A need for using raw acceleration data instead of activity counts for measuring the intensity of physical activity has been expressed widely [18–20]. This study compared the MAD-based cut-points of two different accelerometer brands, a widely used Actigraph GTX3 and a new Hookie device, in classifying the intensity of basic free-living physical activities in adolescents. To our knowledge this is the first study to utilize a raw acceleration data based trait (i.e., MAD) in comparing two accelerometer brands among adolescents.

According to the results both accelerometers provided virtually equal MAD values up to about 700 mg and showed almost perfect agreement in classifying activities into light, moderate and vigorous categories with marginal misclassification rate. Our findings indicate the utility of MAD cut-points in the intensity-based classification of adolescents’ raw acceleration data. Noteworthy, the cut-points observed in the present data of adolescents were similar to those obtained in adults [15], and indeed, the cut-points performed equally well in adults and adolescents. In sum, it may be possible to use same cut-points for classifying locomotion by intensity both in adolescents and adults.

The study raised some concerns, which should receive attention in the future. First, both accelerometers sampled the acceleration signal at sufficient 30 and 100Hz rates for detecting general movements comprising frequencies typically less than 10Hz [14]. However, the wider dynamic range of Hookie (±16,000 mg) outperformed Actigraph (±3000 mg) by covering also the accelerations during vigorous activities such as jogging and running. Also, the higher 100Hz sampling frequency of Hookie accounted at least to some extent for the higher MAD values and thus the cut-points in moderate and vigorous intensity physical activity compared with Actigraph. By using the new Actigraph GT3X+, which can employ 100 Hz sampling frequency and has the dynamic range of ±8000 mg, these differences may disappear and the agreement of the cut-points might have been almost perfect.

Second, the sample size of the present study was too small preventing us from reliable comparison of MAD cut-points between girls and boys although Table 2 indicates that boys may have slightly higher cut-point from sedentary to light activity than girls. The limited sample size hindered us also to stratify the subjects according to common anthropometric and biomechanical variables, such as body mass or height, to evaluate the consistency of the results in various subgroups. For example, the recent treadmill study by Horner et al. [21] involving young, active females indicate that soft tissue (subcutaneous adiposity) under the sensor may explain individual differences and variability in hip-derived accelerometer output in walking. Thus far, the applicability of the findings in various subgroups is left open. However, given the fact that the measurement is based on a direct physical trait i.e., movement induced acceleration and that the universal MAD cut-points, similar to those found in adolescents in the present study, have been proven suitable for classifying the intensity of physical activity in adults [15], it is highly likely that the present findings will apply to other groups of adolescents as well.

Third, it is acknowledged that the array of 10 different activities was quite narrow covering only a fraction of the most usual weight-bearing activities and leaving out a wide spectrum of activities, where the activity-induced accelerations may not be accurately determined such as swimming, cycling, skiing, climbing, gym exercising etc. Moreover, the activities were performed under supervision and in exceptional surroundings, which was likely to lead to some deviation from habitual performance. The correspondence to real life may also have been impaired by the short duration of single activity, which was only 2 min instead of 4 to 5 min suggested by Welk et al. [17]. However, if the activity was performed similarly over the time period of interest the acceleration values would have remained unchanged irrespective of the duration of the activity. Using longer duration would have prolonged the length of the whole protocol up to an hour possibly slowing down the recruitment process and hampering the concentration of teenagers, who may be more impatient than adults in performing rigorous tasks.

Fourth, not measuring the distance walked or run prevented us from standardizing the paces and defining the intensity of ambulatory tasks more objectively. We were, therefore, unable to verify how well our pattern-based intensity classification corresponded to pace-specific intensity levels. Heart rate, which was used in this study to assess the validity of pattern-based classification of intensity, is known to be susceptible for psychological reactions [9]. As more elevated heart rates were observed in sedentary behaviour than in light activity, activities classified to sedentary behaviour had to be excluded from testing the correlation between acceleration levels and heart rates. In terms of light, moderate and vigorous activities the within-individual correlations between the MAD and the heart rate were very high for both accelerometers suggesting a valid physiological association between the MAD value and intensity as judged from the incident heart rate.

Finally, it can be argued that the placement of the monitors to the different sides of the hip may have influenced on MAD values and data interpretation. However, in a previous study by McClain et al. [22], which was conducted in free-living conditions, the intraclass correlations for various accelerometer outputs (activity counts, steps, intensity-specific minutes) between the right and left hip placement were high ranging from 0.97 to 0.99. Thus, it is unlikely that the accelerometer placement had notable impact on MAD values and their interpretation in the present study.