Normalization of the hip abductors rate of activation during a voluntary step task in older adults: Reliability and differences among approaches

Abstract

Background:

Voluntary stepping tasks are used to measure the ability of an individual to step and has been associated with fall risk in older adults. Although electromyography (EMG) amplitude is measured during stepping tasks, there is no consensus about the reference EMG value that should be used to normalize the signal. The purpose of the present study was to 1) investigate the impact of using different EMG parameters as a reference to normalize the rate of activation (RoA) of the hip abductor muscles across lateral voluntary step trials and the differences between trials, and 2) to investigate the reliability among trials of the reference EMG values.

Methods:

Nineteen older adults (>65 years of age) performed ten lateral choice reaction stepping test (CRST), while the gluteus medius and tensor fascia latae EMG were recorded. Three reference EMG values were calculated and used to normalize RoA during the stepping task. A repeated-measures ANOVA was used (normalized RoA[3] x trial[3]) to compare normalized EMG across trials, and an intraclass correlation coefficient and coefficient of variation were used for the inter-trial reliability of the reference EMG values.

Results:

The present study showed that gluteus medius and tensor fascia latae RoA normalized values from the stance and stepping leg (right or left side) measured during CRST are different according to the reference EMG value(P<0.001), with no differences across trials. Overall, the EMG procedures showed high inter-trial reliability, with a few exceptions.

Significance:

Therefore, after careful examination of our results, the peak and mean EMG amplitudes showed consistently higher intraclass correlation coefficients; however, the former may provide a more intuitive reference value.

1. Introduction

Surface electromyography (EMG) is a technique often used as a tool to understand the interaction between the brain and muscle across different movements, including stepping [20,33]. Voluntary stepping tasks (i.e., choice reaction stepping test, CRST) are consistently used to measure an individual ability to step [20,31,32] and present an association with fall risk [24,27]. Taking a rapid step is a common strategy among older adults to avoid a fall [23,29], and EMG amplitude measurements [e.g., rate of activation (RoA)] have been reported during stepping tasks [15,20]. Normalization is used to account for factors that influence EMG amplitude [e.g., subcutaneous fat, cross-talk between muscle and others [9,22]], is regularly used and required depending on the context. The EMG normalization process consists of obtaining a reference EMG value during a reference task (e.g., maximal contraction) and then dividing the EMG amplitude during a task (e.g., submaximal contraction) by the reference EMG value and multiplying by 100 to express as a percentage value [21]. The percent generated can be used as an index of function that allows researchers to compare EMG amplitude among different contexts and subjects [e.g, different days and/or individuals, more can be seen here [4,6]]. Currently, there is no consensus about the reference EMG value and task used to obtain the value for normalizing the EMG amplitude during a stepping task.

There are various methods used for the normalization of isometric and dynamic movements. Researchers have used EMG amplitude from different tasks to normalize the EMG amplitude measured during the stepping task [15,18,20,26], and there are some studies that have not normalized the EMG signals [10,13]. For example, while one study normalized the EMG amplitude during the task by the EMG peak during the stepping task [15], others normalized by the EMG amplitude measured during a maximal voluntary isometric contraction (MVIC) [20,26], or used baseline EMG activity before the task as reference EMG value [18]. Among the examples mentioned above, the EMG amplitude measure during MVIC has shown excellent reliability [3,21] but it is unknown whether other approaches to normalizing EMG amplitude during dynamics tasks are reliable. Moreover, it is unclear if EMG amplitude normalized by different EMG reference values will provide similar outcomes and if the reference values are reliable across trials. Investigating the reliability across trials within the same session is critical since many clinical research studies do not perform tests across days, but rather within the same day.

A well-known and often recommended approach to normalized EMG amplitude is the use of EMG amplitude measured during an MVIC [4,6]. However, it is not always possible to use this approach due to issues such as longer testing sessions and/or limitations in physical function (e.g., stroke individuals). Since stepping tasks are often used in populations with physical limitations, such as older adults or persons post-stroke [12,20], using an approach from the specific task performed that does not require a maximal effort might be an alternative. In this context, current guidelines suggest using a standardized submaximal task or mean EMG amplitude during the task [4]. Additionally, a previous study demonstrated that EMG amplitude measured during dynamic tasks (i.e., jumping and sprinting) in young adults was more reliable than EMG measured during MVIC [30]. Nonetheless, it is unknown how different approaches of reference EMG values influences the normalized EMG amplitude measure during a stepping task (are they providing the same outcome?).

It has often been suggested that the hip abductor muscles play an important role during stepping in the older population, contributing to joint stabilization [1,16,25]. Moreover, to quickly react and avoid a fall by performing a step, the RoA of the hip muscles during the step has shown to be important. Previous research has demonstrated that quicker RoA of the hip abductor [15,20] and adductor [20] muscles influence step performance. Nonetheless, the RoA is a commonly used measure of EMG amplitude in various contexts [e.g., clinical assessments [28]]; hence, further understanding the impact of different normalization procedures is critical. Therefore, the purpose of the present study is twofold. First, investigate the impact of using different EMG parameters as a reference to normalize RoA of the hip abductor muscles across lateral voluntary step trials and the differences between trials. It was hypothesized that normalized RoA would be different between approaches but similar across trials. The second aim is to investigate the reliability of the reference EMG value across trials used to normalize RoA during a voluntary stepping task. We hypothesized that all EMG parameters would present excellent inter-trial reliability.

2. Methods

2.1. Subjects

Nineteen healthy older adults [8 males / 11 females; age= 70.9 ± 4.6 years; height= 1.69 ± 0.1 m; mass= 79.8 ± 14.04 kg; body mass index (BMI) = 28.2 ± 4.94] were volunteers in this study. The inclusion criteria for participation were to be above 65 years old, free of lower limb injuries, and a BMI lower than 40. We performed a post-hoc power analysis [17]. By using G*Power (version 3.0.1, Germany), based on the ANOVA (repeated-measures, within factors), with our sample size of 19 subjects using the effect size of 0.71 (which was the smallest effect size - TFL stepping leg left side); by setting an alpha error of 0.05; the number of groups =3; based on three measurements; correlation among repeated measures= 0.5; and nonsphericity correction= 1, gave us a minimal power of 0.99.

2.2. Procedures

2.2.1. Voluntary step test

Participants performed ten trials of the CRST by standing on two adjacent force platforms (AMTI, Watertown, MA). All volunteers were instructed to take a lateral step as fast as they could following a visual “go” cue without prior knowledge of the stepping leg. The “go” cue was preceded by a warning signal of an inter-stimulus delay of 100 ms. A Matlab code was used to trigger the light cue and also to trigger the recording of the EMG recordings in order to synchronize the EMG signals with the light cue signal for off-line analysis. The direction of the step was randomly assigned (right vs. left), and ten steps were performed (5 trials x 2 sides), Fig. 1. Participants were provided with at least two trials (one on each side) to familiarize them to the task. A footprint was drawn on a piece of paper attached to the force platform to ensure the participants started each trial in the same position. The force platform data were collected at a frequency of 1500 Hz.

Fig. 1.

Sketch of the choice reaction step test. Gray dots indicate the lights for the light cue. A) Initial position; B) Step to the right; C) Step to the left.

2.2.2. Surface electromyography

Surface EMG electrodes were placed on the target muscles before testing on the right and left legs. After skin preparation, surface bipolar pre-gelled disposable electrodes (Ag/AgCl – 2 cm center to center, Noraxon) were positioned parallel to the muscle fibers on the gluteus medius (GM), and tensor fascia latae (TFL) muscles on both legs accordingly to SENIAM guidelines [14]. EMG signals were recorded using a wireless EMG system (NORAXON, USA Inc.) at a frequency of 1500 Hz. The EMG signals were collected in MyoMuscle XP software (Noraxon, USA Inc.) and exported to Matlab software for off-line analysis. The EMG data were filtered (4th-order Butterworth band-pass filter of 20–500 Hz). On the next step, EMG signals were full-wave rectified and down sampled to 500 Hz. After that, the RoA was calculated from EMG onset to EMG peak for each CRST trial and each muscle, and the average rectified value was divided by the respective time period [20]. The onset of the EMG signal was defined as an increase exceeding three standard deviations above the mean baseline EMG, recorded after the light cue activation. The baseline was established using the 100 ms period immediately preceding the appearance of the light cue for each trial.

The reference EMG values (procedures) used for EMG normalization, and performed for each trial, were the following three: 1) EMG peak across the whole step trial (EMG_TrialMax) measured as the highest value from step onset to heel strike. 2) EMG mean measured over the whole step trial (EMG_TrialMean) was calculated as an average value from step onset to heel strike. 3) EMG measured at baseline (EMG_Baseline) was calculated as the mean EMG RMS value of 100 ms before the light cue. The normalized EMG was calculated for the RoA by dividing the RoA by the reference EMG value and multiplying by 100 (e.g., GM RoA / GM EMG_TrialMean * 100). Thus, normalized EMG in the present study is expressed as a percentage of the EMG reference value.

2.3. Statistical analyses

Given that the muscle activation is expected to differ between legs (stance vs. stepping leg) during the CRST, statistical analyses were performed for the stance leg and the stepping leg. To answer the first aim of this study, a two-way repeated-measures ANOVA [3 procedures (EMG_TrialMax, EMG_TrialMean, EMG_Baseline) x 3 trials] was used to analyze the differences between RoA normalized by the reference EMG values across step trials. A separate ANOVA was run for each muscle (GM and TFL) of the stance andthe stepping leg andfor the right and left side. Post hoc comparisons were conducted using Bonferroni post hoc test with corrections. To answer the second aim of the study, the relative reliability of the normalization procedures was tested among the three trials by intraclass correlation coefficient with the 3,1 model with an absolute agreement (ICC_3,1) [2] and coefficient of variation [CV%, (SD/mean) x 100 %]. The ICC values were interpreted as weak (<0.4), moderate (0.4–0.59), good (0.6–0.74) or excellent (0.75–1.0) [8]. Three trials were selected among the five based on the time to complete the step (foot contact after light cue), with the fastest trials (better performance) being selected for analysis. Statistical analyses were performed using SPSS program version 29.0 (IBM Inc., Chicago, IL, USA), and the level of significance was set at P<0.05.

3. Results

3.1. Gluteus medius (GM)

When acting as the stance leg, normalized GM RoA showed a main effect of procedure on the left side (P< 0.001) and right side (P< 0.001), but there was no main effect of trial or an interaction (P≥ 0.073). Similar, when acting as the stepping leg, normalized GM RoA demonstrated a main effect of procedure on the left side (P< 0.001) and right side (P< 0.001), but there was no main effect of trial or an interaction (P≥ 0.793). All normalized values of the GM RoA were different from each other among procedures (P< 0.001) for all conditions (stance and stepping leg, and left and right side). GM RoA normalized by EMG_Baseline showed higher values than the other normalized procedures, followed by GM RoA normalized by EMG_TrialMean being higher than GM RoA normalized by EMG_TrialMax; Fig. 2 (A and B) and 3 (A and B). Reliability: ICC was excellent (above 0.842) for stance and stepping leg (right and left side) for all reference procedures, with CVs ranging from 12 up to 28 %; Tables 1 and 2.

Fig. 2.

Normalized rate of activation (RoA) mean values for EMG_TrialMax (black columns), EMG_TrialMean (white columns), and EMG_Baseline (gray columns) from the choice reaction step test step from gluteus medius (GM), and tensor fascia latae (TFL) muscles for each trial. (A, left side) and (B, right side) are the GM RoA values from stance leg, while (C, left side) and (D, right side) are the RoA values from stance leg for TFL. * Symbol indicate that RoA normalized to EMG_Baseline was higher than RoA normalized by EMG_TrialMax and EMG_TrialMean; # indicates that RoA normalized to EMG_TrialMean was higher than RoA normalized by EMG_TrialMax.

Fig. 3.

Normalized rate of activation (RoA) mean values for EMG_TrialMax (black columns), EMG_TrialMean, (white columns), and EMG_Baseline (gray columns) from the choice reaction step test step from gluteus medius (GM), and tensor fascia latae (TFL) muscles for each trial. (A, left side) and (B, right side) are the GM RoA values from stepping leg, while (C, left side) and (D, right side) are the RoA values from stepping leg for TFL. * Symbol indicate that RoA normalized to EMG_Baseline was higher than RoA normalized by EMG_TrialMax and EMG_TrialMean; # indicates that RoA normalized to EMG_TrialMean was higher than RoA normalized by EMG_TrialMax.

Table 1.

Stance leg inter-trial intraclass correlation coefficient (ICC) and coefficient of variation (CV) for all the reference EMG parameters for right and left gluteus medius (GM), and tensor fascia latae (TFL) muscles.

Stance Leg
		ICC	95 % CI	CV (%)
	Right Side
GM	EMGTrialMax	0.902	0.788–0.959	28.3 ± 4.3
	EMGTrialMean	0.944	0.880–0.977	14.4 ± 3.4
	EMGBaseline	0.920	0.829–0.967	12.0 ± 1.6
TFL	EMGTrialMax	0.950	0.894–0.979	27.9 ± 4.7
	EMGTrialMean	0.917	0.823–0.965	21.0 ± 3.4
	EMGBaseline	0.941	0.875–0.976	13.5 ± 4.2
	Left Side
GM	EMGTrialMax	0.842	0.652–0.936	29.6 ± 5.0
	EMGTrialMean	0.943	0.875–0.977	19.3 ± 2.6
	EMGBaseline	0.966	0.926–0.986	16.1 ± 5.0
TFL	EMGTrialMax	0.922	0.833–0.968	26.7 ± 4.0
	EMGTrialMean	0.947	0.886–0.978	19.3 ± 2.6
	EMGBaseline	0.720	0.408–0.883	16.1 ± 5.0

Table 2.

Stepping leg inter-trial intraclass correlation coefficient (ICC) and coefficient of variation (CV) for all the reference EMG parameters for right and left gluteus medius (GM), and tensor fascia latae (TFL) muscles.

Stepping Leg
		ICC	95 % CI	CV (%)
	Right Side
GM	EMGTrialMax	0.964	0.923–0.985	20.8 ± 2.4
	EMGTrialMean	0.976	0.949–0.990	13.9 ± 2.2
	EMGBaseline	0.855	0.755–0.952	17.5 ± 4.2
TFL	EMGTrialMax	0.776	0.524–0.906	30.1 ± 6.4
	EMGTrialMean	0.925	0.840–0.969	22.0 ± 4.5
	EMGBaseline	0.819	0.598–0.927	19.7 ± 4.6
	Left Side
GM	EMGTrialMax	0.895	0.773–0.957	26.4 ± 2.2
	EMGTrialMean	0.934	0.855–0.973	12.8 ± 2.6
	EMGBaseline	0.966	0.926–0.986	12.6 ± 2.9
TFL	EMGTrialMax	0.771	0.512–0.904	43.9 ± 4.9
	EMGTrialMean	0.541	0.004–0.810	28.4 ± 4.1
	EMGBaseline	0.747	0.460–0.895	27.4 ± 8.8

3.2. Tensor fascia latae (TFL)

When working as the stance leg, normalized TFL RoA showed a main effect of procedure on the left side (P< 0.001) and right side (P< 0.001), but there was no main effect of trial or an interaction (P≥ 0.314). Similarly, when acting as the stepping leg, normalized TFL RoA demonstrated a main effect of procedure on the left side (P< 0.001) and right side (P< 0.001), but there was no main effect of trial or an interaction (P≥ 0.523). TFL RoA normalized values were different from each other among procedures (P< 0.001) for all conditions (stance and stepping leg, and left and right side). TFL RoA normalized by EMG_Baseline was higher than the other normalized values, followed by TFL RoA normalized by EMG_TrialMean being higher than TFL RoA normalized by EMG_TrialMax; Fig. 2 (C and D) and 3 (C and D). Reliability: TFL presented an excellent ICC (above 0.771) for stance and step leg (right and left side) for all reference procedures except for stance leg left side for EMG_Baseline (ICC= 0.720) and stepping leg left side for EMG_TrialMean (ICC= 0.541). The CVs ranged from 13.5 up to 43.9 %; Tables 1 and 2.

4. Discussion

4.1. RoA normalized by different EMG parameters

The first aim was to investigate the impact of using different EMG parameters as a reference to normalize RoA across voluntary step trials and the differences between approaches. Confirming our initial hypotheses, normalized RoA from both muscles (GM and TFL) differed depending on the EMG reference approach, but the EMG reference values did not differ across trials. The results bring important implications for EMG normalization of the hip abductors during a voluntary stepping task. For instance, comparing results between studies that used different EMG approaches should be made carefully since contrasting EMG approaches presented differences in normalized RoA. Yet, the EMG reference values were reproducible across trials within each normalization method, showing that whichever parameter was used the results were consistent across step trials regardless of the parameter.

It does not come as a surprise that the different EMG normalization methods had an impact on EMG amplitude (RoA), since others have reported similar findings [3,19,21]. The mean EMG at baseline produces a small value, and it would be expected to have a higher normalized value than the other approaches (EMG_TrialMax and EMG_TrialMean) as presented here. Additionally, in our study we found that the RoA normalized by EMG_TrialMax and EMG_TrialMean were significantly different, which is contrary to a previous study [21]. In a study by Lanza et al. [21], EMG measurements were recorded from the superficial muscles of the quadriceps femoris during a submaximal knee extension task in young adults. The study compared seven different approaches to normalizing EMG amplitude, including using peak and mean values of the task, which are also utilized here. These contradictory findings reinforce that different tasks (e.g., stepping task vs. knee extension) and muscles (e.g., knee extensors vs. hip abductors) may result in a smaller or larger EMG values. Thus, relying on information from a different task or muscle group to justify the choice for EMG normalization might not be adequate.

The present study showed no differences between trials for each RoA normalized value, regardless of the procedure used for normalizing the RoA, for the stance or stepping leg during the voluntary step. The consistency between trials, independent of the normalization procedure, confirms the robustness of the methods and research design employed here. Moreover, this consistency may occur because the voluntary step is a task that is easy to perform consistently in the setup we used. During a voluntary step, an individual needs to unload one leg, the stepping leg, and transfer weight to the opposite leg, the stance leg, to take the step. The stance leg muscles work to stabilize the body, while the stepping leg muscles assist with push-off, preparing to initiate the step. In our scenario, where there is no unpredictable movement, consistency between trials might be expected, yielding a more consistent EMG signal. However, during an induced step (e.g., pushing or pulling someone off balance) [12], the EMG pattern may vary between trials. Thus, future research should investigate whether the EMG procedures used here to normalize the RoA during the step are still reliable measures during induced steps.

4.2. Reliability of the EMG parameters

An important factor in deciding about the EMG parameter used to normalize EMG amplitude is the reliability of the reference EMG parameter [6]. The second aim was to investigate the inter-trial reliability of the reference EMG parameters used to normalize RoA during the voluntary stepping task. Confirming our hypotheses, overall, the three EMG parameters presented excellent inter-trial reliability (with a couple of exceptions in our results). The TFL EMG_TrialMean and EMG_Baseline at the stepping leg showed good and moderate reliability, respectively. Our study was not designed to explain why there are differences between the various EMG parameters. However, based on our data, we can suggest that this lower reliability, compared to other parameters, is due to the large variability intra-trial within subjects that we are seeing in these two variables. However, considering the excellent reliability of the other TFL EMG parameters, it is difficult to provide a physiological perspective for these differences.

To the best of the author’s knowledge, this is the first study that looked at inter-trial reliability during a voluntary stepping task for the EMG parameters used here. Another study investigated the reliability between days of two parameters (EMG peak during the task and EMG during a MVIC) to normalize surface EMG of the GM during step up and step down tasks [11]. The authors found a excellent reliability for the reference EMG during MVIC but not for the reference EMG peak during the task (poor to excellent reliability). Although we showed an overall excellent inter-trial reliability for a similar parameter, the difference between days may be influenced by other factors (e.g., electrode placement), hence, further research should explore this possibility in a similar scenario then ours.

Additionally, previous research investigating between days reliability using similar parameters, but different tasks (resistance training exercise) presented mixed results [3] but most literature showed a excellent reliability between days [5,7,21,30]. It is important to highlight that in our study, only one parameter (EMG_TrialMean), for one muscle (TFL), at the stepping leg (left side) showed moderate reliability. Hence, overall our results confirm that all the parameters investigated here are reliable and can potentially be used as a reference to normalize EMG amplitude during a stepping task for the stance and stepping leg.

4.3. Strengths, limitations, and practical implications of the study

This study has strengths and limitations that should be highlighted. We presented to researchers three reliable options to normalize EMG amplitude during a lateral voluntary stepping test. Additionally, it is clear that comparisons between studies that use different normalization approaches than our study should be made with caution. A limitation is that we are only reporting inter-trial reliability, and studies that assess muscles after an exercise intervention would benefit from understanding between-day reliability. Also, although in this study we are investigating submaximal levels of EMG activation as an alternative for normalization, the use of EMG during a maximal isometric contraction is a well-known recommended alternative [4,21]. However, in some study designs with some populations (e.g., older adults), performing a maximal isometric contraction may not be possible due to the risk of fatigue. Also, MVC may not be suitable for functional tasks when examining muscle joints with changing joint angles. Thus, the proposed method may be a good alternative for functional tasks.

Another strength of this study is establishing that the EMG reference values used here are reliable in the presented setup. Previous research using a similar setup to ours normalized EMG amplitude by the baseline, peak, or mean of the trial within each trial, which we replicated here. The present result could have significant implications, such as using the highest EMG value (peak or mean of the trial) among all trials as the reference for normalization. The EMG measured at baseline showed overall excellent reliability, but it is not clear if the normalized values of RoA using this procedure have some physiological meaning. For instance, the use of the EMG_TrialMax and EMG_TrialMean generates an index that represents a percentage of the mean or maximal activation during the step. Thus, it gives a more intuitive interpretation of the values generated. The mean EMG measured at baseline is a measure of muscle activity pre-contraction, so the meaning of the of the EMG normalization to actual muscle performance is not evident. Nonetheless, to be consistent between studies in order to compare information, it would be beneficial for researchers to use similar EMG normalization methods. In this study, the EMG_TrialMax was the only reference EMG parameter from both muscles (GM and TFL) that showed excellent reliability across all the condition and side (stance and stepping leg, right and left side). More importantly, it generated a meaningful index relative to the task performance (% of the maximum values). Therefore, based on our results, using the EMG_TrialMax as a parameter to normalize RoA during a CRST is recommended. This recommendation aims to keep consistent across studies from different laboratories and will help others to compare their results to the literature. However, future research should replicate the present results.

5. Conclusion

To conclude, the present study showed that GM and TFL RoA normalized values from the stance and stepping leg, right or left side, measured during a lateral voluntary stepping task are different accordingly to the EMG parameter used to normalize the signal. However, no differences across trials were found for any RoA normalized value. Additionally, the reference EMG procedures showed high inter-trial reliability in virtually all conditions. However, considering the exceptions and based on our results, the peak of EMG amplitude measured during the trial would be recommended to normalize RoA during a CRST.