Relative and absolute within-session reliability of the modified Star Excursion Balance Test in healthy elite athletes

Participants

The sample size was calculated based on the methods described by Walter, Eliasziw & Donner (1998). For α = 0.05, β = 0.2, ρ0 = 0.7, ρ1 = 0.8, using three repetitions of each directions, a sample size of at least 80 would be necessary.

Inclusion criteria for this study were: (1) elite athletes aged between 18 and 35 years; (2) athletes that were familiar with the mSEBT from the previous assessments, tested by the same examiner twice during previous pre-season physical assessments, using four practice trials; (3) minimum score for each of the question of the Oslo Sports Trauma Research Center questionnaire—full participation without problems/no training reduction/no performance reduction/no symptoms in the last week (Clarsen et al., 2014); (4) no lower limb injury or surgery for at least 6 months prior to testing; (5) no vestibular disorder; (6) no concussion in the last 2 years.

First and second division soccer, basketball and volleyball athletes were assessed using the mSEBT as a part of the July and August pre-season physical assessments. Participants who met the inclusion criteria and agreed to participate in the study read and signed the informed consent form. A sample of 122 participants was included in the present study.

The following participants’ data were collected: age, gender, height, body mass, body mass index, leg length, sport (Table 1). Leg length was measured in centimeters with participants lying supine, from ANT superior iliac spine to the ipsilateral medial malleolus, with a standard tape measure, on each lower limb.

Procedure

We used the mSEBT protocol described by Gribble & Hertel (2003), with three reach directions—ANT, PM, and PL, considered in relation with the stance leg. All testing was done in the morning, when dynamic postural balance performance is better, as advocated by Gribble, Tucker & White (2007). One investigator recorded the maximal reach distance in each direction in centimeters. If the participant removed their hands from the hips, lifted the stance leg from the middle of the grid, touched heavily the ground with the reaching foot, did not touch the line with the reaching foot, or did not bring back the reaching leg to the starting position, the trial was considered invalid and an additional trial was performed (Gribble & Hertel, 2003; Gribble, Hertel & Plisky, 2012).

Participants performed one practice trial and after a 1-min rest, completed three trials in the specified directions on each leg. A 5-s rest between trials was allowed. We used the results of each of these three trials to calculate the normalized mSEBT scores. The normalized scores were calculated for each direction according to the formula: normalized score (%) = (reach distance cm/limb length cm) × 100 (Gribble & Hertel, 2003). The testing order for each trial was the following: right ANT, left ANT, right PM, left PM, right PL, and left PL. The composite mSEBT score was calculated as the average of the three normalized reach direction scores for each leg, for each trial, using the formula: composite score = (normalized ANT + normalized PM + normalized PL)/3.

Statistical analysis

Statistical analysis was performed using Medcalc version 17.9.7 (MedCalc Software bvba, Ostend, Belgium). Descriptive statistics (frequencies, percentages, means, and standard deviation) were used to describe the study sample. All mSEBT measurements were tested for normality with the Shapiro–Wilk’s test.

Reliability refers to the degree to which a test provides a free-error measurement over repeated trials (Weir, 2005; Riemann & Lininger, 2018). Intraclass correlation coefficient (ICC) with 95% confidence interval was used as a measure of relative reliability. The ICC (3,1), a two-way mixed-effects model with absolute agreement, was calculated to assess intra-rater reliability (Shrout & Fleiss, 1979; Koo & Li, 2016). We considered ICC values less than 0.75 as moderate, values between 0.75 and 0.90 as good, and values greater than 0.90 as excellent (Koo & Li, 2016).

Standard error of measurement (SEM) reports the measurement error in the same units as the test and it is an indicator of absolute reliability. Absolute reliability refers to the precision of the test by estimating the expected error (Riemann & Lininger, 2018). SEM was calculated according to the formula: $\text{SEM}=\text{SD pooled}\times \sqrt{1-\text{ICC}}$ (Atkinson & Nevill, 1998; Weir, 2005). The measurements are more reliable if the SEM values are smaller. (Atkinson & Nevill, 1998). The smallest detectable change at a 95% confidence interval (SDC₉₅) represents the magnitude of real change between measurements necessary to exceed error. Lower values for SDC indicate higher reliability (Hyong & Kim, 2014). We calculated the SDC₉₅, using the formula: ${\text{SDC}}_{95}=1.96\times \sqrt{2}\times \text{SEM}$ (Weir, 2005).

The significance level was set at p < 0.05 for all tests.

Relative reliability

Using only one practice trial in athletes familiar to the test from previous physical assessments, we observed good to excellent relative within-session intra-rater reliability for the mSEBT for normalized and composite scores, with an ICC (3,1) ranging from 0.90 to 0.95. Our data are comparable with those obtained in previous reliability studies, that used four to six practice trials, on smaller groups of healthy subjects (Kinzey & Armstrong, 1998; Hertel, Miller & Denegar, 2000; Plisky et al., 2006; Munro & Herrington, 2010; Hyong & Kim, 2014; Van Lieshout et al., 2016).

When compared with the results of Hertel, Miller & Denegar (2000), our results showed a slightly higher relative reliability for PM (ICC—0.93 for the right leg, and 0.94 for the left leg) and PL directions (ICC—0.94 for the right leg, and 0.93 for the left leg). In their study, Hertel, Miller & Denegar (2000) used a protocol with six practice trials on a sample of 15 soccer players. In ANT direction, Kinzey & Armstrong (1998) reported slightly lower reliability, in a group of 20 healthy subjects, using five reaching trials.

Our results showed a good to excellent intra-rater reliability using only one practice trial in athletes familiar to the test, while Plisky et al. (2006) found only good intra-tester reliability for ANT, PM, and PL directions, using six practice trials (ICC—0.82–0.87) on a sample of 14 basketball player. Our results based on a larger study sample comprising 122 healthy athletes with different specialization (soccer, basketball, and volleyball) were more optimal than those reported by Plisky et al. (2006).

The relative reliability found in the present study was slightly higher for ANT (right leg ICC—0.9, left leg ICC—0.93), PM (right leg ICC—0.93, left leg ICC—0.94), and PL (right leg ICC—0.94, left leg ICC—0.93) directions than the reliability observed by Munro & Herrington (2010) in a group of 22 healthy volunteers (ICC ranging from 0.84 to 0.92 for the normalized scores of all eight directions of the SEBT, with four practice trials).

The current results were similar with those of Hyong & Kim (2014). They reported high intra-rater reliability for all eight SEBT directions using six practice trials, with an ICC (3,1) ranging between 0.88 and 0.96.

Comparable results were observed by Van Lieshout et al. on a study sample of 55 healthy young adults participating in sports at risk for ankle sprains, using the same three directions as we did in our research. Their results showed good to excellent intra-rater reliability for the composite and separate normalized scores of the SEBT for both legs (ICC—0.87–0.93) (Van Lieshout et al., 2016).

Considering that the values of ICC observed in the current study showed a good to excellent intra-rater relative reliability, our hypothesis that using only one practice trial in athletes familiar to the mSEBT could be considered valid.

Absolute reliability

The SEM and SDC₉₅ provide information about absolute reliability; the lower the values are, the higher the reliability is. With only one practice trial in athletes familiar to the mSEBT we found SEM and SDC₉₅ for normalized and composite scores, ranging from 0.91% to 2.86%, and 2.54% to 7.94%, respectively. The measurement error value is also important in order to know within what range a subject’s true score will lie (Munro & Herrington, 2010).

According to the normalized SDC₉₅ values, only in athletes familiar to the mSEBT evaluated using one practice trial, clinicians should consider important differences between trials higher than 2.54%, and 4.61% in ANT direction for the right leg, and the left leg, respectively. In PM direction, differences between trials should be considered important if they are higher than 7.21% and 7.26% for the right leg and the left leg, respectively. In PL direction, important differences between trials should be higher than 7.02% and 7.94% for the right leg and left leg, respectively. For the composite score, important differences between trials should be higher than 4.78%, and 4.64% for the composite score for right and left leg, respectively.

Our results were lower than the values found by Hyong & Kim (2014) in their study, using a protocol with six practice trials, (intra-tester SEM and SDC of 2.41–3.30, and 7.16–8.99).

The SEM and SDC₉₅ recorded in the present study are lower than the values found by Van Lieshout et al. (2016), except the ANT direction for left leg.

The current SEM and SDC₉₅ values for the three directions were lower for the right leg than the values obtained for the left leg, showing a more stable performance in maintaining balance on the right leg.

In contrast to our results, Van Lieshout et al. (2016) found systematic lower SDC values for the left leg, probably because of the protocol and learning-effects. Their protocol started always with the right leg as the stance leg. In order to eliminate this learning-effect we used the following testing order: right ANT, left ANT, right PM, left PM, right PL, and left PL.

Whereas previous reliability studies have used at least four practice trials on smaller groups of healthy subjects, our data were similar or even superior to those previous studies. The reliability values found in our study on a larger sample have significant implications in sports settings. Using only one practice trial in athletes familiar to the test will reduce the physical assessment time, the burden on athletes and on tester, especially in the pre-season screening of teams with high number of sportsmen.

Our study has some limitations. One limitation is represented by the fact that the sample was largely male. Soccer, basketball, and volleyball athletes were tested, so the results could not be extended to other sports. The within-session evaluations and fatigue could influence the accuracy of the results. The SEM and SDC₉₅ were calculated based on measurements taken from healthy, uninjured young elite athletes and are limited to those who had previous experience with the mSEBT. The results of our study open ways for future work directed at the screening of athletes at risk of lower limb injuries.