Evaluation of the Effect of Seasonality on Biological Age in Adults

Overview

This research aims to provide insights on how seasonal variations influence biological age and enhance the design and analysis of long-term lifestyle interventions targeting biological clocks. Aging is a gradual decline in cellular and organ functions, significantly increasing the risk of non-communicable and infectious diseases. Recent research has focused on identifying aging biomarkers that can better predict functional capability in healthy individuals. Biological age clocks, which can be measured from samples like blood or saliva, are emerging as valuable tools for assessing the pace of aging and calculating age acceleration-the difference between chronological and biological age. These clocks utilize molecular and clinical data, including DNA methylation and plasma proteomics, to predict future health outcomes, such as disease risk and mortality. Various DNA methylation-based clocks have been developed, with the Dunedin Pace of Aging (PoAm) offering a more precise modeling of physiological changes over time. Lifestyle factors, including diet and physical activity, can influence age acceleration, suggesting that lifestyle interventions may impact biological aging. Current evidence indicates that three specific epigenetic clocks-PhenoAge, GrimAge, and Dunedin PACE-are particularly effective in detecting beneficial effects on aging trajectories. However, the stability of these clocks during long-term lifestyle interventions remains unclear, as they can exhibit variability over short periods and may be affected by factors influenced by seasonal changes, such as Vitamin D levels, climate, and white blood cells composition. To investigate these seasonal effects on biological age, a proposed observational study will track changes over a 12-month period in middle-aged and older adults.

Quality measures: the investigator will ensure protection of participant's personal data and that all reports, publications, participant samples, and any other disclosures, except where required by laws. The processing of personal and medical data will follow the Swiss regulation in matter of data privacy, and in conformity with the ORH (Human Research Ordinance) regulation, Art. 5. Participants are identified only by a participant identification number and site identification number to maintain participant confidentiality. On the CRFs and other project specific documents, participants are only identified by their unique participant number. All participant study records will be kept safely in an access-controlled area. Identification code lists linking participant names to participant identification numbers will be stored separate from participant records. The data will be coded. The participant identification list will only be accessible by the site team and identifiable information will only be kept in the investigator file. The investigator file is a secure and certified electronic system (Veeva Vault) dedicated for protecting that study information from unauthorized or accidental disclosure, from alteration, deletion, copying, and theft. The coded data are captured into the Electronic Data Capture (EDC), Medidata RAVE ensures traceability (audit trail). The data recorded into the EDC are only accessible to the Data Manager, site personnel, and study members. After the analyses of biological samples, the results are captured into another system named LabKey in a password-protected manner and only accessible by the laboratory technicians responsible for analyses. The final laboratory data will then be sent from LabKey to the assigned data manager via a secured SharePoint containing audit trail. RAVE and LabKey, as well as the external results are integrated into LSAF, a software dedicated to the statistical analysis, which is password controlled, and accessible only to the statistician and the Data Manager. The servers of each software and their back-up servers are hosted i.e. the United States (Medidata RAVE), Germany (LSAF), and Switzerland (LabKey). Biological material in this project is not identified by participant name but by a unique participant number. Biological material is appropriately stored in a restricted area only accessible to authorized personnel. In case of data transfer, Société des Produits Nestlé S.A. will maintain high standards of confidentiality and protection of participant personal data. Clinical information will not be released without the written permission of the participant, except for monitoring by Regulatory Authorities or the study sponsor. The Principal Investigator or designee will implement and maintain appropriate technical and organizational measures to ensure accessibility to participants of their personal data and their results of biological analysis collected during the study. Upon request, the Principal Investigator or designee will ensure to provide extracts of personal data to participants. The Principal Investigator or designee will implement appropriate procedures to immediately inform the sponsor about any participant's request to rectify or delete their personal data or biological samples during the study. The investigational site files, including the list of codes, will be kept, and archived separately in a secure repository compliant to Nestlé policies with respect to data privacy by the research team for 25 years, even if the candidate is not included in the research project. The Trial Master file will be kept separately for 25 years as per Nestlé policy. In addition to the current study, the participants will be proposed to consider providing consent for "further re-use" of their biological samples and data in the broader scope of Nestlé Center of Biological Resources (NCBR); if they consent to this, the samples and health related data will be handled as described in the biobank regulations (CERVD opinion - Req. 2020-00307). If they do not consent, the data will be used only in the frame of this research. No data will be further used without the Ethics Committee approval and the participants' consent. Sample size calculation: The sample size for this study is based on three sources of information. Koncevičius (2024) showed that biological age clock acceleration is subject to circadian rhythm related variation (over a day) of 1 year of amplitude. Reed (2022) shown that within subject correlation of epigenetic clock is high (up to a correlation of 0.93) across repeats spaced up to 15 years apart. Jokai (2023) shown that PhenoAge clock acceleration variability in middle exercised healthy individuals from both sexes was about 5.4 years (average observed between Male and Female). In order to demonstrate a 1 year variation between Winter and Summer in biological age assuming a variability (standard deviation) of 5.4 years and a within participants correlation of 0.93, using a paired t-test approach, 80% power and alpha set to 0.0166 (0.05/3) in order to account for the multiplicity induced by the repetition across 3 clocks (PhenoAge, GrimAge and DuneDinPace -- Bonferroni correction), one need to end up with 46 subjects in the full analysis set (FAS). Accounting for a 20% drop out rate (subjects enrolled but without biological age calculated at any given point in time), one needs to recruit 46/0.8=58 subjects (rounded up to 60 subjects). SAS 9.4 (LSAF v5.2.1) PROC POWER was used for the calculations. Statistical analyses plan: The main analysis of the primary endpoint will be evaluated on the Full Analysis Set population. A sensitivity analysis will be conducted replicating the main analysis on the PP set and an exploratory analysis will be conducted in order to quantify the individual contribution of various factors on the overall seasonal effect. For PhenoAge and GrimAge, biological age acceleration will be calculated at individual level using a linear mixed model with biological age (years) as a response and Chronological age (years) as a confounder. The subject identifier will be included as a random effect to account for the correlation between consecutive measures of biological age within the subject. The Kenward-Roger method will be used for the degrees of freedom calculation. For a given individual, at a given chronological age, the biological age acceleration will be the fixed regression residual. DuneDinPace is already measuring an age acceleration, therefore the above-mentioned transformation is not applicable. The biological age acceleration is assumed to follow a normal distribution. The difference in biological age acceleration between seasons will be evaluated with a linear mixed model in order to account for the longitudinal nature of the data and intra-subject correlation. Bbiological age acceleration will be used as the outcome variable, and season associated with the sampling, BMI, household income, residence, sex, physical activity, smoking status, and ethnicity used as fixed confounders independent from seasonality. The subject's identifier will be used as random effect to account for the intra-subject correlation of the outcome. Kenward-Roger method will be used for the calculation of degrees of freedom. The mixed linear model will be fitted using the restricted maximum likelihood algorithm, allowing for the use of partial information (subjects with incomplete trajectory, e.g. 1 season available out of the two needed). This method is unbiased if the missing data is missing at random. The presence of seasonal effect will be concluded if the least square means difference of biological age acceleration between winter and summer is statistically significant. P-values inferior to 0.05/3 (Bonferroni correction) will be interpreted as indicative of a difference in biological age acceleration for the clock used. In the presence of far-outliers, a sensitivity analysis will be conducted precluding the observation(s). Consistence in the conclusions between this sensitivity analysis and the main analysis of the primary endpoint will be indicative of the absence of impact of the outliers on the conclusions. Several factors can underly an observed seasonal effect such as variation in physical activity, BMI, smoking status, sleep and diet. An exploratory analysis will be conducted in order to explore the individual contribution of these factors to the overall seasonal effect as quantified with the main analysis. To mitigate potential bias, a variable selection method (elastic net) will be applied to the linear mixed model described above with the additional inclusion of living environment, household income, BMI, Smoking status, Physical activity level, sleep index, diet index and travel abroad (measured at the visit) as fixed covariates. Random 5-fold cross validation will be used for selecting the reduced model and the appropriate value for the ridge regression parameter. The beta coefficients associated towith each fixed covariate retained in the reduced model will be presented.