Randomization to a Provided Higher-Complex-Carbohydrate Versus Conventional Diet in Gestational Diabetes Mellitus Results in Similar Newborn Adiposity

There remains an absence of international consensus on evidence-based diet recommendations for gestational diabetes mellitus (GDM), but it is widely accepted that nutrition therapy is first-line treatment. As prepregnancy BMI has rapidly increased, the global GDM prevalence is estimated at 14%, with some high-risk populations exceeding 20% (1). The United States has reported a 20% increase in GDM cases between 2016 and 2020 alone, with Asian, Native American, and Hispanic/Latina women disproportionately affected (2).

Nutrition during pregnancy creates an early-life exposure that sets the stage for pediatric metabolic health (3). This influence was first described in the hyperglycemia-hyperinsulinemia hypothesis, then by Freinkel (4) as an exacerbation of overall maternal fuel exposure, not only related to glucose, and linked to altered insulin secretion and/or action. In 1990, Jovanovic and Peterson (5) recommended primary treatment of GDM with diet, specifically with energy intake sufficient to avoid ketosis, and total carbohydrate (CHO) restriction to 30–40% of calories to prevent excessive postprandial glucose excursions. In 2005, concerns were raised about exacerbation of maternal insulin resistance (IR) by saturated fat intake and free fatty acid (FFA) exposure (6), and we and others have shown that increased triglyceride (TG) levels contribute to fetal growth in obesity (7). Anthropologists have demonstrated that protein intake is fairly constant; thus, CHO restriction may unintentionally promote an increase in dietary fat due to the abundance of animal fats and inexpensive processed foods (8).

Current evidence-based GDM nutrition guidelines from multiple sources (9,10) recommend intake of ≥175 g/day CHO (11) and an individualized nutrition prescription that limits excessive maternal weight gain (12) without specifying macronutrient composition. Although numerous comparative studies of dietary advice for GDM have been undertaken, CHO restriction as the primary approach to treatment lacks rigorous evidence from randomized controlled trials (RCTs) (13). Across dietary management approaches, the evidence is insufficient and of low quality, and randomized trials are few, with none providing meals to match calories and optimize compliance (13). As a result, nutrition guidelines for GDM remain inconsistent worldwide.

We developed a diet, called CHOICE (Choosing Healthy Options in Carbohydrate Energy), to directly challenge the lower-CHO diet for GDM (14) and promote the inclusion of higher-quality CHOs. Previous data from a randomized crossover trial suggested our CHOICE diet (60% mostly complex CHO, 25% fat) reduced fasting glucose and FFA levels in the short term and reduced newborn adiposity (NB%fat) in those who followed the CHOICE diet (versus lower CHO and higher fat) from ∼33 weeks of gestation through delivery (14,15). In this parallel RCT where all meals were provided, powered for NB%fat as the primary outcome, we tested the hypothesis that compared to a conventional lower-carbohydrate (40%)/higher fat (45%) diet (LC/CONV), 7–8 weeks of CHOICE would improve maternal IR and 24-hour glycemia, resulting in less NB%fat.

This study was approved by the Colorado Multiple Institutional Review Board (approval no. 14-1358). During 2015–2020, women granted their written informed consent and were compensated for their participation. The Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines were followed (Fig. 1, Supplementary Table 1). Participants self-identified their race or ethnicity.

Pregnant women were identified just after GDM diagnosis (∼28–30 weeks’ gestation) at University of Colorado Health and private community obstetric clinics. At 30–31 weeks of gestation, women were provided a washout diet for 2 days (day 1), then were randomly assigned to either the LC/CONV (5,14,16) or CHOICE diet (14). On day 3, women reported to the Colorado Clinical Translational Research Center (CTRC) after an 8- to 10-h fast. After collection of fasting blood, a 75-g, 2-h oral glucose tolerance test (OGTT) was performed and a blinded continuous glucose monitor (CGM) was attached for 72 h. Women began the random diet assignment on day 3, and this was followed until delivery. All meals were provided 3 days at a time, and women were contacted regularly for monitoring (1–2 times per week). After 7–8 weeks (week 36–37), women again wore a CGM for 72 h and a second 75-g, 2-h OGTT was performed.

Cord blood was obtained immediately after delivery. At ∼13 days, the mother and infant reported to the CTRC at Children’s Hospital Colorado, where maternal and infant body composition were measured.

The 1:1 randomization was stratified by ethnicity (Hispanic vs. non-Hispanic) using a random number generator (performed by a biostatistician) and then administered through the Colorado Clinical and Translational Science Institute. The investigators were not involved in the randomization process. Although it was not possible for the frontline study team to be blinded to treatment, the investigators were uninformed of the group assignment.

Including English-speaking pregnant women with singleton pregnancies, who were between 20 and 39 years old, and had BMI of 26–42 kg/m² at the time of diagnosis (expanded from the original BMI range of 26–39 kg/m² after ∼1 year) allowed greater participation in this study. Women with GDM meeting criteria by a 50-g glucose challenge followed by a 3-h 100-g OGTT according to the American College of Obstetricians and Gynecologists (17), supported by the American Diabetes Association (9) and not requiring medication, were randomized. Women with fasting TG level ≥400 mg/dL, who smoked, had preexisting diabetes, or had suspected preexisting diabetes (A1C ≥6.5%, fasting glucose level >125 mg/dL, or random glucose level >200 mg/dL) or women unlikely to achieve glycemic control with diet alone (fasting glucose level >105 mg/dL) were not enrolled. Women with risk factors for placental insufficiency, including hypertension, renal disease, thrombophilias, rheumatologic disease, a history of preeclampsia, or fetal growth restriction were excluded, as were those with a history of preterm labor, and those who used β-blockers or glucocorticoids. Women requiring insulin, oral hypoglycemic agents, or delivery at <36 weeks’ gestation were removed from the study.

Women were randomized to eat either the LC/CONV or CHOICE diet (14,15); the food was prepared in the CTRC Metabolic Kitchen (Supplementary Table 2). Food preferences were assessed by an on-line survey to avoid unliked foods and food allergies and to optimize adherence. After randomization, 100% of estimated energy requirement calories were provided during metabolic testing weeks. All three meals were provided through delivery. Snacks (∼20% of total calories) were chosen by participants according to provided snack lists consistent with the randomized diet. Women recorded their meal intake and snacks and self-monitored blood glucose levels four times per day (standard of care) using provided glucometers (Bayer Contour Next; Ascensia Diabetes Care, Parsippany, NJ) and kept a log.

The initial 2-day washout diet was 50% CHO, 35% fat, and 15% protein. The LC/CONV diet comprised 40% CHO, 45% fat, and 15% protein (5,16) and the CHOICE comprised 60% CHO, 25% fat, and 15% protein, consistent with the American Heart Association guideline to reduce cardiovascular disease risk. Protein, fat, and sugars in both diets were as follows: 15% protein provided ≥71 g/day, which met Institute of Medicine (IOM) guidelines (11); daily fat was ∼35% saturated fatty acid, ∼45% monounsaturated fatty acid, and ∼20% polyunsaturated fatty acid; and simple sugars were fixed at 70 ± 5 g/day (absolute grams) in both diets. Daily calories were partitioned as 25% (breakfast), 25% (lunch), and 30% (dinner), and the remaining 20% was divided into snacks (16). Both diets consisted of mainly low (≤55) to moderate (56–69) glycemic index foods (18).

Diets were matched in kilocalories per day on the basis of IOM recommendations to support weight gain of 7–11.5 kg for overweight women and 5–9 kg for women with obesity (12). Weight was measured at each visit. The estimated energy requirement was revisited at 34 weeks’ gestation on the basis of weight gain (Supplementary Table 3).

Prepregnancy weight was self-reported and verified at the first-trimester prenatal visit. Perinatal outcomes were recorded from the electronic health record. Small for gestational age (≤10th percentile) and large for gestational age (≥90th percentile) conditions were calculated using local and international (19) standards; rates were identical using both methods.

Diet logs were collected and analyzed weekly (Nutrition Data System for Research, University of Minnesota, 2010) to assess adherence. Compliance was calculated by the percentage of provided meals consumed per day. Participants reported any meal deviation; >90% of meals provided were consumed, confirmed by record review.

On day 1 and at 36–37 weeks’ gestation, a fasting (8–10 h) blood sample was collected to measure hemoglobin A_1c (30–31 weeks’ gestation only) glucose, insulin, TG, and FFA levels. Each woman wore a blinded Dexcom G4 Platinum CGM (Dexcom, Inc.). Glucose variables were defined and data were collected, handled, and analyzed as we have published (20). Venous blood samples were obtained at −15, 0, then at 30, 60, 90, and 120 min after a 75-g glucose load for glucose and insulin. Total area under the curve (AUC) values for glucose and insulin were calculated by the trapezoidal method. Whole-body insulin sensitivity was calculated using the Matsuda Index (21,22) as was the liver IR index. Fasting IR was estimated by the HOMA-IR, calculated as (fasting insulin [mU/L] ⋅ fasting glucose [mmol/L])/22.5.

A cord blood sample was collected at delivery to measure C-peptide, glucose, and insulin levels. Samples were available for ∼60% of newborns.

Infant and maternal fat mass and fat-free mass were assessed at 13.4 (SEM, ±0.4) days of life using air-displacement plethysmography (PEA POD and BOD POD; Cosmed, Concord, CA). Anthropometry included length (body and limb), weight, and circumferences (head, abdominal, chest, and midarm). Of the 46 neonates, the PEA POD failed in four instances (LC/CONV, n = 1; CHOICE, n = 3). Skinfold measures were performed in triplicate (Harpenden calipers, British Indicators, St. Albans, U.K.; resolution, 0.2 mm) at five sites (triceps, biceps, subscapular, suprailiac, and midthigh) by one of two highly skilled coinvestigators. Percent fat mass (%FM) was calculated by two equations: that of Slaughter et al. (23) and of Catalano et al. (24).

Glucose (hexokinase; Beckman Coulter, Brea, CA), insulin (radioimmunoassay; Millipore, Billerica, MA), TG (enzymatic; Beckman Coulter), FFA (enzymatic; WaKo Chemicals, Richmond, VA), C-peptide (radioimmunoassay; Diagnostic Products Corp., Los Angeles, CA), and HbA_1c (potassium ferricyanide; Siemens DCA Vantage, Siemans Healthcare) values were measured in the CTRC Core Lab.

Women were advised to walk 30 min/day. PA was assessed at 30–31 and 36–37 weeks’ gestation using the 36-item Pregnancy Physical Activity Questionnaire (intraclass correlations: 0.78–0.93; 1-week test–retest reliability) (25).

The primary end point for this study was the difference in NB%fat between diet groups. A priori, power was based on our preliminary data (14,26) and other published data to support a clinically significant difference in NB%fat. To detect a 3.4% difference in NB%fat by PEA POD (SD, 4%), 46 participants (n = 23/group) would be required. Power calculations were derived using a two-sided two-sample t test assuming α = 0.05 and 1 − β = 0.8. Data are reported as mean ± SEM. No preliminary analyses on powered outcomes were conducted; the outcomes were analyzed after the last infant measurement at ∼13 days.

To confirm balanced groups after randomization, two-sample t tests were performed on baseline measures for the 46 women who completed the study, and differences between post-diet measures between groups were assessed with two-sample t tests at α = 0.05. Multiple imputation (27) was performed for four missing PEA POD measures using predictive mean matching as implemented in the R package mice (28), with infant sex, %FM by the Slaughter et al. equation (23), %FM by the Catalano et al. (24) equation, baby length, baby weight, baby day of life, gestational age at birth, maternal prepregnancy BMI, and diet group as predictors. We created, analyzed, and combined 100 data sets to obtain group mean PEA POD measures and test for differences between the groups. As a sensitivity analysis, Bayesian linear-regression multiple imputation was also performed. For OGTT and CGM measures, point estimates of the between-group differences at 36–37 weeks’ gestation (CHOICE − LC/CONV), and within-group changes (36–37 − 30–31 weeks’ gestation) are presented, along with corresponding 95% CIs.

For this study, 59 women with newly diagnosed GDM were randomized, 28 to the LC/CONV diet and 31 to the CHOICE diet (Table 1, Fig. 1). A total of 46 women (n = 23/group) completed the study after exclusions (Table 1); three were removed due to diet failure (n = 2 in the LC/CONV group; n = 1 in the CHOICE group). Results are presented as LC/CONV versus CHOICE. At 36–37 weeks’ gestation, the mean total energy intake for LC/CONV (2,092 kcal/day) and CHOICE (2,132 kcal/day) was similar, but mean CHOs were different by design (214 vs. 316 g/day at randomization), as were total fats (106 vs. 59 g/day) (Supplementary Table 2, Fig. 2A). Importantly, total gestational weight gain (GWG) (11 ± 1 vs. 10 ± 1 kg) and weight gained during the intervention (1.8 ± 0.3 vs. 2.0 ± 0.4 kg) were not different between groups (Table 1). Fasting TG levels increased similarly in both groups with no between-group differences in TG or FFA levels at 36–37 weeks of gestation (Table 2).

NB%fat (10.3 ± 1 vs. 10.8 ± 1), the primary outcome, was not different between groups at 13.4 ± 0.4 days of life (Table 1). Using predictive mean matching multiple-imputation models, NB%fat remained no different (summary of mean ± SEM; 10.1 ± 1 vs. 10.5 ± 1; P = 0.748). The sensitivity analysis using Bayesian linear-regression imputation procedure produced similar results without between-group differences (data not shown). Furthermore, maternal outcomes; birth weight; anthropometrics; cord blood glucose, insulin, and C-peptide levels; and newborn hypoglycemia were not different between groups (Table 1).

Mean glucometer readings over the course of the intervention were not different between groups (composite average of breakfast, lunch, and dinner). For the LC/CONV group, the mean fasting glucose level was 84 ± 1 mg/dL and the 2-h postprandial level was 108 ± 2 mg/dL. For the CHOICE group, the fasting glucose level was 85 ± 1 mg/dL and the 2-h postprandial glucose level was 109 ± 2 mg/dL.

At 36–37 weeks’ gestation, across a number of CGM metrics, there was no difference in glucose levels under the following conditions: fasting (86 ± 3 vs. 90 ± 3 mg/dL), 1-h (119 ± 3 vs. 117 ± 3 mg/dL), 2-h postprandial (106 ± 3 vs. 108 ± 3 mg/dL) (Fig. 2B); 24-h glucose AUC (Table 2, Fig. 2C); or percent time in range (%TIR; 92 ± 1 vs. 91 ± 1) (Table 2, Fig. 2D). The %time >120 mg/dL was statistically higher (8%) in the CHOICE group, as was the nocturnal glucose AUC, but the nocturnal %TIR (defined as 63–100 mg/dL) was not different between groups (Table 2, Fig. 2C and D). Of interest, both diet groups demonstrated poorer %TIR 63–100 mg/dL during the nocturnal hours at 30–31 and 36–37 weeks of gestation (Table 2, Fig. 2D). Several within-group changes where the 95% CI did not include zero are noted for clinical relevance in Table 2.

At 30–31 and 36–37 weeks’ gestation, there were no between-group differences in fasting or post–glucose challenge plasma glucose or insulin levels, Matsuda Index, or the glucose or insulin AUC (Table 2). By the 36–37-week OGTT, fasting glucose level decreased within both groups (LC/CONV: −5 mg/dL [−6%], 95% CI −8.4, −2.5; CHOICE: −7 mg/dL [−9%], 95% CI −9.2, −4.9). In the CHOICE diet group, there was a within-group reduction in glucose AUC between 30–31 and 36–37 weeks’ gestation (−1,050 mg/dL ⋅ min; 95% CI −1,740, −516) (Table 2). By 36–37 weeks’ gestation, there was a within-group decrease (improvement) in HOMA-IR (Table 2) in the CHOICE diet group. However, there were no between-group differences in HOMA-IR at 36–37 weeks of gestation.

As judged by the investigators, an independent safety monitor, and the institutional review board through annual continuing review, no serious adverse events or harms due to the interventions or study participation occurred in this randomized trial. The trial proceeded as planned and recruitment ceased when enrollment targets were reached. No participants who did not meet a study exclusion were lost to follow-up at ∼13 days postpartum.

In this controlled, prospective RCT of a conventional lower-CHO diet compared with the CHOICE diet in the setting of GDM, we did not find an effect on NB% fat, the primary outcome, contrary to our hypothesis. We have further demonstrated that randomization to a conventional lower-CHO diet compared with CHOICE led to no difference in maternal 24-h glycemia or IR indices. The trial was designed to overcome three of the biggest limitations in nutrition studies within diabetes in pregnancy (8): 1) low adherence due to the lack of provided diets; 2) differences in dietary simple sugars, protein, fiber, total energy, and GWG; and 3) the confounding effect of treatment with insulin or a hypoglycemic agent. While both diets controlled maternal glucose within conventional targets (Fig. 2B–D), those in the LC/CONV group had lower %time >120 mg/dL (8%) and lower nocturnal AUC, although the nocturnal %TIR was not different. The 24-h glucose AUC and %TIR (nocturnal, 24 h) were no different between groups, nor were fasting, 1-h, and 2-h postprandial glucose levels by CGM. Notably, only two women in the LC/CONV arm and one in the CHOICE arm required pharmacologic treatment.

Grounded in the principle that some polysaccharides and starches (i.e., whole grains, legumes) tend to mitigate a sharp increase in postprandial glucose (8), we powered this study to detect a between-group difference in NB%fat. This was based on the hypothesis that CHOICE could blunt postprandial glucose and the increase of IR in pregnancy by decreasing total and saturated fats, and might, thereby, reduce fetal-placental glycemia and FFA and TG exposure over 24 h (7), resulting in lower NB%fat. The lack of difference in NB%fat between groups was somewhat surprising to us but is likely supported by the fact that despite the 20% difference in CHO and fat content, other known drivers of fetal overgrowth were similar between groups: maternal glycemic control was excellent; 24-h CGM, a surrogate for total potential fetal glucose exposure (20), was not different; there were no between-diet differences in simple sugars; and saturated fat was limited to 35% in both groups. This may partially explain the lack of statistical difference in FFA and TG levels and estimates of IR between diet groups at 36–37 weeks of gestation. Importantly, total GWG and weight gain during the intervention were similar (Table 2), the diets were eucaloric, and women were highly adherent (>90%) to the treatment.

Although these diets targeted a lower CHO (40%), higher fat (45%) percentage of total calories in the LC/CONV diet versus the CHOICE diet (60% CHO and 25% fat), higher BMI in these participants required higher total calories, which resulted in total CHO intake above the ≥175 g/day IOM threshold (214 g/day in the LC/CONV diet compared with 316 g/day CHO in the CHOICE diet), with a total fat difference of ∼47 g/day (Supplementary Table 2). Despite an ∼100 g/day difference of mainly complex CHO, women in both diet arms achieved similar glycemic control within conventional targets of fasting glucose <95 mg/dL and 1- and 2-h postprandial concentrations of 140 and 120 mg/dL, respectively (9). The 72-h CGM data demonstrated no differences in fasting, postprandial, and 24-h glycemia between groups at 30–31 and 36–37 weeks’ gestation. Although women in both groups lowered their 1-h postprandial and 24-h glucose AUC, those in the CHOICE group statistically lowered their 1-h postprandial glucose over time (−10 mg/dL), and women in the LC/CONV group lowered their 24-h AUC. We have previously suggested lower glycemic targets (1-h, 122 mg/dL; 2-h postprandial, 110 mg/dL) based on patterns of glycemia in normal pregnancy (29); in this study, women in both diet arms met those targets (Fig. 2B–D).

We believed it important to separate nocturnal %TIR and define it as lower (63–100 mg/dL) than 24-h %TIR (63–140 mg/dL). The hours of 2330–0630 h were chosen to represent the nocturnal period, when women were most likely to be fasting. If a participant had breakfast before 0630, the nocturnal period was adjusted. Although the nocturnal AUC was lower between groups at 36–37 weeks’ gestation in the LC/CONV group, the nocturnal %TIR was not different. We recognize challenges in accurately defining a nocturnal period, which may change from day to day. Interestingly, when defining the nocturnal targets as 63–100 mg/dL, women in both diet arms had similar but poorer nocturnal %TIR compared with the 24-h %TIR (Fig. 2D). Nocturnal %TIR is not a standard clinical metric, but %TIR over 24-h, when defined as 63–140 mg/dL, may dismiss nocturnal hyperglycemia up to 140 mg/dL. This may be critically important to fetal overgrowth, because the Hyperglycemia and Adverse Pregnancy Outcomes study (30) demonstrated that fasting hyperglycemia appears to drive large for gestational age more than a post–oral glucose load. Future research tied to maternal–fetal outcomes in which nocturnal versus daytime %TIR at trimester-specific time periods is indicated.

In adapting a nutrition pattern more liberal in complex CHO for GDM, several considerations bear mention. Fears that a diet of 50–60% CHO might result in excessive weight gain, postprandial hyperglycemia (31), fetal hyperinsulinemia, and excess fetal fat accretion have been raised. Three previous nutrition RCTs in the setting of GDM reported no difference in weight gain between a lower versus a higher CHO diet during the intervention period (8). Of note, in the present study, women in the LC/CONV group gained 2 kg over the IOM guideline (12) for total GWG, and in the CHOICE group, women gained 1 kg above that guideline weight, explained by weight gained before the GDM diagnosis. Once enrolled, weight gain was nearly identical (1.8–1.9 kg) given the diets were matched in calories based on BMI, simple sugars were fixed, and CHOs were low-medium in glycemic index. Studies of low glycemic index diets in pregnancy outside of GDM support less GWG from reduced energy intake (32), probably linked to increased satiety. A heightened insulin response to higher simple CHOs may further tax the β-cells in women already at risk for later type 2 diabetes (33). It is also possible that higher dietary CHO from simple sugars or higher fat could result in higher plasma TG levels (34), an increasingly recognized driver of fetal overgrowth in pregnant women with obesity (7). In this study, where simple sugars were fixed and saturated fats were limited, IR indices were not different; fasting TG increased similarly, likely under the influence of rising placental estrogen.

In both diet groups, the data generated in this RCT demonstrate that nutrition therapy alone was successful when weight gain, saturated fats, and simple sugars were limited. A recent systematic review of 18 RCTs (n = >1,000 women) revealed that any improvement in dietary pattern (versus the control) after GDM diagnosis was associated with lower fasting and postprandial glucose levels, less need for insulin, and lower infant birth weight (35). Although women in the CHOICE group demonstrated a within-group improved glucose response to a 75-g OGTT, we acknowledge that recent data suggest the higher CHO intake, particularly within 24 h preceding the OGTT, may partially explain this finding (36). Women in the CHOICE group demonstrated a within-group, reduced basal IR by HOMA-IR, whereas those in the LC/CONV group demonstrated a within-group decrease in 24-h glucose AUC by CGM. Of concern with CHO restriction is increased fat intake that exacerbates IR, particularly if calories are not limited, promoting overnutrition and driving fetal overgrowth. This concern was not validated here; fasting FFA and TG levels were not different between groups, but women randomized to the LC/CONV diet demonstrated a within-group reduced fasting FFA level. We speculate that by limiting simple sugars to 68–70 g/day in both diets and restricting saturated fats to 35% of total fat (34 g/day in LC/CONV vs. 18 g/day in CHOICE; Supplementary Table 2), both diets would be considered heart healthy, which is important because women with GDM have an increased risk for type 2 diabetes and cardiovascular disease.

One of the biggest barriers to successful nutrition therapy in GDM is diet adherence. Rigid CHO restriction has been described as confining and arduous, infringes on cultural food patterns, and may contribute to depression and anxiety in GDM (8). In the only RCT to date in which women with GDM were randomized to 135 g vs. 180–200 g/day of CHO, only 20% and 65%, respectively, met the intervention target (37). The striking increase in GDM among multiracial women demands that customizable options for nutrition therapy be identified. We have identified by systematic review that a range of 47–70% dietary CHO content supports normal fetal growth patterns (38), which is further suggested by the present study findings. We have published adaptability of the CHOICE diet to multicultural diet patterns (8) and suggest that substituting some calories from fat and liberalizing complex CHOs may present new alternatives for personalized nutrition therapy in GDM that promote adherence and food enjoyment in a culturally acceptable way.

This study has some limitations. The study was highly controlled, all foods were provided, and energy intake was controlled, underscoring that the data demonstrate dietary efficacy but possibly not effectiveness in a free-living population. Furthermore, both diets were healthy in that simple sugars and saturated fats were not in excess. The study diets were designed to meet targets as percentages of total energy intake, not grams of CHO per day. Because of this, in the context of higher BMI, the CHO intake in both diets exceeded 200 g/day; by some GDM nutrition guidelines that define a low CHO diet by absolute grams/day, the conventional diet here might not be considered low CHO. The NB%fat measures were taken at ∼13 days of life in both groups rather than at 48 h, due to some deliveries at other hospitals. Breastfed newborns lose some adiposity during the first days of life and regain it by 5 days according to a study by Roggero et al. (39); notably, 80% of our participants exclusively breastfed. Furthermore, it has been shown that NB%fat by PEA POD at 1–2 weeks for girls and boys are within 1–2% of 48-h measures (40). We were not powered to detect sex differences in our primary outcome, although sex was balanced between groups. Heterogeneity in the response to diet therapy is likely and high-quality, real-world comparative-effectiveness trials during pregnancy across women are needed. Although our confidence in the lack of NB%fat difference is high, we acknowledge that this does not demonstrate noninferiority, because we were not powered to address this question.

In this prospective RCT during which all meals were provided, our a priori hypotheses that the two diets would result in different maternal and infant outcomes were not supported by the data. Although the %time >120 mg/dL was statistically higher (8%) in the CHOICE group, as was the nocturnal glucose AUC, we demonstrated no between-group differences in NB%fat (primary outcome), %TIR, or 24-h glycemia after 7–8 weeks of nutrition therapy with a conventional lower-CHO or a higher-complex-CHO CHOICE diet. Despite an ∼100 g/day difference in CHO intake, glycemia was well controlled in both arms. These data suggest that flexibility in dietary CHO is possible while limiting simple sugars, saturated fats, and excess calories, paving the way for expanded and personalized options for nutrition therapy in GDM.

Acknowledgments. The authors acknowledge the tremendous expertise of the Colorado Clinical and Translational Sciences Institute (CCTSI) Nutrition Core; Archana Mande, Karen Morgenthaler, and Pamila Allen oversaw the meticulous handling and sample analysis in the CCTSI Core Laboratory. The CCTSI Perinatal Nursing Team and the expertise of Lucy Fashaw, Lisa Lewis, Patrisha Adkins, Lauren Zausmer, and Amy Lamprecht are further acknowledged. We especially thank Laura K. Moss for the development, execution, and management of our database. The investigators further acknowledge the support of Dr. Gabriele Ronnett. The contributions by these individuals and the CCTSI were important to the success of this study.

Funding. This study was supported by the National Institutes of Health (NIH), National Institute of Diabetes and Digestive and Kidney Disease (grant R01DK101659), and by a cooperative grant from Janssen Research and Development. S.S.F. was supported by NIH grant T32DK007446 for part of the study period. Continuous glucose monitors were provided at a reduced cost by Dexcom, Inc. Additional support was provided by the NIH National Center for Advancing Translational Sciences Colorado Clinical and Translational Science Award Grant UL1 TR002535.

Contents are the authors’ sole responsibility and do not necessarily represent official NIH views.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. T.L.H. and L.A.B. researched data, contributed to discussion, and wrote, reviewed, and edited the manuscript. J.P.C. reviewed and edited the manuscript. S.S.F., N.H., E.Z.D., K.R., E.H., T.M., J.H., and J.E.F. researched data and reviewed and edited the manuscript. B.K.F. ran statistical analyses and reviewed and edited the manuscript. All authors approved the final version of the manuscript. T.L.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Portions of these data were presented as an oral abstract at the virtual American Diabetes Scientific Sessions, June 2020 and June 2022.