Birgit Linkohr

Proteomic aging clocks have emerged as biomarkers of aging, but most rely on cross-sectional data. Here, we present a proteomic aging clock based on longitudinal measurements of liquid chromatography–mass spectrometry plasma proteomics at up to three time points in the population-based KORA cohort, comprising 9,105 samples from 4,100 individuals in the final dataset aged 25–88 years. The model predicts chronological age and yields a measure of proteomic age acceleration associated with lifestyle and clinical risk factors. Risk factor–stratified analyses reveal shared proteomic signatures among cardiometabolic exposures and distinct signatures for smoking and physical activity. Within individuals, dyslipidemia and impaired kidney function track transitions to increases in age acceleration, while higher age acceleration is associated with increased all-cause mortality. Mediation analysis suggests that risk factors are linked to mortality indirectly through age acceleration. Affected proteins and pathways highlight extracellular matrix remodeling, complement and coagulation pathways, and liver-derived plasma proteins as key contributors to aging. Mendelian randomization identifies proteins with evidence consistent with a causal role in aging-related traits. Together, these findings are consistent with biological aging being dynamically responsive to changes in modifiable risk factors.

Aims/hypothesis Environmental exposures have been adversely associated with the risk of diabetes but their role in the remission and progression of prediabetes (impaired glucose tolerance [IGT] and/or impaired fasting glucose [IFG]), especially different phenotypes, is poorly understood. This study aimed to investigate the relationship between environmental exposures and remission or progression of prediabetes phenotypes. Methods Data were derived from the KORA (Cooperative Health Research in the Region of Augsburg) cohort in Southern Germany (2006–2022). Glucose tolerance status was defined according to the 1999/2006 WHO criteria. We included annual values of particulate matter (PM), light at night (LAN), normalised difference vegetation index (NDVI), imperviousness (IMP) and air temperature variation (T_sd). We fitted complementary log–log regression models for associations between IQR changes in exposures and the incidence of prediabetes phenotypes (isolated impaired fasting glucose [iIFG], isolated impaired glucose tolerance [iIGT], or their combination [IFG+IGT]), remission to normal glucose tolerance (NGT) or progression to type 2 diabetes. We also applied Quantile g-computation regression models for joint association analysis. Results During the follow-up, 370/1618 participants developed prediabetes (124 iIFG, 199 iIGT and 47 IFG+IGT). Among participants with prediabetes, 136/367 (without medication) regressed to NGT and 133/420 progressed to type 2 diabetes. We observed consistent associations of air pollutants and the built environment with iIGT or IFG+IGT. For example, an IQR increase in PM_2.5 (aerodynamic diameter ≤2.5 μm) showed an HR of 1.89 (95% CI 1.07, 3.31) for the risk of IFG+IGT and 1.59 (95% CI 1.03, 2.46) with a decrease in NDVI. Co-exposure to environmental mixtures (higher air pollution, T_sd, IMP and LAN, and lower NDVI) was associated with increased risks of iIFG, iIGT and IFG+IGT. Additionally, lower NDVI (HR 0.52 [95% CI 0.28, 0.96]), higher LAN (HR 0.22 [95% CI 0.08, 0.62]) and higher IMP were associated with reduced remission of iIFG but we found no statistically significant associations with the progression to type 2 diabetes. Conclusions/interpretation The findings indicate potential heterogeneity in environmental exposures and prediabetes phenotypes. Specifically, they are associated with increased incidence of IGT and decreased remission of iIFG. Graphical Abstract

Objective: Upper reference values for thyroid volume are 25 mL for men and 18 mL for women. Thyroid volume alters with age, body weight, body height, and iodine status, which is not considered in the current limits. The aim was to develop reference equations, considering age, body weight, and height to calculate individual reference values for thyroid volume. Design: This cross-sectional study used data from 3 independent cohorts (SHIP-START, SHIP-TREND, and KORA-F4) in Germany. SHIP-START-0, a population-based health survey, was carried out in Northern Germany, from 1997 to 2001. SHIP-TREND-0, a second independent sample of the same study region, was carried out between 2008 and 2012. KORA-F4, a population-based health survey, was conducted between 2006 and 2008 in Southern Germany. Methods: A total of 11 549 individuals (51% women) were included in the data analysis. Eight thousand six-hundred and six individuals (45% women) were used as the thyroid-healthy reference population when developing equations. Sex-stratified quantile regression models for the 95th percentile using age, body weight, and height as explanatory variables were performed. Results: The overall reference value was 38.7 mL for men and 28.6 mL for women. According to the established cut-offs, 34% of the overall population would have had goitre compared with 7% when using our equations. Conclusion: Upper reference values for thyroid volume are too low for an adult, previously iodine-deficient population and do not consider age, body weight, and height. Using individualised equations reduces the prevalence of thyroid enlargement substantially and can lead to a decrease in overdiagnoses and the use of medical resources.

Molecular analytics increasingly utilize machine learning (ML) for predictive modeling based on data acquired through molecular profiling technologies. However, developing robust models that accurately capture physiological phenotypes is challenged by the dynamics inherent to biological systems, variability stemming from analytical procedures, and the resource-intensive nature of obtaining sufficiently representative datasets. Here, we propose and evaluate a new method: Contextual Out-of-Distribution Integration (CODI). Based on experimental observations, CODI generates synthetic data that integrate unrepresented sources of variation encountered in real-world applications into a given molecular fingerprint dataset. By augmenting a dataset with out-of-distribution variance, CODI enables an ML model to better generalize to samples beyond the seed training data, reducing the need for extensive experimental data collection. Using three independent longitudinal clinical studies and a case–control study, we demonstrate CODI’s application to several classification tasks involving vibrational spectroscopy of human blood. We showcase our approach’s ability to enable personalized fingerprinting for multiyear longitudinal molecular monitoring and enhance the robustness of trained ML models for improved disease detection. Our comparative analyses reveal that incorporating CODI into the classification workflow consistently leads to increased robustness against data variability and improved predictive accuracy.

Molecular analytics increasingly utilize machine learning (ML) for predictive modeling based on data acquired through molecular profiling technologies. However, developing robust models that accurately capture physiological phenotypes is challenged by a multitude of factors. These include the dynamics inherent to biological systems, variability stemming from analytical procedures, and the resource-intensive nature of obtaining sufficiently representative datasets. Here, we propose and evaluate a new method: Contextual Out-of-Distribution Integration (CODI). Based on experimental observations, CODI generates synthetic data that integrate unrepresented sources of variation encountered in real-world applications into a given molecular fingerprint dataset. By augmenting a dataset with out-of-distribution variance, CODI enables an ML model to better generalize to samples beyond the initial training data. Using three independent longitudinal clinical studies and a case-control study, we demonstrate CODI’s application to several classification scenarios involving vibrational spectroscopy of human blood. We showcase our approach’s ability to enable personalized fingerprinting for multi-year longitudinal molecular monitoring and enhance the robustness of trained ML models for improved disease detection. Our comparative analyses revealed that incorporating CODI into the classification workflow consistently led to significantly improved classification accuracy while minimizing the requirement of collecting extensive experimental observations. SIGNIFICANCE STATEMENT Analyzing molecular fingerprint data is challenging due to multiple sources of biological and analytical variability. This variability hinders the capacity to collect sufficiently large and representative datasets that encompass realistic data distributions. Consequently, the development of machine learning models that generalize to unseen, independently collected samples is often compromised. Here, we introduce CODI, a versatile framework that enhances traditional classifier training methodologies. CODI is a general framework that incorporates information about possible out-of-distribution variations into a given training dataset, augmenting it with simulated samples that better capture the true distribution of the data. This allows the classification to achieve improved predictive performance on samples beyond the original distribution of the training data.

Endoscopic screening for Barrett's esophagus as the major precursor lesion for esophageal adenocarcinoma is mostly offered to patients with symptoms of gastroesophageal reflux disease (GERD). However, other epidemiologic risk factors might affect the development of Barrett's esophagus and esophageal adenocarcinoma. Therefore, efforts to improve the efficiency of screening to find the Barrett's esophagus population “at risk” compared with the normal population are needed. In a cross-sectional analysis, we compared 587 patients with Barrett's esophagus from the multicenter German BarrettNET registry to 1976 healthy subjects from the population-based German KORA cohort, with and without GERD symptoms. Data on demographic and lifestyle factors, including age, gender, smoking, alcohol consumption, body mass index, physical activity, and symptoms were collected in a standardized epidemiologic survey. Increased age, male gender, smoking, heavy alcohol consumption, low physical activity, low health status, and GERD symptoms were significantly associated with Barrett's esophagus. Surprisingly, among patients stratified for GERD symptoms, these associations did not change. Demographic, lifestyle, and clinical factors as well as GERD symptoms were associated with Barrett's esophagus development in Germany, suggesting that a combination of risk factors could be useful in developing individualized screening efforts for patients with Barrett's esophagus and GERD in Germany.