Lower extremity overuse injuries are a significant problem facing active individuals, particularly athletes and military recruits. One common injury is stress fracture; in runners 15−20% of all musculoskeletal injuries have been attributed to stress fractures.1 Stress fractures have been reported at several locations in the lower extremities, but the tibia is the most common site of injury. In runners, tibial stress fractures (TSFs) have been found to account for between 35% and 49% of all stress fractures.2^–4 A similar incidence of 41% has been reported in military recruits.5

It has been hypothesised that TSFs occur when the accumulation of mechanical forces causes the rate of bone resorption to exceed that of bone remodelling and repair.6 7 During running, greater loading in the lower extremities has been associated with an increased risk of TSFs.8 It has been suggested that tibial stress reactions are a precursor to TSFs, with a stress reaction eventually developing into a stress fracture without intervention.9 10 In their study of military recruits, Milgrom et al11 initially reported negative bone scans in three recruits with documented tibial pain. Bone scans taken a month later revealed positive stress fractures. If a stress reaction is a precursor to a fracture, the two injuries are likely caused by the same injury mechanisms. Thus, in the present review, the term “tibial stress injury” (TSI) will include both TSFs and tibial stress reactions. Importantly, other injuries in the same region, including undefined shin splints, medial tibial stress syndrome and compartment syndrome, were not considered under this term or in this review.

It has been suggested that several extrinsic factors are linked to an increased risk of TSI. These include footwear, running surface, weekly mileage, gender, training adaptation and injury history.5 12^–15 However, intrinsic factors have been associated with overuse running injuries in up to 40% of cases.16 One such intrinsic factor is foot morphology.17^–19 The human foot is an extremely complex structure, with foot characteristics varying widely between individuals.20 The current literature regarding foot type and injury is somewhat contradictory. Different foot characteristics have been considered as potential risk factors associated with lower extremity injury.17 18 21 The present review is concerned only with foot type as a risk factor for developing TSI. Different types and locations of injury probably have different injury mechanisms and, therefore, risk factors associated with them. In an attempt to reduce the incidence of TSI, it is important to identify definitively those risk factors specific to this injury mechanism. This is a critical step in the development of preventative measures to help reduce the incidence of TSI among high-risk populations.

The aims of the present review were twofold. The primary aim was to determine by conducting a systematic review of the available literature whether foot type and foot structural characteristics are risk factors for developing TSI. Second, the study undertook to highlight specific areas in which further research is needed.

METHODS

Searching

A search of the following electronic databases was used to identify relevant papers for inclusion in the review: Amed 1985–2006, Cinahl 1982–2006, Index to UK theses, Medline (SilverPlatter) 1950–2006, PubMed 1966–2006, Scopus 1966–2006, Sports discus 1975–2006 and Web of science 1970–2006. The reference lists of review articles were also searched by hand for relevant articles. The search only included articles available in the English language. The following search terms were used: stress fracture, stress injuries, overuse injuries, running injuries, impact injuries, arch height, medial longitudinal arch, high arch, low arch, foot arch, pes cavus, pes planus, anatomical factors, etiological factors, foot type, foot structure and lower extremity alignment. An example of the search strategy used in Medline is outlined in table 1; similar strategies were used when searching other databases.

Study inclusion

Based on the title and abstract, the first reviewer (AB) identified potentially relevant articles and the full papers were retrieved for further review. The first reviewer excluded studies that were not relevant after an initial screening of the full text. In the case of lack of clarity, studies were advanced to the next stage of screening for further examination. The remaining studies were assessed independently by two reviewers (AB, JW) and scored based on six separate inclusion criteria (table 2). The appraisal tool used was developed by the authors, given that validated assessment tools for studies that are not randomised controlled trials do not exist. The criteria were based on those within existing appraisal tools,22 23 as well as key criteria identified as being specific to this review. The appraisal key was based on previous keys developed by The Cochrane Collaboration Injuries Group. The maximum inclusion assessment score available was 18, with 3 representing the maximum, and 1 the minimum score for each question. A scoring system of 1–3 was used for all six questions so as not to weight the scoring towards any one inclusion criteria. If disagreements concerning the scoring of studies occurred, discussion was used as a resolution tool. If agreement could still not be reached, a third reviewer (CM) was available to resolve the issue. The score of each study was converted to a percentage and the assessment criteria of McKay used, whereby a score of 0–49% was classed as poor, 50–89% moderate and >90% good.24 Studies of 50% or above were deemed acceptable for inclusion in this review. Studies that did not define TSF or stress reactions as specific injuries were not included. Further, care was taken not to include studies that grouped TSI with other injuries in the same region. These included shin splints that were not clearly defined, medial tibial stress syndrome and compartment syndrome.

Data extraction and appraisal

Data were extracted using a custom-designed data extraction form. The forms were piloted on a sub sample of the studies and adapted accordingly before standardized data extraction was completed. The form included details of study design, inclusion criteria, participants and aspects of methodology as well as the study results. All data extraction was completed by the first reviewer, a sample of which was checked independently by the second reviewer. Given that none of the included studies were randomised controlled trials, and differed in population and statistical procedures, it was considered inappropriate to carry out a statistical meta-analysis. Further, given the considerable methodological variations between studies, it was felt a meta-analysis would be unable to correct for these confounding factors. Instead, a descriptive account of studies was formulated to characterize this research and identify potential strengths and weaknesses in this literature.

RESULTS

Searches in all databases identified 479 unique studies. Based on title and abstract, 57 of these were identified as potentially relevant and their full texts retrieved. After an initial review, 32 of these were deemed to fall outside the parameters of this review and were excluded. The remaining 25 studies were assessed against the inclusion criteria by both reviewers independently. Of the 25 articles assessed, nine studies achieved inclusion scores of greater than 50% and were therefore included in this review. In line with the assessment criteria of MacKay,24 all included studies were classed as moderate (50–90%), with scores ranging from 56%25 26 to 78%.27 Details of the included studies are presented in table 3.

Of the nine studies, six involved sporting populations,3 25 26 28^–30 and three involved military recruits.15 19 27 The three military studies were limited to male participants, while the athlete studies had cohorts comprising both male and female participants. The three military investigations were all prospective studies, with follow-up periods ranging from 3.519 to 24 months.27 The six athlete studies were retrospective designs.3 25 26 28^–30 The number of TSIs reported ranged from 630 to 157.3 It should be noted that study participants were injury-free at the time of participation, in both prospective and retrospective studies.

Injury diagnosis

Clinical examination by a medical professional, coupled with injury questionnaires, were used to diagnose stress fractures in two of the included studies.15 30 However, Montgomery and colleagues15 validated these methods with confirmation of a positive fracture on a sample of injured athletes using criterion methods. All other studies used the presentation of clinical symptoms confirmed by imaging technologies to diagnose injury.3 19 25 27 29 One study used triple bone scans,3 while another used either nuclear bone scan or radiography to confirm the presence of a TSI.27 Three further studies used a combination of radiography and bone scintigrams as diagnostic tools.19 25 29 The two remaining studies,26 28 did not state explicitly the methods used to diagnose injury, however, the use of “appropriate imaging methods” was reported in one of these articles.26

Foot type classification

There are numerous methods for classifying foot type and this can prove problematic when comparing studies. We found wide variation in both the methods used to classify foot type and the way in which the methods were reported. The classification of foot type across studies ranged from subjective determination to more detailed anatomical measurements. Three of the included trials assessed foot type subjectively through visual inspection of the participants.3 15 26 In two of these, feet were classified as pes cavus, normal or pes planus,3 15 while in the third, arch height was grouped as low, normal or high.26 Another study29 obtained footprints using a podoscopic mirrored table, and classified feet as pes cavus, normal or pes planus based on these observations. Static arch height based on anthropometric measures was used in three of the reviewed articles.19 27 30 Simkin et al19 measured calcaneal angle based on lateral radiographs of the foot. In other studies, external measures of the feet, specifically navicular height27 30 and dorsum height30 divided by foot length were both used as quantitative measures of arch index. Three further studies used foot pressure analysis to provide a measure of foot type.25 27 28 Of these, one study25 used pressure distribution under the tarsal region of the foot to classify feet, while Kaufman et al27 calculated an “arch ratio” defined by midfoot contact area to total foot contact area. A third study28 calculated rearfoot to forefoot angle, using this measure to indicate either a pronated or an open foot type. The three studies using pressure analysis were the only ones to include dynamic analysis of the foot, with two studies analysing pressures during walking,25 27 and one during running.28 Furthermore, only one study incorporated dynamic measures while walking shod.27

Variation between studies was also seen in the methods used to classify foot measures for comparison purposes. In the studies in which subjective grouping of foot type was conducted,3 15 26 29 those classed as having normal feet were used as a reference group for comparison. After measuring arch parameters, two of the studies subdivided the population arbitrarily in two (high/low arches)19 and three (high/normal/low)27 equal groups. These subgroups provided the basis for comparison, with Kaufman et al27 using the normal arched group as a reference for all comparisons. A further study30 made direct comparisons of injury incidence between those with very high and low arches. In this study arch height was determined relative to a normative database. This ensured that the arch height was determined relative to the population, not arbitrarily assigned relative to the sample recruited into the study. Regression comparison of measured foot parameters between injury and control groups formed the basis for analysis in the two remaining investigations.25 28

Foot type as a risk factor for the development of tibial stress injuries

Of the nine studies included, two failed to find any association between foot type and TSI.25 26 One of these studies26 found pes planus and pes cavus feet to be present in 11% and 7%, respectively, of subjects with TSFs. However, foot type distribution could not be compared to a similar uninjured population. A further study25 reported a similar incidence of high arched feet in both injury and control groups (approximately 30%).

Four of the studies presented data suggestive of an increased risk of TSI associated with a more planus or low arched foot.3 15 27 28 Matheson et al3 found that in those with previous TSFs, considerably more of the population had pronated (approximately 53%) compared to cavus (approximately 2%) feet. Another study reported 20% of those with TSFs as having planus feet, while none were classified as having cavus feet.15 Although neither study3 15 conducted a statistical analysis, subjective comparisons suggest an association between low arches and TSFs. Busseuil et al28 found significantly lower static rearfoot to forefoot angles when comparing those with TSI to healthy controls. Similar differences were seen for dynamic measures, and although not significant these results are suggestive of a more pronated foot in those with TSI. One included study27 found evidence that is suggestive of an increased risk of TSFs associated with both planus and cavus feet. Increased injury risk was reported for both extremes of foot type compared to the normal group. These findings, however, were only significant for a pes planus foot type in the dynamic shod condition (risk ratio of 2.45).

Three further included studies also present data that suggest a high arched or cavus foot may increase the risk of TSFs.19 29 30 When classifying arch height based on calcaneal angle, Simkin et al19 reported a TSF incidence of 9.8% in the low-arched group compared to 17.3% in those with high arches. These differences were not found to be significant, but they do suggest an association between high arches and TSFs. In their study, Williams et al30 adopted a different study design based on recruiting runners with very high or very low arches. Despite low subject numbers in the study, high-arched runners reported twice as many TSFs as those with low arches. In addition, Korpelainen et al29 found high arches to be more prevalent in those with stress fractures (40%) than in the control group (13%).

DISCUSSION

The aim of the present review was to determine if foot type is a risk factor in developing TSI. Based on the nine studies reviewed, there is insufficient evidence to indicate a definitive link between foot type and TSI. Limited evidence can be found in support of an increased risk associated with either cavus or planus foot types. The present findings indicate that the measures of foot structure used currently provide at best a limited indication of TSI risk. This finding lends evidence to the multiplicity of risk factors, particularly external factors, which are likely to relate to the development of TSI.

The relationship between arch height and arch flexibility is one that acts on a continuum. However, it has been suggested that high-arched feet tend to be more rigid compared to low-arched feet, which are considered more flexible.31 It is this relationship that forms the basis for one suggested mechanism for an association between foot type and TSI. Typically, the foot pronates in the early period of the stance phase, an action associated with deformation of the medial longitudinal arch.32 The action of pronation has long since been suggested to act as a shock attenuation mechanism, by allowing energy absorption at foot strike.33 34 There is no direct evidence to fully support this hypothesis. However, indirect support was provided by Perry and Lafortune,35 who found that constraining the degree of pronation resulted in greater impact loading during running. Therefore, a flexible low arch with which more pronation is associated may be better able to reduce the impact loading around heel strike than a stiffer high arch. Despite this, recent evidence found only a weak relationship between the parameters of arch height and stiffness.36 While a higher arch did tend to relate more to a stiffer one, only 9% of the variance in arch height could be attributed to stiffness measures.

During routine biomechanical analysis the foot is often represented as a single rigid segment.37 However, during gait a considerable amount of foot motion occurs at the joints of the midfoot and medial longitudinal arch. Significant movement at these joints in both the sagittal and frontal planes has been reported.38^–40 This suggests that a rigid segment approach is inadequate for analysing the complex three-dimensional sequences within the foot. It is now generally accepted that the foot can be modelled successfully as several smaller segments.41 Many authors have proposed multi-segment foot models, which allow multiple joint rotations to be quantified. It is only recently that such models have been applied to clinical applications and problems.42 43 Quantifying the relative motions at the midfoot joints may be of great importance in understanding foot function and assessing foot mobility characteristics in relation to TSI.

Assumptions relating to injury risk are often based on static measurements of the foot. Several of the included studies based their findings on qualitative static foot type,3 15 26 29 while two further studies used quantitative measures taken solely in static postures.19 30 In the present review only three studies incorporated dynamic foot type measurements,25 27 28 and only two of these27 28 measured both static and dynamic characteristics. Evidence provided by Kaufman et al27 suggests a greater associated injury risk with dynamic as opposed to static measures of foot type. It has been suggested that static measures are of little use for inferring relative motions within the foot during dynamic situations.44^–46 This may account for the lack of consensus regarding foot type and TSI risk in the literature. While static measures tell us much about the anatomy of different foot types, they offer little information regarding how these foot types function during dynamic activities. Although there are clear advantages to being able to quantify injury risk via simple static measures, interaction between the foot and the environment may be overlooked. A more complex measure, for example one that incorporates both arch height and arch stiffness, may be more strongly related to TSI risk.

One crucial factor that might account for the conflicting findings of this review is the various methods used to classify foot type. In the present review, four studies used experienced testers to classify foot type in a subjective manner.3 15 26 29 Such static qualitative procedures offer a simple and efficient grouping method, and one that is particularly useful for clinicians.20 However, subjective methods such as these have been shown to introduce a degree of error associated with misclassification.47 48 When grouping feet, clinicians must base their judgement on previous experience of a wider population, which may result in significant variation between testers. Dahle and colleagues48 reported only a 73% agreement between clinicians when classifying feet into three groups, whilst Cowan et al47 observed high inter-tester variability using a five-point grouping scale. Qualitative assignment has the further problem of being potentially skewed towards the grouping of more planus than cavus feet, as they are considered more prevalent within the population. Evidence of this can be found in one included study, which reported more than double the number of feet classified as pes planus (11.8%) compared to pes cavus (5.4%) in the total population observed.15 Results suggesting a relationship between foot type and TSI based on subjective classification should therefore be treated with caution.

Quantitative foot assessment methods have been shown to offer improved measurement reliability.49 However, the classification of foot type based on these measures is crucial to the validity of study outcomes. Ekenman et al25 used plantar pressure patterns from literature sources to assign foot type to three groups. The finding of approximately 30% cavus and no planus feet in both groups suggests skewness in either the sample population or the measurement method.25 Two studies classified foot type by dividing the measured population arbitrarily into groups.19 27 Such approaches, however, may not represent populations beyond that of the study, as classification is not in relation to a wider sample. Williams et al30 deliberately sampled foot type extremes for comparison based on a larger normative database. This is a strength of the study and a factor that may account for the large observed difference in TSF incidence between the groups.

The definition of TSI and the methods used to diagnose them are also important. Diagnostic methods were not limited to imaging technologies in the present review, with three studies using clinical diagnosis techniques.15 28 30 In the present review, care was taken to exclude other leg injuries such as non-specific shin splints, medial tibial stress syndrome and compartment syndrome. While some misdiagnosis may be expected using clinical methods, even the use of conventional radiographs can fail to give an accurate diagnosis. Evidence of stress reactions cannot be observed, with TSFs only evident at an advanced stage of development. Further, in military recruits Montgomery et al15 suggested injury incidence may go under-reported due to the high level of motivation to continue training.

Prospective studies are often considered the best study design for determining the aetiology of injuries, as they allow the mechanics of the lower extremities to be studied prior to injury occurring. When assessing research design, the present review awarded prospective studies higher inclusion scores than retrospective designs. In this review, only three of the studies were prospective type designs.15 19 27 However, it should be noted that some initial evidence suggests retrospective and prospective studies produce similar results when relating anatomical factors to TSFs.50 Further, more confidence can be placed in retrospective studies if we can assume that the anatomical structure and functional mobility of the foot is not affected by TSI. While this may be true for the majority of retrospective designs, in one such study29 subjects with a history of TSFs were recruited from up to 23 years previously. It has been suggested that arches tend to fall with age, and a greater incidence of low-arched feet in older adults has been observed.51 After such a long period post fracture, it is likely that foot characteristics may have changed such that the foot structure being measured is not the same as the foot structure when the injury was sustained.

Implications for future research

This review has highlighted the need for research regarding intrinsic foot parameters as risk factors for TSI. Studies using multi-segment foot models to investigate functional mobility and flexibility within the foot in dynamic situations such as running are essential. Robust quantitative but simple measures of the foot need to be employed as opposed to the more traditional subjective classification methods. In the present review only one investigation studied foot type when shod.27 While the assessment of shod gait may be more ecologically valid than barefoot assessments, it is probably highly dependent on footwear type. Future attention should be given to how the foot functions during dynamic activities and how mobility characteristics interact with external conditions such as footwear. This may enable the development of interventions designed to reduce the risk of TSI within high-risk populations.

CONCLUSION

This review adopted a systematic approach in which strict selection criteria were used to assess the body of literature surrounding foot type and TSI. The outcomes of the nine investigations included were difficult to compare due to varying methods. Results proved conflicting, with limited evidence implicating any one foot type as a potential risk factor. However, it is likely that extremes of foot type (very low and very high arched) pose an increased risk of TSI compared to normal feet. Dynamic measures of foot function may prove to be more useful in predicting the risk of TSI.