................................................................................................................... 10 Post Hoc Power Analysis............................................................................................................... 11 RESULTS––––––––––––––––––––––––––––––––––––––––––..––––..............–.–. 12 DISCUSSION––––––––––––––––––––––––––––––––––––––––..............––––––.– 15 CONCLUSIONS–––––––––––––––––––––––––––––––––––..............––––––.––––. 18 APPENDIX–––––––––––––––––––––––––––––––––––––..............–––––––.–––– 19 REFERENCES –––––––––––––..–––––––––––––––––––––..............–––––..–––......... 31 vi LIST OF TABLES Table 1. Distribution of variables by included and excluded subjects*, ARCH Cohort 2010-2015 –........ 21 Table 2. Distribution of Variables of subjects, ARCH Cohort 2010-2015–––––––––........................... 23 Table 3. Bivariate Distribution of Birthweight Z-Scorea and Covariates by Physical Activity, ARCH Cohort 2010-2015––..........................................................................................................................................– 25 Table 4. Bivariate Distribution of Physical Activity and Covariates by Birthweight Z-scorea, ARCH Cohort 2010-2015.................................................................................................................................................. 27 Table 5. Odds ratios for the odds of having birthweight z-score greater than the 70th percentilea, ARCH Cohort 2010-2015.–––––––––––––––––––––––––––––––––––..............................–... 29 Table 6. Summary of backward elimination a, ARCH Cohort 2010-2015–––––––––––––............... 30 Table 7. Association of predicted probabilities and observed responses of adjusted models, ARCH Cohort 2010-2015––––––––––..––............................................................................................................... 30 vii LIST OF FIGURES Figure 1: Flow chart of eligibility of subjects in the ARCH Cohort for analysis. Observations included in analysis based on inclusion criteria...––––...–––......––––.............................................................. 20 viii KEY TO ABBREVIATIONS ARCH Archive for Research on Childhood Health CI Confidence Interval DHHS Department of Health and Human Services HBW High Birthweight LGA Large for Gestational Age NHB Non-Hispanic Black or African American NHW Non-Hispanic White or Caucasian PA Physical Activity T2DM Type 2 Diabetes Mellitus 1 INTRODUCTION High Birthweight (HBW) has been described as a current challenge in obstetrics practice, and preventing birthweights in the high end of the distribution is an important step in controlling adverse pregnancy outcomes and chronic disease.(1)Adverse pregnancy outcomes can include complications such as shoulder dystocia (odds ratio: 2.4. 95% confidence interval [CI}: 1.6, 3.6), and neonatal hypoglycemia (odds ratio: 4.2. 95% CI: 3.3, 5.4).(2) Odds of type II diabetes mellitus (T2DM) may be increased in persons whose birthweight was greater than 4000 grams (odds ratio: 1.3. 95% CI: 1.0, 1.6). (3) A variety of definitions of HBW have been used across the literature, which has contributed to the debate about the relationship between Physical Activity (PA) and birthweight. (5-7) Measures of birthweight include mean birthweight (8), Large for Gestational Age (LGA)(7), HBW(9), and fetal growth ratio.(10) LGA, which highlights one end of the birthweight distribution, focuses only on the 90th percentile of birthweight and above. HBW is often defined as a birthweight greater than 4000 grams (see CDC Pediatric and Pregnancy Nutrition Surveillance System for surveillance definition: http://www.cdc.gov/pednss/what_is/pednss_health_indicators.htm). Nevertheless, prevention of birth weights in the extreme ends of the distribution is a public health priority; Healthy People 2020 has a goal of reducing the maternal mortality rate due to labor and delivery complications(11), some of which have been closely associated HBW pregnancy outcomes.(2, 12, 13) An association between birthweight greater than 4000 grams and excessive bleeding (odds ratio: 1.41. 95% CI: 1.37, 1.45), prolonged labor (odds ratio: 1.38. 95% CI: 1.35,1.41), cephalopelvic distortion (odds ratio: 2.42. 95% CI: 2.41, 2.45), and C-section delivery (odds ratio: 1.62. 95% CI: 1.61, 1.63).(12) Other important US public health goals include reducing obesity rates and diseases associated with obesity, such as T2DM and heart disease. (11) There is evidence that suggests these diseases are influenced by birthweight. Associations between birthweights and health problems associated with obesity, such as childhood obesity (14, 15), metabolic syndrome (16), T2DM(3, 17, 18), and cardiovascular disease(16), have been found. One study found for 2 each kilogram increase in birthweight, childhood body mass index (BMI) increased by half a unit (95% CI: 0.47, 0.51). (15) The hazard ratio for metabolic syndrome in children who were born LGA was 2.2 (95% CI: 1.3, 3.8) when compared to children who were born average for gestational age. (19) Mechanism of High Birthweight Maternal glucose levels during pregnancy are associated with birthweight. (20-22) It has been shown that odds of LGA were 3.6 times higher for women whose fasting glucose was greater than 130 mg/dl when compared to women whose fasting glucose was less than 99 mg/dl. (21) Gestational diabetes and chronic diabetes have both been associated with high birthweight outcomes, and this association has been noted for decades.(23, 24) Glucose crosses the placenta and the fetus responds by producing insulin.(25) Insulin in the fetal environment promotes growth, and it has been proposed that in the presence of excess glucose, the fetus will produce more insulin than a fetus exposed to optimal glucose, thereby promoting excessive growth.(26-28) Effective maternal glucose control, achieved through dietary advice, medication, and glucose monitoring during pregnancy has been demonstrated as an important step toward disrupting this pathway to high birthweight outcomes in clinical trials.(29, 30) One trial demonstrated a 100 gram reduction in mean birthweight through control of mild gestational diabetes.(29) Importantly, adverse delivery outcomes, including serious perinatal complications and shoulder dystocia, were reduced in glucose intervention groups in some of these trials, which demonstrated that the interruption of the maternal glucose-fetal insulin growth pathway is important for reducing morbidity and mortality associated with high birthweight.(29, 31) PA during pregnancy may be an effective intervention to help prevent women from delivering pathologically large infants, but few trials have tested this. A 2012 Cochrane Review of PA trials and perinatal outcomes found only two trials that analyzed exercise interventions and macrosomia, defined as birthweight greater than 4000 grams.(32) The estimated effect size was modest and statistically not significant.(32) 3 Risk Factors that May Confound or Modify the Relationship between PA and Birthweight Important risk factors associated with high birthweight include overweight or obese pre-pregnancy BMI classification (23, 33, 34) , maternal height (23, 33), parity greater than two (23, 33), and excessive gestational weight gain.(35, 36) Some of these risk factors are modifiable, while others are not, but effective targeted interventions for reducing the associated risk are essential. Maternal pre-pregnancy BMI is believed to affect birthweight through modification of placental hormones that regulate metabolism and delivery of nutrients to the fetus, such as insulin like growth factor.(37) One study found that mothers who weighed 80 or more kilograms had a 1.9 (95% CI: 1.7, 2.2) fold increase in odds of giving birth to an infant who weighed more than 4000 grams. (33) Maternal height is believed to influence birthweight through genetic pathways(38), but the relationship between fetal growth ratio and PA varies significantly by maternal height.(10) Mothers whose height was between 1.81 and 1.9 meters had an increased odds of delivering an infant greater than 4000 grams of 1.2 (95% CI: 1.1, 1.3), while mothers whose height was in excess of 1.9 meters had odds of delivering a HBW infant of 1.4 (95% CI: 1.2,1.5).(33) Excessive gestational weight gain has been associated with increased risk of delivering high birthweight infants (36, 39), but an exact mechanism has not been elucidated. One study found that regardless of pre-pregnancy BMI classification, excessive gestational weight gain was associated with greater than two-fold odds of delivering a HBW infant.(36) Glycemic load during pregnancy has been associated with both excessive gestational weight gain and high birthweight outcomes, which suggests a nutritional pathway associated with maternal glucose levels.(40) Maternal education, used as a proxy for socio-economic status, has been shown to be associated with birthweight, but a biological mechanism linking the two has not been elucidated.(23, 33) Physical Activity and Pregnancy PA during a healthy pregnancy has been recognized as an asset to both mother and fetus; the United States Department of Health and Human Services (DHHS) recommends healthy pregnant women 4 engage in at least 150 minutes of moderate PA per week.(41) The literature has demonstrated that PA during pregnancy does not increase the risk of delivering a low birthweight baby.(42-46) In non-pregnant populations, PA has been demonstrated to be effective at increasing insulin sensitivity; accumulated daily PA has a direct and positive effect on insulin sensitivity, and that this process may be similar in pregnant women.(47, 48) One study showed that moderate PA and vigorous PA had a positive impact on insulin sensitivity in non-pregnant women (beta coefficient for moderate PA: 0.08. beta coefficient for vigorous PA: 0.22. p-value for trend: 0.006).(47) Modifiable risk factors for high birthweight, such as excessive gestational weight gain, may also be positively impacted by PA.(49, 50) It has been observed that the fetal growth ratio of the offspring of active tall women was similar to the fetal growth ratio of the offspring of both active and inactive shorter women, while inactive taller women delivered offspring with statistically significant higher fetal growth ratios than inactive shorter women, active shorter women, and active taller women, which suggests PA may modify the relationship between maternal height and birthweight.(10) Physical Activity and Birthweight It has been proposed that PA can be leveraged to prevent high birthweight outcomes; however, results from trials have shown mixed effects. Most trials utilized a composite outcome, though one trial specified fetal macrosomia, one trial specified LGA, and one specified mean birthweight as primary endpoints.(8, 49, 51) The trial that specified macrosomia, defined as birthweight greater than 4000g, found no association between birthweight and the PA program that was tested.(8) A second trial stratified analyses by pre-pregnancy BMI, but found no significant difference between LGA offspring in the intervention groups regardless of pre-pregnancy BMI.(49) One trial found PA reduced birthweight, but only in mothers who had gestational diabetes (mean difference: 302 grams, p-value = 0.02).(51) A fourth trial tested the effect of a lifestyle counseling intervention on birthweight; the intervention included counseling on PA, and found a 100 gram reduction in mean birthweight in the intervention 5 group. (52) Observational studies have yielded mixed results for the relationship of PA and birthweight. (4, 6, 7, 9, 10, 34, 48, 53-55). Studies that have found inverse associations between birthweight and PA during pregnancy frequently dichotomize birthweight, using measures such as birthweights greater than 4000 grams, or LGA. Large observational studies, such as the Danish National Birth Cohort and the Norwegian Mother and Baby Cohort (MoBa), have demonstrated no difference in mean birthweight when comparisons between exercising women™s infants were made to non-exercising women, but have found PA reduces risk for LGA.(7, 9) A study from the Danish National Birth Cohort found no difference in mean birthweight, but the hazard ratio for delivering LGA was 0.93 (95% CI: 0.88, 0.98).(7) Other studies have found PA during pregnancy exerts no effect on birthweight, or has less effect than modifying pre-pregnancy risk factors like overweight or obese BMI.(34, 56) Challenges to interpreting these study results include varying measures of birthweight, including grams(34), fetal growth ratio(10), and birthweight z-score(4), as well as different definitions of high birthweight, including LGA and a cut point of greater than 4000 grams.(7, 9) Variation in the timing of PA assessment methods is another challenge to interpreting the results of these studies, with the timing of assessment ranging from prospective to retrospective.(4, 7, 9, 10, 34, 46) Many prospective studies have assessed PA during the second or third trimester of pregnancy, or during both trimesters; others have not reported the timing of data collection for PA.(34, 53, 57) One study found PA in the first trimester was associated with lower risk of high birthweight, but it was a small investigation conducted in urban dwelling Indian women and may not be generalizable to the United States.(58) Women in this study who spent the most time doing physical activity had increased odds of delivering an infant in the lowest tertile; women were 1.6 times (95% CI: 1.0,2.4) more likely to deliver in the lowest tertile compared to women who performed the least physical activity.(58) In spite of variation across the literature, association between PA and birthweight is most likely captured in the extreme ends of the birthweight distribution. Studying the variation in birthweight at 6 different cut-points in the high end of the distribution, such as the 70th percentile, may help further describe the association between PA and HBW. Gaps in Knowledge Mudd et al. suggest that birthweight may be impacted by leisure time PA during pregnancy above the 70th percentile of the birthweight z-score distribution, however, this study used retrospective PA recall three to nine years after the study pregnancy.(4) It remains unknown if this relationship between PA and birthweight z-score holds in a cohort characterized by prospectively collected PA data. It also remains unclear if the timing of PA during pregnancy is an important factor in the relationship between PA and birthweight. Additionally, it is unknown if this relationship holds for all daily PA, as opposed to leisure time only PA. Objective Our objective was to describe the association between PA measured at a mean gestational age of 13.8 weeks and birthweight z-score above the internally calculated 70th percentile. (4) 7 METHODS We obtained data from the Archive for Research on Childhood Health (ARCH). Briefly, ARCH is a biotrust which prospectively collects maternal questionnaire data, birth certificate information, hospital discharge data, and biological samples as early as four weeks into pregnancy. Women are recruited from three prenatal care clinics in the Lansing, MI area, and mother and child dyads are followed for at least five years post-birth. Mothers must be at least 18 years of age at enrollment and speak English. Mean gestational age at enrollment for the entire cohort is 13.2+ 6.3 weeks, but women are eligible to enroll at any time prior to birth. The mean gestational age at enrollment for the study sample was 13.8+ 6.3 weeks. The Michigan State University Institutional Review Board for Human Subjects Research approved this study. Conceptual Model The conceptual model for this study was based on work by Mudd et al. Their study used quantile regression, and identified the 70th percentile of birthweight z-score to be the point in the distribution where birthweight z-score differed by retrospectively measured leisure time PA.(4) For this study, our approach was to dichotomize the sample™s birthweight z-scores at the 70th percentile. We used logistic regression to estimate the odds ratio for birthweight z-score above the 70th percentile for prospectively measured PA categories. Categories were a) active, but not meeting Department of Health and Human Services recommendation of 150 minutes of moderate activity per week (here after referred as active, not meeting recommendation), b) active, meeting Department of Health and Human Services recommendation of 150 minutes per week (here after referred to as active, meeting recommendations), and C) women who reported performing no PA (here after referred to as sedentary). PA was not limited to leisure time in this study, but could include PA attributable to work or leisure. 8 Inclusion For this analysis, we included mother and child dyads if they had recorded PA data as well as birthweight and gestational age data recorded for the calculation of birthweight z-score. Additionally, subjects included had to have been enrolled in ARCH during the index pregnancy; the gestational age at enrollment had to be plausible, and were limited to gestational age at enrollment between 1 and 42 weeks. Subjects with gestational age at enrollment greater than 42 weeks were excluded. Subjects must also have had complete data available for covariates, including maternal race, maternal age, maternal height, pre-pregnancy BMI, parity status, educational attainment, household income, and maternal smoking prior to or during pregnancy. Dyads with one of the above variables missing were excluded (see Figure 1). Exposure Our exposure of interest was variable PA measured at intake interview for the ARCH cohort. ARCH investigators assessed PA by using questions modified from the 2009 Behavior Risk Factor Survey.(59) Pregnancy PA was ascertained at the recruitment interview using the following questions: During the past month, did you do any moderate activities for more than ten minutes that caused a small increase in your breathing and heart rate? o If yes, how many days a week do you usually do these moderate activities? o How much time do you usually spend doing these moderate activities in one day? During the past month did you do any vigorous activities for more than ten minutes that caused a small increase in your breathing and heart rate? o If yes, how many days a week do you usually do these vigorous activities? o How much time do you usually spend doing these vigorous activities in one day? 9 The number of moderate PA minutes per session was multiplied by the number of moderate PA sessions per week reported to calculate the number of moderate PA minutes per week. The number of vigorous PA minutes per session was multiplied by the number of vigorous PA sessions per week to calculate the number of vigorous PA minutes per week. The number of vigorous minutes performed per week was then doubled to roughly equate vigorous and moderate minutes of PA. This calculation assumes the intensity of vigorous activity is approximately twice that of moderate activity. The total number of minutes usually performed per week was calculated by adding the number of moderate PA minutes per week and the doubled number of vigorous PA minutes performed per week. Outcome Our outcome was birthweight z-score dichotomized at the 70th percentile of the sample. We derived birthweight z-score from the birthweight in grams reported on the birth certificate subtracted from the United States population mean birthweight for gestational age and sex, and divided by the United States population standard deviation proposed by Talge, Mudd et al.(60) The sample™s 70th percentile of birthweight z-score was calculated, and a dichotomous variable was created that classified infants as at or below the 70th percentile or above it. Covariates We calculated pre-pregnancy body mass index (weight in kg/height in meters, squared) and women were categorized into one of three groups: less than 25 kg/m2, 25-29.9 kg/m2, and 30 kg/m2 or greater. The range of pre-pregnancy BMI was 16.5 - 56.5. We categorized maternal age in five groups: Less than 20 years of age, 20-24 years of age, 25-29 years of age, 30-34 years of age, and 35 years of age or older. We categorized maternal height into women less than the 75th percentile of height, women between the 75th and 90th percentile, and women above the 90th percentile, based on the United States reference for women over 20 years of age.(61) We categorized marital status as unmarried, unmarried but living with baby™s father, and married. We categorized parity as nulliparous, primiparous, and 10 multiparous. Due to small numbers, we divided maternal race into three categories: Non-Hispanic White or Caucasian (NHW), Non-Hispanic Black or African American (NHB), and Other Race or Ethnicity, including Latina. We dichotomized household income into less than US$25,000 per year versus US$25,000 a year or more. We used covariates that were collected from the intake interview, except maternal smoking before or during pregnancy, parity and maternal age at birth, which were collected from birth certificate data. Statistical Analysis We conducted univariate analysis of frequencies of the birthweight z-score above the 70th percentile, PA categories, and the following covariates of interest: pre-pregnancy BMI, maternal age, maternal height, parity, maternal race/ethnicity, and household income. We considered excessive gestational weight gain a mediator and did not include it in modeling. Bivariate analysis for the outcome of greater than the 70th percentile birthweight z-score and the exposure of active, meeting recommendations, active not meeting recommendation, and sedentary were conducted using frequencies and chi-square tests. We assessed frequency tables of covariates and PA with chi-square tests. We used logistic regression to assess the impact of PA on birthweight z-score by modeling the probability of having a birthweight greater than the 70th percentile given PA status, while controlling for the following covariates: maternal height, maternal age, household income, maternal race, maternal education, parity status, pre-pregnancy BMI, smoking status before or during pregnancy, and marital status. We used backward selection to reduce the model, with a p-value of 0.2 set as the threshold for retention. After backward selection, we tested interactions among retained variables, and used backward selection to determine which interaction terms to retain. All analyses were performed with SAS 9.4; the primary procedure was Proc Logistic (SAS Institute, Cary, NC). 11 Post Hoc Power Analysis We conducted a post hoc power analysis to determine the power detecting a difference in birthweight z-score above the 70th percentile in women who reported active, meeting recommendations PA and women who were classified as sedentary. Analysis was performed with Stata 13. We set the alpha level at p< 0.05. Calculation for power was calculated from a sample size of 331, the control proportion of 21.5% of births greater than the 70th percentile, and an odds ratio of 1.2. 12 RESULTS At the time of analysis (October 2016), ARCH had enrolled 841 women. Of the 841 observations, 223 observations were missing birthweight z-score. Of the remaining 618 observations, 276 did not have data available for PA. Covariates were missing for 11 observations, so the final model included 331 women (Figure 1). Median gestational age at enrollment was 13.8 weeks, which was not different from the gestational age at enrollment of women not included. Observations included in the sample varied significantly from those that were excluded (Table 1) by maternal age (p-value <0.0001), marital status (p-value = 0.004), household income (p-value = 0.007), and maternal smoking status before or during pregnancy (p-value <0.0001). Table 2 shows the distribution of the outcome, birthweight z-score, the main exposure, PA performed during pregnancy, and selected covariates. Overall, 43.2% of the sample was classified as sedentary, 25.7% was classified as active, not meeting recommendation, and 31.1% was classified as active, meeting recommendation. More than half the sample was classified as overweight or obese prior to pregnancy, while 44.1% of the sample had a pre-pregnancy BMI of 25 kg/m2 or lower. NHW women comprised 53.4% of the sample, while NHB women comprised 25.7% and women who reported another race or ethnicity, including Latina, comprised 20.9% of the sample. 48.3% of the sample was under the age of 25, while 27.8% were 25 to 29 years of age. Mothers age 30 and older made up 24.9% of the sample. Women whose height was less than the US 75th percentile made up 64.4% of the sample. Women whose height was between the 75th and 90th US percentile made up 19.9% of the population, and women whose height was greater than the US 90th percentile accounted for 15.7% of the sample. In this sample, 40.5% were nulliparous, 31.4% were primiparous, and 28.1% were multiparous. More than 70% of the sample participants were unmarried at the time of the interview, 36.9% were unmarried but living with the baby™s father, while 37.2% reported being unmarried; 26.0% of the sample was married at the intake interview. The majority of the sample, 72.5%, lived in households with incomes of less than US$25,000 per year. About 50% of the sample had at least some 13 college education, 18.3% did not finish high school, while 30.2% had a high school diploma or GED. The majority of the sample, 69.2%, did not smoke prior to or during pregnancy. Table 3 gives the frequencies for the covariates by PA status. There were few differences across PA groups. PA performed during pregnancy was not associated with birthweight z-score; however, key covariates that did show a significant difference in frequency of PA status were maternal race (p-value =0.0005), marital status (p-value = 0.0003), household income (p-value = 0.0003), and maternal education (p-value= 0.04). Table 4 gives the results of bivariate analysis of the covariates by birthweight z-score status. Frequency of birthweight z-score above the 70th percentile was significantly different across maternal race (p-value= 0.003) and maternal height status (p-value = 0.02). Table 5 provides the beta estimates of the adjusted logistic regression modeling the probability of a birthweight z-score above the 70th percentile given PA status. The final model was adjusted for maternal height, and maternal race and ethnicity, pre-pregnancy BMI, smoking before or during pregnancy, and marital status. Table 6 shows the odds ratio for birthweight z-score greater than the 70th percentile. For the active, meeting recommendation women compared to sedentary women was 1.1 (95% CI: 0.6, 2.0) when holding all other variables constant, but was not statistically significant. The odds ratio for birthweight z-score for active, not meeting recommendations women compared to sedentary women was 1.3 (95% CI: 0.7, 2.4) when holding all other variables constant and was not statistically significant. Birthweight z-score above the 70th percentile did vary by maternal height when holding all other variables constant. For mothers whose height was between the 75th percentile and the 90th percentile for US women 20 years and older, there was a 2.2 fold increase in odds of delivering an infant whose birthweight z-score was greater than the 70th percentile (odds ratio: 2.2 [95% CI: 1.2, 3.9]). Mothers whose height was greater than the 90th percentile did not show a statistically significant difference in odds of delivering an infant greater than the 70th percentile of birthweight z-score. Mothers who 14 reported being NHB had a 50% reduction in odds of delivering an infant with a birthweight z-score above the 70th percentile (0.5 [95% CI: 0.2, 0.9]). The other covariates were not significantly associated with birthweight z-score. The c-statistic was 0.66 indicating the model was a poor predictor of birthweight z-score above the 70th percentile. Results from the post hoc power analysis showed the power to detect a difference was 13%, based on a proportion in the reference group (sedentary women) of 41.3%, an odds ratio of 1.2, and a sample size of 331. 15 DISCUSSION These results did not support the hypothesis that PA measured early in pregnancy and meeting the DHHS recommendations of 150 minutes per week performed during pregnancy will impact birthweight above the 70th percentile. In the literature, PA measured at multiple time points or after pregnancy is associated with a lower probability of delivering infants greater than 4000 grams, or who are LGA.(4, 7, 9) Multiple measures or recalled measures of PA may be representative of PA across pregnancy. This study used one measure of PA taken at a mean of 13.8 weeks gestation at enrollment. The early pregnancy timing of the interviews does not capture PA over the course of the entire pregnancy. The cumulative effect of PA across pregnancy may be necessary for affecting birthweight; this needs to be tested in longitudinal cohort studies that take repeated measures of PA across pregnancy with a diverse sample of women. Perkins et al used multiple time points to assess the relationship between PA and fetal growth ratio (birthweight/United States median birthweight for sex, gestational age, and race), but was limited to a sample focused on White, college educated women.(10) Further research with a diverse sample to allow generalizability is an important next step. The measure of PA did not discern leisure time from work associated PA; work associated PA has been associated with small for gestational age, while leisure time PA has not.(62) It is possible that without separating the two, the effect of work and the effect of leisure time PA may cancel each other out. Compared to self-reported leisure time PA data from the 2000 Behavioral Risk Factor Survey, more women in this sample reported sufficient PA levels (31% of this sample versus 15%).(63) This study allowed women to include work associated PA, which could account for the higher prevalence of PA compared to the national estimates presented by Evenson et al. However, given the allowance for inclusion of work physical activity, the proportion of women in this sample who reported no physical activity was 43.2%, which is slightly higher than the proportion (38.8%) reported by in the non-subcohort women who recalled their PA three to nine years after pregnancy in the study by Mudd et al. 16 (4) Different measures of PA may contribute to this difference, but there may be other differences between the samples of the two studies that may be equally or more important. This highlights the challenge of consistently and accurately measuring PA using self-report. Maternal height between the US 75th and 90th height percentile for women over age 20 had a significant odds ratio of 2.2 (95% CI: 1.2, 3.9). This relationship did not demonstrate an interaction with PA. This is different from the results of a recent study of the relationship between PA level and fetal growth ratio.(10) The authors found the relationship between PA and fetal growth ratio to be most prominent in the tallest women in the study.(10) The study used accelerometry and PA diaries to ascertain the relationship between daily PA during pregnancy and fetal growth ratio. Measures of PA were collected over the entire day, made at 20 and 32 weeks gestation, and provided a range of PA levels that provided sufficient variation to detect differences.(10) Women who reported NHB race were 50 % less likely to deliver a neonate with birthweight z-score above the 70th percentile compared NHW women (OR: 0.5, 95% CI: 0.2, 0.9). The pattern of NHB women having smaller babies at term has been noted elsewhere.(64) Women who reported another race or ethnicity, including Latina, did not have significantly different odds of delivering an infant above the 70th percentile of birthweight z-score compared to NHW women. Interestingly, other covariates that typically predict differences in birth weight, such as maternal education, household income, parity, smoking status before or during pregnancy, and marital status, did not have significant associations with birthweight z-score in this analysis. Limitations of this study include the sample; women for whom data were collected early in the ARCH recruitment period did not report PA; thus almost three hundred women were not available for analysis in this study. Women who did not report PA at their initial interview were also excluded from this analysis. Along with this, data collection for ARCH is ongoing, and birth certificate data were not yet available for 223 women, reducing the available sample size from 841 births to 342 births. Post hoc 17 analysis shows these women are significantly different from the women who did not meet inclusion criteria by important variables including maternal smoking status before or during pregnancy, marital status, and household income. The ARCH cohort was recruited from a limited geographic area, and is not necessarily generalizable to all populations. It is possible the recruitment strategy for ARCH generated a sample too similar to make meaningful comparisons of PA during pregnancy. This study has many strengths. Prospective ascertainment of PA is important for accuracy.(4) The ARCH protocol collects PA data at the initial enrollment interview, which is administered prior to birth. ARCH collects data from multiple sources, allowing for more complete data collection, and the ability to check for mistakes across sources. We used birthweight z-score for our outcome, which controls for sex and gestational age, which allowed us to use all gestational ages at birth, instead of limiting the analysis to infants born at term. This analysis differed significantly from the work by Mudd et al. in terms of timing of PA assessment and ascertainment of PA.(4) Data were prospectively collected in this study. A different analysis was also conducted to assess the association between birthweight z-score above the 70th percentile.(4) This analysis used a dichotomized variable that divided the entire sample at the 70th percentile regardless of PA level. The analysis from Mudd et al. used quantile regression, which quantifies the difference between the birthweight z-score at the 70th percentile for women who were active and meeting PA recommendations, and compares them to the women who were not meeting PA recommendations.(4) This is a subtle difference, but one that is important for interpreting the results of this study. Quantile regression was attempted for this study, but the small sample size limited the ability to test the adjusted model. 18 CONCLUSIONS This study does not support the hypothesis that PA performed early in pregnancy is associated with reduced odds of birthweight greater than the 70th percentile. Measuring PA in a sample that encompasses a wide variety of PA, and follows the participants across pregnancy may be more informative for studying the relationship between PA and birthweight outcomes. Determining the longitudinal impact of PA on birthweight in diverse populations is an important step in determining the ability for PA to reduce risk of HBW outcomes. 19 APPENDIX20 21 Table 1. Distribution of variables by included and excluded subjects*, ARCH Cohort 2010-2015 Variable Included Percent Excluded Percent Chi-Square p-Value BMIb Category Less than 25 147 17.5 241 28.8 0.8 0.7 25-29.9 86 10.3 123 14.7 30 or Greater 98 11.7 143 17.1 N = 838 Maternal Race NHW 177 21.4 276 33.3 0.8 0.7 NHB 85 10.3 114 13.8 Other including Latina 69 8.3 108 13.3 N = 829 Maternal Age <25 Years 160 19.0 363 43.2 45.9 <0.0001 25-29 Years 92 10.9 76 9.0 30-34 Years 50 6.0 51 6.1 Greater than 35 Years 29 3.5 20 2.4 N = 841 Maternal Height At or Below 1.67 meters 213 25.3 353 42.0 2.5 0.3 Between the 1.67 and 1.71 meters 66 7.9 93 11.1 Above 1.71 meters 52 6.2 64 7.6 N = 841 Parity Statusa Nulliparous 134 21.7 119 19.3 0.2 0.9 Primiparous 104 16.8 85 13.8 Multiparous 93 15.1 83 13.4 N = 618 Marital Status Married 86 10.3 182 21.7 10.8 0.004 Unmarried, Living with Baby's Father 123 14.7 145 17.3 Unmarried 122 14.6 180 21.5 N = 838 Household Income Under US$25,000 240 29.7 303 37.5 7.4 0.007 US$25,000 or Higher 91 11.3 175 21.6 N = 809 Maternal Education Did not Finish High School 60 7.3 71 8.6 4.4 0.2 22 Table 1. (cont™d) High School Graduate or GED 100 12.1 169 20.4 Some College 112 13.5 152 18.4 College Graduate or Higher 59 7.1 104 21.0 N = 827 Maternal Smoking Before or During Pregnancya Yes 102 16.6 47 7.6 17.1 <0.0001 No 229 37.2 238 38.6 N = 616 Gestational Diabetesac Yes 10 1.6 6 1.0 Fisher's Exact Test 0.6 No 320 51.9 280 45.4 Total = 594 Gestational Hypertensionac Yes 22 3.6 10 1.6 Fisher's Exact Test 0.02 No 308 49.9 276 44.7 Total = 594 Gestational Age at Enrollment Mean Std Dev Mean Std Dev Mean Diff p-value 13.2 6.3 13.8 6.5 0.6 0.2 Sample Size/ Equality of Var Test 288 327 Folded F 0.7 Abbreviations: ARCH, Archive for Research of Childhood Health; PA, physical activity; NHW, non-Hispanic/Latina White; NHB, non-Hispanic/Latina Black; BMI, body mass index; kg, kilogram; m, meter; diff, difference; var, variance. * To be included subjects had to have complete data for birthweight, sex of neonate, gestational age at birth, physical activity, and covariates. a Covariates obtained from birth certificates. Subjects whose birth certificate data had not yet been received were excluded. b Weight (kg)/height (m)2 c Not included in covariates necessary for inclusion 23 Table 2. Distribution of Variables of subjects, ARCH Cohort 2010-2015 N= 331 Variable Frequency Percent Birthweight Z-Scorea Below 70th Percentile 234 70.7 Above 70th Percentile 97 29.3 Physical Activity Category Sedentary 143 43.2 Inadequate 85 25.7 Adequate 103 31.1 BMIb Category Less than 25 147 44.1 25-29.9 86 26.0 30 or Greater 98 29.6 Maternal Race Non-Hispanic White 177 53.4 Non-Hispanic Black or African American 85 25.7 Other including Hispanic and Latina 69 20.9 Maternal Age Less than 25 Years 160 48.3 25-29 Years 92 27.8 30-34 Years 50 15.1 Greater than 35 Years 29 8.8 Maternal Height At or Below 1.67 meters 213 64.4 Between the 1.67 and 1.71 meters 66 19.9 Above 1.71 meters 52 15.7 Parity Status Nulliparous 134 40.5 Primiparous 104 31.4 Multiparous 93 28.1 Marital Status Married 86 26.0 Unmarried, Living with Baby's Father 122 36.9 Unmarried 123 37.2 Household Income Under US$25,000 240 72.5 US$25,000 or Higher 91 27.5 Maternal Education Did not Finish High School 57 18.3 High School Graduate or GED 94 30.2 Some College 105 33.8 24 Table 2. (cont™d) College Graduate or Higher 55 17.7 Maternal Smoking Before or During Pregnancy Yes 102 30.8 No 229 69.2 Gestational Diabetesc Yes 10 3.0 No 320 96.7 Total 330 Gestational Hypertensionc Yes 22 6.7 No 308 93.1 Total 330 Abbreviations: ARCH, Archive for Research of Childhood Health; PA, physical activity; NHW, non-Hispanic/Latina White; NHB, non-Hispanic/Latina Black; BMI, body mass index; kg, kilogram; m, meter; diff, difference; var, variance. a Birthweight z-score was calculated using the United States™ gestational age and sex specific means and standard deviations. The sample™s 70th percentile was calculated and used as the cut-point. b Weight (kg)/height (m)2 c Not included in covariates necessary for inclusion because they are mediators and were not intended to be used in model. 25 Table 3. Bivariate Distribution of Birthweight Z-Scorea and Covariates by Physical Activity, ARCH Cohort 2010-2015 Physical Activity Category Chi-Square P-Value Sedentary Active, less than 150 minutes Active, 150 minutes or more Birthweight Z-Score Below 70th Percentile 106 74.1 57 67.1 71 68.9 1.5 0.5 Above 70th Percentile 37 25.9 28 32.9 32 31.1 Pre-Pregnancy BMIb Category Less than 25 65 45.5 39 45.8 43 41.7 6.7 0.2 25-29.9 37 25.9 16 18.8 33 32.0 30 or Greater 41 28.7 30 35.3 27 26.2 Maternal Race Non-Hispanic White 59 41.3 54 63.5 64 62.1 20.2 0.0005 Non-Hispanic Black or African American 47 32.9 22 25.9 16 15.5 Other including Latina 37 25.9 9 10.6 23 22.3 Maternal Age Less than 25 Years 83 58.0 40 47.1 37 35.9 13.4 0.04 25-29 Years 29 20.3 25 29.4 38 36.9 30-34 Years 19 13.3 14 16.5 17 16.5 Greater than 35 Years 12 8.4 6 7.1 11 10.7 Maternal Height At or Below 1.67 meters 96 67.1 57 69.5 60 58.3 3.2 0.5 Between the 1.67 and 1.71 meters 27 18.9 17 20.0 22 21.4 Above 1.71 meters 20 14.0 11 12.9 21 20.4 Parity Status Nulliparous 60 42.0 38 44.7 36 35.0 3.2 0.5 Primiparous 42 29.4 28 32.9 34 33.0 Multiparous 41 28.7 19 22.4 33 32.0 Marital Status Married 25 17.5 33 38.8 28 27.2 21.2 0.0003 Unmarried, Living with Baby's Father 49 34.3 26 30.6 47 45.6 Unmarried 69 48.3 26 30.6 28 27.2 Household Income Under US$25,000 120 83.9 55 64.7 65 63.1 16.5 0.0003 US$25,000 or Higher 23 16.1 30 35.3 38 36.9 26 Table 3. (cont™d) Maternal Education Did not Finish High School 36 25.2 10 11.8 14 13.6 13.4 0.04 High School Graduate or GED 47 32.9 26 30.6 27 26.2 Some College 42 29.4 30 35.3 40 38.8 College Graduate or Higher 18 12.6 19 22.4 22 21.4 Maternal Smoking Before or During Pregnancy Yes 42 29.4 23 27.1 37 35.9 2 0.4 No 101 70.6 62 72.9 66 64.1 Gestational Diabetesc Yes 3 2.1 5 6.0 2 1.9 Fisher's Exact Test 0.2 No 140 97.9 79 94.0 101 98.1 Gestational Hypertensionc Yes 8 5.6 9 10.7 5 4.9 Fisher's Exact Test 0.2 No 135 40.8 75 89.3 98 95.1 Abbreviations: ARCH, Archive for Research of Childhood Health; PA, physical activity; NHW, non-Hispanic/Latina White; NHB, non-Hispanic/Latina Black; BMI, body mass index; kg, kilogram; m, meter; diff, difference; var, variance. a Birthweight z-score was calculated using the United States™ gestational age and sex specific means and standard deviations. The sample™s 70th percentile was calculated and used as the cut-point. b Weight (kg)/height (m)2 c Not included in covariates necessary for inclusion because they are mediators and were not intended to be used in model. 27 Table 4. Bivariate Distribution of Physical Activity and Covariates by Birthweight Z-scorea , ARCH Cohort 2010-2015 Birthweight Z-Score Variable At or Below 70th Percentile Above 70th Percentile Chi- Square P-value Physical Activity Category Sedentary 106 45.3 37 38.1 1.5 0.5 Active, did not meet recommendation 57 24.4 28 28.9 Active, did meet recommendation 71 30.3 32 33.0 Pre-Pregnancy BMIb Category 18.5-24.9 111 47.7 36 37.1 3.6 0.2 25-29.9 55 23.5 31 32.0 30 or Greater 68 29.1 30 30.9 Maternal Race/ Ethnicity Non-Hispanic White 116 49.6 61 62.9 6.8 0.03 Non-Hispanic Black 69 29.5 16 16.5 Other, Including Hispanic and Latina 49 20.9 20 20.6 Maternal Age Less than 25 Years 117 50.0 43 44.3 1.4 0.7 25-29 Years 61 26.1 31 32.0 30-34 Years 36 15.4 14 14.4 Greater than 35 Years 20 8.5 9 9.3 Maternal Height At or Below 1.67 meters 160 68.4 53 54.6 7.5 0.02 Between the 1.67 and 1.71 meters 38 16.2 28 28.9 Above 1.71 meters 36 15.4 16 16.5 Parity Status Nulliparous 96 41.0 38 39.2 0.1 0.9 Primiparous 73 31.2 31 32.0 Multiparous 65 27.8 28 28.9 Marital Status Married 54 23.1 32 33.0 3.7 0.2 Unmarried, Living with Baby's Father 91 38.9 31 32.0 Unmarried 89 38.0 34 35.1 Household Income Under US$25,000 175 74.8 65 67.0 2.1 0.1 US$25,000 or Higher 59 25.2 32 33.0 Maternal Education Did not Finish High School 45 19.2 15 15.5 1.6 0.7 High School Graduate or GED 73 31.2 27 27.8 Some College 77 32.9 35 36.1 College Graduate or Higher 39 16.7 20 20.6 28 Table 4. (cont™d) Maternal Smoking Before/During Pregnancy Yes 66 28.2 36 37.1 2.6 0.1 No 168 71.8 61 62.9 Gestational Diabetesc n= 330 Yes 6 2.6 4 4.1 Fisher's Exact Test 0.6 No 227 97.0 93 95.9 Gestational Hypertensionc n= 330 Yes 15 6.4 7 7.2 Fisher's Exact Test 0.9 No 218 93.2 90 92.8 Abbreviations: ARCH, Archive for Research of Childhood Health; PA, physical activity; NHW, non-Hispanic/Latina White; NHB, non-Hispanic/Latina Black; BMI, body mass index; kg, kilogram; m, meter; diff, difference; var, variance. a Birthweight z-score was calculated using the United States™ gestational age and sex specific means and standard deviations. The sample™s 70th percentile was calculated and used as the cut-point. b Weight (kg)/height (m)2 c Not included in covariates necessary for inclusion because they are mediators and were not intended to be used in model. 29 Table 5. Odds ratios for the odds of having birthweight z-score greater than the 70th percentilea, ARCH Cohort 2010-2015. Variable OR 95% CI PA Category Sedentary 1.0 Referent Active, did not meet recommendations 1.3 0.7 2.4 Active, met recommendations 1.1 0.6 2.0 Maternal Height Category Less than or equal to 1.67 meters 1.0 Referent Between 1.67 and 1.71 meters 2.2 1.2 3.9 Above 1.71 meters 1.4 0.7 2.8 Maternal Race and Ethnicity NHW 1.0 Referent NHB 0.5 0.2 0.9 Other Race including Latina 0.8 0.4 1.6 Pre-Pregnancy BMIb Category Less than 25 kg/m2 1.0 Referent 25 - 29.9 vs Less than 25 kg/m2 1.7 0.9 3.2 30 or greater vs Less than 25 kg/m2 1.5 0.8 2.7 Maternal Smoking Before or During Pregnancy No 1.0 Referent Yes 1.5 0.9 2.5 Marital Status Married 1.0 Referent Unmarried 0.6 0.3 1.1 Unmarried, living with baby`s father 0.9 0.4 1.7 Abbreviations: ARCH, Archive for Research of Childhood Health; PA, physical activity; NHW, non-Hispanic/Latina White; NHB, non-Hispanic/Latina Black; BMI, body mass index; kg, kilogram; m, meter. a Birthweight z-score was calculated using the United States™ gestational age and sex specific means and standard deviations. The sample™s 70th percentile was calculated and used as the cut-point. b Weight (kg)/height (m)2 30 Table 6. Summary of backward elimination a, ARCH Cohort 2010-2015 Step Effect Removed DF Number In Wald Chi-Square Pr > ChiSq 1 Education Level 3 9 0.22 0.97 2 Parity Category 2 8 0.53 0.77 3 Household Income 1 7 0.13 0.72 4 Maternal Age category 3 6 1.41 0.70 Abbreviations: ARCH, Archive for Research of Childhood Health; DF, degrees of freedom; Pr, Probability. a To remain in model, variable had to have a p-value of less than 0.2. Table 7. Association of predicted probabilities and observed responses of adjusted models, ARCH Cohort 2010-2015 Percent Concordant 65.4 Somers' D 0.31 Percent Discordant 34.3 Gamma 0.31 Percent Tied 0.3 Tau-a 0.13 Pairs 22698 c 0.66 31 REFERENCES32 REFERENCES 1. Henriksen T. The macrosomic fetus: a challenge in current obstetrics. Acta obstetricia et gynecologica Scandinavica 2008;87(2):134-45. 2. Weissmann-Brenner A, Simchen MJ, Zilberberg E, et al. Maternal and neonatal outcomes of macrosomic pregnancies. Medical science monitor : international medical journal of experimental and clinical research 2012;18(9):Ph77-81. 3. Harder T, Rodekamp E, Schellong K, et al. Birth weight and subsequent risk of Type 2 Diabetes: a meta-analysis. American journal of epidemiology 2007;165(8):849-57. 4. Mudd LM, Pivarnik J, Holzman CB, et al. Leisure-time physical activity in pregnancy and the birth weight distribution: where is the effect? Journal of physical activity and health 2012;9(8):1168-77. 5. Zhang X, Decker A, Platt RW, et al. How big is too big? The perinatal consequences of fetal macrosomia. American journal of obstetrics and gynecology 2008;198(5):517.e1-.e6. 6. Krogsgaard S, Gudmundsdottir SL, Nilsen TI. Prepregnancy physical activity in relation to offspring birth weight: a prospective population-based study in Norway-The HUNT Study. Journal of pregnancy 2013;2013:780180. 7. Juhl M, Olsen J, Andersen PK, et al. Physical exercise during pregnancy and fetal growth measures: a study within the Danish National Birth Cohort. American journal of obstetrics and gynecology 2010;202(1):63.e1-.e8. 8. Haakstad LA, Bo K. Exercise in pregnant women and birth weight: a randomized controlled trial. BMC pregnancy and childbirth 2011;11:66. 9. Owe KM, Nystad W, Bo K. Association between regular exercise and excessive newborn birth weight. Obstetrics and gynecology 2009;114(4):770-6. 10. Perkins CC, Pivarnik JM, Paneth N, et al. Physical activity and fetal growth during pregnancy. Obstetrics and gynecology 2007;109(1):81-7. 11. About healthy people. Washington DC: United States Department of Health and Human Services; 2012. (http://healthypeople.gov/2020/about/default.aspx). (Accessed February 19 2014 2014). 12. Boulet SL, Alexander GR, Salihu HM, et al. Macrosomic births in the united states: determinants, outcomes, and proposed grades of risk. American journal of obstetrics and gynecology 2003;188(5):1372-8. 13. Stotland NE, Caughey AB, Breed EM, et al. Risk factors and obstetric complications associated with macrosomia. International journal of gynecology and obstetrics 2004;87(3):220-6. 33 14. Hediger ML, Overpeck MD, McGlynn A, et al. Growth and fatness at three to six years of age of children born small- or large-for-gestational age. Pediatrics 1999;104(3):e33. 15. Toselli S, Zaccagni L, Celenza F, et al. Risk factors of overweight and obesity among preschool children with different ethnic background. Endocrine 2014:1-9. 16. Huang RC, Burke V, Newnham JP, et al. Perinatal and childhood origins of cardiovascular disease. International Journal of Obesity 2006;31(2):236-44. 17. Sugihara S, Sasaki N, Amemiya S, et al. Analysis of weight at birth and at diagnosis of childhood-onset type 2 diabetes mellitus in Japan. Pediatric diabetes 2008;9(4pt1):285-90. 18. Wei J-N, Sung F-C, Li C-Y, et al. Low birth weight and high birth weight infants are both at an increased risk to have Type 2 Diabetes among schoolchildren in Taiwan. Diabetes care 2003;26(2):343-8. 19. Boney CM, Verma A, Tucker R, et al. Metabolic syndrome in childhood: association with birth weight, maternal obesity, and gestational diabetes mellitus. Pediatrics 2005;115(3):e290-e6. 20. Liu B, Chen H, Xu Y, et al. Fetal growth is associated with maternal fasting plasma glucose at first prenatal visit. PloS one 2014;9(12):e116352. 21. Scholl TO, Sowers M, Chen X, et al. Maternal glucose concentration influences fetal growth, gestation, and pregnancy complications. American journal of epidemiology 2001;154(6):514-20. 22. Yang X, Zhang H, Dong L, et al. The effect of glucose levels on fetal birth weight: a study of Chinese gravidas in Tianjin, China. Journal of diabetes and its complications 2004;18(1):37-41. 23. Lindell G, Marıál K, Källén K. Impact of maternal characteristics on fetal growth in the third trimester: a population-based study. Ultrasound in obstetrics and gynecology 2012;40(6):680-7. 24. Heckbert SR, Stephens CR, Daling JR. Diabetes in pregnancy: maternal and infant outcome. Paediatric and perinatal epidemiology 1988;2(4):314-26. 25. Jansson T, Cetin I, Powell TL, et al. Placental transport and metabolism in fetal overgrowth Œ a workshop report. Placenta 2006;27, Supplement(0):109-13. 26. Fowden AL. The role of insulin in fetal growth. Early human development 1992;29(1-3):177-81. 27. Walsh JM, Mahony R, Byrne J, et al. The association of maternal and fetal glucose homeostasis with fetal adiposity and birthweight. European journal of obstetrics and gynecology and reproductive biology 2011;159(2):338-41. 28. Kieffer EC, Alexander GR, Kogan MD, et al. Influence of diabetes during pregnancy on gestational age-specific newborn weight among US black and US white infants. American journal of epidemiology 1998;147(11):1053-61. 34 29. Landon MB, Spong CY, Thom E, et al. A multicenter, randomized trial of treatment for mild gestational diabetes. New England journal of medicine 2009;361(14):1339-48. 30. George A, Mathews JE, Sam D, et al. Comparison of neonatal outcomes in women with gestational diabetes with moderate hyperglycaemia on metformin or glibenclamide Œ a randomised controlled trial. Australian and New Zealand journal of obstetrics and gynaecology 2015;55(1):47-52. 31. Crowther CA, Hiller JE, Moss JR, et al. Effect of treatment of gestational diabetes mellitus on pregnancy outcomes. New England journal of medicine 2005;352(24):2477-86. 32. Han S, Middleton P, Crowther CA. Exercise for pregnant women for preventing gestational diabetes mellitus. The Cochrane database of systematic reviews 2012;7:Cd009021. 33. Ørskou J, Henriksen TB, Kesmodel U, et al. Maternal characteristics and lifestyle factors and the risk of delivering high birth weight infants. Obstetrics and gynecology 2003;102(1):115-20. 34. Fleten C, Stigum H, Magnus P, et al. Exercise during pregnancy, maternal prepregnancy body mass index, and birth weight. Obstetrics and gynecology 2010;115(2 Pt 1):331-7. 35. Ferraro ZM, Barrowman N, Prud™homme D, et al. Excessive gestational weight gain predicts large for gestational age neonates independent of maternal body mass index. The journal of maternal-fetal and neonatal medicine 2012;25(5):538-42. 36. Chihara I, Hayes D, Chock L, et al. Relationship between gestational weight gain and birthweight among clients enrolled in the special supplemental nutrition program for Women, Infants, and Children (WIC), Hawaii, 2003Œ2005. Maternal and child health journal 2014;18(5):1123-31. 37. King JC. Maternal obesity, metabolism, and pregnancy outcomes. Annual review of nutrition 2006;26(1):271-91. 38. Zhang G, Bacelis J, Lengyel C, et al. Assessing the causal relationship of maternal height on birth size and gestational age at birth: a Mendelian randomization analysis. PLoS medicine 2015;12(8):e1001865. 39. Cox Bauer CM, Bernhard KA, Greer DM, et al. Maternal and neonatal outcomes in obese women who lose weight during pregnancy. Journal of perinatology : official journal of the California Perinatal Association 2016. 40. Horan MK, McGowan CA, Gibney ER, et al. Maternal low glycaemic index diet, fat intake and postprandial glucose influences neonatal adiposity Œ secondary analysis from the ROLO study. Nutrition journal 2014;13(1):1-12. 41. 2008 Physical activity guidelines for Americans summary. United States Department of Health and Human Services; 2008. (http://www.health.gov/paguidelines/guidelines/summary.aspx). (Accessed 6 January 2015 2015). 35 42. Hegaard HK, Pedersen BK, Bruun Nielsen B, et al. Leisure time physical activity during pregnancy and impact on gestational diabetes mellitus, pre-eclampsia, preterm delivery and birth weight: a review. Acta obstetricia et gynecologica Scandinavica 2007;86(11):1290-6. 43. Mudd LM, Owe KM, Mottola MF, et al. Health benefits of physical activity during pregnancy: an international perspective. Medicine and science in sports and exercise 2013;45(2):268-77. 44. Kramer MS, McDonald SW. Aerobic exercise for women during pregnancy. The Cochrane database of systematic reviews 2006(3):Cd000180. 45. Rose NC, Haddow JE, Palomaki GE, et al. Self-rated physical activity level during the second trimester and pregnancy outcome. Obstetrics and gynecology 1991;78(6):1078-80. 46. Duncombe D, Skouteris H, Wertheim EH, et al. Vigorous exercise and birth outcomes in a sample of recreational exercisers: a prospective study across pregnancy. The Australian and New Zealand journal of obstetrics and gynaecology 2006;46(4):288-92. 47. Balkau B, Mhamdi L, Oppert J-M, et al. Physical activity and insulin sensitivity: the RISC study. Diabetes 2008;57(10):2613-8. 48. Hayes L, Bell R, Robson S, et al. Association between physical activity in obese pregnant women and pregnancy outcomes: the UPBEAT pilot study. Annals of nutrition and metabolism 2014;64(3-4):239-46. 49. Hui AL, Back L, Ludwig S, et al. Effects of lifestyle intervention on dietary intake, physical activity level, and gestational weight gain in pregnant women with different pre-pregnancy body mass index in a randomized control trial. BMC pregnancy and childbirth 2014;14:331. 50. Stuebe AM, Oken E, Gillman MW. Associations of diet and physical activity during pregnancy with risk for excessive gestational weight gain. American journal of obstetrics and gynecology 2009;201(1):58.e1-.e8. 51. Barakat R, Pelaez M, Lopez C, et al. Exercise during pregnancy and gestational diabetes-related adverse effects: a randomised controlled trial. British journal of sports medicine 2013;47(10):630-6. 52. Luoto R, Kinnunen TI, Aittasalo M, et al. Primary prevention of gestational diabetes mellitus and large-for-gestational-age newborns by lifestyle counseling: a cluster-randomized controlled trial. PLoS medicine 2011;8(5):e1001036. 53. Hegaard HK, Petersson K, Hedegaard M, et al. Sports and leisure-time physical activity in pregnancy and birth weight: a population-based study. Scandinavian journal of medicine and science in sports 2010;20(1):e96-102. 54. Tomic V, Sporis G, Tomic J, et al. The effect of maternal exercise during pregnancy on abnormal fetal growth. Croatian medical journal 2013;54(4):362-8. 36 55. Voldner N, Froslie KF, Bo K, et al. Modifiable determinants of fetal macrosomia: role of lifestyle-related factors. Acta obstetricia et gynecologica Scandinavica 2008;87(4):423-9. 56. Ruifrok AE, Althuizen E, Oostdam N, et al. The relationship of objectively measured physical activity and sedentary behaviour with gestational weight gain and birth weight. Journal of pregnancy 2014;2014:567379. 57. Owe KM, Nystad W, Skjaerven R, et al. Exercise during pregnancy and the gestational age distribution: a cohort study. Medicine and science in sports and exercise 2012;44(6):1067-74. 58. Dwarkanath P, Muthayya S, Vaz M, et al. The relationship between maternal physical activity during pregnancy and birth weight. Asia Pacific journal of clinical nutrition 2007;16(4):704-10. 59. Behavioral Risk Factor Surveillance System: BRFSS 2009 survey data and documentation Centers for Disease Control and Prevention [CDC]; 2009. (http://www.cdc.gov.proxy1.cl.msu.edu/brfss/annual_data/annual_2009.htm). (Accessed 31 July 2015 2015). 60. Talge NM, Mudd LM, Sikorskii A, et al. United States birth weight reference corrected for implausible gestational age estimates. Pediatrics 2014. 61. Fryar CD, Gu Q, Ogden CL. Anthropometric reference data for children and adults: United States, 2007-2010. Vital and health statistics Series 11, Data from the national health survey 2012(252):1-48. 62. Croteau A, Marcoux S, Brisson C. Work activity in pregnancy, preventive measures, and the risk of delivering a small-for-gestational-age infant. American journal of public health 2006;96(5):846-55. 63. Evenson KR, David, Savitz A, et al. Leisure-time physical activity among pregnant women in the US. Paediatric and perinatal epidemiology 2004;18(6):400-7. 64. Alexander GR, Kogan M, Martin J, et al. What are the fetal growth patterns of singletons, twins, and triplets in the United States? Clinical obstetrics and gynecology 1998;41(1):114-25.