Sarah Kozey Keadle1, Eli S Kravitz2, Charles E Matthews3, Marilyn Tseng1, Raymond J Carroll2,4. 1. Department of Kinesiology and Public Health, California Polytechnic State University, San Luis Obispo, CA. 2. Department of Statistics, Texas A&M University, College Station, TX. 3. Metabolic Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD. 4. School of Mathematical and Physical Sciences, University of Technology Sydney, Sydney, AUSTRALIA.
Abstract
PURPOSE: Interest in a variety of physical behaviors (e.g., exercise, sitting time, sleep) in relation to health outcomes creates a need for new statistical approaches to analyze the joint effects of these distinct but inter-related physical behaviors. We developed and tested an integrated physical behavior score (PBS). METHODS: National Institutes of Health-American Association of Retired Persons Diet and Health Study participants (N = 163,016) completed a questionnaire (2004-2006) asking about time spent in five exercise and nonexercise physical activities, two sedentary behaviors (television and nontelevision), and sleep. In half of the sample, we used shape-constrained additive regression to model the relationship between each behavior and survival. Maximum logit scores from each of the eight behavior-survival functions were summed to produce a PBS that was proportionally rescaled to range from 0 to 100. We examined predictive validity of the PBS in the other half-sample using Cox Proportional Hazards models after adjustment for covariates for all-cause and cause-specific mortality. RESULTS: In the testing sample, over an average of 6.6 yr of follow-up, 8732 deaths occurred. We found a strong graded decline in risk of all-cause mortality across quintiles of PBS (Q5 vs Q1 hazard ratio [95% CI] = 0.53 [0.49, 0.57]). Risk estimates for the PBS were higher than any of the components in isolation. Results were similar but stronger for cardiovascular disease (Q5 vs Q1 = 0.42 [0.39, 0.48]) and other mortality (Q5 vs Q1 = 0.42 [0.36, 0.48]). The relationship between PBS and mortality was observed in stratified analyses by median age, sex, body mass index, and health status. CONCLUSIONS: We developed a novel statistical method generated a composite physical behavior that is predictive of mortality outcomes. Future research is needed to test this approach in an independent sample.
PURPOSE: Interest in a variety of physical behaviors (e.g., exercise, sitting time, sleep) in relation to health outcomes creates a need for new statistical approaches to analyze the joint effects of these distinct but inter-related physical behaviors. We developed and tested an integrated physical behavior score (PBS). METHODS: National Institutes of Health-American Association of Retired Persons Diet and Health Study participants (N = 163,016) completed a questionnaire (2004-2006) asking about time spent in five exercise and nonexercise physical activities, two sedentary behaviors (television and nontelevision), and sleep. In half of the sample, we used shape-constrained additive regression to model the relationship between each behavior and survival. Maximum logit scores from each of the eight behavior-survival functions were summed to produce a PBS that was proportionally rescaled to range from 0 to 100. We examined predictive validity of the PBS in the other half-sample using Cox Proportional Hazards models after adjustment for covariates for all-cause and cause-specific mortality. RESULTS: In the testing sample, over an average of 6.6 yr of follow-up, 8732 deaths occurred. We found a strong graded decline in risk of all-cause mortality across quintiles of PBS (Q5 vs Q1 hazard ratio [95% CI] = 0.53 [0.49, 0.57]). Risk estimates for the PBS were higher than any of the components in isolation. Results were similar but stronger for cardiovascular disease (Q5 vs Q1 = 0.42 [0.39, 0.48]) and other mortality (Q5 vs Q1 = 0.42 [0.36, 0.48]). The relationship between PBS and mortality was observed in stratified analyses by median age, sex, body mass index, and health status. CONCLUSIONS: We developed a novel statistical method generated a composite physical behavior that is predictive of mortality outcomes. Future research is needed to test this approach in an independent sample.
Authors: Deborah Rohm Young; Marie-France Hivert; Sofiya Alhassan; Sarah M Camhi; Jane F Ferguson; Peter T Katzmarzyk; Cora E Lewis; Neville Owen; Cynthia K Perry; Juned Siddique; Celina M Yong Journal: Circulation Date: 2016-08-15 Impact factor: 29.690
Authors: A Schatzkin; A F Subar; F E Thompson; L C Harlan; J Tangrea; A R Hollenbeck; P E Hurwitz; L Coyle; N Schussler; D S Michaud; L S Freedman; C C Brown; D Midthune; V Kipnis Journal: Am J Epidemiol Date: 2001-12-15 Impact factor: 4.897
Authors: S Chasan-Taber; E B Rimm; M J Stampfer; D Spiegelman; G A Colditz; E Giovannucci; A Ascherio; W C Willett Journal: Epidemiology Date: 1996-01 Impact factor: 4.822
Authors: A M Wolf; D J Hunter; G A Colditz; J E Manson; M J Stampfer; K A Corsano; B Rosner; A Kriska; W C Willett Journal: Int J Epidemiol Date: 1994-10 Impact factor: 7.196
Authors: Kara M Whitaker; Matthew P Buman; Andrew O Odegaard; Katie C Carpenter; David R Jacobs; Stephen Sidney; Mark A Pereira Journal: Am J Epidemiol Date: 2018-02-01 Impact factor: 5.363
Authors: Shirley M Bluethmann; Sarah K Keadle; Tonya S King; Charles E Matthews; Frank M Perna Journal: J Cancer Surviv Date: 2021-06-10 Impact factor: 4.062
Authors: Mitch J Duncan; Anna T Rayward; Elizabeth G Holliday; Wendy J Brown; Corneel Vandelanotte; Beatrice Murawski; Ronald C Plotnikoff Journal: Int J Behav Nutr Phys Act Date: 2021-03-25 Impact factor: 6.457