Chrono-Nutrition, Chrono-Type, and the Prevalence of Type 2 Diabetes Mellitus in a Cross-Sectional Study from the EuroPean Prospective Investigation into Cancer and Nutrition (EPIC) Study

This cross-sectional study was established using a sample of the Spanish cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC) study. The Spanish cohort included 41 437 Caucasian subject, ages 29–69 years who were recruited between 1992 and 1996 from five Spanish regions (Asturias, Granada, Murcia, Navarra, and Gipuzkoa). Further details on the study design and sample characteristics can be found elsewhere [12,13].

The eligible participants for this analysis were described previously [6,14] as younger than 67 years (in the case of men) and younger than 60 years (in the case of women) as of December 2015 (n = 8000). The age restriction was based on evidence showing that the effect of chronotype and social jetlag diminishes in older generations [15], while the different sex-specific cut-off points for ages allowed us to obtain a balanced sample by sex. From them, 5600 were invited to participate, and finally, a total of 4224 (75.4%) accepted to take part in this study. Furthermore, a total of 4031 (95.4%) participants underwent a physical examination by a nurse, and 3772 (89.3%) agreed to a blood sampling procedure extraction.

This study was conducted according to the guidelines of the Declaration of Helsinki, and the study protocol was approved by the Medical Ethical Committee of the Bellvitge Hospital (Barcelona). The approval code is PR073/16 and approval date is 7 April 2016. All the participants signed informed consent.

Data collection was previously described [6,14]. Briefly, between 2017 and 2018, phone interviews were performed by trained professionals to collect information on socio-demographic, reproductive, and lifestyle factors, as well as medical history, meal timing, chronotype [16], and sleeping patterns (used to define chronotype). The Horne–Ostberg’s Morningness–Eveningness Questionnaire (MEQ) and the Munich Chrono-type Questionnaire (MCTQ) were used to collect information on sleeping habits [16]. The sleeping habit information allowed us to calculate mid-sleep time corrected for sleeping on working and weekend days (MSFc), which was used to categorize participants into three chronotypes. The cut-off points of MSFc were at percentiles 10 (2:20 a.m.) and 90 (6:30 a.m.) and classified individuals’ chronotypes as follows: early type, normal type, and late type to calculate the amount of social jetlag accrued by each participant. Social jetlag was determined by estimating the difference between average sleep duration on free days and sleep duration on working days. Sleep quality information was assessed through the Pittsburgh Sleep Quality Index (PSQI) questionnaire [17], a questionnaire of 19 items that gave us information on sleep quality over the previous month. The final score ranges from 0–21 points, and poor sleep quality is considered when the punctuation is greater than five.

Blood samples (n = 3772) were collected to determine serum levels of specific biochemical parameters (i.e., glucose, insulin, triglycerides, and cholesterol) using standard laboratory techniques. The samples were obtained in a fasting state (10–12 h since the last intake), fractioned, aliquoted, and stored at −80 °C following standard protocols [6].

Comprehensive dietary data were collected using the validated diet history questionnaire from the Study on Nutrition and Cardiovascular Risk in Spain (DH-ENRICA^® [DH-E]) [18]. The DH-E questionnaire is based on a previous EPIC-validated dietary questionnaire, but it included larger numbers of items and traditional dishes and cooking methods from Spain. The DH-E considered ten occasions of eating, including those occurring immediately after waking up during the night. Meals were classified according to frequency: less than 3, 4 to 5, and more than 5 [6]. Additionally, the DH-E used food composition tables from Spain and other countries to convert foods to nutrients. Nutrient intakes were assessed for total and simple CHO, fiber, proteins (including animal and plant-based), and lipids (monounsaturated, polyunsaturated, and saturated) for the overall and specific daily intake. The percent of energy from CHO, proteins, and fats consumed at breakfast, lunch, and dinner was calculated, as well as time-dependent energy percentages relative to the daily total energy intake.

The prevalence of T2DM was estimated based on laboratory and self-reported data. Fasting blood serum samples were used to determine levels of glucose, and self-reported diabetes was assessed during the telephone-based interview. A prevalent case of T2DM was considered when one of the following criteria was present: serum glucose concentration ≥126 mg/dL or self-reported diabetes.

For continuous variables, descriptive statistics are presented as N (%) or as the median and percentiles 25 (p25) and 75 (p75). Adjusted means for dietary intake by diabetes status variables were adjusted for the center, sex, age at recruitment (years), educational level (none, primary, technical school, secondary, university, or higher), smoking history (never, former, and current), non-laboral physical activity (MET-h/week), BMI (<25, 25–<30, and ≥30), fat percentage, and hypertension (yes or no).

To estimate the odds ratio (OR) and their confidence intervals (CI), logistic regression was used to determine the associations between different exposure variables (sleep quality, chronotype, chrono-nutrition variables, and macronutrients) and the prevalence of T2DM. Macronutrients were evaluated both as categorical variables, expressed in grams/day (quintiles), and as continuous variables (1-SD increment of controls). For the assessment of the trend test, the categorical variable was used as a continuous one. All models were adjusted for potential confounders, including age at recruitment, sex, center, educational level, smoking history, non-laboral physical activity, BMI, fat percentage, and hypertension. Moreover, in the models assessing the main groups of macronutrients, adjustments were made for additional factors such as alcohol intake (yes/no) and quantity (grams/d), total energy intake (kcal/day), and sleep quality (poor/bad). We initially evaluated additional variables, including laboral and vigorous physical activity, chronotype, social jetlag, number of eating occasions, breakfast time, lunch time, and dinner time. These variables were considered based on their potential role as confounders; however, they were not included in the final models; their inclusion did not materially change the results.

Multiplicative interactions were modeled separately by sex, BMI (<25, 25–<30, and ≥30), smoking status (never, former, and current), and sleep quality (poor or good) with T2DM and evaluated using the log-likelihood ratio test. All statistical analyses were two-sided and evaluated at a significance level of 5%. Statistical analyses were performed using R 3.6.2 and SAS v 9.4.

This study was based on data from 3465 participants (2068 women, 60%), with a median age of 65 (61–68) years. The prevalence of T2DM was 20% (689 cases: 368 males and 321 females) when we conducted this study. The prevalence of T2DM was higher when the sleep quality was poor (28.7%) compared to good sleep quality (19.6%). It was also higher in late-type chronotype (24.2%) compared to normal-type chronotype (19%). Additionally, T2DM prevalence was higher when social jetlag was less than 30 min (21.5%) compared to more than 1:30 h (14%). Furthermore, T2DM prevalence was greater when breakfast and dinner were consumed later than 9 a.m. and later than 9 p.m., respectively, compared to earlier hours. Furthermore, prevalent T2DM cases were one year older on average, had a lower educational level, and had a higher body mass index than subjects without T2DM (). Table 1

provides a comprehensive overview of dietary information by T2DM status, accounting for potential confounders. In terms of carbohydrates, T2DM cases showed lower adjusted means for % energy from CHO (non-cases: 41.61% vs. cases: 40.90%), total carbohydrates (non-cases: 256.62 vs. cases: 250.00), and simple carbohydrates (non-cases: 96.62 vs. cases: 90.10). Conversely, higher means were observed for the percentage of energy from proteins (non-cases:18.50% vs. cases: 18.90%) and animal protein (non-cases: 69.34 vs. cases: 71.43). When we explored meal-specific occasions, we observed that at breakfast, non-cases were characterized by a higher energy intake (non-cases: 378.42 vs. cases: 357.97) and a higher energy contribution (non-cases: 15.46% vs. cases: 14.78%). Non-cases also had a higher percentage of CHO (non-cases: 20.53% vs. cases: 19.09%), simple CHO (non-cases: 27.52% vs. cases: 24.56%), proteins (non-cases: 11.70% vs. cases: 11.20%), and plant-based proteins (non-cases: 14.25% vs. cases: 13.45%) from the total. The same differences were observed when macronutrients were evaluated as grams/breakfast. Additionally, prevalent T2DM cases showed a higher percentage of lipids at breakfast (non-cases: 28.08% vs. cases: 29.45%). At lunch, prevalent cases had a higher percentage of simple CHO from the total (non-cases: 30.20% vs. cases: 31.29%) and a higher percentage of proteins (non-cases: 20.74% vs. cases: 21.10%). However, simple CHO measured in grams/lunch was higher in no-cases. Finally, at dinner, prevalent T2DM cases showed a higher percentage of CHO (non-cases: 21.28% vs. cases: 22.24%), simple CHO (non-cases: 20.44% vs. cases: 22.14%), plant-based proteins (non-cases: 20.62 vs. cases: 21.53), and fiber (non-cases: 19.48 vs. cases: 20.86) compared to the total. When macronutrients were measured in grams/dinner, higher intakes were observed in T2DM cases for total and animal proteins and fiber. Table 2

In the multivariable logistic analysis, as shown in, a higher prevalence of T2DM was associated with poor sleep quality (OR= 2.90, 95% CI = 1.30, 6.28). We observed no difference in T2DM prevalence in relation to time of sleep, chronotype, social jetlag, and chrono-nutrition variables. Table 3 poorgood vs

Association between macronutrients by eating occasions and the prevalence of T2DM () was nearly 60% lower for participants in the highest quintile of carbohydrate intake at breakfast (OR= 0.40, 95% CI = 0.27, 0.59; p-trend = 0.01). Increasing 1-SD from carbohydrates at breakfast was inversely associated with the prevalence of T2DM by 25% (OR = 0.75, 95% CI = 0.66, 0.85; p-trend ≤ 0.01). On the contrary, the OR for lipids at breakfast was 1.13 (95% CI = 1.01, 1.26) per 1-SD increase. No statistical significance was found for associations between macronutrients at lunch and dinner and T2DM. Table 4 Q5Q1 vs

Our findings showed a positive association between poor sleep quality and T2DM prevalence (OR = 2.90, 95% CI = 1.30, 6.28), which is in line with the results of a large cross-sectional study from China [19] and a cohort study from Korea [20], both of which reported that poor sleep quality was associated with higher odds of being diagnosed with T2DM. Further, the US NHANES linked poor sleep quality with the prevalence of clinically identified prediabetes [21]. Briefly, this positive association could be attributed to disruptions in circadian rhythm, influencing insulin sensitivity and consequently leading to an increased risk of T2DM [2]. Another possible explanation is that people who sleep poorly are generally more likely to have an unbalanced diet and eat more foods that raise blood sugar, which correlates with obesity, a risk factor for T2DM [22].

In relation to chronotype, it has been seen that chronotype changes with age, becoming more stable and earlier with increasing age [23]. In that line, we observed that 79.7% of our participants had a similar chronotype (normal type); then, we were not able to find a statistically significant association between chronotype and the prevalence of T2DM.

Our findings regarding macronutrient intake and the prevalence of T2DM showed that a higher intake of lipids at breakfast had a positive association with the prevalence of T2DM; particularly, we observed that for each 1-SD increase in lipid intake, the prevalence of T2DM was up to 13% higher (OR = 1.13, 95% CI = 1.01, 1.26). We observed comparable results with higher intakes of protein, although the association was not statistically significant. On the contrary, T2DM prevalence was 60% lower in participants who had a higher intake of carbohydrates at breakfast (OR_Q5vsQ1 = 0.40, 95% CI = 0.27, 0.59; p-trend = 0.01). The influence of this macronutrient (lipids and carbohydrates) distribution throughout the day on T2DM remains relatively underexplored. The China Health and Nutrition Survey (CHNS), an ongoing cohort study, observed that a higher intake of lipids at dinner compared to breakfast increased the risk of T2DM. In addition, increasing energy from carbohydrates at breakfast (5%) by reducing energy from lipids at dinner was associated with a reduced risk of T2DM [24]. A British cohort showed that eating more carbohydrates in the morning while reducing the consumption of lipids was related to a lower risk of T2DM [25] and, in general, to a lower risk of metabolic syndrome and its components [26]. A recent comprehensive literature review tried to figure out this aspect, concluding that an earlier consumption of carbohydrates might mitigate the risk of obesity, which is one of the main risk factors of T2DM. Moreover, the timing of carbohydrate intake significantly impacts glycometabolic control, with a higher proportion of carbohydrates consumed in the evening, potentially negatively influencing it [27]. These findings underscore the direct association between patterns of carbohydrate consumption and susceptibility to the development of T2DM.

The main limitation of our study was that the cross-sectional design does not allow for the determination of causality. Information on antidiabetic medication use was not available, so the prevalence of T2DM could have been underestimated. Information on genetic predisposition was not available in this study. Consequently, our findings may be affected by residual confounding due to genetic factors. Participants for this study were recruited from a convenience sample, so selection and participation bias may be present in our study. Finally, our study included participants from a Mediterranean cohort, which might hamper the extrapolation of our findings to other populations. However, we obtained information on the diet using a validated diet history, including the timing of intake and specific times of day when foods were consumed, to gather accurate dietary data.

The strengths of this study include its novelty as the first one to examine the relationship between chrono-nutrition and chronotype with T2DM prevalence in a Spanish study. Second, we assessed chronotype and sleep quality using a validated method. Third, we adjusted for potential confounding factors in the analyses, making the association reported robust.