Impact of Meal Frequency on Insulin Resistance in Middle-Aged and Older Adults: A Prospective Cohort Study

Data source and study population

Initiated in 2001 by the Korean Center for Disease Control and Prevention, the Korean Genome and Epidemiologic Study (KoGES) embarked on a long-term exploration focusing on two distinct South Korean communities: the rural Ansung cohort and the urban Ansan cohort. Biennially conducted from the baseline in 2001–2002 through the ninth follow-up in 2019–2020, KoGES targeted individuals aged 40 to 69 years. The details of KoGES and the methodology employed therein have been published previously [13]. The selection process of the study population is depicted in Fig. 1. Among the initial 10,030 participants, 4,570 were selected for inclusion after excluding individuals: (1) with a baseline HOMA-IR ≥2.5 (n= 1,314); (2) with implausible total energy intake (<600 kcal/day or >5,000 kcal/day) (n=105); (3) with incomplete baseline data (n=4,181); and (4) those who did not follow-up after the baseline survey (n=869). The KoGES study protocol was reviewed and approved by the Institutional Review Board (IRB number: 9-2024-0115) of the Korea Centers for Disease Control and Prevention, with informed consent obtained from all participants. Additionally, this study received approval from the IRB of Yong in Severance Hospital.

Fig. 1.

Flow chart of the study population. KoGES, Korean Genome and Epidemiologic Study; HOMA-IR, homeostasis model assessment of insulin resistance.

Assessment of diet

A semi-quantitative 103-food item food frequency questionnaire (FFQ), which is a validated and commonly used nutritional assessment tool in several prospective cohort studies, was administered by trained dietitians [14]. The Korea National Health and Nutrition Examination Survey data, representing a cross-section of the Korean population, was utilized to determine serving sizes based on the median quantities of various food items [15]. A questionnaire assessed the amount of each food item consumed per serving, with serving sizes classified as small (1/2 serving), medium (1 serving), and large (2 or more servings). Visual aids were provided to enhance participants’ understanding of serving sizes for different foods.

The food intake frequency, which assessed how often specific foods were consumed over the past year, was categorized into nine groups: never or less than once in a month, once a month, 2–3 times per month, once or twice a week, 3–4 times a week, 5–6 times a week, once daily, twice daily, or 3 times daily. Frequency data obtained from the FFQ were converted into daily frequencies for the calculation of daily nutrient intake. Daily energy intake was determined by converting the reported weight based on the frequency of food consumption and portion sizes. MF was assessed through a dietary habits survey in which participants selected from the options of once, twice, three times, or four or more times per day. Participants were divided into two groups based on a previous study’s cutoff of three daily meals for MF [6,16]. Additionally, a 24-hour recall method was employed to assess participants’ food and nutrient intake over the past day. Participants were asked to recall all foods and beverages consumed during the previous 24 hours, with meal and snack distinctions applied based on participant responses. The total grams of carbohydrates, fat, protein were determined using data from the FFQ. The intake of carbohydrates, protein, and fat was subsequently converted into energy intake in calories (1 g of carbohydrates=4 kcal, 1 g of protein=4 kcal, and 1 g of fat=9 kcal). The proportion of carbohydrates, protein, and fat in the diet was calculated as follows: (calories from carbohydrates, protein, and fat/total calorie intake)×100.

Covariates

Trained medical personnel conducted all health examination procedures. Participants were instructed to wear lightweight clothing and remove their shoes prior to height, weight, and waist circumference (WC) measurements, which were recorded to the nearest 0.1 cm or 0.1 kg. Body mass index (BMI) was calculated by dividing weight in kilograms by height in meters squared. WC was measured three times at the narrowest point between the lowest rib and the uppermost border of the iliac crest, and the average was used. Blood pressure was measured using a Baumanometer mercury sphygmomanometer (W.A. Baum Co., Copiague, NY, USA) following standardized protocols, with participants resting for 5 minutes in a seated position before measurements. Blood pressure was defined as the average of the last two of three measurements. Blood tests were conducted after an 8-hour fast. Total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), C-reactive protein (CRP), fasting plasma glucose (FPG), and insulin levels were determined using a Chemistry Analyzer (Hitachi 7600, Hitachi, Tokyo, Japan until August 2002; and ADVIA 1650, Siemens, Tarrytown, NY, USA from September 2002). Low-density lipoprotein cholesterol (LDL-C) level was calculated using the Friedewald equation [17]. A 75-g oral glucose tolerance test (OGTT) was performed with fasting, and post-load glucose was measured at 1 and 2 hours. The 2-hour post-load glucose level was used to interpret OGTT results.

Smoking habits (every day, some days, former, and never smoker), alcohol intake, and physical activity were self-reported via questionnaire. Total alcohol intake in grams per day was calculated by summing the alcohol consumption of each type of drink. The metabolic equivalent of task (MET) per day was computed for each participant, accounting for the types and intensity of physical activity. DM was defined as FPG ≥126 mg/dL, a plasma glucose level of ≥200 mg/dL at 2 hours after a 75-g OGTT, glycosylated hemoglobin (HbA1c) ≥6.5%, diagnosis by a physician, or current use of oral anti-diabetic medication or insulin therapy. Hypertension was defined as systolic blood pressure (SBP) of ≥140 mm Hg, diastolic blood pressure (DBP) of ≥90 mm Hg, diagnosis by a physician, or use of antihypertensive medication. Dyslipidemia was defined as TC ≥240 mg/dL, TG ≥200 mg/dL, HDL-C <40 mg/dL, LDL-C ≥160 mg/dL, diagnosis by a physician, or use of dyslipidemia medication.

Definition of IR

For assessing IR, HOMA-IR was computed using the following equation: [fasting plasma insulin (μIU/mL)×FPG (mg/dL)/405] [11]. Participants were identified as IR if their HOMA-IR value was ≥2.5, as indicated by previous studies [11,12,18,19].

Statistical analysis

Continuous variables were presented as mean±standard deviation, while categorical variables were expressed as number (percentage). Differences in continuous variables between two for were assessed using independent t-tests or one-way analysis of variance (ANOVA). The chi-squared test was employed for categorical variables.

To compare the cumulative incidence rate of new-onset IR between the two groups, Kaplan-Meier curves and the log-rank test were utilized. Multiple Cox proportional hazard regression analyses were conducted to determine the hazard ratio (HR) and 95% confidence interval (CI) for IR incidence. Model 2 adjusted for age, sex, BMI, WC; Model 3 additionally adjusted for physical activity, smoking, and alcohol consumption; Model 4 further adjusted for hypertension, DM, dyslipidemia, and total energy intake. Finally, subgroup analyses based on sex, DM, and BMI were conducted. We used a linear mixed model was used to identify MF group difference of changes in clinical variables. The Satterthwaite’s method was used to calculated P values for time, MF group, and interaction effect.

All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and R software version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria). A significance level of P<0.05 was set to determine statistically significant associations or differences.

Availability of data and materials

The dataset used in this study was obtained after the review and evaluation of the research plan by the Korea Centers for Disease Control and Prevention (https://www.kdca.go.kr/contents.es?mid=a40504020100).

Baseline characteristics of the study population

Baseline characteristics of the study population are shown in Table 1. Participants within the MF ≥3 group exhibited significantly higher mean values in age, WC, SBP, DBP, serum TG levels, physical activity (METs), and total energy intake. Additionally, they had a higher proportion of men, daily and former smokers, hypertension, and dyslipidemia were present in the MF ≥3 group than in the MF <3 group. Conversely, individuals with MF <3 demonstrated significantly higher mean values in BMI, serum TC, HDL-C, LDL-C levels, and a higher proportion of women, never and someday smokers. There were no significant between-group differences in the mean values of CRP, FPG level, insulin, OGTT 2-hour, HbA1c, HOMA-IR, alcohol consumption, and snack frequency, or in the proportion of individuals with DM. Regarding meal composition, the MF ≥3 group had a significantly higher proportion of carbohydrate intake and significantly lower proportions of fat and protein intake compared to the MF <3 group (Supplementary Table 1).

Table 1.

Baseline characteristics of the study population based on the meal frequency

Longitudinal association between MF and HOMA-IR

In total, 2,257 of 4,570 individuals, with a median follow-up period of 9.75 years, developed IR during the observation period. Table 2 illustrates the incidence of IR over the follow-up period, with 2,257 new cases arising over 48,442.0 person-years, a biennial incidence rate ranging from 3.2 to 14.8.

Table 2.

Incidence of insulin resistance during the follow-up period

Fig. 2 depicts Kaplan-Meier curves alongside the log-rank test, illustrating the cumulative incidence rate of IR based on MF. The cumulative incidence rate of IR was significantly higher in individuals with MF <3 compared to those with MF ≥3 during the follow-up (log-rank test, P=0.037).

Fig. 2.

Kaplan-Meier curve showing the difference in cumulative incidence of insulin resistance risk between the two meal frequency groups.

Table 3 presents the results of the Cox proportional regression analysis for incident IR relative to MF groups. The HRs and 95% CIs for IR incidence in the MF ≥3 group were 0.868 (95% CI, 0.775 to 0.972) compared to those in the MF <3 group. After adjusting for confounding variables, including age, sex, BMI, WC, smoking and drinking status, physical activity, hypertension, DM, dyslipidemia, and total energy intake, the adjusted HRs for IR incidence in the MF ≥3 group were 0.880 (95% CI, 0.782 to 0.990) compared to those in the MF <3 group.

Table 3.

Multiple Cox proportional hazard regression analysis for HOMA-IR incidence according to meal frequency

We conducted a linear mixed model analysis to assess changes over time BMI, WC, SBP, DBP, glucose, insulin, TG, and HDL-C—components of metabolic syndrome—based on MF (Supplementary Fig. 1). Significant group*time interactions were observed for BMI (P=0.007), WC (P<0.001), DBP (P<0.001), glucose (P=0.015), and TG (P=0.003). Both BMI and glucose levels consistently remained lower in the MF ≥3 group compared to the MF <3 group. Additionally, increases in BMI, WC, and glucose were smaller over time, while reductions in DBP and TG were more pronounced in the MF ≥3 group compared to the MF <3 group. These findings suggest that individuals in the MF ≥3 group tend to experience more favorable metabolic outcomes compared to those in the MF <3 group.

Subgroup analysis

Fig. 3 shows the results of the subgroup analyses by sex, DM, and BMI, adjusting for the same confounding variables as in model 4 of Table 3. The adjusted HRs and 95% CIs for the incidence of IR in the MF ≥3 group versus the MF <3 group were 0.747 (95% CI, 0.611 to 0.914) in the male subgroup, 0.879 (95% CI, 0.777 to 0.995) in the subgroup without DM, and 0.815 (95% CI, 0.686 to 0.969) in the BMI <25 subgroup. No significant associations were observed between MF and HOMA-IR in the women, those with DM, and those with a BMI ≥25.

Fig. 3.

Forest plot of subgroup analysis stratified by sex, diabetes mellitus (DM), and body mass index (BMI). HR, hazard ratio; CI, confidence interval.

Insulin and IR

The development of IR is most closely associated with dietary habits [5,20]. The ingestion of glucose and amino acids during a meal stimulates the secretion of insulin from the β-cells of the pancreatic islets of Langerhans. This hormone is intricately involved in maintaining glucose homeostasis. In individuals with normal metabolic function, blood glucose levels are tightly regulated through the control of glucose production by the liver and kidneys, as well as glucose uptake by peripheral tissues such as skeletal muscle, adipose tissue, and the liver. In addition, insulin is involved in lipid metabolism as it both promotes lipid synthesis in the liver and adipocytes and inhibits lipolysis, the breakdown of TG into fatty acids [21]. Chronic inflammation is a precursor to serious health conditions, including cancer, Alzheimer’s disease, and CVD [22,23], and some lipids act as messengers in inflammation. Therefore, efforts to break this vicious cycle through dietary strategy are of paramount importance for mitigating the progression of IR and its associated complications.

Association between MF and IR

MF, which was the main focus of this study, has been proposed as a manageable aspect of dietary behavior that can influence metabolic health management, as it is considered an influencing factor in weight management and various health indicators in the blood [24]. More specifically, high MF is generally thought to be beneficial for maintaining metabolic health [25-27]. In a study involving overweight Hispanic youth, it was confirmed that higher eating frequency, defined as consuming three or more meals per day, was associated with decreased BMI, reduced WC, lower fasting insulin levels, improved HOMA-IR values, and lower TG levels [26]. Additionally, in their study, Holmback et al. [27] reported that higher daily eating frequency was associated with a reduced likelihood of both general and central obesity among middle-aged men. This association persisted after adjusting for relevant confounders such as total energy intake, lifestyle factors, and dietary habits.

These results, which are consistent with our study findings, suggest potential benefits for appetite regulation. A previous study involving healthy males showed that individuals who consumed a single large meal per day consumed more energy during later meals compared to those who consumed smaller, evenly spaced meals [28]. This finding indicates that increasing MF improved satiety control, leading to reduced energy intake during subsequent ad libitum meals. Better appetite control may therefore promote the adoption of “good eating habits.” Several prior studies have suggested that consuming a larger proportion of calories in the morning and eating regularly are beneficial [29-31]. Owing to data constraints, our study did not examine meal timing and regularity because of data constraints. However, it is plausible that individuals who eat more frequently exhibit better appetite control, avoid late-night eating, and maintain regular eating patterns [32]. Conversely, individuals who consume fewer meals per day may be more prone to irregular meal timing and overeating during meal times because of prolonged fasting periods between meals, resulting in postprandial glucose spikes [33]. This increased blood glucose variability may, in turn, cause metabolic detriments [34]. Additionally, postprandial thermogenesis, which comprises 5% to 15% of overall daily energy expenditure and involves an acute rise in resting metabolic rate after eating, has been shown to increase with higher MF [35]. Therefore, dividing the same total caloric intake into smaller, more frequent meals, rather than consuming it all in one large meal, improves glucose utilization [35].

In contrast, some studies have shown no association between MF and metabolic impact, or lower MF resulting in better metabolic outcomes [36,37]. These contradictory results are likely because of the complexity and diversity inherent in dietary research. Studies vary in their definitions of MF, such as whether snack consumption is included, as well as in their study populations. In addition, different studies use various metabolic markers for evaluation, such as BMI, lipid profiles, and insulin levels.

Moreover, dietary habits encompass multiple factors, including meal timing and regularity, as well as differences in diet quality that cannot be explained by caloric intake alone. In a study among Koreans examining the relationship between eating frequency and obesity indicators across different diet quality groups, a significant inverse correlation was observed between eating frequency and body fat percentage, WC, and BMI in the high diet quality group, whereas no such association existed in the low diet quality group [38].

Furthermore, the subgroup analysis in our study indicates that certain individual factors may influence the relationship between MF and metabolic outcomes. Firstly, our study found significant associations only in males, consistent with previous studies [7,39]. Although the exact reasons for this gender difference remain unclear, one possible explanation could be the varying patterns of fat distribution between men and women. Men tend to accumulate excess fat as visceral fat in the central abdominal area, whereas women are more likely to store fat as subcutaneous fat in the gluteofemoral region [40]. Visceral fat in the central region is metabolically more detrimental than fat stored in peripheral areas. If individuals with lower MF are more prone to fat storage, this could explain why reduced MF is linked to cardiometabolic risk factors specifically in men.

Additionally, no significant association was observed in individuals with existing metabolic risks, such as those with DM or obesity. A prior study of patients with type 2 DM indicated that consuming two larger, rather than six smaller, meals per day is more beneficial [37]. Another randomized controlled trial involving overweight/obese women showed that although eating frequency did not affect weight loss, groups with higher eating frequency exhibited better metabolic outcomes in body composition analyses [41]. While these studies share some similarities with our findings, the results of the previous study on DM patients, which suggested that lower MF is metabolically beneficial, somewhat contradict our findings. This discrepancy suggests that individuals with pre-existing metabolic conditions may have altered metabolic responses to MF. A BMI of 25 or higher is classified as obesity and is associated with adverse changes in adipose tissue. These changes include the activation of inflammatory processes, a reduction in lipid metabolism, and the deposition of fat in inappropriate areas [42]. Such metabolic disturbances may alter the body’s usual response to energy intake, potentially leading to unexpected outcomes. Consequently, it is implied that dietary guidelines should be customized for different population groups based on their specific metabolic profiles.

Limitations

This study has several limitations. Firstly, because of the limitations of the collected data, we could not account for dietary patterns such as meal timing and regularity. Secondly, dietary intake was only assessed once at the beginning of the study, which may introduce measurement bias during follow-up and limits our ability to consider factors like meal nutritional composition that could affect IR. Thirdly, despite adjusting for many potential risk factors, we cannot exclude the possibility of unknown confounders. Fourthly, the use of retrospective self-reported questionnaires for collecting lifestyle data, including diet, may have introduced recall bias. Lastly, the participants were all of a single ethnic origin thus limiting the generalizability of the results.

Despite these limitations, the significance of this study lies in its longitudinal examination of the impact of MF on insulin sensitivity, a key factor in metabolic health, and in providing insights into the need for personalized dietary patterns. In an era where metabolic diseases are increasingly prevalent, the importance of dietary habits is more emphasized, underscoring the need for further research based on tailored medical approaches.

In conclusion, among middle-aged and older adults, a higher MF was associated with better insulin sensitivity. However, this difference was significant only in men, individuals without DM, and those with a BMI <25. These results emphasize the importance of dietary behavior in preventing and managing metabolic disorders, suggesting that addressing MF in dietary interventions may improve insulin sensitivity and reduce the risk of related complications. In the future, randomized trials are needed to clarify the causal relationship between these factors and to determine the most effective meal patterns for promoting metabolic health.