Abstract P122: The Predictive Ability and Validity of a Composite Physical Activity Score in Relation to Adiposity Measures: Coronary Artery Risk Development in Young Adults (CARDIA)
Background: The predictive ability and validity of combining self-report and accelerometer data into composite physical activity scores has not yet been evaluated.
Methods: CARDIA is a multi-site, prospective cohort study of black and white men and women (n=5,115), initiated in 1985-86. Participants were randomly assigned to the derivation (2/3; n=3,411) or validation (1/3; n=1,703) dataset. At Year 20, physical activity was assessed via self-report and accelerometry (ActiGraph 7164). Principal components analysis (PCA) was used to create a composite score reflecting Year 20 physical activity. In the derivation dataset, generalized linear regression models were constructed to explore the variance explained by the composite score with adiposity measures (R2 value), collected at Years 20 and 25, compared to self-report and accelerometer estimates alone. Adiposity measures included: body mass index (BMI) and waist girth. At Year 20, fat and fat-free mass via dual-energy x-ray absorptiometry were also assessed. This process was repeated in the validation dataset and the difference between the R2 values was computed as the cross-validation value.
Results: At Year 20, 3,549 participants attended the clinic exam; a similar number attend at Year 25 (n=3,499). At Year 20, participants were aged 45.2±3.6 years and the majority were female (54.5%) and black (51.6%). Year 20 and 25 BMI was 29.5±7.2 and 30.2±7.2 kg/m2, respectively. The first four principal components, representing the linear combination of reported TV viewing, moderate and vigorous intensity activity, and accelerometer-detected sedentary to vigorous intensity activity (Matthews cutpoints), explained 72.3% of the variance. In the derivation dataset, after modeling the composite score regressed on BMI, the R2 values for Year 20 and 25 BMI were 5.2% and 6.3%, respectively (both p<0.001), which were higher compared to accelerometer (2.1% and 2.2%, respectively) or self-report (2% and 2.8%, respectively) alone (all p<0.001). The R2 values for Year 20 and 25 waist girth were higher with the composite score compared to self-report and accelerometer, respectively (Year 20: 3.8%,1%, and 1.3%, respectively; Year 25: 4.1%, 3.3%, and 1.3, respectively; all p<0.001). Finally, the R2 values for Year 20 fat and fat-free mass were higher with the composite score versus self-report and accelerometer, alone (Fat Mass: 10.5%, 6.2%, and 5.1%; Fat Free Mass: 5.8%, 3.4%, and 1.3%; all p<0.001). The cross-validation values for all Year 20 and 25 adiposity measures were <10%, suggesting the composite score is externally valid.
Conclusions: Findings support the continued use of self-report and device-based methods to assess physical activity in population-based studies. However, investigators should consider statistical approaches, like PCA, to create a more robust estimate that reflects the complexities of physical activity behavior.
Author Disclosures: K. Pettee Gabriel: None. A. Perez: None. J. Lee: None. H.W. Kohl: None. D.R. Jacobs: None. B. Sternfeld: None.
This research has received full or partial funding support from the American Heart Association, South Central Affiliate (Arkansas, New Mexico, Oklahoma & Texas).
- © 2016 by American Heart Association, Inc.