Foot characteristics and mechanics in individuals with knee osteoarthritis: systematic review and meta-analysis

Background Foot characteristics and mechanics are hypothesized to affect aetiology of several lower extremity musculoskeletal conditions, including knee osteoarthritis (KOA). The purpose of this systematic review was to identify the foot characteristics and mechanics of individuals with KOA. Methods Five databases were searched to identify relevant studies on foot characteristics and mechanics in people with KOA. Meta-analyses were performed where common measures were found across included studies. Included studies were evaluated for data reporting quality using the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) checklist. Results Thirty-nine studies were included in this systematic review. Two studies reported participants with KOA had statistically significantly (P < 0.05) more pronated foot postures than those without. Meta-analyses for foot progression angle (FPA) and peak rearfoot eversion angle found no difference between those with and without KOA (FPA mean difference:-1.50 [95% confidence interval − 4.20-1.21]; peak rearfoot eversion mean difference: 0.71 [1.55–2.97]). Conclusion A more pronated foot posture was noticed in those with KOA. However, it was not possible to establish a relationship between other foot characteristics or mechanics in people with KOA due to heterogeneity between the included study and limited number of studies with similar measurements. There is need for identifying common measurement techniques and reporting metrics when studying the foot in those with KOA.

to people without KOA. There were two aims of the study: 1) to provide an overview of the foot characteristics and mechanics that have been evaluated in the extant literature in people with KOA, and 2) to investigate whether foot characteristics and mechanics vary between people with and without KOA.

Methods
This systematic review was submitted and approved through the PROSPERO registry of systematic reviews (CRD42015023946), and it followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [12].

Search strategy and study selection
Five electronic databases were searched: MEDLINE, Web of Science, Current Nursing and Allied Health literature (CINAHL), Physical Education Index, and Physiotherapy Evidence Database (PEDro). The searches were conducted in May 2020, with no restrictions by language, year of publication or study design. The Medical Subject Headings (MeSH) search terms adopted were "foot" and "knee osteoarthritis" using the Boolean operator AND.
Studies were evaluated for relevance by applying specific inclusion and exclusion criteria (see Table 1). At the title stage, one reviewer (RA) eliminated publications, with a second reviewer (JLR) verifying the results. At the abstract stage, two reviewers (RA and JLR) independently reviewed abstracts for inclusion, and reference lists of prior KOA review articles were searched to include relevant studies. For manuscripts included following the abstract stage, full-text articles were obtained and independently reviewed for inclusion by reviewers (RA and JLR).

Data extraction
Data from the included manuscripts were extracted (RA) and checked (JLR). For each manuscript, the data extracted was as follows: the country, year of study, sample size, age, gender, body mass index (BMI), diagnostic and inclusion criteria for participants, footwear condition (i.e., barefoot, shod), foot-related outcome measures, and foot-related outcome data. For intervention studies, the baseline data were extracted for analysis. The level of agreement was determined using weighted kappa statistics for inclusion/exclusion.

Assessment of study quality
Study quality of the information reported in the included manuscripts were based on the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) checklist criteria [13], which is a reliable quality rating tool for observational studies [14]. Each criterion was scored "Yes", "No", or not applicable (NA). A criterion received a "Yes" if it was applicable and met in the study, "No" if it was applicable but not met, and "NA" if it was not relevant to the study. The number of "Yes" criterion divided by the number of applicable criterions per manuscript yielded a percentage of the applicable STROBE criteria. Articles were dichotomized by their rating scores, with ≥65% regarded as high-quality studies, and < 65% deemed low-quality. The 65% cut-off point is similar to work conducted by Andrews et al. [15] in dichotomizing high and low quality studies. The 65% cut-off point is lower than the recommended cutoff point of 80% [16] as the reported foot characteristics and mechanics were often not the study's primary outcome measure.

Data analysis
Meta-analyses were performed to estimate the differences between the foot characteristics of participants, with and without KOA, for foot progression angle and peak rearfoot eversion angle. Mean differences (MD) with 95% confidence intervals (95% CI) were calculated. The standard deviation (SD) was extracted or estimated from the standard error of the mean, the 95% CI, P value, or other methods as recommended by the Cochrane Collaboration [17]. Meta-analyses were performed in STATA (16.1) using the 'meta' command.

Criteria Description
Study design Studies with cross-sectional data or intervention data if the baseline data were available.
Study participants Studies were included if they recruited participants with KOA; where a control group was included, they had to be otherwise healthy and free from KOA.
The effect sizes of the meta-analyses are reported in degrees.

Results
Following the implementation of the outlined search strategy, MeSH search yielded 12,736 articles, of which 1837 were duplicate publications ( Fig. 1), leaving 10,899 articles for the title stage. Screening at the title stage excluded 10, 696 of these articles, leaving 203 articles eligible for the abstract stage. At the abstract stage, 43 titles were added from reference lists and other sources, making a total of 246 articles eligible for the abstract stage, and 136 articles were excluded. A total of 110 articles were then reviewed at the full-text stage and 72 articles were excluded, while one article matching the eligibility criteria was added in the full-text stage from other sources, leaving 39 articles found to have evaluated foot characteristics and/or mechanics in individuals with KOA. Kappa agreement values between the reviewers were 0.79, 0.79, and 0.73 for the title, abstract, and full-text stage, respectively.

Participant characteristics Participant age
The mean age of the study participants was 61.5 years, ranging from 47 years [51] to 74 years [50] in the control groups, and 53 years [48] to 75 years [50] in KOA groups ( Table 2).
Foot posture FPI was reported in six studies [18,35,38,47,53,55]. However, the study outcomes were not presented comparably between these studies, limiting the possibilities of meta-analysis. Two studies measured differences in foot posture using FPI in KOA and non-KOA populations [18,38]. Both of them noted that participants with KOA had statistically significant (P < 0.05) highly pronated foot postures, with a difference of 0.78 [18] and 0.61 [38] between the groups (Table 4). Four additional studies measured FPI in individuals with KOA [35,47,53,55], with the results reported here in Table 5 as they were measured differently, with two reporting results as median and interquartile ranges [47,53] and two categorising and reporting the prevalence of individuals into categories. The first study categorised individuals into three categories: normal, pronated, or highly (severely) pronated [55], while the other study added two categories: supinated, and severely supinated [35]. The highest prevalence in both studies was in the pronated foot posture category, with 52% of participants (N = 11) in one study [55] and 37% (N = 30) in the other [35] (Table 5).

Pes planus
Two studies reported on the prevalence of pes planus in individuals with KOA. Pes planus was measured with reference to the medial arch index in one study, and it showed a statistically significant greater prevalence of pes planus in participants with KOA (42% vs. 22%) [18]. Another study measured pes planus by the lateral talometatarsal angle, where it was defined as an angle > 4°, Fig. 2 Forest plot for the differenuihjhjce in FPA during walking between KOA people and healthy controls. 95% CI = 95% Confidence Interval, SD = standard deviation Fig. 3 Forest plot for the difference in peak rearfoot eversion angle during walking between KOA people and healthy controls. 95% CI = 95% Confidence Interval, SD = standard deviation and reported that 38.3% of participants with KOA had pes planus [32].

Other outcomes
Other foot characteristics and mechanics measured in individuals with KOA were divided into two categories and reported in two different tables: static foot variables (Table 5) and dynamic foot variables ( Table 6). The medial arch of the foot was assessed and reported in four studies using different methods (vertical navicular height, navicular drop, and arch index), with different tools (arch index, static footprint, goniometer, and navicular drop test). Of those four studies, two studies compared the results of the KOA group to a control group [38,46]. When participants with KOA were compared to those without, they were found to have a more significant navicular drop (0.03 ± 0.01 vs 0.02 ± 0.01), a significantly greater arch index (0.26 ± 0.04 vs 0.22 ± 0.04) [38], and significantly lower navicular height in sitting (5.22 ± 0.94 cm vs 5.28 ± 0.89 cm) [46] and standing (4.69 ± 0.83 cm vs. 4.73 ± 0.98 cm) [46]. Plantar pressure was measured during walking while wearing plantar pressure sensor insoles embedded inside lab shoes in two studies [50,56]. One study [50] assessed and reported the percentage of partial foot pressure per body part, and reported that plantar pressure was statistically lower in participants with KOA compared to those without KOA in the heel (27.1 ± 11.2% vs. 41.7 ± 8.5%), and hallux (1.5 ± 2.2% vs. 3.5 ± 3.0%), and statistically greater at the midfoot (central) (33.1 ± 11.2% vs. 16.5 ± 13.8%) [50]. In the other study [56], a significantly greater plantar pressure was reported in the midfoot (132.8 ± 28.3 kPa vs. 116.5 ± 30.0 kPa), and the first metatarsophalangeal joint (295.1 ± 100.4 kPa vs. 224.3 ± 62.4 kPa) when compared to a control population [56].
One study [51] investigated the vibratory perception threshold (VPT) in specific foot areas and reported significant deficits in vibratory sensation in participants with KOA. Compared to participants without KOA, those with KOA demonstrated significantly greater VPT in the first metatarsophalangeal joint (15 ± 9.9 V vs. 6.4 ± 3.3 V), medial malleolus (22 ± 11.7 V vs. 12.3 ± 5.2 V), and lateral malleolus (22.3 ± 10.5 V vs. 10.4 ± 3.2 V) [51]. Another study which explored Achilles tendon thickness reported significantly thicker tendons in the KOA group compared to the control [29] (17.1 mm vs. 15.1 mm), with thickness associated positively with KOA severity.

Discussion
The purpose of this review was to evaluate foot characteristics and mechanics in individuals with KOA and compare them to people without KOA where possible. Variations in foot characteristics and mechanics in people with KOA were found in the included studies. These variations included differences in FPA, peak rearfoot eversion angle, pronated foot posture, and incidence of pes planus in people with KOA. Several studies compared foot characteristics and mechanics in individuals with KOA to those without KOA; however measurement techniques and outcome measures were not homogenous across studies. Therefore, meta-analyses were conducted on two foot variables only, FPA and peak rearfoot eversion angle. However, these revealed no statistical difference in FPA or peak rearfoot eversion angle. The results across the included studies were inconsistent, a situation which can be attributed to three main reasons: 1) several studies had no control group without KOA, limiting the ability to report between group differences; 2) studies employed different measurement techniques or methods of reporting, limiting the ability to combine data in meta-analyses; and 3) foot characteristics or mechanics were reported by only one study (e.g., VPT, prevalence of hallux valgus deformity, Achilles tendon thickness), making it impossible to draw robust conclusions. Therefore, further work is needed to fully understand the differences in foot characteristics and mechanics in individuals with KOA.
Results of the present work suggest that the prevalence of pes planus and pronated foot posture is higher among participants with KOA. Zhang et al. (2017) reported significantly greater plantar pressure in the midfoot in those with KOA compared to those without. The increase of midfoot and central plantar pressure aligns with the increased incidence of pes planus [18,32] and greater foot pronation [18,38] associated with KOA. Further, the positive association noted between pes planus and lower vertical navicular height [38] may explain the high pressure in the midfoot area and the absence of a medial longitudinal arch in the foot [50]. The greater peak rearfoot eversion angles evident in individuals with KOA [21,25,31] also align with the reported FPA differences between those with and without KOA [22,34,37,52,54], as these measurements are hypothesised to influence each other biomechanically.
As the included studies measured foot characteristics and mechanics in those with KOA at a single time point, it is unclear if foot posture or incidence of pes planus is a cause or effect of KOA. Nonetheless, the presence of the biomechanical foot differences (pronated foot posture, greater peak rearfoot eversion angle, and incidence of pes planus) associated with KOA highlight the importance of the kinetic chain and biomechanical influence of one joint on another, which may indicate that foot characteristics may be related to KOA progression. However, further longitudinal studies are required to confirm this. As foot posture and foot function have previously been associated with knee joint loading [38,57], a cause of primary progressive KOA [9], it is possible that changing the foot posture or function may be an appropriate intervention for KOA.
Conservative interventions targeting a biomechanical change to address KOA have included foot-related interventions [58,59]. The most common foot-related interventions used to manage KOA are gait modifications and lateral wedge insoles [58]. Toe-out gait has been widely deployed as a conservative intervention in order to reduce knee adduction moment (KAM) and symptoms in people with KOA [59]. Walking with a greater toe-out angle as a mechanical intervention changes the knee joint load in individuals with KOA, shifting the KAM into a flexion moment and reducing knee pain [60]. Furthermore, a greater toe-out degree during walking has been associated with a reduced likelihood of disease progression in participants with KOA for over 18 months [27]. Therefore, this intervention can be limited to targeting people with KOA who walk with a toe-in gait pattern. However, the findings of this systematic review also revealed a diversity in walking patterns among people with KOA (toe-in vs. toe-out gait); thus, this intervention cannot be applied widely in people with KOA.
Lateral wedge orthoses are another common footrelated intervention for KOA [58]. A recent systematic review and meta-analysis demonstrated a reduction in knee joint load, reported as a significant small reduction in first peak of external KAM (standardized mean difference [SMD]: − 0.19; 95% confidence interval [95% CI] -0.23, − 0.15) and second peak external KAM (SMD -0.25; 95% CI -0.32, − 0.19) with a low level of heterogeneity (I 2 = 5 and 30%, respectively) and small but favourable reduction in knee adduction angular impulse during walking in people with KOA (SMD = − 0.14; 95% CI -0.21, − 0.07, I 2 = 31%) [58]. However, the biomechanical changes reported as resulting from lateral wedge orthoses were considered minimal, thus limiting the efficacy of this intervention [58]. Furthermore, the impact of this intervention is still unknown for people with KOA who have pronated foot posture as lateral wedge orthoses were reported to significantly increase subtalar joint valgus moment [61]. Therefore, defining foot characteristics and mechanics in individuals with KOA is extremely important, as doing so can play an essential role in selecting the most appropriate foot-related interventions to fit the individual's own foot characteristics and mechanics.
This systematic review has identified several gaps and areas where future research is needed. Intrinsic foot muscle strength, which affects gait and balance [62], remains an unknown characteristic in the KOA population. Future work evaluating the association between foot muscle strength and KOA may prove beneficial in determining if foot strength or its improvement may be an effective KOA intervention. Further, only one study [51] to date has investigated and reported a loss of vibratory sensation in the foot and ankle with KOA, a measure also affecting gait [63]. Understanding if there is a loss in vibratory sense loss or proprioception as well as how it affects those with KOA may also inform the type of rehabilitation deemed appropriate for this population. It has been suggested that poor neuromuscular control affects injury risk and prevention [64], and neuromuscular control has been associated with KOA severity [65]. Therefore, improving foot neuromuscular control may potentially lessen the risk of knee injury and decrease the impact of KOA.

Strengths and limitations
As with any study, the systematic review and metaanalyses presented here should be evaluated with respect to their strengths and limitations. This review set out a wide range of foot characteristics and mechanics in people with KOA. However, most of the measures were only reported in one or two studies with a small sample of participants, which may limit their generalisability to the wider KOA population. Further, this study has evaluated foot characteristics and mechanics in individuals with KOA and suggested a potential relationship between some of the foot measures and KOA. However, the potential cause and effect relationship of foot characteristics and mechanics outcome measures to KOA is still unknown, as this work has reported foot-related data collected at one time point from observational studies, or data at baseline from intervention studies. Future researchers are advised to investigate the relationship between KOA and foot characteristics and mechanics in more depth via longitudinal studies.
One strength of this study is its robust design, which allowed for the breadth of foot characteristics published to be included in the systematic review and metaanalysis, providing a strong background for researchers to develop longitudinal and intervention studies. However, the wide variety of techniques used to measure similar outcomes prevented the possibility of conducting multiple meta-analyses. Therefore, future studies are advised to develop and follow standardized techniques with which to measure foot characteristics and mechanics in order to facilitate further meta-analyses.
The foot characteristics and mechanics reported in this systematic review were assessed and measured using a range of specific measurements. These could be divided into two categories: 1) laboratory-based measurement (e.g., 3D motion capture, static footprint, force platform, and Biodex); and 2) visual observation and objective manual measurements (e.g., navicular drop test, knee to wall test, FPI, Staheli arch index, and digital caliper). Many of the included studies omitted to provide sufficient details on how the measurements were taken. Moreover, due to the heterogeneity in measurement methods used to investigate foot characteristics and mechanics between the included studies, the process of pooling results for comparison was limited.
One of the limitations identified during this review was the lack of quality in the included studies, as only ten studies attained 65% on the STROBE checklist and could thus be considered high-quality studies. A lower cut-off point of 65% was utilized during the assessment of study quality because foot characteristics and mechanics were not generally the primary outcome measure in the included studies; thus, a cut-off point higher than 65% would not have been achievable by the included studies.

Conclusion
In conclusion, despite the large body of prior research investigating foot characteristics and mechanics in individuals with KOA, many studies lacked a comparison group without KOA. Five foot characteristics and mechanics measures were commonly reported in the included studies (FPA, rearfoot peak eversion angle, peak rearfoot inversion angle, foot posture, and prevalence of pes planus). A more pronated foot posture was noticed in the presence of KOA. Further, of these five common foot characteristics and mechanics, two were of similar design, enabling a meta-analysis to be conducted -FPA and peak rearfoot eversion angle. Meta-analysis of these two variables demonstrated no significant differences between participants with and without KOA. Thus, the implications of the present work suggest a need to adopt and adhere to unified measurement techniques of common foot characteristics and mechanics to make metaanalyses more viable. Lastly, longitudinal studies are needed to identify the potential causal relationship between foot characteristics and mechanics and KOA in people with KOA.