Subjects and protocol
In this experimental study, 47 subjects with different frailty levels were asked to perform the 30-s CST. Specifically, 13 frail subjects (4 males and 9 females, aged 85 ± 5 years, body mass 67.5 ± 8.6 kg, and height 1.54 ± 0.05 m), 16 pre-frail (8 males and 8 females, aged 78 ± 3 years, body mass 71.6 ± 10.5 kg, and height 1.61 ± 0.08 m) and 18 healthy subjects (14 males and 4 females, aged 54 ± 6 years, body mass 75.2 ± 3.4 kg, and height 1.76 ± 0.04 m) volunteered to participate in this study. The frail and pre-frail subjects were selected from the population used for the baseline data of the Toledo Study for Healthy Aging (TSHA) [37]. According to the criteria defined by Fried et al., [2], frailty was determined as the presence of three or more of the following criteria: slowness, weakness, weight loss, exhaustion, and low physical activity. Subjects were classified as pre-frail if one or two criteria were present and as non-frail if no criteria were present. All of the subjects were thoroughly informed about the experimental procedure; the purpose, nature, and possible risks associated with the study; and their right to terminate participation at their discretion. Subsequently, the subjects provided their written informed consent to participate. These experimental procedures were approved by the Institutional Review Committee of the Public University of Navarra and the Department of Health Sciences of the Government of Navarra, according to the Declaration of Helsinki.
The 30-s CST consists of standing up and sitting down from a chair as many times as possible within 30 seconds. A standard chair (with a seat height of 40 cm) without a backrest but with armrests was used. Initially, subjects were seated on the chair with their back in an upright position. They were instructed to look straight forward and to rise after the “1, 2, 3, go” command at their own preferred speed with their arms folded across their chest. All trials were performed using the same chair and with similar ambient conditions. The medical staff who supervised the performance of the test did not participate in analyzing the kinematic data, and they did not have any knowledge about the analysis whatsoever.
As described in the background section, the present study contains two parts. In the first part, all of the subjects from the three frailty groups were evaluated. Everyone was able to finish the test properly. In the second part, a subset of the data from the initial test was considered. A group of seven pre-frail and eight healthy subjects, with a mean number of 17 sit-stand-sit cycles per group and a range of 15–20 were evaluated.
Instrumentation
An inertial MTx Orientation Tracker (WSENS, Xsens Technologies B.V., Enschede, Netherlands) was attached over the L3 region of the subject’s lumbar spine to provide the kinematic data for each trial. It recorded at a sampling rate of 100 Hz. The L3 position was chosen because of its proximity to the body’s center of mass (CoM) in the standing position. The nine individual MEMS sensors from the MTx provided kinematic data such as the 3D acceleration and the 3D rate of turn (rate gyro). Moreover, the drift-free 3D orientation was also provided by the MTx using Kalman filters and the previously mentioned kinematic data.
Before starting the test, when the subject was sitting on the chair in an upright position, the sensor-fixed reference frame was aligned with the global reference system (X, Y, and Z). This global reference system was defined as the Earth-fixed global reference frame (XYZ), whose Z-axis points vertically upwards, with the X-axis in the lateral direction and the Y-axis in the anterior-posterior direction. The orientation data, consisting of the Euler angles in either XYZ or roll-pitch-yaw order, defined the rotation aligning the global axis to the sensor-fixed reference frame at each time point. The IU provides both the linear acceleration and the rate of turn in its sensor-fixed Cartesian reference frame (xyz). The linear acceleration in the global reference frame can be translated into the global reference frame using the orientation data.
Data analysis
An automated data analysis procedure was implemented using Matlab 7.11 (MathWorks Inc., Natick, MA, USA) to improve the objectivity and simplicity of the current 30-s CST evaluation. The automated analysis provides an accurate count of the number of repetitions, removing failed attempts as determined by the roll rotation angle (X-orientation) in combination with the Z-acceleration signal, and the derived Z-velocity and Z-position, and the kinematic parameters. The procedure was implemented as a three-stage algorithm:
First, the raw signals were processed to obtain the Z-velocity and Z-position. Specifically, single and double integration of the Z-acceleration was performed. Furthermore, a two-step processing method (a fourth-level polynomial curve adjustment followed by baseline interpolation from local maxima and minima) was chosen to correct the inherent drift effect. A fourth-level polynomial fitting was chosen to accommodate for slow changes in the acceleration bias without incurring in over-fitting. Then, remaining baseline fluctuations were estimated by spline interpolation of local maxima and minima.
Second, the corresponding sit-to-stand-to-sit cycles and their main phases (impulse, stand-up and sit-down) [33] were determined using the X-orientation as well as the Z-acceleration, Z-velocity and Z-position. The Z-position signal was used as an indicator of changes in the vertical position of the MTx unit, making it possible to automatically obtain the number of completed cycles (the current standard measurement from the 30-s CST). The X-orientation informs about the body’s sway movements (i.e., forward and backward trunk leans), while the Z-acceleration gives information about the up and down body forces exerted to complete the cycles. The combination of these two signals with the Z-position provides enough markers to clearly detect stand-up and sit-down transitions, as well as failed attempts. A failed cycle was defined as an attempt performed by a subject who did not reach the upright position. In the algorithm, these situations were automatically detected based on a threshold applied to both the time elapsed between a maximum and a minimum of the Z-position and to their difference. The Z-velocity signal was used to establish whether a transition was SitTS or StandTS, (Figure 1), [33].
Finally, to quantify the potential differences between subjects, specific movement-related parameters were derived based on the raw MTx acceleration and orientation signals, as well as additional values obtained after data analysis (i.e., the duration, velocity, and position).
An analysis was performed on the X-orientation and Z-acceleration signals. The X-orientation was selected because it contains information about the way the subjects manage their body (the trunk’s forward and backward tilt), while the Z-acceleration was related to the impulse required to reach the upright position. To evaluate each parameter, the data were first divided into cycles; then, each cycle was separated into its corresponding phases. Therefore, kinematic parameters could be defined in each phase of the performed cycles for any subject. The overall value of a parameter for a subject was obtained by computing its mean value across the subject’s cycles. These parameters describe the subject’s movement performance in terms of the mean, standard deviation, maximum, minimum and range of several features of the subtasks, including the duration of the phases and the orientation, position and acceleration signals. The parameters described below (a, b, and c) were also obtained from our analysis:
a) X-orientation range
Four parameters were defined to characterize the amount of forward and backward trunk tilt occurring during each cycle, (Figure 2, blue line). These parameters were evaluated not according to the phases of the cycles but instead to the trunk movements of the subjects performing SitTS and StandTS transitions. Considering that the cycle starts with the impulse phase, we assumed that the subject was initially in the upright position to define the following ranges of movement:
-
TurnB_Sit is a backward trunk lean while the subject is sitting down that is generally produced to accommodate the weight into the chair (Figure 2, blue line, “1”).
-
TurnF_Sit is a forward trunk lean in the seated position to start the next standing-up (Figure 2, blue line, “2”).
-
TurnB,Up is a backward trunk lean that occurs while the subject is standing-up until he reaches the upright position (Figure 2, blue line, “3”).
-
TurnF_Up is a forward trunk lean in the standing position while the subject is descending that is normally generated to improve balance control (Figure 2, blue line, “4”).
b) Standing-up and sitting-down “modified impulses”
The AUCZacc parameter was defined as the area under the curve of the acceleration for the duration of the movement (1). It was related to the necessary impulse to stand upright and to return to the seat. As previously defined in [33], the AUC was divided into positive and negative components, according to the direction of the displacement. AUC+Zacc referred to the active “modified impulse” used to perform the transition upward, whereas AUC-Zacc referred to the passive transition back to the chair (Figure 2).
(1)
c) Maximum peaks of standing-up and sitting-down velocities
Drift-effect cancellation, which has been previously described, was required to obtain the 3-axis velocity from the corresponding acceleration signals. For simplicity, only the Z-velocity and the Y-velocity were evaluated because these parameters have greater relevance for the transitions. The Z-velocity refers to the vertical movements of each cycle of the 30-s CST, while the Y-velocity is the forward and backward speed when standing-up and sitting-down.
Finally, standard statistical methods were used to calculate the mean and standard deviation (SD) of each phase parameter across both cycles and subjects.
Statistical analysis
The differences among the three groups (frail, pre-frail and healthy) were determined using a one-way analysis of variance (ANOVA) with Newman-Keuls post-hoc comparisons. When the normality test failed (p < 0.05), the Mann–Whitney rank sum test was employed. A p < 0.01 criterion was used to establish statistical significance. Box plots of each parameter for the different movement phases were used to graphically display the variable’s location. The box itself contained the middle 50% of the data. The upper and lower edges of the box indicated the 75th and 25th percentiles, respectively, and the central line was the median value of the data. The ends of the vertical lines, or “whiskers”, were the minimum and maximum data values, and any points outside the whisker ends represented outliers.