Cluster discriminant analysis of relationship between physical examination parameters and blood biochemical indexes in elderly people

Abstract

Objective To investigate the relationship between noninvasive anthropometric parameters and laboratory test results from community healthy elderly people by cluster analysis and generate the discriminant equations. Methods Eleven communities were randomly selected from Pudong New District, Shanghai, China. Individuals aged 65 years and above from the communities were recruited. Body mass, blood pressure, blood cholesterol, and glucose were collected. However, individuals with following conditions were excluded: confirmed high blood pressure, diabetes mellitus, coronary heart disease, and hepatic and renal dysfunction. Four variables were selected to perform cluster analysis and establish discriminant function, which included age, systolic blood pressure (SBP), diastolic blood pressure (DBP), and body mass index (BMI). Comparison of anthropometric and blood biochemical indexes of all kinds of elderly people by analysis of variance. Results Three subgroups were distinguished. Individuals in cluster 1 were youngest with the highest noninvasive parameters including blood pressure and BMI and highest laboratory test results consisting of blood cholesterol and glucose and hepatic function parameters. The results of multiple comparisons using the Fisher′s least significant difference (LSD) procedure showed that there was no statistical difference in SBP or fasting glucose between cluster 1 and cluster 2, and that there was no statistical difference in bilirubin among the three clusters. However, differences in the other parameters and results were statistically significant (P<0.0125). The error rates of discriminant functions established by both retrospective method and cross validation method were less than 0.2. Conclusion Non-invasive anthropometric indexes can be used to estimate the biochemical indexes such as blood glucose, blood lipid and liver function in healthy elderly people in the community. Key words: Cluster analysis; Physical examination; Anthropometric parameters; Blood biochemical indexes