Age-related trends in metabolic and cardiovascular biomarkers in southern Peru using nonparametric regression analysis with LOESS and GAM

Torres Cruz E.; Guerra Lima S.R.; Coyla Idme L.; Tumi Figueroa A.; Huata Panca P.; Quispe Cruz H.; Villasante Saravia F.H. · Nutricion Clinica y Dietetica Hospitalaria (2026) · DOI: 10.12873/461torres-2

Resumen

Introduction: Cardiometabolic diseases represent an increasing burden of comorbidity and mortality in rural populations living at high altitude, where the interaction between physiological, environmental, and sociocultural factors constitutes a key element of risk. The analysis of biomarker trends allows a better understanding of risk factor associations according to age. Objective: To model age-related associations of metabolic, cardiovascular, and anthropometric biomarkers in a rural community of Chucuito, Puno, using LOESS techniques and generalized additive models (GAM). Methodology: A cross-sectional, observational, and analytical study based on secondary records collected during 2025 from the Chucuito Health Center. A total of 345 individuals were included through probabilistic sampling. Anthropometric biomarkers (weight, height, and body mass index), metabolic biomarkers (glucose, triglycerides, and total cholesterol), and cardiovascular biomarkers (systolic and diastolic blood pressure, mmHg) were evaluated. Nonlinear associations were analyzed using LOESS and GAM, considering a statistical significance level of p < 0.05. Results: The population showed a predominance of women (83.8%) and a greater concentration in older age groups. Body mass index exhibited a significant nonlinear association with age (EDF = 2.96; p = 0.002), with an increase during middle age followed by stabilization. Total cholesterol showed the strongest association with age (EDF = 3.78; p < 0.001), with a progressive decline in older ages. Glucose and triglycerides presented significant but smaller associations, while baseline blood pressure did not show a significant association with age (p = 0.268). Although the associations were significant, age explained a limited fraction of the total variability (maximum adjusted R2 = 0.14), suggesting the influence of other biological and environmental factors. Conclusions: LOESS and GAM techniques allow the identification of nonlinear age-related association patterns in cardiometabolic biomarkers in rural high-altitude populations, providing relevant evidence for epidemiological surveillance and for designing preventive strategies adapted to Andean communities.

Instituciones: Universidad Nacional del Altiplano