This study evaluated the validity of the total energy expenditure (TEE) estimated using uniaxial (ACCuni) and triaxial (ACCtri) accelerometers in the elderly. Thirty-two healthy elderly (64-87 years) participated in this study. TEE was measured using the doubly labeled water (DLW) method (TEE(DLW)). TEE(ACCuni) (6.79 +/- 1.08 MJ day(-1)) was significantly lower than TEE(DLW) (7.85 +/- 1.54 MJ day(-1)) and showed wider limits of agreement (-3.15 to 1.12 MJ day(-1)) with a smaller correlation coefficient (r = 0.703). TEE(ACCtri) (7.88 +/- 1.27 MJ day(-1)) did not differ from TEE(DLW) and showed narrower limits of agreement (-1.64 to 1.72 MJ day(-1)) with a larger correlation coefficient (r = 0.835, P < 0.001). The estimated intensities of light activities were significantly lower with ACCuni. Greater mediolateral acceleration was observed during 6-min walk tests. The results suggest that ACCtri is a better choice than ACCuni for assessing TEE in the elderly.