Study design
This paper focuses on the longitudinal analysis of data reported at 30 weeks pregnancy and 6 weeks, 6 and 12 months postpartum. The data collection was conducted via mailed questionnaires from June 2003 to November 2004. Written informed consent was obtained from every respondent. The Medical Ethics Committee of the VU (Vrije Universiteit) University Medical Center, Amsterdam, the Netherlands approved the study protocol.
Study sample
The Municipal Health Services of Amsterdam, which has a registry of almost all pregnant women in Amsterdam and the surrounding area, because they are responsible for blood screening in pregnancy, cooperated in the recruitment phase. First, 550 women of eighteen years or more were randomly selected by the Municipal Health Services to be invited to participate in our study. No upper age limit was used. We were allowed to send one mailed invitation, accompanied with some demographic questions, and one reminder invitation. There was no telephone contact during the recruitment phase for privacy reasons. Written informed consent was obtained from every respondent. After having given consent, reminder phone calls were made for follow-up questionnaires.
Behavioral covariates
Covariates were repeatedly assessed at 6 weeks, 6 and 12 months postpartum and mean postpartum scores of these measures were computed.
Physical activity (PA) was assessed by means of the SQUASH (the Short QUestionnaire to Assess Health-enhancing physical activity) a reliable and reasonably valid questionnaire, that can be used to rank adults according to their PA levels [
19]. Minutes per day and number of days per week spent on activities were assessed in four domains: commuting, work, household work, and leisure time activities. Initially, the mean number of minutes spent per day in the first year postpartum on light PA (<4 Metabolic Equivalents, METs) and on moderate or vigorous PA (≥4 METs) were derived from the questionnaire. Second whether or not participants met the Dutch PA guideline [
20], i.e. accumulating at least 30 minutes of moderate or vigorous physical activity on at least five days per week, was determined. The PA guideline was not deemed applicable six weeks after childbirth, since women are often recommended not resume their normal PA program before that time. The mean postpartum score for meeting the PA guideline was categorized as never/sometimes/always meeting the guideline at the three moments of measurement.
We also determined social comparison for PA to gain insight into how respondents perceived their individual PA level, comparing themselves with other women in the first year postpartum. At each measurement, women responded to the statement: "I think I am a lot less/a little less/equally/a little more/a lot more physically active than other women". The outer categories on both ends were combined with their neighboring category because they were hardly used. The resulting three categories (a little or a lot less/equally/a little or a lot more) were used in the analyses.
Time spent sleeping and time spent sitting and resting (lying down but not sleeping) were assessed. The following questions were used: "In the last week, how many hours did you sit/rest/sleep during a 24-hour day?" This was requested for week and weekend days separately, to reduce reporting bias owing to divergent activities on week and weekend days that are usual for working people. Finally, mean values over the three postpartum time points were computed for total hours per week spent on these behaviors.
The Dutch food frequency questionnaire was developed at the Division of Human Nutrition of the Wageningen University and validated to assess the intake of total energy, total fat and different types of fatty acids [
21,
22]. The original FFQ has been updated twice based on data of Dutch national food consumption surveys in 1992 and 1998 [
23,
24]. It is a 104-item questionnaire in which the women reported their food consumption of the previous four weeks. Frequency of foods used, as well as preparation methods, portion sizes and additions can be filled out. Using the Dutch food composition table of 2001 [
25], the following measures were derived from this questionnaire: (1) total energy intake (kJ/day), (2) percentage of total energy intake derived from fat intake, and (3) percentage of total energy intake derived from saturated fatty acid intake. Furthermore, we determined whether or not participants met the guidelines of the Dutch Health Council for total fat intake (<40% of total energy intake) and for saturated fatty acid intake (<10% of total energy intake) [
26]. Subsequently, mean scores of the three postpartum measurements were computed, illustrating whether participants always, sometimes or never met the guideline. Because the majority (75%) of participants always met the fat intake guideline, the "never" group (3%) was combined with the "sometimes" group (22%) in the analyses. With regard to saturated fatty acid intake, 37% never met the guideline, and the "sometimes" group (53%) was combined with the "always" group (9%).
Finally, in each questionnaire participants were asked to report on their breastfeeding practices. This variable was categorized into the following tertiles: never/up to four months/longer than four months. The cut point of four months postpartum was chosen since this time point is the time many women stop breastfeeding in the Netherlands.
Statistical analyses
The analyses were conducted with SPSS 12.0.2. Dummy variables of categorical variables were generated and included in the regression models. Univariate predictors were selected (p < 0.2) for both outcomes separately. Two multiple regression models were developed by means of the stepwise forward method, with a significance level of p < 0.05: (1) a linear model with the dependent measure total postpartum weight change, and (2) a logistic model with the dependent measure substantial postpartum weight retention of 5 kg or more. All analyses were adjusted for prepregnancy body mass index (BMI). Finally, as a measure of model fit, the adjusted R-square was determined for the linear model.