Interventions for increasing ankle joint dorsiflexion: a systematic review and meta-analysis

Electronic databases were searched without date or language delimiters (Embase, Medline, Cochrane and CINAHL) using keyword searches, as follows:

The only search parameter applied was the human delimiter. Titles and abstracts were screened for relevance to the research question, and full text evaluations were performed on potentially relevant studies using predetermined criteria. A hand search of the reference lists of all relevant studies was undertaken to identify further eligible studies. Studies were selected for inclusion in this review based on the following criteria:

Studies were excluded based on the following criteria:

Two authors (RY, AW) independently assessed the included studies against a modified PEDro scale (Table 1) [35]. The two authors met to discuss the PEDro scale rating system prior to undertaking quality assessments in order to ensure clear understanding of assessment criteria. A consensus meeting resolved disagreements between assessors and a third party was available to provide mediation if required. Consensus on all criteria was reached without need for third party mediation.

The modified PEDro scale featured three additional assessment criteria taken from Law et al. [36]. The additional criteria assessed whether the sample size was justified and whether the outcome measures used were both reliable and valid. The assessment scale featured 14 criterion designed to assess the methodological quality of randomised and non-randomised trials. Trials were awarded a ‘yes’ or ‘no’ rating for each criteria and ‘yes’ responses were then summed to produce an overall quality score for each trial assessed. For consistency, any criteria not directly reported by the authors of each trial were considered to have been unfulfilled and subsequently were awarded a ‘no’ rating.

Data extraction was performed by a single investigator (RY), who recorded details regarding study design, sample characteristics, outcome measures, interventions and follow-up periods. In order to calculate effect sizes, means and standard deviations (SD) were obtained wherever possible for each study group. In two studies [15, 25] a smaller reported value for ankle ROM indicated greater dorsiflexion, so in these cases the means were subtracted from 90 degrees prior to analysis to allow for standardisation against other presented results. Standardised mean differences (SMDs) and 95% confidence intervals (CIs) were calculated, based on the difference between treatment and control groups at the longest period of follow-up. Where studies did not include a comparison group, SMDs were not calculated. SMDs were considered statistically significant if the 95% CI did not contain zero, and interpretation of the magnitude of SMDs was based on previous guidelines [37]: small effect ≥ 0.2, medium effect ≥ 0.5, large effect ≥ 0.8. Positive effect sizes indicated greater increases in ankle joint dorsiflexion in the treatment group compared to the control group. Where more than one intervention group was compared, intervention group A was designated to be the reference group for analysis. Random effects meta-analysis methods were used to pool data where study methods were considered to be homogenous. Sensitivity analyses were performed to investigate the influence of differing intervention or assessment techniques. Heterogeneity was quantified using Chi-squared tests (p < 0.10) and the I² statistic described by Higgins et al. [38], which represents the percentage of total variation across studies due to heterogeneity. To investigate potential bias across studies included in the meta-analysis, effect sizes were plotted against study quality score and sample size, and symmetry of these plots was assessed visually. All statistical analyses were performed using Stata version 10 (StataCorp LP, College Station, TX).

The search strategy returned 3,362 studies total from five databases (Figure 1). A further two potentially relevant studies were sourced through expert consultation. After initial screening to exclude irrelevant studies and to remove duplicates, 541 studies remained for detailed evaluation. The abstracts were read in order to select those that were directly relevant to this review. Three hundred and fifty-six studies were excluded based on abstract screening, and full text evaluations were performed on the remaining 185 studies. Twenty-three studies including a total of 734 participants satisfied the inclusion criteria for this review. Selected characteristics of included studies are presented in Additional file 1.

The 23 included studies received overall quality scores ranging from 1/14 to 11/14 on the modified PEDro scale. Results from quality assessment are presented in Table 1. Of the studies selected for inclusion, 13 featured a control group [2, 3, 5, 12, 16, 18, 19, 21],[23, 25, 29, 31, 32, 38]. Sixteen of the included studies randomly assigned participants to treatment groups [1‐3, 5, 12, 14‐16, 18, 19, 21, 23],[25, 29, 31, 32], nine studies featured blinding of the participants, assessors or therapists [1, 2, 12, 13, 16, 19, 21, 31],[38] and 13 studies reported on the reliability of the measures used [1, 3, 5, 12, 14, 16, 23, 25],[27, 29‐32]. No studies reported on the validity of measures employed.

Eighteen studies investigated stretching interventions [1, 3, 5, 6, 12, 14, 15, 18],[19, 21‐23, 25, 27, 29‐32] and six of the 18 studies investigated the effects of combining stretching with interventions such as ultrasound therapy or strengthening exercises [14, 19, 21‐23, 30]. Two studies investigated manipulation therapy [13, 16], two studies investigated mobilisation therapy [38, 39] and one investigated soleal trigger point therapy [2]. Four studies did not include a comparison group [6, 13, 27, 39] and a further four studies [19, 31, 32, 39] reported insufficient data for SMDs to be calculated. Data comparing the effectiveness of various interventions, including SMDs and 95% CIs, is presented in Table 2. Table 3 provides a synthesis of the body of evidence for each intervention.

Of the 18 studies that investigated stretching interventions, seven studies did not compare to a control group [1, 6, 14, 15, 22, 27, 30] and two studies presented insufficient data [19, 31]. Another study investigating proprioceptive neuromuscular facilitation stretching was excluded from the meta-analysis due to differing stretching technique [25]. The results of eight remaining studies [3, 5, 12, 18, 21, 23, 29, 32] investigating static stretching interventions were combined by meta-analysis methods. The combined effect size indicated that stretching had a statistically significant effect on increasing ankle joint dorsiflexion (SMD 0.68, CI: 0.40 to 0.97) (Figure 2). There was no statistically significant heterogeneity between pooled studies (Chi-squared 4.32, p = 0.74) (Figure 2). Visual inspection of funnel plots revealed that study quality score and sample size did not appear to bias the findings of studies included in the meta-analysis. Further analyses showed that removing studies from the meta-analysis that had used a non-weight bearing stretching technique, or used a different assessment technique (active versus passive ROM or knee flexed versus knee extended) did not substantially alter the findings of the meta-analysis.

One study demonstrated that proprioceptive neuromuscular facilitation stretching was associated with greater increases in ankle dorsiflexion ROM than static stretching (SMD 1.90, 95% CI: 0.92 to 2.88) [15]. Stretching in combination with warming up (SMD 0.87, 95% CI: 0.18 to 1.55) [30], superficial moist heat (SMD 0.84, 95% CI: 0.18 to 1.50) [21], diathermy (SMD 1.12, 95% CI: 0.05 to 2.18) [23], diathermy and ice (SMD 1.16, 95% CI: 0.12 to 2.20) [23] and heel raise exercises (SMD 0.70, 95% CI: 0.04 to 1.37) [21] were also found to be associated with greater increases in ankle dorsiflexion when compared to controls. In contrast, 12 studies showed no significant differences in ankle joint dorsiflexion between the stretching and comparison groups (SMDs: range −0.84 to 1.19) [1, 5, 6, 14, 18, 19, 22, 25],[27, 29, 31, 32].

Two studies investigated the use of ultrasound prior to stretching compared to stretching alone. One reported significantly greater overall gains in ankle joint dorsiflexion ROM compared to stretching alone (SMD 0.95, 95% CI: 0.27 to 1.58) [21]. The second study found that there was no statistically significant difference between these two interventions. (SMD 0.0, 95% CI: -0.62 to 0.62) [14].

Two studies eligible for inclusion in this review assessed the effect of manipulation therapy on ankle joint ROM. One study found that talocrural joint manipulation did not produce a statistically significant increase in dorsiflexion ROM in asymptomatic ankles when compared with a control group (SMD 0.0, 95% CI: -0.44 to 0.44) [16]. The second study reported insufficient data and was not included in the meta-analysis [13].

Two studies eligible for inclusion in this review assessed the effects of mobilisation therapy on ankle joint dorsiflexion ROM [39, 40]. One study assessing the effects of mobilisation therapy did not report sufficient data to be included in the meta-analysis [40]. The second study found that talocrural joint mobilisation did not produce a statistically significant increase in dorsiflexion ROM when compared to a control group (SMD 0.19, 95% CI: -0.37 to 0.75) [39]. The second study assessing mobilisation therapy reported insufficient data and was not included in the meta-analysis [40].

One included study investigated the effect of trigger point therapy on ankle joint dorsiflexion ROM and reported no statistically significant difference between the intervention and control groups (SMD 0.72, 95% CI: -0.18 to 1.63) [2].