Foot orthoses for adults with flexible pes planus: a systematic review

The following databases were searched from inception to June 2013: Ovid Medline® (from January 1966 to date); CINAHL (from 1982 to date); Cochrane Central Register of Controlled Trials (CENTRAL) (issue 5, May 2013); Web of Science (from inception to date); SportDiscus (from inception to date); and Ovid Embase (1988 to date). Medical subject headings (MeSH) were exploded and combined with relevant keywords that were truncated where required. The search was limited to adult human subjects with no language restrictions applied. An example search strategy for Ovid Medline is outlined (Additional file 1). Further articles were sought from review of reference lists, conference proceedings and personal communications with content experts (Figure 1).

Two authors independently reviewed titles and abstracts of all studies generated by the search strategy using the inclusion and exclusion criteria. Consensus was reached for included studies through discussion with all three authors.

Studies were included if:

Studies were excluded if the participants had: a history of significant trauma, disease, inflammatory or systemic condition that may affect lower limb function (e.g. diabetes, neurological dysfunction, rheumatoid arthritis); had previously worn orthoses in the past six months; or the risk of bias of the study was deemed unacceptable (Figure 1).

Risk of bias was assessed using the McMaster critical review form – Quantitative studies [36] and descriptive guidelines [37] which is applicable to RCTs, controlled clinical trials and repeated measure intervention trials. The tool has fifteen individual assessment points within eight domains. These are the study purpose, literature review, sample (described, justified, reliable), outcomes (reliable, valid), intervention (described, not contaminated, co-intervention/s avoided), results (statistically significant, appropriate analyses, clinically important), drop outs and conclusions (including clinical implications). The reporting of drop outs was not applicable for repeated measure trials and therefore was excluded from the critical review. Studies were awarded a ‘yes’ or ‘no’ rating for each individual assessment point of the remaining seven domains. A ‘yes’ rating was applied if the study completely fulfilled the criterion and a ‘no’ rating if the criterion was not completely fulfilled. Domains were required to have a ‘yes’ rating for a minimum of 50% of the individual assessment points to be considered acceptable. An a priori decision was set to include only studies deemed acceptable in four or more of the seven domains (Additional file 3). Risk of bias assessment was completed by two reviewers independently and discussions held until consensus.

The Australian National Health and Medical Research Council’s (NHMRC) evidence hierarchy was used to determine the level of evidence for each study with systematic reviews of RCTs considered high (level I) evidence, RCTs good (level II) evidence, pseudo-RCTs (controlled clinical trials) moderate (level III) evidence and repeated measure trials low (level IV) evidence [38].

Data describing the sample characteristics, intervention characteristics, study protocol and adverse effects were extracted by two reviewers independently, from all studies, with consensus on results. Two reviewers independently extracted data with consensus on results. Novel measures, with no independent reliability or validity data, outcomes that were repetitious (within the same study) and those considered outside the scope of this review (extraneous) were excluded by consensus of three reviewers (Additional file 4). Means and standard deviations (SD) for each group were extracted where data were provided or supplied on request [34, 39‐44].

When more than one type or prescription approach of FOs were investigated, each device within the study was allocated simple identification (Device A, B etc.). Device A from one study is not comparable with Device A from an alternative study (Table 1).

Outcomes are reported for pain, rearfoot kinematics, foot kinetics and physical function. All tabled data are displayed in descending level of evidence, followed by descending levels of assessed risk of bias concerns, followed by alphabetical order unless otherwise noted.

Sample size weighted standardised mean differences (SMD) and 95% confidence interval (CI) were calculated using Cochrane Review Manager (V.5) using the difference in mean scores between the FO and no-FO condition divided by the pooled standard deviation. Where the mean and standard deviations were not available for each condition, SMDs were calculated on mean difference divided by standard deviation of the mean difference, multiplied by the square root of two [45]. SMDs were considered statistically significant if the CI did not cross zero. Results are presented in forest plots where negative numbers favour the FO over the no-FO condition. Interpretations of the strength of the SMDs statistics were based on Cohen’s guidelines [46] with small effect ≥ 0.2, medium effect ≥ 0.5 and large effect ≥ 0.8. Statistically significant SMDs of less than 0.2 were considered very small.

Initial searches identified 2,211 studies. Two thousand, one hundred and sixty three (2,163) were excluded based on title and abstract. Forty eight studies were reviewed in full for eligibility, of which 35 studies were excluded based on inclusion and exclusion criteria (Figure 1) leaving 13 included studies (Table 1). No studies were high level of evidence, two studies were good level of evidence (RCTs), one study was moderate level of evidence (controlled clinical trial) and ten studies were low level evidence (repeated measure intervention trials).

All studies had identified risk of bias (Additional file 3). While all included studies clearly stated their study purpose and the level of statistical significance, of the 13 included studies, only one justified the sample size [41]. Four studies had potential bias of co-interventions [41, 42, 47, 48]. Individual studies had a risk of bias within the literature review [34], sample description [43], outcome measure reliability [39], outcome measure validity [49], intervention description [39, 42, 50], intervention contamination [39, 51], appropriate analysis of results [39, 50], clinical importance of results [49] and overall conclusion [34, 39, 49, 51].

In total, 312 participants were included in the 13 studies (Table 1). Most participants were young adults with 10 of the 13 included studies using cohorts aged 40 years or less. One study only recruited females [50]. Zammit and Payne [42] and Johanson et al. [49] recruited participants from clinical practice. Esterman and Pilotto [34] recruited air force cadets during basic training. Zifchock and Davis [38] and Mündermann et al. [44] both targeted recreational runners. Johanson et al. [46] and Hurd et al. [48] did not recruit from a specific population but required their participants to have forefoot varus of at least eight or five degrees respectively. The remaining studies did not report recruiting from specific cohorts. All studies recruited people with pes planus based on static foot posture. All studies involved 50 or less participants. Zifchock and Davis [41] had both high and low arch participants and did not report the low arch group separately. The authors kindly supplied data to isolate the low arch cohort.

No two studies used the same FOs or approach to prescription (Table 1). Eight of the 13 studies involved two or more types of FOs with three of those comparing the same orthoses with different levels of customisation and/or posting (Table 1). Ten standardised the shell type and the level of rearfoot and forefoot posting across the cohort [34, 39, 40, 43, 44, 47, 48, 50‐52]. Two studies independently prescribed post levels, based on foot morphology, using different approaches [41, 49]. One study investigated a combination of customised and prefabricated FOs individually prescribed [42] (Table 1).

All of the included studies used different outcome measures and measurement approaches (Tables 2, 3, 4 and 5). No study reported on adverse effects. The comparisons (SMD) between FO and no FO conditions for all relevant outcome measures are presented in Figures 2, 3, 4 and 5.

Two studies measured pain changes. One study (level II) reported no significant change in pain [34] and the other study (level IV) reported significant improvement in pain scores with FO use [42] (Table 2). Esterman and Pilotto [34] had a small, non-significant SMD for reducing ‘lower limb pain in previous 24 hours’ when comparing their intervention and control groups over eight weeks of basic training in air force recruits (Table 2). Data from Zammit and Payne [42] indicated a large SMD (mean difference of 21.02 points on the foot health status questionnaire (FHSQ) in reducing foot pain following four weeks of FOs use within a clinical cohort (Figure 2).

From the 13 studies included, data were extracted for 59 relevant outcome measures related to the signs (rearfoot kinematics, foot kinetics and physical function) and symptoms (pain) associated with flexible pes planus (Tables 2, 3, 4 and 5). From these 59 outcome measures, 41 results reported across 13 studies were not statistically significant. Of the 13 studies reporting non-significant results, one was a level 11 [34] and 11 were level IV evidence [39‐44, 47‐49, 51, 52]. Eighteen outcomes reported across eight studies were statistically significant (Table 6). Of the eight studies reporting statistically significant results, one was level 11 [48], one was level III [50] and six were level IV evidence [40, 42, 44, 47, 49, 51]. Available data demonstrated that all statistically significant outcomes were a large or medium SMD effect when comparing the FO to the no FO condition (Table 6).

The largest SMD (-4.11, CI -5.24 to -2.98) was observed in the domain of physical function (Table 6) in Otman et al.’s [50] controlled clinical trial that investigated changes in energy cost during treadmill walking, in a female pes planus group, with and without FOs (Table 6). These results favour the FO condition (Figure 5). The next largest SMD (1.13, CI 0.49 to 1.77) was within the domain of kinetics where Redmond et al. [44] investigated changes in the force-time integral at the midfoot (Table 6). This result favoured the no FO condition (Figure 4). This was the only statistically significant SMD effect that favoured the no FO condition (Table 6). Large effects within individual studies were also observed for reducing peak loading rates, pain and peak foot eversion (Table 6).

Each study used a different measure of pes planus and there was no consistency in recruitment criteria among studies. The majority of studies recruited cohorts of convenience based on static pes planus measures. It would seem more appropriate to investigate effectiveness of an intervention within a participant group who have recognised symptoms associated with the condition. One of the two studies that recruited from clinical practice reported large and medium SMD effects on pain and self-reported function [42] (Figures 2 and 5). These outcomes could, arguably, be the most clinically important measures included within this review.

Concerns around the type of FOs employed were noted. The type of FOs selected for investigation in a study was rarely justified and often not described in detail. Ten of the included studies standardised the shell and approach to posting [34, 39, 40, 43, 44, 47, 48, 50‐52], only three studies individualised their approach [41, 42, 49]. A criticism of FOs intervention studies, anecdotally at least, is the justification of the FOs used and the apparent ad hoc approach to prescription options. This may be related to an absence of appropriate prescription guidelines to direct the researcher [53]. Currently there is no evidence to suggest that individually prescribing FOs offers any benefits over standardised devices, however if the FOs investigated in research are not mirroring those used in clinical practice then the evidence may also not be easily translated to clinical situations. The effect of the diversity of FOs used in the studies on the overall results of this review is unknown, and requires a further review with separate analyses. Further research into the impact of different approaches to the FOs used across all foot types is recommended along with the development of appropriate prescription guidelines to ensure future research outcomes are a genuine reflection of clinical practice results.

The symptom of fatigue was not measured in any study and changes in pain levels were only investigated in two studies. This was surprising given that pain and fatigue are assumed to be common drivers for people with flexible pes planus to seek podiatric intervention [1, 22, 54]. Within this review, pain was significantly reduced when FOs were independently prescribed within a clinical cohort (Table 2); however, with no separate control group (or ‘sham device’ group) improvements in pain levels cannot be attributed to the FOs alone and may simply reflect pain resolving over time, a placebo or a Hawthorne effect [46].

Changes in rearfoot kinematics were frequently investigated (Table 3). Changes were noted predominantly in rearfoot eversion during walking with the measured reduction being significant (Figure 3). It is important to note that the actual magnitude of change is small (1.28 to 2.30°) and falls within accepted levels of measurement error [55]. It has also been suggested that this magnitude of effect on rearfoot kinematics is clinically meaningless and a direct link between rearfoot positioning and functioning has yet to be established [56]. Overall, the majority of rearfoot kinematic measures were not significant (Table 3).

Within the domain of kinetics, only two reported outcome measures demonstrated a statistically significant reduction for loading forces, both observed during running trials (Figure 4). Loading forces during walking were increased across the midfoot with the use of FOs and there was not a lateral shift of force demonstrated as expected (Figure 4). Results may have been influenced by both the use of an in shoe insole placed over the top of the orthoses and by the methods adopted for the quantitative analysis within this domain. The assumption was made that the goal of FO therapy was to reduce the overall force and decrease the time the force was applied [25]. Therefore, the increase in the amount and time of force measured across the midfoot was allocated as favouring the no FO condition (Figure 4). In essence, the actual clinical consequences of the changes in measured force and timing reported is not clearly understood [44] and interpretation of results should be viewed accordingly.

The impact on physical function is where the highest level of evidence was found (Table 5). Both medial-lateral sway during quiet standing and energy cost were positively affected with the use of FOs and offer good and moderate levels of evidence. This supports the historical belief that people with pes planus had reduced efficiency in gait, a belief that restricted entry into military service in the Australian, British and US armed forces for both World Wars [2]. Interestingly, while the use of FOs for improved stability in stance is supported within the literature [57, 58], energy cost studies using FOs in other populations do not concur with the outcomes reported within this review. Hennacy [59] concluded that FOs, within a ‘foot problem’ group, induced an initial increase in energy consumption that returned to normal within three months. In a more recent study, Kelly [54] reported that no statistically significant changes were noted in energy cost during, and following, a one hour run with and without FOs in a non-pes planus cohort. Based on the outcomes of this systematic review, further investigations of energy cost and postural control within a flexible pes planus adult population is warranted.

Only a small number of studies were included in this review. All of the included studies had identified risk of bias with the assessment of risk of bias undertaken with a tool that allocated equal weight to all criteria (Additional file 3). Ten of the included studies are low level evidence (level IV) (Table 1) with all 10 being a repeated measure study design and eight of the 10 studies investigating two or more types of FOs. These studies met the required inclusion criteria; however, their inclusion may have affected the conclusions. To manage data levels and reduce errors from the same participants appearing repeatedly in analyses, several outcome measures were excluded on the basis they were considered repetitious or extraneous (Additional file 4), the effect of excluding these data on the results reported remain unknown.