Association of late‐night carbohydrate intake with glucose tolerance among pregnant African American women

Abstract

Obesity and late‐night food consumption are associated with impaired glucose tolerance. Late‐night carbohydrate consumption may be particularly detrimental during late pregnancy because insulin sensitivity declines as pregnancy progresses. Further, women who were obese (Ob) prior to pregnancy have lower insulin sensitivity than do women of normal weight (NW). The aim of this study is to test the hypothesis that night‐time carbohydrate consumption is associated with poorer glucose tolerance in late pregnancy and that this association would be exacerbated among Ob women. Forty non‐diabetic African American women were recruited based upon early pregnancy body mass index (NW, <25 kg m⁻²; Ob, ≥30 kg m⁻²). Third trimester free‐living dietary intake was assessed by food diary, and indices of glucose tolerance and insulin action were assessed during a 75‐g oral glucose tolerance test. Women in the Ob group reported greater average 24‐h energy intake (3055 kcal vs. 2415 kcal, P < 0.05). Across the whole cohort, night‐time, but not day‐time, carbohydrate intake was positively associated with glucose concentrations after the glucose load and inversely associated with early phase insulin secretion (P < 0.05). Multiple linear regression modelling within each weight group showed that the associations among late‐night carbohydrate intake, glucose concentrations and insulin secretion were present only in the Ob group. This is the first study to report an association of night‐time carbohydrate intake specifically on glucose tolerance and insulin action during pregnancy. If replicated, these results suggest that late‐night carbohydrate intake may be a potential target for intervention to improve metabolic health of Ob women in late pregnancy.

Introduction

As pregnancy progresses, women become less insulin sensitive and glucose concentrations increase (Lain & Catalano 2007). Among healthy women, an increase in insulin secretion from pancreatic β‐cells will compensate for reduced insulin sensitivity in order to maintain circulating glucose concentrations within a normal range. However, when women are obese, the ability of the β‐cells to adequately compensate for reduced insulin sensitivity may be compromised, resulting in higher circulating glucose concentrations. An increasing body of literature suggests that relatively high glucose concentrations during pregnancy, even among non‐diabetic women, may contribute to adverse perinatal outcomes and long‐term risk for obesity in offspring (Catalano & Ehrenberg 2006; HAPO Study Cooperative Research Group 2008; Bush et al. 2011; Chandler‐Laney et al. 2011).

Research concerning the influence of free‐living diet on glucose tolerance during pregnancy has failed to reach a consensus. In a meta‐analysis of interventions to prevent gestational diabetes, Oostdam et al. (2011) found modest evidence that general dietary advice was beneficial, but there was no clear advantage of either of the most common interventions: to reduce intake of fat or to reduce intake of carbohydrate or high glycaemic index foods. Furthermore, timing of caloric intake is rarely, if ever, examined during pregnancy, but may be an important modifier of the association of diet with glucose tolerance. In clinical trials of healthy non‐pregnant adults, glucose concentrations are higher and insulin secretion is lower, following meals consumed in the evening vs. morning despite identical composition (Morgan et al. 2012; Saad et al. 2012). The large body of literature describing adverse metabolic health consequences of shift‐work, circadian rhythm disruption and alteration of typical feeding times (Mendoza 2007; Bray & Young 2012; Morris et al. 2012; Maury et al. 2014) also implies that late‐day energy consumption may be detrimental to glucose tolerance. To our knowledge, no previous study has examined the association of late‐night food intake, and in particular carbohydrate intake, with glucose tolerance during pregnancy.

The overall goal of the current study was to test the hypotheses that late‐night total energy consumption, and consumption of carbohydrate specifically, would be associated with higher circulating glucose concentrations following an oral glucose load and an impaired ability to secrete sufficient insulin to compensate for reduced insulin sensitivity. Furthermore, we hypothesised that the association of late‐night food intake with impaired glucose tolerance would be exacerbated among women who were obese in early pregnancy because they have lower insulin sensitivity in late pregnancy than do women who were normal weight in early pregnancy. These hypotheses were tested in a cohort of low‐income African American women; a population with high risk for complications during pregnancy (Bryant et al. 2010); and for whom late‐night eating is not uncommon (Allison et al. 2012).

Key messages.

• Insulin sensitivity decreases in late pregnancy and obesity exacerbates this problem, increasing the potential for glucose concentrations to become elevated.

• High glucose concentrations during pregnancy are associated with adverse effects for the mother and fetus.

• Studies of non‐pregnant adults have shown that late‐day meal consumption may be detrimental to glucose tolerance.

• In this study, late‐night eating, particularly of carbohydrates, was associated with reduced glucose tolerance and lower early phase insulin secretion in women who were obese, suggesting that an intervention to reduce late‐night intake of carbohydrate may be efficacious.

Materials and methods

Participants

Non‐diabetic African American women were recruited into this study during the third trimester of pregnancy and stratified by body mass index (BMI) in early pregnancy: normal weight (NW) with BMI ≤ 24.9 kg m⁻² or obese (Ob) with BMI ≥ 30.0 kg m⁻². Women were recruited from among those receiving prenatal care at local health departments and planning to deliver at a large urban hospital. Mothers were eligible if they were >16 years old at recruitment, had initiated prenatal care prior to 19 weeks' gestation and were experiencing a healthy singleton pregnancy. Women were excluded if they had pre‐existing type 1 or 2 diabetes, had developed gestational diabetes in the current pregnancy (on the basis of routine clinical gestational diabetes tests administered at 24–28 weeks gestation), had previously delivered a preterm (i.e. <37.0 weeks) or growth‐restricted (i.e. <2500 g) infant, or had any medical condition during pregnancy that is believed to interfere with normal foetal growth.

Procedure

At 32.0–34.6 weeks of pregnancy, participants completed a 3‐day food record and underwent an oral glucose tolerance test (OGTT) following a 12‐h overnight fast. Maternal weight in early pregnancy was retrieved from medical records, and height (cm) and weight (kg) were measured prior to the OGTT. The Institutional Review Board for Human Use at the University of Alabama at Birmingham approved all study procedures. Informed consent was obtained from all participants, and consent was also obtained from a parent or guardian of women who were aged <19.0 years at enrolment (n = 5).

Dietary intake

Participants were provided with a 3‐day food diary and asked to complete the diary with as much detail as they could. Upon return to the clinic, research assistants, who had been trained by an experienced investigator, reviewed the food diary with each participant to clarify portion sizes and any other detail that was missing or inadequate. Research assistants subsequently entered all information from the food diary into ASA24^TM (National Cancer Institute 2011). Dietary outcomes of interest included average total 24 h, daytime (6 am–7:59 pm) and night‐time (8 pm–5:59 am) intake of total kilocalories, and kilocalories from protein, carbohydrate and fat. Daytime intake on the day participants started the food diary was excluded because intake prior to the clinic appointment was not recorded. Consequently, daytime intake was averaged across days 2 and 3 of the food diary. Night‐time intake was averaged across the first and second nights of the food diary, with the third night excluded because of potential confounding due to the requirement to fast prior to the OGTT.

Oral glucose tolerance test

Women returned to the clinic following at least a 12‐h overnight fast, which was verified by the food diary. An intravenous catheter was placed to obtain blood samples during the 75‐g OGTT. Blood draws were obtained at −10 and −5 min (averaged to provide fasting concentrations), and at 10, 20, 30, 60, 90 and 120 min relative to glucose consumption. Sera were separated and stored at −85°C until assayed for glucose, insulin and C‐peptide concentrations.

Serum assays

Glucose, insulin and C‐peptide were analysed by the Core Laboratory of the UAB Nutrition Obesity Research Center, Diabetes Research Center and the Center for Clinical and Translational Science. Fasting glucose was measured with the SIRRUS analyser (Stanbio Laboratory, Boerne, TX, USA). This analysis had a mean intra‐assay coefficient of variation (CV) of 1.28% and an inter‐assay CV of 2.56%. Insulin and C‐peptide were assayed using the TOSOH AIA‐600 II automated analyser (TOSOH Bioscience, Inc., San Francisco, CA, USA). The analyses for insulin had mean intra‐ and inter‐assay CV of 4.42% and 1.49%, respectively. For C‐peptide, the intra‐ and inter‐assay CV were 1.67% and 2.59%, respectively.

Insulin sensitivity and secretion

Glucose, insulin and C‐peptide concentrations throughout the 2‐h test were used to derive insulin sensitivity (SI) using the Matsuda index (Matsuda & DeFronzo 1999) and indices of β‐cell response (Breda et al. 2001). Indices of β‐cell response included total β‐cell response (i.e. overall secretion; PhiTOT), basal β‐cell response (i.e. β‐cell sensitivity to glucose under fasting conditions; PhiB), dynamic β‐cell response (i.e. β‐cell response to the rate of increase in glucose; PhiD) and static β‐cell response (i.e. β‐cell response to glucose concentrations above basal across the test period; PhiS). The overall disposition index (DI; the ability of the β‐cells to secrete enough insulin to compensate for reduced insulin sensitivity) was derived as the product of the Matsuda index and PhiTOT.

Statistical analysis

Gestational weight gain (GWG) was calculated as the difference between measured body weight at the first and last prenatal care visits. Independent groups' t‐tests were used to compare the two groups in terms of GWG, dietary outcomes, fasting and post‐challenge glucose, C‐peptide and indices of insulin action. One participant could not complete the OGTT and so was excluded from any analyses involving post‐OGTT response. A second participant was found to be an extreme outlier (i.e. >3 SD) in terms of SI, and a third was an outlier for DI. Data from these participants were excluded only from analyses involving SI and DI, as applicable. Post‐OGTT response was calculated as the area under the curve (AUC) for glucose, using the trapezoid method (Matthews et al. 1990). Simple Pearson's correlations were calculated to explore associations among day and night total kilocalories, carbohydrate and fat consumption, glucose concentrations and insulin action. Multiple linear regression modelling was then used to examine whether any association of day or night total kilocalories or carbohydrate consumption with glucose concentrations or insulin action was independent of weight status and total 24‐h kcal consumption. Follow‐up regression models within each weight group were calculated to examine whether the association between night‐time total kilocalories or carbohydrate intake and glucose concentration or insulin action were specific to women in the Ob group only. Due to non‐normality of distribution, insulin concentrations and indices of insulin secretion were log₁₀‐transformed prior to analysis. Alpha was set at 0.05 for statistical significance, and all analyses were performed using the Statistical Package for the Social Sciences (spss), version 18 (SPSS Inc., Chicago, IL, USA).

Results

Forty women enrolled in the study (20 per group). Characteristics of the sample are presented in Table 1. The groups did not differ in terms of age at the time of the study, but women in the Ob group tended to have greater parity (P = 0.09). Seventy per cent of the NW women were nulliparous, compared with 25% of the Ob group. On average, women in the Ob group gained less weight across pregnancy in comparison to those in the NW group, and this difference remained after adjusting for the number of days between the first and last prenatal care visits (P < 0.01). Overall, women in the Ob group reported higher 24‐h energy intake (P < 0.05), but the relative macronutrient composition of the diets did not differ (not shown). Women in each group reported that their earliest daytime caloric consumption occurred at approximately 9:30 am, and their latest caloric consumption was at approximately 10 pm, but there was no group difference in these times. Consistent with this meal pattern, food and beverage consumption after 8 pm at night was very prevalent, with all but one woman reporting energy intake on at least one of the two recorded free‐living nights. NW and Ob women consumed an average of 22% and 24% of daily kilocalories at night, respectively (P > 0.40). Women in the Ob group reported greater fat intake during the day (P < 0.05) and greater carbohydrate intake at night (P < 0.05), but protein intake did not differ. When adjusted for total daytime or night‐time kilocalories as appropriate, daytime fat intake did not differ by group, whereas night‐time carbohydrate intake tended to remain higher for women in the Ob group (P = 0.09).

Table 1.

Characteristics of the study participants (data are mean ± SD)

	NW	Obese	P‐value
Age (years)	22.1 ± 3.6	24.0 ± 4.6	0.15
Parity	0.6 ± 1.1	1.1 ± 0.9	0.09
BMI (kg/m2)	21.4 ± 2.5	38.2 ± 6.7	<0.001
GWG (kg)	15.2 ± 5.6	6.4 ± 10.9	<0.01
Glucose at routine GDM screen (mg dL−1)	102.9 ± 19.6	112.6 ± 16.1	0.10
Average energy intake (kcal/day)	2414.8 ± 572.6	3055.5 ± 891.2	<0.05
Daytime kilocalories (6 am–7:59 pm)	1863.7 ± 486.5	2256.9 ± 699.5	<0.05
Night‐time kilocalories (8 pm–5:59 am)	529.8 ± 342.3	782.6 ± 591.9	0.11
Daytime protein (kcal)	259.0 ± 68.5	313.8 ± 111.9	0.07
Daytime fat (kcal)	669.9 ± 212.4	870.8 ± 309.6	<0.05
Daytime CHO (kcal)	952.4 ± 315.6	1090.3 ± 386.6	0.22
Night‐time protein (kcal)	100.9 ± 70.0	115.2 ± 93.1	0.59
Night‐time fat (kcal)	209.7 ± 134.5	293.9 ± 267.0	0.22
Night‐time CHO (kcal)	220.9 ± 171.1	378.4 ± 270.0	<0.05
Time of earliest caloric consumption (hh : mm)	9:22 ± 1:52	9:29 ± 2:00	0.86
Time of latest caloric consumption (hh : mm)	21:43 ± 1:56	22:41 ± 2:17	0.16
Fasting glucose (mg/dL)	72.4 ± 6.0	77.9 ± 5.6	<0.01
Glucose AUC (mg/dL)	13007.5 ± 1959.1	13910.6 ± 1672.8	0.13
Fasting insulin (uU/mL)	10.7 ± 6.7	18.2 ± 7.2	<0.01
Insulin sensitivity	4.1 ± 1.7	2.6 ± 0.9	<0.01
X0 (pmol L−1)	1506.8 ± 608.5	1523.0 ± 597.2	0.93
PhiB (109 min−1)	8.5 ± 3.6	12.7 ± 3.4	<0.01
PhiS (109 min−1)	112.8 ± 58.3	122.3 ± 46.0	0.58
PhiD (109)	488.5 ± 125.6	493.7 ± 134.4	0.90
PhiTOT (109 min−1)	112.9 ± 58.3	122.4 ± 46.0	0.58
DI (Matsuda × PhiTOT)	377.5 ± 137.7	307.6 ± 131.9	0.13

AUC, area under the curve; BMI, body mass index; CHO, carbohydrate; DI, disposition index; GWG, gestational weight gain; GDM, gestational diabetes mellitus; NW, normal weight; PhiB, basal β‐cell response; PhiD, dynamic β‐cell response; PhiS, static β‐cell response; PhiTOT, total β‐cell response.

Fasting glucose was higher in the Ob group as compared to the NW group (P < 0.01; Table 1), but the glucose AUC was not statistically different (P = 0.13). Fasting insulin was higher, and correspondingly, insulin sensitivity was lower, in the Ob group (P < 0.01). Basal insulin secretion (PhiB) was higher in the Ob group (P < 0.05), but this difference weakened when adjusted for insulin sensitivity in a subsequent analysis (P = 0.07; not shown). There was a trend for women in the Ob group to have lower DI (P = 0.13).

Simple Pearson's correlation analyses were used to explore associations of day and night total kilocalorie, and kilocalories from fat and carbohydrate with glucose concentrations, insulin sensitivity, insulin secretion and DI. As shown in Table 2, daytime intake was not associated with glucose concentrations or any of the indices of insulin action to a statistically significant degree. However, night‐time total energy, carbohydrate, and in some cases, fat, intake were associated with greater glucose concentrations following the OGTT, higher fasting insulin and lower PhiD (P < 0.05). Day and night protein intake was not associated with any of these glucose and insulin variables (not shown). Multiple linear regression models were used to examine whether the associations of night‐time total energy or carbohydrate intake with glucose AUC, PhiD and DI were independent of potential confounders. As shown in Table 3 (left panel), night‐time energy intake was not independently associated with glucose AUC or DI after adjusting for weight group and daytime energy intake (models 1 and 2), but it remained inversely associated with PhiD after adjusting for weight group, daytime energy intake and insulin sensitivity (model 3). Results were the same when the duration of the overnight fast prior to the OGTT was added to the models (not shown). To examine whether the associations of night‐time energy intake with glucose tolerance and insulin action were present in either weight group when examined alone, the above linear regression models were repeated within each group. Results are shown in the middle and right‐hand panels of Table 3, and indicate that night‐time energy intake was not associated with glucose tolerance or insulin action for the NW group, but in the Ob group, it remained inversely associated with PhiD after adjusting for insulin sensitivity and daytime energy intake.

Table 2.

Simple Pearson's correlations of day‐ and night‐time total kilocalories and kilocalories from carbohydrate and fat with indices of glucose tolerance, insulin sensitivity and the disposition index

	Day kilocalories	Day CHO	Day fat	Night kilocalories	Night CHO	Night fat
Fasting glucose	−0.03	−0.09	0.02	0.38*	0.38*	0.39*
Glucose 60min	−0.14	−016	−0.06	0.28 †	0.39*	0.17
Glucose 120min	−0.12	−0.15	−0.05	0.18	0.27	0.10
Glucose AUC	−0.17	−0.19	−0.10	0.28 †	0.40*	0.17
Fasting insulin	0.11	0.07	0.14	0.34*	0.33*	0.32*
Insulin sensitivity	−0.09	−0.07	−0.09	−0.27	−0.31 †	−0.25
X0	0.01	−0.08	0.11	−0.11	0.00	−0.18
PhiB	0.18	0.09	0.23	0.32*	0.32*	0.31 †
PhiD	0.27 †	0.18	0.30 †	−0.34*	−0.34*	−0.32*
PhiS	0.25	0.22	0.21	−0.08	−0.18	0.03
PhiTOT	0.25	0.22	0.21	−0.08	−0.18	0.03
DI	0.18	0.24	0.08	−0.30 †	−0.38*	−0.21

AUC, area under the curve; CHO, carbohydrate; DI, disposition index; PhiB, basal β‐cell response; PhiD, dynamic β‐cell response; PhiS, static β‐cell response; PhiTOT, total β‐cell response. *P < 0.05; ^†0.05 < P < 0.10.

Table 3.

Results of multiple linear regression models predicting indices of glucose metabolism from night‐time total energy intake

	Whole cohort				Normal weight				Obese
Variable	Adj. R 2	Unstd. β	Std. β	P	Adj. R 2	Unstd. β	Std. β	P	Adj. R 2	Unstd. β	Std. β	P
Model 1: Glucose AUC	0.01			0.09	0.00			0.38	0.07			0.22
Group		505.19	0.28	0.11		–	–	–		–	–	–
Daytime kilocalories		−0.74	−0.25	0.13		−1.28	−0.33	0.19		−0.49	−0.21	0.37
Night‐time kilocalories		0.79	0.21	0.19		0.30	0.05	0.83		0.94	0.33	0.15
Model 2: DI	0.12			0.07	−0.12			0.88	0.19			0.06
Group		−39.91	−0.29	0.09		–	–	–		–	–	–
Daytime kilocalories		0.06	0.27	0.11		0.03	0.13	0.65		0.07	0.37	0.09
Night‐time kilocalories		−0.06	−0.23	0.16		−0.01	−0.04	0.89		−0.08	−0.35	0.11
Model 3: PhiD	0.22			<0.05	0.03			0.35	0.34			<0.05
Group		−0.02	−0.18	0.31		–	–	–		–	–	–
Insulin sensitivity		−0.03	−0.39	<0.05		−0.03	−0.42	0.10		−0.03	−0.25	0.23
Daytime kilocalories		0.00	0.28	0.08		0.00	0.18	0.47		0.00	0.34	0.09
Night‐time kilocalories		−0.00	−0.42	<0.05		−0.00	−0.10	0.69		0.00	−0.58	<0.01

AUC, area under the curve; CHO, carbohydrate; DI, disposition index; PhiD, dynamic β‐cell response.Bolded and italicized values indicate models that were statistically significant at the P < 0.05 level.

When total night‐time energy intake was replaced in the models with night‐time carbohydrate consumption, the associations between night‐time carbohydrate consumption and glucose AUC or DI were confirmed, after adjusting for weight group and 24‐h intake (Table 4, left panel; models 1 and 2). Although the overall model predicting PhiD was not statistically significant, the association between night‐time carbohydrate intake and PhiD remained after adjusting for weight group, 24‐h intake and insulin sensitivity (model 3). When included in the models, duration of overnight fast prior to the OGTT did not alter the results (not shown). The models were repeated within each weight group with results showing that the aforementioned associations were only present in the Ob group (Table 4, middle and right panels). To further illustrate the results of these linear regression models, the association of night‐time carbohydrate intake with glucose AUC, PhiD and DI, within each weight group and after adjusting for the covariates, is displayed in Fig. 1.

Table 4.

Results of multiple linear regression models predicting indices of glucose metabolism from night‐time carbohydrate intake

	Whole cohort				Normal weight				Obese
Variable	Adj. R 2	Unstd. β	Std. β	P	Adj. R 2	Unstd. β	Std. β	P	Adj. R 2	Unstd. β	Std. β	P
Model 1: Glucose AUC	0.18			<0.05	–0.01			0.43	0.29			<0.05
Group		384.91	0.21	0.20		–	–	–		–	–	–
Total kilocalories		–0.80	–0.35	0.07		–1.11	–0.33	0.23		–0.68	–0.37	0.14
Night‐time CHO		16.75	0.54	<0.01		12.25	0.27	0.32		18.37	0.74	<0.01
Model 2: DI	0.17			<0.05	–0.11			0.82	0.32			<0.05
Group		–32.53	–0.24	0.15		–	–	–		–	–	–
Total kilocalories		0.06	0.35	0.08		0.04	0.16	0.58		0.07	0.48	0.05
Night‐time CHO		–1.20	–0.52	<0.05		–0.50	–0.16	0.60		–1.49	–0.76	<0.01
Model 3: PhiD	0.10			0.13	–0.03			0.49		0.25		0.06
Group		0.00	0.04	0.84		–	–	–		–	–	–
Insulin sensitivity		0.00	–0.20	0.28		0.00	–0.39	0.15		–0.00	–0.06	0.83
Total kilocalories		0.00	0.34	0.12		0.00	0.19	0.51		0.00	0.39	0.17
Night‐time CHO		–0.00	–0.61	<0.01		0.00	–0.12	0.69		–0.00	–0.78	<0.05

AUC, area under the curve; CHO, carbohydrate; DI, disposition index; PhiD, dynamic β‐cell response.Bolded and italicized values indicate models that were statistically significant at the P < 0.05 level.

Figure 1.

Partial correlation plots illustrating that the association of night‐time carbohydrate intake with indices of glucose tolerance and insulin action remains statistically significant for only women in the obese (Ob) group, after adjusting for potential confounders. As shown in the top panel, night‐time carbohydrate intake remained positively associated with glucose area under the curve (AUC), after adjusting for total 24‐h energy intake, for only women in the Ob group: (a) normal weight (NW) (partial r = 0.25, P = 0.32) vs. (b) obese (partial r = 0.61, P < 0.01). In the middle panel, night‐time carbohydrate intake remained inversely associated with PhiD, after adjusting for total 24‐h energy intake and insulin sensitivity, for only women in the Ob group: (c) NW (partial r = −0.26, P = 0.32) vs. (d) obese (partial r = −0.63, P < 0.01). In the lower panel, night‐time carbohydrate intake remained inversely association with the disposition index, after adjusting for total 24‐h energy intake, among only women in the Ob group: (e) NW (partial r = −0.14, P = 0.60) vs. (f) obese (partial r = −0.62, P < 0.01). OGTT, oral glucose tolerance test; PhiD, dynamic β‐cell response.

Discussion

A growing body of literature implicates maternal obesity and relatively high glucose concentrations as potential contributors to complications during pregnancy and delivery, and as predictive of long‐term risk for obesity in the offspring. In the current study, we showed that late‐night caloric consumption was very prevalent among low‐income African American women in late pregnancy. Furthermore, among only those who were obese in early pregnancy, late‐night carbohydrate intake was positively associated with greater glucose concentrations following the oral glucose load and inversely associated with insulin secretion and the ability of insulin secretion to compensate for reduced insulin sensitivity. Together, these results suggest that late‐night carbohydrate intake may be associated with impaired metabolic health during pregnancy, particularly among women who were obese prior to pregnancy.

In this cohort, almost 25% of daily energy intake during the third trimester of pregnancy was consumed at night. This pattern of intake occurred irrespective of whether the women were normal weight or obese in early pregnancy and is similar to results found in a cohort of overweight African American pregnant women (Allison et al. 2012). It was not possible to determine from the data available whether any of the women suffered from night‐eating syndrome, but in the aforementioned study of African American women, just 4% met the criteria for night‐eating syndrome (Allison et al. 2012), and so it seems unlikely that night‐eating syndrome would explain the high prevalence of night‐time eating found in this cohort. Although the reasons for this night‐time intake are not known, it is consistent with research from the general population showing that food is now consumed more frequently in the United States than it was three decades ago (Popkin & Duffey 2010). In addition to population changes in the frequency of energy consumption, another potential reason for the high prevalence of late‐night food intake during pregnancy could be disrupted sleep. Sleep disturbance is common during pregnancy, and some studies report shorter sleep duration and/or more awakenings as pregnancy progresses (Hedman et al. 2002; Signal et al. 2007; Facco et al. 2010). Women with reduced nocturnal sleep duration during pregnancy are more likely to have impaired glucose tolerance, as indicated by high glucose concentration in response to an oral glucose challenge, and/or greater risk for gestational diabetes (Qiu et al. 2010; Reutrakul et al. 2011; Herring et al. 2014). One limitation of these studies is that night‐time energy intake was not reported so little is known about the respective roles of night‐time sleep vs. night‐time food intake on glucose tolerance in pregnancy. It should also be acknowledged that it is not known whether women in the current study engaged in late‐night eating prior to pregnancy, rather than initiating this meal pattern during pregnancy. Given that previous studies in the general population have linked late‐night eating with obesity and impaired metabolic health (Baron et al. 2011; Morgan et al. 2012; Morris et al. 2012), it is possible that women in this cohort engaged in late‐night eating prior to pregnancy, and for some, this late‐night eating may have contributed to obesity and lower insulin sensitivity. It would be interesting in the future to examine changes in meal patterns, and in particular, meal timing, across pregnancy.

Consistent with previously published studies (Catalano 2010), women from this cohort who were obese in early pregnancy had lower insulin sensitivity. Likely as a compensatory response to this reduced insulin sensitivity, basal insulin secretion was higher among the obese women. Average glucose‐stimulated insulin secretion, however, did not differ between the weight groups, a finding that is consistent with the obese women having a trend towards lower DI or, in other words, an impaired ability of pancreatic β‐cells to secrete sufficient insulin to compensate for reduced insulin sensitivity. This finding is consistent with our previous study of non‐pregnant, non‐diabetic, African American women, for which greater adiposity was associated with reduced DI (Chandler‐Laney et al. 2010).

An association was found between free‐living late‐night energy intake with insulin secretion, and between late‐night carbohydrate intake specifically, with indices of glucose tolerance and insulin action, for only those in the obese group. As mentioned previously, it is not possible to know from this study whether late‐night eating existed prior to pregnancy. However, given that late‐night carbohydrate intake was associated with impaired glucose tolerance and insulin secretion in only the obese group suggests that late‐night carbohydrate intake presents a particular challenge to glucose tolerance of women who are obese, or women who are predisposed to become obese. In this study, late‐night carbohydrate consumption was inversely associated with dynamic phase insulin secretion specifically, and was independent of insulin sensitivity, suggesting that the deficit in insulin responsiveness is likely attributable to an inadequacy of the initial β‐cell response. Previous research suggests that this initial deficit in glucose‐stimulated insulin secretion may be part of the aetiology for type 2 diabetes (Cali et al. 2009). Although the mechanisms underlying the association of late‐night carbohydrate intake with β‐cell deficit are unknown, it is possible that persistent evening carbohydrate intake is more taxing on β‐cells due to reduced insulin sensitivity late in the day (Morgan et al. 2012; Saad et al. 2012). Over time, persistent late‐night carbohydrate consumption might accelerate the progressive β‐cell failure and apoptosis that is known to occur with circadian disruption (Gale et al. 2011; Qian et al. 2013; Rakshit et al. 2014). On a more positive note, however, it is encouraging that at least for this cohort, daytime energy intake was not associated with indices of glucose tolerance and insulin action. On the basis of these findings, it would be interesting in future research to explore whether an intervention to reduce late‐night carbohydrate consumption specifically might be beneficial for the metabolic health of women who are obese during pregnancy, and ultimately, for their offspring.

This study has several limitations that should be acknowledged. Sleep duration was not assessed in the current study but may have also contributed to the association of night‐time eating with impaired glucose tolerance in the obese women. The assessment of food intake for this study was limited to two free‐living days and nights, which, although this diminished participant burden, prevented evaluating the consistency of meal patterns. Further, although research assistants were trained to review food diaries by the same senior investigator, no data were obtained to assess inter‐rater reliability. Finally, this study was completed among low‐income African American women, which may have limited the generalisability of these findings and reduced heterogeneity regarding income level and diet quality.

To conclude, we found that late‐night intake was very prevalent among low‐income African American women during late pregnancy, and late‐night intake of carbohydrate was associated with impaired glucose tolerance among women who were obese prior to pregnancy. It will be of interest in the future to examine whether late‐night food intake is common among all stages of pregnancy and across diverse populations. If limited to low‐income African American women, this pattern of intake might partially contribute to some of the health disparities that exist among ethnic groups. Prevalence across diverse populations, however, suggests the need to incorporate messages about avoiding night‐time energy consumption in general public health efforts. Although this study is the first, to our knowledge, that has shown an association of late‐night carbohydrate intake with impaired glucose tolerance and β‐cell function during pregnancy, this finding is consistent with research regarding shift‐work and circadian rhythm regulation. Together with existing studies, this research suggests that a focus on improving night‐time sleep and reducing late‐night energy intake may improve metabolic health of pregnant women and could potentially improve long‐term health outcomes of her children.

Source of funding

This work was supported by pilot funding from the UAB Diabetes Research Center (P30DK079626), by a career development award to PCC from the National Institutes of Health (K01DK090126), and by laboratory support from the UAB Nutrition Obesity Research Center (P30DK056336).

Conflicts of interest

The authors declare that they have no conflicts of interest.

Contributions

PC and MM designed the study and developed the protocol with input from JB and BG. PC collected the data with the assistance of research assistants and research nurses, and with medical oversight from MM and JB. CS and WG derived dietary and insulin action variables. PC conducted the statistical analyses and wrote the manuscript. All authors read and edited the manuscript.

Chandler‐Laney, P. C. , Schneider, C. R. , Gower, B. A. , Granger, W. M. , Mancuso, M. S. , and Biggio, J. R. (2016) Association of late‐night carbohydrate intake with glucose tolerance among pregnant African American women. Maternal & Child Nutrition, 12: 688–698. doi: 10.1111/mcn.12181.