Relationship Between Carbohydrate Intake and Oral Glucose Tolerance Test Results Among Pregnant Women

Abstract

Aims:

Evaluate the relationship between self-reported carbohydrate intake and oral glucose tolerance test (OGTT) results in pregnancy.

Methods:

We measured carbohydrate intake using 24-hour dietary recall and performed a 2-hour 75-gram OGTT in 95 pregnant women with gestational diabetes (GDM) risk factors at a median of 26 weeks’ gestation. We tested for associations between carbohydrate intake in the 24 hours preceding the OGTT and 60-minute OGTT glucose, glucose at other timepoints, and glucose area under the curve (AUC) using linear regression, with adjustment for potential confounders.

Results:

We observed an inverse linear relationship between carbohydrate intake (median 237 grams [interquartile range: 196, 303]) and 60-minute OGTT glucose. For every 50 gram reduction in carbohydrate intake, there was an 8.9 mg/dl increase in 60-minute OGTT glucose (P<0.01) in an adjusted model. Lower carbohydrate intake was also associated with higher 30-(adjusted β=−6.5 mg/dl, P<0.01) and 120-minute OGTT glucose (adjusted β=−8.1 mg/dl, P=0.01) and AUC (adjusted β=−767,P<0.01).

Conclusions:

Carbohydrate intake in the day preceding an OGTT in pregnancy is associated with post-load glucose values, with lower carbohydrate intake predicting higher glucose levels and higher carbohydrate intake predicting lower glucose levels. Carbohydrate restriction or excess before an OGTT may affect GDM diagnosis.

1. Introduction

Pregnant women in the United States (US) undergo universal screening for gestational diabetes mellitus (GDM) between 24 and 28 weeks’ gestation; the diagnosis of GDM is made based on the results of a diagnostic oral glucose tolerance test (OGTT) (1,2). In 2007, the Fifth International Workshop-Conference on GDM recommended a moderate to high carbohydrate diet (≥150g/day for 3 days) prior to OGTTs among pregnant women to obtain accurate diagnostic results (3). The idea of a preparatory OGTT diet dates back to a classic study in non-pregnant individuals, which showed that restricted carbohydrate intake prior to an OGTT was associated with impaired glucose tolerance (4).

Today, GDM guidelines from the American Diabetes Association (ADA) (2020) and the American College of Obstetricians and Gynecologists (ACOG) (2018) do not include recommendations regarding a preparatory diet, and the Endocrine Society (2013) states that an OGTT should be performed “without having reduced usual carbohydrate intake” (1,2,5).Although current US medical society guidelines do not include a preparatory diet, such diets are sometimes recommended in clinical practice (6–9). In light of variability across centers in instruction regarding dietary intake in the days prior to an OGTT as well as the current popularity of low carbohydrate diets (10), we sought to investigate the impact of carbohydrate consumption in the day preceding an OGTT on post-load glucose levels in pregnant women.

Previous studies evaluating the relationship between carbohydrate intake and subsequent OGTT blood glucose among pregnant women have demonstrated mixed results: some studies showed no relationship (11–14), while others found an inverse relationship between carbohydrate intake and OGTT glucose (15–18). Limitations of previous interventional studies that have manipulated carbohydrate content and did not find a relationship include either failure to account for carbohydrate intake in the control arm or lack of a meaningful difference in carbohydrate intake between comparison groups (11–14). Some studies demonstrating a significant relationship between carbohydrate intake and OGTT glucose levels have evaluated dietary intake over several months as opposed to on the days prior to the OGTT (15,17,18). Thus, previous studies have not adequately addressed whether real-life, unmanipulated carbohydrate intake on the day prior to an OGTT has an association with the results of a diagnostic OGTT in pregnancy.

Our goal was to assess the relationship between carbohydrate intake prior to a diagnostic OGTT and OGTT blood glucose levels among pregnant women participating in a longitudinal study of glucose metabolism in pregnancy. Based on previous studies, we hypothesized that lower carbohydrate intake on the day prior to an OGTT would be associated with higher post-load OGTT blood glucose values.

2. Subjects, Materials and Methods

Data are derived from the Study of Pregnancy Regulation of INsulin and Glucose (SPRING), a longitudinal prospective study designed to examine insulin physiology and glucose metabolism in pregnant women at high risk for GDM. In this interim analysis for a secondary purpose, we analyzed data from the visit at 24–32 weeks’ gestation (visit 2) available as of December 2020.

2.1. Participants

Pregnant women, age 18–44 years, were recruited from the obstetric practice at Massachusetts General Hospital (MGH) and through paper and electronic advertisements in the Boston area between 2016 and 2020. At the first visit (in the first trimester), women completed a survey reporting their past medical history, past pregnancy history and family history. Weight was measured at each study visit and height was measured at the first study visit. Women included in the study were required to have risk factors for GDM (e.g., overweight body mass index [BMI] plus the presence of one additional risk factor as described by the ADA, GDM in a previous pregnancy regardless of BMI, or family history of diabetes mellitus regardless of BMI) (2). Women with a history of bariatric surgery, pre-existing diabetes mellitus, with a HbA1c ≥ 6.5% at the first trimester study visit, or taking medications that affect glucose tolerance were excluded. The Partners Committee on Human Research (Institutional Review Board) approved the study, and all participants gave written informed consent.

2.2. Study Procedures

SPRING participants underwent a standardized 2-hour 75-gram OGTT at 7–14 weeks’ gestation (visit 1), 24–32 weeks’ gestation (visit 2) and 6–24 weeks postpartum (visit 3). Participants with elevated post-load glucose in the first trimester (1-hour glucose ≥180 mg/dl, 2-hour glucose ≥153 mg/dl) did not undergo an OGTT later in pregnancy and were referred to their obstetric provider for GDM management (Supplemental Figure 1). The fasting glucose level in the first trimester that excluded participation for later OGTTs changed during the study from 92 mg/dl to 100 mg/dl because of emerging evidence that fasting glucose thresholds for GDM diagnosis at 24–28 weeks’ gestation should not be applied to the first trimester (19). Normal first trimester results were communicated to obstetric providers and if requested, to the participant; abnormal results were communicated to both the obstetric clinical team and the participant. Data from the visit at 24–32 weeks’ gestation (visit 2) were used for the primary analysis in the current study, as this is the gestational age when OGTTs are routinely performed in pregnancy for diagnosis of GDM (2).

We performed 24-hour dietary recalls at the time of the OGTT using the Automated Self-Administered 24-Hour Dietary Assessment Tool (ASA24), a validated self-administered web-based dietary recall tool from the National Cancer Institute, along with other questionnaires (20–23). The ASA24 provides a detailed analysis of 65 nutrients and 37 food groups: the system is based upon the United States Department of Agriculture’s Automated Multiple-Pass Method and the food code assignments are linked to the USDA Food and Nutrient Database for Dietary Studies (24). The ASA24 has been validated in healthy individuals across the body weight/BMI spectrum (20–23). Women completed the dietary recall during the OGTT and reported their food intake from the previous day. Women were asked to fast (other than water) for at least eight hours prior to the OGTT but were not given other dietary or physical activity instructions. During the OGTT, blood samples were drawn fasting, 30 minutes, 60 minutes and 120 minutes following consumption of the 75-gram oral glucose load. Women who underwent an OGTT at the second visit and who completed the ASA24 were included in the analysis (Supplemental Figure 1).

2.3. Lab analyses

Blood samples for glucose were collected in sodium fluoride tubes. Mimicking clinical practice, samples were sent immediately to the MGH core clinical laboratory for analysis. Glucose was measured using the hexokinase method with the Cobas 8000 modular analyzer (Roche Diagnostics, intra-assay coefficient of variation (CV) < 1%). We used the International Association of the Diabetes in Pregnancy Study Groups’ (IADPSG) 2010 criteria to define GDM at the second study visit (fasting glucose ≥92, 1-hour glucose ≥180 mg/dl, 2-hour glucose ≥153 mg/dl) (25).

At the beginning of the study, fresh insulin samples were assayed using a 2-site electrochemiluminescent immunoassay on the Roche automated platform (N=30) (Assay1, intra-assay CV <5%). Starting in July 2018, frozen insulin samples were assayed using the Beckman Coulter Access Ultrasensitive Insulin Assay (N=65) (Assay 2, intra-assay CV <5%). Results on duplicate samples from the two assays were highly correlated (R²=0.9723), but the values from assay 2 were slightly lower. Using these data, a transformation was performed so that results from both insulin assays could be used in the same analyses (Assay 2 result = 0.8886 [Assay 1 result] − 2.2909).

2.4. Statistical Analyses

To compare baseline characteristics in our participants, women were grouped by median carbohydrate intake into a “low carbohydrate intake” and a “high carbohydrate intake” group. We compared continuous variables between the low and high carbohydrate intake groups using a Wilcoxon rank-sum test and categorical variables using a chi-squared test. We also performed a Pearson correlation test to evaluate the relationship between carbohydrate intake at visit 1 and carbohydrate intake at visit 2. We used linear regression models to assess the relationship between the primary exposure variable, total carbohydrate intake over the twenty-four hours preceding the OGTT, and the pre-specified primary outcome variable for this analysis, 60-minute post-OGTT blood glucose level. We chose the 60-minute timepoint for our primary outcome because, amongst the three time points that are used for GDM diagnosis, it typically demonstrates the peak effect of the oral glucose load (26). Secondary outcomes included 30-minute post-OGTT blood glucose level, 120-minute post-OGTT blood glucose level, and glucose area under the curve (AUC). We also included the diagnosis of GDM (IADPSG criteria) as an exploratory outcome. We assessed the relationship between total carbohydrate intake in grams and both primary and secondary outcomes using linear regression (or logistic regression for the GDM outcome).

In regression models, the exposure variable, total carbohydrate intake (in grams) over the twenty-four hours preceding the OGTT, was divided by 50 to assist with model interpretation. Three models types were constructed: 1) no adjustment, 2) adjustment for total caloric intake over 24 hours (kcals/24 hours), and 3) adjustment for age, BMI (measured at the study visit), gestational age, race/ethnicity, education (divided into three groups: completion of middle school or high school, completion of college, and completion of graduate school), and total caloric intake over 24 hours (kcals/24 hours). In addition, we created an adjusted model, which categorized BMI as underweight (<18.5), normal (18.5–24.9), overweight (25–29.9) and obese (≥30). Models were checked for normality of the residuals and homoscedasticity.

We performed three sensitivity analyses evaluating the primary outcome, excluding different groups of participants: 1) individuals at <10^th percentile of total caloric intake, 2) women with multiple gestation, and 3) women with a history of GDM in a prior pregnancy. We also conducted a sensitivity analysis using the percent carbohydrate in the diet [(grams of carbohydrate * 4)/ Total kcal] as the exposure variable.We also assessed the impact of physical activity on our primary analyses by adding amount of time spent sitting per day to our regression models in the subset of participants who reported this information.

In post-hoc analyses, we compared insulin secretory response (Stumvoll 1^st phase estimate) between low and high carbohydrate intake groups using a Wilcoxon rank sum test to understand the mechanism behind our findings (27). In addition, we conducted analyses to evaluate whether potential knowledge of prior glucose values altered reported carbohydrate intake at visit 2 by testing for a correlation between the 60-minute post-OGTT blood glucose values at visit 1 and the difference between the carbohydrate intake at visit 1 and visit 2. We also tested for a correlation between glucose at visit 1 and carbohydrate intake at visit 2; we evaluated the correlation between carbohydrate intake and 60-minute post-OGTT blood glucose values at both visit 1 and visit 2. Lastly, we conducted a two-sample Wilcoxon rank-sum test to evaluate the difference in carbohydrate intake between women with and without a history of GDM to determine if women with previous GDM reported less carbohydrate intake.

Statistical analyses were conducted in STATA/IC version 16 (College Station, TX).

3. Results

3.1. Participant Characteristics

Table 1 displays participant characteristics. Twelve women were diagnosed with GDM at the first-trimester visit and did not undergo an OGTT at the second visit and were therefore ineligible for our study (Supplemental Figure 1). Participants were studied at a median 26.3 weeks’ gestation (interquartile range (IQR) 25.0, 27.4). Labs were drawn at a median time of 8:30 am (IQR 7:50 am, 9:20 am). Women recruited from outside the obstetric practice had higher education, lower body mass index, and consumed fewer total calories than those recruited from within the practice (Supplemental Table 1). Median carbohydrate intake over twenty-four hours was 236.9 grams (IQR 196.0, 302.6) at visit 2. Carbohydrate intake at visit 2 was significantly correlated with carbohydrate intake at visit 1 (r=0.31, P=0.005)

Table 1:

Characteristics of Study Participants

Variable	All participants (N=95) Median (IQR) N (%)	Low carbohydrate (N=48) Median (IQR) N (%)	High carbohydrate (N=47) Median (IQR) N (%)	P-value (Low vs. High Carbohydrate Intake)
Age (years)	33.0 (30.9, 35.8)	33.4 (31.5, 36.6)	32.2 (28.7, 34.8)	0.06
Race/Ethnicity				0.89
Hispanic/Latina	19 (20.0%)	9 (19%)	10 (21%)
White non-Hispanic	57 (60.0%)	29 (60%)	28 (60%)
Black non-Hispanic	11 (11.6%)	5 (10%)	6 (13%)
Asian non-Hispanic	4 (4.2%)	3 (6%)	1 (2%)
Other non-Hispanic	4 (4.2%)	2 (4%)	2 (4%)
Education completed				0.14
Middle school or high school	11 (11.6%)	3 (6%)	8 (17%)
College	42 (44.2%)	25 (52%)	17 (36%)
Graduate school	42 (44.2%)	20 (42%)	22 (47%)
First trimester BMI (kg/m2) *	29.2 (25.4, 33.5)	29.5 (25.8, 33.2)	29.1 (23.9, 33.6)	0.72
Personal History of Gestational Diabetes	11 (11.6%)	6 (12%)	5 (11%)	0.78
Family History of Diabetes	34 (35.8%)	18 (38%)	16 (34%)	0.73
Primigravid	32 (33.7%)	16 (33%)	16 (34%)	0.94
Multiple gestation	5 (5.4%)	4 (9%)	1 (2%)	0.18
Gestational age	26.3 (25.0, 27.4)	26.7 (25.7, 27.5)	26.0 (24.9, 27.3)	0.16
BMI at OGTT visit (kg/m2)	31.1 (27.6, 36.3)	31.2 (27.9, 35.6)	31.0 (26.0, 36.5)	0.91
Gestational weight gain (lbs) **	13.2 (8, 17.4)	13.3 (6.7, 17.1)	13.0 (10.0, 17.4)	0.39
Time spent fasting prior to OGTT (hours)	11.9 (10.6, 13.4)	12.1 (10.8, 13.0)	11.6 (10.3, 13.5)	0.26
Total Calories (kCal/24 hrs)	1968 (1661, 2486)	1689 (1497, 1893)	2471 (2109, 2764)	<0.001
Total Carbohydrate (g/24 hrs)	236.9 (196.0, 302.6)	196.2 (167.9, 221.8)	302.6 (278.3, 354.3)	<0.001
Dietary Percent Carbohydrate ᵻ	48.5 (43.9, 57.1)	46.2 (39.4, 51.5)	52.8 (46.7, 60.2)	<0.001
Fasting glucose (mg/dl) +	79.0 (75.0,84.0)	79.5 (75.0, 84.5)	79.0 (75.0, 84.0)	0.68

BMI=body mass index, OGTT=oral glucose tolerance test

^*BMI measured at first study visit during first trimester

^**Weight gain reported as the difference between measured weight at the first study visit and measured weight at the second study visit

^ᵻPercent carbohydrate of the diet defined as (grams of carbohydrate * 4)/kcals * 100

⁺Fasting glucose at the time of the second study visit OGTT

While there were no statistically significant differences in participants with carbohydrate intake below versus above the median, those with lower carbohydrate intake tended to be older (P=0.06) than those with higher intake. Participants with low carbohydrate intake had significantly lower total energy consumption than those with high carbohydrate intake (P<0.001). Participants in both groups had similar fasting blood glucose levels. Thirty-minute glucose levels were significantly higher in the low carbohydrate group (median 134 mg/dl [IQR 121, 155]) than in the high carbohydrate group (median 124 mg/dl [IQR 110, 145] P=0.01). The 60-minute and 120-minute demonstrated a similar trend (60-minute glucose levels: low carbohydrate group (median 144 mg/dl [IQR 123, 161], high carbohydrate group (median 134 mg/dl [IQR 108, 157], P=0.13); 120-minute glucose levels: low carbohydrate group (median 119 mg/dl [IQR 101, 135], high carbohydrate group (median 108 mg/dl [IQR 91, 127], P=0.06).

3.2. Primary outcome

We observed a graded inverse linear relationship between carbohydrate intake and 60-minute post-OGTT blood glucose level (r=−0.28, P=0.007, Figure 1) in pregnant women. In the fully-adjusted model (model 3), there was an 8.90 mg/dl (P<0.01) increase in the glucose level for every 50-gram reduction in carbohydrate intake reported (Table 2). Changing BMI to a categorical co-variate rather than a continuous co-variate did not change our findings (adjusted β =−10.0, P <0.01). Results remained similar when BMI at visit 1 and gestational weight gain replaced BMI at study visit in the adjusted model.

Figure 1.

Relationship between Carbohydrate Intake and 60-minute OGTT Blood Glucose Level

Scatter plot demonstrates the relationship between 60-minute OGTT glucose and carbohydrate intake over 24 hours prior to the OGTT as ascertained by the ASA24. Circles represent raw data. There was a significant inverse correlation between carbohydrate intake in the preceding 24 hours and 60-minute OGTT glucose (r=−0.28, P=0.007).

Table 2.

Association between Carbohydrate Intake and Post-Load Glucose Levels

	60-minute post OGTT glucose		30-minute post OGTT glucose		120-minute post OGTT glucose		Glucose Area Under the Curve
Model	Coefficient (mg/dl), 95% Confidence Interval	P-value	Coefficient (mg/dl), 95% Confidence Interval	P-value	Coefficient (mg/dl), 95% Confidence Interval	P-value	Coefficient, 95% Confidence Interval	P-value
Unadjusted	−4.85 (−8.30, −1.39)	0.006	−2.96 (−5.53, −0.40)	0.024	−5.58 (−8.92, −2.25)	0.001	−460.65 (−750.02, −171.29)	0.002
Caloric Intake	−9.48 (−15.57, −3.39)	0.003	−7.39 (−11.87, −2.90)	0.002	−7.22 (−13.19, −1.25)	0.018	−790.84 (−1303.28, −278.39)	0.003
Confounder	−8.90 (−15.42, −2.34)	0.008	−6.52 (−11.32, −1.73)	0.008	−8.06 (−14.37, −1.75)	0.013	−766.85 (−1318.49, −215.22)	0.007

In linear regression models described here, the primary exposure was total reported carbohydrate intake in the 24 hours preceding the OGTT (grams). The coefficients for each outcome (60-minute post-load glucose, 30-minute post-load glucose, 120-minute post-load glucose, and glucose area under the curve) are given per 50-gram increase in carbohydrate intake.

Unadjusted model: No adjustments

Caloric Intake Model: Adjustment for total caloric intake (kCals/24 hours)

Confounder Model: Adjustment for age, BMI, gestational age, race/ethnicity, education, total caloric intake (kcals/24 hours)

3.3. Secondary outcomes

Lower carbohydrate intake was associated with higher 30-minute post-OGTT blood glucose (adjusted β=−6.5 mg/dl, P<0.01), 120-minute post-OGTT blood glucose (adjusted β=−8.0 mg/dl, P=0.01), and glucose AUC (adjusted β=−767, P<0.01) (Table 2). Figure 2 displays the predicted glucose levels across the OGTT for the fully adjusted models in women with carbohydrate intake at the 25^th percentile and at the 75^th percentile.

Figure 2.

Predicted OGTT Glucose Values by Carbohydrate Intake

This figure displays the predicted glucose values at each OGTT timepoint (fasting, 30 minutes, 60 minutes, 120 minutes) adjusting for age, BMI, gestational age, race/ethnicity, education, and total caloric intake (kCals/24 hours) for pregnant women at the 25^th percentile for carbohydrate intake and at the 75^th percentile for carbohydrate intake.

Higher carbohydrate intake was not significantly associated with a lower odds of GDM (IADSPG criteria), but the odds ratio point estimate was less than 1 (adjusted OR=0.79, 95% CI=0.46–1.36 P=0.40).

3.4. Sensitivity analyses

Results for the primary outcome remained similar after exclusion of those at the lower end of reported caloric intake (<10^th percentile total kcals) (adjusted β=−8.2 mg/dl). Similarly, results remained unchanged (adjusted β=−7.3 mg/dl) after exclusion of women with multiple gestation. Results were slightly attenuated when we omitted women with a history of GDM (adjusted β=−6.1 mg/dl). Results remained similar when the exposure variable was percent carbohydrate intake: there was a 0.89 mg/dl increase in blood glucose per 1% decrease in carbohydrate intake in the adjusted model (P<0.01). Sedentary behavior was reported in 82 of our 95 participants; adding hours spent sitting per day to our regression models in this subset of participants did not affect results (adjusted β=−8.9).

The median carbohydrate intake for women with a history of GDM was 234.4 grams and the median carbohydrate intake for women without a history of GDM was 238.9 grams (P=0.33). There was no significant correlation between the 60-minute post-OGTT blood glucose level at visit 1 and the difference in carbohydrate intake between visit 1 and visit 2 (r=0.002, P=0.98). There was no significant correlation between glucose at visit 1 and carbohydrate intake at visit 2 (r=−0.17, P=0.13). Both carbohydrate intake at visit 1 and carbohydrate at visit 2 appeared to be negatively correlated with the respective 60-minute post-OGTT blood glucose levels but the relationship at visit 2 (r=−0.25, P=0.02) was stronger than the relationship between carbohydrate intake and 60-minute post-OGTT blood glucose at visit 1 (r=−0.19, P=0.08).

3.5. Post hoc analyses

All participants had data on insulin secretory response, 48 participants in the low carbohydrate group and 47 participants in the high carbohydrate group. We found higher insulin secretory response (Stumvoll 1^st phase estimate) in the high carbohydrate group (1152.2 μIU/mL, IQR 934.9, 1389.5) than the low carbohydrate group (994.1 μIU/mL, IQR 840.3, 1259.4) (P=0.11), but this did not reach statistical significance.

4. Discussion

Here, in pregnant women at a median of 26 weeks’ gestation with risk factors for GDM, we demonstrate that lower carbohydrate intake in the twenty-four hours prior to an OGTT is associated with higher post-load glucose levels. This relationship was linear across the range of carbohydrate intake, suggesting that both excessive and restrictive carbohydrate intake on the day prior to an OGTT could alter the post-load glucose levels and therefore may affect the diagnosis of GDM.

Our results are consistent with studies among non-pregnant individuals (4,28–30). In pregnancy, the results of previous studies evaluating the effect of carbohydrate intake on post OGTT blood glucose levels are mixed (11–18,31,32). Unlike prior studies among pregnant individuals, our study assessed real-life dietary intake with a validated dietary instrument, without instruction, on the day prior to an OGTT in a group of women at high risk for GDM. This makes our results particularly relevant for current clinical practice.

Consistent with our findings, Takizawa et al. studied the impact of a single low and high carbohydrate evening meal on a two-hour 75-gram OGTT the subsequent day among twenty-seven pregnant women and found that the 1- and 2-hour OGTT blood glucose levels were significantly higher following the low carbohydrate meal (16). This led to the diagnosis of GDM in five participants that were not classified as GDM when eating the high carbohydrate evening meal. To our knowledge, this is the only study other than our own demonstrating a relationship between carbohydrate intake specifically in the 24 hours leading up to an OGTT and post-load OGTT glucose levels in pregnant women. The results from Takizawa et al. do not completely address the need for a preparatory diet given that researchers assessed only the effect of the evening meal prior to the OGTT, and the assigned meals were extreme in their carbohydrate content (7% carbohydrate vs. 86% carbohydrate) and therefore less realistic. This study was also limited by a small sample size.

Some other studies utilized food frequency questionnaires and found a relationship between habitual or routine low carbohydrate intake in the first or second trimester of pregnancy and higher glucose levels following an OGTT or diagnosis of GDM in the third trimester (17,31,32). While these studies address habitual diet rather than diet leading up to an OGTT, the dietary information collected may be applicable to the day prior to the OGTT and therefore in line with our results. Consistent with this, we found a moderate strength correlation between carbohydrate intake reported on the 24-hour recall in the first trimester and at 24–32 weeks’ gestation, suggesting that our 24-hour recall may have represented habitual intake to some extent, though women did consume more carbohydrate at the later time point (data not shown).

Other studies have failed to show a relationship between a preparatory diet and OGTT results among pregnant women (11–14). Buhling et al. evaluated 34 pregnant women at 30 weeks’ gestation in a crossover design comparing high carbohydrate intake (diet with carbohydrate composition of 50% of diet) and low carbohydrate intake (diet with carbohydrate composition of 40% of diet) for one week, with dietary compliance assessed by participant logs. There was no difference in OGTT results between the two diets (11). We note that the diets were relatively similar in their carbohydrate content. In three prior interventional studies assessing the difference between moderate to high carbohydrate diets and an “ad lib” diet, results showed no significant difference between subsequent OGTT glucose values or the diagnosis of GDM, but these studies did not assess adherence to the diets or calculate the carbohydrate content of the ad lib diets (12–14). Although the authors of these studies concluded that a recommendation regarding carbohydrate intake prior to an OGTT should be abolished, our data suggest that an “ad lib” or “regular diet” may vary greatly across pregnant women, and on average is ≥ 150 grams of carbohydrate/day. This may explain why no difference in outcome was found between groups compared in studies mentioned above (11–14),in contrast to our own.

The physiologic mechanism by which reduced carbohydrate intake leads to elevated post-load glucose levels has been speculated for many years. The concept of “starvation diabetes” or “hunger diabetes” dates back to the 1800s when it was observed that sugar given to an animal after fasting for 24–36 hours lead to the development of glycosuria (33). Some studies have demonstrated reduced insulin secretion along with higher glucose levels following a carbohydrate restricted diet (28–30,34). In pregnant women, Takizawa et al. showed that the insulin secretory response during an OGTT performed after a low carbohydrate evening meal was lower than that following a high carbohydrate evening meal (16). In line with these results, we found that women with low carbohydrate intake had a trend toward lower insulin secretory response as compared to women with higher carbohydrate intake. This supports the idea that the relationship between low carbohydrate intake and elevated post-load OGTT blood glucose levels is related to lower insulin secretion.

Our study has several strengths, which include prospective observational design, use of a validated 24-hour dietary recall, and enrollment of women with risk factors for GDM. We accounted for appropriate covariates such as total caloric intake as well as BMI, age, and education. The population studied was at high risk for GDM, which is relevant because these women are more likely to require diagnostic OGTT testing; however this may reduce the generalizability of our study.

Limitations do exist. Although the ASA24 is a validated instrument, women may have under- or over-reported carbohydrate intake. Recall was performed by participants and was unsupervised; we excluded women with incomplete recalls and also performed a sensitivity analysis omitting women with the bottom ten percent of caloric intake. We acknowledge that participants were not blinded to their glucose levels from the first trimester, but our results suggested that knowledge of glucose levels from visit 1 did not influence reported dietary carbohydrate intake at visit 2. Of note, we observed no relationship between previous history of GDM and carbohydrate intake suggesting that our findings were not the result of women with a history of glucose intolerance modifying their carbohydrate intake or report of carbohydrate intake in response. Blood samples for glucose were collected in sodium fluoride tubes to inhibit glycolysis and sent immediately to the chemistry laboratory, but were not placed on ice prior to analysis. It is possible that some decline in glucose occurred prior to analysis, but the handling of samples mimicked how most samples are collected and processed in clinical practice. We note that our study used a 2-hour 75-gram OGTT while many clinical practices in the United States use a 3-hour 100-gram OGTT to diagnose GDM (1). Given that the underlying physiologic response to a 75 gram and 100 gram glucose load at 60-minutes is thought to be similar (26), we think this is unlikely to influence the applicability of our results. Our dietary data was limited to 24 hours prior to the OGTT, while an OGTT preparatory diet has been recommended for three days, but the relationship between carbohydrate intake and OGTT results might have been even stronger if our dietary data were extended (30). Unfortunately, we lacked power to adequately ascertain the relationship of carbohydrate intake with diagnosis of GDM.

Older guidelines in the US have advocated for a preparatory diet with at least 150 grams of carbohydrate per day for three days prior to an OGTT in GDM diagnostic protocols (5,35–38),while the most recent ADA and ACOG guidelines do not include recommendations regarding diet prior to an OGTT in pregnant women (1,2). Our data suggest that carbohydrate intake on the day prior to a diagnostic OGTT among pregnant women with risk factors for GDM may impact the diagnosis of GDM: lower carbohydrate intake is associated with higher post-load glucose levels and higher carbohydrate intake is associated with lower post-load glucose levels. Our findings suggest that a standardized preparatory diet may need to be added to recommendations to obtain accurate OGTT results for GDM diagnosis, but continued research in this area should be performed before altering current guidelines. Such future studies should evaluate the effect of this phenomena on GDM diagnosis in a larger population of pregnant women, investigate the relationship between carbohydrate intake and perinatal outcomes, and explore alternative methods to the OGTT to identify women with hyperglycemia in pregnancy.

Supplementary Material

mmc1

Supplemental Figure 1. SPRING Participant Retention and Exclusion

This figure illustrates women in SPRING who dropped out of the study or were excluded from analysis. Our sample size of 95 excludes women diagnosed with GDM before visit 2 and those with missing oral glucose tolerance test or ASA24 dietary data.