Physical activity patterns during pregnancy through postpartum

Interviewers administered a past-week recall on physical activity during telephone and in-home interviews, at pregnancy and postpartum, respectively. The physical activities captured included recreational, occupational, transportation, child and adult care-giving, indoor household, and outdoor household activity that increased breathing and heart rate. For each activity, the participant reported the number of sessions per week, duration of each session, and the perceived intensity level using the following options: 'fairly light,' 'somewhat hard,' and 'hard or very hard.' In addition, the activities were later assigned an absolute intensity level using published metabolic equivalent (MET) tables [20, 21], in which one MET corresponds to energy expended at rest per hour/kilogram [22]. The outcome variables were hours (h/wk) and MET-hours (MET h/wk) in the past week. The detailed measurement of activity is reported elsewhere [23]. In separate validation and reliability studies, the questionnaire has shown substantial agreement for test-retest reliability and from low to substantial agreement for criterion validity [23, 24].

Poisson regression models with generalized estimating equations (GEE) with compound symmetry working correlation [25, 26] for repeated count measures were applied to test whether the significant changes occurred in physical activity (total or by intensity levels) across the 4 time points. The goodness-of-fit statistics of all models indicated over dispersion; therefore the Pearson scaling adjustment was applied with compound symmetry as the working correlation structure. All GEE models were adjusted for potential confounders: age (16-25, 26-34, >= 35 years), race/ethnicity (Non-Hispanic white, Non-Hispanic black, other), education (< = 12, 13-15, >= 16 years), marital status (partnered or not), parity (0, 1, 2 or more), general health status (excellent or very good, good, fair or poor), and pre-pregnancy BMI (underweight or normal weight, overweight or obese). Further adjustment for bedrest during pregnancy did not meaningfully change the results; thus it was not included in the final models. The SAS statistical package was used (version 9.1, SAS Institute Inc., Cary, NC).

During the study period from 2001 to 2005, among the 3203 women eligible for the study, 2006 were successfully recruited and provided informed consent. Of the 2006 women who were enrolled for the PIN3 Study, 37 had multiple birth or lost pregnancy, 159 requested to drop out or no future contact, 84 did not complete the first PIN3 interview, 87 did not deliver at UNC Hospitals or moved out of the area, 8 had medical problems at delivery, and 462 delivered before the Postpartum Study began. Thus, 1169 women were eligible for the postpartum recruitment. A total of 231 were then excluded for the following reasons: 24 had medical constraints, 153 were unreachable, and 54 were greater than 5 months postpartum when we contacted them. Of the 938 women who were asked to participate, 688 (73.3%) agreed to participate and completed a three-month home interview. There were no significant differences (p < 0.05) between the women who completed the 3-month interview and those that were excluded or refused, accounting for women enrolling more than once, for any of the socio-demographic characteristics (age, race/ethnicity, marital status, education), as well as pregravid BMI, parity, bedrest, general health, and total physical activity (in both h/wk and MET h/wk).

Five hundred and fifty (79.9%) of the participants completing the 3 month postpartum visit went on to complete the 12-month component. Participants who became pregnant between the two postpartum time points (n = 45) and those who moved out of the recruitment area (n = 73) were not eligible for the 12-month home visit, thus explaining most of the attrition. The remaining 20 women were excluded for medical complications or were lost to follow-up. After excluding women who participated in the PIN3 Study for a second or third time (n = 64) or had missing information on physical activity variables (n = 15), the final sample comprised 471 women that completed interviews at all 4 time points.

Approximately three-fourths (77.6%) of the women in our sample were non-Hispanic white and 84.1% were married or partnered (Table 1). For about half of the women (54.6%), this was their first pregnancy. The mean education levels were relatively high (16.1 years), as was perceived general health. In the sample, about one-quarter (28.5%) were classified as overweight or obese. Few women reported being advised to stay on bed rest and most women reported at least some moderate to vigorous physical activity.

The total h/wk and MET h/wk of physical activity decreased from 17-22 weeks to 27-30 weeks of gestation, but rebounded up at 3 months postpartum and remained stable at 12 months postpartum among this cohort of women (Table 2). The mean MET h/wk values of total activity levels for each time point were 24.7 (standard deviation, SD 26.8), 19.1 (SD 18.9), 25.7 (SD 29.3), and 26.7 (SD 31.5) and corresponding median values were 17.2 (interquartile range, IQR 8.1-31.9), 14.3 (IQR 6.5-27.0), 15.9 (IQR 6.0-34.6), and 18.3 (IQR 6.3-33.3). A similar pattern of decreasing total activity during pregnancy and rebounding back at postpartum was observed for somewhat hard intensity activities. The hard or very hard intensity activities remained stable during pregnancy, but increased to 3 months postpartum and remained unchanged thereafter. These observed changes were statistically significant, even when adjusting for age, race/ethnicity, education, marital status, parity, general health status, and pre-pregnancy BMI. A large part of total physical activity consisted of fairly light physical activity, which remained stable across the 4 time points, with mean h/wk values of 3.9 (SD 5.0), 3.4 (SD 4.5), 3.8 (SD 5.9), and 3.4 (SD 5.3).

The percentage of women reporting care-giving, outdoor household, recreation, transportation, and occupational activity decreased from the 17-22 weeks to 27-30 weeks of gestation, with an exception of increased percentage in household-related indoor activity (Figure 1). Participation in any household-related indoor activity decreased from the pregnancy level of 64% at the 27-30 weeks' gestation to 47% at 12 months postpartum. In postpartum, the largest increase from pregnancy levels was reported for care-giving activity, from 38% at the 17-22 weeks' gestation to 51% at 3 months postpartum and to 58% at 12 months postpartum. The highest and most stable participation was reported for recreational physical activity, with around 67% participating during all 4 measurement periods.

The separate activity modes that contributed the most to the total amount of physical activity in h/wk were care-giving and recreational activities (Figure 2). Care-giving physical activity constituted 26% of total activity level at the 17-22 weeks' gestation, decreased to 22% at the 27-30 weeks' gestation, but increased thereafter to 30% at 3 months postpartum and to 34% at 12 months postpartum. Recreational physical activity contributed around 31% to total activity and remained fairly stable from pregnancy through postpartum.

One study [10] measured overall physical activity from pregnancy to postpartum, but included only 25 women and did not report activity across intensity levels. Those results suggested, in line with our findings, decreased activity during pregnancy and a rebound at one-year postpartum. They also estimated that at one-year postpartum, 34% of daily energy expenditure was spent on sleeping or lying on a sofa, 35% on sitting, 28% on light household and care, and only 3% on walking, cycling and other sport activities. They did not report the change in the proportions of modality during pregnancy through postpartum. Another study [13] among 63 women reported no change in total activity and measured activity only at pre-pregnancy and 2 and 7 months postpartum, making comparisons to our study difficult.

The percentage of women reporting any activity was the highest and most stable over time for any recreational physical activity, with values of 70%, 62%, 69%, and 68% for the 4 time points. Previous studies [14, 15, 27‐29] have not included comparisons of different modes of activity. Nevertheless, walking is the most commonly reported form of recreational activity during pregnancy [6, 14‐16]. Two prospective studies have reported decreased percentages in exercise or sports from 63% to 39% [27] and from 72% to 37% [28] from pre-pregnancy to pregnancy. One retrospective study reported a change in the percentages from 48% to 42% [29] for exercise from pre-pregnancy to pregnancy and another study [15] reported fairly stable percentages of 81%, 76%, 73%, and 66% for recreational activity and 70%, 65%, 57%, and 49% for structured exercise from pre-pregnancy through all trimesters. One cross-sectional study [14] using a representative sample of the U.S. population reported that 66% of pregnant women engaged in any leisure time physical activity.

Several limitations should be acknowledged. Physical activity measurement in our study was based on self-reports via interviews, which is not as accurate a measurement method as objective methods, such as accelerometers [30]. However, it allowed for the characterization of the types of physical activities, not yet easily detected with accelerometry. Self-reported physical activity measurements can be prone to measurement error and misclassification, which most often incur bias towards the null. To minimize this bias, a seven day recall was selected instead of a longer recall time period and the assessment was done by a trained interviewer instead of using self-administration [22]. The recall was developed for this study and the test-retest reliability and validity were tested in pregnant women [23, 24], but has not been confirmed in postpartum women. The physical activity questionnaire was conducted by telephone during pregnancy and at in-home visits during postpartum, which may have resulted in bias due to the different interview contexts. Furthermore, this physical activity questionnaire was not administered before pregnancy. Thus we cannot speculate whether the women returned to their earlier activity habits after the baby was born.

A commonly reported concern in clinic-based pregnancy cohort studies is the loss to follow-up [31, 32]. The preceding PIN1 and PIN2 studies reported that underrepresented women enrolled during early pregnancy were most often lower educated, younger, African American, had higher parity, and had a higher pregnancy risk profile [31, 32]. For the PIN3 study, we were unable to characterize differences between those who initially enrolled in pregnancy and those who did not. However, we found no meaningful differences between the women who completed the 3-month interview and those who were excluded or refused to complete the 3-month interview. Even so, we cannot rule out the potential for selection bias and reduced generalizability.