- Click here for explanation of classification scheme.

To examine the relationship between macronutrient intake early in pregnancy and the development of glucose intolerance.

• women between 24 and 29 wk of pregnancy
• singleton pregnancy
• >16 y of age
• spoke English
• had access to a phone
• planned to continue their care at 1 of the research sites
• signed an informed conscent. Study protocols approved by Institution Review Boards of the School of Medicine at the University of North Carolina at Chapel Hill and Wake Medical Center.

• racial group other than white or black
• second pregnancy
• preexisting diabetes
• no glucose screening data
• a high glucose screening results without an oral glucose tolerance test, thus unclassified
• missing dietary data
• missing body mass index
• women with total estimated energy intake less than the 5th percentile or greater than the 95th percentile

 

Recruitment:  Data from the Pregnancy, Infection and Nutrition (PIN) Study in central North Carolina.

Design :  Prospective cohort study

Blinding used:  not applicable 

Intervention: Women with GDM received diet treatment, insulin treatment, or both at both clinics and women with IGT were not treated at either clinic.

Statistical Analysis:

• bivariate analysis of sociodemographic characteristics and dietary variables with  glucose status was performed.
• Chi-square analysis was conducted to test the differences in categorical variables
• one-way analysis of variance with a Bonferroni correction for multiple comparisons was used for continuous variables
• multinominal logistic regresssion was used to calculate the likelihood of developing IGT or GDM compared with normal glucose tolerance while adjusting for potential confounders.  Variables that were associated with both glucose status and dietary variables and resulted in at least a 10% change in the IGT or GDM ß coefficients were retained in the models. 
• The partition method was used to evaluate the effect of adding dietary fat  or carbohydrates to the diet.  Nutrients were partitioned into calories from fat, carbohydrates, and protein. Variables were scaled to to represent risk per 100 kcal of the micronutrient.

            - Model 1 reported the effect of adding dietary fat to the diet

            - Model 2 reported the the effects of adding carbohydrates to the diet.

  - Models 3 through 5 evaluated the effect of theoretically substituting one macronutrient for another, allowing the researchers to compare isocaloric diets.  

• To evaluate the potential for effect modification by race or weight gain on the relation between diet and glucose status, researchers evaluated diet-by-race and diet-by-weight gain interaction terms in the models when P values were <=1.0.
• To facilitate interpretation of the results, the predicted probabilities of IGT and GDM were calculated from the coefficients of the models that showed significant dietary effects (Models 2 and 3). The predicted probabilities were calculated for the various dietary scenarios.

Timing of Measurements

• several questionnaires were self-administered at the time of recruitment including a food-frequency questionnaire
• contacted by telephone within 2 weeks of recruitment regarding current and pregravid health behaviors and sociodemographic characteristics 
• dietary intake information assessed during the second trimester
• maternal height measured during clinic visits
• physical activity - participation in a regular or strenuous physical activity was assessed 3 months before becoming pregnant and during the first and second trimesters. 
• cigarette smoking habits assessed during the first 6 months of the pregnancy.

Dependent Variables

Glucose intolerance:

• information obtained from hospital computer data base and medical charts
• glucose tolerance status involved two-steps: 1.) initial screening test measuring the plasma glucose concentration 1 h after a 50-g glucose challenge test (random screen that does not require fasting state); 2.) site specific protocols established the cutoffs for the follow-up testing (these were different for each clinic because disagreement existed about which cutoff represented risk).
• OGTT - the test was conducted in the fasting state with glucose analysis performed at fasting and 1, 2, and 3h after the oral glucose load. A value of >=140 mg.dL at the Univ of North Carolina sites and a value of >=130 mg/dl at the Wake Medical Center sites indicated the need for a full 3 h,100 g oral glucose-tolerance test. Tests are performed on serum samples with the use of with the use of the glucose oxidase method. The Carpenter and Coustan cutoff of 95mg/dL for fasting, 180 mg/dL for 1-h, 155 mg/dL for 2-h, and 135 mg/dL for 3-h were used for abnormal values.
• IGT was defined as one abnormal value form the OGTT.
• Normal glucose tolerance was defined as having a normal or high result on the glucose challenge test but no high values on the OGTT.
• GDM were defined by 2 abnormal values on the OGTT.

  Dietary data - variables of interest: macronutrients and total energy.   

• dietary intake (2nd trimester) assessed by using a modified food frequency questionnaire originally developed by Block et al. Modifications were made to include local foods, make the questionnaire specific to the pregnancy time period, and include more food specific portion sizes.
• the validity of the food-frequency questionnaire in the PIN population was assesed among 99 women by comparing nutrient results from the food frequency questionnaire with three 24-hr dietary recalls colllected at random on nonconsecutive days.
• the deattenuated Pearson correlation coefficients for total energy and micronutrients intake for the food frequency questionnaire and the 24-h recall  were 0.35 for total energy, 0.43 for fats, 0.44 for protein, and 0.26 for carbohydrates. These results for energy, fat, and protein were similar to a recent comparison of a food frequency questionnaire and 10 day food record intakes in pregnant women; the results for carbohydrates were lower (energy, r= 0.24; fats, r=0.48; protein, r=0.55; and carbohydrates, r=0.49).

Independent Variables

• prepregnancy BMI (kg/m²) was classified according to guidelines established by the Institute of Medicine: underweight, <19.8;normal weight, 19.8-26; overweight, >26-29; and obese, >29. Recalled prepregnancy weight was obtained from the medical record or the screening questionnaire. If the first weight measurement was after 15 weeks of gestation, a prepregnancy weight could not be imputed. In those cases, underweight and normal weight were combined and used as the reference category. Height was measured during clinic visits.
• maternal age
• weight gain
• physical activity-participation in a regular or strenuous physical activity was assessed 3 months before becoming pregnant and during the first and second trimesters. 
• race-self identified during phone call or taken from medical chart
• family income was represented as a percentage of the poverty index according to U.S. Bureau of Census 1996 poverty guidelines
• cigarette smoking habits assessed during the first 6 months of the pregnancy.

Control Variables

 

Initial N: 2898 pregnant women

Attrition (final N):  1698 used in analysis

Age: See Table I

Ethnicity: white or black

Other relevant demographics:

Anthropometrics

Location: Department of Nutrition and Maternal and Child Health, School of Public Health and the Carolina Population Center, University of North Carolina, Chapel Hill.

 

The overall prevalence of IGT in the cohort was 2.6%, and that of GDM was 5.2% (see Table I).

Table 1. Characteristics of the participants in the Pregnancy, Infection, and Nutrition Study by glucose tolerance status¹ 

Variable	Normal(n=1565)	IGT(n=44)	GDM(n=89)	p2
Race (%)	-	-	-	0.002
White	60	73	76	-
Black	40	27	24	-
Marital status (%)	-	-	-	0.008
Single	41	25	25	-
Married	52	61	65	-
Other	7	14	10	-
Poverty index, n=1507 (%)	-	-	-	0.78
<185%	53	59	57	-
185-350%	20	22	17	-
>350%	27	19	26	-
Smoked, n=1594 (%)	25	25	24	0.97
Physical activity, n=1600 (%)	-	-	-	-
During 3 mo before pregnancy	25	28	17	0.22
During first trimester	16	16	10	0.33
During second trimester	9.5	9.3	3.6	0.19
Pregravid BMI (kg/m2)	25.0 ±6.53	28.4±7.9	30.1±7.7	0.001
BMI categories (%)	-	-	-	0.001
Underweight or	-	-	-	-
normal weight (BMI<26)	67	45	35	-
Overweight (BMI>26-29)	11	16	15	-
Obese (BMI>29)	22	39	50	-
Mothers' age (y)	26±6.2	29±6.0	28±5.6	0.001
Mothers'education (y)	13.6±2.9	13.7±2.2	14.0±2.8	0.54
Parity, n=1692	0.8±1.1	1.1±1.1	0.9±0.9	0.16
Mothers' height (m)	1.65±0.07	1.64±0.07	1.64±0.07	0.07

¹ IGT,impaired glucose tolerance; GDM, gestatational diabetes mellitus

² P values are from overall chi-square test for categorical variables or from ANOVA  with Bonferroni correction for continuous variables.

³ Mean±SD

 

Only the intake of carbohydrates and fat differed by glucose status (Table 2).

Table 2 Mean dietary composition by glucose tolerance status of women in thr Pregnancy, Infection, and Nutrition Study ¹

	Normal	IGT	GDM	P2
Total energy (kcal)	2603±995	2431±844	2572±939	0.51
Protein (% of energy)	14±2.8	15±2.5	14±2.7	0.15
Carbohydrate (% of energy)	53±7.4	50±6.5	51±7.1	0.0004
Fat (% of energy)	33±6.3	35±5.9	35±5.9	0.001

¹ Mean ±SD. n=1698. IGT, impaired glucose tolerance; GDM, gestational diabetes mellitus.

² P values are from an overall chi-square test for macronutrients or from an ANOVA with Bonferroni correction for total energy.

The relative risk ratios and 95% confidence intervals are presented in Table 3. Model 2: showed that adding 100 kcal carbohydrates was associated with a 12% decrease in risk of IGT and a 9% decrease risk of GDM; Model 3 showed that substituting fat for carbohydrate resulted in significant increase in i risk of both IGT and GDM.

Table 3 Relative risk ratios (RRs) of impaired glucose tolerance (IGT) and gestational diabetes mellitus (GDM)¹

	IGT		GDM
	RR(95% CI)	P	RR(95% CI)	P
Additional models: calories of micronutrients2	-	-	-	-
Model 1: Adding fat	1.1(0.97,1.30)	0.14	1.1(1.0,1.20)	0.06
Model 2: Adding carbohydrate	0.9(0.79,0.98)	0.02	0.9(0.85,0.98)	0.01
Substituting models: percentage of macronutrients3	-	-	-	-
Model 3: Substituting fat for carbohydrate	1.1(1.02,1.12)	0.006	1.1(1.02,1.10)	0.002
Model 4: Substituting fat for protein	1.0(0.93,1.15)	0.58	1.0(0.94,1.10)	0.68
Model 5: Substituting carbohydrates for protein	0.9(0.90,1.06)	0.41	0.9(0.90,1.03)	0.24

 ¹n=1698. All models were adjusted for BMI, maternal age, and race.

² Additional models are per 100 kcal for carbohydrate.

³ Substitution models are adjusted for total calories and per 1%

Other Findings

• The predicted probabilities of IGT and GDM were higher for white women than for black women at every fat intake level.    
•  The predicted probabilities of IGT and GDM decreased with the additional energy from carbohydrates.

 

• In an isocaloric diet, an increase in carbohydrate intake as a percentage of energy with a simultaneous decrease in fat intake as a percentage of energy significantly reduced the risk of glucose intolerance.
• Increasing carbohydrates without decreasing fat and thus not controlling for total energy intake also significantly reduces the risk of both IGT and GDM.    
• the validity of the food-frequency questionnaire in the PIN population was assesed among 99 women by comparing nutrient results from the food frequency questionnaire with three 24-hr dietary recalls colllected at random on nonconsecutive days.
• The interaction terms were not significant.
• Separate predicted probabilities were presented for black and white women because the prevalences of IGT and GDM were significantly different by race.

Limitations

• the assessment of dietary exposure could result in a bias estimate effect
• the correlation coefficient for carbohydrates was 0.26. Validation studies have shown that correlation coefficients for carbohydrates was 0.26. Validation studies have shown that correlation coefficiients <0.4 can attenuate an association between an exposure and an outcome, and thus the estimate of a carbohydrate effect may be conservative.
• the effect of residual confounding from sources not measured could influence the estimates
• the analysis population somewhat underrepresented black women, which could influence the estimates.      

Government:	NIGMS, NICHD
University/Hospital:	Dept. of Nutrition Clinical Nutrition Research Center

The limitations and critique of the study, as stated by the authors appear to be very appropriate.

Analytical longitudinal surveys refer to what epidemiologists term prospective or cohort studies. A Cohort Study is a study in which patients who presently have a certain condition and/or receive a particular treatment are followed over time and compared with another group who are not affected by the condition under investigation. Studies of this kind provide a better opportunity than one time cross sectional studies to examine whether certain behaviors do in fact lead to (or cause) the disease.