Maternal prepregnancy waist circumference and BMI in relation to gestational weight gain and breastfeeding behavior: the CARDIA study^,,

Abstract

Background: Studies suggest that gestational weight gain (GWG) and breastfeeding behavior may influence long-term maternal abdominal fat mass. However, this could be confounded by abdominal fat mass before pregnancy because it is unknown whether abdominal fat mass, independently of body size, affects GWG and breastfeeding behavior.

Objective: We investigated how maternal prepregnancy fat distribution, described by waist circumference (WC) and body mass index (BMI), is associated with GWG and breastfeeding behavior.

Design: We analyzed 1371 live births to 1024 women after enrollment in the Coronary Artery Risk Development in Young Adults study (1985–1996). For each birth, maternal prepregnancy BMI and WC were measured at year 0 (baseline), 2, 5, or 7 examinations. Recalled GWG and breastfeeding behavior were collected at years 7 and 10. GWG was analyzed by using linear regression and breastfeeding behavior by using logistic regression and discrete-time logistic regression.

Results: Adjusted for potential confounders, a 1-cm larger WC adjusted for BMI was associated with a 0.19-kg (95% CI: −0.29-, −0.10-kg) lower GWG. In contrast, a 1-unit higher BMI adjusted for WC was associated with a 0.27-kg (95% CI: 0.06-, 0.47-kg) higher GWG. The OR for ever breastfeeding compared with never breastfeeding was 0.93 (95% CI: 0.90, 0.97) per 1-cm larger WC after adjustment for BMI, whereas it was 1.10 (95% CI: 1.02, 1.19) per 1-unit higher BMI adjusted for WC.

Conclusions: Maternal prepregnancy body size was differently associated with GWG and breastfeeding behavior depending on the location of the fat mass. Thus, maternal fat distribution may be a more important determinant of GWG and breastfeeding behavior than BMI alone. This trial was registered at clinicaltrials.gov as NCT00005130.

INTRODUCTION

Childbearing is associated with increased maternal waist circumference and visceral fat mass later in life (1, 2). This might be due to excessive gestational weight gain (GWG), lack of breastfeeding, or shorter breastfeeding duration (3–7). However, these studies had not controlled for measures of maternal abdominal fat mass before pregnancy. We previously conducted a path analysis to study the association between GWG, breastfeeding duration, and maternal abdominal fat mass 7 y after delivery, adjusted for BMI (in kg/m²) (7). The observed pathways suggested that confounding from prepregnancy abdominal fat mass may be present because abdominal fat mass before pregnancy may be a determinant of GWG as well as breastfeeding duration, thereby affecting the validity of the findings. Thus, maternal measures of abdominal fat mass before pregnancy may be needed to evaluate the independent effects of GWG and breastfeeding behavior on subsequent maternal abdominal fat mass.

Very little is known about the influence of maternal abdominal fat mass before pregnancy on reproductive outcomes mainly because this measure is rarely available. It has been suggested that abdominal fat mass may be a better predictor of cesarean delivery, large-for-gestational age, gestational diabetes, birth weight, and maternal hypertension than prepregnancy BMI (8, 9). It is unknown whether this is also the case for GWG and breastfeeding behavior. Higher maternal BMI before pregnancy has been associated with lower GWG and shorter duration of breastfeeding or no breastfeeding in some (10–15) but not all studies (16, 17). The inconsistency might result from differences in maternal fat distribution because BMI provides information only about overall body size (18). Among nonpregnant women, abdominal fat mass is related to an adverse metabolic profile and an increased risk of mortality (19, 20), whereas the opposite has been observed for lower body fat mass (21, 22). Studying both waist circumference and BMI simultaneously seems to make it possible to distinguish between regional fat distributions and their different effects (19, 20, 23).

The aim of the present study was to examine how maternal prepregnancy fat distribution, described by waist circumference and BMI, was related to GWG and breastfeeding behavior. We hypothesized that larger prepregnancy waist circumference was associated with lower GWG, reduced probability of ever breastfeeding, and shorter duration of breastfeeding.

METHODS

Study participants

Details of the Coronary Artery Risk Development in Young Adults (CARDIA) study have been reported elsewhere (24, 25). Briefly, CARDIA is a multicenter, longitudinal observational study designed to describe the development of risk factors for coronary artery disease in young adults. In 1985–1986, a total of 2787 women (52% black, 48% white) aged 18–30 y were enrolled from 4 geographic areas: Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California. For this study, we used data from follow-up examinations conducted at 2, 5, 7, and 10 y after the baseline (year 0) examination, because information on GWG was available only within the first 10 y of follow-up. Retention rates were high at the follow-up examinations (∼91%, 86%, 81%, and 79% of surviving participants) (26).

Of the 2787 women enrolled at baseline (1985–1986), 1103 women gave birth 1669 times during the 10-y follow-up period (deliveries after week 20 of gestation are defined as births) (Figure 1). Of these, we excluded 8 births with no available date of delivery and 156 subsequent births to the same woman within each of the 4 time intervals (0–2 y, 2–5 y, 5–7 y, and 7–10 y), so that each woman could contribute only one birth per interval but more births in other intervals. Furthermore, we excluded 30 sets of twin births and 35 births because the mother was pregnant (>60 d of gestation) at the time of examination, and her previous examination was missing or unusable because of pregnancies in the previous interval. Finally, we excluded 17 births ending in a stillbirth or unknown status and 52 births because maternal prepregnancy anthropometric measures were missing. The final sample included 1024 women (49% black) with 1371 births. In total, 715 women contributed with one birth, 273 with 2 births, 34 with 3 births, and 2 with 4 births. Institutional review boards at each participating study center approved the study, and participants gave written informed consent at each examination.

FIGURE 1.

Flow diagram of sample selection criteria for the CARDIA study, 1985–1996. A birth was defined as a delivery after 20 wk of gestation. CARDIA, Coronary Artery Risk Development in Young Adults.

Definition of prepregnancy

The examinations at years 0, 2, 5, and 7 provided prepregnancy measurements for births in the subsequent time intervals (e.g., prepregnancy information for births that occurred between years 2 and 5 was obtained from the year 2 examination). The period between prepregnancy measurements and estimated date of conception therefore varied among women and was included in the models as a covariate. Date of conception was estimated as date of delivery minus gestational age.

Data collection

At each examination, the characteristics of participants, including lifestyle, sociodemographics, medical conditions, and reproductive events since last examination (number of pregnancies and births plus pregnancy outcome, gestational age, and birth weight), were assessed by self- and interviewer-administered questionnaires.

Anthropometric measurements

Anthropometric measurements were obtained by certified technicians according to standardized protocols at each examination (24, 25). Body weight (wearing light clothing) was measured to the nearest 0.2 kg by using a calibrated balance beam scale. Height (without shoes) was measured to the nearest 0.5 cm by using vertical ruler and waist circumference to the nearest 0.5 cm at the minimal abdominal girth (27).

GWG and breastfeeding behavior

Information on GWG for all births within the 10 y of follow-up was obtained from the year 10 Medical History Questionnaire (28). We conducted a validation study for a subsample of pregnancies (n = 133) within the study population for whom GWG was measured and obtained from medical records. Women tended to overreport their measured GWG by a mean ± SD of 0.41 ± 3.72 kg. A Bland-Altman plot of the difference against the mean of the 2 estimates showed no systematic trend. Thus, the estimated mean difference was similar over the entire GWG scale, and the agreement between the self-reported and measured GWG was good. Recall was slightly better for births within the last 5 y than the first 5 y of the 10-y period. However, a small overreporting of GWG within each period and no significant difference between the 2 periods were observed.

Breastfeeding behavior for all live births before the year 7 follow-up was obtained from the year 7 Medical History Questionnaire (29) and for all live births between years 7 and 10 from the year 10 Pregnancy Questionnaire (30). For each birth, women were asked whether they breastfed (yes or no), which defined ever/never breastfeeding. If women responded yes, then they were asked “how long” they breastfed for each child by choosing one of the following duration categories: <6 wk, 6–11 wk, 3–6 mo, or >6 mo. Therefore, ever breastfeeding included women with any duration of breastfeeding. Breastfeeding duration was defined based on the 4 duration categories.

Covariates

Potential covariates to include in the analyses of this study were selected a priori among those available in the CARDIA study. The CARDIA study collected information on race (black and white) at each study center. The study was designed to recruit equal numbers of black and white women. From each prepregnancy examination, the following time-dependent covariates were obtained: smoking (never, former, or current), marital status (married or unmarried), work status (full-time, part-time, or unemployed), years of education (≤12, 13–15, or ≥16), and parity (1, 2, 3, or ≥4); on the continuous scale were age (y), time from prepregnancy measurements to conception (d), alcohol intake (mL/d), and physical activity (race-specific physical activity score quartiles) (31, 32). Child birth weight (g) and gestational age (wk) were reported at the examination at the end of each time interval. Dietary intake was measured at years 0 and 7, and thus year 0 provided information for the first 3 intervals and year 7 for the fourth interval. Daily intakes of total fat, protein, and carbohydrate in grams were calculated as percentages of intake in kilojoules and total fiber intake as grams per 1000 kJ.

Statistical methods

Maternal prepregnancy anthropometric measures, reproductive factors, and potential confounders for each birth were described by tertiles of prepregnancy waist circumference.

All statistical analyses were performed by using STATA/SE 13 (StataCorp LP), and to address the problem of missing data (Table 1 shows the number of missing values), we used multiple imputation (33). The imputation and subsequent analyses were conducted by using mi procedures available in STATA 13. Because of our exclusion criteria, some variables had complete data, such as prepregnancy weight, height, age, and waist circumference, and were included in the imputation step as explanatory variables in addition to the other variables included for imputation. For women with no information about breastfeeding (n = 247), we first imputed information on whether she breastfed and then the duration in days, with a universal lower limit of 1 d and an upper limit of 365 d. The breastfeeding duration was then categorized as <6 wk, 6–11 wk, 3–6 mo, or >6 mo in accord with the self-reported categories. A total of 50 copies of the data set were created, each of which had had its missing values imputed, with an appropriate level of randomness, by chained equations (33, 34). For comparison, we also performed complete case analyses, including only women with no missing information. These results did not differ substantially from those presented (see Supplemental Tables 1–3).

TABLE 1.

Maternal characteristics before and after each birth presented by prepregnancy waist circumference tertiles: 1024 women with 1371 births from the CARDIA study, 1985–1996¹

	Prepregnancy waist circumference
Maternal characteristic	≤67 cm (n = 340)	>67–<80 cm (n = 679)	≥80 cm (n = 352)	Missing
Black race, n (%)	128 (38)	300 (44)	249 (71)	0
Age, y	27.0 ± 4.42	27.8 ± 4.2	27.9 ± 4.2	0
Education <12 y, n (%)	97 (29)	184 (27)	134 (38)	0
Employment (unemployed), n (%)	59 (17)	185 (27)	102 (29)	5
Marital status (married), n (%)	181 (53)	385 (57)	176 (50)	0
Time from conception and exam, d	399 ± 293	415 ± 312	383 ± 304	3
Primiparous, n (%)	209 (61)	311 (46)	122 (35)	0
Weight, kg	52.8 ± 5.1	63.3 ± 6.7	87.7 ± 15.3	0
Height, cm	162.7 ± 6.3	165.0 ± 6.6	164.9 ± 7.0	0
BMI, kg/m2	20.0 ± 1.7	23.3 ± 2.3	32.3 ± 5.3	0
Overweight or obese (BMI ≥25), n (%)	0 (0)	137 (20)	338 (96)	0
Waist circumference, cm	64.1 ± 2.3	72.7 ± 3.4	91.6 ± 9.7	0
Smoking (current), n (%)	70 (21)	174 (26)	86 (25)	6
Physical activity score3	276 (342)	258 (292)	209 (260)	8
Alcohol intake,3 mL/d	2.4 (8.8)	2.4 (9.7)	0 (7.3)	75
Dietary fat, % kJ	37.3 ± 6.5	37.1 ± 6.7	37.0 ± 6.2	40
Dietary CHO, % kJ	47.2 ± 7.4	47.2 ± 7.6	47.8 ± 7.6	40
Dietary protein, % kJ	14.8 ± 2.6	15.1 ± 2.7	14.8 ± 2.9	40
Fiber, g/1000 kJ	0.9 ± 0.8	0.9 ± 0.9	0.9 ± 0.8	40
Gestational age, d	268.5 ± 26.8	268.7 ± 29.1	268.9 ± 27.9	3
Child birth weight, g	3255 ± 553	3354 ± 637	3369 ± 719	175
Gestational weight gain, kg	14.5 ± 5.3	16.2 ± 6.1	13.3 ± 7.9	209
Breastfeeding duration, n (%)				247
None	70 (25)	148 (27)	134 (48)
<6 wk	28 (10)	75 (13)	31 (11)
6–11 wk	37 (13)	46 (8)	18 (6)
3–6 mo	70 (25)	126 (23)	53 (19)
>6 mo	77 (27)	165 (29)	46 (16)

¹CARDIA, Coronary Artery Risk Development in Young Adults; CHO, carbohydrate.

²Mean ± SD (all such values).

³Values are medians; IQRs in parentheses.

Multiple linear regression was used to examine the association of prepregnancy waist circumference and BMI with GWG and multiple logistic regression to examine the association with ever breastfeeding compared with never breastfeeding. A discrete-time logistic hazards model was used to study breastfeeding duration within the 4 categories (<6 wk, 6–11 wk, and >6 mo) among women who ever breastfed. Thus, we estimated the OR of breastfeeding cessation in a given breastfeeding duration category, according to larger waist circumference and higher BMI, conditionally on not having ceased breastfeeding in the previous interval. To allow different associations over time, we included interaction terms between breastfeeding duration categories and waist circumference and BMI. All models were adjusted for time between examination and conception. In the fully adjusted models, we further adjusted for maternal age, race, study center, parity, educational level, work status, child birth weight, gestational length, physical activity level, dietary nutrient intake, alcohol intake, and smoking. Prepregnancy BMI, race, and parity were examined as effect measure modifiers of the association of waist circumference with GWG and ever breastfeeding by including an interaction term. We used robust variance estimation to account for the clustering of several births within a given woman (35, 36). We performed several sensitivity analyses. We restricted our study population to the first birth for each woman or to term births; however, the results did not differ from those presented (see Supplemental Tables 4–9). We examined any contribution of GWG to the association of prepregnancy maternal waist circumference and BMI with ever breastfeeding by adjusting for GWG. Also, we adjusted the analyses of ever breastfeeding for calendar year of birth because of secular changes in breastfeeding behavior. No changes in the results were observed (data not shown). Finally, we adjusted the analyses of maternal prepregnancy BMI and waist circumference on ever breastfeeding and GWG, respectively, for prepregnancy insulin resistance determined by the HOMA-IR based on fasting serum insulin and glucose measures at years 0 and 7 in the CARDIA study (37, 38). The observed associations were similar to those presented (data not shown).

It is well known that waist circumference and BMI are correlated; in the present study, the correlation coefficient was 0.91, as observed elsewhere (39). However, there was still variation in waist circumference for a given BMI, and the mutually adjusted estimates examined the scientific question we were interested in—namely, the contribution of maternal fat distribution (23). To address any potential problem of multicollinearity, we examined the associations of GWG and ever breastfeeding by categories of waist circumference (<80 or ≥80 cm) and BMI (<25 or ≥25) and by waist circumference residuals when regressed on BMI. None was influenced by potential multicollinearity, and results were comparable to those presented (see Supplemental Tables 10–13).

We also studied waist circumference residuals in relation to GWG and ever breastfeeding by using restricted cubic splines with 4 knots to avoid restrictive assumptions on the associations. The waist circumference residuals express the deviation from the predicted value by BMI and reflect the variation in waist circumference (e.g., abdominal fat mass) that cannot be attributed to variation in overall adiposity as measured indirectly by BMI. Therefore, a positive value reflects a waist circumference larger than predicted by BMI and a negative value a smaller waist circumference than predicted by BMI.

RESULTS

For the 1371 births included in the study sample, the median time between the prepregnancy examination and conception was 371 d (5th, 95th percentile: 4, 939), the mothers’ mean ± SD prepregnancy BMI was 24.8 ± 5.6, and the prepregnancy waist circumference was 75.4 ± 11.6 cm. Compared with mothers in the middle or lower waist circumference tertiles, mothers in the highest tertile were more often of black race, were less often primiparous, had a higher prepregnancy BMI, gained less weight during pregnancy, and were less physically active, less educated, and less prone to breastfeed (Table 1).

GWG

In the crude models, prepregnancy BMI and waist circumference were inversely associated with GWG (Table 2). In the mutually adjusted model, waist circumference remained inversely associated with GWG, whereas BMI did not. In the fully adjusted model, a 1-cm larger waist circumference adjusted for BMI was associated with a 0.19-kg (95% CI: −0.29-, −0.10-kg) lower GWG. In contrast, a 1-unit higher BMI adjusted for waist circumference was associated with a 0.27-kg (95% CI: 0.06-, 0.47-kg) higher GWG. There were no statistically significant interactions by race (P = 0.58) or parity (primiparous compared with multiparous) (P = 0.40) on the association between waist circumference and GWG. When we divided women into parity 1, 2, 3 or ≥4, the same tendencies were observed for all parity groups. Prepregnancy BMI (<25 compared with ≥25) modified (P-interaction <0.01) the association between waist circumference and GWG. GWG was inversely associated with waist circumference for women who were overweight or obese but not for those who were normal weight or underweight, for whom the association was not statistically significant (Table 2).

TABLE 2.

Differences in gestational weight gain (kg) according to maternal prepregnancy waist circumference and BMI¹

	BMI, β (95% CI)
Prepregnancy anthropometric factor	All (n = 1371)	<25 (n = 896)	≥25 (n = 475)
Crude
Waist circumference, per cm	−0.07 (−0.11, −0.03)	0.06 (−0.02, 0.14)	−0.17 (−0.24, −0.10)
BMI (per unit)	−0.11 (−0.19, −0.02)	0.27 (0.06, 0.49)	−0.26 (−0.41, −0.11)
Mutually adjusted2
Waist circumference, per cm	−0.14 (−0.23, −0.05)	−0.03 (−0.14, 0.08)	−0.23 (−0.35, −0.11)
BMI (per unit)	0.16 (−0.02, 0.34)	0.33 (0.03, 0.63)	0.16 (−0.11, 0.42)
Fully adjusted3
Waist circumference, per cm	−0.19 (−0.29, −0.10)	−0.09 (−0.22, 0.04)	−0.27 (−0.41, −0.13)
BMI, per unit	0.27 (0.06, 0.47)	0.43 (0.11, 0.76)	0.27 (−0.03, 0.57)

¹All models were adjusted for time from examination to conception. P < 0.01 for interaction by prepregnancy BMI groups in the fully adjusted model.

²Waist circumference and BMI mutually adjusted.

³Waist circumference and BMI mutually adjusted and further adjusted for race, study center, gestational age, child birth weight, and maternal prepregnancy height, age, parity, marital status, education, work status, smoking, physical activity, alcohol intake, and intake of fat, protein, carbohydrate, and fiber.

Positive waist circumference residuals were associated with a lower GWG and negative waist circumference residuals with a higher GWG, as illustrated by the restricted cubic spline in Figure 2 (stratified by prepregnancy BMI <25 or ≥25 in Supplemental Figure 1). This figure shows that women who had a larger-than-predicted waist circumference based on their BMI gained less during pregnancy, whereas women who had a smaller-than-predicted waist circumference based on their BMI gained more.

FIGURE 2.

Difference in GWG according to centimeter deviation from the predicted waist circumference for a given BMI. The reference value is set to 0; to the left side is a lower waist circumference than predicted by BMI, and to the right side is a larger than predicted waist circumference. The solid line shows the estimated difference in GWG, and the dotted line the 95% CIs. Adjusted for race, study center, gestational age, child birth weight, time from prepregnancy measure to conception, and maternal prepregnancy height, age, parity, marital status, education, work status, smoking, physical activity, alcohol intake, and intake of fat, protein, carbohydrate, and fiber. GWG, gestational weight gain.

Breastfeeding behavior

In the crude models, larger waist circumference and higher BMI were associated with lower odds of ever breastfeeding than never breastfeeding (Table 3). In the mutually adjusted model, larger waist circumference showed unchanged lower odds, whereas the OR for BMI was not statistically significant. In the fully adjusted model, a 1-cm larger waist circumference for the same BMI unit was associated with a reduced odds (0.93; 95% CI: 0.90, 0.97) for ever breastfeeding compared with never breastfeeding (Table 3). In contrast, women who had a 1-unit higher BMI for the same waist circumference had a higher odds (1.10; 95% CI: 1.02, 1.19) for ever breastfeeding than never breastfeeding. No statistical interactions were observed with BMI <25 compared with ≥25 (P = 0.68), race (P = 0.67), or parity (primiparous compared with multiparous) (P = 0.10), but slightly stronger associations were observed for primiparous women than for multiparous women (see Supplemental Table 14). Further stratification into parity 1, 2, 3, or ≥4 showed the same tendencies among all parity groups. Adjustment for GWG in the analyses of maternal prepregnancy BMI and waist circumference on ever breastfeeding did not change the associations (see Supplemental Table 15).

TABLE 3.

ORs and 95% CIs for ever breastfeeding compared with never breastfeeding according to maternal prepregnancy waist circumference and BMI¹

Prepregnancy anthropometric factor	Never (n = 430), OR	Ever (n = 941), OR (95% CI)
Crude
Waist circumference, per cm	(Ref)	0.97 (0.95, 0.98)
BMI, per unit	(Ref)	0.94 (0.92, 0.97)
Mutually adjusted2
Waist circumference, per cm	(Ref)	0.95 (0.92, 0.98)
BMI, per unit	(Ref)	1.04 (0.98, 1.10)
Fully adjusted3
Waist circumference, per cm	(Ref)	0.93 (0.90, 0.97)
BMI, per unit	(Ref)	1.10 (1.02, 1.19)

¹All models were adjusted for time from examination to conception.

²Waist circumference and BMI mutually adjusted.

³Waist circumference and BMI mutually adjusted and further adjusted for race, study center, gestational age, child birth weight, and maternal prepregnancy height, age, parity, marital status, education, work status, smoking, physical activity, alcohol intake, and intake of fat, protein, carbohydrate, and fiber.

The OR for ever breastfeeding compared with never breastfeeding was lower with positive waist circumference residuals and higher with negative waist circumference residuals (Figure 3). These results mean that women who had a larger-than-predicted waist circumference based on their BMI were less likely to breastfeed, whereas women with a smaller-than-predicted waist circumference were more likely to breastfeed.

FIGURE 3.

The OR for ever compared with never breastfeeding according to centimeter deviation from the predicted waist circumference for a given BMI. The reference value is set to 0; to the left side is a lower waist circumference than predicted by BMI, and to the right side is a larger than predicted waist circumference. The solid line shows the OR, and the dotted line the 95% CIs. Adjusted for race, study center, gestational age, child birth weight, time from prepregnancy measure to conception, and maternal prepregnancy height, age, parity, marital status, education, work status, smoking, physical activity, alcohol intake, and intake of fat, protein, carbohydrate, and fiber.

Among women who breastfed (initiation with or without persistent breastfeeding), cessation within the first 6 wk was positively associated with prepregnancy BMI and waist circumference in the crude models (Table 4) and after adjustment for the covariates [OR per SD in waist circumference: 1.27 (95% CI: 1.04, 1.55); OR per SD in BMI: 1.25 (95% CI: 1.03, 1.53)]. When mutually adjusted and in the fully adjusted models, no statistically significant associations were observed for prepregnancy waist circumference and BMI with breastfeeding cessation in any of the breastfeeding duration categories (Table 4).

TABLE 4.

OR and 95% CIs for breastfeeding cessation in each category of breastfeeding duration according to maternal prepregnancy waist circumference and BMI among women who initiated breastfeeding (n = 941)¹

	Breastfeeding duration
Prepregnancy anthropometric factor	<6 wk (n = 145)	6–11 wk (n = 118)	3–6 mo (n = 327)
Crude
Waist circumference, per cm	1.03 (1.01, 1.04)	0.99 (0.97, 1.01)	1.01 (0.99, 1.03)
BMI, per unit	1.05 (1.02, 1.09)	0.99 (0.95, 1.04)	1.02 (0.99, 1.06)
Mutually adjusted2
Waist circumference, per cm	1.02 (0.94, 1.11)	0.96 (0.91, 1.01)	0.99 (0.95, 1.03)
BMI, per unit	1.02 (0.97, 1.06)	1.07 (0.97, 1.17)	1.05 (0.97, 1.13)
Fully adjusted3
Waist circumference, per cm	1.02 (0.97, 1.06)	0.96 (0.91, 1.02)	0.99 (0.94, 1.03)
BMI, per unit	1.01 (0.93, 1.10)	1.05 (0.95, 1.16)	1.03 (0.95, 1.13)

¹All models were adjusted for time from examination to conception, and interaction terms between waist circumference and breastfeeding duration categories and BMI and breastfeeding duration categories were included in the models. All women ceased breastfeeding in the category >6 mo, and no estimation was possible and thus not included in the table. P = 0.13 for interaction between prepregnancy waist circumference and BMI and breastfeeding duration categories in the fully adjusted model.

²Waist circumference and BMI mutually adjusted.

³Waist circumference and BMI mutually adjusted and further adjusted for race, study center, gestational age, child birth weight, and maternal prepregnancy height, age, parity, marital status, education, work status, smoking, physical activity, alcohol intake, and intake of fat, protein, carbohydrate, and fiber.

DISCUSSION

In this longitudinal study, a larger maternal prepregnancy waist circumference adjusted for BMI (an indication of more abdominal fat mass) was associated with a lower GWG and lower probability of breastfeeding. In contrast, a higher maternal prepregnancy BMI adjusted for waist circumference (an indication of increased lower body fat and muscle mass) was associated with a higher GWG and greater probability of breastfeeding.

In previous studies, BMI has mainly been used to determine prepregnancy maternal obesity, but this measure does not distinguish between regional fat distributions. Instead, waist circumference alone or adjusted for BMI has been positively associated with visceral fat mass and negatively with lower body fat mass (23, 39, 40). In contrast, adjusted for waist circumference, BMI has been positively associated with lower body fat, subcutaneous fat, and muscle mass and negatively associated with visceral fat mass (23, 39).

Pregnancy is a relatively short period with a rapid increase in fat mass. From 14 to 37 wk of gestation, maternal fat mass increases by ∼4 kg (15). Women gain fat mass, especially in the early period, which meets fetus-placental demands later in pregnancy (41), and the body tries to accomplish this by a small increase in insulin sensitivity (41) in very early pregnancy and an accentuated lipoprotein lipase activity in the femoral tissue (42). Fat is gained especially at the thighs and the trunk with individual variation (15, 43).

Researchers have hypothesized that abdominal fat accumulation is a consequence of limited storage capacity of triglyceride in the periphery and lower body (44, 45). Thus, women with more abdominal fat mass before pregnancy might be closer to their capacity of storage in the periphery, and they might store less, especially at the thighs, during pregnancy. In contrast, women with more lower body fat mass or muscle mass might have greater storage capacity and store more, perhaps at the thighs. Also, a greater peripheral fat mass accretion during pregnancy is observed among lean women, whereas a more central fat mass accretion is observed among obese women (46, 47).

Insulin sensitivity is inversely associated with abdominal fat mass in nonpregnant women (48), and lower insulin sensitivity before pregnancy is related to less fat accumulation early in pregnancy (49). Therefore, insulin resistance might explain part of our association between maternal abdominal fat mass and GWG. However, adjusting for prepregnancy insulin resistance did not change the associations. Thus, other pathways, such as restricted peripheral storage capacity, may be more likely to explain our findings.

Several studies have linked prepregnancy obesity defined by BMI to less breastfeeding (10, 12, 50). By including prepregnancy waist circumference, our results indicated that maternal abdominal fat mass was associated with reduced probability of ever breastfeeding but not duration. As a result of a low number of women in each breastfeeding duration category, lack of power may have limited our ability to study breastfeeding duration. In nonpregnant women, the adverse effect of obesity is mainly associated with increased abdominal fat mass and hence increased exposure to fatty acid and metabolic dysfunction (18, 51). It is unknown whether this may explain the lower probability of ever breastfeeding with larger waist circumference. However, maternal metabolism and obesity may interfere negatively with mammary gland development, which may affect lactogenesis, as suggested in animal studies (10); breast size increment during pregnancy (52); and later onset of stage II lactogenesis (53, 54). In women with gestational diabetes, heightened insulin resistance may interfere with the pathways for initiation of lactogenesis (55). However, in our study, adjustment for prepregnancy insulin resistance did not change our findings. Also, overweight/obese women have a blunted prolactin response to infant suckling (56). This might be even more pronounced with more abdominal fat mass because visceral fat is positively correlated with daily release of prolactin and a higher basal concentration (57). Moreover, greater abdominal fat mass and breast size might also be related to problems with proper positioning of the infant for breastfeeding as well as latching problems to a greater extent than lower body fat mass and muscle mass. Other factors that may reduce the probability of breastfeeding might be previous breastfeeding problems or racial differences, with black women tending to be less likely to initiate breastfeeding (58). However, in our study, adjustment or stratification by race or parity (primiparity compared with multiparity) did not support this. Future studies are needed to examine the mechanisms that might link maternal fat distribution to breastfeeding success.

Our findings indicate that the previously observed protective effect of breastfeeding on maternal abdominal fat mass later in life (5, 7, 59, 60) should be interpreted with caution because it might be confounded by maternal prepregnancy abdominal fat mass. Within the CARDIA study, changes in waist circumference within a 3-y interval from prepregnancy to a mean of 13 mo after delivery did not differ between women who breastfed and women who did not after adjustment for prepregnancy waist circumference, BMI, and other covariates (61). This indicates a limited effect of breastfeeding on abdominal fat mass after delivery when prepregnancy abdominal fat mass was taken into account. It is unknown whether this potential confounding also influences the previously observed protective effect of breastfeeding on long-term maternal metabolic diseases (62, 63). However, in a previous analysis based on the CARDIA study, a protective association for longer breastfeeding duration across all births with subsequent incidence of the metabolic syndrome was observed independent of prepregnancy components of the metabolic syndrome, among these waist circumference (64).

Limitations of our study include the period from prepregnancy measurements to conception because of the potential changes in waist circumference, BMI, and other covariates. However, the length of this period, which was accounted for in the analyses, did not differ systematically across waist circumference tertiles. Another limitation might be recall of breastfeeding behavior and GWG at the year 7 and 10 examinations. Our validation of GWG observed a fairly accurate maternal recall with no systematic differences across GWG or waist circumference. We were unable to validate breastfeeding information, but the categorical reporting might limit the degree of misclassification. In another study, 20-y recall of breastfeeding duration showed a modest median overestimation of only ∼2 wk (65). Although we were able to adjust for several potential confounding factors, confounding by unknown or unmeasured covariates, such as weight loss and change in behavioral factors from examination to conception, intention to breastfeed, and smoking behavior during pregnancy, cannot be ruled out.

The strengths of our study include the longitudinal cohort design with available standardized measurements of waist circumference and BMI before pregnancy in a diverse sample of women. Time to pregnancy is hard to predict, and enrolling women before pregnancy is difficult. Therefore, very few existing studies have collected prepregnancy measurements prospectively, and measured prepregnancy anthropometric factors are rarely obtained. In the CARDIA study, anthropometric factors were measured by certified technicians and thus more accurately determined than self-reported information that is used in most studies.

In conclusion, our findings show that maternal body size before pregnancy may be associated with GWG and ever breastfeeding differently depending on the location of the maternal fat mass and the amount of maternal overall fat mass. Women who had more abdominal fat mass before pregnancy than predicted on their BMI had a lower GWG and a lower probability of breastfeeding. In contrast, women who had less abdominal fat mass and a greater BMI had a higher GWG and a higher probability of breastfeeding. Thus, maternal fat distribution might be more informative than BMI alone in identifying women who may be at risk of high GWG and lack of breastfeeding. Future research should examine how maternal fat distribution affects postpartum changes in fat mass and to what degree prepregnancy or early pregnancy abdominal fat mass might explain a potential association of breastfeeding and GWG on long-term maternal abdominal fat mass and obesity-related diseases.