Gait parameters associated with hallux valgus: a systematic review

Comprehensive searches of electronic databases (Medline, Embase, and CINAHL) were conducted by the first author for all years available up to October 2011, without language restriction. A highly sensitive search strategy was used and has previously been reported in detail [1]. Search terms included subject headings specific to each database, as well as keywords including “hallux valgus,” “bunion,” and “foot deformity” with truncation and proximity symbols. The search was limited by a second string of terms, including synonyms relating to cross-sectional, case–control or prospective study designs. Reference lists of relevant publications were hand-searched by the same investigator to retrieve all available studies.

Assessment of study eligibility was performed by one investigator. Titles and abstracts of all records identified by the search strategy were scanned for eligibility using the screening question: “Does the study discuss factors associated with HV?” Eligible full-text articles were then retrieved for detailed evaluation according to the following inclusion criteria: 1) clear definition of HV using angular criteria or categorical rating scale; 2) investigated association between HV and gait parameters; 3) study population of adults free of systemic disease; 4) cross-sectional or longitudinal study design with non-HV comparison group. Translations were obtained for articles published in languages other than English to determine their eligibility for inclusion. Authors were contacted for clarification of study methodology as required.

Included studies were assessed for methodological quality by two independent raters, with any disagreements remaining after a consensus meeting resolved by third party consultation. Title, journal, and author details were removed to de-identify articles prior to rating. Quality ratings were performed using the Epidemiological Appraisal Instrument (EAI) [30], which has been validated for assessment of observational studies. Thirty-one items from the original EAI were used, after removing items specifically relating to interventions, randomization, follow-up period, or loss to follow-up, that were not applicable to cross-sectional observational studies. Items were scored as “Yes” (score = 2), “Partial” (score = 1), “No” (score = 0), “Unable to determine” (score = 0), or “Not Applicable” (item removed from scoring). Scores for all applicable items were summed and an average score was determined, with a maximum possible score of 2 (range 0 to 2). To assess potential publication bias across included studies, visual inspection of funnel plots was conducted with effect sizes plotted against study quality scores, sample size and publication year.

For all included studies, the following information was extracted by one investigator: publication details (author, year, publication type, country), sample characteristics (sampling frame, inclusion criteria, number of HV cases, number of control subjects, age, sex), and study methodology (study design, examiner details, definition of HV, associated factors investigated, reliability of measurement methods). In order to calculate effect sizes, means and standard deviations (SD) were recorded for HV and control participants for continuous variables, and raw counts for dichotomous variables. If a study reported data for subgroups of HV severity (e.g. mild and severe), these subgroups were combined for analysis by calculating a weighted average. Authors were contacted and requested to provide additional data where means and SDs were not provided in the original publication.

Pooling of data by meta-analysis was not performed due to lack of homogeneity of study methods and factors investigated. Where sufficient data was provided for continuous variables, standardized mean differences (SMD) were calculated as the difference between HV and control group means, divided by the pooled standard deviation. Interpretation of SMDs was based on previous guidelines [31]: small effect ≥ 0.2, medium effect ≥ 0.5, large effect ≥ 0.8. Where dichotomous data were reported, risk ratios (RR) were calculated as the number of participants with HV in the group with the associated factor present, divided by participants with HV in the group without the associated factor; thus, HV was considered the “event” for the purposes of calculating RR [32]. A RR of > 1.0 indicated that HV was more likely to be found in subjects with the associated factor present. Interpretation of RRs followed previous guidelines, with a small effect represented by RR ≥ 2.0, and a large effect represented by RR ≥ 4.0 [33]. Effect sizes were considered statistically significant if the 95% confidence interval (CI) did not contain zero for SMD or one for RR. Effect sizes and 95% CIs were calculated using Stata Version 10 (StataCorp LP, College Station, TX).

A total of 7833 records were retrieved by the electronic database search. Reference list searches identified another 211 potentially relevant titles. After screening all 8044 titles and abstracts, 532 full text articles relating to HV were examined. Translations were obtained for four studies relating to gait parameters that were published in languages other than English (1 Chinese, 1 Italian, 2 Japanese). Figure 1 outlines studies excluded at each stage of the selection process. After excluding literature reviews, case studies, cadaveric investigations, and studies that did not evaluate gait parameters (k = 498), 23 additional studies were excluded from analysis as they did not adequately define HV (k = 15) [19, 20, 23, 34‐45], did not include a control group (k = 4) [21, 22, 46, 47] or did not perform comparison of HV subjects with controls in their analysis (k = 4) [18, 48‐50]. Comparison with a control group was required to answer the research question being addressed, and a clear definition of HV was essential to enable comparison between studies and to ensure validity of conclusions drawn from this review. Of the eleven studies meeting our inclusion criteria, two reported previously published data [51, 52]; therefore, nine unique studies were evaluated [53‐61].

Inter-rater agreement on the quality appraisal tool was 83.5% (46 disagreements out of 279 quality assessment items rated). Overall quality scores ranged from 0.26 to 1.19 out of a possible score range from 0 to 2 (Additional file 1). The majority of studies (7/9) [53‐58, 60] clearly reported their aims or objectives. All studies adequately reported sample characteristics (age, sex), and seven studies clearly reported subject inclusion criteria [53, 54, 56‐60]. In contrast, no study provided an adequate description of the sampling frame, participation rate or non-responder characteristics. Reporting of statistical methodology was quite poor, with only three studies scoring “Yes” [56‐58], and none providing details of sample size or power calculations. Basic data such as means and SDs were presented for all associated factors in six studies [53‐57, 60]; however, three studies reported insufficient data for some associated factors and therefore scored “Partial” [58, 59, 61]. Only two studies reported effect sizes that represented the magnitude or strength of association [56, 60]. Three studies accounted for covariates such as sex and age, either by matching HV and control groups or statistical adjustment in their analyses [56‐58]. All studies fulfilled our predetermined criterion for defining HV (using angular criteria or categorical rating scale); however, only four studies clearly described a quantitative measurement method for HV (a score of “Yes”) [53, 55, 56, 60]. The remaining five studies provided incomplete details for measurement of HV angle, or used a scale that relied upon visual observation (a score of “Partial”) [54, 57‐59, 61]. Measurement reliability was considered separately for assessment of HV and associated factors. Only three studies reported adequate reliability (coefficient > 0.7) for assessment of HV angle [53, 57, 60], and only two studies reported adequate reliability for all associated factors [53, 60]. Another three studies [56‐58] scored “Partial” for this item as reliability coefficients > 0.4 were reported, or the study made reference to measurement reliability documented in previous literature. Regarding potential publication bias, when SMDs for all associated factors were plotted against study quality scores, sample size and publication year, resulting funnel plots appeared symmetrical, indicating that publication bias was unlikely to have impacted findings from this review.

Additional file 2 presents selected characteristics of the nine studies, including a total of 589 subjects (287 HV, 302 controls). Table 1 outlines age and sex characteristics of the study samples. In six included studies, the HV group comprised predominantly women [53‐55, 59‐61], and one study only included female participants [56]. One study recruited a relatively even ratio of males to females (17:19) [58], while another study did not report male:female ratio [57]. Only one study matched HV and control groups by sex and age [58], although two studies demonstrated that the mean age of their HV and control groups were within three years [56, 60]. Seven studies utilised a case–control study design [53‐56, 59‐61], while two studies used cross-sectional designs and compared participants with moderate to severe HV to those with mild or no HV deformity [57, 58]. The definition of HV varied between studies (see Additional file 2). Six studies used radiographic HV angle to define HV cases [53, 55, 56, 59‐61], two studies used the Manchester Scale [54, 58], and the final study [57] reported visual observation using the following angular criteria: mild (HV angle <15°), moderate (HV angle 15-45°) or severe (HV angle >45°).

Six studies investigated dynamic plantar pressure parameters [53, 55, 56, 58, 59, 61], while five studies investigated spatio-temporal parameters, including angle of gait [54, 56‐58, 60]. One study investigated gait stability, using accelerometers at the head and pelvis to measure the degree of rhythm of subject’s stride patterns [57], and another study investigated intersegmental joint kinematics throughout the gait cycle [54]. Only one included study investigated dynamic electromyographic (EMG) activity of the intrinsic foot muscles [59].

While limited kinematic data are available, Deschamps et al. [54] showed that individuals with HV (n = 20) displayed reduced ankle dorsiflexion during terminal stance compared to controls (n = 22). This observation supports the concept that restricted ankle dorsiflexion may contribute to HV development via an early and increased forefoot loading. There may also be a tendency to compensate by externally rotating the foot, subsequently increasing valgus forces on the hallux [26]. Deschamps et al. [54] also reported less internal rotation of the rearfoot with respect to the tibia during terminal stance, which suggests less rearfoot supination during terminal stance in those with HV compared to controls. This is consistent with the suggestion that late stance phase pronation may contribute to the development of HV via disruption to first ray mechanics [26]. The terminal stance phase is of particular clinical relevance in HV, as this is when the highest ground reaction forces are exerted on the forefoot, and altered alignment of joint axes and moments may lead to pathology [25, 62]. Another kinematic study was excluded from our analysis due to an inadequate definition of HV [35], which may have affected the validity of their study findings. However, it should be noted that Canseco et al. [35] found no significant differences between groups in hindfoot position throughout the gait cycle, although significantly reduced forefoot and hindfoot ranges of motion during certain phases of gait were noted in those with symptomatic HV (n = 33) compared to controls (n = 25). Further studies are warranted to investigate kinematic parameters in HV, particularly kinematics of the first ray [46] and first MTPJ during toe-off [62], as improvements in technology and foot modelling overcome some of the challenges involved with kinematic analysis of the foot.

An important finding of this systematic review is the lack of evidence regarding dynamic muscle function in HV, with only one included study having investigated muscle activity during gait [59]. The main findings of this study showed that individuals with HV had earlier onset of intrinsic muscle activity at heel strike. Abductor hallucis is known to have an important role in supporting the medial longitudinal arch [29, 63], and early activation of intrinsic muscles may be an attempt to stabilise a hypermobile first ray. A further study by Hoffmeyer et al. [39] found abnormal muscle biopsies (abductor hallucis or first dorsal interosseous) in 53 out of 57 HV patients undergoing surgery, as well as abnormal surface EMG recordings during gait in HV patients (n = 19) compared to controls (n = 19); however, this study did not clearly define HV and was therefore excluded from our analysis. There are a number of obvious difficulties associated with recording EMG of intrinsic foot muscles dynamically, including the potential for cross-talk if using surface electrodes, and gait pattern alterations due to discomfort if using fine wire electrodes. It is also difficult to normalise EMG data in populations with HV due to reduced intrinsic muscle strength and the inability of participants to perform meaningful maximum voluntary contractions. Two previous EMG studies [27, 28] have utilised static standing tasks and isometric contractions to quantify muscle imbalance in HV. Both studies found that abductor hallucis became less active in abduction of the hallux and more active in flexion during isometric tasks; however, muscle activity during gait was not evaluated. From the limited data available, it appears that muscle imbalance may be a significant factor in HV. Further studies are warranted to investigate how the timing and magnitude of muscle activity during gait may differ in those with HV compared to controls, particularly muscles related to function of the medial longitudinal arch and first ray (e.g. tibialis posterior, peroneus longus and intrinsic foot muscles). A better understanding of impaired muscle function in HV could guide clinical interventions aimed at retraining muscle activation patterns.

Findings of plantar pressure studies to date are not in agreement regarding forefoot loading in HV. While effect sizes for two studies [53, 58] demonstrate increased pressure under the medial forefoot in those with HV, Komeda et al. [64] report significantly lower first metatarsal pressures. Similarly, while two studies in this review reported increased pressure under the hallux in those with HV [53, 56], previous reports have shown an inverse correlation between hallux plantar loading and increasing HV severity [17, 18, 38]. It should also be noted that some studies found no significant differences in hallux [58] and medial forefoot pressures [56] in HV subjects compared to controls.

Several considerations may help explain these inconsistent study findings. It is plausible that different plantar loading patterns may be found in different stages of HV progression, as soft tissues adapt to forefoot deformity and joint degeneration may develop in the first MTPJ [16]. The presence of foot pain may also lead to inconsistent plantar pressure findings in HV [65], as individuals with first MTPJ pain may adopt strategies to offload the painful area during gait. This tendency to adopt a more cautious or antalgic gait pattern secondary to painful foot deformity has been previously discussed by Crosbie et al. [66]. Although HV subjects were not compared with controls, Morag and Cavanagh [49] discussed several other structural and functional factors influencing loading under the first metatarsal and hallux, including hallux and first metatarsal length, range of motion at the ankle and first MTPJ, and sesamoid height. Future studies investigating plantar pressures in HV should consider severity of deformity, as well as presence of foot pain and other structural factors that may influence plantar pressures.

Another consideration when comparing results between plantar pressure studies is the different systems used for data collection and analysis. Two studies included in this review used the EMED system (Novel, Germany) [53, 58], while other systems included Biofoot (IBV, Valencia, Spain) [56] and F-Scan (Tekscan, South Boston) [55] in-shoe systems, as well as older methods [59, 61]. These systems each use different sensor technologies with varying sensor sizes and responsiveness, which may impact on results [67]. Furthermore, in-shoe plantar pressure analysis results will vary compared to barefoot pressure analysis due to the influence of footwear [67].

Regarding spatio-temporal parameters, angle of gait, or toe-out angle, was not significantly different between HV and control subjects in the two studies that investigated this [58, 60]. These studies provide no evidence to support the proposition that an abducted angle of gait contributes to the development of HV by increasing abduction forces on the hallux during propulsion [25]. However, to date no prospective studies have investigated this parameter and future studies could consider angle of gait due to its theoretical link with HV development. Several studies have shown that basic spatio-temporal parameters do not differ significantly between those with and without HV [54, 56‐58, 60]. However, one study by Menz and Lord [57] found that older adults with moderate to severe HV walked slower along an irregular surface, had a reduced stride length when walking along level and irregular surfaces (Table 3), and demonstrated less stable gait patterns when walking along an irregular surface (Table 4) compared to controls. Therefore, while basic spatio-temporal parameters appear to be largely unaffected in individuals with HV, these parameters may be affected in older individuals with moderate to severe HV, especially during more challenging walking tasks such as on uneven surfaces.

Based on these findings, conservative interventions that target biomechanical foot function during gait are warranted in HV management. Such interventions may include orthoses designed to alter rearfoot motion and assist with efficient forefoot loading and toe-off. Retraining of muscle activitation patterns may be relevant, particularly targeting muscles that dynamically support the medial longitudinal arch. Stretching of the gastrocnemius-soleus complex and manual therapies aimed at improving talo-crural joint motion may facilitate increased ankle dorsiflexion during terminal stance. Clinical trials are needed to investigate the effects of such conservative interventions on gait parameters in populations with HV.

Studies included in this review were somewhat limited in number, as stringent inclusion criteria were used to ensure the validity of overall conclusions drawn. The number of gait studies excluded for methodological reasons highlights the need for future studies to use rigorous study methodology. However, gait parameters investigated by the excluded studies were similar to those reported by included studies, with the majority of excluded studies investigating plantar pressures (k = 22) [18‐23, 34, 36‐50]. As discussed, one excluded study investigated three-dimensional segmental kinematics [35], and one study investigated muscle activity during gait using surface EMG [39]. Since the parameters investigated by excluded studies are similar to those reported by included studies, the authors believe that our systematic review thoroughly summarises the best available evidence regarding gait factors associated with HV.

Our quality assessment revealed several limitations of the available literature investigating gait parameters in HV. First, there was poor reporting of study recruitment methods, making it difficult to assess the generalisability of study results. Furthermore, no studies reported sample size calculations, meaning that null findings might be due to a lack of statistical power required to detect significant differences between groups. The importance of an adequate definition of HV has been discussed previously [1], and was therefore a predetermined inclusion criteria for this review. Differences in HV definition still existed between included studies and may have contributed to inconsistent study findings. Future studies should use a validated approach to HV assessment and diagnosis, such as the Manchester Scale [68] or measurement of HV angle using digital photographs or radiographs [69]. Another pertinent issue is the lack of age and sex-matching of HV and control groups, with only three studies adequately adjusting for both age and sex. Although peak pressures have been shown not to differ between men and women [70], gait parameters including plantar pressures vary significantly with age [50]. Finally, due to the cross-sectional designs utilised by these studies, causal relationships cannot be inferred from this data. Until prospective studies can be conducted in this area, the level of evidence for altered gait parameters in HV should be considered low, and results interpreted with appropriate caution. However, inherent difficulty exists in conducting a prospective study with sufficient follow-up to investigate a slowly progressive deformity such as HV.