INTRA-SESSION AND INTERSESSION RELIABILITY OF ELECTROMYOGRAPHY IN LEG EXTENSION DURING MAXIMUM VOLUNTARY ISOMETRIC CONTRACTIONS OF QUADRICEPS : THE EFFECT OF KNEE ANGLE

Numerous factors can influence the reliability of the signal obtained by electromyography EMG among which the type of contraction is practically fundamental. This study aimed to investigate intraand inter-session reliability of EMG for maximal voluntary isometric contraction (MVIC) of the knee extensors at different joint angles. Nine healthy male students from the Faculty of Sport and Physical Education voluntarily joined the experiment. Main characteristics of the subjects were: age 23±1 years, body weight 80.8±7.8 kg and body height 182±7 cm. EMG signal from three surface heads of quadriceps femoris was recorded – vastus lateralis (VL), vastus medialis (VM), and rectus femoris (RF). Subjects had to perform 3 MVIC as fast as possible, as strong as possible at 6 different knee joint angles on 2 occasions separated by 7-8 days. The intra-class coefficient of correlation (ICC) was used to identify relative intraand inter-session reliability, while standard error of measurement (SEM) was used to calculate the absolute reliability of each tested muscle. All muscles showed high intra-session reliability (ICC=0.488-0.988 and SEM=1.38-11.35). VL showed good inter-session reliability for most of the conditions (ICC=0.603-0.948), VM for two knee angles (ICC120°=0.764 and ICC130°=0.788), RF was not reliable for any knee angle. This study indicates that RF does not need to be used in EMG testing in leg extension MVIC due to lack of reliability between two sessions, and If EMG RMS from VL and VM will be tested, knee angle of 120° should be used for the testing.


INTRODUCTION
The surface electromyography (EMG) signal represents motor unit's neural activity recorded on the muscle's surface.Activation of the motor unit leads to contraction of every muscle fibre innervated by that motor unit which brings the production of the muscle force.Furthermore, the force produced by muscle partially is a result of the amount and timing of the motor unit activity (Wang and Butchman, 2002;Farina et al., 2014;Enoka and Duchateau, 2015).Numerous extrinsic and intrinsic factors can influence the quality of the signal obtained by the EMG: technical error, equipment error, learning effects, biological variance (Ball and Scurr, 2010;Burden, 2010;De Luca, 1997;Larsson, 2003).To overcome some of the factors and to make EMG signal comparable between the muscles and between tests, the signal should be normalized (Mirka, 1991;Burnett et al., 2007;Dankaerts et al., 2005).Because muscle activation varies with a change in joint angle, comparison of the EMG signal in different muscle lengths should be done with a normalized signal as a relativized level of muscle activity (Mirka, 1991;Merletti et al., 1999).Regarding practical applicability, the use of normalisation method is needed for the reliability evaluation of the EMG signal.
In recent history, different approaches to reliability were reported.Mostly, reliability was investigated for different intensities (Campy et al., 2009;Larsson et al., 2003;Mathur et al., 2005;Rainoldi et al., 2001;Smith et al., 2012).Campy et al (2009) examined the hamstring muscles at different contraction intensities and found high test-retest reliability coefficients (ICC=0.70-0.92 and SEM=9.72-24.94%)only for the medial hamstring across low to moderately high isometric contraction intensities, 10-60% of maximal voluntary contraction (MVC).In the same study, lateral hamstring muscle showed moderate reliability coefficient (ICC=0.57-0.68 and SEM=13.12-21.35%)only at low contraction intensities, 10-30% of MVC (Campy et al., 2009).Another study demonstrated high reliability (ICC=0.80-0.95 and SEM=9-69) for knee extensor muscles for two sets of 100 dynamic maximum concentric right knee extensions separated by 7-8 days (Larsson et al., 2003).Mathur et al (2005) investigated the reliability of median frequency and amplitude of EMG at 80% and 20% of maximal voluntary contraction held to exhaustion.For both, median frequency and amplitude, the initial, final and normalized EMG showed moderate to high reliability for vastus lateralis (VL), vastus medialis (VM) and rectus femoris (RF) at both contraction levels: ICC = 0.59-0.88for MDF; ICC = 0.58-0.99 for amplitude (Mathur et al., 2005).Similarly, Rainoldi et al (2001) showed a high level of reliability (ICC>70%) for 50 seconds sustained contraction at 50% MVC.In the weight-bearing control study, Smith et al (2012) reported high inter-session reliability of the quadriceps muscles in most of the tested conditions, ICC=0.60-0.94.In the same study authors found similar results for the intra-session reliability because ICC was from 0.63-0.95.
When it comes to intra-session reliability, many studies showed moderate to high level of reliability, depending on the task used (Larsson et al., 1999;Fauth et al., 2010, Smoliga et al., 2010;Smith;2012;Worrel et al., 1998).Larsson et al (1999) aimed to investigate the during-the-day reproducibility by using the protocol that involves 10 maximal dynamic extensions with a one-hour rest.They followed the EMG signal of VL, RF and VM and the results showed a high level of reliability for all muscles, ICC=0.83-0.98 but VL was the most reliable (Larsson et al., 1999).Another study of reliability included 3 different tasks maximal voluntary isometric contraction (MVIC), jump from the height and cutting and the reliability of EMG was very high with all ICC values greater than 0.80 (Fauth et al., 2005).Smoliga et al (2010) investigated the reliability of the EMG in competitive runners.They calculated ICC for 25 consecutive strides and found ICC>0.80 for 7 out of 13 tested muscles where VL and RF had ICC<0.80 with RF more reliable than VL, ICC for RF=0.794 and ICC for VL=0.394(Smoliga et al., 2010).In the study with lateral step-up conditions, Worrel et al (1998) found high intra-session reliability with ICC ranging from 0.89-0.98 for VM and VL and 0.67-0.99 for hamstring and gluteus maximus.On the other side in the same study was reported that inter-session reliability ranged from poor to moderate with ICC ranging from 0.06-0.67 and 0.34-83 for the same muscle groups (Worrel et al., 1998).
Only 2 out of 9 above mentioned studies dealt with both, the inter-and intra-session reliability (Worrel et al., 1998;Smith et al., 2012) and they had different results.Both studies reported good EMG reliability for the intra-session repetitions, but one study reported poor to moderate reliability EMG for inter-session (Worrel et al., 1998) while another study reported moderate to high reliability of the EMG for inter-session design (Smith et al., 2012).It seems that there is a certain amount of the studies that explained reliability of the EMG in various study designs with different tasks.However, to our knowledge, there is a lack of the evidence exist regarding the reliability for the basic relationship muscle activity -muscle length (EMG -L), during MVIC.Also, not many studies dealt with both intra-and inter-session reliability, and there is an existing inconsistency with the results.Thus, more data regarding that matter is needed.Therefore, the purpose of this study was to investigate the effect of knee angle on intra-session and inter-session reliability of the EMG for MVIC.Four hypotheses will be tested: 1) intra-session will show high reliability across all knee angles and muscles; 2) inter-session reliability will have moderate to high level of reliability depending on the knee joint angle, and the muscle followed among which VL will be the most reliable.

METHODS
Since this study included 9 participants and two testing session, this study could be classified as a pilot study conducted by the laboratory principles using cluster sample (Hopkins, 2000).

Participants
Nine healthy male students from the Faculty of Sport and Physical Education voluntarily joined the experiment.Main characteristics of the subjects were: age 23±1 years, body weight 80.8±7.8 kg and body height 182±7 cm.All subjects were involved in physical activity minimum 3 times a week for last 3 years.There was no evidence of any knee injury or neurological disorders in their medical history.All participants were informed about the purpose of the study, and they could leave the study at any moment.The study was approved by the ethical committee of the Faculty of Sport and Physical Education following the Declaration of Helsinki regarding human experimentation.

Equipment
Electrode placement EMG signal from three surface heads of the quadriceps femoris were recorded: VL, VM, and RF using surface Delsys Single differential Surface EMG sensors (Delsys Ink., Boston, MA).Electrodes were placed by Delsys manual guide (Delsys technical note 101: EMG sensor placement).A ground electrode was placed on the hip joint of the opposite leg.The skin was shaved at the location of the electrode placement, cleaned with alcohol, and lubricated with ultrasound gel for better electrical conductivity.The electrodes were placed on the most prominent point of the muscle belly following Delsys manual guide (De Luca, 2002).Electrode positions were marked on the skin by the waterproof marker so the day to day position can be controlled.To avoid cross-talk, electrodes were placed at the minimum of 3 cm from each other and tightly fixed with medical tape around the leg and over the electrodes, so there is no movement of the electrodes.Additional tapes were used to fix the cables.

EMG signal acquisition and processing
Single differential multichannel EMG amplifier of 1000Hz was used to register EMG signal.Sampling frequency was 2000Hz, and A/D converter with 12bit precision in the voltage range of ±0.002V.
Original Delsys software (Delsys EMG works 4.1) was used for signal processing.To estimate muscle activity level, root mean square (RMS) calculation was used.According to the De Luca (1997), RMS is a measure of the power of the EMG signal.Window length was 100ms with 80ms of overlapping.For further statistical analysis, first 2 seconds of each contraction were normalized relative to highest maximal voluntary contraction (MVC).The beginning of the contraction was followed by the fourth channel where foot switcher was connected.Foot switcher was positioned on the front edge of the tibia where the lower leg was in contact with the lever arm.The change in voltage caused by the pressure of the tibia on the lever arm was recorded, and it presented the beginning of the contraction.

Familiarization and settings
Subjects came to the laboratory on 2 occasions: one week before the first testing day and then 3 days before the first testing day to get familiar with the testing procedures.First familiarization session was used for the setting of the isokinetic chair (Kin-Com, Chattanooga Group, Inc. Chattanooga TN) according to subject's anthropometry and to get familiar with the contraction mode.The horizontal position of the of the thigh was fixed for all participants, while the back support and whole chair were adjusted to fix the hip joint position in 90°.The axis of rotation of the dynamometer and the axis of the rotation of the knee joint were aligned by moving the head of the dynamometer.The same principle was used to parallelly align dynamometer lever arm and tibia.Two separate crossbelts were used to fix upper body to the armchair, one belt fixed hips to the chair, and one belt was used to fix the thigh to the chair.Force detection system was positioned 1cm above the lateral malleolus.The chair positions for every knee joint angle were marked and written down.Keeping the same position of the chair for both testing days kept technical error of the measurement at the minimum.Second familiarization session was used for getting more familiar with the contraction mode and for marking the places for electrodes placement.

Testing procedure
Before the testing, all participants did warm up protocol of 10 minutes cycling on the bicycle ergometer, followed by 3-5 minutes dynamic stretching.After stretching, electrodes were positioned and with fixed electrodes participants were positioned in already individually prepared Kin-Com dynamometer chair.Subjects had to perform 3 maximal voluntary isometric contractions (MVIC) as fast as possible, as strong as possible in 6 different, randomly selected, knee joint angles.Tested angles were 80°, 90°, 100°, 110°, 120°, and 130° of knee extension (angle between femur and tibia).Duration of each contraction was 3 seconds.Rest period between contractions was 45-60s.After the testing in one joint angle was finished, participants had 3-5 minutes of rest, and during that time, the next joint angle was set.Contraction correctness was controlled by the force output.Every contraction that contained stretch-shortening cycle or lasted less than 3s was repeated until 3 successful contractions were recorded.Subjects could see the force output so they can give their best in every contraction.Also, for better motivation, they were highly supported by the testing team.The identical protocol with the same order of knee angles has been conducted in both testing sessions.

Statistical analysis
All RMS values were written in Microsoft Excel for further analysis.The descriptive statistics for mean and standard deviation was conducted in Microsoft Excel.For reliability analysis software SPSS Statistics 17 was used.Intraclass Coefficient of Correlation with 95% confidence interval (ICC 2,1 two-way random effect) was used to identify within and between-day reliability (Larsson et al., 2003), using p<0.05 level of significance.ICC has been chosen as a measure of relative reliability because it considers between and within-subject variance (Larsson et al., 2003;Mathur et al., 2005;Wier, 2005).ICC values of more than 0.75 represent high reliability, values of 0.60-0.74represents medium reliability, and low or lack of the reliability would be for all values below 0.60 (Smith et al., 2012).For acceptable reliability level, the value of 0.75 was taken as per previous study (Dankaerts et al., 2004).Sometimes a negative ICC can occur which means that the within-subject variance exceeds the between-subject variance (Larsson et al., 2003;Rain-oldi et al., 2001).Variability among subjects must be significant.If that is not the case, limits of the ICC do not match the theoretical limits of 0.0-1.0, and it cannot be considered valid.Practically, it is possible for ratios to range from negative to positive infinity.In that case, the alternative measure of reliability -standard error of measurement (SEM) was used to express the absolute reliability of the measure (Smith et al., 2012).Since there were two testing sessions and three trails for each knee angle, additionally paired-sample T-test was used to determine if differences in normalized EMG magnitudes occurred between testing sessions (Smith et al., 2012).One-way ANOVA was used to test if differences existed between three trails (Larsson et al., 1999).In both statistical analyses, the level of significance was set at a<0.05.The standard error of measurement (SEM) was used to calculate the absolute reliability between two testing sessions and is determined from the standard deviation of the scores.The smaller the SEM, the better result is (Wier, 2005; Mathur et al., 2005).

RESULTS
Simple descriptive statistics shows the differences between test-retest means and standard deviations of normalized EMG signal for each tested muscle.Figures 1-3 show that VL has the smallest differences between two testing sessions comparing to other two tested muscles.The smallest differences in VM occurred at the knee angle of 120° and 130° (Figure 2).In RF only knee angle of 110° seem to have a small difference between means (Figure 3).

Intra-session reliability
Table 1 shows the results of absolute and relative intra-session reliability among 3 consecutive trials.ANOVA detected significant differences between means only in VL and VM at the knee angle of 110° (Sig=0.021for VL and 0.011 for VM).Results for all other tested angles for each muscle did not show any significant difference in trail-trail means.1).Almost all SEM results were in range of 1.38-7.28,except VL at 130° which was slightly higher (11.73).In other words, 95% confidence interval in which the true results of 3 consecutive trails are likely to fall is relatively narrow, which could also be an indicator of the good validity of EMG.Practically, when it comes to trail-trail analysis, use of EMG is a very precise method.

Inter-session reliability
Table 2 shows the results of absolute and relative inter-session reliability.According to T-test statistical analysis, VL mean values from 2 testing sessions are not significantly different for all tested knee joint angles with the highest p-value at 110° (p=0.853) and smallest at 100° (p=0.163).VM and RF also did not show a significant difference between means from two testing session except RF at the angle of 80° (p=0.006).The smallest difference for VM was found at the knee angle of 120°, and for RF at 110° (p=0.869).In terms of consistency, VL showed the most consistent reliability relative to knee joint angles, while in terms of the best reliability with the smallest difference between means VM showed the highest p-value at 120° of knee angle.Overall, T-test showed that VM had the lowest p values among 3 followed muscles except at the 120° of knee angle where VM was the most reliable muscle.Comparing to T-test, ICC results showed similar results for VL muscle (Table 2).The only joint angle of 80° and 130° were slightly below the criterion value of 0.75 which was chosen as the lowest acceptable level of reliability for this study.In other words, VL showed very consistent reliability across the tested range of muscle length.On the other side, VM and RF differences between two sessions were insignificant in most of the angles in T-test which is not the case with ICC.According to ICC results, VM showed sufficient reliability only at the knee angles of 120° and 130° (ICC=0788 and 0.764, respectively), while RF did not show reliability at all in any of tested knee angles.Additionally, results from SEM showed the lowest values right on the knee angles which were reliable either for VL or VM.The highest reliable SEM value found for VL at the angle of 110° (ICC=0.867,SEM=12.08) and the lowest was at 130° for VM (ICC=0.764,SEM=7.14)(Table 2).In other words, all reliable ICCs were in range of SEM from 7.14-12.08.Comparing to intra-session 95% confidence interval for inter-session is slightly but acceptably wider.However, it should be considered when using EMG analysis in testing sessions divided by one week.

Intra-session reliability
The results of this study showed that intra-session reliability is high for all tested muscles across all joint angles (ICC=0.795-0.988and SEM=1.38-11.71),except RF at 120° (ICC=0.448).Similar results for ICC were reported in study that tested within-day reliability in voluntary isometric contraction (ICC=0.95for VL, 0.97 for VM and 0.96 for RF), jump landings (ICC=0.90for VL and VM, and 0.93 for RF) and cuttings with ICC=0.94 for VL, 0.88 for VM, and 0.78 for RF (Fauth et al., 2010).Additionally, Larsson et al (1999) investigated intra-session reliability on 3 sets of 10 dynamic contractions and found high reliability for VL, VM and RF with ICC=0.89-0.96for the mean of ten contractions, and ICC=0.83-0.97 for the contractions with the highest peak torque.Moreover, their ANOVA results showed no difference among the sets in both cases, which goes in line with our results that showed insignificant differences between means in 3 consecutive trials in most of the cases.Only two p values in our study (VL and VM at 110°) showed significant differences but followed with high ICC which could be due to small sample size and corresponding big standard deviation.Smith et al (2012) found that intra-session reliability is moderate to high for most of the weight bearing force control conditions that were used.The lowest ICC was 0.34 and highest 0.94, but the majority of the results was above ICC of 0.75 with corresponding SEMs of not more than 0.20.Intra-session reliability in running for EMG RMS of VL and RF showed good reliability for RF but insufficient for VL with ICC=0.394 and 0.794, followed with SEM of 0.126 and 0.075 or normalized SEM=34.77and 30.18 (Smoliga et al., 2012).It could be due to large standard deviations of the mean which reflects inter-subject variations due to differences in body composition or differences in motor patterns among individuals.(Smoliga et al., 2010).Furthermore, dynamic activity depends on variations in muscle length and shape, and therefore this affects EMG throughout soft tissue filtering since the distance between electrode placement and the active motor unit is cyclically changing (Smoliga et al., 2010).
Even though there are some variations in absolute and relative intra-session reliability results, our and other studies conducted in dynamic and isometric conditions showed high reliability of EMG for RMS between consecutive trails, when the electrodes once positioned had not been moved.When comparing isometric to dynamic contractions of leg extensor muscles, isometric conditions undoubtedly showed high reliability for all three surface heads.In that regard, the first hypothesis of this study has been shown to be true.

Inter-session reliability
Inter-session reliability is shown be moderate to high for VL at 90°, 100°, 110°, 120° (ICC=0.872-0.948with SEM=7.30-9.09,respectively), and moderate for VM at 120° and 130° (ICC=0.764and 0.788 with SEM=7.14 and 10.01).It was also observed that T-test generally indicated the smallest differences between means of VL compering to VL and RF, which is additional evidence of EMG signal from VL being the most consistent among three tested muscles.The problem with T-test was that it showed non-significant differences between two testing sessions for most of the conditions.More precisely, only RF muscle at one knee angle was shown to have significant differences in means.This could be due to the nature of T-test which measures the straight-line association and not agreement between two measures which is not the appropriate way of estimating reliability (Larsson et al., 2001).That is why this study additionally employed ICC analysis that considers between and within-subject variance for relative reliability, and SEM for the absolute reliability which is determined from standard deviation of the scores from all subjects and ICC reliability coefficient (Larsson et al., 2003;Mathur et al., 2005;Wier, 2005).
Knee angle of 120° of showed highest reliability values for VL and VM while lengthening the muscle further resulted in lowered reliability levels for VM and lack of reliability for VL.Rectus femoris did not show reliable activation on any of tested angles.Mathur et al. ( 2005) reported day to day reliability to be low to very high (ICC=0.68for VL, 0.88 for VM, and 0.91 for RF) during sustained contraction of 20% of MVC.They also reported a decreased level of reliability for RF and increased for VL, when the contraction intensity was set to 80% of MVC (ICC=0.66 for RF, 0.83 for VM, and 0.84 for VL).This was explained by the inverse correlation between force production and force variability -higher force shows higher within-subject variability which may affect day to day reliability (Mathur et al., 2005).In the study of Larsson et al. (2003), the results showed lack of reliability of three surface quadriceps heads when the RMS signal was normalized (ICC=-0.02 for VM, 0.43 for RF and 0.52 for VL).Even it was very low, normalized EMG reliability for VL in this study was higher than for VM and RF (Larsson et al., 2003).However, protocol in this study consisted of 100 dynamic movements where electrode position could be compromised, and it cannot be certain if the activation from the same motor units were recorded in every consecutive contraction.Also, authors used different normalization method which, as they said, should be further investigated (Larsson et al., 2003).Rainoldi et al (2001) reported similar results as ours.They used knee angle of 130° for isometric contraction and for amplitude variable only VL showed reliable activation.The probable cause for lack of repeatability of the activation pattern in VM could lie in the muscle fibre length, in the portion of VM where the electrode should be placed.Vastus medialis oblique is a small part of the VM close to the knee, and those fibres are short.Thus, day to day placement could influence the signal obtained (Rainoldi et al., 2001).The overlapping of the actin and myosin filaments seem to be the most optimal on these muscle lengths as well as the angle of the muscle attachment to the bone.Thus, Golgi tendon organ will not be neither excited nor inhibited.Furthermore, joint surfaces are in such position, so there is no high intra-articular pressure and anterior cruciate ligament is relatively relaxed which again removes excitation and inhibition reflexes that could influence the muscle activation (Gandevia & McKenzie, 1998;McGinty et al., 2000).It could be concluded that these knee joint angles might be the best positions for the acquisition of the EMG signal in quadriceps testing when the leg extension is planned to be used.These muscle lengths allow maximal muscle activation while other factors are minimalized.
Standard Error of Measurement has frequently been used as a measure of absolute reliability (Larsson et 2009) conducted the inter-session reliability for medial and lateral hamstring muscles at different contraction intensities.For every intensity that showed reliable results (ICC=0.77-0.92),SEM values were below 14 (SEM=9.72-13.57).Comparing to mentioned study, our results showed better relative reliability followed with the low corresponding absolute measure of reliability as well.Furthermore, results of Mathur et al (2005) suggest that 80% MVC contraction intensity could be a better choice than 20% MVC for EMG investigations because relative reliability was similar for both intensities, but SEM showed better absolute reliability at 80% MVC.Even though ICC for RF at 20% MVC was higher, its SEM was lower as well as VM's and VL's SEM values (9.8-17.2 for 80% MVC vs 14.8-22 for 20% VMC) (Mathur et al., 2005).Note that most of the moderate to high ICC values were followed by SEM values less than 15, which is also the case in our study.Thus, this might be the evidence of the good validity of the approach and precision of the electrode placement.
It cannot be assumed that two muscles within a region or similar function will be equally reliable.The same holds true for the knee extensor muscles, namely with the VL being more reliable in RMS than other two tested muscles.However, VM also showed good reliability on 120° and 130° of knee joint angle.The discrepancy between reliable and non-reliable muscles could be due to consistency of muscle activation patterns between muscles crossing one joint versus the one crossing two joints.Comparing to the VL and VM, RF is a two-joint muscle which could cause the redundancy of the muscle activation in non-specific contraction conditions, and increased redundancy lowers the reliability.Contrary, natural pattern of the RF requires the adjustment of its activation in accordance with the hip-knee relation, and RF has been shown to doing it very reliably (Mathur et

CONCLUSION
Our study dealt with the effect of muscle length on reliability, and the conclusion could be made that reliability between consecutive trails is not affected by the change of muscle length in any tested muscle.On the other side, when the testing sessions are divided by 7-8 days, change in muscle length does affect the reliability of both, single-joint and multi-joint muscles of quadriceps femoris during leg extension.Furthermore, the most reliable surface head of quadriceps femoris is shown to be vastus lateralis followed by vastus medialis, while rectus femoris did not show any reliability in activation.Thus, the second hypothesis of this study is also proved to be true because the reliability of all tested muscles was affected by knee angle and VL was the most reliable among the tested muscles.Additionally, the angle of 120° of knee extension is shown to be the only angle reliable for both VL and VM which suggests that this angle might be a good choice for the leg extension testing if it includes MVIC and EMG.
This study was a short experimental investigation of possible methodological issues when electromyography is used in leg extension MVIC at different knee angles, and 2 practical advices could be made: 1) RF can be excluded due to lack of reliability; 2) If EMG RMS from VL and VM will be tested, knee angle of 120° should be used for the testing.

Table 1 .
Intra-session reliability statistics for all knee angles and muscles.

Table 2 .
Inter-session reliability statistics for all knee angles and muscles.
al., 2003; Campy et al., 2009; Mathur et al., 2005; Smith et al., 2012).Results of Larsson et al (2003) were very different after they normalized EMG signal.When RMS was expressed in µV, RF, VM and VL showed high reliability (ICC=0.89for RF, 0.88 for VM and 0.83 for VL) but the SEM values were very high (SEM=49, 49, 63).In other words, relative reliability was high but absolute was low.On the other side, when RMS was normalized, ICC showed lack of reliability with VL having the highest ICC, but the SEM values were much lower (SEM=17 for RF, 19 for VM, and 16 for VL).Campy et al (