Introduction
Overuse injuries of the lower limb associated with intensive weight bearing exercise are a significant problem for athletes and military recruits, with estimated incidence of running-related injuries reported to range from 20% to 79% [
1]. Lower limb overuse injuries are generally recognised as having multifactorial aetiologies [
2]. Some of the most common injuries, such as Achilles tendinopathy, medial tibial stress syndrome, patellofemoral pain and lower limb stress fractures, are reported to be more prevalent in those with altered foot function [
3],[
4].
The potential mechanisms linking variations in dynamic foot function with lower limb overuse injury may be related to altered lower limb biomechanics and subsequent changes in tissue stress [
5]. This is supported by laboratory-based research using uninjured participants, which suggests that variations in foot posture (flat- and normal-arched feet) are associated with systematic differences in lower limb kinematics [
6]-[
8], kinetics [
4],[
9],[
10], muscle function [
11]-[
16] and tendon morphometry [
17].
While laboratory-based research is important for understanding potential mechanisms linking foot function and lower limb overuse injury, field-based prospective studies are required to determine whether foot function is a risk factor for lower limb overuse injury. Our accompanying systematic review [
18] found that static measures indicating greater foot pronation were associated with an increased risk of patellofemoral pain and medial tibial stress syndrome. However, the small effects suggest that static measures may not adequately represent dynamic foot function. A substantial number of prospective studies have utilised a variety of measurement techniques in order to quantify dynamic foot function and its relationship with lower limb overuse injury [
19]-[
46]. However, it is unclear if there are consistent findings across different measures, or whether particular foot function characteristics are risk factors for specific overuse injuries. Enhanced knowledge regarding this may lead to the development of targeted preventative strategies.
Therefore, the aim of this systematic review was to: (i) identify and appraise the current evidence for the prospective link between dynamic foot posture and lower limb overuse injury; and (ii) provide guidance for future research in this area. This review represents the second component of a two-part systematic review on foot posture-related risk factors for lower limb overuse injury.
Methods
The systematic review protocol was developed in consultation with guidelines provided by the Preferred Reporting of Systematic Reviews and Meta-Analysis (PRISMA) Statement [
47].
Search strategy
MEDLINE, CINAHL, Embase and SPORTDiscus were searched from inception until April 2014. Medical Subject Headings (MeSH) were exploded to include relevant subheadings, in addition to keywords specific to the research question (Additional file
1). The search was limited to adult human participants and English language publications. To ensure identification of all relevant studies, reference lists of appropriate narrative and systematic reviews were hand searched, and discussion with field experts (e.g. physiotherapists, podiatrists) was conducted regarding known important publications. A cited reference search for each included paper was also completed in Google Scholar.
Eligibility criteria
All studies identified by the search strategy were exported to Endnote version X5 (Thomson Reuters, Philadelphia), by a single investigator (GJD). Abstracts and then full text versions were reviewed by two authors (GJD, MMFS) to determine eligibility. Discrepancies were resolved in consultation with a third reviewer (GSM). Initial eligibility criteria were: (i) prospective cohort study design; (ii) quantitative measurement of foot posture or function at baseline (static or dynamic); and (iii) prospective collection of specific or non-specific lower limb overuse injury surveillance data over a specified time period. Specific lower limb overuse injuries were defined as injuries with a single diagnosis, while non-specific lower limb overuse injuries included injuries without a specific diagnosis or where multiple overuse types of injuries were pooled by the study reviewed. After retrieval of studies that fulfilled the initial eligibility criteria, suitable studies were separated into those that investigated dynamic measures of foot function (i.e. measured during walking or running), and those that investigated static measures of foot posture. This review focused on dynamic measures as risk factors, while static measures are addressed in the accompanying review [
18].
Quality assessment
Assessment of the methodological quality of the included studies was performed using the Epidemiological Appraisal Instrument (EAI) [
48]. This instrument is designed to assess the quality of cohort (prospective and retrospective) studies. The EAI consists of 43 items separated into five domains — (i) reporting, (ii) subject/record selection, (iii) measurement quality, (iv) data analysis and (v) generalisability of results [
48]. Items on the EAI were scored as “Yes” (score of 2), “Partial” (score of 1), “No” (score of 0), “Unable to determine” (score of 0) or “Not Applicable” (item excluded). The EAI has demonstrated good/excellent validity, and good to excellent intra-rater (Kappa coefficient range 52 to 60), and inter-rater reliability (Kappa coefficient = 90% [95% CI; 87 to 92%]) [
48]. For the purpose of this review, the wording of all 43 items was modified slightly to improve clarity and rater interpretation. No items were removed or modified, in order to maintain validity (Additional file
2).
Two raters (GJD, NJC) independently evaluated each study while blind to author and publication details. For any discrepancies in assessment of items between the two raters, a meeting occurred and consensus was achieved. To evaluate the overall quality of the studies, average scores across the 43 items were calculated, with a maximum possible score of two (i.e. as individual items are scored ‘0’, ‘1’ or ‘2’, the maximum ‘average’ score across 43 items is two). A ranking system was used to evaluate the quality of evidence, whereby studies were classified as being high (EAI ≥ 1.4), moderate (EAI 1.1 to <1.4), or low quality (EAI < 1.1) [
47].
Data management
Two investigators (GJD, GSM) extracted data regarding study characteristics, including publication details (year, author, country), participant characteristics (number of injured and uninjured, age, sex, inclusion and exclusion criteria, population [i.e. military]) and study methods (dynamic foot function measurement, examiner details, injury outcome, duration of study and covariates investigated). To facilitate calculation of effects, means and standard deviations (SD) were extracted for injured and uninjured participants for continuous foot function variables, while raw counts were extracted for nominal variables.
Where appropriate data was not provided in the publication, authors were contacted with a request to provide additional data. Where studies described specific variables but did not publish data, it was recorded as ‘not reported’ (NR) and, for the purpose of the analysis, assumed that the variable investigated was not significantly different between the injured and the uninjured population.
Statistical methods
Inter-rater reliability of the raters’ EAI scores was evaluated using a descriptive analysis. Differences between rater scores for “Yes”, “Partial”, “No”, and “Unable to determine” were calculated, with a difference of zero indicating perfect agreement and a difference of 1 indicating near perfect. The rating “not applicable” was excluded from analysis because no interpretation was required for this rating.
For continuous foot function variables, standardised mean differences (SMD) were calculated as the difference between injured and uninjured group means, divided by the pooled standard deviation [
49]. SMDs and 95% confidence intervals (CI) were calculated using the ‘Effect Size Calculator’ from the Centre for Evaluation and Monitoring [
50]. Interpretation of the SMD was based on previous recommendations, where > 1.2 was considered large, 0.6 to 1.2 moderate, and < 0.6 small [
51]. For nominal scaled foot function variables, risk ratios (RR) and 95% CI were calculated using the ‘Confidence Interval Calculator’ from the Physiotherapy Evidence Database (PEDro) [
52]. This was represented as the number of participants with lower limb overuse injury in the group with the associated factor (e.g. delayed time to peak force), divided by participants with lower limb overuse injury in the group without the associated factor. A RR > 1.0 indicated that the lower limb overuse injury was more likely to be found in participants with the risk factor present. A small effect was indicated by a RR ≥ 2.0, and a large effect ≥ 4.0 [
53]. Effects were considered statistically significant if the associated 95% CI did not contain zero for the SMD, or one for RR.
Evidence-based recommendations
In order to provide recommendations based on statistical findings, while incorporating the methodological quality of included papers, a scale regarding levels of evidence was utilised, based on previous work by van Tulder et al. [
54].
Strong evidence: pooled results derived from three or more studies, including a minimum of two high quality studies that are statistically homogenous; may be associated with a statistically significant or non-significant pooled result.
Moderate evidence: statistically significant pooled results derived from multiple studies that are statistically heterogeneous, including at least one high quality study; or from multiple moderate quality or low quality studies which are statistically homogenous.
Limited evidence: results from one high quality study or multiple moderate or low quality studies that are statistically heterogeneous.
Very limited evidence: results from one moderate quality study or one low quality study.
No evidence: pooled results insignificant and derived from multiple studies regardless of quality that are statistically heterogeneous.
Discussion
This systematic review evaluated current evidence for dynamic foot function as a risk factor for the development of lower limb overuse injuries. From six of the twelve studies included, we found very limited evidence that plantar pressure and kinematic variables representing dynamic foot function are associated with an increased risk of patellofemoral pain, Achilles tendinopathy and non-specific lower limb overuse injury [
35],[
38],[
42]-[
45]. Notably, significant findings reported across the studies had small to moderate effect sizes, and many 95% confidence intervals included zero, indicating non-significant findings.
Plantar pressure patterns associated with patellofemoral pain differed for walking and running gait. Risk factors in walking gait included greater lateral COP displacement and lower maximal displacement of the medio-lateral COP [
42], whereas for running gait risk factors included earlier time to peak force in the lateral heel and greater peak force in the second and third metatarsal region [
43]. While it is difficult to suggest a mechanism linking these plantar pressure differences with the development of patellofemoral pain, Thijs et al. [
42],[
43] speculated that these findings may indicate a resultant reduction in foot pronation during the loading phase of gait, and subsequent reduction in shock attenuation at the foot. This could increase transfer of ground reaction forces to more proximal structures, such as the patellofemoral joint.
Plantar pressure patterns associated with Achilles tendinopathy were evident from one study investigating jogging gait, and included earlier time to peak force in the lateral heel, less posterior COF displacement/more posterior COF position, greater laterally directed force and delayed time to initial contact in the second metatarsal region [
44]. Van Ginkel and colleagues [
44] speculated that these findings may indicate a more lateral foot roll-over following heel strike and diminished forward force transfer from the rearfoot to the forefoot. It is plausible that differences in force transfer across the foot may lead to altered loading of the Achilles tendon and contribute to injury, but this requires further evaluation.
Another consideration is that increased lateral loading at the foot is an adaptive response to proximal mechanics that increase medial lower limb loading. Prospective studies have shown that increased hip adduction during overground running [
46] and increased hip internal rotation when landing from a drop jump [
22] are risk factors for the development of patellofemoral pain. Furthermore, cross-sectional studies have reported deficits in neuromuscular control of the hip in those with patellofemoral pain [
55]-[
61] and Achilles tendinopathy [
62],[
63]. Further research is required to better understand the relationship between proximal and distal mechanics during gait, and risk of overuse injury development.
In contrast to evidence we found regarding plantar pressure, we found very few kinematic risk factors for lower limb overuse injuries. Our search strategy identified only one study that investigated kinematic risk factors for patellofemoral pain, which presented contradictory findings, no prospective studies that investigated kinematic risk factors for Achilles tendinopathy and two studies that reported differences in rearfoot eversion and forefoot abduction as risk factors for non-specific injuries [
35],[
45]. Whilst cross-sectional findings indicate differences in foot kinematics in people with patellofemoral pain [
64] and Achilles tendinopathy [
49], we found a lack of prospective kinematic data to indicate the temporal relationship between foot kinematics and overuse injury. Thus, at this time it is difficult to draw conclusions as to whether altered foot kinematics is a clear risk factor for lower limb overuse injuries.
In addition to necessitating more kinematic studies, consideration needs to be given to the method of measuring foot kinematics. Considering that overuse injuries generally involve cumulative exposure to load, it is plausible that those who develop overuse injuries demonstrate subtle kinematic differences that are not detectable by current kinematic measures. This is supported by previous findings regarding a lack of biomechanical coupling of plantar pressure indices and angular movements recorded between the calcaneus and the tibia [
65]. Further studies are required to increase understanding of this relationship, which could be achieved using more sophisticated three-dimensional and multisegment foot modeling techniques, and more clinically applicable measures of foot function.
Not surprisingly, it was difficult to identify a systematic pattern of plantar loading and kinematic risk factors for the category of ‘non-specific injuries’. For example, significant risk factors were evident for greater lateral and medial directed COP, as well as increases and decreases in pressure-related outcomes in the fifth metatarsal region. While these findings indeed add evidence of a relationship between dynamic foot function and lower limb injury, the nature of the relationship is unpredictable, and likely relates to the variability of injuries evaluated under the term ‘non-specific injuries’. Therefore, with the advancement and availability of diagnostic algorithms and imaging for lower limb injury, future research should avoid pooling all injuries, and instead focus efforts on exploring conditions that are discrete and well-defined. This is likely to enhance identification of injury-specific risk factors.
Interestingly, we found no evidence that dynamic foot function is a risk factor for iliotibial band syndrome or lower limb stress fractures including the foot. Findings from Noehren et al. [
46] indicated that aberrant hip mechanics may be a stronger risk factor for iliotibial band syndrome than dynamic foot function. They reported that increased hip adduction during running, but not rearfoot eversion, was a predictor of patellofemoral pain development in a cohort of 400 female runners [
46]. This is logical given the proposed mechanism of iliotibial band syndrome, where increased tension on the iliotibial band compresses the lateral femoral epicondyle [
46].
The lack of foot specific injuries (e.g. plantar fasciitis, metatarsal stress fracture) associated with dynamic foot function is another unexpected finding. Although Kaufman et al. [
29] reported that dynamic pes planus in shoes, measured as the ratio of midfoot contact area to total contact area, was a significant predictor of lower limb stress fracture (one third of which involved the foot), our effect size calculations were not significant. This is because the authors set significance at 0.10, whereas we used the more conventional alpha of 0.05. Because of the large number of variables evaluated, this is the more conservative approach to reduce the risk of type II error. An earlier study also reported that pronated foot type (i.e. static foot posture) was a significant risk factor for metatarsal stress fractures, while a supinated foot type was a risk factor for tibial and femoral stress fractures [
66]. However, the static x-ray measure of foot type used in this study may not correlate with dynamic foot function. It is plausible that lower limb stress fractures are more a function of bony overload due to the application of external loads, rather than the biomechanical characteristics of the foot. This is in part supported by the use of military cohorts in both studies [
29],[
66]. The influence of dynamic foot function on the development of lower limb stress fractures should be investigated in civilian populations to ascertain this.
Plantar loading variables were the most abundant risk factor identified for lower limb injury, albeit a relatively low risk with small to moderate effect sizes. In terms of the clinical application of these findings, it is difficult to map the plantar pressure risk factors to specific static foot types. De Cock et al. [
67] reported that participants with low arched feet had a more laterally directed COP across the gait cycle. This is consistent with our plantar pressure findings relating to patellofemoral pain and Achilles tendinopathy. Conversely, Wong et al. [
68] investigated the effect of foot morphology on center-of-pressure excursion during barefoot walking. Their findings indicated that more supinated foot types displayed a larger area of lateral COP excursion, and, conversely, more pronated foot types displayed a smaller area of lateral COP excursion. However, these findings were taken over the entire gait cycle, rather than the discrete phases evaluated in the prospective studies included in this review. In light of the volume of studies that use plantar pressure measures to evaluate dynamic foot function, there is a clear need for further studies to investigate methods of transferring plantar pressure information to clinically relevant measures.
Nevertheless, having some limited knowledge of the pattern of plantar loading risk factors may serve to inform the design of new and existing interventions that may redistribute or counter-balance plantar loading patterns observed in people at risk of injury. For example, arch-contoured foot orthoses alter plantar pressure systematically by reducing pressure in the forefoot and heel regions, and redistributing pressure to the midfoot [
69]. With this in mind, there is evidence from pooled data from randomised clinical trials (RCTs) that foot orthoses are effective in preventing lower limb overuse injuries [
70], as well as evidence from high-quality RCTs that foot orthoses reduce symptoms associated with patellofemoral pain [
71]. In the absence of evidence regarding kinematic effects, our findings suggest that foot orthoses may exert their clinical effects by redistributing plantar pressure (i.e. alter the magnitude, location and temporal patterns of reaction forces at the foot-orthosis interface). However, this requires further investigation.
Whilst this review has highlighted specific measures of dynamic foot function that are risk factors for lower limb overuse injuries, there are several limitations to the identification of these risk factors in a clinical practice setting. Firstly, while findings indicate the direction of altered plantar loading that may increase the risk of development of Achilles tendinopathy or patellofemoral pain, there are no reported thresholds of when an individual is deemed at risk (e.g. peak force in forefoot region exceeding 150 N). Future investigations are required to establish clinical guidelines and screening criteria for these risk factors. Secondly, the assessment of plantar pressures and three-dimensional kinematics requires expensive and sophisticated equipment that is not readily available in clinical practice settings, as well as specialised training in performing and processing these measurements. Future studies should investigate the translation of these laboratory-based measures to clinically applicable measures.
There are also limitations associated with the included studies. The majority of studies evaluated foot function while walking or running barefoot, which may limit the generalisability of findings to shod gait. While it is acknowledged that there are limitations associated with measuring plantar pressures and kinematics while wearing shoes, this is the condition that most closely resembles gait during daily and sporting activities. There were also differences between studies in the evaluation of overground versus treadmill gait analysis. As different gait patterns have been observed for treadmill and overground gait [
72],[
73], it may be inappropriate to measure dynamic foot function during treadmill gait in habitual overground runners, and vice versa. This may lead to a discrepancy between dynamic foot function measured during testing, and foot function during cumulative usual activity. A further limitation of this systematic review is that the methodological quality of the majority of included studies was generally poor. This was largely related to inadequate reporting of foot function measures, covariates, and non-significant results. Thus, the findings should be considered with this in mind. In order to enhance the overall quality of research in this field, future prospective studies should comply with published guidelines for minimum standards of reporting [
74].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
GSM, MMFS, BSN, IBG, CB, SEM and NJC conceived the idea for this review. GSM, MMFS, BSN, IBG designed and piloted the search strategy. GJD undertook the search. Title and abstracts were reviewed by GJD and GSM. Quality appraisal was undertaken by GJD and NJC. Study information and data was extracted by GJD and GSM. The manuscript was drafted by GJD, GSM, MMFS, BSN, IBG, CB, SEM, and NJC. All authors have read and approved the final manuscript.