Early meal timing improves nocturnal glucose in pregnancies complicated by gestational diabetes.

Introduction

One in ten pregnancies in the USA is affected by gestational diabetes (GDM), which is associated with significant adverse outcomes for the pregnant individual (e.g. pre-eclampsia) and the neonate (e.g. increased birthweight and hyperinsulinaemia) [1, 2]. Lifestyle modifications, including exercise and medical nutrition therapy, are first-line interventions in GDM management to reduce hyperglycaemia and associated pregnancy complications [3, 4]. While many lifestyle interventions have been shown to be efficacious for non-pregnant people with type 2 diabetes, they have not been sufficiently explored in the context of GDM. Currently, lifestyle interventions have primarily focused on modifying carbohydrate quality or amount to achieve pregnancy-specific glucose targets [5]. More recently, studies have demonstrated the potential benefits of lifestyle modifications that shift the timing of caloric intake in non-pregnant people, for example, by restricting the eating interval, eating earlier, or shifting the distribution of calories within the day [6–8]. However, the evidence for the role of temporal eating patterns (e.g. chrononutrition) in achieving glycaemic targets for GDM is insufficient.

The endogenous circadian (~24 h) clock regulates the timing of physiology and behaviour, including hormone release and metabolic responses to food [9, 10]. Endogenous insulin is under circadian control, and its concentration peaks in the morning [9]. Consequently, the human body is better prepared for food intake in the morning when insulin sensitivity is highest. Glucose concentrations are also rhythmic across the day, with predictable responses to fasting and meals [9]. Outside of pregnancy, the timing of when meals are consumed is known to affect blood concentrations of glucose and insulin [11–15]. Food consumption early in the day, often experimentally implemented by consuming breakfast as opposed to skipping breakfast, is associated with lower 24 h and fasting glucose [16, 17], improved insulin response to glucose [18], and better glycaemic management for people with type 2 diabetes [19]. While similar studies have yet to be experimentally completed with pregnant individuals, one study showed that skipping breakfast is associated with an increased risk of GDM [20], and another showed that, among people with GDM, individuals had lower mean blood glucose if they ate a greater proportion of total carbohydrates in the morning (50% vs 10% of the daily total) [11]. These positive health outcomes could be partially explained by the alignment of food consumption to the circadian rhythm [21]. This suggests that interventions that shift meal initiation to the morning may be beneficial for glucose management and time in range (TIR).

Continuous glucose monitoring (CGM) devices have transformed the way pregnant individuals with GDM are able to monitor their glucose levels [22]. People with CGM devices can receive real-time feedback about their glucose metabolism and response to meals and activity. This improves per cent TIR, defined as 3.5–7.8 mmol/l, among pregnant individuals with GDM; this reduction appears to be primarily driven by decreasing time above range (>7.8 mmol/l) [22–24]. CGM also enables monitoring of nocturnal glucose, previously inaccessible with self-capillary blood glucose (SCBG) monitoring. Recent research shows that nocturnal glucose management is important for fetal outcomes [25].

Few studies have examined temporal eating patterns in pregnancies complicated by GDM. We hypothesised that eating the first meal early in the morning may lower overall 24 h interstitial glucose, which could be an intervention to improve 24 h glucose metrics among people with GDM. Therefore, the objective of this study was to investigate the association of temporal patterns of eating and CGM glucose metrics represented as 24 h, diurnal and nocturnal profiles.

Methods

This is a secondary analysis to compare CGM metrics between pregnant individuals with GDM who start their first meal early vs late in the day. The primary study was a single-centre, open-label RCT comparing real-time CGM to SCBG for the glycaemic management of pregnant people with GDM to achieve greater glucose TIR between June 2021 and November 2023 (ClinicalTrials.gov number NCT04605497). Study procedures were approved by the Oregon Health & Science University Institutional Review Board and were performed according to the principles of the Belmont Report and the Declaration of Helsinki. The detailed methods and primary study results are published [24]. Briefly, pregnant people ≥20 and <35 weeks of gestation with a GDM diagnosis were eligible and randomised 2:1 to either real-time or blinded CGM with both groups measuring conventional SCBG at least 4 times per day for glycaemic management. Because the primary study was performed prior to United States Food and Drug Administration clearance of CGM in pregnancy, all participants, regardless of randomisation, were instructed to measure SCBG once daily in the morning (fasted) and 1 h after all meals (postprandial) or when the sensor glucose was outside of range (3.5–7.8 mmol/l) for safety and medication adjustments. Dexcom G6 (real-time) and G6 Pro (blinded) CGM monitoring devices (Dexcom, San Diego, CA) were used to measure CGM, and Contour Next blood glucose monitoring devices (Ascensia Diabetes Care, Parsippany, NJ) were used for all SCBG recordings.

For this study, all participants who had explicitly marked their SCBG values as a ‘meal’ on their glucometer to infer meal timing were included for analysis, irrespective of the primary study’s group assignment (i.e. real-time vs blinded). Additionally, participants with night shift worker status, SCBG measured <2 per day postprandial, outliers, and short intervals between first and last meal were excluded from the analysis (Fig. 1). To prevent outliers from driving group differences, we excluded outliers that were >5 SD from the mean CGM glucose values and SCBG fasting mean; this process only excluded one participant.Fig. 1Flow chart depicting the selection of participants for the study

Flow chart depicting the selection of participants for the study

We generated a mean time for the first meal of the day for each participant by averaging all first-of-the-day postprandial SCBG recording times. We inferred that meals occurred 1 h prior to the SCBG measurement time, as participants were instructed to measure their glucose 1 h postprandial. We conducted a median split using the mean times of first SCBG to create two distinct groups: (1) early eating: early morning eaters whose first mean meal was before 09:56 hours; and (2) late eating: late morning eaters whose first mean meal was after 09:56 hours.

Our primary outcomes are 24 h, diurnal, and nocturnal glucose temporal patterns. Our secondary outcomes are the CGM metrics, including mean glucose, per cent TIR (3.5–7.8 mmol/l), per cent time in hyper- and hypoglycaemia (above 7.8 mmol/l or below 3.5 mmol/l, respectively), glucose AUC for a priori time periods (24 h; daytime from 06:00−22:59 hours, and nocturnal from 23:00−05:59 hours), CV (SD of glucose readings divided by the mean glucose), and mean amplitude of glycaemic excursions (MAGE, the mean magnitude of glycaemic excursions that are greater than the SD of the glucose values) [26, 27]. In the primary study, participants in the real-time group placed a new CGM sensor every 10 days, and the blinded group placed their sensor approximately every 20 days (10 days on- followed by 10 days off-sensor) during the study period. The blinded group was instructed to have no-sensor wear periods to minimise participant burden and optimise retention while still collecting data throughout pregnancy. We removed days with data signal loss greater than 10% to avoid a sensor change period or sensor malfunction [28].

Statistical analyses were performed using R Studio (version 4.3.2) [28] and ClockLab (version 6.1.15, Actimetrics Software, Wilmette, IL). CGM data were imported into ClockLab in 5 min mean bins. The resulting 24 h glucose profile shows the daily waveform composed of the mean glucose values at each bin across all days. Multiple mixed-effect models were used to analyse these data. To analyse the 24 h glucose profiles, we used mixed-effect cosinor analyses with factors of group (early vs late), time-of-day parametrised as cosine and sine, and group × time-of-day interactions [29]. We evaluated a two-harmonic model with 24 h and 12 h components, which can better fit rhythmic data that are not precisely sinusoidal [30, 31]. The primary outcome of interest was the interaction effect (group × time-of-day). Given the more linear nature of the separated nocturnal (23:00−05:59 hours) and diurnal (06:00−22:59 hours) glucose patterns, we evaluated both with a mixed-effect linear model with factors of group (early vs late), and time-into-night or time-into-day parametrised linearly. All models were adjusted for maternal age at enrolment, mean gestational age during study participation, GDM subtype (diet-managed or medication-managed), and primary study group assignment (real-time or blinded). Participant ID was included as a random factor in the models to account for repeated measures.

ANCOVA was used to explore meal timing’s influence on TIR, glycaemic variability, hyper- and hypoglycaemic events, glucose AUC, CV, and MAGE. The primary exposure of interest for these analyses is the group (early vs late) effect on CGM metrics. The ANCOVA models were adjusted for maternal age at enrolment, mean gestational age during study participation, GDM subtype (diet-managed or medication-managed ), and primary study group assignment (real-time vs blinded).

To test the effects of late-night meal timing, we conducted exploratory analyses with both early and late meal timings. Using the mean time for the first meal of the day and the last meal of the day for each participant, we conducted a median split at 09:56 hours for the first meal and 18:44 hours for the last meal to generate four distinct groups: early first meal-early last meal, early-late, late-early and late-late. We repeated the cosinor and linear (nocturnal and diurnal) analyses, adjusting for maternal age at enrolment, mean gestational age during study participation, GDM subtype (diet-managed or medication-managed ), and primary study group assignment (real-time or blinded). Participant ID was included as a random factor in the models to account for repeated measures. For each model, the t value and associated degrees of freedom are reported. Welch’s two-sample t tests were used to analyse continuous variables. Data are presented as mean ± SD unless otherwise indicated. Dichotomous variables are presented as number (per cent) with p values from Pearson’s χ2 test or Fisher’s exact test if any cell count is less than 5. Statistical significance was defined using an alpha a priori as p < 0.05.

Results

Of the 111 participants in the parent study, 71 participants were included in analyses after exclusions (Fig. 1). Participant characteristics, including age, race/ethnicity and glycaemic status, were similar between the two groups and reflect the overall study population of the parent trial characterised as primarily white, with obesity and with early diagnosis of GDM before 24 weeks of gestation (Table 1). The early eating group was more likely to have a first-degree family history of diabetes (58.3% vs 28.6%, p=0.022) compared with the late eating group. Table 1Participant demographics and GDM characteristicsCharacteristicCombined sample(n=71)Early eating(n=36)Late eating(n=35)p valueAge, years32.8 ± 5.233.9 ± 4.131.7 ± 5.90.079Self-identified race and ethnicity0.274 Asian9 (12.7)5 (13.9)4 (11.4) Black1 (1.4)1 (2.8)0 (0.0) Hispanic or Latinx7 (9.9)2 (5.6)5 (14.3) Other8 (11.3)2 (5.6)6 (17.1) White46 (64.8)26 (72.2)20 (57.1)Insurance status0.098 Medicaid14 (19.7)3 (8.3)11 (31.4) Private insurance47 (66.2)28 (77.8)19 (54.3) No insurance2 (2.8)1 (2.8)1 (2.9) Other2 (2.8)1 (2.8)1 (2.9) Unknown/missing6 (8.5)3 (8.3)3 (8.6)Nulliparity32 (45.1)16 (44.4)16 (45.7)>0.999Pre-pregnancy BMI, kg/m231.8 ± 9.030.2 ± 7.933.4 ± 9.80.140Approximate total gestational weight gain, kga9.9 ± 10.110.0 ± 8.49.8 ± 11.70.9419History GDM28 (39.4)12 (33.3)16 (45.7)0.410First-degree family history of diabetes31 (43.7)21 (58.3)10 (28.6)0.022*Substance useb7 (9.9)3 (8.3)4 (11.4)0.710Chronic hypertension11 (15.5)4 (11.1)7 (20.0)0.480Early HbA1c level, mmol/mol34.3 ± 4.534.3 ± 4.434.1 ± 4.50.895Early HbA1c level, %5.3 ± 0.45.3 ± 0.45.3 ± 0.40.895Gestational age at GDM diagnosis, weeks20.8 ± 7.122.0 ± 7.419.6 ± 6.60.147OGTT, mmol/l Fasting5.29 ± 0.795.24 ± 0.915.34 ± 0.640.623 1 h9.56 ± 1.679.47 ± 1.519.64 ± 1.830.686 2 h7.64 ± 1.487.85 ± 1.287.44 ± 1.650.269Gestational age at enrolment, weeks27.5 ± 4.927.8 ± 5.227.1 ± 4.60.505Primary study ‘real-time’ group assignment, count48 (67.6)26 (72.2)22 (62.9)0.556Continuous variables are presented as mean ± SD with p values from Welch’s t test. Dichotomous variables are presented as number (per cent) with p values from Pearson’s χ2 test or Fisher’s exact test if any cell count is less than 5Seven participants were missing early HbA1c values, four participants were missing any OGTT results, and an additional three were missing 1 and 2 h OGTT resultsaThe approximate total gestational weight gain was calculated as the weight at delivery minus the pre-pregnancy weight (defined as weight within 3 months of conception or in the first trimester) that was self-reported or abstracted from the medical record. One participant in the early eating group was missing delivery weightbAny tobacco, alcohol, illicit drug, or cannabis use in pregnancy*Significant results, p<0.05

Participant demographics and GDM characteristics

Continuous variables are presented as mean ± SD with p values from Welch’s t test. Dichotomous variables are presented as number (per cent) with p values from Pearson’s χ2 test or Fisher’s exact test if any cell count is less than 5

Seven participants were missing early HbA1c values, four participants were missing any OGTT results, and an additional three were missing 1 and 2 h OGTT results

aThe approximate total gestational weight gain was calculated as the weight at delivery minus the pre-pregnancy weight (defined as weight within 3 months of conception or in the first trimester) that was self-reported or abstracted from the medical record. One participant in the early eating group was missing delivery weight

bAny tobacco, alcohol, illicit drug, or cannabis use in pregnancy

Electronic supplementary material (ESM) Table 1 summarises compliance between the primary study’s group assignments. As expected, significantly more data are available for the real-time group, who placed a new CGM sensor every 10 days, than the blinded group, who placed their sensor approximately every 20 days (10 days on- followed by 10 days off-sensor) during the study period. Notably, there was no difference in the proportion of real-time vs blinded assignment in the early and late eating groups (Table 1).

Postprandial SCBG times fell into typical breakfast, lunch, and dinner clusters (Fig. 2a), but with variability in the mean time of first meal and last meal (Fig. 2b). The first mealtime for early eating ranged from 06:22 to 09:45 hours, and for late eating, the first mealtime ranged from 09:56 to 14:32 hours (p<0.001) (Table 2). The last mealtime was earlier in the early eating group compared with late eating (18:23 ± 01:12 hours vs 19:12 ± 01:28 hours, p=0.013). However, the mean interval between the first meal to the last meal was significantly longer in the early eating group (9 h 55 min ± 1 h 06 min vs 8 h 18 min ± 2 h 02 min, p<0.001). See Table 2 for SCBG recording times.Fig. 2(a) Histogram of all postprandial SCBG recording times for every participant. Colours indicate the first recording of the day (blue), last recording of the day (red), and recordings that are neither first nor last (grey). (b) Mean eating intervals for each participant (n=71) in order by mean first (blue) to last (red) SCBGTable 2Glucose concentration characteristics from CGM and SCBG recordingsCharacteristicCombined sample(n=71)Early eating(n=36)Late eating(n=35)p value24 h %TIR (3.5–7.8 mmol/l)92.4 ± 6.993.1 ± 6.291.6 ± 7.60.384Mean 24 h glucose, mmol/l5.73 ± 0.55.64 ± 0.55.83 ± 0.60.127Mean CV, %16.8 ± 2.916.6 ± 2.917.1 ± 2.90.472MAGE, mmol/l1.70 ± 0.371.66 ± 0.361.74 ± 0.390.365Days CGM worn48.5 ± 31.553.9 ± 31.642.9 ± 30.70.139Percentage of CGM wear days to potential days79% ± 22%84% ± 15%74% ± 27%0.055Total glucose data points15,027 ± 962716,613 ± 976013,397 ± 93470.161Total included glucose data points13,863 ± 902715,417 ± 912412,264 ± 87680.1422Percentage of glucose data used92% ± 7%93% ± 3%91% ± 9%0.283Mean time: first meal (hh:mm, range), hours09:56(06:22–14:32)08:31(06:22–09:45)11:23(09:56–14:32)<0.001*Mean time: last meal (hh:mm, range), hours18:48(13:42–22:41)18:23(13:42–21:00)19:12(16:27–22:41)0.013*Mean interval: first to last meal9 h 07 min ± 1 h 49 min9 h 55 min ± 1 h 06 min8 h 18 min ± 2 h 02 min<0.001*SCBG fasting glucose, mmol/l4.94 ± 0.384.92 ± 0.374.96 ± 0.390.727First postprandial SCBG reading, mmol/l6.60 ± 0.596.53 ± 0.526.68 ± 0.650.278Last postprandial SCBG reading, mmol/l6.65 ± 0.576.62 ± 0.566.70 ± 0.590.521Continuous variables are presented as mean ± SD with p values from Welch’s t test*Significant results, p<0.05

(a) Histogram of all postprandial SCBG recording times for every participant. Colours indicate the first recording of the day (blue), last recording of the day (red), and recordings that are neither first nor last (grey). (b) Mean eating intervals for each participant (n=71) in order by mean first (blue) to last (red) SCBG

Glucose concentration characteristics from CGM and SCBG recordings

Continuous variables are presented as mean ± SD with p values from Welch’s t test

The mean 24 h glucose concentration profile stratified by group is plotted in Fig. 3a; raw data are provided in ESM Fig. 1. There were significant main effects of time-of-day (24 h component: −0.32 mmol l−1 min−1, t102,232=−188.9, p<0.001; 12 h component: −0.11 mmol l−1 min−1, t102,232=−65.2, p<0.001) and of group × time-of-day (24 h component: 0.09 mmol l−1 min −1, t102,232=37.9, p<0.001; 12 h component: 0.04 mmol l−1 min−1, t102,232=15.3, p<0.001), but there was no group effect (0.11 mmol/l, t65=0.99, p=0.324). The significant interaction term indicates a change in phase or amplitude that depends on group: in the early eating group, the trough glucose concentration was lower, and the 24 h profile was phase advanced (leftward shift towards earlier) compared with the late eating group (Fig. 3b). The model for early eating had a minimum glucose concentration at 04:51 hours and a peak at 20:25 hours. The model for late eating had later minimum and maximum glucose concentrations (05:56 and 21:24 hours, respectively).Fig. 3(a) 24 h mean glucose concentration for early eating (blue, n=36) vs late eating group (red, n=35). Data are presented as mean ± SEM. (a, b) Shading demarcates the nocturnal (23:00–05:59 hours) intervals. (b) Double harmonic cosinor model fit for 24 h glucose concentration for early eating (blue) vs late eating (red). The model was adjusted for maternal age at enrolment, mean gestational age during study participation, GDM subtype (diet-controlled or medication-controlled), and primary study group assignment (real-time or blinded). Participant ID was included as a random factor to account for repeated measures

(a) 24 h mean glucose concentration for early eating (blue, n=36) vs late eating group (red, n=35). Data are presented as mean ± SEM. (a, b) Shading demarcates the nocturnal (23:00–05:59 hours) intervals. (b) Double harmonic cosinor model fit for 24 h glucose concentration for early eating (blue) vs late eating (red). The model was adjusted for maternal age at enrolment, mean gestational age during study participation, GDM subtype (diet-controlled or medication-controlled), and primary study group assignment (real-time or blinded). Participant ID was included as a random factor to account for repeated measures

During the daytime, there was a significant time-of-day (7.0 × 10−4 mmol l−1 min−1, t72,418=150.8, p<0.001) and group × time-of-day effect (7.0 × 10−5 mmol l−1 min−1, t72,418=10.0, p<0.001), but no group effect (0.01 mmol/l, t65=0.06, p=0.950). This was evident by an earlier rise in the early eating group resulting from the leftward shift in the 24 h profile.

Overnight, glucose levels decreased by approximately 0.67 ± 0.39 mmol/l in both groups. The late eating group, however, had significantly higher nocturnal glucose compared with the early eating group. This was reflected by a significant group (0.26 mmol/l, t65=2.3, p=0.023), time-of-day (−0.09 mmol l−1 min−1, t29,818=−119.0, p<0.001), and group × time-of-day effect (−0.01 mmol l−1 min−1, t29,818=−11.8, p<001). Controlling for maternal age, mean gestational age during study participation, medication usage and primary study group assignment, the late group decreased at a faster rate than the early group (−0.11 mmol l−1 min−1 vs −0.09 mmol l−1 min−1, respectively) but started at a higher glucose concentration. By comparison, the nocturnal glucose concentration in the early eating group reached the minimum earlier and remained level for the remainder of the nocturnal interval.

Compared with the analysis of the 24 h glucose concentration profile, group was not associated with other daily mean metrics generated from CGM, including 24 h, nocturnal or diurnal TIR, glycaemic variability, per cent time spent in hyper- and hypoglycaemia, and glucose AUC (ESM Table 2). Furthermore, the postprandial SCBG glucose readings were not different between early and late eating groups for either the first (near daily minimum) or the last (near daily maximum) SCBG.

Our exploratory analyses of early-early, early-late, late-early, and late-late eaters are summarised in ESM Table 3. For 24 h and diurnal glucose concentration, there were significant time and group × time-of-day effects (see ESM Table 3 for values), with no significant group effects. In comparison, there was a significant group effect for nocturnal glucose, whereby the late-late eating group was associated with significantly higher nocturnal glucose compared with the early-early eating group (0.40 mmol/l, p=0.004).

Discussion

Lifestyle modifications are first-line interventions for GDM management. However, specific, feasible lifestyle interventions that are not focused primarily on carbohydrate modifications are needed. Glucose concentration reliably follows a circadian rhythm that rises during the day and falls at night with superimposed postprandial excursions. We found that early eating is associated with a glucose rhythm that is shifted earlier and downwards, leading to significantly lower nocturnal glucose concentrations (23:00−05:59 hours), despite a longer eating interval in the early eating group. This work suggests that meal initiation during the day is a potential lifestyle choice to improve glycaemic goals. This is the first study to examine how the timing of the first meal affects 24 h glucose concentration measured via CGM during pregnancy in individuals with GDM.

The benefits from early eating are thought to arise from the alignment of food consumption to the internal circadian rhythm. Food ingestion is a powerful signal, directing circadian rhythms in peripheral organs [12, 32]. Temporal misalignment between food ingestion and the central circadian day/night cycle is thought to contribute to metabolic dysregulation [32, 33]. Indeed, late and night-time food consumption, which commonly occurs with shift work, are associated with worse glycaemic responses and development of long-term health consequences in both the non-pregnant [15, 18, 19, 34–36] and the pregnant population [37–39]. Our results support that early eating within the day may benefit glucose management, specifically through lowering nocturnal glucose levels. This is complemented by literature linking the increased incidence of GDM with skipping breakfast [20].

Time-restricted eating (TRE) is an intervention that encourages long daily fasting intervals by curtailing the time during which food is consumed to 8, 6 or even 4 h per day. TRE is effective in improving glucose levels in people with, and at risk for, type 2 diabetes [40–48], but restricting the eating interval and enforcing long daily fasts may not be appropriate or tolerable during pregnancy. Previous qualitative research reported that the primary concerns for adopting a TRE regimen during pregnancy are safety, hunger and nausea [49]. Skarstad et al explored the influence of a 5 week TRE intervention on cardiometabolic outcomes in a pregnant population at risk for GDM [6]. While they found no influence of TRE in cardiometabolic measures, notably over half of the participants self-selected their TRE interval to begin later than their baseline eating interval. Participants were able to adhere to the TRE interval for 4.7 days/week, and no adverse events were reported. In the current study, interstitial glucose was lower in the early eating group without the imposition of TRE; early eaters actually had a longer eating interval than the late eaters. Importantly, this emphasises that interventions that are effective for type 2 diabetes may not immediately translate to GDM. Furthermore, our results support meal initiation earlier in the day as a potential lifestyle choice to improve glycaemic goals without the need to restrict the eating interval.

We further examined the temporal eating distribution patterns by splitting our cohort into four groups with combinations of early and late eating for the first and last meal. These analyses support the finding that the timing of meals is an important regulator of daily glucose concentration, because the late-late group had higher nocturnal glucose compared with the early-early group. Nevertheless, these exploratory results should be interpreted with caution, given the small and unequal sample size of the groups. Further studies would be needed to corroborate the results and to fully parse out the contributions of early and late meal combinations.

There are a wide variety of chrononutrition techniques that can be implemented to reduce hyperglycaemia during pregnancy. Our work suggests early eating is one possible intervention for improving glucose management. It may be beneficial to use in combination with other eating interventions, such as the distribution and quality of macronutrients throughout the day. Previous interventions with positive results include a high complex carbohydrate diet [5] and consuming 50% of daily carbohydrates in the morning [11]. However, neither of these interventions affected nocturnal glucose. While we did not explore either the quantity of food or macronutrients in this study, it may be prudent to explore the intersection between mealtime and meal composition in future research studies to develop multimodal lifestyle approaches to glycaemic optimisation in pregnancy.

An important factor of this study is the implementation of CGM during pregnancy. The conventional method of SCBG misses nocturnal glycaemic events as the individual is sleeping. The introduction of CGM to the management of GDM allows individuals to improve their prognosis and assess interventions by receiving real-time feedback [22]. Our findings suggest that analysing the temporal patterns of nocturnal glucose concentrations may provide more valuable insights than simply relying on a single, early morning fasting value. We observed that early eating significantly impacted nocturnal glycaemic levels. Increased nocturnal glucose is associated with increased fasting glucose [8] and large-for-gestational age infants [25]. Therefore, by targeting night-time glucose concentrations, we may be able to improve GDM outcomes.

The present study has several limitations to consider. First, this was an observational study with no intervention, limiting any conclusions for the efficacy of a mealtime intervention during pregnancies complicated by gestational diabetes. Furthermore, meal timing was inferred from SCBG timing, and we depended on participant compliance with study instructions (i.e., measure SCBG 1 h postprandial). There were no formal measurements of participant compliance, which could reduce the accuracy of the mealtime estimates. To minimise this effect, we chose to conduct a median split instead of treating first mealtime as a continuous variable. By dichotomising meal timing, we could minimise temporal noise. As a result of this, we are unable to interpret incremental changes in the relationship between first meal timing and 24 h glucose. Additionally, the primary study did not collect data regarding an individual’s circadian timing, meal composition, or sleep time, which may be important contributing factors to the effects of meal timing and glucose levels. Recent work has suggested that improving sleep health in individuals with GDM may be protective against developing type 2 diabetes [50]. This may be important because in healthy adults, one night of sleep restriction and circadian misalignment can increase postprandial glucose response and may contribute to metabolic dysregulation [15]. Despite these limitations, our study demonstrates that meal timing may contribute to nocturnal glucose management in pregnancies complicated by GDM.

Nutrition modifications and incorporation of exercise are key interventions for GDM management. However, expanding behavioural modification strategies that impact glucose metabolism are sorely needed for individuals with GDM to reduce hyperglycaemia and reverse health risks for the pregnant individual and fetus. The present study demonstrates that early eating is associated with lower nocturnal glucose as measured by CGM in individuals with GDM. As treatment options for this at-risk population are limited, meal timing may be a viable, non-invasive lifestyle intervention.