High-protein diet promotes a moderate postpartum weight loss in a prospective cohort of Brazilian women

Introduction 

Several studies have addressed the association between child-bearing and weight retention [1], [2], [3], [4]. Nearly 20% of women do not return to their prepregnancy body weight, and some women retain significant weight after delivery [2]. Many predictors have been associated with body weight change during the reproductive cycle: age, physical activity, smoking, lactation, and diet [1]. Also, the weight retained postpartum is highly correlated with weight gain during pregnancy [3], [4]. Despite the growing prevalence of obesity worldwide and the role of the reproductive period in weight gain, few studies have addressed dietary factors associated with weight change postpartum [5], [6].

Recommendations for lactating women are highly focused on dietary intake as it relates to breast-feeding, which poses the greatest energy demand during the reproductive cycle. However, nutritional recommendations so far have not included the composition of the diet as a strategy to decrease overweight.

For non-pregnant women, systematic reviews have shown that short-term high-protein (HP) diets seem to be more effective in promoting weight loss than diets rich in refined carbohydrates [7], [8]. According to Westman et al. [9], an HP diet led to a decrease in body weight in 41 of 51 overweight or obese volunteers who completed a 6-mo follow-up program following a very low-carbohydrate diet [9]. In that study, 95% of subjects lost weight, representing mean decreases of 10.3±5.9% in body weight and 2.9±3.2% in fat mass. Positive results were also observed in a clinical trial performed by Farnsworth et al. [10], in which the replacement of carbohydrate with protein seemed to have caused the preservation of lean body mass and the decrease in serum triacylglycerol concentrations. Two other randomized studies also showed that HP diets seemed to improve body composition, blood lipids [11], [12], and satiety [11].

Although these studies have provided some evidence that an HP diet improves body composition, there are no studies of this sort performed with postpartum women. As a matter of fact, lactating women are usually excluded from these types of trials [9], [13] due to possible safety concerns [14]. In general, HP diets are associated with a Western dietary pattern that is rich in red meat, saturated fat, processed and refined foods, and restricted in fruits and vegetables [15]. This dietary combination increases the risk of coronary heart disease [16], [17] and the incidence of cancer [18], [19], making safety a major concern in the long-term use of HP diets.

Nonetheless, it is interesting to note that the recommendation regarding the consumption of protein of 0.8g/kg of body weight daily (grams per kilogram per day) for non-pregnant women is increased to 1.3g · kg−1 · d−1 during lactation [20]. Thus, prospective studies looking at the effects of a protein-rich diet during the reproductive period seem to be a feasible and acceptable way of evaluating the role of dietary composition in body weight change postpartum. The aim of this study was to evaluate whether women with a protein intake above 1.2g · kg−1 · d−1 of body weight, considered an HP diet by the Food and Nutrition Board [21], would show greater body weight loss postpartum when compared with a low-protein (LP) diet group.

Materials and methods 

Data were obtained using a prospective cohort design with three follow-up evaluations (2, 6, and 9 mo after delivery). Baseline was considered at 0.5 mo postpartum. Data were gathered from May 1999 to April 2001 (15-mo recruitment and 9 mo of follow-up) at a primary health service center (Marcolino Candau Municipal Health Center, Rio de Janeiro City, Brazil). Participants were recruited 1) during prenatal routine care at the municipal health center, 2) at the time of the newborn's bacillus Calmette-Guérin routine immunization at the same health center, and 3) at the main maternity hospital in the area of study, 1 to 3 d after delivery. All study protocols met the guidelines of the research ethics committee and were approved by the Center for Collective Health Studies of the Federal University of Rio de Janeiro. Informed consent was obtained from all participants. Additional details can be obtained from Kac et al. [22], [23].

Eligibility criteria for enrollment in the cohort were an age 15 to 45 y, the first interview being conducted within 30 d after delivery (baseline was approximately at 0.5 mo postpartum), absence of chronic diseases, no history of actual multiple births, gestational age at delivery at least 35 wk, and a household address within the area of the local health center. A total of 709 women were invited to participate and 479 accepted (Fig. 1a). Women younger than 18 y (n=47) were excluded from analyses. Other exclusion criteria were subjects with an energy intake (EI) below 500kcal and above 5000kcal [24]. Two subjects with an EI above 6000kcal during pregnancy were excluded from analysis. To avoid a wide range of under- or over-reporting and provide more confidence for ours results, we repeated the statistical analyses described above, excluding women with an implausible ratio of reported EI to basal metabolic rate (MBR). We adopted 1.14 and 2.39 as cutoff points for under- and over-reporting, respectively [25].

View Large Image Download to PowerPoint

Fig. 1 (a) Recruitment and selection of the 430 Brazilian women in the cohort. (b) Study design and FFQ interview. FFQ, food-frequency questionnaire.

The pattern of loss to follow-up has been discussed previously [2], [22] and we demonstrate that the losses to follow-up were random with respect to all variables except age: women who dropped out from the study were younger. For the present analysis, loss to follow-up at the end of the study came to 34%, but more than 70% of women were followed until 6 mo (Fig. 1a). Body weight at baseline (P=0.30) and prepregnancy weight (P=0.27) were not different between those who were followed and those who dropped out. No differences of loss to follow-up according to the exposure variable were observed among women who were followed until 6 mo postpartum and women at baseline (P=0.61). Protein intake per body weight was 1.43±0.52g · kg−1 · d−1 versus 1.47±0.62g · kg−1 · d−1, respectively.

Results 

Women with an HP diet had shorter stature and were thinner at baseline compared with those using an LP diet. Mean age of the HP group was younger than that of the LP group; 16% of women in the HP group were at least 30 y old and 31% in the LP group were of the same age. Also, there was a lower proportion of whites in the HP group and a higher proportion of non-smokers in the LP group. No other differences were observed between diet groups (Table 1).

Table 1. Time since delivery, breast-feeding duration, and other characteristics at baseline∗ of Brazilian women according to postpartum HP or LP intake

	Variable	HP (n=261)	LP (n=160)	P†
	Time since delivery (d)	16.9±7.2	16.9±7.0	0.99†
	Breast-feeding (d)	81.9±59.6	81.0±63.7	0.88†
	Weight (kg)	59.4±10.4	67.9±11.6	<0.0001†
	Stature (cm)	157.7±6.9	159.8±6.1	0.0012†
	BMI (kg/m2)	22.5±3.3	25.6±4.2	<0.0001†
	Prepregnancy BMI (kg/m2)	20.9±3.2	23.7±4.1	<0.0001†
	Body fat (%)	28.2±6.3	32.4±6.1	<0.0001†
	Age (y)	25.5±6.0	26.7±5.9	0.05†
	Family income (US$)	359±356	356±384	0.94†
	Leisure physical activity (score)	1.3±1.3	1.3±1.2	0.80†
	Marital status			0.15‡
	Married	191 (73.2)	127 (79.4)
	Single	70 (26.8)	33 (20.6)
	Race			0.04‡
	White	84 (32.2)	70 (43.8)
	Mulatto	128 (49.0)	61 (38.1)
	Black	49 (18.8)	29 (18.1)
	Smoking			0.06‡
	Yes	49 (18.8)	19 (11.9)
	No	212 (81.2)	141 (88.1)
	Parity (n)			0.65‡
	1	151 (57.9)	89 (55.6)
	≥2	110 (42.1)	71 (44.4)
	Schooling (y)			0.08‡
	≤4	97 (37.2)	59 (36.9)
	5–8	104 (39.8)	50 (31.2)
	≥9	60 (23.0)	51 (31.9)

BMI, body mass index; HP, high protein (≥1.2g · kg−1 · d−1); LP, low protein (<1.2g · kg−1 · d−1)

∗ Baseline is the first evaluation that occurred approximately 0.5 mo postpartum. Values are means±SDs or numbers of subjects (percentages). † Student's t test. ‡ Chi-square test.

The total EI was higher in the HP group than in the LP group (2623 versus 1791kcal, respectively, P<0.0001). EI varied from 610 to 4294kcal (Table 2). Mean intakes of protein per kilogram of body weight (grams per kilogram per day) and total protein intake (grams per day) were higher in the HP group postpartum: 1.54±0.32g · kg−1 · d−1 in the HP group and 0.83±0.20g · kg−1 · d−1 in the LP group. Also, the density of the protein from the diet was higher in the HP diet. Women in the HP group had a greater consumption of all HP food items in the FFQ. Except for chicken and beans, the overall protein density, measured by the protein as a percentage of EI, was also greater in the HP group, but the consumption of a few food items was statistically different between groups (Table 2).

Table 2. Daily protein intake and selected foods in grams and density of intake for Brazilian women with HP intake (n=97) and LP intake (n=181) from the food-frequency questionnaire taken at 6 mo postpartum (n=278)∗

	Variable	Nutrient/food items			Intake as percentage of energy
		HP†	LP‡	P§	HP†	LP‡	P§
	Energy (kcal)	2623±591	1791±522	<0.01	2.62±0.59	1.79±0.52	<0.01
	Protein (g)	83.9±18.8	54.2±14.7	<0.01	32.2±3.33	30.8±4.88	<0.01
	Protein (g · kg−1 · d−1)	1.54±0.32	0.83±0.20	<0.01	0.60±0.11	0.48±0.12	<0.01
	Egg (units)	0.32±0.28	0.20±0.24	<0.01	0.13±0.12	0.11±0.14	0.47
	Chicken (piece)‡	0.59±0.41	0.53±0.37	0.31	0.24±0.18	0.32±0.27	<0.01
	Meat (steak)‡	1.05±0.67	0.59±0.37	<0.01	0.40±0.21	0.37±0.21	0.02
	Fish (fillet)‡	0.28±0.37	0.15±0.24	<0.01	0.11±0.14	0.08±0.13	0.06
	Milk (glass)‡	1.62±1.30	1.08±0.95	<0.01	0.63±0.50	0.62±0.56	0.93
	Beans (spoon)‡	2.04±1.07	1.50±0.95	<0.01	0.80±0.41	0.86±0.55	0.35

∗ Values are means±SDs. † Protein intake ≥1.2g · kg−1 · d−1. ‡ Protein intake <1.2g · kg−1 · d−1. Whole portion=1.00. § Student's t test.

The study showed that women who had an HP diet postpartum lost more total body weight (1.96±4.14kg) compared with women who had an LP diet (1.12±5.18kg, P=0.13). During the 9 mo postpartum, the mean body weight loss among all women was 1.34±4.9kg. More than 60% of women had lost weight at 2, 6, and 9 mo of follow-up. At the end of the study, 35% of women had a body weight that did not change or increased from baseline. Analysis of data regarding the changes in body weight postpartum indicates that women with an HP diet lost more weight than women with an LP diet at each time point after baseline (Fig. 2). Based on an average body weight loss at 6 mo postpartum, women with an HP diet lost more body weight than did women with an LP diet (1.69±3.7 versus 0.77±4.3kg, P=0.08). Comparing the mean weight at baseline with the mean weight at the 9-mo follow-up, we observed that the mean difference in body weight loss between the HP and LP groups was 0.84kg (P=0.17). These crude observations were later confirmed in the adjusted analysis presented in Table 3.

View Large Image Download to PowerPoint

Fig. 2 (a) Mean body weight difference in women with a high-protein diet (≥1.2g · kg−1 · d−1; n=97, 97, 97, 87; squares) or a low-protein diet (<1.2g · kg−1 · d−1; n=181, 181, 181, 164; circles) postpartum; significant time-by-group interaction. (b) Adjusted mean body weight loss by energy intake, energy intake by time, stature, percentage of body fat, percentage of body far by time, age, race, smoking, smoking by time, schooling, and schooling by time in women with a high-protein diet (≥1.2g · kg−1 · d−1; n=97, 97, 97, 87, squares) or a low-protein diet (<1.2g · kg−1 · d−1; n=181, 181, 181, 164; circles) postpartum; 1significant time-by-group interaction.

Table 3. Regression coefficients±SE estimates from longitudinal data analyses predicting body weight loss (n=278) for high (≥1.2g · kg−1 · d−1) and low (<1.2g · kg−1 · d−1) protein intakes postpartum through nested models (A to E)∗

			Parameter	Model A	Model B	Model C†	Model D‡	Model E‡
	Fixed effects
	Initial status	Intercept	Weight	62.2±0.59||	62.8±0.59||	66.7±0.87||	62.9±2.03||	62.7±2.03||
			HP/LP			−10.15±1.47||	−4.28±1.03||	−4.28±1.03||
	Rate of change	Intercept	Weight		−0.153±0.02||	−0.123±0.03||	−0.608±0.12||	−0.548±0.12||
			HP/LP			−0.104±0.05§	−0.313±0.06||	−0.316±0.06||
	Variance components
	Level 1 (within)			5.91±0.28||	5.59±0.25||	5.71±0.29||	5.14±0.26||	5.16±0.28||
	Level 2 (between)			147.5±10.21||	147.8±10.22||	131.6±11.33||	30.0±2.70||	30.0±2.70||
	Goodness-of-fit
	−2 res log likelihood			8337.8	8289.8	6225.3	5818.4	5799.1
	Akayke information criterion			8341.8	8293.8	6229.3	5822.4	5803.1

∗ Models were fitted using a random intercept. Model A is an unconditional means model. Model B is an unconditional growth model. Model C adds the interaction between linear time and the dummy variable protein by kilogram of weight (high protein ≥1.2g · kg−1 · d−1) as a fixed effect of the level-two predictor postpartum. Model D is a conditional model with the controlled effect of the high-protein diet predictor variable adjusted by energy intake, percentage of body fat at baseline, stature, age, race, smoking, schooling, and their interaction with time. Model E is a conditional model with the controlled effect of the high-protein diet predictor variable adjusted by energy intake, percentage of body fat, stature, age, race, smoking, and schooling and between the interaction of time with energy intake, percentage of body fat, smoking, and schooling. † Unadjusted model. ‡ Adjusted model. § P≤0.05. || P≤0.01.

Repeated measurement analysis built through nested models indicated a systematic variation of body weight within and between women in model A (Table 3). Model B showed, from a pseudo Rε2 statistic (pseudo Rε2=[(σε2 model A−σε2 model B)/σε2 model A]), that time accounts for only 5.3% of body weight loss (P<0.0001). The estimate of an intraclass correlation coefficient (ŷ=σ02/σ02+σ02) indicated that 96% of the total variation in body weight loss was attributable to difference between women. Therefore, inclusions of predictor variables of the second level explain most of this variation. Adjusted model E showed that there was a significant difference in body weight loss over time between the HP and LP groups (Table 3). The estimated average weight reduction was 316g/mo greater in the women with an HP diet postpartum. Women who had an HP diet lost more weight over time than women who had an LP diet postpartum (863 versus 548g/mo).

The EI:MBR ratios were 2.02±0.49 in the HP diet group and 1.24±0.32 in the LP group. About 18% of women with an HP diet were classified as over-reporters and 35% of the women with an LP diet were classified as under-reporters. None of the women with an HP and LP diet reported an EI:MBR below 1.14 or over 2.39, respectively. Further analyses showed that th exclusion of women with an EI/MBR below 1.14 or above 2.39 did not change the results that women with an HP diet lost more (324±0.06g) per month than women with an LP diet (P<0.0001).

Discussion 

Although the difference in body weight loss due to the HP diet was relatively small, this difference could become significant over time, constituting a possible strategy to lose and subsequently maintain an adequate body weight after pregnancy. Different from randomized trials [11], [13], the subjects of our study were not assigned to an HP or LP diet group. The pattern of HP intake that was observed among women was their usual diet. It means that adherence to the diet and significant difference of body weight loss can be maintained over time.

Mean body weight losses for both dietary groups were lower than the Institute of Medicine (IOM) [37] recommendation of about 2kg/mo after the first month postpartum for women with low or normal prepregnancy BMI. This fact could corroborate with idea that the diet was the primary determinant of the small weight change over time among women. Data not described in the present report showed that all covariables included in the multivariate model were not different (P>0.05) between diet protein groups at 6 mo postpartum, except prepregnancy BMI and %BF variables. Also, previous analysis of these data comparing intake during pregnancy with postpartum intake showed that those women with greater energy restriction from pregnancy to postpartum changed to the greatest concentration of protein by EI postpartum [38]. This suggests that it was the specific dietary change, and not the general lifestyle, that was associated with the greater weight change.

However, this is the first study examining the effects of an HP diet intake on postpartum body weight loss, and this dietary approach should be tested in controlled studies. There is increasing evidence regarding the advantages of HP diets in promoting body weight loss [9], [12], [13], [39] based on short-term studies as reviewed by Halton and Hu [40]. Also, an HP diet appears to be an effective tool to preserve body lean mass [10], [41]. Metabolic improvements on satiety [11], [35], [42], [43], insulin sensitivity [7], [39], and blood lipid profiles [8], [9], [13], [14] have also been observed.

A recent meta-regression analysis used to evaluate the effects of dietary composition on body mass and composition during energy restriction showed that HP diets improve body composition and present metabolic advantages when compared with LP diets [44]. Also, HP diets appear to have a significant effect on energy expenditure. A randomized, single-blind, three-way cross-over study with three isoenergetic interventions (with a low-fat and HP diet of pork meat, a low-fat and HP diet of soy protein, and a low-fat and high-carbohydrate diet) concluded that lower levels of carbohydrate increased energy expenditure [42]. A randomized cross-over trial compared the dynamic effects of two experimental diets, an HP diet and a normal-protein diet, and showed that the thermogenesis caused by an HP diet can explain the efficacy of this diet for weight loss [45]. Postprandial resting energy expenditures were 8 and 14kcal higher in the HP diet group after breakfast and lunch, respectively.

As a function of the greater nutritional requirements during lactation, additional protein intake has been recommended. The WHO recommends 0.75g of animal protein per kilogram of body weight [46], which corresponds to 0.91g of protein per kilogram in the Brazilian diet. The WHO recommends daily increase of approximately 16.0g of protein intake during the first 6 mo postpartum. Our results showed that women in the HP group consumed a greater amount of protein per kilogram of body weight than recommended by the WHO. Nonetheless, a much larger amount of protein was recently proposed by the IOM, with a recommended dietary allowance of 1.3g of protein per kilogram of body weight during lactation [20]. The mean amount of protein intake in our HP group was 1.54g, slightly larger than the new recommended values.

An important methodologic aspect of our analyses concerns the breast-feeding and physical activity effects on body weight loss. Although these variables were not considered confounders in our theoretical causal model, we decided to test them separately in model D because of previous results [23]. The present results from final the model D confirms previous analyses [23] that have shown an effect of breast-feeding on body weight loss (P=0.07) over time. No statistical significance was observed between leisure physical activity (P=0.84) and body weight loss, probably due to the low score of leisure time between the HP and LP groups postpartum. It is important to state that the inclusion of both variables in model D did not change the effect of HP diet on body weight loss (P<0.01).

The thermogenesis of food intake was not measured and there may have been a significantly higher contribution in the HP group than in the LP group. Another fact that might have contributed to increased energy expenditure was the higher percentage of lean body mass in the HP group than in the LP group (P<0.0001) at 9 mo postpartum. Furthermore, at the end of follow-up, %BF was lower among women with an HP diet (P<0.0001). According to Layman et al. [11], [12], HP diets have positive effects on body composition. It seems that women with an HP diet preserved more lean body mass than those with an LP diet.

Although most of our analyses were adjusted for EI, differences between nutrient densities in obese and non-obese groups in a validation study of the Brazilian FFQ questionnaire [27] indicated that the reporting of low EI does not predict a proportional reduction of nutrients. In that study, obese women reported lower carbohydrate and higher protein levels. Therefore, a possible information bias in the present study may support the alternative hypothesis of a greater difference in protein intake between groups. Women in the LP group were fatter than those in the HP group and overweight women tend to under-report more frequently [47], [48]. Under-reporting in the LP group may be partly responsible for the higher EI in the HP than in the LP group. Probably women in the LP group reported a higher protein intake than their usual diet. In this regard, it is important to state that all possible confounding variables concerning baseline weight differences between the LP and HP diets were considered at analyses and the effect of protein intake on body weight loss over time remained even after these variables were introduced, which confirms consistency of the data. It is also important to mention that the effect of a higher-protein diet on body weight loss remained even after an exclusion criterion based on EI/MBR was applied, although the LP group tended to under-report and the HP group over-report EI.

An FFQ was used to evaluate usual intake postpartum and was used to rank subjects according to their food consumption and nutrient intake. Studies have shown that the FFQ has acceptable validity and reproducibility [49], [50] and present correlation coefficients between 0.5 and 0.7 for protein intake, after adjustment for total caloric intake. Also, FFQs have been largely used to collect dietary data and measure nutritional assessment postpartum [51], [52], [53]. The strength of our prospective study is that dietary information was collected during follow-up, before any dietary diagnosis, and the HP diet was the usual diet of a large proportion of participants. Usually women postpartum were motivated subjects and had a desire to lose weight. The possibility of residual confounding explaining our findings cannot be completely rejected, but our final analysis included all possible variables that could explain the dietary option of the women, such as schooling, smoking, and race. Also, loss to follow-up, a major issue in prospective and randomized studies, was not high in our analysis, because data were based on questionnaires at 6 mo postpartum, when 72% of the sample was still present, when no differences of loss to follow-up were observed according to the exposure variable. Although this investigation had a considerably high dropout rate at 9 mo, a previous analysis showed that loss to follow-up in this cohort was random considering the baseline socioeconomic variables [20] and body weight at baseline and before pregnancy. Also, the women who were followed for 9 mo postpartum had the same food intake as women who dropped out from the cohort, except for the consumption of beer and alcohol, which was greater in the former group [38].

Some investigators have observed differences between diets groups as in the women in our study at 9 mo postpartum. Examples include higher calcium intake [54], heme iron [55], [56], and fatty acid and cholesterol [8] in the HP group than in the LP group (P<0.0001). Considering that the main objective of this study was to test the hypothesis of an effect of a higher-protein diet on body weight loss, this result will not be discussed. Only calcium, saturated fatty acid, and cholesterol intake as percentages of energy were higher in the HP group.

In conclusion, this study describes a positive relation between an HP diet and body weight loss, opening a pathway of investigation that may help establish a new strategy for women's diet postpartum. Additional randomized long-term studies are needed to determine the risks and benefits of this type of diet postpartum.