The role of footwear in improving running economy: a systematic review with meta-analysis of controlled trials

Abstract

This systematic review aimed to explore the impact of different types of footwear and footwear characteristics on the running economy (RE) of long-distance runners and providing guidance for running enthusiasts and clinical practice. A comprehensive search of Web of Science, PubMed, SPORTDiscous, SCOPUS, and the China National Knowledge Infrastructure (CNKI) databases from inception to April 2024 was performed. Trials evaluating the RE of adults participating in long-distance running included comparing different footwear characteristics. This review followed the PRISMA statement. Two reviewers screened titles and abstracts to make a relevant shortlist, then retrieved and evaluated full texts against inclusion criteria for eligibility. Two independent reviewers evaluated the methodological quality of each included analysis by employing the Physiotherapy Evidence Database Scale (PEDro scale). The standardized mean difference (SMD) for the results of RE studies in each study was calculated. Of the 1338 records retrieved, 26 studies were identified in the systematic review and meta-analysis. Limited evidence indicated that compared with shod running, barefoot running (SMD = − 0.50 [95% CI − 0.86, − 0.14], P = 0.006) and minimalist running (SMD = − 0.62 [95% CI − 0.83, − 0.42], P < 0.00001) had a positive impact on RE. Compared with barefoot running, minimalist running did not show a beneficial effect (SMD = 0.37 [95% CI − 0.07, 0.81], P = 0.10) on RE. Additionally, compared with the control condition, RE showed some improvement with increased footwear longitudinal bending stiffness (SMD = − 0.53 [95% CI − 0.90, − 0.17], P = 0.005) and cushioning (SMD = − 0.33 [95% CI − 0.61, 0.06], P = 0.02). However, compared with control, RE did not change with footwear comfort (SMD = − 0.11 [95% CI − 0.42, 0.21], P = 0.51). Barefoot running or minimalist running may be more economical than shod running, high longitudinal bending stiffness and high cushioning shoes could improve RE.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-88271-2.

Introduction

Running economy (RE) refers to the energy demand of an individual during a constant-sub-maximum speed run, it can be measured by the amount of oxygen uptake (VO₂) per kilogram of body weight and minute (i.e. mL/kg/min) at that speed^1,2. It is an effective indicator of aerobic capacity. At the same steady-state speed, runners with good RE have lower VO₂ than runners with poor RE³. A previous study has shown that RE is an important predictor of long-distance running performance, especially for runners with similar maximum oxygen uptake (VO₂max)⁴. Research has shown that among elite runners with similar VO₂max, RE can explain 65.4% of the observed changes in 10 km of race performance, exhibiting high inter-individual differences (15–30%)^4,5. So, RE is a complex and multifactorial indicator influenced by various factors, including morphology, biomechanics, training, neuromuscular, metabolic, and cardiopulmonary function².

In recent years, the impact of different footwears on RE has become a topic of great interest^6,7. For runners, choosing suitable footwear is considered a basic requirement of running and a mean of improving RE^8,9. The advancement of footwear technology aims to maximize RE and minimize energy loss¹⁰. Different footwear characteristics, such as mass, cushioning performance, motion control performance, and comfort, can affect RE^6,7,11,12. Among them, the mass of footwear has been proven to be one of the important factors affecting RE⁶. Some scholars have studied the impact of footwear mass on RE and showed that for every 100 g increase in footwear mass, the VO₂ required for exercise increases by about 1%¹¹.

Barefoot running or minimalist running (running in minimalist shoes and imitating barefoot running) without cushioning may be more economical than some shod running¹². Scholars have found that over 50% of marathon runners choose minimalist running or barefoot running¹³. Reducing the footwear mass during minimalist running or barefoot running is suggested to help improve RE compared with shod running¹⁴. Some scholars believe that minimalist running or barefoot running can cause runners to transition from landing on their rearfoot to forefeet, thus increasing rhythm and reducing energy consumption, which may help improve RE¹⁵. However, at present, no consensus exists on the impact of footwear in RE among papers published in this field. Some scholars have proposed that barefoot running can improve RE more than shod running, but the effect of the research results was very small¹⁶. In addition, improvement in footwear comfort can improve RE. Protruding patterns can more effectively improve RE than no protruding patterns on the front half of the sole¹². However, no systematic review has fully determined the impact of footwear cushioning performance on RE.

Hall and Cheung et al. investigated the effects of footwear on RE in distance runners and systematically reviewed the impact of different footwears on RE^17,18. The results indicated that barefoot running or minimalist running may require lower utilization of oxygen than shod running. However, the types and characteristics of varying footwears and sports injuries were not discussed^17,18. Moreover, no meta-analysis was conducted in the past 5 years. Due to the high incidence of sports injuries and the increasingly rapid development of footwear technology, organizing and analyzing existing literature on the impact of footwear on the RE and provide scientific support through comprehensive research information are necessary. In addition, a systematic review can provide valuable information for long-distance runners to help them choose suitable footwear and minimize the risk of sports injuries.

Therefore, the present system review and meta-analysis aimed to explore the impact of footwear types (shod running, minimalist running, barefoot running) and footwear characteristics (mass, cushioning, longitudinal bending stiffness, and comfort) on the RE of long-distance runners, while considering sports injuries and providing guidance for running enthusiasts and clinical practice.

Methods

This review followed the PRISMA statement for improved reporting of systematic reviews with meta-analysis¹⁹ and is registered on the INPLASY website under registration number INPLASY2024120032.

Information sources and literature search

The search was conducted on 30 April 2024. The search databases include Web of Science, PubMed, SPORTDiscous, SCOPUS, and the China National Knowledge Infrastructure (CNKI) databases. This study searched for literature from establishment of the database to April 2024. It supplemented the search by citation search and manual selection of relevant references that met the selection criteria. The search strategy was to refine the inclusion and exclusion of pertinent literature, and all terms were searched in the form of free text and keywords. In each database, the following types of keywords were searched through “title, summary”: Keywords 1, Running economy (running economy OR VO₂ OR oxygen consumption* OR metabolic cost) AND Keywords 2, Footwear (footwear OR shoe* OR shod OR barefoot* OR minimalist*) AND Keywords 3, Running (run* OR jog* OR treadmill OR overground OR marathon). The detailed search strategy can be found in Appendix 1.

Inclusion and exclusion criteria

The inclusion of literature was based on the evidence-based medicine PICOS framework, mainly considering five factors: subjects, intervention measures, control groups, research results, and research design²⁰. The inclusion criteria for the present study were as follows: (1) healthy adult long-distance runners (aged > 18 years); (2) exercise data on barefoot running or minimalist running for the intervention group; (3) exercise data on shod running for the control group; (4) reported RE value and measured steady-state oxygen consumption or calculate energy consumption using indirect calorimetry; (5) the research design was strictly limited to randomized controlled trials (RCTs) published in peer-reviewed journals.

Trials that met one of the following exclusion criteria were excluded: (1) research without available statistical data; (2) conference abstracts, observational studies, papers, or letters; (3) literature other than English or Chinese; (4) conduct experiments in special environments (such as water) or unusual road surfaces (such as sand) that may alter running style or performance.

Study selection and data extraction

The included studies were screened by the Endnote X9 citation manager. Two researchers (LX and YW) independently reviewed this literature, excluded duplicate and irrelevant studies, and then independently decided whether to include them in the study. In the case of disagreement, a third researcher (XW) was consulted to make the final decision. First, the titles and abstracts were reviewed to exclude studies that clearly did not meet the inclusion criteria. Then, the full texts of the remaining research were screened and read, the eligibility was evaluated, and the final literature to be included was determined. Two researchers (LX and YW) independently extracted studies that met the inclusion and exclusion criteria, including article publication information, participant characteristics, study design, research content, intervention-related information, and RE results. The authors were contacted for clarification of incomplete or missing data.

Quality assessment

The two researchers (LX and YW) evaluated the methodological quality of the included literature by using the PEDro scale²¹. The classic PEDro scale consists of 11 items related to scientific rigor. A score of < 6 represents the threshold for studies with a low risk of bias²². Item 1 is rated as Yes/No, items 2–11 are rated using 0 (absent) or 1 (present), and a score out of 10 is obtained by summation. Any differences arising from the evaluation process were settled by the third arbitrator (XW).

Statistical analyses

Statistical analyses were conducted using Review Manager 5.4 (Cochrane Collaboration). The standardized mean difference (SMD) for the results of RE studies in each study were calculated while calculating the standardized mean and standard error²³. 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)²⁴. This study analyzed the RE levels on barefoot running, minimalist running, or shod running and compared the RE under different footwear longitudinal bending stiffness, cushioning degrees, and comfort levels. However, compared with barefoot running, minimalist running, or shod running, other characteristics, such as shoe cushioning and longitudinal bending stiffness, cannot be controlled, so statistical analysis included the average impact of shoe mass only. The heterogeneity across the studies was evaluated using the I² statistic. I² ≤ 25% indicated insignificant heterogeneity, I² ≤ 50% and I² > 25% indicated moderate heterogeneity, and I² ≤ 75% and I² > 50% indicated high heterogeneity²³. If heterogeneity exists, a subgroup analysis of regulatory variables was performed²⁵. Visual analysis of funnel plot symmetry was used to test publication bias²⁵.

Results

Search results

The literature selection process is shown in Fig. 1, and a total of 1338 articles were obtained through a preliminary search. After 587 duplicate reports were deleted, the titles and abstracts of 751 articles were evaluated, further excluding 717 records. After the entire article was read, 26 articles were finally included in the meta-analysis^{11,16,26–49}. All of the included articles reported the required outcome data, which enabled us to conduct a comprehensive meta-analysis without the need to separate articles for qualitative analysis only. The main reasons for being excluded were as follows: candidates did not meet the inclusion and exclusion criteria, the intervention method between orthosis and shoes was compared, and the RE outcome indicators were not provided^50–56.

Fig. 1.

Flow diagram of the study selection process.

Characteristics of included studies

The characteristics of the 26 studies included in the review are shown in Tables 1 and 2. All articles were published between 2008 and 2023, with 483 participants, and all reported their gender. Among them, 19 studies included male subjects only^{11,16,28–32,35–39,42–46,48,49}, two studies included female subjects only^27,40, and five studies included male and female subjects^{26,33,34,41,47}. The participants were between 20 and 40 years old and casual or skilled long-distance runners. The studies reported the VO₂max of the participants, and according to the guidelines of De Pauw et al. for sports science research⁵⁷, it can be considered that the subjects had above average VO₂max.

Table 1.

Characteristics of included studies.

Year	N (M/F)	Age (years)	Subjects	Study design	Comparison	Landing pattern	Footwear adaptation period (min)	Running economy (ml/kg/min) (W/kg)
Barnes et al. 2019	24 (12/12)	M: 24.3 ± 4.5 F: 23.0 ± 3.1	Motorized treadmill; slope was set at 0%	4 × 5 min at 14 km/h	SR1 vs. SR2	19 rearfoot 5 forefoot	15 min	SR1: 43.38 ± 2.21; 42.55 ± 2.58 (F) SR2: 45.43 ± 2.21; 44.44 ± 2.41 (F)
Berrones et al. 2016	14 (0/14)	27.6 ± 1.6	Motorized treadmill; slope was set at 1%	3 × 5 min at 8.85, 10.46, 12.07 km/h	SR vs. BR	SR: rearfoot BR: forefoot	5 min	SR: 41.8 ± 3.2 BR: 38.6 ± 2.7 P = 0.09
Cochrum et al. 2019	8 (8/0)	25.5 ± 4.9	Motorized treadmill; slope was set at 0%	2 × 5 min at 3.35 m/s	SR vs. MR vs. BR	SR: rearfoot MR: forefoot BR: forefoot	6 min	BR: 43.14 ± 3.07 MR: 41.11 ± 3.53 SR: 44.84 ± 4.35
Dinato et al. 2021	11 (11/0)	33.1 ± 7.2	Motorized treadmill; slope was set at 0%	2 × 3 min at 12 km/h	SR vs. MR	All rearfoot	3 min	MR: 43.6 ± 2.1 SR: 42.5 ± 2.6 P = 0.01
Divert et al. 2008	12 (12/0)	24 ± 5	Motorized treadmill; slope was set at 0%	6 × 4 min at 3.61 m/s	SR vs. MR vs. BR	SR: rearfoot MR: forefoot BR: forefoot	10 min	BR: 40.7 ± 2.9 MR: 40.6 ± 3.1 SR: 42.1 ± 2.3 P < 0.05
Franz et al. 2012	14 (14/0)	29.8 ± 7.3	Motorized treadmill; slope was set at 0%	7 × 5 min at 3.35 m/s	MR vs. BR	All forefoot	10 min	MR: 39.43 ± 0.75 BR: 40.28 ± 0.88 P < 0.05
Fuller et al. 2017	50 (50/0)	27 ± 8	Motorized treadmill; slope was set at 0%	2 × 3 min at 11.71 ± 1.07 km/h	SR vs. MR	All rearfoot	5 min	SR: 45.47 ± 2.32 MR: 42.62 ± 2.12
Hébert-Losier et al. 2022	18 (18/0)	33.5 ± 11.9	Motorized treadmill; slope was set at 1%	3 × 3 min at 12.9 ± 0.7 km/h	SR vs. MR	14 rearfoot 4 forefoot	4 min	SR: 38.9 ± 2.8 MR: 39.5 ± 3.4 P = 0.002
Hunter et al. 2019	19 (19/0)	23.3 ± 6.1	Motorized treadmill; slope was set at 0%	2 × 5 min at 4.44 m/s	SR1 vs. SR2	All rearfoot	5 min	SR1: 48.11 ± 2.49 SR2: 49.48 ± 2.60 P < 0.001
Hunter et al. 2022	18 (10/8)	25 ± 10	Motorized treadmill; slope was set at 0%	2 × 5 min at 3.83 m/s	SR1 vs. SR2	All rearfoot	5 min	SR1: 46.7 ± 3.8 SR2: 47.4 ± 4.8 P = 0.004
Jaén-Carrillo et al. 2022	41 (25/16)	28.5 ± 6.9	Motorized treadmill; slope was set at 0%	2 × 3 min at 11.71 ± 1.07 km/h	SR/BR	SR: rearfoot BR: forefoot	5 min	SR: 3.08 ± 0.32 BR: 3.07 ± 0.32 P = 0.568
Knoepfli-Lenzin et al. 2014	37 (37/0)	36.1 ± 7.6	Motorized treadmill; slope was set at 1%	2 × 15 min at 11 km/h	SR1 vs. SR2	All rearfoot	5 min	SR1e: 48.61 ± 2.71 SR2: 49.79 ± 2.52 P < 0.05
Knopp et al. 2023	14 (14/0)	22.7 ± 3.2	Motorized treadmill; slope was set at 1%	2 × 3 min at 9 and 13 km/h	SR1 vs. SR2	All rearfoot	15 min	SR1: 47.7 ± 2.6 SR2: 46.1 ± 3.2 P = 0.043
Lindorfer et al. 2020	15 (15/0)	26 ± 4	Motorized treadmill; slope was set at 0%	2 × 6 min at 11.3 ± 1.7 km/h	SR1 vs. SR2	All rearfoot	5 min	SR1: 44.4 ± 8.9 SR2: 44.2 ± 9.0 P = 0.2
Lussiana et al. 2013	14 (14/0)	23.4 ± 4.4	Motorized treadmill; slope was set at 0%	7 × 5 min with at 10 km/h	SR/MR	SR: rearfoot MR: forefoot	10 min	SR: 43.02 ± 2.5 MR: 40.30 ± 2.8 P > 0.05
Pace et al. 2020	14 (14/0)	20–30	Motorized treadmill; slope was set at 1%	2 × 5 min at 7.40 ± 0.34 km/h	SR/MR	All rearfoot	10 min	MR: 34.4 ± 3.3 SR: 35.5 ± 3.5
Paulson et al. 2014	8 (0/8)	20.1 ± 1.4	Motorized treadmill; slope was set at 0%	2 × 7 min at 3.13 m/s	SR/MR/BR	NR	5 min	BR: 42.94 ± 3.62 MR: 44.05 ± 3.04 SR: 45.45 ± 2.84 P = 0.04
Rodrigo-Carranza et al. 2020	11 (6/5)	M: 20.64 ± 1.60 F: 22.12 ± 1.03	Motorized treadmill; slope was set at 1%	2 × 5 min at 3.83 m/s	SR vs. MR	All rearfoot	15 min	MR: 47.19 ± 5.68 SR: 49.77 ± 5.24
Rodrigo-Carranza et al. 2023	28 (28/0)	28.1 ± 6.4	Motorized treadmill; slope was set at 1%	2 × 3 min at 9 and 13 km/h	SR1 vs. SR2	All rearfoot	10 min	SR1: 15.89 ± 1.24 SR2: 16.39 ± 1.24 P = 0.001
Sanno et al. 2021	18 (18/0)	24.4 ± 3.7	Motorized treadmill; slope was set at 0%	NR	SR vs. MR	All rearfoot	5 min	SR: 43.9 ± 3.8 MR: 41.5 ± 2.4
Sinclair et al. 2016	12 (12/0)	22.4 ± 2.2	Motorized treadmill; slope was set at 0%	2 × 6 min at 12 km/h	SR1 vs. SR2	All rearfoot	3 min	SR1: 41.8 ± 3.2 SR2: 43.6 ± 3.7 P = 0.008
Sinclair et al. 2016	10 (10/0)	23.4 ± 2.1	On a motorized treadmill; slope was set at 0%	2 × 6 min with at 13 km/h	MR1 vs. MR2	Rearfoot	10 min	MR1: 35.9 ± 3.4 MR2: 37.8 ± 5.2 P < 0.05
Vercruyssen et al. 2016	13 (13/0)	38.2 ± 4.8	Motorized treadmill; slope was set at 0%	2 × 5 min at 2.77 m/s	SR vs. MR	NR	10 min	SR: 46.7 ± 3.6 MR: 45.6 ± 3.8
Vincent et al. 2014	21 (16/5)	30 ± 10.9	Motorized treadmill; slope was set at 1%	2 × 6 min at 3.1 ± 0.3 m/s	SR vs. BR	All forefoot	5 min	SR: 40 ± 5.2 BR: 39.4 ± 4.7
Warne et al. 2015	23 (23/0)	43 ± 10	Motorized treadmill; slope was set at 1%	2 × 6 min at 11 km/h	SR vs. MR	All rearfoot	5 min	SR: 42.13 ± 4.25 MR: 40.60 ± 4.29 P = 0.002
Whiting et al. 2022	16 (16/0)	27.4 ± 5.4	Motorized treadmill; slope was set at 3%	6 × 5 min at 13 km/h	SR1 vs. SR2	All rearfoot	10 min	SR1: 13.52 ± 0.87 SR2: 14.06 ± 0.89

M male, F female, SR shod running, BR barefoot running, MR minimalist running, LBS longitudinal bending stiffness, NR not reported, P significance level; Data are mean ± SD.

Table 2.

Footwear characteristics of included studies.

Year	Footwear mass (g)	LBS (report whether LBS is present or absent)	Footwear cushioning (report the level of cushioning (high or low))	Footwear comfort (report the level of comfort (high or low))
Barnes et al. 2019	SR1: 205 SR2: 236	SR1: Yes SR2: No	SR1: High SR2: Low	SR1: High SR2: Low
Berrones et al. 2016	SR: 590 ± 20 BR: 0	NR	NR	NR
Cochrum et al. 2019	NR	NR	NR	NR
Dinato et al. 2021	MR: 275 SR: 320	NR	MR: Low SR: High	MR: Low SR: High
Divert et al. 2008	MR: 150 BR: 0 SR: 350	No	NR	NR
Franz et al. 2012	MR: 135 BR: 0	No	MR: High BR: Low	MR: High BR: Low
Fuller et al. 2017	SR: 333 ± 25 MR: 138 ± 10	NR	NR	SR: High MR: Low
Hébert-Losier et al. 2022	SR: 211 MR: 153	NR	SR: High MR: Low	SR: High MR: Low
Hunter et al. 2019	SR1: 184 SR2: 230	SR1: Yes SR2: No	NR	NR
Hunter et al. 2022	SR1: 213 SR2: 167	NR	SR1: High SR2: Low	NR
Jaén-Carrillo et al. 2022	NR	NR	SR: High BR: Low	SR: High BR: Low
Knoepfli-Lenzin et al. 2014	SR1: 343 ± 30 SR2: 345 ± 48	NR	SR1: High SR2: Low	NR
Knopp et al. 2023	SR1: 197 SR2: 255	NR	SR1: High SR2: Low	NR
Lindorfer et al. 2020	NR	NR	NR	SR1: 85 ± 15 High SR2: 28 ± 10 Low
Lussiana et al. 2013	NR	NR	NR	NR
Pace et al. 2020	MR: 300 SR: 801	NR	NR	MR: High SR: Low
Paulson et al. 2014	NR	NR	NR	NR
Rodrigo-Carranza et al. 2020	SR1: 178–247 SR2: 278–347	NR	NR	NR
Rodrigo-Carranza et al.2023	SR1: 248.3 ± 5.5 SR2: 240.0 ± 4.0	SR1: Yes SR2: No	NR	NR
Sanno et al. 2021	MR: 170 SR: 348	MR: No SR: No	MR: No SR: Yes	NR
Sinclair et al. 2016	NR	NR	SR1: High SR2: Low	SR1: High SR2: Low
Sinclair et al. 2016	NR	NR	NR	MR1: High MR2: Low
Vercruyssen et al. 2016	SR: 331 ± 14 MR: 209 ± 10	NR	NR	NR
Vincent et al. 2014	SR: 390 ± 20 BR: 0	NR	NR	NR
Warne et al. 2015	SR: 400 MR: 150	NR	NR	NR
Whiting et al. 2022	SR1: 203 SR2: 209	NR	SR1: High SR2: Low	NR

SR shod running, BR barefoot running, MR minimalist running, LBS longitudinal bending stiffness, NR not reported.

In all 26 articles, all participants completed the measurements of RE under laboratory conditions, with constant-sub-extreme running based on individual lactate concentration being a common measurement method. The RE of the reports was evaluated using ml/kg/min or W/kg. Previous studies have confirmed that including different units does not impact the analysis results¹⁴. Before measurements were taken, washout period of 3^29,44, 4³¹, 5^{27,30,32–35,37,40,43,47,48}, 6²⁸, 10^{11,16,38,39,42,45,46,49} or 15 min^26,36,41 were conducted to avoid errors caused by unfamiliarity with the treadmill or shoes.

Two studies compared the impact of shoe mass on RE^32,41, five studies compared shod running with barefoot running^{11,27,28,40,47}, 12 studies compared shod running with minimalist running^{11,28–31,38–41,43,46,48}, and four studies compared minimalist running with barefoot running^11,16,28,40.

Among the seven studies that considered barefoot running, two studies included participants without barefoot running experience^27,28, two studies included participants with barefoot running experience^16,40, and three studies did not clearly indicate participants’ barefoot running experience^11,34,47.

Twenty-five studies reported foot landing pattern, of which 14 were in the rearfoot category^{29–31,33,35–37,39,41–44,48,49}, two were in the forefoot category^16,47, and two did not report foot landing pattern^40,46. Three studies evaluated the impact of footwear longitudinal bending stiffness^26,32,42, five studies evaluated the effectiveness of footwear cushioning^{33,35,36,43,49}, and four studies evaluated the effect of footwear comfort^34,37,44,45.

Quality assessment

According to the PEDro scores, the selected studies have good quality (PEDro score ≥ 6). Most studies have a high risk of bias in condition concealment, because the subjects and investigators cannot be blinded to footwear conditions. The detailed scoring information can be found in Appendix 2.

Meta-analysis

Impact of shod running, barefoot running and minimalist running on RE

Six studies were included to compare the RE between barefoot running and shod running. As shown in the forest plot using a fixed model, the SMD was − 0.28 ([95% CI − 0.56, − 0.01], 104 participants), with significant statistical differences (P = 0.04, Fig. 2)^{11,27,28,34,40,47}. For minimalist running versus shod running, data were pooled from 12 studies by using a fixed model, and the SMD was − 0.62 ([95% CI − 0.83, − 0.42], 200 participants), with statistically significant differences (P < 0.00001, Fig. 3)^{11,28–31,38–41,43,46,48}. By contrast, for barefoot running versus minimalist running, the SMD was 0.37 ([95% CI − 0.07, 0.81], 42 participants) with no statistically significant differences (P = 0.10, Fig. 4), based on combined data from four studies using a fixed model^11,16,28,40. The heterogeneity of the above three analyses was all in the medium range (I² = 28% for barefoot running versus shod running; I² = 41% for minimalist running versus shod running; I² = 42% for barefoot running versus minimalist running).

Fig. 2.

Forest plot of the effects of barefoot running (BR) vs. shod running (SR) on RE.

Fig. 3.

Forest plot of the effects of minimalist running (MR) vs. shod running (SR) on RE.

Fig. 4.

Forest plot of the effects of barefoot running (BR) vs. minimalist running (MR) on RE.

Impact of footwear longitudinal bending stiffness, cushioning or comfort on RE

RE showed some improvement with increased footwear longitudinal bending stiffness (SMD = − 0.53[95% CI − 0.90, − 0.17], 59 participants, P = 0.005) compared with the control condition (Fig. 5). Similarly, RE demonstrated improvement with increased footwear cushioning (SMD = − 0.48 [95% CI − 0.76, 0.20], 103 participants, P = 0.0007, Fig. 6). In addition, RE did not change with footwear comfort (SMD = − 0.11 [95% CI − 0.42, 0.21], 78 participants, P = 0.51) compared with control (Fig. 7). The I² test showed a nonsignificant heterogeneity for footwear longitudinal bending stiffness (I² = 0%), footwear cushioning (I² = 0%), and footwear comfort (I² = 0%).

Fig. 5.

Forest plot of the effects of footwear longitudinal bending stiffness on RE.

Fig. 6.

Forest plot of the effects of footwear cushioning on RE.

Fig. 7.

Forest plot of the effects of footwear comfort on RE.

Subgroup analysis

To further analyze the causes of heterogeneity in the studies of barefoot running versus shod running, we conducted a subgroup analysis on the results with heterogeneity. When performing a subgroup analysis based on the gender of the included subjects, we found that the heterogeneity in the barefoot running and shod running groups was reduced (I² = 0%).

According to previous research analysis of minimalist running versus shod running, we did not conduct a subgroup analysis of the results due to the small number of included studies. After excluding the literature one by one, we found that the heterogeneity was significantly reduced (I² = 0%) after excluding the study by Fuller et al.³⁰.

Similarly, for the studies on minimalist running and barefoot running, the number of studies after grouping by gender did not meet the requirements for meta-analysis, so no subgroup analysis was performed on this result. After excluding the literature one by one, we found that the heterogeneity was significantly reduced (I² = 0%) after excluding the study by Franz et al.¹⁶. The detailed forest plot can be found in Appendix 3.

Quality of the included studies

The visual inspection of the funnel plots of the six different results indicated that there was no publication bias in the studies included in the meta-analysis (Appendix 4).

Discussion

This meta-analysis investigated the impact of various footwear types and characteristics on the RE of long-distance runners. The results revealed distinct effects on RE. Firstly, among different footwear conditions, compared with traditional shod running, both barefoot running and minimalist running exhibited a positive influence on RE. However, the comparison between barefoot running and minimalist running did not show a statistically significant difference in RE. Secondly, in terms of footwear characteristics, the improvement of RE was associated with increased footwear longitudinal bending stiffness and cushioning, while footwear comfort had no significant impact on RE. Notably, these effects were influenced by multiple factors related to the footwear and the runners themselves, providing valuable insights into the relationship between footwear and RE in long-distance running.

Impact of different footwear on RE

Despite the overall standardized mean difference (SMD) falling within a moderate or small range, barefoot running and minimalist running demonstrated statistically more favorable RE compared to shod running. Jaén-Carrillo et al. evaluated 41 endurance runners (including 16 females) in both barefoot and shod running conditions³⁴. Their findings revealed that barefoot running led to improved RE, which was consistent with lower form power and vertical oscillation values. This suggests that the biomechanical characteristics of barefoot running may contribute to enhanced RE. The deformation of shoes caused by foot impact may result in a loss of elastic energy, which does not occur in barefoot running nor minimalist running. Shod running provides energy return when recoiling, but relatively speaking, barefoot running and minimalist running can more efficiently store and release elastic energy in the achilles tendon and longitudinal arch of the foot⁴¹.

Berrones et al. selected 14 female long-distance runners with no barefoot running experience to conduct three 5 min sub-extreme running experiments to examine the influence of barefoot running and shod running on RE, representing 65%, 75%, and 85% of VO₂max, respectively²⁷. The results showed that barefoot running had a significant improvement in RE at 85% VO₂max (P = 0.018), but no improvement was observed in 65% or 75% of the trials (P > 0.05, both). This finding indicated the direct improvement of RE by barefoot running in a relatively high proportion of VO₂max. However, running barefoot or in flat, thin-soled minimalist shoes is suggested to cause runners to increase cadence and adopt a forefoot strike, which, in turn, improves RE¹². Therefore, the difference in RE may be due to differences in running styles.

In contrast, the comparison of RE between barefoot running and minimalist running did not yield statistically significant differences. The weight of shoes is a factor that cannot be ignored in the impact of footwear on RE. A study on the impact of shoe mass on RE reported that RE deteriorated with increasing shoe mass⁴¹. The results showed that adding 100 g per shoe can damage the RE and performance of well-trained runners. This finding was positively correlated with the research results of Hoogkamer et al., who found that for every 100 g increase in shoes, the VO₂ increased by about 1%⁶, probably because lighter running shoes reduce energy consumption during running by reducing the inertia of leg swinging⁵⁸.

If based on mass only, people may intuitively believe that barefoot running consumes lower metabolic costs than minimalist running. On the contrary, Perl et al. reported that while controlling shoe mass and stride frequency, the oxygen requirement in minimalist running was significantly reduced (2.4–3.2%), although small³⁸. Franz et al. studied 12 males with extensive barefoot running experience and reported that minimalist running had lower VO₂ and metabolic energy requirements of 3–4% than BR (P < 0.05)¹⁶. This finding indicates that barefoot running offers no metabolic advantage over running in lightweight, cushioned shoes. The heightened somatosensory feedback associated with running in minimalist shoes, which lack cushioning, is believed to prompt runners to increase cadence and land with a more anterior foot strike⁵⁹.

Impact of footwear characteristics on RE

The study results showed that under the premise of controlling the mass of shoes, footwear with high longitudinal bending stiffness and high cushioning displayed better RE, although the overall SMD was in a small range. Rodrigo et al. studied 28 male runners wearing experimental shoes with carbon fiber plate (increased longitudinal bending stiffness) and control without carbon fiber board and showed that increasing footwear longitudinal bending stiffness can improve RE during training⁴². Longitudinal bending stiffness refers to the bending resistance around the outer and inner axes of the shoe. It is believed to be most affected by midsole additives, especially the presence of carbon fiber board. However, other midsole modifications, such as increasing thickness, selecting midsole materials, and adding flexible grooves or thermoplastic polyurethane inserts, may also have an impact⁷.

Increasing the cushioning was usually associated with RE improvement⁴³. Tung et al. found that 10 mm footwear surface cushioning can considerable improve the RE on a treadmill, whereas 20 mm surface cushioning had no significant effect⁶⁰. This finding may be attributed to the optimal cushioning for individuals, which can minimize the metabolic costs of running. This shows that if there is too much cushioning, it will not improve RE. Sanno et al. studied the specific contribution of footwear cushioning to lower limb joint work in 18 male runners during a 10 km fatigue run⁴³. The results showed that the group with high cushioning significantly increased ankle-negative work (P < 0.01). So, future research should determine whether predicting the footwear cushioning required to optimize RE is possible on the basis of the characteristics of runners, such as body composition.

The inclusion of literature in this study showed that footwear comfort did not significant change RE. As an important parameter of running shoes, comfort is increasingly valued in the research and development of running shoes. It is proposed in the example of comfort filters to reduce injury rates and improve metabolic needs^61,62. Footwear comfort is generally evaluated subjectively, and Luo et al.’s study found that among the subjectively most comfortable footwear, VO₂ was significantly reduced⁶¹. Sinclair et al. conducted a 6 min steady-state run on 12 male runners wearing shoes with different comfort levels⁴⁴. The results showed that the VO₂ of high-comfort shoes and the respiratory exchange rate were significantly reduced, and the RE improved⁴⁴.

Under comfortable shoe conditions, stride frequency has been proven to be reduced⁶³. An explanation may be that the reduced stride frequency is closer to the personal optimal stride frequency that has been proven to reduce VO₂. However, this finding was not found in Lindorfer et al.’s study. Lindorfer et al. conducted running experiments on 15 male runners and found no reduction in oxygen consumption or improvement in RE under the optimal comfortable shoe conditions (P = 0.2)³⁷. Moreover, footwear comfort is generally considered a mediator of muscle mechanical work output, and muscle activity is positively correlated with overall VO₂, thereby improving RE⁶². So, previous suggestions regarding the positive effect of enhancing footwear comfort, whether from an economic or injury prevention perspective, cannot be supported, consistent with the results of the present study.

Statistical heterogeneity

Moderate heterogeneity was noted in some comparisons, such as those between barefoot running and shod running, minimalist running and shod running, and barefoot running and minimalist running. The sources of this heterogeneity could be attributed to several factors. When performing a subgroup analysis based on the gender of the included subjects, we found that the heterogeneity in the barefoot running and shod running groups was reduced. Differences in lower limb strength distribution and joint range of motion between genders might cause variations in the RE results with the same footwear, contributing to heterogeneity²⁷.

The heterogeneity of minimalist running versus shod running was significantly reduced after excluding the study by Fuller et al.³⁰. We speculate that the adaptability of the subjects to minimalist shoes in Fuller et al.’s study could account for the observed heterogeneity³⁰. Fuller et al. gradually introduced minimalist shoes during a six-week training period to explore the impact on RE, and it may not be possible to determine whether the improvement came from training or adaptation to the new shoes³⁰. The participants included in other studies may have no experience in wearing minimalist shoes, resulting in differences in biomechanics and muscle memory among participants, thus leading to different RE results.

The heterogeneity of minimalist running and barefoot running was significantly reduced after excluding the study by Franz et al.¹⁶. In the study by Franz et al., the subjects had substantial barefoot running experience (running barefoot or in minimalist running shoes for at least 8 km per week for at least 3 months in the last year)¹⁶. However, the subjects in other studies may have different levels of barefoot running experience. Some studies may include participants with less experience or no specific experience, which may lead to differences in adapting to minimalist running shoes or in the RE performance.

Sports injury

When exploring the impact of footwear on RE, potential sports injuries should be considered. Five studies comparing shod running and minimalist running reported that minimalist running leads to not only improvement in RE but also increased ankle pressure (P = 0.005)⁴³, increased calf pain (P = 0.034)⁴⁶, mild surface pain of triceps femoris⁴⁸, knee valgus⁴⁰, and increased knee angle (P < 0.01) in subjects³⁸. Under shod running, for runners to land with a dorsiflexing ankle, relieving local pressure underneath the heel by cushioning with the elevated heel is necessary⁵⁹. In barefoot initial contact, the dorsiflexion of the foot and ankle decreases to reduce this local pressure⁶⁴. However, research showed that a decreased knee extension moment may have implications for knee injuries in barefoot running or minimalist running, because the length of the ground reaction force moment arm increases^64,65. The increase in ankle power generation and absorption may be related to ankle overuse injuries in barefoot running or minimalist running, such as Achilles tendinopathy^64,66. Therefore, joint moments and power changes may be considered risk factors for knee or ankle overuse pathology¹⁸. In addition, from a practical perspective, the long-term effect of barefoot running training may lead to an increase in leg stiffness for runners, thereby affecting RE.

Limitation and future perspectives

Several limitations should be acknowledged in this current study.

First, in this study, the PEDro scale assessment indicated that most studies had a high risk of bias in condition concealment because it was difficult to blind the participants and assessors to the footwear conditions. Since footwear was the core intervention factor, it was challenging to implement double-blind procedures for both the participants and the researchers. The participants could clearly perceive the differences in the footwear they were wearing, and the researchers could hardly avoid knowing the footwear conditions during the assessment process. This lack of blinding may lead to expectancy bias. For example, the participants might have a psychological positive impact because they believed that barefoot running or a specific type of footwear was more beneficial for RE, thus affecting the experimental results. However, to minimize this bias, we formulated strict inclusion criteria during the study design to ensure the quality and comparability of the included studies. Additionally, we comprehensively retrieved literature from multiple databases to cover as many relevant studies as possible and reduce the bias caused by literature selection.

Second, the studies used different methods to measure RE (e.g., VO₂/kg/min vs. W/kg). Although the manuscript stated that this did not affect the results, no detailed analysis supported this claim. In actual measurements, different devices and measurement environments could lead to measurement errors. For example, the accuracy and calibration methods of the gas analysis equipment used to measure VO₂, as well as the stability of the subjects’ breathing patterns during the measurement process, could all cause differences in VO₂ measurements under the same footwear conditions in different studies. Similarly, the accuracy of the equipment used to measure power (W/kg) and the selection of measurement points could also affect the results. Future studies could simultaneously adopt multiple measurement methods in the same experiment to compare the consistency of the measurement results or standardize the calibration of different measurement methods to ensure the reliability and comparability of the research results.

Third, the short follow-up periods (up to 15 min in most studies) in this study limited our ability to draw conclusions about the long-term effects of different footwear types on running economy. In the long term, runners may experience adaptive changes in muscle strength and joint stability over time, which could further affect RE. Additionally, long-term wearing of different footwear may have cumulative effects on the body, such as joint wear and muscle fatigue recovery, which could not be reflected in short-term studies. Therefore, extending the follow-up time is crucial for comprehensively evaluating the long-term impact of footwear on running economy.

Perspective and recommendations

The results indicated that barefoot running or minimalist running may be more economical than shod running. Furthermore, high longitudinal bending stiffness and high cushioning also improve RE. In addition, barefoot running or minimalist running for long-term training may lead to a series of lower limb injuries risk. Future research still needs to develop standard testing plans to explore biomechanical factors and improve RE while avoiding the occurrence of sports injuries.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Designed the study and wrote the protocol, L.X.; Independent screening and data extraction, L.X. and Y.W.; Quality scoring, L.X. and Y.W.; Statistical analysis and wrote the first draft, L.X. and Y.W.; Revised the manuscript, X.W. All the authors made significant contributions to the final manuscript and approved its publication.The study received financial support from vivo Mobile Communication Co., Ltd. The funders had no role in study design, data analysis, data interpretation, writing of this report, or in the decision to submit this paper for publication.

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.