FormalPara Key Points

Running shoes with greater shoe cushioning, greater longitudinal shoe stiffness and greater shoe comfort were associated with improved running economy.

Running in light shoes or running barefoot reduced metabolic cost compared with running in heavy shoes but there was no difference in metabolic cost between running in light shoes and running barefoot.

No studies have investigated the effect of footwear on running performance measured using a time-trial or time-to-exhaustion test.

1 Introduction

Selection of appropriate footwear (or lack of footwear, i.e. barefoot) is often advocated as an essential requirement for distance running [1] and as a means for improving running performance [2]. Performance enhancement is also a primary motivating reason that runners try new footwear [3]. However, a systematic search conducted in 2007 found no studies that had investigated the effect of footwear on running performance [4] and there is a lack of consensus amongst the literature on what should be considered appropriate footwear for distance running [4–6]. Despite the lack of research investigating the effects of footwear on running performance, several studies have investigated the effect of footwear on running economy, a surrogate measure of running performance [5, 7]. Running economy is determined from the oxygen demand at a given velocity of submaximal running and is a good predictor of distance running performance [8].

Several different footwear characteristics such as shoe mass, cushioning, motion control, longitudinal bending stiffness, midsole viscoelasticity, drop height and comfort have been proposed to influence running economy and in turn influence running performance [2, 6, 9–14]. Shoe mass has been shown to be important for determining running economy, with additional shoe mass predictably increasing metabolic cost at a given workload [9]. The effect of shoe cushioning on running economy is less clear [13, 15]. Increased shoe cushioning does not always reduce metabolic cost [13] and running barefoot or in minimalist shoes with no cushioning has been shown to be more economical than some cushioned shoes [2, 6]. Indeed, running barefoot or in minimalist shoes that have a flat shoe sole with no cushioning can cause runners to make acute, short-term changes in running gait from a rearfoot strike to a forefoot strike, increase cadence and reduce vertical oscillation of the centre of mass, which can contribute to improved running economy [2, 6].

A review of the current available literature concerning the effects of footwear on distance running performance and running economy is important given the increasing amount of research being published in the area and the need for synthesis of this information to provide direction for ongoing research. There is also an increasing desire on the part of athletes to understand the effects of different footwear [3] and a systematic review could help determine the optimal footwear for distance running. As a result, the aim of this review was to investigate the effect of different types of footwear (heavy, light or minimalist) and footwear characteristics (shoe mass, cushioning, motion control, longitudinal bending stiffness, midsole viscoelasticity, drop height and comfort) on running performance and running economy in distance runners, by reviewing controlled trials that compare different footwear conditions or compare footwear with barefoot.

2 Methods

This review followed the PRISMA statement for improved reporting of systematic reviews [16].

2.1 Information Sources

A literature search was conducted on 5 April 2014. The following databases were searched: Web of Science, Scopus, MEDLINE, EMBASE, AMED (Allied and Complementary Medicine), CINAHL (Cumulative Index to Nursing and Allied Health), SPORTDiscus and CENTRAL (Cochrane Central Register of Controlled Trials). Databases were searched from inception up until April 2014. Searches were supplemented by forward citation searching and hand searching the reference lists of eligible studies.

2.2 Search Strategy

In each database the title, abstract and keyword search fields were searched using the following search strategy:

run* AND shoe* OR footwear OR shod AND performance OR race* OR racing OR marathon* OR time OR distance OR speed OR endurance OR economy OR efficiency OR oxygen OR VO2 NOT orthotic OR pain OR injury

Where possible, limits were placed on searches according to publication type so that only controlled trials, which provide the highest quality of scientific evidence, were included. Additionally, searches were limited to human participant and English language only publications. Eligibility criteria are shown in Table 1.

Table 1 Eligibility criteria
Full size table

2.3 Study Selection

Eligibility and risk of bias assessment were performed independently by two reviewers (JTF and CRB) with disagreement settled by consensus. All records were examined by title and abstract in order to exclude obviously irrelevant records. Full-text articles for the remaining records were retrieved and assessed for eligibility. Data including the publication details, study design, participant characteristics, randomisation, allocation, blinding, testing procedures, description of intervention and results of any analysis of running performance or running economy outcomes were extracted from all eligible studies. If insufficient information was reported (e.g. shoe mass not reported) authors were contacted to seek clarification or additional information about the included studies.

2.4 Risk of Bias Assessment

This review used the Cochrane Collaboration’s tool for assessing risk of bias in controlled trials [17]. Additionally, as all studies identified by the search were of crossover design, the appraisal of bias also considered the appropriateness of using a crossover design and whether appropriate statistical analysis had been performed on the paired data.

2.5 Statistical Considerations

No studies concerning the effect of footwear on distance running performance were identified. As a result, statistical analysis was confined to the effect of footwear on running economy. For each running economy study outcome, standardised mean differences (SMDs) were calculated. Mean differences were standardised using the pooled between-subject standard deviation for the two footwear conditions being compared. Effects were quantified as trivial (<0.2), small (0.2–0.6), moderate (0.61–1.2), large (1.21–2.0) and very large (>2.0) [18].

To investigate the effect of shoe mass on running economy, the SMD for all studies comparing shod (heavy, light or minimalist shoe) and barefoot conditions were plotted against the respective shoe mass, calculated as the combined mass of both shoes per pair. Studies that controlled for shoe mass in the comparison between shod and barefoot were not included because they had controlled for the shoe mass that was the focus of this analysis. The association between running economy and shoe mass was explored using bivariate correlation analysis and linear regression.

Meta-analysis was undertaken for studies that compared running economy between heavy shoes, light shoes and barefoot without controlling for shoe mass across conditions. It was not possible to control for other footwear characteristics (e.g. cushioning, motion control, longitudinal bending stiffness, etc.) when comparing between heavy and light shoes. As a result, statistical analysis considered only the average effect of a heavy or light shoe. Meta-analysis also compared running economy between minimalist and conventional running shoes in studies that controlled for difference in shoe mass and between soft and hard cushioned heavy running shoes. Statistical significance was set at P < 0.05.

Random-effects meta-analysis was performed in Review Manager (RevMan) software (version 5.2, Cochrane Collaboration, Oxford, UK) using the inverse-variance method. Where not reported, the standard error of mean difference and correlations between treatment outcomes were estimated from P values using the equivalent T-statistic or F-statistic. When this was not possible, standard error of mean difference was estimated according to the methods described by Elbourne et al. [19], using the lowest correlation estimate among other studies. Presence of statistical heterogeneity was determined using the I 2 and Cochran’s Q statistics [20].

3 Results

After removal of duplicates, the initial search identified 634 records. An additional six records were identified through hand searching of the reference lists of articles identified in the electronic search. A summary of the search, including number of studies suitable for meta-analysis and reasons for exclusion, is shown in Fig. 1. All 19 studies included in the review were of crossover design and are summarised in Table 2 [2, 5–7, 9–15, 21–28].

Fig. 1
figure 1

Literature search flow chart. n number of studies

Full size image
Table 2 Summary of included studies
Full size table

3.1 Reasons for Exclusion

Five studies were excluded for not using a study sample of distance runners only [29–33], one study was excluded for only comparing running shoe with running shoe plus orthotic [34], two studies were excluded for comparing running shoe with military boots [35] or spring boots [36], three studies were excluded for using short-distance running tests (18, 20 and 60 m) that were not considered representative of distance running performance [37–39], two studies were excluded for measuring running economy while running on an underwater treadmill [40, 41] and one study was excluded for not reporting running performance or economy data [42].

3.2 Risk of Bias

All eligible studies used a crossover design and produced paired data. Only one study [13] did not report use of appropriate pairwise analysis. Thomson et al. [13] reported use of a two-sample t test for analysis of data obtained from repeated measures on the same participants and was at a high risk of detection bias. No studies reported sufficient information regarding randomisation or allocation concealment. No studies provided information regarding blinding of participants, personnel or outcome assessors, and it is unclear what influence lack of blinding would have had on running performance and running economy outcomes. Four studies [9, 23, 24, 26] did not provide sufficient information regarding the number of participants assessed and analysed and were at an unclear risk of attrition bias. Two studies [11, 21] excluded participants from analysis and were at a high risk of detection bias. Nigg et al. [11] excluded one participant from analysis due to unreliable oxygen consumption measurement and Burkett et al. [21] excluded two participants from analysis without providing a reason. It is unclear what effect selective reporting may have had on the results of this review.

3.3 Participants

In all, 243 distance runners had running economy compared between differing footwear conditions. Eleven studies [6, 9–11, 13, 15, 21, 22, 25–27] included male participants only, two studies [2, 23] included female participants only, three studies [5, 7, 28] included both male and female participants, and three studies [12, 14, 24] did not report participant sex. Of the seven studies that considered barefoot running economy, three studies [22, 27, 28] included only participants who were experienced barefoot runners, one study [2] excluded participants with barefoot running experience and in three studies [7, 9, 21] participant experience with barefoot running was unclear.

3.4 Footwear

A variety of footwear conditions and footwear characteristics were compared (Table 2). Five studies [2, 9, 22, 27, 28] compared a light shoe with barefoot, five studies [2, 7, 9, 21, 27] compared a heavy shoe with barefoot, eight studies [2, 5, 6, 9, 23, 25–27] compared a heavy shoe with a light shoe, five studies [2, 5, 6, 25, 27] compared minimalist shoes with conventional shoes, four studies [13, 15, 24, 26] assessed the effect of sole cushioning, one study [14] compared a heavy cushioned shoe with a motion control shoe, one study [10] assessed the effect of shoe comfort, one study [12] assessed the effect of longitudinal bending stiffness and one study [11] assessed the effect of shoe sole viscoelasticity.

3.5 Study Outcomes

No studies provided information concerning the effect of footwear on running performance. All eligible studies provided information concerning the effect of footwear on running economy. All studies expressed oxygen uptake (VO2) relative to body mass. Fifteen studies [2, 6, 7, 9–14, 21, 23, 24, 26–28] expressed VO2 relative to time, three studies [5, 15, 25] expressed VO2 relative to distance and two studies [22, 28] converted VO2 to caloric expenditure. The unit of measure chosen to assess running economy did not appear to affect study findings (Table 2).

The washout period between assessments in different footwear conditions ranged from 2 min to 7 days (Table 2). The length of the washout period was unclear in two studies [13, 23]. All studies, except one [7], assessed running economy during submaximal running bouts on a treadmill. Hanson et al. [7] compared running economy between barefoot and heavy shoe conditions during submaximal running bouts on a treadmill and overground. The authors reported an SMD in running economy for barefoot compared with heavy shoes during overground running that was three times that reported for treadmill running (Table 2). This suggests that footwear might affect running economy differently for treadmill compared with overground running. However, the overground running results reported by Hanson et al. [7] have been challenged in the literature, with concerns about the presence of systematic error in the experimental procedures used to assess overground running economy [43, 44]. Due to the potential difference in running economy outcomes tested on treadmill and overground, only study outcomes assessed on a treadmill were included in the meta-analysis.

SMD in running economy ranged from 0 to 0.79 (Table 2). Two studies [6, 27] reported SMDs for light minimalist shoe compared with heavy shoe that were of moderate effect (0.65–0.79). Two other studies reported SMDs for light shoe [9] and barefoot [2, 9] compared with heavy shoe that were close to moderate effect (0.52–0.56). The remaining studies reported SMDs that were of trivial to small effect.

3.6 Regression Analysis

There was a strong correlation between the combined mass of a shoe pair and change in VO2 relative to running barefoot (R = 0.85, P < 0.01) [7, 9, 21, 22, 27, 28]. The metabolic cost of running increased linearly with increasing shoe mass (Fig. 2). The linear relationship predicted that there would be no difference in VO2 between shod and barefoot running for footwear with a combined shoe mass of 440 g per pair (Fig. 2). Extrapolation of this linear relationship to the hypothetical situation where shoe mass was zero indicated that shoe characteristics other than shoe mass had a theoretical combined small beneficial effect on running economy (SMD = 0.58; Fig. 2).

Fig. 2
figure 2

Change in oxygen uptake for shod running in shoes of different mass compared with barefoot. Shoe mass values are the combined mass of a shoe pair [7, 9, 21, 22, 27, 28]. SMD standardised mean difference, VO2 oxygen uptake

Full size image

3.7 Meta-Analysis

A summary of within-study comparisons and methods used to calculate the individual study standard error of mean difference is shown in Table 3. Results of meta-analysis are shown in Fig. 3. Shoe conditions were grouped into two categories based on shoe mass:

Table 3 Available data and results for 18 outcomes across 14 studies included in meta-analysis
Full size table
Fig. 3
figure 3

Results of meta-analysis. CI confidence interval, n number of participants, SE standard error, SMD standardised mean difference

Full size image

  • Light shoe (combined shoe mass per pair >0 to ≤440 g);

  • Heavy shoe (combined shoe mass per pair >440 g).

The shoe mass of 440 g was selected as the demarcation between heavy and light shoes based on the results of the linear regression analysis, which predicted that footwear with a combined shoe mass greater than 440 g per pair would increase the metabolic cost of running (Fig. 2).

Light shoes (>0 to ≤440 g per pair) and barefoot significantly reduced the metabolic cost of running compared with heavy shoes (>440 g per pair) (P < 0.01), but there was no significant difference between light shoes (>0 to ≤440 g per pair) and barefoot (P = 0.34). When shoe mass was controlled for, minimalist shoes were significantly more economical for running than conventional running shoes (P < 0.01). There was no significant difference between running economy in soft shoes and running economy in hard shoes of the same shoe mass (P = 0.40). However, this result was significantly affected by statistical heterogeneity (P = 0.04, I 2 = 69 %).

4 Discussion

The effect of footwear on running performance is an area of increasing interest, with almost half of the 19 studies identified by the search published in the last 3 years (Table 2). However, despite the increased interest, no studies identified by the search determined the effect of footwear on time or distance measures of running performance. Instead, all studies identified used running economy measured at submaximal running speeds as a measure of running performance. This choice has logical validity given that running performance is dependent to some extent on running economy [8] and running economy is likely to be affected by footwear. Indeed, several authors have previously observed a strong association between running economy and running performance [45, 46]. However, a strong association between running economy and running performance has not always been found [47, 48] and time or distance measures should be considered the reference standard for assessment of running performance.

All studies included in this review used crossover designs and only one study [6] included a longer-term follow-up of 4 weeks. To the authors’ knowledge, there have been no studies investigating what the appropriate washout period for footwear intervention studies is and it is unclear whether the washout periods of several minutes to several days used by studies included in this review were appropriate (Table 2). Additionally, all but one study [6] focused on only the acute, short-term effects of footwear on running economy. Focusing on acute effects ignores potential long-term effects on running economy that may be associated with running in certain footwear over time. This possibility has not been thoroughly tested, but it would seem reasonable to expect that, over time, learning and training effects would occur in response to running in footwear. Indeed, a study measuring running economy in a novel shoe condition with and without a 4-week familiarisation found significant differences between the effect of the novel footwear when tested with and without the familiarisation [6]. Knowledge of these possible learning and training effects would be valuable information for runners who spend extensive time running in different footwear.

A variety of different footwear characteristics were investigated across the studies included in this review (Table 2). The effect of motion control characteristics [14], midsole longitudinal bending stiffness [12], heel viscoelasticity [11] and shoe comfort [10] was trivial to small (SMD = 0.0–0.30) and each was only investigated by an individual study so could not be pooled for meta-analysis. Despite the small effects observed for these characteristics, improvements in running economy were significant for shoes with stiff midsole components (38–45 N·mm) [12] and comfortable shoes [10]. These significant effects were of a magnitude that was classified as trivial (SMD = 0.08–0.12) [10, 12]. However, even these small effects on running economy may be important for high-performance athletes for whom relatively small improvements in performance can have large effects on the outcome of major competitive events [49].

The largest individual study effect sizes were reported by studies investigating the effect of light shoes, minimalist shoes or barefoot on running economy [2, 6, 9, 27]. This suggests that shoe mass is a critical consideration for designing and selecting shoes for use in distance running competition. The importance of shoe mass in determining running economy is intuitive. If one were to consider the simple inertial differences between a heavy and light shoe that must be accelerated with and against gravity with each step taken, it is logical that the reduced muscular effort will lead to improved running economy. Indeed, a positive association between shoe mass and the oxygen cost of running has been previously reported [22] and our linear regression model found a similar positive association (Fig. 2). However, interestingly, our model suggested that the detrimental effect on running economy for shoe mass compared with barefoot was only evident for shoes weighing greater than 440 g per pair and shoes weighing less than this would have a beneficial effect on running economy.

When using 440 g as the demarcation between light and heavy shoes, meta-analysis found light shoes and barefoot to be significantly more economical than heavy shoes (SMD = 0.24–0.34), but found no difference between light shoes and barefoot. The reason that the mass of a light shoe (<440 g per pair) does not have detrimental effects on running economy relative to barefoot, and may even improve running economy, remains untested. It would seem likely that, for footwear weighing less than 440 g per pair, any disadvantage due to having to repeatedly accelerate and decelerate the shoe against gravity might be balanced by the beneficial effects on running economy derived from the shoe cushioning [15], stiffness [12] and comfort [10]. Indeed, our linear regression model predicted that if shoe mass could be zero, then the combined effect of other shoe characteristics would have close to a moderate beneficial effect on running economy (Fig. 2).

Although still considered a light shoe, minimalist shoes differ from conventional running shoes in regards to drop height, sole thickness and toe box structure. Two studies [2, 5] compared the effect of these differences on running economy by controlling for the effect of shoe mass. Meta-analysis of these two studies found a significant small improvement (SMD = 0.29) in running economy for minimalist shoes compared with conventional running shoes (Fig. 3). It has been suggested that the flat, thin-soled minimalist shoes cause runners to increase cadence and adopt a forefoot strike, which in turn improves running economy [6]. However, both studies included in the meta-analysis of minimalist shoes found significant improvements in running economy for minimalist shoes compared with conventional running shoes even when controlling for changes in foot strike and cadence [2, 5]. As a result, there must be further reason for the observed difference in running economy. Perl et al. [5] suggested that flat, thin-soled shoes increase the storage and release of elastic energy in the Achilles tendon and longitudinal arch of the foot. However, to date, this hypothesis has not been thoroughly tested. Nonetheless, when controlling for shoe mass, running in minimalist shoes has a beneficial effect on running economy.

When shoe mass and running gait were not controlled for between minimalist shoes and conventional running shoes, the beneficial effect of minimalist shoes on running economy increased (SMD = 0.12–0.79) [6, 25, 27] (Table 2). This larger beneficial effect could be partly explained by the reduction in shoe mass associated with minimalist shoes compared with conventional shoes but could also be due to changes in foot strike and cadence that have previously been associated with running in a minimalist shoe [6, 27, 50]. It is thought that the heightened somatosensory feedback associated with running in minimalist shoes, which lack cushioning, prompts runners to increase cadence and land with a more anterior foot strike [2]. Indeed, Squadrone and Gallozzi [27] and Warne and Warrington [6] observed SMD improvements in running economy of 0.65 and 0.79 for the minimalist shoe when runners also adopted a forefoot strike [6, 27] and increased cadence [6]. The participants were either experienced barefoot runners [27] or were given 4 weeks to familiarise themselves with the minimalist shoes [6]. Given the size of these effects, future research should further explore the long-term effects of running in minimalist shoes on running economy and biomechanics.

Although running in minimalist shoes may be associated with large improvements in running economy [6, 27], running in these shoes may have negative effects on injury risk [51] and this effect on injury should not be ignored. Running in minimalist shoes is associated with increased peak tibial acceleration [26], which is known to be significantly greater in runners who have sustained a recent tibial stress fracture than in healthy controls [52]. Additionally, changing from a rearfoot strike to a forefoot strike is associated with increased ankle joint contact forces and increased plantar flexor muscle forces [53]. These unaccustomed high forces could increase the risk of injury until the associated muscular and articular tissue has had time to adapt [53]. The long-term safety of minimalist shoes should be investigated before they are advocated as a means for runners to improve running economy.

Meta-analysis was also possible for studies that compared running economy between soft- and hard-soled shoes (Table 3) [13, 15, 24]. Meta-analysis found no significant difference in running economy between soft- and hard-soled shoes of similar mass (Fig. 3). However, this result was significantly affected by statistical heterogeneity and should be interpreted with caution (Fig. 3). It is likely that the heterogeneity in results was due to differences in the extent of cushioning provided by each of the shoe conditions considered across the different studies. Tung et al. [28] showed that while 10 mm of surface cushioning significantly improved barefoot running economy on a treadmill, 20 mm of surface cushioning had no significant effect. The authors hypothesised that there may be an optimum amount of surface cushioning for each individual, which minimises the metabolic cost of running [28]. Future research should investigate this hypothesis to determine if it is possible to predict the amount of shoe cushioning needed to optimise running economy based on the characteristics of a runner (e.g. body size, body composition, etc.).

In addition to the aforementioned limitations associated with use of short washout periods between testing in different shoe conditions and lack of long-term follow-up, two additional limitations should be considered when interpreting the findings of this review. Firstly, all study findings were based on running economy measurements at submaximal running speeds and assume that the response to footwear will be the same at the faster speeds that may be used in competition. Secondly, all results were based on running economy assessed on a treadmill and different levels of treadmill cushioning between laboratories may have influenced the findings [28]. Additionally, the single study [7] that compared overground running with treadmill running found significant differences in economy which favoured overground running. Although the accuracy of this finding has been debated in the literature [43, 44], overground running has the greatest external validity for investigating the effect of footwear on running performance and future research should further explore the validity and reliability of overground running economy assessment.

5 Conclusion

This review found trivial and small effects on running economy, such that greater longitudinal shoe stiffness, greater shoe cushioning and greater shoe comfort were associated with improved running economy. Light shoes and barefoot also had a small effect on running economy when compared with heavy shoes. Shoe mass was positively associated with metabolic cost of running. However, for footwear with a combined shoe mass of less than 440 g per pair, there was no detrimental effect on running economy. When controlling for differences in shoe mass, foot strike and cadence, minimalist shoes had a small beneficial effect on running economy compared with conventional running shoes. This beneficial effect appeared to increase further in response to gait adaptations resulting from training in minimalist shoes. However, further research is required to confirm this finding and any long-term beneficial effects on running economy associated with running in these shoes must be considered against the potential to affect injury risk. Future research in footwear performance should include time or distance measures of running performance and include a long-term follow-up.