Body mass index and physical activity in seven-year-old children whose mothers exercised during pregnancy: follow-up of a multicentre randomised controlled trial

This is a follow-up study of a Norwegian multicentre RCT that investigated whether regular exercise during pregnancy could prevent gestational diabetes [25]. Pregnant women booking appointments for routine ultrasound scans at St. Olavs Hospital, Trondheim University Hospital, and Stavanger University Hospital were invited to participate.

In total, 875 of approximately 12,000 women accepted participation during the inclusion period from April 2007 to June 2009. Of these, 855 women in week 18–22 of pregnancy were randomly assigned to receiving a 12-week standard exercise program (intervention group) or standard antenatal care (control group). Inclusion criteria were Caucasian women aged 18 years or older with a singleton live foetus. We included only Caucasian women to ensure an ethnically homogenous sample as ethnicity is a risk factor for gestational diabetes [26, 27]. The age limit was set due to age of majority in Norway, and singleton live foetus to avoid pregnancy complications. Exclusion criteria were high-risk pregnancies, diseases that could interfere with participation and women living too far away (> 30 min’ drive) from the hospitals. Randomisation in blocks of 30 was performed by a web based computerised procedure at Unit for Applied Clinical Research, Norwegian University of Technology and Science. Because of the nature of the study, physiotherapists leading group sessions and study participants were not blinded. The women were examined at baseline (week 18–22 of pregnancy), at end of intervention (week 32–36 of pregnancy) and 3 months after delivery [28]. Pregnancy outcome and newborn data were registered at time of delivery [25]. The children were clinically assessed at 18 months of age [10] and by parent-report at 7 years of age [11].

Data for the 7-year follow-up were collected electronically using the CHECKWARE software (CheckWare AS, Trondheim, Norway) from October 2014 to December 2016. Parents of all 855 children received the questionnaire in the autumn term of their child’s second year of primary school.

Women in the intervention group received a standardised exercise program including aerobic activity of moderate intensity, strength training and balance exercises as recommended by the American College of Obstetricians and Gynecologists and the Norwegian Directorate of Health [6, 29]. Training sessions of 60 min instructed by a physiotherapist were offered once per week over a period of 12 weeks (between week 20 and 36 of pregnancy). In addition, the women were encouraged to follow a written 45-min home exercise program at least twice per week, including 30 min of endurance training and 15 min of strength and balance exercises. Exercises were shown in text and pictures in a booklet and instructed at group sessions. Adherence to the protocol was defined as exercising 3 days per week or more at moderate to high intensity and participating in at least one group session per week [25]. Moderate intensity was defined as getting sweaty and a little out of breath during exercise. Performing the exercise programme was strongly emphasised and recorded in the women’s personal training diaries and through reports from physiotherapists leading the training groups. In addition, PA was recorded in a questionnaire for both groups.

Women in the control group received standard antenatal care in Norway and the customary information given by their midwife or general practitioner. They were not discouraged from exercising on their own [25].

At the 7-year follow-up, we used questions regarding height, weight, diseases and health problems from a questionnaire developed by the National Institute of Health for the Norwegian Mother and Child Cohort (MoBa) study [30, 31]. The children’s height was reported in cm and their weight in kg. BMI was calculated as weight in kilogram divided by the square value of height in meters (kg/m²). We calculated the children’s iso-BMI, i.e. BMI adjusted for sex and age, by using a weight calculator [32]. We used iso-BMI as a continuous variable and as a dichotomised variable with 25 kg/m² as a cut off for overweight [32, 33]. Parents reported whether their child had any of the following diseases or health problems; rheumatoid arthritis, cancer, diabetes, cerebral palsy, attention-deficit hyperactivity disorder, coeliac disease, bone fractures, epilepsy, mental retardation, autistic traits, Asperger’s syndrome, chronic fatigue syndrome/myalgic encephalomyelitis, tonsillectomy, ear drainage or other conditions or congenital diseases [31].

The PA questions were the same as used in a Norwegian cohort study in the county of Telemark [34, 35]. Two of the questions were equal to questions from the WHO Health Behaviour in School-aged Children questionnaire [36], also used in the large Norwegian Young-HUNT study [37, 38]. In order to assess whether the child met the recommendation from the Norwegian Directorate of Health [6] to perform 1 h or more of daily moderate to vigorous PA (MVPA), parents reported total (including school, after-school program and leisure time) daily MVPA for their child as 1) Less than 1 h or 2) 1 h or more [34]. Frequency of leisure time MVPA, i.e. PA performed outside school and after-school program where the child was out of breath or sweaty, was reported as number of times per week (never/once a month or less/once a week/2 to 3 times a week/4 to 6 times a week/every day). Weekly leisure time MVPA was reported as approximate hours per week (none/1 h/2 to 3 h/4 to 6 h/7 h or more). Intensity of PA was reported as 1) takes it easy without getting out of breath and/or sweaty, 2) gets out of breath and/or sweaty, 3) gets almost exhausted [34]. The questions on frequency of leisure time MVPA and duration, i.e. weekly leisure time MVPA, have been examined for reliability and validity [37]. Intraclass correlation coefficients for reliability were 0.71 for frequency and 0.73 for duration. Validity against VO_2peak was acceptable with Spearman correlation coefficients (r_s) of 0.39 with frequency and 0.33 with duration, but low against total energy expenditure and physical activity level (PAL) measured by accelerometers [37]. We also assessed time spent on TV, video, electronical devices, DVD or PC outside school, with response options 1) less than ½ hour, 2) ½ to 1 h, 3) 2 to 3 h a day [34]. Finally, approximate hours of sleep at night on weekdays was reported (8 h or less/9 h/10 h/11 h/12 h or more) [31].

Mothers’ height in cm was self-reported at baseline. They reported their own weight in kg at the 7-year follow-up and their BMI was calculated (kg/m²). Questions regarding the mothers’ exercise were part of the Physical Activity and Pregnancy Questionnaire (PAPQ) [39], previously used in this study [25]. The mothers reported whether they met the recommendation from the Norwegian Directorate of Health valid at the time to perform 30 min of daily PA (yes/no) and if they currently exercised regularly, i.e. performed weekly physical activity to maintain or improve physical fitness (yes/no). If so, they reported the average frequency of exercise per week (1/2/3/4/5 days or more) and average duration of each exercise session with response options 1) 0–30 min, 2) 31–60 min, 3) 61–90 min and 4) > 90 min. Intensity of mother’s PA was reported as 1) a little strenuous, 2) strenuous or 3) very strenuous, and she reported minutes per day spent cycling/walking/running to and from work (none/20–30 min/31–60 min/> 60 min). The PAPQ has shown good correlation with PAL measured by accelerometers for time spent in high intensity activities (r_s = 0.59), moderate for time spent standing/moving (r_s = 0.36) and fair for sitting/lying (r_s = 0.29) [40].

At baseline (week 18–22 of pregnancy), the women’s age, height, pre-pregnancy weight, exercise, parity, birth weight of previous children and diabetes in near family (parents, siblings or children) were self-reported. Weight was measured, and pre-pregnancy BMI and baseline BMI was calculated. Socioeconomic status (SES) was calculated based on mother’s education and occupation, according to Hollingshead Two-Factor Index of Social Position [41]. Information about the children’s sex, birth weight, gestational age, length, head circumference, type of delivery and admittance to the neonatal intensive care unit (NICU) were retrieved from medical charts after birth.

In a power calculation prior to the follow-up study, we assumed that approximately 50% of the eligible children (i.e. n = 400) would attend the follow-up study. With a power (β) of 80% and significance level (α) of 0.05, we would be able to detect group differences in BMI of 0.4 kg/m² based on a standard deviation (SD) of 1.3 for 7-year-old Norwegian boys and girls [42].

Table 2 shows height, weight, iso-BMI, and PA of the children at 7 years. There were no differences in height or weight, and mean iso-BMI in both groups were 19.4 kg/m². There were no group differences in proportion of children with overweight. In total, 109 (66.5%) children in the intervention group and 68 (59.1%) children in the control group had at least 1 h per day of MVPA. There were no differences in frequency of leisure time MVPA, weekly leisure time MVPA or intensity of PA, daily use of electronical devices or hours of sleep on a weeknight (Table 2). Parents reported that 36 (22.4%) children in the intervention group and 19 (16.8%) children in the control group had a disease or health problem (p = 0.259). None had cancer, diabetes, cerebral palsy, mental retardation, Asperger’s syndrome, or chronic fatigue syndrome/myalgic encephalomyelitis.

Stratified by sex (Additional file 1: Tables S1a and S1b), a higher proportion of boys in the control group (n = 21; 40.4%) spent 2–3 h a day on electronical devices compared with 18 (18.8%) boys in the intervention group (p = 0.030). When we performed subgroup analysis including the children of women who adhered to the protocol (n = 93) compared with the control group, iso-BMI and distribution of PA were similar with no significant group differences (data not shown). When we excluded children born preterm (n = 10), children admitted to the NICU (n = 7) and children with diseases or health problems at the 7-year follow-up (n = 55), results were unchanged (data not shown).

There were no significant group differences in the mothers’ height, weight, BMI or exercise at follow-up (Table 3). In all, 114 (69.9%) women in the intervention group, and 90 (76.9%) women in the control group reported at least 30 min of daily PA, and 114 (69.9%) women in the intervention group and 76 (65.0%) women in the control group performed regular exercise.

When we performed subgroup analysis of women who adhered to the protocol, mothers in the intervention group had a lower mean weight (64.0 kg, SD8.0 vs. 67.1 kg, SD11.3; p = 0.026) and BMI (22.6 kg/m², SD2.7 vs. 23.7 kg/m², SD3.7; p = 0.019) compared with mothers in the control group.

Table 4 shows the correlation between children’s and mothers’ BMI and PA in the total sample. There was an association between children’s iso-BMI and mothers’ BMI (r = 0.254, p < 0.001). The children’s weekly leisure time MVPA correlated with the mothers’ daily PA (r_s = 0.129, p = 0.030) and the children's intensity of PA correlated with mothers’ intensity of exercise (r_s = 0.212, p = 0.003).

Stratified by group, children’s iso-BMI correlated with mothers’ BMI in both groups. In the intervention group, both daily MVPA and PA intensity of the children correlated with mothers’ intensity of exercise (Additional file 2: Table S2a). In the control group, both frequency of leisure time MVPA and PA intensity of the children correlated with mothers’ intensity of exercise (Additional file 2: Table S2b).

There was a lower proportion of non-participants in the intervention group (n = 265; 61.8%) than in the control group (n = 309; 72.5%) (p < 0.001). There were no differences between participants and non-participants in either group regarding maternal age, age > 40 years, height, weight, pre-pregnancy or baseline BMI, parity, birth weight of previous baby or diabetes in near family at baseline or children’s gestational age, birth weight, length, head circumference, sex, mode of delivery, proportion of preterm birth or older siblings (data not shown). In the intervention group, mean exercise sessions per week before pregnancy was slightly higher among mothers of participating children compared with non-participants (1.8, SD1.4 vs. 1.7, SD1.4, p = 0.048). In the control group, mean SES was 4.0 (SD0.8) among mothers of participating children compared with 3.8 (SD1.0) among non-participants (p = 0.004). Further, only one (0.9%) participating control child had been admitted to the NICU compared with 18 (5.9%) non-participating control children (p = 0.026).

This RCT follow-up study reports on anthropometry and PA of 7-year-old children whose mothers were randomised to moderate intensity exercise at least three times per week during pregnancy or standard antenatal care. As hypothesised, there were no group differences in BMI or PA of the children. Stratified analyses by sex, subgroup analyses including only children of women who adhered to the exercise protocol and sensitivity analyses excluding children born preterm, children admitted to the NICU and children with diseases or health problems at follow-up, gave essentially the same results. We found that children’s iso-BMI correlated with mothers’ BMI. Further, there was an association between the children's leisure time MVPA and intensity of PA with the mothers' daily PA and intensity of exercise.

A strength of the RCT was the large number of participants and the computerised randomisation procedure used to allocate the women. This entails that misclassification is unlikely to have affected the results. The included women exercised regularly and had baseline BMI within the normal range, indicating they were active and healthy women. Their baseline characteristics were comparable to the large Norwegian Mother and Child Cohort Study [25, 43], indicating a representative selection of Norwegian pregnant women. However, results should be interpreted with caution in pregnant populations with higher BMI, less physically active women and in ethnically diverse populations.

The present follow-up study was based on parent-report. Data was collected electronically using a software, which is a feasible, efficient and low-cost way of collecting data for larger groups of people, ensuring good data quality, and reducing the risk of data entry errors [44]. However, respondents might interpret questions differently or answers could be exaggerated, misremembered, or affected by social desirability bias. We cannot exclude the possibility that this, or other factors such as community or culture, may have influenced the results. Nevertheless, it is unlikely to have affected the groups differently. We do not know whether the parents estimated or measured the child’s weight and height. On an individual level, a Belgium study reported less accuracy when parents estimated their child’s weight and height, and accordingly BMI, than when measurements were performed at home by the parents of children aged 3–7 years [45]. However, on a group level there was no important differences between estimated and measured parent-reports [45]. Moreover, the BMI of the children in our study were similar to parent-reported values in the large national MoBa study with a mean BMI in 7-year old girls (n = 1839) and boys (n = 1932) of 15.9 kg/m² (SD2.0 and 1.8, respectively) [46]. The PA questions on frequency and duration have shown acceptable reliability and validity [37], and all questions are in line with questions used in other Norwegian and international studies [31, 34‐36, 38]. The PAPQ is considered an acceptable method for assessing habitual physical activity and exercise among women at group level [40]. In a priori power calculations we assumed that approximately 50% of the eligible children would attend the follow-up study. Thus, the low follow-up rate of 33%, limits our power to demonstrate differences and correlations, and non-significant findings should therefore be interpreted with caution. However, the mean values were highly similar between the groups and the non-significant correlations generally low, indicating that type II errors were less likely. Furthermore, the low follow-up rate may threaten the representativeness and limit the generalisability of results. A higher proportion of children in the intervention than in the control group attended the follow-up study, however, there were few baseline differences between participants and non-participants. Thus, we can assume that our results are representative for the sample initially included in this study.

Our study is one of few long-term follow-up studies of exercise during pregnancy. A small RCT of home-based stationary cycling or regular activity during pregnancy of 84 sedentary women reported no differences in height, weight, or BMI of the children at 1 or 7 years. However, children in the intervention group had increased body fat at 7 years, measured by dual-energy x-ray absorptiometry [21]. In three different case-control studies of 65 [18], 40 [19], and 104 [20] exercising women, children of women who continued to exercise in pregnancy weighed less and had less body fat than control children at birth [18] and at 5 years of age [19], suggesting that exercise during pregnancy reduced subcutaneous fat mass of the offspring. However, at 1 year, all anthropometric measurements were similar [20]. Limitations with these studies include several potential confounders and highly selected and relatively small groups of healthy, exercising women. A large population-based study of 5125 Greek children reported that retrospectively recalled PA during pregnancy were significantly associated with obesity in the offspring at 8 years of age [47]. However, this study included a large proportion of pre-pregnancy overweight and obese women, not necessarily representative for a healthy pregnant population. This may indicate that preconceptual health is more important than prenatal exercise or that other mechanisms of childhood obesity play a role. In the present RCT follow-up study, we found no evidence that children of mothers receiving regular exercise during pregnancy had a more unfavourable body composition than children of mothers receiving standard care. Consistent with our results, a meta-analysis of 135 studies (n = 166,094 women) from 32 countries revealed no association between prenatal exercise and neonatal outcomes or body composition in terms of body fat percentage, body weight and BMI in childhood [48]. Further, data from the large Danish National Birth Cohort Study, including 40,280 mother-child pairs, supports our findings that maternal exercise during pregnancy are not related to children’s BMI or risk of overweight [49].

Not meeting the recommendation of 1 h daily MVPA [6] is associated with an increased chance of overweight and obesity [50]. Parents reported that 55.6–60.3% of the girls and 63.5–70.8% of the boys in the two groups, respectively, met this recommendation. These proportions seem lower than what has been found by use of accelerometer measurements of PA in Norwegian children, where 87% of girls and 94% of boys met the recommendation at 6 years of age, but the proportions fell to 64 and 81% at 9 years of age [51]. In stratified analyses by sex, boys in the control group spent significantly more time on electronical devices than boys in the intervention group. This is probably a random finding. Even though longer screen viewing time in children has been associated with more sedentary behaviour, which might displace PA during early childhood [52], we have no reason to believe that it affected the children’s PA in this study. We found that daily MVPA, frequency of leisure time MVPA, weekly leisure time MVPA and intensity of PA were similar in the two groups.

The present results also showed that most women in both groups performed daily PA and exercised regularly 7 years after the intervention period. When we performed subgroup analyses of women adhering to the exercise protocol, their weight and BMI were lower compared with the control group, consistent with other studies suggesting that an exercise intervention during pregnancy has a positive effect on mothers in the long term [21, 53‐55]. However, the subgroup analysis did not change the children’s results.

The associations between children’s and mothers’ BMI and PA both in the total material and in each group, suggest that mother’s adherence to a healthy lifestyle could influence her child’s health. These findings are supported by a British study, reporting a direct association between PA levels measured by accelerometery in 554 mothers and their 4-year-old children [56]. Furthermore, a healthy lifestyle including regular exercise and a healthy BMI of the mother during their offspring’s childhood and adolescence has been associated with a substantially reduced risk of obesity in the children [22].

Our results add to the existing knowledge on exercise during pregnancy. Both in previous [10, 11] and the current follow-up study, we have shown that exercise during pregnancy does not adversely affect childhood outcomes. As pregnancy may be a time when the level of PA declines [43, 57], interventions for encouraging pregnant women to be physically active seem beneficial, given the adverse health effects of inactivity, overweight and excessive gestational weight gain [16, 43, 58‐60]. Thus, promoting a healthy lifestyle, both before, during and after pregnancy, may prevent chronic disease risk in more than one generation.