Behavioral biomarkers of aging: illustration of a multivariate approach for detecting age-related behavioral changes.

The goal of the current project is to develop a multivariate statistical strategy for the formation of behavioral indices of performance and, further, to apply this strategy to establish the relationship between age and important characteristics of performance. The strategy was to begin with a large set of measures that span a broad range of behaviors. The behavioral effects of the following variables were examined: Age (4, 12, 24, and 30 months), genotype [Fischer 344 and a hybrid (F1) of Fischer 344 and Brown Norway (F344xBN)], gender (Fischer 344 males and Fischer 344 females), long-term diet (ad lib diet or dietary restriction beginning at 4 months of age), and short-term diet (ad lib diet or dietary restriction during testing). The behavioral measures were grouped into conceptually related indicators. The indicators within a set were submitted to a principal component analysis to help identify the summary indices of performance, which were formed with the assumption that these component scores would offer more reliable and valid measures of relevant aspects of behavioral performance than would individual measures taken alone. In summary, this approach has made a number of important contributions. It has provided sensitive and selective measures of performance that indicated contributions of all variables: psychological process, age, genotype, gender, long-term and short-term diet and has increased the sensitivity of behavioral measures to age-related behavioral impairment. It has also improved task-manageability by decreasing the number of meaningful variables without losing important information, consequently providing a simplification of the pattern of changes.