Objective
Our objective was to examine the associations of total and trimester-specific gestational weight gain (GWG) rate with postpartum maternal weight and cardiometabolic risk. We hypothesized that first-trimester GWG would be most strongly associated with long-term maternal health.
Study Design
We studied 801 women enrolled during the first trimester of pregnancy in the Boston-area Project Viva cohort 1999 through 2002. At 3 years postpartum we measured maternal weight, waist circumference (WC), and systolic blood pressure (SBP) and collected fasting blood from a subset. At 7 years postpartum we again measured weight and WC. We used multivariable linear regression to evaluate relations of total and trimester-specific GWG rate with weight change (vs self-reported prepregnancy weight) and WC at each time point, stratified by prepregnancy weight, as well as associations with SBP and insulin resistance at 3 years.
Results
Median age at enrollment was 34.0 years (range, 16.4–44.9); 65% were white. Mean (SD) total GWG rate was 0.38 (0.14) kg/wk. Women gained weight faster during the second (0.47 [0.19] kg/wk) and third (0.44 [0.22] kg/wk) trimesters than the first (0.22 [0.22] kg/wk). Total and first-trimester GWG rate were most strongly associated with postpartum weight change. Among normal-weight women, each 1-SD increase in total and first-trimester GWG rate corresponded with 0.85 (95% confidence interval [CI], 0.07–1.63) kg and 2.08 (1.32–2.84) kg greater weight change at 3 and 7 years postpartum, respectively, but there was not strong evidence of association for either second- (–0.30 kg; 95% CI, –1.08 to 0.48) or third- (–0.26 kg; 95% CI, –1.08 to 0.55) trimester GWG. First-trimester GWG rate also related to 3-year postpartum weight change in overweight (2.28 kg; 95% CI, 0.95–3.61) and obese (2.47 kg; 95% CI, 0.98–3.97) women. Greater total and first-trimester GWG rate were associated with larger WC and higher SBP but not insulin resistance.
Conclusion
In this observational cohort, first-trimester weight gain was more strongly associated with maternal weight retention as well as higher WC and blood pressure than second- or third-trimester gain. Interventions targeting GWG beginning very early in pregnancy may benefit long-term maternal health.
An estimated 40% of normal-weight women and 60% of overweight women exceed Institute of Medicine (IOM) recommendations for gestational weight gain (GWG). Excess GWG is associated with greater short-term postpartum weight retention and adverse cardiometabolic outcomes for both mother and child, including increased risk of obesity, central adiposity, and higher systolic blood pressure (SBP).
Clinicians and researchers typically focus on total GWG. However, weight gained across specific intervals of pregnancy has differential contributions to maternal, placental, and fetal growth. Greater early GWG is associated with higher risk of pregnancy complications, including hypertensive disorders of pregnancy, gestational diabetes, and postpartum weight retention, independent of body mass index (BMI) entering pregnancy. To our knowledge only 1 study has examined GWG timing with respect to longer-term maternal outcomes. In that analysis, midpregnancy GWG (19-28 weeks) was more strongly related to adiposity and blood pressure at 16 years postpartum, but only among normal-weight women. However, in that United Kingdom cohort recruited 1991 through 1992, both maternal BMI (mean, 22.5 kg/m 2 ) and GWG (mean, 12.6 kg) were substantially lower than is common in the United States today, and obese women were not studied separately. Categorizing by prepregnancy BMI is both practically important given that current IOM guidelines stratify by prepregnancy weight status, and physiologically important given known differences in gestational metabolism between normal-weight and obese women.
In this study, we hypothesized that first-trimester GWG would be most strongly associated with long-term maternal cardiometabolic health measures, including overall and central adiposity, SBP, and insulin resistance at 3 and 7 years postpartum.
Materials and Methods
Study population
We studied participants from Project Viva, a prospective prebirth cohort of women enrolled 1999 through 2002 from Harvard Vanguard Medical Associates, a multispecialty practice in eastern Massachusetts. All women provided written informed consent at each visit and the Harvard Pilgrim Health Care Institutional Review Board approved this study.
Of 2128 mothers who delivered a live singleton infant, we conducted in-person visits with 1287 at 3 years postpartum (median, 3.2; range, 2.8–6.2 years) and 1107 at 7 years postpartum (median, 7.7; range, 6.6–10.9 years). We excluded from this analysis women who were underweight entering pregnancy (BMI <18.5 kg/m 2 ) because of insufficient sample size for stratified analysis, or who had an intervening pregnancy after the index delivery, which might affect outcomes, which left 741 eligible participants in the 3-year analysis and 525 women in the 7-year analysis. The 801 women total (some women were included at both visits) were similar to those excluded although the analyzed cohort was slightly older (mean age 34 vs 31 years), had higher prepregnancy BMI (∼1.2 kg/m 2 ), and accordingly gained less total weight during pregnancy (0.9 kg).
Exposure: total and trimester-specific GWG rate
Women reported their prepregnancy weight at study enrollment. We obtained serial clinical weights from prenatal records, with a median of 13 (3-28) repeated measurements per woman. We calculated total GWG rate as the difference between the last clinically measured weight (within 4 weeks prior to delivery) and self-reported prepregnancy weight divided by gestation length in weeks. Since literature varies for specific periods of GWG, we took a data-driven approach to identify population-specific cutoff points that demarcate significant differences in GWG rate throughout pregnancy. Using methods described by Fraser et al, we identified linear slope changes at 15 and 28 weeks’ gestational age. However, these time points were not substantially superior to those used for trimester cutoffs, so we proceeded with more clinically relevant trimesters. We defined first trimester as the date of last menstrual period to day 91, second trimester as day 91-182, and third trimester as day 182 to the date of delivery. We performed linear interpolation between the 2 closest weight measures to estimate weight at day 91 and day 182 and calculated trimester-specific GWG rates (kg/wk) for each period.
Postpartum health outcomes
During in-person 3-year postpartum visits, trained research assistants measured women’s weight to the nearest 0.1 kg using an electronic scale (model 881; Seca Corp, Hanover, MD); waist circumference (WC) to the nearest 1 mm using a nonstretchable measuring tape; and SBP 5 times, 1 minute apart, with a Dinamap Pro 100 (Critikon Inc, Tampa, FL). Phlebotomists collected blood specimens, which were processed within 24 hours. We assayed glucose and insulin from plasma only from women who were fasting at least 12 hours at the time of the visit (n = 181); we did not require that all women attend the visit in a fasted state. We used the homeostasis model assessment (HOMA) to estimate insulin resistance from these samples: HOMA-IR = (fasting insulin [μU/mL] × fasting glucose [mmol/L]/22.5). At the 7-year visit, research assistants again measured weight (scale model TBF-300A; Tanita Corp of America Inc, Arlington Heights, IL) and WC, but not blood pressure and did not collect blood, as we did not have funding at this visit to assess maternal outcomes. We calculated postpartum weight change as the difference between weight at each time point (3 or 7 years postpartum) and prepregnancy weight.
Assessment of covariates
At enrollment and the midpregnancy visit, women completed questionnaires and interviews inquiring about race, ethnicity, education, parity, and lifestyle characteristics including smoking habits, diet, and physical activity. We calculated prepregnancy BMI using self-reported prepregnancy weight and height. Medical records provided prenatal glucose tolerance status, delivery date, and infant sex. Mothers reported breast-feeding duration in postpartum questionnaires.
Data analysis
First, in bivariate analyses, we evaluated the distributions of total weight gain rate and gain in each trimester across categories of sociodemographic, anthropometric, perinatal, and lifestyle characteristics to identify potential confounders of the association of GWG rate with postpartum outcomes. We assessed their significance using the Wald χ 2 test for categorical variables and a test for linear trend for ordinal variables.
Next, we examined the relations of total and trimester-specific weight gain rate with change in weight and WC by 3 and 7 years postpartum using multivariable linear regression models. We conducted separate analyses for normal-weight (BMI 18.5–24.9 kg/m 2 ), overweight (BMI 25.0–29.9 kg/m 2 ), and obese (BMI ≥30 kg/m 2 ) women in light of evidence of an interaction ( P -interaction < .05) between prepregnancy weight status and GWG rate for the anthropometric outcomes. We quantified the difference and 95% confidence interval in each postpartum outcome according to quintiles of total and trimester-specific GWG rate. Because the associations were generally linear and monotonic, we opted to examine GWG rate continuously, scaled to a 1-SD increment to optimize power. In our base model (model 1), we accounted for confounders including age, race, education, prepregnancy BMI, parity, and partner’s BMI. We adjusted for GWG rate in the previous period(s) when second- or third-trimester GWG rate was the exposure of interest. Unadjusted results are not presented since they were not notably different from base model results. We then additionally adjusted for pregnancy smoking habits, gestational glucose tolerance, child’s sex, and breast-feeding duration (model 2). Covariate selection was based on bivariate associations and a priori knowledge of GWG predictors. All models met assumptions for linear regression.
For cardiometabolic outcomes (SBP and HOMA-IR) at 3 years postpartum, we examined associations among all women because associations did not differ by prepregnancy weight status ( P -interaction > .10 for all models). We accounted for the same set of covariates as described above for models 1 and 2, but we also included a third model (model 3) that accounted for current adiposity (BMI at 3 years postpartum). In models for HOMA-IR, we excluded women with type 1 and type 2 diabetes, or gestational diabetes mellitus, which can influence both patterns of weight gain and postpartum hyperglycemia.
Adjustment for physical activity before pregnancy, and adherence to the prudent or Western dietary patterns during early pregnancy did not change results so they were not included in the final models. We conducted sensitivity analyses excluding preterm births (gestational age <37 weeks; n = 56) and women with preeclampsia (n = 26) and observed no substantial differences in our findings.
Some subjects were missing covariate information ( Appendix ; Supplemental Table 1 ), so we used chained equations in SAS PROC MI to impute values for these covariates. To avoid incorrect imputations, all Project Viva subjects were used in generating imputed values. We generated 50 imputed datasets and our reported results combine the estimates across the results. We included only participants with observed, nonimputed outcomes. The results from analyses using imputed data were not materially different from those using original data (data not shown). All analyses were performed using software (SAS, version 9.3; SAS Institute Inc, Cary, NC).
Results
Median age at enrollment was 34.0 years (range, 16.4–44.9) and 65.4% of women were white. Mean (SD) total GWG was 15.0 (5.6) kg over 39.4 (1.9) weeks’ gestation, or an average gain of 0.38 (0.14) kg/wk throughout pregnancy. Women gained weight at faster rates during the second (6.1 [2.5] kg at 0.47 [0.19] kg/wk) and third (6.0 [2.9] kg at 0.44 [0.22] kg/wk) trimesters than in the first (2.9 [2.9] kg at 0.22 [0.22] kg/wk). The mean rate of GWG for normal, overweight, and obese women in the cohort exceeded 2009 IOM recommendations for the second and third trimesters ( Table 1 ).
| Variable | Observed in Project Viva | IOM guidelines, a | |
|---|---|---|---|
| Mean | Range | Range | |
| Total GWG rate, kg/wk | |||
| Normal weight (BMI 18.5–24.9) | 0.39 | 0.08 to 0.78 | 0.28–0.40 |
| Overweight (BMI 25–29.9) | 0.40 | 0.06 to 0.86 | 0.17–0.28 |
| Obese (BMI ≥30) | 0.30 | −0.19 to 0.76 | 0.12–0.23 |
| First-trimester GWG rate, kg/wk | |||
| Normal weight (BMI 18.5–24.9) | 0.22 | −0.67 to 1.08 | 0.04–0.15 |
| Overweight (BMI 25–29.9) | 0.24 | −0.69 to 1.20 | 0.04–0.15 |
| Obese (BMI ≥30) | 0.21 | −0.63 to 1.22 | 0.04–0.15 |
| Second-trimester GWG rate, kg/wk | |||
| Normal weight (BMI 18.5–24.9) | 0.51 | 0.04 to 1.06 | 0.36–0.45 |
| Overweight (BMI 25–29.9) | 0.50 | 0.01 to 1.20 | 0.23–0.32 |
| Obese (BMI ≥30) | 0.31 | −0.41 to 0.97 | 0.18–0.27 |
| Third-trimester GWG rate, kg/wk | |||
| Normal weight (BMI 18.5–24.9) | 0.45 | −0.24 to 0.98 | 0.36–0.45 |
| Overweight (BMI 25–29.9) | 0.46 | −0.34 to 1.15 | 0.23–0.32 |
| Obese (BMI ≥30) | 0.39 | −0.45 to 1.00 | 0.18–0.27 |
Associations of GWG rate with sociodemographic and behavioral characteristics differed by trimester ( Table 2 ). Second- and third-trimester GWG rate varied with many sociodemographic characteristics including maternal age (third trimester), marital status and race/ethnicity (second trimester), and education and household income (both). In contrast, first-trimester GWG rate did not vary with these characteristics, but was associated with behaviors including smoking, prepregnancy physical activity, and diet pattern. GWG rate varied across prepregnancy BMI categories for second- and third-trimester gain, but not for first-trimester gain.
| Variable | Mean ± SD rate of GWG, kg/wk | ||||
|---|---|---|---|---|---|
| n (%) | Total | First trimester | Second trimester | Third trimester | |
| (0-13 wk) | (14-26 wk) | (≥27 wk) | |||
| Overall | 801 | 0.38 ± 0.14 | 0.22 ± 0.22 | 0.47 ± 0.19 | 0.44 ± 0.22 |
| SOCIODEMOGRAPHIC CHARACTERISTICS | |||||
| Age at enrollment, y | |||||
| 15-24 | 61 (8) | 0.40 ± 0.16 | 0.20 ± 0.31 | 0.48 ± 0.22 | 0.50 ± 0.25 |
| 25-34 | 403 (50) | 0.38 ± 0.15 | 0.22 ± 0.24 | 0.47 ± 0.21 | 0.45 ± 0.22 |
| 35-44 | 337 (42) | 0.37 ± 0.13 | 0.23 ± 0.19 | 0.47 ± 0.18 | 0.42 ± 0.21 |
| P value trend a | .25 | .36 | .83 | .003 | |
| Marital status | |||||
| Married/cohabitating | 725 (91) | 0.38 ± 0.14 | 0.22 ± 0.22 | 0.47 ± 0.19 | 0.44 ± 0.22 |
| Single | 76 (9) | 0.37 ± 0.17 | 0.24 ± 0.27 | 0.41 ± 0.25 | 0.45 ± 0.25 |
| P value b | .39 | .60 | .01 | .82 | |
| Race/ethnicity | |||||
| Black | 142 (18) | 0.36 ± 0.15 | 0.23 ± 0.29 | 0.42 ± 0.23 | 0.42 ± 0.24 |
| Hispanic | 65 (8) | 0.36 ± 0.16 | 0.23 ± 0.29 | 0.44 ± 0.21 | 0.40 ± 0.26 |
| White | 524 (65) | 0.39 ± 0.14 | 0.22 ± 0.19 | 0.48 ± 0.19 | 0.46 ± 0.21 |
| Asian | 35 (4) | 0.36 ± 0.09 | 0.20 ± 0.17 | 0.45 ± 0.11 | 0.45 ± 0.21 |
| Other | 35 (4) | 0.41 ± 0.16 | 0.27 ± 0.33 | 0.49 ± 0.18 | 0.47 ± 0.22 |
| P value b | .11 | .53 | .01 | .13 | |
| Annual household income | |||||
| <$20,000 | 51 (6) | 0.37 ± 0.18 | 0.23 ± 0.37 | 0.43 ± 0.25 | 0.44 ± 0.27 |
| $20,000–39,999 | 92 (11) | 0.35 ± 0.18 | 0.23 ± 0.31 | 0.42 ± 0.24 | 0.39 ± 0.26 |
| $40,000–69,999 | 180 (22) | 0.37 ± 0.16 | 0.22 ± 0.27 | 0.45 ± 0.24 | 0.45 ± 0.24 |
| $70,000–100,000 | 173 (22) | 0.37 ± 0.14 | 0.22 ± 0.19 | 0.48 ± 0.20 | 0.41 ± 0.23 |
| >$100,000 | 306 (38) | 0.40 ± 0.12 | 0.23 ± 0.19 | 0.49 ± 0.17 | 0.48 ± 0.20 |
| P value trend a | .003 | .87 | .0003 | .02 | |
| Education | |||||
| Primary | 75 (9) | 0.35 ± 0.15 | 0.22 ± 0.29 | 0.43 ± 0.23 | 0.40 ± 0.21 |
| Secondary | 486 (61) | 0.38 ± 0.15 | 0.22 ± 0.23 | 0.47 ± 0.21 | 0.44 ± 0.24 |
| University | 240 (30) | 0.39 ± 0.12 | 0.22 ± 0.18 | 0.48 ± 0.16 | 0.46 ± 0.19 |
| P value trend a | .08 | .89 | .05 | .05 | |
| Smoking habits | |||||
| Never | 558 (70) | 0.38 ± 0.14 | 0.22 ± 0.22 | 0.47 ± 0.18 | 0.44 ± 0.21 |
| Quit before pregnancy | 155 (19) | 0.38 ± 0.14 | 0.22 ± 0.21 | 0.47 ± 0.21 | 0.45 ± 0.24 |
| Smoked in early pregnancy | 88 (11) | 0.40 ± 0.16 | 0.28 ± 0.26 | 0.46 ± 0.23 | 0.45 ± 0.24 |
| P value b | .64 | .02 | .92 | .81 | |
| Partner’s BMI, kg/m 2 | |||||
| Not overweight (<25 kg/m 2 ) | 283 (35) | 0.39 ± 0.13 | 0.23 ± 0.21 | 0.48 ± 0.19 | 0.45 ± 0.22 |
| Overweight (≥25 kg/m 2 ) | 518 (65) | 0.38 ± 0.15 | 0.22 ± 0.24 | 0.46 ± 0.20 | 0.44 ± 0.23 |
| P value b | .39 | .79 | .14 | .91 | |
| PERINATAL CHARACTERISTICS | |||||
| Prepregnancy BMI, kg/m 2 | |||||
| Normal (18.5–24.9 kg/m 2 ) | 454 (57) | 0.39 ± 0.11 | 0.22 ± 0.18 | 0.51 ± 0.15 | 0.45 ± 0.19 |
| Overweight (25.0–29.9 kg/m 2 ) | 195 (24) | 0.40 ± 0.14 | 0.24 ± 0.24 | 0.50 ± 0.19 | 0.46 ± 0.24 |
| Obese (≥30.0 kg/m 2 ) | 152 (19) | 0.30 ± 0.19 | 0.21 ± 0.31 | 0.31 ± 0.25 | 0.39 ± 0.28 |
| P value trend a | < .0001 | .91 | < .0001 | .01 | |
| Parity | |||||
| 0 | 234 (29) | 0.39 ± 0.14 | 0.23 ± 0.21 | 0.48 ± 0.20 | 0.47 ± 0.22 |
| 1 | 368 (46) | 0.38 ± 0.14 | 0.22 ± 0.23 | 0.46 ± 0.19 | 0.45 ± 0.21 |
| ≥2 | 199 (25) | 0.37 ± 0.14 | 0.23 ± 0.24 | 0.47 ± 0.20 | 0.41 ± 0.23 |
| P value trend a | .10 | .86 | .71 | .001 | |
| Gestational glucose tolerance | |||||
| Normoglycemic | 648 (81) | 0.39 ± 0.14 | 0.22 ± 0.22 | 0.48 ± 0.19 | 0.46 ± 0.20 |
| Isolated hyperglycemia | 82 (10) | 0.35 ± 0.15 | 0.21 ± 0.23 | 0.44 ± 0.20 | 0.40 ± 0.24 |
| Impaired glucose tolerance | 32 (4) | 0.40 ± 0.16 | 0.30 ± 0.27 | 0.48 ± 0.24 | 0.41 ± 0.27 |
| Gestational diabetes | 39 (5) | 0.31 ± 0.16 | 0.25 ± 0.25 | 0.39 ± 0.22 | 0.29 ± 0.34 |
| P value trend a | .003 | .20 | .02 | < .0001 | |
| Offspring sex | |||||
| Male | 412 (51) | 0.39 ± 0.14 | 0.23 ± 0.23 | 0.48 ± 0.20 | 0.46 ± 0.23 |
| Female | 389 (49) | 0.37 ± 0.14 | 0.22 ± 0.22 | 0.46 ± 0.19 | 0.43 ± 0.21 |
| P value b | .02 | .36 | .12 | .05 | |
| Prudent dietary pattern | |||||
| Q1 (lowest) | 199 (25) | 0.38 ± 0.15 | 0.20 ± 0.27 | 0.47 ± 0.20 | 0.46 ± 0.23 |
| Q2 | 201 (25) | 0.37 ± 0.15 | 0.20 ± 0.21 | 0.47 ± 0.22 | 0.44 ± 0.22 |
| Q3 | 200 (25) | 0.38 ± 0.15 | 0.22 ± 0.21 | 0.47 ± 0.20 | 0.45 ± 0.24 |
| Q4 (highest) | 201 (25) | 0.39 ± 0.15 | 0.28 ± 0.25 | 0.47 ± 0.19 | 0.43 ± 0.24 |
| P value trend a | .19 | .001 | .72 | .21 | |
| Prepregnancy physical activity | |||||
| Q1 (lowest) | 169 (21) | 0.35 ± 0.15 | 0.18 ± 0.25 | 0.44 ± 0.21 | 0.43 ± 0.22 |
| Q2 | 178 (22) | 0.40 ± 0.13 | 0.23 ± 0.23 | 0.49 ± 0.20 | 0.48 ± 0.21 |
| Q3 | 244 (30) | 0.39 ± 0.14 | 0.24 ± 0.21 | 0.47 ± 0.20 | 0.45 ± 0.23 |
| Q4 (highest) | 210 (26) | 0.38 ± 0.16 | 0.24 ± 0.26 | 0.48 ± 0.21 | 0.42 ± 0.25 |
| P value trend a | .12 | .01 | .19 | .47 | |
a From test for linear trend in which ordinal predictor is entered into model as continuous variable
In adjusted regression models, the rate of first-trimester GWG was more strongly associated with postpartum weight change and WC than rate of gain in either of the other 2 trimesters, regardless of prepregnancy weight status ( Figure ). Results for second- and third-trimester gain were similar even without adjustment for weight gain in the previous period(s) (data not shown). Associations of total GWG rate with both weight change and WC followed similar trends to those of the first trimester, more so than the second and third trimesters at both 3 and 7 years postpartum ( Figure ). Additional adjustment for smoking, gestational glucose tolerance, offspring sex, and breast-feeding duration did not attenuate associations ( Supplemental Tables 2 and 3 ).
