Cross-sectional associations between meal timing patterns and diet quality indices in Iranian women

Table of Contents

Study participants

In this cross-sectional study, 574 healthy women without chronic diseases were recruited from healthcare facilities in southern Tehran between November and December 2024. The target sample size of 495 participants was determined via the formula²⁴ n = ((z_1−α/2+z_1−β ×\(\:\sqrt1-r^2\))/r)² +3, accounting for the correlation between eating frequency and energy intake (r = 0.21)²⁵, with an α level of 0.05 and 95% power (1-β). To accommodate potential under- and overreporting (15%), the sample size was adjusted to 570 participants. The participants were women aged 20–60 years with a BMI below 40 kg/m². Those with a history of chronic illnesses (e.g., diabetes, cardiovascular disease, liver or kidney disorders, stroke, or cancer), pregnant or breastfeeding women, and individuals following specialized diets such as vegan or ketogenic were excluded. Written informed consent was obtained from all participants, and the study was approved by the Research Ethics Committee of Islamic Azad University, Tehran Medical Sciences—Pharmacy and Pharmaceutical Sciences Faculty (IR.IAU.REC.1403.487).

Questionnaires and anthropometric measurements

Key sociodemographic and lifestyle data, including age, education level, marital status, smoking habits, sleep duration (the time of going to bed and waking up was recorded for each recall day), menopausal status, were collected through face‒to-face interviews, and all information was recorded on standardized forms. Total sleep duration was calculated via the following formula²⁶: total sleep duration = (5 × sleep duration weekday + 2 × sleep duration weekend)/7.

Socioeconomic status (SES) was determined through principal component analysis (PCA) based on household income and assets (e.g., cars, homes, and electronics). Physical activity levels were measured via the short form of the validated International Physical Activity Questionnaire (IPAQ)²⁷ where participants reported weekly durations of walking and moderate or vigorous activities. These were expressed as metabolic equivalent minutes per week (MET-min/week) and classified as low (< 600 MET-min/week), moderate (600–3000 MET-min/week), or high (> 3000 MET-min/week)²⁸.

Weight was measured via a Seca scale with participants dressed in light clothing and without shoes (Model: 874 1321009, Seca & Co. KG, Hamburg, Germany). Height was assessed with a wall-mounted stadiometer, which has a precision of 0.1 cm, and participants were asked to be barefoot. Body mass index (BMI) was calculated by dividing weight in kilograms by height in meters squared. Waist circumference (WC) was measured with a non-stretch fiberglass tape positioned at the midpoint between the lower margin of the rib cage and the top of the iliac crest, according to WHO guidelines. Additionally, hip circumference was measured at the widest part of the hips, usually around the level of the greater trochanters²⁹. The waist‒hip ratio (WHR) was calculated by dividing the circumference of the waist by the circumference of the hips.

Dietary assessment

The dietary data were gathered through three nonconsecutive 24-hour dietary recalls conducted on two weekdays and one weekend day. These recalls were administered by trained dietitians in one-on-one sessions to ensure accuracy. The first recall occurred during the participants’ initial visit to the healthcare center, while the remaining two were completed via telephone on randomly assigned days. The eating event of participants identified as—breakfast, lunch, dinner, or snacks—on the basis of their habitual eating patterns. All reported foods were converted into grams per day via standard portion sizes and household measures for consistency³⁰. Nutritionist IV software (First Databank, San Bruno, CA, USA), which was modified to reflect Iranian foods, was used. We identified 430 distinct food items, which were then categorized into 35 specific food groups (Supplementary Table S1).

Meal timing assessment

Variables of meal timing were derived from the average of three 24-hour dietary recalls. The following variables were examined: the proportion of energy intake before 11:00 a.m. and after 6:00 p.m., as well as the duration of pre-sleep fasting, the overall eating window, and the number of EOs throughout the day. The number of EOs was defined as events where participants consumed at least 50 kilocalories, with a minimum separation of 15 min from the previous or subsequent eating event³¹. The duration of the eating window was determined by measuring the number of hours between the first and last EOs on a given day³². Pre-sleep fasting was defined as the interval between the time of the last eating occasion (EO) and the time the participant went to bed. The averages of the meal timing variables were calculated to provide a summary measure for analysis.

Healthy eating index-2015 (HEI-2015)

The HEI-2015 was a measurement used to evaluate diet quality on the basis of how well it aligned with the Dietary Guidelines for Americans. It included 13 components, which were grouped into two categories: adequacy and moderation. Adequacy components, such as fruits, vegetables, whole grains, dairy, and protein foods, reward higher intake, whereas moderation components, such as refined grains, sodium, added sugars, and saturated fats, award higher scores for lower consumption. For example, eating at least 0.8 cup equivalents of fruit per 1000 kcal earned the maximum score for total fruits, while consuming none scored zero¹³. The HEI-2015 offered a detailed method for assessing overall dietary quality. The food groups and scoring criteria are detailed in Table S2.

Carbohydrate food quality score (CFQS-4)

The Food and Nutrient Database for Dietary Studies (FNDDS) 2017–2018 supplied by the US Department of Agriculture (USDA)³³ and Iranian food composition table³⁴ were utilized to categorize foods for this score. The analysis focused on primary carbohydrate sources, grouping foods into grains, snacks, and sweets. It included cooked grains, breads, cereals, savory snacks, sweet bakery products, candies, and desserts. Legumes, certain vegetables, and whole fruits were also counted as carbohydrate sources (Table S3). The excluded items were fruit juices, sugar-sweetened beverages, most dairy products, and baby foods, allowing for a targeted look at foods most relevant to carbohydrate intake and metabolic health. Free sugars were defined as added sugars; sugars from 100% fruit juice; sugars in sweetened beverages, jams, and jellies; and honey, sugars, and syrups. Table S4 presents the components and scoring system of the CFQS-4. The whole grain content of the foods was converted to g/100 g. In brief, the food scoring system assesses nutritional quality on the basis of four criteria: fiber (1 point if ≥ 10 g per 100 g carbohydrate), free sugar (1 point if < 10 g per 100 g carbohydrate), sodium (1 point if < 600 mg per 100 g dry weight), and potassium (1 point if > 300 mg per 100 g dry weight). Each component either met the criteria for a point or scored 0, allowing for a total quality score ranging from 0 to 4, which identified foods with beneficial nutrient profiles¹⁴.

Cholesterol-saturated fat index (CSI)

Food intakes extracted from the average of 24-hour dietary recalls were used to calculate the CSI. The CSI was calculated according to the following formula: CSI = (1.01 × g saturated fat) + (0.05 × mg cholesterol)¹⁵.

Statistical analysis

Statistical analyses were conducted via SPSS version 26 (IBM). Descriptive statistics are presented as the means ± standard deviations (SDs) for continuous variables and percentages for categorical variables. Differences in general characteristics and dietary habits across MTPs were assessed via the chi-square (χ²) test for categorical variables and one-way analysis of variance (ANOVA) for continuous variables. Socioeconomic status (SES) was determined via principal component analysis (PCA), which incorporated household income and asset data. On the basis of the PCA results, the participants were categorized into three SES groups: high, medium, and low. We used the intraclass correlation coefficient (ICC) to measure the consistency in the consumption of foods across days³⁵. The range for the ICC is from 0 (no agreement in food consumption over the days) to 1 (perfect agreement in food consumption over the days). The usual meal timing exposures (energy % intake before 11:00 a.m. and after 6:00 p.m.), the duration of pre-sleep fasting, the overall eating window, and the number of EOs were calculated as described. K-means cluster analysis was utilized to classify MTPs by analyzing the z scores of five meal timing variables³⁶. This method facilitated the identification of distinct participant groups with similar meal timing characteristics, enabling the study of their associations with dietary and health outcomes. A three-cluster solution was selected on the basis of interpretability and theoretical relevance.

Generalized linear regression (GLM) models were applied to examine associations between key variables, adjusting for potential confounders to ensure a comprehensive understanding of the relationships among the studied factors. Directed acyclic graph (DAG) theory was used to define the best adjustment set (Fig. S2)³⁷, i.e., adjustment for age, total energy intake (kcal/d), SES, physical activity, smoking status, education, BMI (as a continuous variable), and total sleep duration (as a continuous variable). In sensitivity analyses, over- and underreporting of energy intake (EI) were assessed via the ratio of the EI to the basal metabolic rate (BMR), which was calculated via the Harris–Benedict formula. Underreporting was defined as EI: BMR < 1.35, and overreporting was defined as EI: BMR > 2.40³⁸. After excluding participants with over- or underreporting (n = 8 underreporting and n = 22 overreporting), the data were reanalyzed. We used AI-assisted tools for language editing, ensuring clarity and grammatical accuracy. The final manuscript was reviewed by the authors for precision and adherence to scientific standards.

A description of the MTPs is presented in Table 1 and illustrated in Fig. 1. The three resulting MTPs from the k-means cluster analysis can be found. Cluster 1, labeled “pre-sleep fasting”, had the longest pre-sleep fasting duration (mean ± SD, 1:31 ± 0:48) and the lowest energy proportion before 11:00 a.m. and after 6:00 p.m. (24.50 ± 8.45 and 26.22 ± 8.23), respectively or the highest energy intake in the midday. Cluster 2, termed “long-frequent”, featured the longest eating window (15:37 ± 1:12) and the most frequent meals (6.97 ± 0.70). Cluster 3 “late-short” had the shortest eating window (12:43 ± 1:01), fewer EOs (6.23 ± 0.75), and the highest energy proportion after 6:00 p.m. (36.59 ± 9.54). The cluster sizes were 24%, 35%, and 39%, respectively. These patterns highlight the varied energy distributions and eating behaviors among participants. By omitting over- and underreporting cases (n = 30), we identified very similar MTPs (Fig. S2, Table S5), and Table S6 presents the demographic and lifestyle characteristics of Iranian women (n = 544).

Table 1 Mean z-scores of usual intakes in the three-meal timing pattern (MTPs).