2.3.2. Joint Position Sense Test

The SYSTEM 3 PRO dynamometer (Biodex Medical Systems, Shirley, NY, USA) was used to conduct a joint position sense test measuring active and passive repositioning. The dynamometer exhibited medium-to-high reliability in an assessment of joint kinesthesia ability (r = 0.6–0.8) [20]. The participants laid in a supine position while blindfolded, and the involved ankle was placed on the ankle inversion-eversion footplate of the dynamometer with a plantar flexion of 15◦. Three reference angles were established for ankle inversion, neutral ankle position, and ankle eversion (15◦, 0◦, and 10◦, respectively), and the participants were asked to actively and passively reproduce the angle. The absolute value of the joint angle error represents the actual difference between the reference angle and the matching angle [21].

In active repositioning, the involved ankle was first placed in a neutral position and then moved to the inversion or eversion reference angle for 10 s. The participants were asked to actively reproduce the angle three times; the corresponding angles produced by the participants were then recorded. In passive repositioning, the involved ankle was also first placed in a neutral position and then moved to the inversion or eversion reference angle for 10 s. The involved ankle was then passively inverted and everted through a full range of motion by using 5◦/s of angular velocity and stopped at the original reference angle by using a hand-held switch. Three passive repositioning trials were completed to reproduce the reference angle, and the corresponding angles were recorded. The average absolute values of the joint angle errors were subsequently analyzed.

#### 2.3.3. Isokinetic Strength Test

The SYSTEM 3 PRO dynamometer was also used to conduct the isokinetic strength test. The dynamometer exhibited high reliability (ICC coefficients = 0.87–0.96) and was effective in measuring the isokinetic strength of ankle joints [22]. Ankle invertor and evertor muscle strength was measured in terms of concentric contraction (CON) and eccentric contraction (ECC) at velocities of 30◦/s and 120◦/s, respectively [23]. Prior to testing, the participants warmed up for 10 min using general range-of-motion exercises for ankle joints. The participants sat in a chair with the backrest at a seatback tilt of 70◦, and their trunks and pelvises were fixed with straps. The involved leg was fixed with a strap, and the involved foot was secured to ankle attachments with two straps. The tested ankle was positioned with 20◦ of plantar flexion, and the rotational axis of the dynamometer was leveled at the subtalar joint. Three repetitions were performed at velocities of 30◦/s and 120◦/s with 1-min rest intervals between each repetition [24]. The CON and ECC of the ankle inversion and eversion were calculated as peak torque normalized according to body weight. The respective ratios of inversion and eversion for ECC or CON at 30◦/s and 120◦/s were also calculated.

### *2.4. Statistical Analysis*

Statistical analysis was performed using SPSS version 25.0 (SPSS, Chicago, IL, USA). The Shapiro–Wilk test was used to verify the normality of the data to ensure the normal distribution of all assessed variables (*p* > 0.05). Descriptive statistics were used, and all data are presented as the mean ± standard deviation. An analysis of variance (ANOVA) and chi-squared test were used for continuous and categorical variables, respectively, to compare the differences within groups. The results of the SEBT, joint position sense test, and isokinetic strength test were analyzed using a two-way repeated-measures ANOVA (three groups × two times) followed by a Bonferroni post hoc test. Effect size (d) was classified according to the scale of Cohen [25] into very small (<0.2), small (0.2–0.5), medium (0.5–0.8), and large (>0.8), and the effects of Groups A and B were determined to respectively compare Group C. Multivariable linear regression analysis was used to evaluate the association of main outcome measurements between the two training groups and the non-training group. Multivariable modelling was performed with R<sup>2</sup> and β coefficients and specified as

the change values of assessment variables for the prediction. The α level for all statistical analyses was set at 0.05.

#### **3. Results**

Sixty-three female athletes (50 basketball players and 13 volleyball players) with dominant-leg CAI participated following the inclusion criteria in our study. The participants were randomly divided using a random number generator into Groups A, B, and C (all *n* = 21, Figure 1). After the study process, no participants dropped out, and no participants reported adverse reactions. All of the participants completed the study. The demographics are presented in Table 1, and no significant differences were observed within the three groups (all *p* > 0.05).

**Table 1.** Demographics of the participants.


Left, Lt; Right. Rt; Cumberland Ankle Instability Tool, CAIT.

The ANOVA outcomes are summarized in Table 2. In the results of multivariable linear regression (Table 3), the R2 values were 0.80 and 0.93 in Group A and B, respectively. Significant relationships between the change in values in active repositioning and 30◦/s of ECC ankle inversion were noted (*p* < 0.05). Results of the SEBT among the three groups are presented in Table 4. To calculate the composite score of SEBT, the main effects of group (F(2, 60) = 5.30, *p* = 0.03), time (F(2, 60) = 67.78, *p* = 0.001), and time × group (F(2, 60) = 17.84, *p* = 0.001) were observed. Post hoc tests indicated no significant differences in the SEBT within the groups before assessment in terms of composite score and individual directions (*p* > 0.05). Between Groups A and C, the anteromedial (*p* = 0.01, effect size: d = 1.25, 95% CI = 0.59–1.91), posterolateral (*p* = 0.03, effect size: d = 1.05, 95% CI = 0.41–1.70), and lateral (*p* = 0.03, effect size: d = 1.09, 95% CI = 0.44–1.74) directions in the SEBT were significantly different. However, no significant difference was observed in composite scores on the SEBT within the two groups (*p* > 0.05). Moreover, the results indicated that the SEBT composite score and individual directions were higher in Group B than in Group C (all *p* < 0.05). Within the two groups, a small effect size in composite score on the SEBT was observed (d = 2.34, 95% CI = 1.55–3.12), and very small to small effect sizes were observed for all individual directions (anterior, d = 1.70, 95% CI = 1.00–2.41; anterolateral, d = 1.53, 95% CI = 0.84–2.22; anteromedial, d = 1.88, 95% CI = 1.15–2.60; posteromedial, d = 1.21, 95% CI = 0.55–1.86; posterior, d = 1.74, 95% CI = 1.03–2.45; posterolateral, d = 2.13, 95% CI = 1.38–2.89; medial, d = 1.32, 95% CI = 0.65–1.99; and lateral, d = 2.18, 95% CI = 1.41–2.94).


**Table 2.** Outcome of ANOVA with the factors for the assessed variables.

COM: concentric contraction; ECC: eccentric contraction.

**Table 3.** The main outcome measurements in two training groups using linear regression analysis.


COM: concentric contraction; ECC: eccentric contraction.


**Table 4.** Results of star excursion balance test in three group.

\* *p* < 0.05, Group A vs. Group C, between-group using Bonferroni test; <sup>+</sup> *p* < 0.05, Group B vs. Group C, between-group using Bonferroni test.

The changes in active and passive repositioning data before and after assessment for the three groups are presented in Table 5. During active repositioning, for an ankle inversion of 15◦, the main effects of group (F(2, 60) = 3.35, *p* = 0.08), time (F(2, 60) = 12.37, *p* = 0.006), and time × group (F(2, 60) = 0.48, *p* = 0.63) were observed. For the neutral ankle position, the main effects of group (F(2, 60) = 6.59, *p* = 0.01), time (F(2, 60) = 12.26, *p* = 0.006), and time × group (F(2, 60) = 1.58, *p* = 0.25) were observed. For an ankle eversion of 10◦, the main effects of group (F(2, 60) = 0.51, *p* = 0.61), time (F(2, 60) = 8.27, *p* = 0.01), and time × group (F(2, 60) = 1.60, *p* = 0.25) were observed. During passive repositioning, for an ankle inversion of 15◦, neutral ankle position, and an ankle eversion of 10◦, no statistical significance was observed in the main effects of group, time, and time × group (*p* > 0.05).

**Table 5.** Results of joint position sense measurements in three group.


\* *p* < 0.05, Group A vs. Group C, between-group using Bonferroni test; <sup>+</sup> *p* < 0.05, Group B vs. Group C, between-group using Bonferroni test.

For active repositioning data, the within-group analysis of Group A revealed a significant decrease in neutral ankle position (*p* = 0.04) but no significant decreases for an ankle inversion of 15◦ (*p* = 0.16) and an ankle eversion of 10◦ (*p* = 0.13). In Group B, significant decreases for an ankle inversion of 15◦ (*p* = 0.01), neutral ankle position (*p* = 0.02), and an ankle eversion of 10◦ (*p* = 0.01) were observed. Post hoc tests indicated no significant differences within the three groups before assessment. Compared with Group C, Groups A and B exhibited significant decreases for an ankle inversion of 15◦, neutral ankle position, and an ankle eversion of 10◦ (*p* < 0.05). Very small to small effect sizes were observed for an ankle inversion of 15◦ (d = −0.97, 95% CI = −1.61 to −0.33), neutral ankle position (d = −2.18, 95% CI = −2.94 to −1.41), and an ankle eversion of 10◦ (d = −0.95, 95% CI = −1.59 to −0.31) after assessment between Groups A and C. Between Groups A and C, small to medium effect sizes were also observed for an ankle inversion of 15◦ (d = −0.90, 95% CI = −1.54 to −0.27), neutral ankle position (d = −2.20, 95% CI = −2.97 to −1.43), and an ankle eversion of 10◦ (d = −0.89, 95% CI = −1.52 to −0.25) after assessment.

The CON and ECC data before and after assessment for the three groups are presented in Table 6. For 30◦/s of CON ankle inversion, the main effects of group (F(2, 60) = 3.01, *p* = 0.10), time (F(2, 60) = 0.04, *p* = 0.82), and time × group (F(2, 60) = 4.49, *p* = 0.04) were observed. For 30◦/s of ECC ankle inversion, the main effects of group (F(2, 60) = 14.02, *p* = 0.002), time (F(2, 60) = 5.56, *p* = 0.04), and time × group (F(2, 60) = 0.71, *p* = 0.51) were observed. Post hoc tests indicated no significant differences within the three groups for 30◦/s of CON and ECC ankle inversion before assessment, but the number of ankle invertor muscle contractions in Group C was significantly lower than those in Groups A and B for 30◦/s of CON (*p* = 0.01 and *p* = 0.02, respectively) and 30◦/s of ECC (*p* = 0.01 and *p* = 0.001, respectively). Very small effect sizes were observed in 30◦/s of CON ankle inversion after assessment in Group A (d = 1.24, 95% CI = 0.58–1.90) and Group B (d = 1.13, 95% CI = 0.47–1.78) compared with Group C. Small effect sizes were also observed in 30◦/s of ECC ankle inversion after assessment in Group A (d = 1.15, 95% CI = 0.49–1.80) and Group B (d = 1.37, 95% CI = 0.69–2.04) compared with Group C. However, no significant main effects of group, time, and time × group (*p* > 0.05) were observed in 30◦/s of ECC and CON ankle eversion. In addition, for 120◦/s of CON and ECC ankle inversion or eversion, no statistical significance was observed in the main effects of group, time, and time × group (*p* > 0.05). Regarding the analysis of the ratios of inversion and eversion for ECC or CON at 30◦/s and 120◦/s, no isokinetic parameters were statistically significant after using two-way repeated-measures ANOVA (*p* > 0.05).

**Table 6.** Results of isokinetic strength test in three groups.


\* *p* < 0.05, Group A vs. Group C, between-group using Bonferroni test; <sup>+</sup> *p* < 0.05, Group B vs. Group C, between-group using Bonferroni test; concentric contraction, CON; eccentric contraction, ECC.
