This retrospective cohort study was conducted using data from the Birthing Outcomes System (BOS) at a major tertiary institution in eastern Australia between the 1st of January 2008 and the 31st of December 2017. There were 30,121 births over this period, with the hospital being the major maternity centre for a catchment population of 550,000. However, stillbirths and multiple pregnancies were excluded as were births where maternal BMI had not been recorded. This left 27,814 birth events for analysis. Ethical approval was obtained from the relevant Health Research Ethics and Governance Office. (no: ETHLR.18.048).
Data Assessment
Maternal BMI was derived from measured height (cm) and weight (kg) recorded at the first antenatal appointment (usually at 12-14 weeks gestation) [26]. Classification of BMI was defined, according to WHO cut-offs, into four groups: underweight (<18.5kg/m2); healthy weight (18.5-24.9kg/m2); overweight (25-29.9kg/m2); obese class I (30-34.9kg/m2); obese class II (35-39.9kg/m2) and obese class III (>40kg/m2) [2].
Other demographic information which was collected included maternal age, maternal country of birth, relationship status, employment, smoking (both maternal and paternal), parity, and obstetric outcomes such as GDM, hypertensive disorders of pregnancy and premature rupture of membranes.
Maternal place of birth is recorded in the BOS database. Women were categorised into three broad groups: ‘all’(regardless of ethnicity) ‘Australian-born’ and ‘Asian-born’. The Standard Australian Classification of Countries (SACC), Second Edition [23] was used to define the nations to be included in this final category (for example China, India, Pakistan).
Gestational age was calculated from either the last menstrual period or the earliest ultrasound examination. The Australian national birthweight percentiles published by Dobbins et al., were used to calculate LGA defined as a birthweight > 90th percentile for gestational age [26, 31, 32]. Birthweight results were expressed as SD (z) scores corrected for gestation at time of birth.
Maternity complications such as GDM and hypertensive disorders of pregnancy were defined according to the World Health Organisations (WHO’s) International Statistical Classification of Diseases and Related Health Problems manual [25]. Screening for GDM is universally conducted at the study hospital between 24- and 28-weeks’ gestation with a 75 g oral glucose tolerance test (OGTT). A positive diagnosis is made if the fasting plasma glucose is 5.1 - 6.9 mmol/L or if the 2-hour post glucose load is 8.5 - 11.0 mmol/L. Women with GDM receive group education from experienced dietitians and diabetes educators. This includes blood glucose monitoring, carbohydrate counting and recommended physical activity levels. Women are strongly encouraged to attend individual follow-up appointments either weekly or fortnightly in line with the Australasian Diabetes in Pregnancy Society (ADIPS) consensus guidelines for the testing and diagnosis of hyperglycaemia in pregnancy [26]. Data are entered into the database by clinicians contemporaneously or as soon as practicable after an episode of care with regular validation checks by the system administrator. Mandatory reporting fields are validated by the Epidemiology Section of the Department of Population Health at the jurisdiction level.
Statistical Analysis
Descriptive analysis was reported using means and standard deviations for continuous variables and frequencies and percentages for categorical variables. Binary logistic regression was performed, to assess the relationship between maternal BMI, GDM and LGA. Following this, multivariate binary logistic regression, using the forced entry method, was applied to associations found to be significant at the bivariate level. All models were adjusted for parity, baby gender, marital status, smoking, maternal country of birth, employment and premature rupture of membranes. These covariates are considered by clinicians working in the filed as important and have been used in similar published analyses on this topic [8, 26, 33, 34].
Cook's distance values were used to examine for multivariate outliers and influential data points. All cases included in the study had Cook’s D values below one. No signs of multicollinearity were observed, and an acceptable goodness of fit model was found. Statistical significance was set at p <0.05. Analyses were conducted using SPSS version 24 (SPSS Inc., Chicago, USA) [27].