Study design
Efficacy of the Melbourne INFANT Program (2008–2010; ISRCTN81847050) was assessed mid- and post-intervention using a cluster-randomised controlled trial design and these results have been reported elsewhere [12, 19,20,21, 23]. Participants were followed-up two (2011–2012) and 3.5 (2013) years after conclusion of the intervention with these post-intervention follow-ups being the focus of this paper, the protocol for which has been published [24]. The CONSORT 2010 statement extension for cluster randomised trials is used in the reporting of this study [25].
Sample
Recruitment for the initial trial has been previously reported [12, 24]. Briefly, 14 local government areas (LGAs) were randomly selected from those within a 60 km radius of the study centre in Melbourne, Australia. From these, 50% of all eligible maternal and child health centres (a free universal service available to all parents in Victoria, Australia) running first time parent groups within each LGA were randomised to the intervention arm. If either an LGA or centre/group declined, they were replaced with the next on the randomly ordered list. Eligible groups included a minimum of eight consenting parents which was reduced to a minimum of six in disadvantaged areas.
For the follow-up phase, all intervention and control families who remained enrolled in the program at the end of the intervention (n = 480; 88.6% of those originally recruited) were recontacted and invited to participate. Renewed written consent was required for participation. Ethics approval for the initial trial and the follow-up (as an extension) were received from Deakin University’s Human Research Ethics Committee (EC 175–2007).
Intervention
The Melbourne INFANT Program [12] was a 15-month health behavior intervention delivered to first-time parents of infants from approximately 4 m of age. The six 2-h group sessions (occurring approximately once every 3 months) were delivered by dietitians within existing first-time parent groups. The intervention was informed by social cognitive theory and utilised an anticipatory guidance framework [18] to facilitate parents’ acquisition of knowledge and strategies to promote healthy dietary intake, physical activity and sedentary behaviors in line with their infants’ developmental phases over the next 15 months. It was structured around key messages [26] which were reiterated throughout the program and provided a framework for parents to understand how these same messages could be applied at different developmental phases. Such intervention strategies could potentially facilitate longer term maintenance, however, no explicit discussion of the preschool developmental phase was incorporated into the program. During the follow-up phase no further intervention occurred.
The control group received six newsletters delivered approximately once every 3 months over the same period as the intervention. Newsletter topics were unrelated to any of the behaviors under investigation e.g. literacy, common childhood illnesses. Control group participants may have received information on topics related to the intervention from their maternal and child health nurse or other health professional during usual care but this was not assessed.
Measures
At each follow-up, parents were provided with a questionnaire for self-completion prior to their home visit. This covered demographic information, screen time, health care usage and a number of tertiary outcomes not reported here. At a prearranged time, researchers visited each participant’s home to collect child anthropometric data, fit accelerometers, and collect questionnaires. Dietary recalls were conducted at unscheduled times following the home visit. Baseline trial data collected when infants were 4 m of age (parent reported demographics and researcher measured infant length and weight) were included as covariates in analyses.
Anthropometry
Height was measured to the nearest 0.1 cm using a portable stadiometer (Seca 220/217, Hamburg). Weight was measured in light clothing to the nearest 10 g using digital scales (Tanita BWB-800/InnerScan 50, Illinois). Waist circumference was measured to the nearest 0.1 cm using a steel non-stretch tape (Lufkin Executive Thinline W606PM, Maryland). All measures were taken twice by trained research staff, with a third taken if the difference exceeded 0.5 cm/0.5 kg. The mean of the closest two measures was used in analyses. Body mass index (BMI; kg/m2) and BMI z-scores (zBMI) were calculated based on World Health Organisation BMI-for-age growth charts [27].
Dietary intake
Using the 24-h recall method [28], trained researchers conducted unscheduled telephone interviews with parents on three non-consecutive days, including one weekend day, to capture all food and drink consumed by the child on the previous day. Food measurement booklets were provided to assist parents with estimating quantities. Data were categorised according to the Australian Food, Supplement & Nutrient Database (AUSNUT2007) (Food Standards Australia New Zealand, Canberra, Australia, 2008), with additional infant-specific products added. Coding of all food items was checked for accuracy and completeness by a dietitian. Average daily intakes of fruits (excluding juice), vegetables (excluding potatoes), noncore sweet foods (eg, chocolate, cakes), noncore savory foods (eg, crisps, savory biscuits), noncore drinks (ie, fruit juice, soft drinks), and water were calculated. Variety of fruits was calculated by summing the number of specific fruits reported by parents across all recalls. Juice was excluded. Where multi-variety foods were reported (e.g. fruit salad) a score of 2 or 3, rather than 1, was assigned for the purposes of calculation. The same process was used to calculate vegetable variety. Potato was excluded.
Physical activity
Children wore ActiGraph™ accelerometers (Model GT1M, Pensacola) on an elasticised belt at the right hip during waking hours for 8 consecutive days. Movement counts were recorded in 15-s epochs. Light-intensity physical activity (LPA) was defined as > 25 and < 420 counts [29]. Moderate- to vigorous-intensity physical activity (MVPA) was defined as ≥420 counts [29]. Non-wear time, defined as ≥20 min of consecutive zero counts, was removed. Children with at least 4 days of ≥7.4 h of recorded data were included in analyses, which has been shown to provide a reliable estimate of habitual physical activity [30].
Sedentary behaviors
Parents reported the time (hours and minutes) their child usually spent watching television/DVDs on a typical weekday and typical weekend day, using items with established reliability [31]. Average daily minutes of television viewing was generated ((weekdays × 5 + weekend × 2)/7). To assess time spent sitting, children wore activPAL™ monitors (PAL Technologies Ltd., Glasgow) on an elasticised belt on the front of the right thigh, midway between the knee and hip, for 8 consecutive days during waking hours. Non-wear time, defined as ≥20 min of consecutive zero counts based on the vertical axis of the accelerometer in the activPAL™, was removed. Children with at least 2 days of ≥6 h of recorded data were included in analyses. Two days of data were selected to maximise inclusion and were shown to be not substantially different from 3+ days of data in sensitivity analyses.
Economic analysis
Economic evaluation considered the incremental costs of the intervention from a healthcare perspective (costs accrued in the intervention compared to the control arm) and net of any reduction in service use over the period of follow-up. The pre-specified definition of cost-effectiveness assessed costs only in terms of additional costs per unit change in BMI [24]. Intervention costs were calculated at the end of the intervention period [12], and converted to 2018 figures. Over the period of follow-up, parents reported any services used due to concerns over their own or their child’s weight, diet or physical activity. Parents were first asked to indicate whether they had accessed any services for these purposes since last completing a survey. If they responded ‘yes’, they were asked to report the average number of visits, average total cost and out of pocket cost for visits to each of eight commonly utilised health services (e.g. GP, mother-baby or parenting centre) plus any other health professionals they may have visited. Reported service use was valued in 2018 Australian dollars using existing national estimates.
Data analyses
Participants were analysed on an intention-to-treat basis. Effects of the intervention were assessed at each time point by testing for differences between trial arms on the outcome variables using maximum likelihood linear mixed models with random intercepts for parent groups, to account for clustering. Due to skewness among outcome variables, bootstrapping with 2000 resamples was used to calculate model standard errors and normal-approximation confidence intervals were constructed. Unadjusted models were tested for all outcomes with the exception of the BMI z-score outcomes where baseline values were available and included as covariates. Behavioral outcomes were not available at baseline given children were 4 m old. All physical activity models (unadjusted and adjusted) included average accelerometer wear time. Adjusted models included child age, child sex, and maternal education as covariates. Additionally, mothers’ pre-pregnancy BMI (reported at baseline) was included in models for the BMI and waist circumference z-score outcomes, and child overall energy intake was included in the dietary outcomes models. Since the study outcomes were measured in a variety of units, standardised effect sizes (d) were calculated for the forest plots by dividing adjusted mean differences between treatment groups by the standard deviation of the outcome in the control group. An effect size of d = 1 would indicate the intervention caused an estimated increase of one standard deviation in the outcome, relative to the level of the outcome without intervention. Stata/SE 15.0 (StataCorp, Texas) was used for analyses.