Study design and population
This study was conducted as part of the nation-wide 10K Project study, the full details of which have been previously described elsewhere [17]. Briefly, this ongoing project involves a large cohort of healthy adult participants with deep multi-omics profiling and long-term follow-up, including onsite meetings every two years. Registration for the study began on October 28, 2018, with participants aged between 40 and 70 years at baseline. The novelty of the project is the combination of innovative medical tests and advanced artificial intelligence methods to discover personal characteristics that can help predict future medical conditions, even before they manifest. The 10K Project is conducted according to the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of the Weizmann Institute of Science.
At the time of this study, 9,988 individuals had been recruited for the 10K Project, and a baseline visit was made. Of these, 8,548 participants had adequate dietary data recorded, with inclusion criteria requiring three days or more of logged 500 to 4000 kcal/day. Furthermore, 1,961 of these participants attended an onsite visit for a two-year follow-up and had subsequent dietary recordings. Six participants were excluded due to a change in weight beyond 25% since it might represent an underlying illness or pathological process.
Dietary assessment
Participants in the 10k Project were instructed to log their food intake in real time over a two-week period during each visit using a designated smartphone app (“Project 10K app”). In total, dietary data included 394,801 days of logging, with a median of 17 days and 1585 ± 606 kcal per day per participant. This app, specifically developed for the cohort, features a database of more than 7,000 foods with full nutritional value and is based on the Israeli Ministry of Health database, which was further expanded with additional items from certified sources. Participants select each food item from the database, noting its weight or portion size, and log it into their user profile. Notably, the application has been employed in multiple studies over the past decade [19,20,21]. The logging data underwent a quality control process, including removing items with improbable weights or improbable timing (e.g., many meals logged within a short time period). Moreover, the dietary data has been correlated with serum metabolites linked to diet, providing an objective validation [22].
Our analyses encompassed 21 macronutrients and micronutrients and 34 distinct food categories. Macronutrients included carbohydrates, protein, and total dietary fat. Values were computed as the absolute amount consumed in grams per day as well as the percentage of total daily energy. Total dietary fiber was evaluated as grams per 1000 kcal consumed per day. Subtypes of dietary fats, including saturated fatty acids (SatFat), monounsaturated fatty acids (MUFAs), and polyunsaturated fatty acids (PUFAs), were computed by weight. The micronutrients included: cholesterol, calcium, magnesium, iron, potassium, sodium, vitamin A (RAE), vitamin B1 (thiamin), vitamin B3 (niacin), vitamin B6 (pyridoxine), vitamin B9 (total folate), vitamin B12, vitamin C, and vitamin E. These values were computed as absolute daily consumption (mg or µg). Certain micronutrients with negligible counts or inconsistent annotations in the food database were excluded from our analysis (e.g., vitamin K, vitamin B7 (biotin), iodine, and trans fatty acids).
All food items logged by participants in the mobile app (with more than 10 counts) were classified into 34 common food categories based on their botanical and nutritional properties (Supplementary Material 1: Table S1-S2). The average daily energy intake was used to evaluate consumption within these categories. Importantly, from a practical perspective, we computed both the energy intake from each food category and its proportion of the total daily intake. Specifically, the category of ultra-processed food (UPF) represented the proportion of calories from foods classified as UPF (based on NOVA classification) out of the total energy intake.
Anthropometric measurements and definitions
Height and weight were measured both at baseline and at the two-year follow-up visit, from which BMI was calculated. BMI was classified according to WHO recommendations as normal weight (18.5-25 kg/m2), overweight (25-30 kg/m2), or obese (≥30 25-30 kg/m2). A small subset of 13 women with a BMI less than 18.5 kg/m2 were grouped into the normal BMI category. Given their near-normal BMI and healthy status at baseline, we believe this minor deviation does not signify actual malnutrition. For the purpose of this study, no weight change was defined as less than a 2% change at the 2-year follow-up visit. Weight loss and weight gain were defined as a 5% or greater weight reduction or gain at the two-year follow-up, respectively.
Other measurements
Monthly household income, highest level of education, back pain for more than 3 months, moderate physical activity for at least 3 days a week, current smoking status, and recent depressive symptoms were self-reported as part of a pre-baseline visit online questionnaire. The recent depressive symptoms score (RDS) was calculated by summing four items (each scored on a 1–4 scale, where 1 = not at all, and 4 = nearly every day) assessing the presence of the following self-reported depressive symptoms over the past 2 weeks: depressed mood, unenthusiasm/disinterest, tenseness/restlessness, and tiredness/lethargy. The resulting sum score ranged between 4 and 16, with higher scores indicating more frequent and severe depressive symptoms. The RDS score has been previously validated against several commonly used depression scales, including the 9-item Patient Health Questionnaire [23].
Blood pressure was obtained at each visit while the participants were in a seated position after 5 minutes of rest. Laboratory results, including low-density lipoprotein (LDL) cholesterol, high-density lipoprotein cholesterol, total cholesterol, triglycerides, glucose, hemoglobin A1C, hemoglobin, albumin, and creatinine, were provided by participants prior to the baseline visit.
Statistical analysis
Statistical analyses were performed using IBM SPSS Statistics for Windows, Version 29.0 (IBM Corp., Armonk, NY) and the Python programming language, Version 3.7 (Python Software Foundation, Wilmington, DE). We compared dietary habits between groups at baseline and changes at the two-year follow-up. Changes in diet at follow-up were assessed as the difference between daily consumption during the two-year follow-up and baseline. We employed the chi-squared test for binary variables and the Kruskal‒Wallis rank sum test for continuous variables. The Mann‒Whitney test was used for post hoc pairwise comparisons between the different groups. We constructed linear models by sex, to predict BMI at baseline incorporating food categories and popular foods as variables (48 food items that accounted for the highest percentage of calories logged by the participants; Supplementary Material 1: Table S3), along with age, education, income, RDS score, physical activity, and smoking status as covariates. For weight change at the 2-year follow-up, we included baseline weight as a variable, along with food categories and popular foods, with age and income as covariates (Supplementary Material 1: Table S4). A correlation matrix between the food categories and popular foods was constructed (Supplementary Material 1: Figure S1). To assess whether our study cohort is representative of healthy adults in Israel, we conducted a comparative analysis of their dietary intake against data from the Israeli national survey, MABAT, a comprehensive dietary assessment of healthy adults in Israel [15]. We focused on the 45-65 age group to closely align with our cohort. Using the mean intake of specific dietary parameters (mean daily energy intake, the percentage of total daily energy intake derived from carbohydrates, protein, and fats and absolute values for sodium) documented in MABAT [15], we established a common intake range (mean ± 25%). We then calculated the number of participants within this range, stratified by sex.
link