Objective
Document weight change trajectories that lead to gestational weight gain or postpartum weight loss outside clinical recommendations established by the Institute of Medicine.
Study Design
Women aged 14-25 receiving prenatal care and delivering singleton infants at term (n = 427). Medical record review and 4 structured interviews conducted: second and third trimester, 6- and 12-months postpartum. Longitudinal mixed modeling to evaluate weight change trajectories.
Results
Only 22% of participants gained gestational weight within Institute of Medicine guidelines. There were 62% that exceeded maximum recommendations—more common among those overweight/obese (body mass index ≥25.0; P < .0001). 52% retained ≥10 lb 1-year postpartum. Increased weight gain and retention documented among smokers and women with pregnancy-induced hypertension; breastfeeding promoted postpartum weight loss (all P < .02). Body mass index by race interaction suggested healthier outcomes for Latinas ( P = .02).
Conclusion
Excessive pregnancy weight gain and inadequate postpartum weight loss are highly prevalent among young low-income ethnic minority women. Pregnancy and postpartum are critical junctures for weight management interventions.
Inappropriate patterns of gestational weight gain and postpartum weight loss have important public health implications. Inadequate gestational weight gain is primarily associated with intrauterine growth restriction (IUGR) and preterm delivery. Excessive maternal weight gain is associated with cesarean delivery, maternal anemia, preterm labor, macrosomia, and traumatic birth injury. Postpartum weight retention can contribute to maternal long-term obesity and associated adverse health sequellae of cardiovascular disease, hypertension, diabetes and degenerative joint disease, among others. Moreover, excessive gestational weight gain has been associated with higher body mass index (BMI) and blood pressure in children and young adults. Several population-based cohort studies have documented that only 30-40% of women meet recommended weight gain guidelines during pregnancy.
In 2009, the Institute of Medicine (IOM) published new target gestational weight gain recommendations to address gestational weight gain based on prepregnancy BMI. Briefly, although women who begin pregnancy as “underweight” (BMI <18.5 kg/m 2 ) are still encouraged to gain between 28 and 40 lbs during pregnancy to avoid serious complications including IUGR, women who are overweight (BMI 25.0-29.9 kg/m 2 ) should restrict their weight gain to between 15 and 25 lbs and obese women (BMI ≥30 kg/m2) should be further encouraged to gain no more than 20 lbs over the 40 weeks of gestation. The IOM panel responded to data noting that, by 2009, the distribution of BMI among women of childbearing age had substantially increased compared with what was measured in the 1980s—the benchmark used for the prior 1990 IOM recommendations, such that a greater proportion of women are entering pregnancy overweight or obese, increasing health risks for both mother and child. IOM also explicitly stated the need for research to identify factors that affect weight gain and weight retention in diverse populations. To date, few studies have focused on young minority women, and data examining gestational weight gain and retention in light of the new IOM guidelines are limited.
Epidemiologic investigations that have evaluated determinants of excessive gestational weight gain suggest that this is more likely among African American women, women with hypertension, and among those who ceased smoking coincident with or immediately preceding the index pregnancy. Postpartum weight retention has been most strongly associated with increased gestational weight gain. Other factors such as exercise frequency, dietary choices, smoking status, educational attainment, and bottle feeding have been inconsistently associated with postpartum weight retention.
The purpose of this study was to document the trajectory of weight change from prepregnancy baseline through pregnancy, delivery, and the first year postpartum among a cohort of young, predominantly African American and Hispanic women. The application of longitudinal modeling techniques permits inclusion of correlation structures across repeated measures that reduce standard errors and increase the power to detect the effect of predictors—a substantial advantage over traditional cross-sectional data analysis. Controlling for relevant demographic, physiologic, and behavioral variables, we examined the role of prepregnancy BMI on trajectories of weight change that lead to either gestational weight gain or postpartum weight loss outside clinical recommendations recently established by the IOM.
Materials and Methods
Patient selection and data collection
Data are from a randomized clinical trial that enrolled pregnant women ages 14-25 at the largest publicly-funded university-affiliated hospital clinics in New Haven, CT, and Atlanta, GA, (2001-2004). The trial was aimed at promoting improved reproductive outcomes through group prenatal care, and complete details of the clinical protocol have been published elsewhere. Inclusion criteria consisted of: (1) pregnant between 16-24 weeks’ gestation; (2) no severe medical problems at time of enrollment that would require individualized assessment and tracking as a “high-risk pregnancy”; and (3) conversant in either English or Spanish.
Participants completed structured interviews via “audio computer self-interviewing” (A-CASI) on laptop computers assisted by trained study staff. Interviews were conducted during the second (average gestational age = 18 weeks, standard deviation [SD] = 3.3) and third (average gestational age = 35 weeks, SD = 3.1) trimesters of pregnancy as well as 6 and 12 months postpartum. Labor, delivery, and birth outcomes were obtained through medical chart review available for 95% of participants (990/1047). There were no systematic differences between those who were retained vs those lost to follow-up. The study was approved by the Institutional Review Boards at both study sites, and all patients provided informed written consent before randomization. This analysis is restricted to participants enrolled in New Haven who delivered singleton infants at term (≥37 weeks’ gestation) and had BMI data available by medical record review (n = 427). Comprehensive medical record review for the Atlanta-based participants was not feasible; consequently, this subpopulation was excluded from this secondary analysis.
Definition of variables
Outcome variable
Change in weight was calculated as the difference in weight from prepregnancy at each study time point. Usual prepregnancy weight and current weight during the second and third trimesters and at 6 and 12 months postpartum was obtained by self-report at each study interview. For participants whose last prenatal care visit occurred within 2 weeks of delivery, maximal predelivery weight was the measured weight using a standard triple-beam balance at that visit. For participants whose last obstetric visit was greater than 2 weeks before delivery, maximal predelivery weight was that reported at a postpartum interview. Early postpartum weight was obtained from the medical record for women who returned to the clinic between 5 and 9 weeks postpartum.
Measurement of time
To account for variability in the timing of delivery and interviews, time was designated as a continuous variable measured in weeks since last menstrual period.
Time-invariant covariates: demographic and baseline clinical factors
Demographic factors were obtained by self-report at the baseline interview. Race was categorized as African American/black, Latina/Hispanic, other. As many participants were still of school age, a single variable that combined educational and employment status was created, categorizing participants as: both in school and working, in school only, working full-time only, working part-time only, and neither in school nor working. Gravidity and parity were obtained by self-report at the intake visit. Prepregnancy lifestyle covariates of tobacco and alcohol use were dichotomized into never- and ever-users. Study group referred to the study condition randomly assigned for each participant. Baseline BMI (kg/m 2 ) was calculated from self-reported prepregnancy height and weight. Incidence of gestational diabetes, hypertension, and cesarean delivery were extracted from obstetric records. Total number of weeks of breastfeeding was determined from self-report 6 and 12 months postpartum.
Time-varying covariates: clinical and psychosocial factors
Time-varying covariates are collected serially at each interview and change with the progression of time. Lifestyle variables included use/nonuse of tobacco, alcohol, marijuana, and other illicit drugs during the prior interval. A modified version of the Pregnancy-Relevant Health Behaviors Scale (PRHBS) that included all items on the exercise, nutrition, and sleep subscales was used to evaluate the scope of health-promoting behaviors pursued. Use/nonuse of hormonal contraception (oral contraceptive pills, depot progesterone injections, transdermal patch) was assessed at each postpartum visit. Repeat pregnancy described whether the participant was newly pregnant at either the 6- or 12-month follow-up visit. Psychosocial factors were measured using validated and reliable scales: perceived stress (Perceived Stress Scale [PSS]), depression (Centers for Epidemiologic Study-Depression [CES-D]), and self-esteem (Rosenberg Self-Esteem Scale [RSES]).
Statistical analyses
Descriptive statistics
The entire study population was subdivided according to prepregnancy BMI category, using the recently published IOM-defined cut points. Within each BMI category, the number of participants gaining below, within, or above the IOM guidelines as well as the number of participants returning to within 10 lbs (4.5 kg) of their prepregnancy weight by 1-year postpartum was tabulated. The distributions and associations between each fixed and time-varying covariate and baseline BMI category were evaluated. Exposures occurring in <5% of the study population were dropped from further analysis.
Longitudinal data modeling
Weight change over time was modeled using linear mixed models that included random variables for intercept (ie, subject) and slope (ie, time) that accounted for all higher orders of the time parameter with an unstructured correlation between the elements in the variance/covariance G matrix. As these modeling techniques are robust to unbalanced data, all available data were included. The underweight group (n = 23) were excluded from the longitudinal modeling secondary to their small number.
First, the best relationship between weight change and time was established by beginning with the linear model and then adding successively higher-order terms until the last term added had P > .05. Next, bivariate relationships between each fixed or time-varying covariate and the outcome variable of weight change were evaluated in a mixed model that contained all significant parameters for time as well as random effects for individual and time. Interaction terms were considered in a bivariate model that included the interaction between the selected parameter and each polynomial term for time in addition to the main effects for both the selected parameter and the time polynomial expression. Multivariable mixed modeling was executed using several model selection strategies in parallel. Backward elimination was initiated with the full model that included the time polynomial, main effects for all fixed and time-varying covariates, any significant bivariate interactions with time, and the random variables described previously. All covariates with type 3 likelihood ratio P > .10 were eliminated. Then, model parsimony was achieved through minimization of the Akaike Information Criterion (AIC). In parallel, a forward selection strategy was performed where, beginning with a model that included the time polynomial, variables were added in order of their significance on bivariate modeling. Variables were initially retained if P ≤ .15 but required P < .05 to remain in the model following evaluation of all eligible variables. Interactions between the remaining covariates were evaluated with those significant at P < .05 retained. All analyses were performed using SAS version 9.1.3 (SAS Institute, Cary, NC).
Results
Description of study participants
We included 427 of the 503 women originally enrolled at the New Haven site in these analyses. Six women were excluded for multiple gestations, 46 women were excluded for delivering preterm infants, 17 were missing birth outcomes data, and 7 did not have a calculable baseline BMI. Per inclusion criterion, they ranged in age from 14-25 years. Sixty percent were black, 25% Hispanic/Latina, 12% white, and 3% other racial/ethnic categories; there is no difference in racial distribution by BMI category.
The validity of self-reported weights was confirmed via comparison of actual weights recorded in medical records during the clinical visit occurring closest in time, with Pearson correlation coefficients ranging from 0.96 to 0.99 and corresponding linear regression coefficients ranging about ± 0.95 for all study time points. A paired t test conducted for the 12-month postpartum timepoint revealed a mean difference of −2.66 ( t = −4.01; P < .001) indicating that, on average, self-reported weights were only underreported by 2.66 lbs compared with the medical record. No interaction with BMI was noted.
Participant demographics stratified by prepregnancy BMI were calculated for both fixed and time-varying covariates ( Table 1 ). Overweight and obese women had greater parity ( P = .02), were more likely to have had a cesarean delivery ( P = .02), and less likely to use depot progresterone 6 and 12 months postpartum ( P ≤ .05). There was no difference in prepregnancy classification of BMI on the basis of race/ethnicity, age, employment/education, relationship status, intervention condition, substance use, breastfeeding interval, or other postpartum contraception choices (all P > .05). Our sample also did not reveal an association between baseline BMI and subsequent development of either gestational diabetes ( P = .75) or pregnancy-induced hypertension ( P = .14), both shown by others to be associated with baseline overweight or obesity. Although we observed relatively few incident cases of gestational diabetes (n = 18) or pregnancy-induced hypertension (n = 25), these were proportionately distributed across the selected BMI categories, as reflected by their nonsignificant Fisher’s exact test P value. Finally, there was no significant difference in the time-varying behavioral metrics, including prenatal health behaviors, prenatal stress, or depression captured at any of the follow-up time points by BMI category.
| Variable | BMI <18.5 (n = 23) | BMI 18.5-24.9 (n = 184) | BMI 25.0-29.9 (n = 100) | BMI ≥30.0 (n = 120) | P value |
|---|---|---|---|---|---|
| Fixed covariates | |||||
| Race | |||||
| African American/Black | 13 (56.5%) a | 112 (60.9%) | 58 (58.0%) | 68 (56.7%) | .93 |
| Latina/Hispanic | 6 (26.1%) | 45 (24.5%) | 28 (28.0%) | 32 (26.7%) | |
| White | 3 (13.0%) | 24 (13.0%) | 11 (11.0%) | 19 (15.8%) | |
| Other | 1 (4.4%) | 3 (1.6%) | 3 (3.0%) | 1 (0.8%) | |
| Age, y | 20.48 ± 2.37 | 20.60 ± 2.79 | 20.99 ± 2.61 | 21.11 ± 2.47 | .32 |
| Employment/education status | |||||
| Currently in school and working | 3 (13.0%) | 12 (6.5%) | 13 (13.0%) | 13 (10.8%) | .39 |
| Not in school, working full time | 1 (4.4%) | 26 (14.1%) | 9 (9.0%) | 12 (10.0%) | |
| Not in school, working part time | 7 (30.4%) | 30 (16.3%) | 15 (15.0%) | 13 (10.8%) | |
| Currently in school, not working | 4 (17.4%) | 41 (22.3%) | 20 (20.0%) | 28 (23.3%) | |
| Neither in school nor working | 8 (34.8%) | 75 (40.8%) | 43 (43.0%) | 54 (45.0%) | |
| Relationship status | |||||
| Single, never married | 11 (47.8%) | 79 (42.9%) | 53 (53.0%) | 55 (45.8%) | .52 |
| Unmarried, living with partner | 9 (39.1%) | 79 (42.9%) | 35 (35.0%) | 48 (40.0%) | |
| Married | 3 (13.0%) | 24 (13.0%) | 11 (11.0%) | 11 (9.2%) | |
| Separated | 0 (0.0%) | 1 (0.5%) | 1 (1.0%) | 5 (4.2%) | |
| Widowed | 0 (0.0%) | 1 (0.5%) | 0 (0.0%) | 1 (0.8%) | |
| Study condition | |||||
| Individual/standard prenatal care | 13 (56.5%) | 74 (40.2%) | 36 (36.0%) | 49 (40.8%) | .54 |
| Centering pregnancy | 3 (13.0%) | 55 (29.9%) | 35 (35.0%) | 36 (30.0%) | |
| Centering pregnancy + HIV | 7 (30.4%) | 55 (29.9%) | 29 (29.0%) | 35 (29.2%) | |
| Parity | |||||
| Nulliparous | 19 (82.6%) | 130 (70.7%) | 60 (60.0%) | 64 (53.3%) | .02* |
| 1 previous birth | 3 (13.0%) | 38 (20.7%) | 24 (24.0%) | 33 (27.5%) | |
| 2+ previous births | 1 (4.4%) | 16 (8.7%) | 16 (16.0%) | 23 (19.2%) | |
| Total weeks breastfed this infant | 4.33 ± 5.82 | 3.09 ± 5.00 | 2.89 ± 4.58 | 3.26 ± 5.48 | .70 |
| Smoked cigarettes in the year before the index pregnancy | |||||
| No | 14 (60.9%) | 96 (52.2%) | 57 (57.0%) | 65 (54.2%) | .79 |
| Yes | 9 (39.1%) | 88 (47.8%) | 43 (43.0%) | 55 (45.8%) | |
| Drank alcohol in the year before the index pregnancy | |||||
| No | 12 (52.2%) | 91 (49.5%) | 53 (53.0%) | 72 (60.0%) | .35 |
| Yes | 11 (47.8%) | 93 (50.5%) | 47 (47.0%) | 48 (40.0%) | |
| Gestational diabetes with index pregnancy | |||||
| No | 21 (91.3%) | 177 (96.2%) | 96 (96.0%) | 115 (95.8%) | .75 |
| Yes | 2 (8.7%) | 7 (3.8%) | 4 (4.0%) | 5 (4.2%) | |
| Pregnancy-induced hypertension with index pregnancy | |||||
| No | 20 (87.0%) | 177 (96.2%) | 91 (91.0%) | 114 (95.0%) | .14 |
| Yes | 3 (13.0%) | 7 (3.8%) | 9 (9.0%) | 6 (5.0%) | |
| Type of delivery | |||||
| Vaginal delivery | 21 (91.3%) | 156 (86.2%) | 87 (87.0%) | 89 (74.2%) | .02 b |
| Cesarean section | 2 (8.7%) | 25 (13.8%) | 13 (13.0%) | 31 (25.8%) | |
| Time-varying covariates | |||||
| Depression (CES-D) | |||||
| Mean CES-D score: visit 1 | 11.57 ± 7.20 | 11.84 ± 8.13 | 12.54 ± 8.45 | 13.48 ± 9.01 | .39 |
| Mean CES-D score: visit 2 | 8.82 ± 5.72 | 11.54 ± 8.99 | 11.77 ± 8.93 | 11.66 ± 9.73 | .56 |
| Mean CES-D score: visit 3 | 8.61 ± 6.67 | 9.70 ± 7.65 | 9.16 ± 7.41 | 10.13 ± 8.23 | .75 |
| Mean CES-D score: visit 4 | 7.45 ± 5.04 | 9.60 ± 8.11 | 9.89 ± 7.71 | 10.68 ± 9.06 | .37 |
| Self-esteem (RSES) | |||||
| Mean RSES score: visit 1 | 33.30 ± 5.51 | 33.67 ± 5.04 | 32.83 ± 4.96 | 32.67 ± 5.44 | .34 |
| Mean RSES score: visit 2 | 34.68 ± 4.98 | 34.17 ± 4.88 | 34.45 ± 4.96 | 33.85 ± 5.87 | .82 |
| Mean RSES score: visit 3 | 33.96 ± 4.71 | 33.77 ± 5.32 | 34.24 ± 5.26 | 33.80 ± 5.22 | .91 |
| Mean RSES score: visit 4 | 34.64 ± 4.73 | 34.13 ± 5.15 | 33.74 ± 5.50 | 33.80 ± 5.44 | .85 |
| Perceived stress (PSS) | |||||
| Mean PSS score: visit 1 | 17.22 ± 6.40 | 17.47 ± 6.51 | 17.30 ± 7.37 | 18.07 ± 7.22 | .83 |
| Mean PSS score: visit 2 | 16.68 ± 5.24 | 16.38 ± 7.25 | 16.43 ± 8.36 | 16.33 ± 7.23 | > .99 |
| Mean PSS score: visit 3 | 16.48 ± 6.73 | 16.23 ± 7.00 | 15.78 ± 7.16 | 16.25 ± 7.39 | .95 |
| Mean PSS score: visit 4 | 16.55 ± 6.26 | 16.04 ± 7.54 | 16.45 ± 7.49 | 15.82 ± 6.90 | .93 |
| Prenatal health behaviors (PRHBS) | |||||
| Mean PRHBS score: visit 1 | 35.74 ± 5.97 | 34.01 ± 6.99 | 32.68 ± 7.74 | 32.74 ± 6.95 | .12 |
| Mean PRHBS score: visit 2 | 36.55 ± 5.21 | 34.96 ± 6.83 | 33.71 ± 7.16 | 34.02 ± 6.22 | .18 |
| Mean PRHBS score: visit 3 | 25.74 ± 6.38 | 27.83 ± 7.36 | 27.02 ± 6.98 | 27.73 ± 5.75 | .48 |
| Mean PRHBS score: visit 4 | 27.64 ± 7.43 | 26.77 ± 7.25 | 27.60 ± 6.79 | 27.77 ± 6.18 | .64 |
| Smoked cigarettes since pregnant/last interview | |||||
| n (%) positive in visit 1 | 5 (21.7%) | 51 (27.7%) | 25 (25.0%) | 33 (27.5%) | .90 |
| n (%) positive in visit 2 | 5 (22.7%) | 34 (18.9%) | 17 (17.2%) | 25 (21.2%) | .84 |
| n (%) positive in visit 3 | 7 (30.4%) | 56 (34.4%) | 32 (35.6%) | 37 (33.6%) | .97 |
| n (%) positive in visit 4 | 5 (22.7%) | 67 (40.9%) | 29 (31.9%) | 40 (37.4%) | .26 |
| Drank alcohol since pregnant/last interview | |||||
| n (%) positive in visit 1 | 1 (4.4%) | 13 (7.1%) | 7 (7.0%) | 10 (8.3%) | .92 |
| n (%) positive in visit 2 | 1 (4.6%) | 4 (2.2%) | 4 (4.0%) | 3 (2.6%) | .80 |
| n (%) positive in visit 3 | 9 (39.1%) | 55 (33.7%) | 33 (36.7%) | 27 (24.6%) | .22 |
| n (%) positive in visit 4 | 6 (27.3%) | 64 (39.0%) | 32 (35.2%) | 29 (27.1%) | .20 |
| Used marijuana since pregnant/last interview | |||||
| n (%) positive in visit 1 | 1 (4.4%) | 16 (8.7%) | 12 (12.0%) | 6 (5.0%) | .25 |
| n (%) positive in visit 2 | 1 (4.6%) | 8 (4.4%) | 6 (6.1%) | 2 (1.7%) | .44 |
| n (%) positive in visit 3 | 1 (4.4%) | 12 (7.4%) | 11 (12.2%) | 7 (6.4%) | .38 |
| n (%) positive in visit 4 | 1 (4.6%) | 18 (11.0%) | 8 (8.8%) | 8 (7.5%) | .66 |
| Used hard drugs since pregnant/last interview | |||||
| n (%) positive in visit 1 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | — |
| n (%) positive in visit 2 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | — |
| n (%) positive in visit 3 | 0 (0.0%) | 0 (0.0%) | 1 (1.1%) | 0 (0.0%) | — |
| n (%) positive in visit 4 | 0 (0.0%) | 2 (1.2%) | 0 (0.0%) | 1 (0.9%) | — |
| Used oral contraceptive pills since delivered/last interview | |||||
| n (%) positive in visit 3 | 3 (13.6%) | 33 (22.8%) | 20 (24.1%) | 21 (21.9%) | .77 |
| n (%) positive in visit 4 | 2 (10.5%) | 35 (23.0%) | 17 (20.0%) | 24 (25.8%) | .48 |
| Used depot progresterone since delivered/last interview | |||||
| n (%) positive in visit 3 | 9 (40.9%) | 46 (31.9%) | 15 (18.1%) | 22 (22.9%) | .04b |
| n (%) positive in visit 4 | 8 (42.1%) | 44 (29.0%) | 17 (20.0%) | 17 (18.3%) | .053 |
| Used contraceptive patch since delivered/last interview | |||||
| n (%) positive in visit 3 | 0 (0.0%) | 16 (11.1%) | 9 (10.8%) | 4 (4.2%) | .10 |
| n (%) positive in visit 4 | 2 (10.5%) | 18 (11.8%) | 11 (12.9%) | 7 (7.5%) | .66 |
| Repeat pregnancy in postpartum period | |||||
| n (%) repeat pregnancy at visit 3 | 1 (4.4%) | 15 (9.2%) | 8 (8.9%) | 17 (15.5%) | .23 |
| n (%) repeat pregnancy at visit 4 | 4 (18.2%) | 31 (18.9%) | 22 (23.9%) | 32 (29.6%) | .21 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
