Age Differences, Age Changes, and Generalizability in Marathon Running by Master Athletes

This study examines the world’s Top 100 age class performance times by Master athletes in marathon running. The predominant paradigm for this type of research assumes that the outcomes represent a “virtual” cross-sectional study with important implications about aging. This article critiques this perspective and presents alternative models that include temporal dimensions that relate to cohort differences, age changes and historical transitions. One purpose of this study is to compare these models with respect to goodness of fit to the data. A second purpose is to evaluate the generalizability of findings from the fastest divisional age class quartile to the slower quartiles. Archival listings by the Association of Road Racing Statisticians include a maximum of 100 fastest age class performances in marathon running performances by men and women. This database includes 937 performances by 387 men performances and 856 performances by 301 women. The mean ages are 62.05 years for men and 60.5 years for women. The mean numbers of performances per runner are 6.64 for men and 6.4 for women. Analysis by mixed linear modeling (MLM) indicates best goodness of fit for logarithms of performance time by a model that includes linear and quadratic expressions of age at entry into the database (termed “entry cohort”) and subsequent age changes (termed “elapsed age”) as variables. Findings with this model show higher performance times in women than men. Rates of increase in performance time are higher at older cohort ages and elapsed ages. Performance time increases with interactions between cohort age and elapsed age, cohort age and gender, and elapsed age and gender (i.e., with greater increases in women than men). Finally, increases in performance time with cohort age and elapsed age are higher in slower than faster performance quartiles, with athletes in the faster quartiles more likely to have multiple data entries and athletes in the slower quartiles single data entries. Implications of these findings are discussed.