Predicting Free-living Activity Energy Expenditure With Hip And Wrist Accelerometry Versus Doubly Labeled Water

Abstract

While accelerometry-based activity monitors (AMs) have been validated to predict activity energy expenditure (AEE) in controlled lab settings, their use for predicting free-living AEE is not well established. PURPOSE: The purpose of this study was to compare the ability of hip- and wrist-worn AMs to predict free-living AEE in overweight and obese adults as compared with AEE from the doubly labeled water (DLW) technique. METHODS: 28 subjects (20 women, 8 men; Mean±SD: 44±14 yrs age, 31.6±3.7 kg/m2 BMI) wore the hip AM while a subset of 23 subjects (18 women, 5 men; 45±14 yrs age, 30.2±7.4 kg/m2 BMI) also wore the wrist AM for seven complete days. AM output from both hip and wrist monitors were converted into average AEE over the 7-day period (AEEHP and AEEWR, respectively; kcals/day) using established 1R- and 2R-algorithms. Wrist AM data were also transformed with a smoothing algorithm prior to the AEE transformation (smoothed AEEWR). Direct measures of AEE (AEEDLW, kcals/day) were derived from total daily energy expenditure (TDEE) using the DLW technique after adjusting for measured resting metabolic rate (RMR) and the thermic effect of food (TEF), where AEEDLW = TDEE - RMR - 0.1(TEF). 3-factor repeated measures ANOVA and correlations were used to compare 1R and 2R estimates of AEEHP with AEEDLW, while 1R and 2R estimates of both AEEWR and smoothed AEEWR were compared with AEEDLW (a=0.05). RESULTS: 1R and 2R estimates of AEEHP (Mean±SE: 749±98 and 876±118, respectively) were similar to AEEDLW (782±88) but correlated poorly (r=0.31-0.32; P=0.11). In contrast, using the subset of subjects, 1R and 2R estimates of AEEWR (3243±176 and 3119±174, respectively) differed significantly from AEEDLW (881±95; P<0.001) but correlated moderately (r=0.49-0.53; P<0.02). Lastly, 1R and 2R estimates of smoothed AEEWR (1076±58 and 1040±58, respectively) were statistically similar to AEEDLW and correlated moderately (r=0.49-0.57; P<0.02). CONCLUSIONS: Mean values of AEE from DLW were predicted accurately with AM-derived measures of AEEHP and smoothed AEEWR, but only smoothed AEEWR also correlated moderately with AEEDLW. These results indicate that both hip- and wrist-worn AMs may be capable to predicting free-living AEE for groups of overweight and obese adults, but only the wrist-worn AM rank ordered subjects reasonably well.