The reliability of a restraint sensor system for the computer-supported detection of spinal stabilizing muscle deficiencies

Twenty healthy individuals with ages ranging from 21 to 43 years, equally distributed between sexes, were included in this study. As the functionality and safety of the CTT Centaur have not yet been proven, a young and healthy study population was recruited to eliminate confounders such as secondary diseases and minimize unforeseen hazards for the participants. The subjects were recruited from March 2018 to August 2018 using various approaches (e.g., an announcement on our homepage, social networks, public announcements). The volunteers included met all of the inclusion and exclusion criteria (Table 1). These criteria were determined on the basis of the specifications of the certified investigator, local ethics committee, and device manufacturer (BfMC GmbH, Leipzig, SN, Germany). The presence of any contraindications prevented volunteers from participating in the study. To create nearly the same test conditions and reduce the number of confounding factors, the subjects were instructed to maintain a constant lifestyle during the study, especially regarding medication consumption, nutrition and sports participation. Each subject’s health status was verified by a medical examination and anamnesis.

The study was approved by the local ethics committee (493/16-ek), and each participant gave written informed consent.

All tests were conducted in a separate laboratory at the University of Leipzig Medical Center. Within the scope of this prospective study, the test-retest reliability of the 3D spatial rotation device (BfMC GmbH) was evaluated using intraclass correlation (ICC) analysis. Therefore, three measurements (t1, t2, t3) were carried out under the same conditions. The “Centaur” (Fig. 1) is a computer-supported test and training system (CTT) that is used to train autochthonous back muscles by means of applying a precisely reproducible magnitude of strain. After the subject’s lower body is fixed at the hips and thighs, a restraint bracket is positioned over his or her shoulders. When the whole body is tilted in the sagittal plane from the neutral upright position, forces are applied to the unsupported trunk. The subject’s muscle strength was tested by a static endurance test in eight randomized positions (0°, + 45°, − 45°, + 90°, − 90°, + 135°, − 135°, + 180°; “-”: clockwise, “+”: counter clockwise). Figure 2 shows how the subject was oriented in these positions. The red arrows represent the direction of the applied gravitational force. The different positions in the transverse plane were defined by using CTT Centaur software. They were chosen so that strain was applied to all muscle areas in the upper body. The signs (+/−) indicate the Centaur’s direction of rotation. During the test, the participant remained tilted at 90 ° for 1 min at each position. In doing so, the subject permanently tried to stabilize his or her upper body along the body axis against gravitational forces. Meanwhile, a multicomponent force sensor (strain gauges), which was connected with the shoulder restraint system, quantified all movements (x, y coordinates) generated by body contact. These values included momentum values calculated on the basis of the tilt angle, trunk length, patient weight, and force exerted, which was based on the trunk’s deviation from the body axis. By using a feedback system, the movements of the upper body were continuously tracked and displayed dynamically in a coordinate system (x-coordinate: movement in the frontal plane, y-coordinate: movement in the sagittal plane; Fig. 5) to the subject. Thus, the subject could react adequately to adjust his or her posture. Each trial was followed by a total recovery period of 2 min to reduce symptoms of fatigue. During the test, the subject’s arms were crossed in front of the chest. Prior to data collection, the CTT Centaur measurement values were set to zero to calibrate the measurement system. Furthermore, all tests were performed by the same study members, with a two-day interval between the measurements to avoid habituation effects and the impact of aching muscles.

After the data collection was completed, the momentum dataset representing the subject’s trunk movement in the frontal plane and sagittal plane was analyzed. Since the momentum values were generated by sensor-trunk contact, all data were standardized (std.) to each subject’s upper body mass (62% of total body mass) [18, 21]. To fully include the sagittal and frontal movements (x, y coordinates), the 95% confidence ellipse (Fig. 3) of each test position was calculated by a custom evaluation procedure using MATLAB, version R2016b (The MathWorks Inc., Natick, MA, USA). These 95% confidence ellipses (Fig. 3) enclosed an area (mm²) in which 95% of all movements of the upper body were located. Finally, all confidence ellipses for each measurement were merged to a summarized quantity. This approach allows the whole range of motion to be considered independent of the subject’s spatial orientation, and it is more robust against outliers. Prior to statistical analysis, a data transformation (square root transformation) was performed to obtain normally distributed data and to stabilize the variance. The normality of the data was assessed visually with a histogram and quantitatively with the Shapiro-Wilk test. The ICC estimates and their 95% confidence intervals were calculated using IBM SPSS Statistics, version 23.0 (IBM Corp., Armonk, NY, USA), based on a single-rating, absolute agreement, two-way mixed-effects model (ICC(A,1) [13]; ICC (3, 1) [17, 20]. The statistical significance level was set to be α = 0.05. In this context, values less than 0.50 are indicative of poor reliability, values from 0.50 to 0.69 indicate moderate reliability, values from 0.70 to 0.89 indicate good reliability, and values greater than 0.89 indicate excellent reliability [15]. To determine the level of agreement between different measurements [6], Bland-Altman plots were constructed with the associated limits of agreement (LOA) using SPSS software. Moreover, regression lines including the 95% confidence interval (CI) were added to the Bland-Altman plots. The p-values for the effect estimates were also calculated.