Modeling the frailty phenotype by using biochemical markers measured in blood samples and multiple regression analysis

Abstract

Background Human frailty syndrome is a complex phenotypic trait and the result of multiple sources of influence. Its assessment is not a simple task; however, it is important to preventing disability in adult people. However, it requires retrieval of clinical data and calculations that are time-consuming, which may be unfeasible in routine clinical practice. We propose a diagnostic tool based on multiple regression analysis and aimed to have a preliminary assessment of the frailty condition in senior adults. Because it depends on blood sample analyzes, our predictive functions are less invasive, expensive, time consuming, and less physically and psychologically discomforting. Method Data came from the Toledo Study for Healthy Aging, a prospective cohort study designed for evaluating the determinants, and consequences, of the frailty syndrome in older people living in the province of Toledo, Spain. The Garcia´s Frailty Trait Scale (FTS12) was considered as the dependent variable in the statistical analyzes. Forty-six traits assessed from blood samples were included as independent variables to be tested. They included 15 variables from hemograms, 19 variables from biochemical profiles, and 12 extra biochemical characteristics (not included in standard biochemical profiles). The stepwise regression method was used in deciding which explanatory variables were to be included in the predictive models. Results Two predictive functions of the observed frailty for each gender were obtained: a “full model” and a “reduced model”. A range of 0.615 to 0.649 of moderately high Pearson's correlations was recorded between observed and predicted FTS12 values across the four equations. The correlation between all functions of risk factors and FTS12 was larger than 0.4 and highly significant in all four study cases. With single risk factors, the highest correlation coefficient was 0.283 (p-value < 0.0001). Conclusions The frailty syndrome can be moderately well predicted by using biochemical information. However, full models showed a narrow superior predictive capability. Albumin was present in all functions of risk factors selected by the modeling. It showed to be a risk factor that helps in predicting the frailty phenotype. We expect that exploring Albumin’s potential as part of a predictive equation might help to design anti-aging therapeutic measures.