Changes in leisure-time physical activity among Brazilian pregnant women: comparison between two birth cohort studies (2004 – 2015)

The present study analyzed data from two population-based birth cohort studies carried out in the city of Pelotas, Southern Brazil, in 2004 and 2015. Strategies to recruit participants were identical in both surveys. All five maternity hospitals (attending patients from both private and public insurance) located in Pelotas were visited daily from 1 January to 31 December of each year, and all births of mothers living in the urban areas of the city were identified (99% of deliveries are performed at hospitals). A total of 4244 and 4271 mothers were interviewed after delivery in 2004 and 2015, respectively. The non-response rate at recruitment was below 1.5% in both studies. Face-to-face interviews took place in the hospital within 24 h after the delivery. Trained interviewers collected information on mother-child health using a structured questionnaire. Interviews lasted 60 min, on average. A fieldwork supervisor repeated 10% of the interviews to check the quality of the information collected. Further methodological details of the 2004 study are available elsewhere [13, 18].

Type, frequency and average duration of sessions of LTPA in a typical week (7-day recall) were investigated in four time periods during the perinatal interview: the three-month period prior to pregnancy as well as the first, second and third trimesters of pregnancy. Up to three different physical activities were recorded for each period. Women were asked not to report commuting, household or occupational activities as LTPA. The instrument used to assess LTPA in 2015 was the same employed in 2004 [13]. The total LTPA score was generated by the sum of minutes per week spent on each activity. A cut-off point of 150 min per week was used to classify women as active or not in each period. We also explored trends in any LTPA practice by comparing the proportion of women who were engaged in LTPA regardless of weekly amount.

Maternal covariates assessed in the studies and used in the comparison analysis were maternal age, schooling, parity, household income, pre-pregnancy BMI, and pre-pregnancy LTPA. Maternal age was collected as a continuous variable and divided into four categories (13–19, 20–29, 30–39 and 40–47 years). Schooling was assessed as maternal years of formal education and categorized into four categories (0–4, 5–8, 9–11, 12 or more years). Family monthly income was assessed as the sum of incomes of household members in the past month and categorized into quintiles. Parity was categorized into three categories (1/2/≥3) according to the total number of live births, including the birth from the cohort. Pre-pregnancy BMI was categorized according to the World Health Organization criteria into underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²); overweigh (25.0–29.9 kg/m²) and obese (≥30 kg/m²) based on self-reported height and weight. Pre-pregnancy LTPA (≥150 min/week) was also considered as a covariate. Independent variables were identically collected in both surveys for comparability.

Statistical analyses were carried out using Stata version 13.0 (StataCorp, College Station, TX, USA). Data analysis initially included the comparison of study populations in terms of sociodemographic, behavioral and health characteristics. Percentages and confidence intervals were used to compare the data between the studies. The prevalence of LTPA was estimated for each study, and its changes in the period were evaluated, including comparisons according to subgroups of the independent variables. The prevalence of each type of LTPA practiced was also compared between the studies. Chi-square test for difference in proportions across the study period was used. To identify correlates of LTPA in each period, the proportion of active women was described according to the subgroups of the independent variables and confidence intervals were calculated. In each year, logistic regression models were used to provide estimates for the adjusted odds ratios within each subgroup (with adjustment for the other variables studied). Associations between LTPA during pregnancy (outcome) and maternal age, income, schooling and parity were not adjusted for pre-pregnancy variables (BMI and LTPA) because they were considered mediators in the causal chain of its determination. Statistical significance was set at p < 0.05. We excluded from our analysis a group of 142 pregnant women from the 2015 Pelotas (Brazil) Birth Cohort Study who were randomly enrolled in the intervention group of a randomized controlled trial nested in the cohort to study the effects of an exercise program during pregnancy on mother and child health outcomes [19].

Data from a total of 4244 mothers in 2004 and 4129 mothers in 2015 were analysed in this study. Table 1 presents the mothers’ sociodemographic, behavior and health-related characteristics in both cohorts. The proportion of adolescent pregnancies decreased from 19.0% (95%CI 17.9–20.2) in 2004 to 15.0% (14.0–16.1) in 2015, while the proportion of mothers aged 30 to 39 years increased from 28.0% (95%CI 26.7–29.3) to 34.7% (95%CI 33.3–36.2). We observed an increase in the proportion of women with 12 or more years of formal education (9.9 to 30.0%). Employment during pregnancy increased from 40.1% (95%CI 38.6–41.5) to 55.5% (95%CI 58.5–61.4) during the period. A decrease in the proportion of women who reported being underweight before pregnancy from 7.4% (95%CI 6.5–8.4) in 2004 to 3.9% (95%CI 3.3–4.5) in 2015, accompanied by an increase in the proportion of overweight (20.1 to 27.4%) and obese (8.9 to 20.6%) was observed. Smoking during pregnancy significantly decreased between 2004 and 2015 (27.6 to 17.0%) while gestational diabetes nearly tripled (3.0 to 8.6%). Physical activity advice received from health professionals during prenatal care significantly decreased from 72.2% (95%CI 70.8–73.5) to 61.3% (95%CI 59.8–62.8) from 2004 to 2015. Over the period there was an increase in the median of family income from $206 in 2004 to $649 in 2015 (data not shown).

Changes in LTPA patterns are presented in Fig. 1. There was an increase in the proportion of women practicing any pre-pregnancy LTPA between 2004 and 2015, from 15.3% (95%CI 14.2–16.3) to 21.3% (95%CI 20.1–22.6); as well as in the proportion of those engaged in recommended levels of LTPA, from 11.2% (95%CI 10.0–12.2) to 15.8% (95%CI 14.6–16.9). Regarding pregnancy LTPA patterns, no changes were observed for the first and second trimesters in the 11-year period. However, we observed a significant decrease in the prevalence of any LTPA from 6.6% (95%CI 5.9–7.5) to 3.4% (95%CI 2.9–4.0) as well as in recommended LTPA during the third trimester of pregnancy, declining from 5.0% (95%CI 4.3–5.7) to 2.4% (95%CI 1.9–2.8). In both cohort studies, the prevalence of LTPA (any or recommended) markedly declined from pre-pregnancy to the third gestational trimester (p < 0.001).

Among women engaged in LTPA during pregnancy, walking was the most reported type of LTPA in both surveys (Fig. 2). The second most commonly reported LTPA was cycling and weight training in 2004 and 2015, respectively. Regarding changes over time, reductions in walking (from 77.2% in 2004 to 47.4% in 2015) and cycling (from 8% in 2004 to 3.1% in 2015) and increases in weight training (from 6.1% in 2004 to 21% in 2015), water gymnastics (from 3.9% in 2004 to 9.7% in 2015), aerobics (from 2.7% in 2004 to 5.7% in 2015) and dancing (from 2% in 2004 to 4.9% in 2015) were observed.

Tables 2 and 3 describe detailed time changes in the prevalence of recommended and any LTPA by maternal age, income, schooling, parity, pre-pregnancy BMI and pre-pregnancy LTPA from 2004 to 2015. The proportion of women considered active increased in the pre-pregnancy period among all groups of maternal age, income (except for the richest) and parity, from 2004 to 2015. Marked increases were also observed among those mothers with 5 to 8 years of formal education and those classified as underweight before the pregnancy. During the first and second trimesters of pregnancy, despite the overall stability in prevalence, decreases in recommended LTPA were observed for some subgroups of women. During the first gestational trimester, LTPA declined between 2004 and 2015 among women with 9 to 11 years of schooling and among those considered active before pregnancy. Decreases in the prevalence of recommended levels of LTPA in the second trimester of pregnancy were observed among mothers aged 20 to 29 years, classified in the intermediate quintile of family income, with at least 9 years of education, who were giving birth to the first child, obese and, among those engaged in recommended levels of LTPA before pregnancy. Declining changes for the third gestational trimester followed the same patterns observed for the second trimester of pregnancy, except that LTPA also declined among mothers belonging to the fourth quintile of family income and among those with a normal pre-pregnancy BMI.

The proportion of women engaged in any LTPA before pregnancy also increased from 2004 to 2015 and among all subgroups of maternal age, income, schooling, parity and, pre-pregnancy LTPA (Table 3). In terms of pre-pregnancy BMI, marked increases in the prevalence of any LTPA were observed only among mothers classified as normal according to their pre-pregnancy BMI. During pregnancy the same patterns described for recommended LTPA were observed for any LTPA, except that in the third trimester of pregnancy it also decreased among adolescent mothers and those with intermediate schooling.

Adjusted associations between recommended LTPA and the independent variables in 2004 and 2015 are presented in Table 4. During the pre-pregnancy period, mothers aged 30 to 39 years were more likely to reach recommended LTPA levels when compared to adolescent mothers in 2004 (OR 1.69; 95%CI 1.17–2.44) but no association between maternal age and pre-pregnancy LTPA was found in 2015. In both 2004 and 2015 studies, income and schooling were positively associated with pre-pregnancy recommended LTPA while parity was negatively associated. Pre-pregnancy LTPA was the strongest correlate of LTPA during pregnancy in all gestational trimesters and in both surveys. However, the magnitude of associations was smaller in 2015. During pregnancy, maternal age was associated with LTPA in the first trimester of pregnancy in 2004; with mothers aged 30 to 39 years being more likely to be engaged in recommended LTPA. In 2015, all confidence intervals included the null value. A significant positive association between recommended LTPA and income was identified only for the first trimester in the 2004 cohort. A positive association between recommended LTPA and schooling was observed in all trimesters in 2004, but in 2015 it was only observed for the first trimester. Parity was negatively associated with LTPA in all trimesters of pregnancy in both studies, except in the third trimester in 2015.

In terms of any LTPA, similar patterns of associations were observed for all independent variables (Table 5). However, positive associations with income were observed in all pregnancy trimesters in 2015 and negative associations with parity were observed in all periods for both studies. Besides, negative associations between pre-pregnancy BMI and any LTPA were observed for all pregnancy trimesters in 2015.

The similarity of data collection methods between both surveys combined with the high response rates, are certainly the major strengths of the present study. The use of the same inclusion criteria and methodologies over an 11-year period provides reliable findings. Besides, LTPA patterns were measured during the different trimesters of pregnancy allowing the distinction of specific changes that happened across gestational trimesters.

Some limitations of the study need to be taken into consideration while interpreting the findings. First, instruments based on self-report such as the one used in this study, might overestimate physical activity levels. However, as this limitation was equally present in both surveys, comparability over time was not impaired. Second, the intensity of the reported leisure-time physical activities was not measured and for this reason we could not assess its changes over time. Yet, the lack of information on intensity was purposeful given that intensity prompts (amount of moderate or vigorous-intensity activities) would not be ideal in a retrospective analysis—women might remember how much they practiced some months ago, but probably not the intensity of the activities performed. Third, the retrospective assessment of LTPA could result in recall bias. But in the worst scenario mothers had to report their LTPA patterns 9 months before the interview (pre-pregnancy LTPA), which is considered a reasonable period since long-term maternal recall of pregnancy-related events has been suggested to be highly accurate [45]. Lastly, it should also be noted that the prevalence estimates reported in the present study might have being influenced by the fact that we excluded 141 mothers from the 2015 Birth Cohort Study who took part in an exercise intervention during pregnancy and who were not reaching recommended levels of LTPA during recruitment (inclusion criteria in the trial). Nevertheless, given the low probability of previously inactive women starting to practice physical activity during the pregnancy period [46] we expect a slight overestimation in the observed prevalence of LTPA during pregnancy in 2015. Consequently, the decline trends observed could be worse if including this sample of women.