Demographics and study background

Recruitment to the ORIGINS cohort commenced in October 2016 and targeted pregnant women intending to give birth at Joondalup Health Campus, a tertiary maternity service in Perth, Western Australia, incorporating both public and private care pathways. Recruitment concluded in October 2024, resulting in a cohort of 10,450 mothers and their children. The majority of births (9473; 90.7%) occurred at Joondalup Health Campus, while 977 children (9.3%) were born at other hospitals across Western Australia. Of the enrolled cohort, 3933 participants consented to active participation, which involved ongoing follow-up from pregnancy through birth and early childhood, via questionnaires, biological sample collection, and research clinic visits. ^{Reference Davis, Talati and Whalan17}

Data collection

The ORIGINS participants completed antenatal online questionnaires, specifically the International Physical Activity Questionnaires (IPAQ) and the Godin Leisure-Time Exercise (Godin) questionnaire, which provided information on the types and intensity of physical activities engaged in by the participants. ^{Reference Booth18,Reference Sikes, Richardson, Cederberg, Sasaki, Sandroff and Motl19} Various studies have assessed the validity of these questionnaires, yielding modest agreement with objective measures, ^{Reference Helmerhorst, Brage, Warren, Besson and Ekelund20} which is considered acceptable for self-reported physical activity measures in large population studies. The IPAQ and Godin questionnaires were included in this study because they capture complementary aspects of physical activity. The IPAQ assesses overall physical activity across multiple domains, whereas the Godin questionnaire focuses on the frequency and intensity of leisure-time exercise. Using both measures enabled assessment of whether associations with infant body composition were consistent across different dimensions of maternal physical activity.

International Physical Activity Questionnaires (IPAQ)

The IPAQ is a widely used self-report questionnaire designed to assess an individual’s physical activity levels over a specific period. ^{Reference Noda, Yoshida and Kawakami21} It includes questions on various types of physical activities, including leisure time, work-related, transportation, and domestic activities. Physical activity levels derived from the IPAQ were expressed as Metabolic Equivalent of Task (MET)-minutes per week, calculated from reported walking, moderate-intensity, and vigorous-intensity activities. Participants were classified into low, moderate, or high physical activity categories according to standard IPAQ scoring protocols based on the frequency, duration, and intensity of activity reported during the previous seven days. ^{Reference Booth18}

High physical activity was defined as vigorous activity on at least 3 d accumulating ≥1500 MET-minutes/week or any combination of activities achieving ≥3000 MET-minutes/week. Moderate physical activity was defined according to standard IPAQ criteria, including ≥600 MET-minutes/week, while participants not meeting these criteria were classified as having low physical activity.

The Godin Leisure-Time Exercise questionnaire (Godin)

The Godin Leisure-Time Exercise questionnaire is a brief and easy-to-administer self-report tool to assess leisure-time physical activity over the past seven days. ^{Reference Sikes, Richardson, Cederberg, Sasaki, Sandroff and Motl19} The Godin questionnaire categorises physical activities into three intensities: strenuous (e.g., running, soccer), moderate (e.g., swimming, bicycling), and mild/light exercise. Participants reported the frequency of engaging in strenuous (e.g., running, vigorous swimming, hockey), moderate (e.g., fast walking, easy cycling, recreational tennis), and mild/light (e.g., easy walking, yoga, bowling, golf, light stretching) leisure-time activities for periods of at least 15 min during a typical week. A total Leisure Score Index (LSI) was calculated by weighting the frequency of strenuous, moderate, and mild activities by metabolic equivalents (9, 5, and 3 METs, respectively) and summing the resulting values, with higher scores reflecting greater leisure-time physical activity. The Godin questionnaire is particularly useful for assessing moderate and vigorous physical activity because it quantifies the frequency of leisure-time exercise performed at different intensity levels. ^{Reference Sikes, Richardson, Cederberg, Sasaki, Sandroff and Motl19}

Additional covarites

Information was also collected on additional covariates including socioeconomic status (SEIFA, IRSAD), Mediterranean diet (MDI), ethnicity, maternal age, pre-pregnancy BMI, and parity. ^{Reference Ashwin, Gibson, Hagemann, D’Vaz, Bear and Silva22}

The Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD) is a part of the Socio-Economic Indexes for Areas (SEIFA), developed by the Australian Bureau of Statistics. ²³ IRSAD scores measure the socioeconomic status of geographic areas in Australia based on indicators like income, education, employment, and housing. ²³ Higher scores indicate more socioeconomic advantage, while lower scores indicate greater disadvantage. ²³

The MDI is a measurement tool used to evaluate adherence to the Mediterranean diet, a healthful eating pattern inspired by the traditions of Mediterranean countries. ^{Reference Ashwin, Gibson, Hagemann, D’Vaz, Bear and Silva22,Reference Amati, Hassounah and Swaka24} The MDI assesses how well someone follows the diet’s principles, which include consuming abundant fruits and vegetables, whole grains, healthy fats like olive oil, lean proteins, and limiting red meat. ^{Reference Ashwin, Gibson, Hagemann, D’Vaz, Bear and Silva22} The index assigns scores based on adherence to these guidelines, offering insights into a diet associated with various health benefits, such as reduced risk of chronic diseases. ^{Reference Ashwin, Gibson, Hagemann, D’Vaz, Bear and Silva22} The Mediterranean Diet index employed in this study comprised 13 questions (excluding the alcohol-related item). Participants received a positive score for consuming (1) a high ratio of monounsaturated fats to saturated fats, (2) high olive oil intake, (3) high vegetable intake, (4) high fruit intake, (5) low intake of meat and meat products, (6) low intake of saturated fats, (7) low intake of sweetened carbonated beverages, (8) high legume intake, (9) high fish or seafood intake, (10) less sweets/pastry intake, (11) high intake of nuts, (12) a higher consumption of white meat to red meat ratio, and (13) high sofrito (tomato/onion-based) sauce use when cooking. The medians, stratified according to sex within the sample, served as cutoff points, ^{Reference Costacou, Bamia, Ferrari, Riboli, Trichopoulos and Trichopoulou25} and grams per day were used as the intake measurement. ^{Reference Trichopoulou, Kouris-Blazos and Wahlqvist26}

Body composition assessment

Infant body composition was assessed within 7 d of birth using the PEA POD® Infant Body Composition System (COSMED, Concord, CA, USA), an ADP device that estimates body composition through whole-body densitometry in infants weighing between 1 and 8 kg. ^{Reference McCarthy, Khashan and Murray16} The PEA POD is a non-invasive, validated method that has been widely used in neonatal research due to its accuracy, ease of use, and ability to provide rapid measurements. ^{Reference McCarthy, Khashan and Murray16}

For each assessment, infants were undressed and weighed using the integrated electronic scale. Body volume was then determined by measuring air displacement within the chamber. Standard corrections were applied for air contained within the lungs and for the isothermal air layer adjacent to the skin. Body density was calculated from the measured mass and volume, and body composition was subsequently estimated using a two-compartment model comprising body fat percentage and fat-free mass. Percentage body fat was derived using established assumptions regarding fat density and age-and sex-specific fat-free mass densities based on reference data from Fomon et al. ^{Reference Fomon, Haschke, Ziegler and Nelson27} Once body density had been calculated, body fat percentage and fat-free mass percentage were estimated automatically by the PEA POD software (version 3.0.1).

Participant selection and exclusion criteria

The study focused on a subset of the ORIGINS participants (n = 871) who met specific eligibility criteria, resulting in 323 (37%) participants being included in the analysis (Fig. 1). These criteria included consent to ORIGINS, singleton pregnancies, birthing at Joondalup Health Campus, completion of the IPAQ or Godin questionnaire during the second trimester, mismatch between Godin and IPAQ (those with a inactive score for Godin and high activity level for IPAQ), newborns with a gestational age of at least 37 weeks, and availability of PEA POD measurements within 7 d of birth. Participants with discordant physical activity classifications between the IPAQ and Godin questionnaires were excluded to minimise potential exposure misclassification and increase confidence in physical activity categorisation.

Figure 1.

Inclusion and exclusion criteria.

Data analysis methods

Categorical data was presented as frequencies and percentages. For continuous outcomes and predictors, the data was reported as means and standard deviations when normally distributed and medians and interquartile ranges when data was skewed.

As MDI could be both a confounder and a mediator in the relationship between maternal physical activity and infant body fat percentage, we assessed associations between physical activity levels and MDI, which was done using chi-squared tests. Differences in body fat percentage between female and male infants was analysed using an independent t test.

To analyse the association of physical activity with infant body fat percentage (Peapod fat percent), two linear regression models where used. Model 1 examined the IPAQ as the main predictor, with low physical activity as the reference category. Model 2 examined the Godin, with inactive behaviour as the reference. In both models an interaction term between physical activity and infant sex was included to assess sex-specific differences in association. The models were adjusted for the potential confounding effects of socioeconomic status (SEIFA, IRSAD), MDI, ethnicity, maternal age, pre-pregnancy BMI, and parity. Multiple imputations using chained equations was performed to handle missing data for MDI (missing n = 14) and ethnicity (Caucasian, missing n = 1), with 20 imputed datasets generated. We tested for a linear trend in the outcome across the three ordered physical activity categories within sex using the orthogonal polynomial linear contrast of covariate-adjusted marginal means. Statistical significance was set at p < 0.05. All statistical analysis was performed using Stata v 16.1 (StataCorp, College Station, Texas).

Ethics approval

Prior to commencing this study, ethics approval was obtained from the Ramsay Health Care WA|SA Human Research Ethics Committee (#1440).