Age-specific differences in the time-frequency representation of surface electromyographic data recorded during a submaximal cyclic back extension exercise: a promising biomarker to detect early signs of sarcopenia

The study protocol was approved by the Ethics Committee of the City of Vienna - Number: “EK 11–064-VK-NZ” Full address: Ethikkommission der Stadt Wien Magistratsabteilung 15, Gesundheitsdienst der Stadt Wien 3., Thomas-Klestil-Platz 8 Town Town Tel.: 01/4000–87,523. All participants provided written informed consent. The data collection was carried out in accordance with the directives of the Declaration of Helsinki. Participants received financial compensation for their participation in the study.

Study volunteers were recruited by word of mouth, by presenting the study to groups of older adults attending institutions with a focus on leisure and social activities in and around the city of Vienna, and by posting flyers in appropriate areas at companies near the Karl-Landsteiner Institute of Outpatient Rehabilitation Research where the data collection took place. A total of 86 asymptomatic volunteers (40 females, between 18 and 90 years of age) were enrolled in the study.

Subjects were eligible to participate in the study if: they were generally healthy, i.e., were free of co-morbidities (mild diabetes, well controlled hypertension, mild osteoarthritis of lower limb weight bearing joints were included); had no major general health problems that would interfere with testing; were free from any functional limitations (reported independent walking distance exceeded 800 m, timed up-and-go test less than 10 s, Tandem-stand exceeded 10 s, chair rise test less than 15 s); were independent in their activities of daily life; performed a normal level of physical activity (but did not participate in competitive sports more than two times per week); and, if they were free from any risk factors that could preclude them from participating in vigorous exercise. Physical Medicine & Rehabilitation physicians screened all participants.

Volunteers were not considered eligible to participate in the study if: 1) they were unable to follow instructions given in German; 2) experienced more than five mild back or referring back pain episodes (Visual Analog Scale] > 30) lasting more than two days each within the past year; 3) had a history of spine surgery or any kind of specific diseases of the spine (tumor, inflammation, osteoporosis, fracture); 4) were pregnant; 5) complained about any medical condition that might interfere with maximum strength or submaximal endurance testing; or, 6) had a body mass index (BMI) exceeding 35 kg/m2.

MATLAB routines (The MathWorks, Inc., Natick, MA, USA) were used for EMG data processing. All data was preprocessed by skipping the first three seconds of the static task and the first eight seconds (first two cycles) of the dynamic task, and by removing artifacts. EMG data was filtered using a 20 Hz high-pass and 500 Hz low-pass Butterworth filter design. RMS-EMG calculations were performed from the second to the fifth second of the sustained 80% MVC contraction, as described previously [24]. These values were used for normalization of the RMS-EMG values recorded from the dynamic contractions.

EMG signals of the dynamic fatigue task were sequenced according to their concentric and eccentric phases using acceleration data from one electrode positioned on the lever arm of the Total Trunk device. These on-average 1 s EMG intervals corresponded in equal portions to trunk ROM segments during flexion and extension to avoid acceleration, deceleration, and turning phases during the exercise. Thus, movement velocity and muscle length - torque geometry were comparable in the EMG representations not only between the shorting and lengthening of the cyclic exercise but also between the individuals tested (see Fig. 1).

Root mean square (RMS) and instantaneous median frequency (IMDF) SEMG variables were computed for all sequenced SEMG intervals recorded during the dynamic shortening and lengthening contractions. SEMG sections, consisting of approximately 2000 samples, were split up into windows of 500 samples. These windows were overlapping by 50%, whereas IMDF and RMS calculations of the first and last 125 samples of the approximately 1 s intervals were rejected to ensure the extraction of the central portion of the angular range of motion. IMDF calculations are based on Cohen Class Time-Frequency representations [42] for computing instantaneous frequency distributions. Our approach is based on findings from comparisons of five different time-frequency transformations, namely, the Wigner-Ville, the smoothed Wigner-Ville, the Cone Kernel, the reduced interference, and the Choi-Williams. These techniques were applied to computer-synthesized realizations of stochastic processes, as well as to real signals detected during different types of contractions in healthy human volunteers [30]. Cohen Posch representations were obtained by first computing a Cohen Class [30] time-frequency representation of the EMG data and then applying an iterative algorithm to remove undesirable oscillating terms that mark Cohen Class representations [43]. Cohen Class transformations may be seen as an extension of the correlogram technique [30]. The autocorrelation function is replaced by the instantaneous autocorrelation function [42]. Its Fourier transformation is time-dependent and constitutes a representation of the instantaneous frequency content of the EMG signal [30]. As Cohen Class representations provide a relatively unstable estimate of the EMG frequency content, Cohen-Posch representations [44] were derived from them via iterative adjustments of the time and frequency marginals [43]. The IMDF-EMG value was calculated for one portion of each phase of the exercise cycle, across all the cycles of the exercise [33].

The electrical manifestations of muscle fatigue were assessed for each series of contractions by computing the least squares regression line through the 25 IMDF-estimates relative to the 25 signal bursts of each contraction series. The corresponding IMDF SEMG values of one exercise session were averaged to obtain one IMDF value per cycle and phase. Figure 1 shows the procedure described above applied to the signal collected from the multifidus muscle of a subject during a series of back flexion-extension cyclical movements (legend to Fig. 1). The RMS-SEMG values were normalized by the RMS-SEMG obtained during 4 s (i.e. seconds 3 to 7) of the 80% static, sustained contractions.

Linear regression was applied to the time series of SEMG values to calculate the rate of change in RMS-SEMG and IMDF-SEMG over time; these measures are indicative of muscle fatigue. Following previous recommendations [45], both the RMS and the IMDF slopes were divided by their corresponding intercept values to achieve normalized SEMG indices of muscle fatigue, which accounted for differences in subcutaneous tissue thickness amongst test participants. In order to improve reliability, the scores of each of the three electrode pairs at the three lumbar levels (L1, L2, and L5) were averaged to increase their reliability [46].

The following dependent variables were used in the analysis: MVC as a measure of trunk extensor strength, mean RMS-SEMG and IMDF-SEMG values derived from the initial values of linear regression models, mean RMS-SEMG and IMDF-SEMG slopes normalized by the RMS-SEMG, and IMDF-SEMG initial values. The independent variables were subjects’ age (older or younger than 50 years of age) and sex. All statistical analyses were done in the R environment for statistical computing [48].

For the analysis of age and sex specific differences, biomechanically comparable (velocity and acceleration) dynamic tasks and artefact-free SEMG data from the full set of electrodes was available from 27 older and 29 younger participants (Fig. 2). This was done to control for any confounders that could occur from group imbalances in the biomechanics of task performance or due to missing EMG variables. Our primary analysis was performed with fitted data of missing SEMG variables of individual electrodes. Fitting of data was performed if the SEMG recordings of one of the two (paired) electrodes per segment were undisturbed. Consequently, the undisturbed SEMG was duplicated and onset and slopes values were corrected by the left-right imbalance ratios of the respective lumbar segment obtained from the 56 individuals with the full data set. Consequently, data from a total of 35 older and 41 younger individuals were available for analysis. A secondary analysis was performed with the full set of electrodes available. Results of the secondary analysis are not shown but can be provided upon request.

Descriptive statistics served to summarize the participants’ characteristics. To evaluate whether maximum back extension moment or SEMG fatigue variables differed between the two age-specific and sex-specific groups, mixed effects model with “age”, “sex” and “test number” as fixed factors and the random factor “persons” were computed for each outcome variable. P-values were Bonferroni corrected (five electrode levels) and considered significant if p ≤ 0.01. The sample size estimate for the mixed ANOVAs was calculated with Monte Carlo power simulation [49] and was based on findings from two previous studies [24, 50]. Considering a total of five comparisons at an alpha of 0.01 and a power of 0.9 (1-beta), a sample size of 80 participants was necessary to detect medium age- or sex-specific differences between groups.

Generalizability Theory (G-Theory) [51‐53] was used to examine reliability of the SEMG measures. As compared to classical test theories, the G-theory takes into account how the consistency of outcomes may change if a measure is used to make absolute (criterion referenced) versus relative (norm referenced) decisions. Using a “multi-factorial random-effects Analysis of Variance (ANOVA)” model that included several sources of measurement error related to: subject, day, side, subject x day, and subject x side, the absolute standard error of measurement (SEM) values as well as the respective coefficients of dependability (D), the latter is a type of Intra Class Correlation Coefficient (ICC), were calculated. Note that absolute decisions are criterion-referenced and occur if the subject’s measurement results are independent of the performance of other subjects.

The normalized RMS-SEMG onsets derived both from the concentric and the eccentric phases of the cyclic exercise did not significantly differ between younger and older individuals, nor between males and females. It is worthwhile to note that the initial RMS-SEMG values would be approximately twofold in size if the concentric portions of the exercise were compared with the eccentric ones.

If the concentric portions of the exercise were considered, the RMS-SEMG of all muscles, the longissimus and iliocostalis lumborum recording sites, significantly increased in individuals of both age and sex groups over the course of the exercise routine and did not significantly differ between younger and older participants, nor between males and females. As opposed to the younger participants, no significant RMS-SEMG fatigue increases were observed for the multifidus muscle of the older participants but differences between age groups did not reach a level of significance. For the eccentric portions of the exercise, the RMS-SEMG slopes were found to significantly increase in younger participants in the iliocostalis lumborum electrode recording site. However, neither the respective age nor sex specific differences reached a level of significance. Interestingly, those electrode sites that revealed the most negative IMDF slope showed a significant decrease in the normalized RMS-SEMG changes for the eccentric portions of the exercise, but did not differ significantly between age and sex groups. Tables 2 and 3 display the median RMS-SEMG onsets and changes over time normalized by the RMS-SEMG values derived from the 80% isometric MVC test. Our second analysis with the full, unfitted data set revealed similar findings.

The IMDF-EMG onsets obtained from the cyclic back exercise tended to be slightly higher in older as compared to younger individuals. Neither age-specific nor sex-specific differences reached a level of significance for both the concentric and eccentric contraction phases.

The IMDF-SEMG slopes normalized to the IMDF onset values significantly decreased with the duration of the exercise. Although these slopes were generally more pronounced in younger than in older participants, significant age specific differences were predominantly observed when the IMDF-SEMG slopes were derived from the concentric phases of the exercise. This was true for all electrodes, the pair of electrodes with the most negative slope, and the multifidus and longissimus (in a tendency significant) electrode recording sites. For the eccentric phases of the cyclic exercise, significant age specific differences were observed for the multifidus muscle and the electrode with the steepest IMDF-SEMG slope. No sex specific differences were observed, except for the most negative electrode recording site. There were neither age nor sex specific differences observed for the imbalance scores. The results of the age and sex specific differences of the IMDF-EMG are presented in Table 4 for the concentric phase and in Table 5 for the eccentric phase. Our secondary analysis including the full, unfitted data set revealed age specific differences for the IMDF onsets and slopes that were similar to those obtained from our primary analysis.

Overall, the older participants recruited in this study were more physically active than their younger counterparts. In fact, all older individuals clearly exceeded the recommended minimum physical activity levels [54]. Such high physical activity levels amongst the older individuals likely explains why the maximum back extension strength scores of our older participants were similar to those of our younger participants, and appeared to be 25 to 30% higher than would normally be expected [3]. Moreover, the scores obtained from our younger participants widely corresponded with those observed from healthy males and females who had been tested using similar equipment [55]. Therefore, the back scores obtained from the younger as well as the older participants who participated in this study likely reflect close to true maximum sores. The lack of any age specific differences in back extension strength further suggests that the consequences of an assumed aging-induced motor neuron loss in our older participants [8, 56] may have been functionally well compensated, most likely through their relatively high exercise levels.

The biomechanical variables “range of motion” and “mean movement velocity” that had been monitored to control the biomechanical quality of exercise performance were well matched with the overall requirements of the repetitive flexion extension test and kept constant throughout the exercise. Moreover, as neither the onsets nor the changes over time of these variables differed between the younger and older participants in a relevant way, no major confounders on EMG recordings from variable biomechanical performance of the cyclic exercise could be expected. Although the work performed during the exercise was comparable between the two age groups, the younger individuals reported significantly more back muscle fatigue at the end of the cyclic back exercise, even though the respective scores still indicated fatigue to be moderate. This result is inline with those from systematic reviews that found the back muscle endurance function to be better preserved in older as compared to younger individuals [4, 57]. In addition, findings from a study wherein healthy young individuals were able to perform a maximum of 34 cycles of back extensions on a Roman chair, with the back loaded to a similar extent and the cyclic exercise performed within the same range of movement and at comparable movement velocity [45], seems to externally validate the fatigue ratings observed amongst the younger participants in this study.

Research that investigated age specific differences in motor unit control using the needle electromyographic method observed a proportionally higher number of motor units activated that fired at similar [58] or modestly lower rates [8, 59] when older individuals produced a sub-maximum force comparable to that of younger persons. Against our expectations, this study’s results revealed a lack of relevant age-specific differences for the normalized back extensor RMS-SEMG onset score. Our results suggest that this variable is likely not sensitive enough to detect aging induced adaptive alterations in neuromuscular activation. Several confounders known to affect the RMS-SEMG provide an explanation. Among these, amplitude cancellation that preferentially occurs if motor unit action potentials (MUAPs) of simultaneously active motor neurons are more homogeneous in shape and have a longer duration [60, 61] may have significantly confounded age specific differences for this EMG variable. As with muscles of the extremities, aging back muscles are probably also composed of motor units whose MUAPs are, if compared to young muscles, larger in area and amplitude and more homogeneous in shape but reduced in absolute number [3, 6, 8, 9, 26, 56]. These characteristics expose the amplitude SEMG variables, resulting in a higher risk for amplitude cancellation with lower than expected normalized values in older than younger individuals.

As was found in previous studies [45, 62], the RMS-SEMG significantly increased over the course of the concentric phases of the exercise and tended to be more pronounced in the younger than in the older participants, when the concentric mode of the exercise was considered. This consistent result supports the theory that such fatigue related increases in SEMG amplitude reflect both a progressive recruitment and an increase in firing rate of motor units in order to compensate for fatigue induced muscle fiber contraction failure [63‐65]. The back muscles of the younger participants in this study had a comparatively smaller percentage of motor units activated when moving the 50% of maximum load, followed by a relatively stronger pronounced progressive recruitment of motor units and a larger increase in firing rate. This was likely due to a higher reserve capacity to adjust for an increase in fatigue driven compensatory excitatory drive to the alpha motor-neuron pool in younger individuals. These factors could theoretically explain the more pronounced rise in SEMG amplitude in the younger participants. However, the differences in RMS-EMG increases failed to demonstrate sufficient age specific discriminative power. Therefore, this variable may, as with the amplitude SEMG onset variable, be assumed to be of minor importance if the early detection of aging muscle function is of interest.

It is interesting that, when the eccentric contraction mode of the exercise was considered, a significant decrease rather than an increase in RMS-SEMG was observed over the course of the exercise for both the elderly and younger group of participants. In eccentric contractions muscle fibers have a greater intrinsic force capacity [66] and the generated force is smaller than the load that is to be resisted. This is accompanied by recruitment of fewer MUs that fire at lower rates than during concentric contractions, if the muscle length vs. force relationship is comparable [67‐69]. A more pronounced increase in amplitude cancellation with the duration of exercise in the older as compared to younger participants could likely be due to the fatigue-induced widening in the duration of the MUAPs, as well as a progressive recruitment of motor units that would be more homogeneous in shape in elderly persons with muscle fatigue.

In static back extensions at increasing strength, the median of the frequency spectrum (MF) only increased up to 30–50% of maximum [70]. If the strength further increases, then the MF-SEMG decreases. This is in a clear contrast to muscles of the upper and lower extremities whose mean MF-SEMG spectrum is known to increase with increasing muscle force production until progressive motor unit recruitment has finished [70, 71]. Considering all of the following: 1) a close relationship between median frequency of the EMG signal to the muscle fiber conduction velocity, 2) the shape of the action potential, and 3) the temperature of the muscle, such atypical changes of the MF-EMG from back extensors with increasing force would be best explained by the recruitment of the in-cross section larger slow twitch fibers at lower force thresholds than the smaller fast-fatiguing muscle fibers [35, 71, 72]. Likewise, confounding variables of the MF-EMG such as spatial filtering, differences in recruitment ranges of different muscles detected in the electrode recording area overlying muscles, and interference phenomena of the many concurrently active MUAP trains may further explain non-linear force- MF-SEMG behavior in increasing strength of back extensor muscles. As this study did not reveal any age or sex specific significant differences for the IMDF-EMG onset values, this variable likely provides a useful measure for normalizing the IMDF-EMG fatigue slopes. Such a normalization procedure would control for: 1) overall smaller muscle fiber conduction velocities of the older muscles; 2) other confounders of the SEMG signal including variability in subcutaneous fatty tissue under the electrode [64]; 3) age specific differences in maximum recruitment thresholds as well as differences in maximum recruitment thresholds between the short, deep, and longer superficial back extensor muscles [73]; and, 4) the loss of orderly recruitment of motor units, with larger motor units who fire at lower rates being recruited earlier in a contraction when force increases [6, 9].

The only SEMG variable that clearly indicated aging-induced muscle functional changes in cyclically activated back extensor muscles was the IMDF-SEMG fatigue slope. This result corresponds with our previous observation made from static back extensions that had been performed at 80% of maximum for 30s [24]. It is important to note that such age specific differences in SEMG spectral compression occurred although the load of the cyclic exercise equated only to that of 50% of maximum, and the older patients’ back muscle strength had been comparable in the two different age groups. If the fatigue dependent degree of compression of the IMDF-SEMG spectrum toward lower frequencies was mainly dependent on the slowing in muscle fiber conduction velocity, then an overall smaller percentage of fast fatiguing motor units had contributed to the significantly less pronounced IMDF-SEMG slopes in our older individuals. The flatter IMDF fatigue slopes observed from older adults were likely due to either: 1) adaptive changes in back muscles resulting from an increasingly forward-leaning posture of the trunk in people older than 50 years old [74] with an increased mechanical loading of these muscles during moving around, and/or 2) compensatory adaptive processes caused by an aging induced loss of motor neurons. For the latter explanation, orphaned fast twitch fibers that remain from the apoptotic neurons of previous high threshold motor units will be re-innervated by the axons of low threshold motor units, which are typically comprised of fatigue resistant muscle fibers. Such re-innervation would force a formerly fast fatiguing, yet re-innervated muscle fiber to adapt its behavior through altered mechanic demands towards those required for low threshold motor units. If such re-innervated fast twitch, fast fatiguing muscle fibers could not translate towards a slow twitch, fatigue resistant one [75], then apoptosis caused by overuse might occur. Indeed, co-expression of two instead of one myosin heavy-chain (MHC) isoforms has been demonstrated in at least 40% of the total muscle fibers within a muscle, [76] which suggests that such a fast to slow muscle fiber translation is likely an adaptive process. However, such a compensatory re-innervation after axonal degeneration may remain widely incomplete [36], as shown from the vastus lateralis muscles where 40% of muscle fibers had been lost in persons aged approximately 75 years old. The maintenance of the maximum back extension torque despite loss of fast twitch muscle fibers in our older participants may be explained through regeneration via the muscle fiber satellite cell pool, thereby further enlarging the motor units and compensating for lost muscle fibers. This would be consistent with research that found failure to expand the motor units to be associated with a loss in muscle mass [27], as well as research that muscle fiber atrophy can be overcome at least in part by physical exercise [77‐79] and drug interventions [80]. It is worthwhile noting that with advancing age, regeneration through the type II fiber satellite cell pool seems more limited than that of type I muscle fibers [81]. In addition, more mechanic stress would train the enlarged motor units of older individuals who would comprise less glycolytic type II muscle fibers (but more aerobic type I and hybrid type II a muscle fibers), and thus increase the intrinsic twitch force output the muscle fibers of the remaining motor units [82].

The IMDF–SEMG fatigue slopes appeared to be more sensitive to detecting age specific differences during the concentric than in the eccentric phases of the exercise. The SEMG accompanying concentric versus eccentric contraction has been extensively discussed in a recent study [47]. In short, less recruitment of motor units and a lower firing rate during the eccentric phase of the contraction (as indicated by a reduced RMS SEMG) seems to lower the age specific discriminative power of the IMDF-SEMG slopes. This result suggests that a minimum threshold load relative to maximum may exist for conducting a diagnostically accurate test. In addition, our findings further suggest that normalization of the IMDF-EMG slope to the initial IMDF frequency is advisable, as it would compensate for confounders that may occur from spatial filtering due primarily to individual differences in the distance from the muscles to the surface electrodes.

This study also investigated the reproducibility of the variables of interest and found the relative day-to-day retest reliability for the SEMG variables to be good to excellent, depending on the recording site. Absolute reliability, however, was highly variable for both the normalized slopes and the EMG onset values. This was true for both age specific subgroups. These observations concur with those from earlier research where healthy individuals performed cyclic free lifting [34] or a sustained 80% of maximum contractions [24]. Future research will have to address the need to improve the precision of the IMDF EMG fatigue variables or introduce new IMDF fatigue-based variables if the SEMG cyclic back extension exercise test is intended to be further developed to a sensitive standard biomarker to detect the very early cases of sarcopenia.

One may suspect that the fatigue slopes observed from our older population were biased by significantly higher physical activity scores as assessed with the IPAQ and higher than expected back extension strength scores. It is worth noting that the spectral MF-SEMG fatigue method is independent from an individual’s absolute maximum strength score, but the submaximum load needs to be derived from a true maximum performance. Given this, the age specific differences observed with this method would be widely unaffected by the training state of the back muscle unless the respective strength and power scores were considered in addition. This is important because training induced compensation in strength and power decline following an aging related loss of glycolytic type II fibers is mainly achieved through mechanisms leading to an increase in mechanic output in less fatigable type II and fatigue resistant type I muscle fibers [5, 77]. This all likely suggests that a decline in the proportion of high threshold motor units with age as indicated by less IMDF-SEMG fatigue would be widely unaffected by the training state of an individual’s back extension strength. This conclusion seems supported by very recent findings from age specific differences in SEMG-fatigue observed in cLBP patients a static 80% of maximum sustained isometric back extension during 30s. In this recent study, older cLBP patients were able to produce 85% MVC back extension strength of that observed from younger patients, however, older and younger cLBP patients were comparable in their physical activity scores as assessed with the IPAQ [83].

It could be argued that differences in blood perfusion of the back extensor muscles confounded the findings from the spectral SEMG observed in this study [64]. We did not monitor blood perfusion, nor did we take any muscle probes that would have confirmed a similar vascularization between age groups. However, this confounder seemed well controlled by the relatively highly trained state of the back muscles observed amongst our older healthy participants. The high level of training amongst older participants caused a capillarization of muscle fibers that was as good as that observed in young individuals [84]. However, even if back muscles had been clearly weaker, less trained and thus probably less capillarized in our older participants, the IMDF-SEMG slopes - as opposed to the IMDF onsets - would have remained widely unaffected as evidenced from individuals with peripheral occlusive disease [85, 86]. Thus, the age specific discriminative power of this variable has only a small risk of being confounded by altered capillarization of the back extensors.

It is also possible that sex specific differences could interfere with the IMDF-SEMG fatigue changes. In fact, fast fatiguing high threshold motor units of the back muscles were found to be smaller in females than males [87], and for both the RMS and IMDF fatigue SEMG recorded during static sustained back extensions, sex specific differences have repeatedly been observed in static sustained back extensions [88]. In this study, however, we did identify sex specific differences within slopes of the IMDF-EMG of the most negative electrode only; this suggests that the age specific effects observed for the IMDF-EMG in this study are widely unaffected by sex differences.

We attempted to record a full set of EMG data during the cyclic exercise from a total of 87 individuals. Unfortunately, artifact free recording from all electrode sites was feasible in only 56 individuals. Intermittent loss of electrode skin contact leading to major EMG artifacts, which occurred most frequently at the L5 and L2 recording sites, was one of the main reasons for exclusion of data. This may have reduced the statistical power to detect age specific differences in this explorative study. We therefore analyzed the fitted EMG data that considered side related differences in pairwise electrode recordings. Results between primary and secondary analyses were similar. In fact, future research will have to compare different types of electrodes with defined inter-electrode distance that would be unlikely detach during the performance of such a cyclic exercise.

Loss of neurons may start in an individual’s later 40s, and seems to be independent from any muscle training performed [3]. As back muscle loss has been demonstrated to occur at a comparatively high rate [19], and impaired lumbar extensor muscle function (weakness and fatigue) is associated with loss of balance, postural alignment and control [20, 21, 23], and osteoporosis [89], functional assessment based on strength and endurance scores is of utmost importance. Proving the early loss of back-extensor muscles with a simple test like ours, before a decrease in back muscle strength and endurance becomes overt, seems key to early interventions intended to decrease the higher risk of future falls and frailty. However, before this functional back muscle test can be offered as a preventive diagnostic tool, data processing would need to be further developed to a fully automatic procedure, and the absolute reliability of the IMDF-EMG fatigue slope-based variables needs to be further improved. As our novel EMG-based test is based on a sub-maximum cyclic exercise protocol at moderate load, testing can likely be considered safe even in patients at risk for increased vertebral fractures, and it is simple and quick to perform [90]. Therefore, this test would meet the basic requirements for implementation of nation-wide screening to detect early forms of deterioration in muscles that are the most likely to predispose elderly persons to future falls and fall-related injuries, frailty and early death.