Descriptive differences in physiological
and biomechanical parameters between
running shoes
- a pilot study with a single-subject experimental
design
Alexander Wolthon
THE SWEDISH SCHOOL OF SPORT
AND HEALTH SCIENCES
Master’s degree Project 52:2020
Sports Science: 2018-2020
Supervisor: Filip Larsen
Co-supervisor: Toni Arndt
Examinor: Erik Hemmingsson & Magnus Lindwall
Abstract
Running performance has increased immensely during the last few years, coinciding with multiple shattered world records in relatively short amount of time. Improvements in footwear material and design are likely reasons for this increase in running performance. Previous studies on the effect of footwear on running economy (RE), a determinant of running performance, have not included participant-blinding. Furthermore, they have yet to compare multiple carbon-fiber plated running shoes available for purchase, what differences there are across price ranges and shoe categories, and if there is such a thing as a placebo-effect.
Aim: (1) Descriptively compare a set of heterogeneous running shoes, with regards to running economy, Foot Strike Type (FST), vertical oscillation, ground contact time, stride length and cadence; including (2) a ‘sham’ and ‘normal condition’ of the same running shoe model; and (3) explore the participant’s perception of the study-specific blinding protocol.
Method: A Single Subject Experimental Study (N=1), comparing nine different shoe conditions using a crossover design. The assessment of RE was conducted using indirect calorimetry with mixing-chamber in a climate-controlled facility. Spatiotemporal parameters were assessed using a Garmin HRM-Run™, and foot strike type was visually assessed using a frame-by-frame approach based on 2D-video at 240 fps.
Results: The average running economy across all shoe tests varied between 16.02 to 17.02 W/kg, with the ‘worst’ shoe costing 6.24% W/kg more than the ‘best’ shoe. The descriptive difference between the ‘sham’ and ‘normal condition’ were negligible and within the range of measurement error. Spatiotemporal parameters were overall descriptively similar between the shoes, with a few outliers who differed with regards to measure of central tendency or dispersion. FST differed between the shoes including the ‘sham’ and ‘normal condition’, but were overall consistent with the participant’s habitual FST. The study-specific blinding procedure was perceived to work well, but may also be improved in some remarks.
Conclusion: Descriptive difference in some, but not all, physiological and biomechanical parameters were observed between the shoe conditions in this study, including the ‘sham’ and ‘normal condition’. Blinding procedures in experimental footwear research may be feasible and adopted with future studies.
Sammanfattning
Prestation inom löpning har ökat drastiskt de senaste åren, tillsammans med flera nya världsrekord under relativt kort tid. Förbättrade löparskor med avseende på material och design är en sannolik orsak till denna ökning i prestation. Tidigare studier som studerat effekten av löpskor på löpekonomi (RE), en avgörande faktor för löpprestation, har inte genomförts med deltagarblinding. Dessutom har tidigare studier inte undersökt flera olika löpskor med kolfiberplatta, vilka skillnader det kan finnas mellan pris- eller skokategori, eller ifall placeboeffekter kan påverka utfallet.
Syfte: (1) Deskriptivt jämföra ett heterogent urval av löpskor, med avseende på löpekonomi, fotisättning, höjdförflyttning, markkontakttid, steglängd och stegfrekvens; (2) hur dessa variabler skiljer sig mellan ’sham’ och ’normal condition’; och (3) undersöka deltagarens uppfattning om det studie-specifika blindningsprotokollet.
Metod: En Single Subject Experimental Study (N=1), som jämför nio olika skointerventioner baserat på en crossover-design. Bedömningen av RE utfördes via indirekt kalorimetri med blandningskammare i ett klimatkontrollerat rum. Höjdförflyttning, markkontakttid, steglängd och stegfrekevens bedömdes via en Garmin HRM-Run™, ochfotisättning bedömdes via en frame-by-frame metod baserat på 2D-videoinspelning i 240 fps.
Resultat: Den genomsnittliga löpekonomin för respektive skotest varierade från 16.02 till 17.02 W/kg, där den ’sämsta’ skon kostade 6.24% W/kg mer än den ’bästa’ skon. Den deskriptiva skillnaden mellan ’sham’ och ’normal condition’ var försumbar och inom ramen för tekniskt mätfel. Skotesterna var deskriptivt lika varandra med avseende på spatiotemporala parametrar, men några skor dokumenterades som outliers med avseende på antingen central- eller spridningsmått. Fotisättning skiljde sig mellan skorna, inklusive ’sham’ och ’normal condition’, men generellt var FST likt deltagarens habituella FST. Det studiespecifika blindningsprotokollet uppfattades fungera väl av studiedeltagaren, men kan möjligtvis förbättras i några avseenden.
Konklusion: Deskriptiva skillnader mellan vissa, men inte alla, fysiologiska och biomekaniska variabler observerades mellan interventionsskorna i denna studie, inklusive mellan ’sham’ och ’normal condition’. Det kan eventuellt vara genomförbart med blindningsprotocol inom experimentell skoforskning, vilket kan tänkas bli vanligt i framtida forskning.
Table of Contents
1 Background ... 1
1.1 A New Running Paradigm ... 1
1.2 Running Performance Physiology ... 1
1.2.1 Maximal Oxygen Uptake ... 2
1.2.2 Lactate Thresholds and Fractional Utilization of VO2max ... 3
1.2.3 Running Economy... 5
1.3 Running Shoes ... 6
1.3.1 Performance and Regulation... 6
1.3.2 Effect on Running Economy and Mechanisms of Action ... 6
1.3.3 Previous Studies ... 7
1.4 Aims and Research Questions ... 8
2 Method ... 8
2.1 Design ... 8
2.2 Population... 8
2.3 Ethics ... 8
2.4 Intervention ... 9
2.5 Equipment and Calibration ... 9
2.6 Preparation...10 2.7 Experimental Procedure ...10 2.7 Statistics ...11 3 Results ...12 3.1 Running Economy ...11 3.2 Spatiotemporal Variables ...15
3.3 Foot Strike Type ...20
3.4 Self-Reported Outcomes ...20
3.5 Qualitative Assessment ...21
4 Discussion ...22
4.1 Blinding Protocol ...22
4.2 Physiological and Biomechanical Parameters ...23
4.3 External Validity and Inter-Individual Variation ...25
4.5 Limitations...26
4.6 Future Directions ...27
5 Conclusion ...29
6 References ...30
Table and figures Table I. Descriptive of person-related characteristics ... 12
Table II. Descriptive of shoe-related characteristics ... 13
Figure 1. Raincloud plot of Running Economy ... 14
Table III. Cardiorespiratory variables... 14
Figure 2. Raincloud plot of Ground Contact Time ... 16
Table IV. Pairwise comparison for Ground Contact Time ... 16
Figure 3. Raincloud plot of Cadence ... 17
Table V. Pairwise comparison for Cadence ... 17
Figure 4. Raincloud plot of Stride Length ... 18
Table VI. Pairwise comparison for Stride Length ... 18
Figure 5. Raincloud plot of Vertical Oscillation ... 19
Table VII. Pairwise comparison for Vertical Oscillation ... 19
Figure 6. Graphical illustration of foot strike classification ... 20
Figure 7. Graphical illustration of self-reported items ... 21
Table VIII. Shoe-condition sequences based on a Williams design ... 27
Supplementary Appendices
Supplementary Appendix 1: Shoes and Setup
Supplementary Appendix 2: Rating Scales, Questionnaire and Shoe Catalogue Supplementary Appendix 3: Frame-By-Frame Analysis of FST
1
1 Background
1.1 A New Running Paradigm
Participation in the New York City Marathon increased between the years of 2006 and 2016 (Nikolaidis et al., 2018). Similarly, half-marathon and marathon participation in Switzerland increased by more than 500% between the years of 1999 and 2014 (Knechtle et al., 2016). These findings highlight the immense growth in running popularity, and that more people now than ever participate in competitive running races. Moreover, running performance is at a level never seen before, which is illustrated by the male world record (WR) of 02:01:39 (hours:minutes:seconds) at the Berlin Marathon, September 2018; the unofficial marathon of 01:59:40 in Vienna, October 2019; the female WR of 02:14:04 at the Chicago Marathon, October 2019; the male 10-km WR of 00:26:24 in Valencia, January 2020; and more (World Athletics, 2020).
1.2 Running Performance Physiology
Numerous physiological parameters (Saltin & Åstrand, 1967; Costill et al., 1973; Farrell et al., 1979; Conley & Krahenbuhl, 1980), and peak treadmill velocity (Noakes et al., 1990) have been found to be strongly associated with running performance. Based on the framework by Bassett and Howley (2000), as far as physiological parameters go, Maximal Oxygen Uptake (VO2max), Fractional Utilization of VO2max (%VO2max), Lactate Threshold (LT), and Running Economy (RE), are determinants of endurance performance. Indeed, Farrell et al. (1979) found that the running velocity at LT (LTV) could predict marathon performance to 98%.
Furthermore, McLaughlin et al. (2010) found that VO2max, %VO2max and RE could explain 95.4% the variance in 16-km performance among well-trained distance runners. After including peak treadmill velocity in their prediction model, they were able to explain 97.8% of the variance in performance. Consequently, physiologist have attempted to model the best possible marathon performance based on the aforementioned parameters and have come up with 01:57:48 (Joyner, 1991), although this is purely hypothetical and based on the assumption that it is possible to have exceptional values in all relevant physiological parameters simultaneously. Interestingly, this could be improbable as there may be a trade-off between VO2max and RE
2 1.2.1 Maximal Oxygen Uptake
Cellular metabolism enabling running activity is supported by both anaerobic and aerobic metabolic pathways, with a shift in the predominant pathway depending on the relative running intensity (De Feo et al., 2003; Nilsson et al., 2019). During long-distance running, oxidative phosphorylation is perhaps the most important metabolic pathway (Spriet, 2007), and is as it sounds, related to oxygen. Therefore, VO2max correspond with the upper limit of the cardiorespiratory system to transport oxygen to working muscles to maintain aerobic metabolism which long-distance running is dependent on (Bassett & Howley, 2000). Oxygen uptake (VO2) is linearly and curvilinearly related to running velocity at submaximal speeds, among average- and sub elite runners respectively (Batliner et al., 2017); and it is therefore not surprising that higher VO2max is associated with better long-distance running performance, and alone able to explain 90.2% of the performance in a 16-km race (McLaughlin et al., 2010).
VO2max is defined as a plateau in VO2 despite increased workload (Saltin & Åstrand, 1967), hence ‘maximal’ oxygen uptake. Numerous exercise methods exist in order to assess this (Beltz et al., 2016), using either maximal- (Andersen, 1995), or submaximal (Åstrand & Ryhming, 1954; Björkman et al., 2016) workload tests. VO2max is trainable and improvements can be achieved by different training protocols, with similar improvements when comparing training intensities above 60% of VO2max according to a recent meta-regression and meta-analysis
(Scribbans et al., 2016). However, another meta-analysis showed that there may be a small beneficial effect with training at higher intensity (Milanović et al., 2015), although this should be interpreted with caution as the 95% confidence interval (95% CI) for the mean difference between low- and high-intensity training was close to zero.
It has been extensively debated which factors limit VO2max and consequently prevent runners from exercising at maximal capacity for longer periods of time, with proposed mechanisms ranging from central oxygen delivery, to neural control systems within the brain and spinal cord (Basset & Howley, 1997; Bergh et al., 2000; Levine, 2008; Robergs, 2017; Noakes, 2018). It is however most probable that oxygen delivery and transport capacity, best explained by cardiac output (CO) and the oxygen transport coefficient for blood, are the major limiting factors of VO2max (Di Prampero, 2003), although other factors (e.g. mitochondrial function) play an important role as well (Cardinale et al., 2019). Interestingly, the limiting
3
capacity and importance of the cardiac system on VO2max is something Nobel Laurate Hill et
al. (1923; 1924) proposed almost a century ago.
Therefore, the physiological mechanisms which contribute to the improvements in VO2max after a training period could be a product of increased CO (Daussin et al., 2007), increased blood volume (Sawka et al., 2000), and/or mitochondrial adaptation (McInnis & Gibala, 2017; Cardinale et al., 2019; Larsen et al., 2020), and be dependent on the training modality used. For example, despite similar improvements in VO2max, differences in CO adaptation were observed in a randomized trial comparing training at different intensities, with higher intensity providing no improvements in CO (Macpherson et al., 2011). However, contradictory results have been published, and a recent review discussed that it is difficult to draw strong conclusions on how certain training methods relate to specific physiological adaptation pathways leading to improvements in VO2max(McInnis & Gibala, 2017).
1.2.2 Lactate Thresholds and Fractional Utilization of VO2max
Although VO2max is a strong predictor of long-distance performance (McLaughlin et al., 2010), equally interesting is what fraction of VO2max that can be utilized and maintained for longer periods of running. This physiological parameter is strongly linked with LTV (Costill et al.,
1973) which itself is a better predictor of endurance performance than VO2max in a homogenous group of runners (Farrell et al., 1979). Additionally, it is speculated that runners would be required to stop running due to fatigue before having covered a fourth of a marathon if they were running 5-10% faster than their ideal marathon pace (i.e. running faster than their LTV)
(Coyle, 2007).
The concept of LT is not straightforward as there are at least 25 definitions available (Faude et al., 2009). Commonly used definitions are: LT1, sometimes called the aerobic threshold, which correspond to the first increase in blood lactate (BLa) above resting values; and LT2, sometimes called theanaerobic threshold (Svedahl & MacIntosh, 2003), onset of blood lactate accumulation, lactate turnpoint or maximal lactate steady state, which correspond to a rapid increase of BLa (Faude et al., 2009). To complicate things further, some define these parameters as fixed BLa values (+ 0.2, + 0.5, or + 4.0 mmol/L) above resting values; others visually inspect a given lactate curve composed of BLa and intensity, and select the point on the curve that corresponds to an inclination angle of 45 or 51°; others use mathematical models (e.g. Dmax) to
4
determine which running intensity best correspond to the LT (Faude et al., 2009). Associated with the LT is an increase in pulmonary ventilation (VE) and CO2 output (VCO2). It is debated whether increased VCO2 is a cause or effect of increased VE (Hopker et al., 2011; Whipp &
Ward, 2011), however this increase is defined as the gas exchange threshold, the anaerobic ventilatory threshold, or simply the ventilatory threshold (Gosh, 2004).
The importance of the LT is somewhat understood when reviewing the possible physiological explanations behind it (Walsh & Banister, 1988; Ferguson et al., 2018). The metabolic energy systems providing runners with energy are not solely aerobic or anaerobic, instead there will be a shift in the predominant pathway depending on intensity of effort (De Feo et al., 2003; Nilsson et al., 2019). Exercise intensity above LT will lead to a more predominant anaerobic pathway activation (glycolysis) because the aerobic pathways (oxidative phosphorylation) cannot provide enough energy to supply the increased energy demand. BLa will begin to accumulate as an end-product from the accelerated glycolysis (Ferguson et al., 2018), and act as an oxidizable energy substrate to support energy production (Brooks 2007; 2018). Moreover, BLa accumulation contribute to a decrease in pH, causing the working muscle to become acidic, although it does not appear that acidosis itself is a cause of fatigue (Westerblad, 2016). The physiological mechanism underlying the fatigue-phenomenon above LT is not well understood, however it is a reality which do inhibit running performance.
The LT or %VO2max is not determined by the cardiac system as VO2max is, but more so influenced by the metabolic environment within the working muscles. Therefore, mitochondrial- (Hashimoto & Brooks, 2008), capillary- (Tesch & Wright, 1983), and muscle fiber characteristics (Ivy et al., 1980) all play their part, and can be improved by training although adaptation may differ depending on the training stimulus applied (Joyner & Coyle, 2007; MacInnis & Gibala, 2017). For example, adaptation of the mitochondria may be more pronounced when performing training at high intensity compared with low intensity at similar training volume (Gibala, 2009; Popov, 2018). Specifically, it is observed that exercise-induced lactate-rich metabolic environments which is more common at higher intensities, is associated with a greater mitochondrial biogenesis (Norrbom et al., 2004), which may be due to the lactate molecule itself acting as a signal promoting adaptation (Hashimoto & Brooks, 2008). Consequently, it is therefore reasonable that high intensity training is better at improving LT than low intensity training (Londeree, 1997).
5 1.2.3 Running Economy
RE is traditionally defined as the amount of oxygen cost per bodyweight per minute or km (ml
O2/kg/min, or ml O2/kg/km) (Barnes & Kilding, 2015), however it is argued that it should be
defined as energy expenditure (W/kg, instead (Shaw et al., 2014; Beck et al., 2018). Therefore, in the context of this study it will be defined as energy expenditure (W/kg; Joule/(kg·s)), derived from the following formula {RE [W/kg] = (VO2 [L/min] · RER caloric equivalent [kcal/L O2]
· 4184 [Joule/kcal]) / (60 [seconds/min] · bodyweight [kg])}, which is based on RER caloric equivalents (Peronnet & Massicotte, 1991).
As previously stated, RE is a determinant of long-distance performance (Bassett & Howley, 2000), and like %VO2max it is not as dependent on the cardiac system as VO2max is. Instead, RE is more dependent on biomechanics and peripheral physiological factors (Saunders et al., 2004; Barnes & Kilding, 2015; Fletcher & MacIntosh, 2017). For example, ground contact time and vertical oscillation have been found to be associated with running economy (Folland et al., 2017), and ankle kinematics at initial-contact could be a relevant parameter as well although there is currently limited evidence for this (Anderson et al., 2019). Moreover, dietary strategies may be of importance (Burke et al., 2019),for example seen by reductions in the oxygen cost of exercise by the ingestion of nitrates (Larsen et al., 2007; Larsen et al., 2010; Jones et al., 2018), which seem to affect mitochondrial function (Larsen et al., 2011a; Larsen et al., 2012).
Based on the framework by Hoogkamer et al. (2017) and Kipp et al. (2019) it was proposed through mathematical modelling and specific biomechanical strategies, that improvements in RE would allow runners to maintain a faster running velocity, and thus enable mankind to break the 2-hour marathon barrier. Not long after, with some biomechanical aid (Hoogkamer et al. 2017),the Nike sponsored Eliud Kipchoge managed to do so, averaging ~21.16 km/h (02:50 min:sec/km) (World Athletics, 2020). Knowing that Nike was in contact with physiologist Andrew Jones (University of Exeter, 2019), it can be speculated that dietary factors had an important role to play as well.
RE is commonly assessed using indirect calorimetry, either by the breath-by-breath or mixing-chamber method. During these assessments, participants are required to run below their LT to achieve a physiological steady-state (Barnes & Kilding, 2015). The VO2 kinetic response
6
to exercise depends on the intensity of the effort, but can be broken down in three phases; 1) a fast increase in VO2 which last for 15-25 seconds, possibly due to increased cardiac output and pulmonary blood flow (Weissman et al., 1982); 2) an exponential increase in VO2 uptake which last for about 3 minutes, possibly due to metabolic change within the working muscles (Whipp & Wawrd, 1990); and 3) a physiological steady-state where VO2 remain at a stable level. However, running above the LT will lead to bLA accumulation, and VO2 will continue to increase at a slow rate preventing a physiological steady-state. This slow increase of VO2 is called the ‘slow-component of VO2’ (Xu & Rhodes, 1999).
1.3 Running Shoes
1.3.1 Performance and Regulation
Running shoes are constructed with many purposes, of which one has been to mitigate abnormal mediolateral movements in an attempt to reduce the risk of running-related injuries (RRI) (Asplund & Brown, 2005; Hamill, 2017). Although studies have failed to show beneficial effects on RRI from matching shoe type with foot type (Nigg et al., 2015; Napier & Willy, 2018), the running shoe effect on performance has been shown both in laboratory studies (Hoogkamer et al, 2016), as well as during real world running events (Nike, 2017; New York Times, 2019), and is a consequence of new cushioning materials, carbon-fiber plates, and shoe construction (Hoogkamer et al., 2017).
These advances in materials science and running shoe design are most certainly contributors to how mankind was able to achieve a sub 2-hour marathon in October 2019. New technology and advances in science are welcomed; however, there may be an issue when the new technology is not available to all; giving some an unfair advantage over others. Unsurprisingly, World Athletics (2020) renewed their regulations regarding running shoes in January 2020, which were in accordance with the opinion of some researchers (Burns & Tam, 2019) but not all (Edward, 2019; Hoogkamer 2020).
1.3.2 Effect on Running Economy and Mechanisms of Action
These findings on improved performance are in particular thought to be explained by improvements in RE, where ‘better’ shoes nowadays reduce the energy cost of running to a great extent (Hoogkamer et al., 2017). The most prominent running shoe with new energy-return midsole material and a carbon-fiber plate, the Nike Vaporfly 4%, has been reported to
7
improve RE by an average of 4%, compared with the Nike Zoom Streak and the Adidas Adizero Adios (Hoogkamer et al., 2018). This finding has been replicated in other studies (Barnes & Kilding, 2018; Hunter et al., 2019). Interestingly, on the extreme side of this spectrum of technology innovation, researchers have found that a powered ankle exoskeleton could improve RE by 14.6% on average, and up to 23.9% (Witte et al., 2020).
This effect on RE could perhaps be explained by the combination of (1) reduced shoe-weight, as it has consistently been found that higher weight on the shoes increase energy cost of running (Frederick et al., 1984; Franz et al., 2012); (2) the improvements in cushioning material, allowing for increased material resiliency (i.e. the ability to return stored mechanical energy) as it has been shown that more this type of cushioning material improves RE (Worobets et al., 2014; Sinclair et al., 2016); (3) increased longitudinal bending stiffness (LBS), achieved by the insertion of a stiff plate (e.g. carbon-fiber plate) within the midsole material (Sun et al., 2020), although the effect from the stiff plate could be dependent on running velocity (Day & Hahn, 2020); Lastly (4), possibly influenced by a small amount, by improvements in shoe-design to help reduce air resistance (Hoogkamer et al., 2018).
Although there is evidence for the beneficial shoe-effect on RE, not all individuals experience the same effect. In the study on the Vaporfly 4%, compared with the Nike Zoom Streak, one individual experienced an improvement in RE by ‘only’ 1.59%, while another individual experienced an improvement by 6.29% (Hoogkamer et al., 2018). This inter-individual difference has been shown in other studies as well (Barnes & Kilding, 2018; Hunter et al., 2019). It is not fully understood why this is, and what separates ‘responders’ from ‘non-responders’, but possible reasons could be differences in comfort or ride perception (Luo et al., 2009; Nigg et al., 2017), differences in foot strike pattern (Hoogkamer et al., 2018; McLeod et al., 2020), or other factors that influence which amount of LBS is ‘optimal’ for a given individual, as too much stiffness can be detrimental (Oh & Park, 2017; McLeod et al., 2020).
1.3.3 Previous Studies
Numerous studies have been published regarding the effect of running shoes on biomechanics and physiological parameters (Sun et al., 2020). However, many studies are performed on custom study shoes (e.g. modified by weight or sole inserts) (Roy & Stefanyshyn, 2006; Worobets et al., 2014; Madden et al., 2016; Hoogkamer et al., 2016; Hoogkamer et al., 2018;
8
Oh & Park, 2017; McLeod et al., 2020), confidential study shoes simply described as ‘neutral protective running shoes’ (Squadrone & Gallozzi, 2009), or shoes with characteristics that no longer represent the currently available selection of running shoes (Hamill et al., 1988). The only studies that are performed on identifiable non-custom study shoes within the last five years compare a few models only (Hollander et al., 2015; Au et al., 2018; Mercer et al., 2018; Barnes & Kilding, 2018; Hunter et al., 2019), and all lack participant blinding. As such, it is currently unknown what differences there are across current shoe models and categories (race shoes and training shoes; or neutral and motion control shoes), or across price categories (cheap and expensive), and if the previous reported results have been influenced by participant bias due to a lack of blinding.
1.4 Aims and Research Questions
Therefore, the aim of this study is to (1) descriptively compare a set of heterogeneous running shoes, with regards to running economy, foot strike type, vertical oscillation, ground contact time, stride length and cadence; including (2) a “normal” and “sham” condition of the same running shoe model; and (3) explore the participant’s perception of the study-specific blinding protocol.
2 Method
2.1 Design
A single-subject experimental design in laboratory setting.
2.2 Population
Recruitment took place in running groups, universities, and running-shoe retail stores in Stockholm. Inclusion criteria were: Previous finishing times on 10k, half-marathon, or marathon of ≤ 00:45:00, 01:41:00, and 03:31:00 respectively; and having a shoe size equivalent to US Women 9.5-10.5, and US Male 8.5-9.5. Exclusion criteria were: Age below 18 years old; running-related injury during the last 12 weeks; any cold- or flue like symptom (e.g. cough, sore throat); and international travel during the last 2 weeks.
9
2.3 Ethics
According to Swedish law (The Ethical Review Act, 2003:460), this project did not require formal ethical approval due to it being carried out as a master’s project (2§). However, according to the Swedish Higher Education Act (1992:1434) ethical consideration was required (3a§), and the declaration of Helsinki was therefore adhered to (World Medical Association, 2013). For example, informed written consent was obtained from the participant. It was within their right as a participant to stop the procedure whenever they wanted, and/or have their data removed. Because the treadmill running can be exhausting, the study participant was offered water to rehydrate during the rest pauses between each test, and if they required additional rest that would have been given to them.
2.4 Intervention
Eight different running shoe models were tested, which were: (A) Kalenji Run 100, (B) ASICS
Gel-Foundation 13, (C) Xtep 160X, (D) Adidas Adizero Adios 5, (E) Skechers GOrun Speed Elite, (F) Brooks Hyperion Elite, (G) Nike Vaporfly 4%, and (H) Nike Next%. Kalenji Run 100
was tested twice [(A), (I)], with one of the times (I) being presented as a ‘lightweight high-end
performance shoe prototype designed to improve RE by 5%’, as per the second research
question. Thus, nine shoe conditions were tested, in the following order: DCEBFAGIH.
2.5 Equipment and Calibration
This pilot-study took place at the Laboratory of Applied Sport Science (The Swedish School of Sport and Health Sciences), in a climate-controlled facility with a temperature, humidity, and barometric pressure of 19ºC, 40%, and 740 mmHg, respectively. A computerized metabolic system (Oxycon Pro™, Erich Jaeger GmbH & Co KG, Friedberg, Germany) was used to analyze respiratory gases, with mixing-chamber due to the excellent concurrent validity compared with the gold-standard Douglas Bag Method, only underestimating VO2 by 0.8% (Foss & Hallén, 2005). Moreover, the mixing-chamber method has better reproducibility compared with the popular breath-by-breath method, with a coefficient of variation of 1.2% instead of 4.4-6.5% (Carter & Jeukendrup, 2002; Foss & Hallén, 2005).
Respiratory ventilation was measured within the metabolic system using a digital volume sensor, and the concentration of oxygen and carbon dioxide were measured by a paramagnetic and infra-red analyzer, respectively. A face-mask (7450 Series Silicone V2™ Oro-Nasal Mask,
10
Hans Rudolph Inc., Kansas, USA) was connected to the metabolic system through a 2.0 m long urethane/nylon flexible ducting (U62, Senior Aerospace BWT, Macclesfield, United Kingdom). Blood lactate was obtained by capillary blood samples (20 μl) and analyzed in a lactate analyzer (Biosen C-line Clinic, EKF Diagnostics, Cardiff, United Kingdom), to control the participant staying below a fixed LT of 4 mmol/L. The treadmill which was used was motor-driven, 2 m long, 0.6 m wide (RL2500E, Rodby Innovation AB, Vänge, Sweden), and partially covered in a cloth to prevent the participant from identifying which shoe models were worn (see supplementary appendix 1 for pictures of the shoes and setup).
Before use, the climate-controlled facility was calibrated an hour in advance; the metabolic system was calibrated to ambient conditions, volume flow, and to a high precision gas of 15.00% O2 and 6.00% CO2 (accuracy O2 ± 0.04% and CO2 ± 0.1%; Air Liquide AB, Kungsängen, Sweden); and the blood lactate analyzer calibrated to a sample of known lactate concentration. Ground contact time, vertical oscillation, cadence and stride length were obtained through a heart rate monitor (HRM-Run™, Garmin Ltd., Kansas, USA) connected to a sports watch (Forerunner® 735XT, Garmin Ltd., Kansas, USA). Foot strike type were recorded in 240 fps 1080p (Galaxy S20+, Samsung Electronics Co., Ltd., Suwon, South Korea), and analyzed using a frame-by-frame approach (Meyer et al., 2018). Overall shoe comfort was assessed using a 100 mm VAS scale (Mills et al., 2010); perception of ride (weight, yield and energy return) was assessed using a 5-point likert scale (Agresta et al., 2020).
2.6 Preparation
The participant arrived at the laboratory after having abstained from intense training, alcohol and caffeine for 48-, 72-, and 12 hours respectively. Once at the laboratory, they were introduced to the equipment, reminded of the study protocol and potential risks of participating. The participant gave their informed written consent, was introduced to the rating scales, shoe-catalogue, and had their body weight, length, and recommended shoe size (Centimeter Adult Brannock Device, The Brannock Device Co., Inc., New York, USA) measured. Once that was complete, they were equipped with the face-mask and connected to the metabolic system and remained seated for 10 minutes to obtain resting VO2 values.
2.7 Experimental Procedure
11
(2) Determination of appropriate individual running speed: The warm-up transitioned into a discontinuous incremental treadmill test with 2 stages, and continued for 5 minutes per stage. It started at comfortable running speed of 12.0 km/h, and the second stage was performed at 13.5 km/h. Within 30 seconds after each stage, a capillary sample was obtained. Both the warm-up and the incremental treadmill test were performed in the participant’s own shoes.
(3a) Shoe experiment preparation: The participant was equipped the 1st shoe condition with the help of AW, having their feet behind a study-specific apparatus to blind the participant to the shoes being equipped. The participant tied the shoes themselves but was not allowed to touch the upper material. Once fitted with the shoe model, it was covered by a piece of cloth, and the participant moved onto the treadmill where shoe-cloth was removed.
(3b) Shoe experiment test: The participant ran on the treadmill at 13.5 km/h. The treadmill run lasted for 5 minutes to allow the participant to achieve a physiological steady-state of 2 min. After each treadmill run, the participant rated their perceived exertion (Borg RPE 6-20), shoe comfort (100 mm VAS), ride (0-5 likert scale), and also reviewed a shoe catalogue of 100 models to select the running shoe they believe they ran in (see supplementary appendix 2 for the instruments used). Between each shoe experiment, the participant was given a rest pause of 3 min, which consisted of standing rest and changing of shoe condition. After every third treadmill run (3rd, 6th, 9th), the participant had their body weight re-assessed.
2.8 Statistics
Physiological and biomechanical variables are based on the last 2 minutes of running, from each shoe condition. Physiological measures were sampled every 15th second, except HR which was sampled every 0.5 second. Stride Length was sampled very 5th second, while Cadence, VO, and GCT were sampled every 0.5 second. Foot strike type was manually analyzed. The assumption of normality was assessed using a three-step process, first by visual inspection of histograms, followed by Shapiro-Wilk and Kolmogorov-Smirnov, and lastly assessment of skewness and kurtosis by comparing them to their respective standardized error value. Descriptive statistics are presented as mean (M) and standard deviation (SD) for parameters where most shoes are normally distributed, and median (Mdn) and interquartile range (IQR) for parameters where most shoes are non-normally distributed. Raincloud plots (Allen et al., 2019) are used to illustrate the data. Descriptive statistics and data visualization were performed in R (v.4.0.2) for windows, using RStudio.
12
3 Results
Out of 47 interested applicants (35 males, 12 females), the individual included in this study was selected by chance as he (#14) was the only one to schedule their tests before week 13. Additional applicants were initially planned to be included during the following weeks, with both time and day having been assigned to them for data collection. However, these plans were revoked due to local research restrictions at the Swedish School of Sport and Health Sciences due to the Covid-19 pandemic, and no further participants were included. The participant performed the treadmill test with a pace of 04:26 min/km (13.5 km/h). See Table I for person-related characteristics, and Table II for shoe-person-related characteristics.
Table I. Descriptive of person-related characteristics. Sex Male Age (years) 25 Weight (kg) 75.7 Length (cm) 180.5 BMI (kg/m2) 23.2
Recommended shoe size length (cm) x 27 – 28.5
Recommended shoe size width x D
Training volume per week (km) 100 Foot Strike Type FFS
Achilles Tendon Resting Angle (angle) Left 60, Right 67 Achilles Tendon Length (cm) Left 21.8, Right 22 Have you heard that certain running shoes seem to have a
significant influence on performance? Yes
Race times 3k: 00:09:39, 10k: 00:35:59 List of current running shoe models Saucony: Kinvara 10,
Fastwitch; Adidas: Tempo List of previous running shoe models New Balance: 1080
x Measured using a Brannock Device; FFS = Forefoot Strike. Foot Strike Type assessed from 2D-video
13 Table II. Descriptive of shoe-related characteristics.
D C E B F A G I H
Price (USD) 140 190 138 120 250 13 250 13 295 MI (%) 46% 62% 48% 22% 40% 62% 38% 62% 34% Weight (g) 222 162 193 352 192 167 188 174 188 Stack Height Heel (mm) 27 26 33 37 33 25 38 25 38
Drop (mm) 10 4 12 8 8 10 10 10 10
Stability features (0-5) 2 1 0 4 1 1 0 1 1
Longitudinal Flex. (0-5) 4 3 3 3 0 5 0 5 0
Torsional Flex. (0-5) 3 3 3 2 2 4 1 4 1
Midsole TPU/EVA EVA ? EVA EVA EVA PEBA EVA PEBA Plate (Yes/No) No Yes Yes No Yes No Yes No Yes Weight measured by Mettler Toledo PB3002-S DeltaRange Balance. MI = Minimalist Index, higher value corresponding to more ‘minimalistic’ shoe. Stability features (0-5) = higher value corresponding to more stability features; Flex (0-5) = higher value corresponding to higher flexibility. TPU = Thermoplastic Polyurethane; EVA = Ethylene-Vinyl Acetate; PEBA = Polyether Block Amide. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
3.1 Running Economy
Both measures of central tendency and dispersion were descriptively different across the shoe conditions with regards to running economy. The mean energy cost of running for each of the shoe models ranged from 16.02 to 17.02 W/kg, with corresponding standard deviation ranging from 0.32 to 0.76. Ranking the shoes in order from highest to lowest energy cost revealed the following order (% difference compared with the aforementioned model): A, I 0.61%), H (-0.03%), F (-1.67%), B (-0.92%), G (-0.64%), C (-0.14%), D (-0.71%), E (-1.33%). The largest difference between the shoes were therefore between E and A, where the latter shoe cost 6.24% W/kg more than the former shoe.
The ‘sham condition’ shoe (I; Kalenji Run 100) was descriptively similar to the normal shoe (A; Kalenji Run 100), with an absolute and relative difference of 0.10 W/kg and 0.61% respectively. See Figure 1 for a graphical illustration of the results. For other cardiorespiratory variables, see Table III.
14
Figure 1. ‘Raincloud plot’ of running economy for each shoe condition. Black circle = Mean values;
Point range = Standard deviation; White circles = raw data points (sampled every 15th sec during the last two minutes of steady-state running). Light gray line = intersects mean values. Shoe order represent the order the shoes were tested in. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
Table III. Cardiorespiratory variables.
D C E B F A G I H VE 82.0 (4.5) 86.6 (6.0) 83.0 (4.2) 87.4 (3.8) 89.5 (2.7) 91.9 (2.4) 86.0 (3.7) 90.8 (4.1) 91.0 (3.9) BF 33.2 (2.6) 37.9 (3.7) 35.9 (1.7) 38.2 (1.9) 39.8 (1.5) 39.5 (1.6) 36.8 (2.1) 39.9 (1.9) 40.6 (1.4) VO2 Abs. 3.57 (.17) 3.60 (.15) 3.53 (.13) 3.62 (.11) 3.65 (.08) 3.72 (.09) 3.59 (.10) 3.70 (.09) 3.70 (.14) VO2 Rel. 47.4 (2.3) 47.7 (2.0) 46.9 (1.7) 48.2 (1.4) 48.7 (1.0) 49.8 (1.2) 48.1 (1.3) 49.6 (1.2) 49.7 (1.8) RER .893 (.014) .893 (.023) .863 (.016) .884 (.014) .884 (.015) .886 (.013) .870 (.015) .873 (.010) .864 (.019) HR 159.9 (1.5) 159.8 (1.7) 157.9 (0.8) 159.1 (1.1) 159.6 (1.3) 161.1 (1.7) 160.4 (1.3) 163.8 (1.4) 163.7 (1.9) Values are presented as mean (standard deviation). VE = Respiratory Minute Volume, BF = Breathing Frequency , VO2 Abs = Absolute VO2 (L/min), VO2 Rel = Relative VO2 (ml/kg/min), RER =
Respiratory Exchange Ratio, HR = Heart Rate. Shoe order represent the order the shoes were tested in. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
15
3.2 Spatiotemporal Variables
All parameters were overall descriptively similar. Median values for ground contact time were within the range of 219.6 – 222.8 milliseconds for all shoes except D and C which yielded median values of 212.5 and 216 respectively. Similarly, these two shoe models had the largest interquartile range of 7.8 and 9.3, while the other shoes had relatively smaller interquartile ranges of 1.7 – 4.3. For graphical illustration and descriptive pairwise comparison of the ground contact time data, see Figure 2 and Table IV.
Cadence data varied more between the shoes compared with the other biomechanical measures, and overlapped rarely from one test to the next. The participant ran mostly with cadence (median) values of 86.2 – 87.5 steps per minute per leg during the tests, although with two shoes the participant performed the tests with a slower cadence of 85 and 85.5 (C and G, respectively). The fastest cadence of 88.5 was observed for shoe A. The interquartile range for the cadence data varied from 0.2 to 0.6, with one of the shoes (C) having a wider interquartile range of 1.0. For graphical illustration and descriptive pairwise comparison of the cadence data, see Figure 3 and Table V.
The participant performed all but one test with average stride lengths of 1447 – 1478 millimeter. For the other test (shoe G), however, a higher average stride length of 1504 millimeter was observed. Similar standard deviations were noted between the tests (SD: 5.5 – 9.7), although the standard deviation for the test with shoe C were noticeably larger (SD: 14.7). For graphical illustration and descriptive pairwise comparison of the stride length data, see Figure 4, Table VI.
Vertical oscillation data for the shoes varied between (median) values of 112 – 117 millimeter for all shoes except A and C which yielded median values of 107 and 120 respectively. Furthermore, shoe model C had the highest interquartile range of 0.46, while data for the other shoes varied between 0.16 and 0.37. For graphical illustration and descriptive pairwise comparison of the vertical oscillation data, see Figure 5, Table VII.
16
Figure 2. ‘Raincloud plot’ with boxplots (median and interquartile range) of the ground contact data,
for each shoe condition. White circles = raw data points (sampled every 0.5th sec during the last two minutes of steady-state running). Light gray line = intersects median values. Shoe order represent the order the shoes were tested in. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
Table IV. Descriptive pairwise comparisons for Ground Contact Time D (Mdn = 212.5) C (Mdn = 216.0) E (Mdn = 221.0) B (Mdn = 221.0) F (Mdn = 222.0) A (Mdn = 222.3) G (Mdn = 222.6) I (Mdn = 222.8) C (Mdn = 216) -3.5 E (Mdn = 221.0) -8.5 -5.0 B (Mdn = 221.0) -8.5 -5.0 +0.0 F (Mdn = 222.0) -9.5 -6 -1.0 -1.0 A (Mdn = 222.3) -9.8 -6.3 -1.3 -1.3 -0.3 G (Mdn = 222.6) -10.1 -6.6 -1.6 -1.6 -0.6 -0.3 I (Mdn = 222.8) -10.3 -6.8 -1.8 -1.8 -0.8 -0.5 +0.2 H (Mdn = 219.6) -7.1 -3.6 +1.4 +1.4 +2.4 +2.7 +3.0 +3.2
The value reported is the median difference in ground contact time between the column and row shoe model. Positive value indicates column model > row model; negative value indicates column model < row model. Mdn = median. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
17
Figure 3. ‘Raincloud plot’ with boxplots (median and interquartile range) of the cadence data, for each
shoe condition. White circles = raw data points (sampled every 0.5th sec during the last two minutes of steady-state running). Light gray line = intersects median values. Shoe order represent the order the shoes were tested in. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
Table V. Descriptive pairwise comparisons for Cadence D (Mdn = 86.5) C (Mdn = 85.0) E (Mdn = 87.1) B (Mdn = 86.2) F (Mdn = 87.5) A (Mdn = 88.5) G (Mdn = 85.5) I (Mdn = 87.0) C (Mdn = 85.0) +1.5 E (Mdn = 87.1) -0.6 -2.1 B (Mdn = 86.2) +0.3 -1.2 +0.9 F (Mdn = 87.5) -1.0 -2.5 -0.4 -1.3 A (Mdn = 88.5) -2.0 -3.5 -1.4 -2.3 -1.0 G (Mdn = 85.5) +1.0 -0.5 +1.6 +0.7 +2.0 +3.0 I (Mdn = 87.0) -0.5 -2.0 +0.1 -0.8 +0.5 +1.5 -1.5 H (Mdn = 87.0) -0.5 -2.0 +0.1 -0.8 +0.5 +1.5 -1.5 +0.0
The value reported is the median difference in cadence between the column and row shoe model. Positive value indicates column model > row model; negative value indicates column model < row model. Mdn = median. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
18
Figure 4. ‘Raincloud plot’ of stride length for each shoe condition. Black circle = Mean values; Point
range = Standard deviation; White circles = raw data points (sampled every 5th sec during the last two minutes of steady-state running). Light gray line = intersects mean values. Shoe order represent the order the shoes were tested in. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
Table VI. Descriptive pairwise comparisons for Stride Length D (M = 1455.4) C (M = 1470.7) E (M = 1478.1) B (M = 1463.7) F (M = 1466.8) A (M = 1504.2) G (M = 1473.8) I (M = 1455.2) C (M = 1470.7) -15.3 E (M = 1478.1) -22.7 -7.4 B (M = 1463.7) -8.3 +7.0 +14.4 F (M = 1466.8) -11.4 +3.9 +11.3 -3.1 A (M = 1504.2) -48.8 -33.5 -26.1 -40.5 -37.2 G (M = 1473.8) -18.4 -3.1 +4.3 -10.1 -7.0 +30.4 I (M = 1455.2) +0.2 +15.5 +22.9 +8.5 +11.6 +49.0 +18.6 H (M = 1447.2) -8.2 +23.5 +30.9 +16.5 +19.6 +57.0 +26.6 +8.0 The value reported is the mean difference in stride length between the column and row shoe model. Positive value indicates column model > row model; negative value indicates column model < row model. M = mean. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
19
Figure 5. ‘Raincloud plot’ with boxplots (median and interquartile range) of the vertical oscillation data,
for each shoe condition. White circles = raw data points (sampled every 0.5th sec during the last two minutes of steady-state running). Light gray line = intersects median values. Shoe order represent the order the shoes were tested in. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
Table VII. Pairwise comparisons for Vertical Oscillation D (Mdn = 114.4) C (Mdn = 121.0) E (Mdn = 113.2) B (Mdn = 114.6) F (Mdn = 112.2) A (Mdn = 107.2) G (Mdn = 116.6) I (Mdn = 113.2) C (Mdn = 121.0) -5.6 E (Mdn = 113.2) +1.2 +6.8 B (Mdn = 114.6) -0.2 +5.4 -1.4 F (Mdn = 112.2) +2.2 +7.8 +1.0 +2.4 A (Mdn = 107.2) +7.2 +12.8 +6.0 +7.4 +5.0 G (Mdn = 116.6) -2.2 +3.4 -3.4 -2.0 -4.4 -9.4 I (Mdn = 113.2) +1.2 +6.8 +0.0 +1.4 -1.0 -6.0 +3.4 H (Mdn = 113.2) +1.2 +6.8 +0.0 +1.4 -1.0 -6.0 +3.4 +0.0
The value reported is the median difference in vertical oscillation between the column and row shoe model. Positive value indicates column model > row model; negative value indicates column model < row model. Mdn = median. D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, E = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, I = Kalenji Run 100 ‘Sham condition’, H = Nike Next%.
20
3.3 Foot Strike Type
The participant was classified with a forefoot strike (FFS) in their own shoes (Saucony Fastwitch), and FST varied between the study shoes from rearfoot strike (RFS) to FFS. While six shoes (DCEAGI) were classified as either RFS (I), midfoot strike (MFS) (DE), or FFS (CAG), two shoes were classified as both MFS and FFS (BF). Due to high camera battery consumption during this slow-motion video recording, the last shoe model (H) was not included as the camera ran out of battery. See figure 2 for a graphical illustration of initial-contacts for each shoe, and supplementary appendix 3 for a more detailed frame-by-frame analysis.
Figure 6. Graphical illustration of foot strike classification. Each picture represents a unique
initial-contact. RFS = Rearfoot Strike, MFS = Midfoot Strike, FFS = Forefoot Strike. Top row from left to right: MFS (Adidas Adizero Adios 5), FFS (Skechers GOrun Speed Elite), MFS (Xtep 160X), FFS (ASICS Gel-Foundation 13), MFS (ASICS Gel-Foundation 13). Bottom row from left to right: FFS (Brooks Hyperion Elite), MFS (Brooks Hyperion Elite), FFS (Kalenji Run 100 ‘normal condition’), FFS (Nike Vaporfly 4%), RFS (Kalenji Run 100 ‘sham condition’).
3.4 Self-rated Outcomes
Across all shoes, the average self-rated value for yield, energy return, weight, RPE, and comfort were 2.22, 2.67, 2.56, 11.44, 52.89 respectively. Greatest variance was observed for comfort (1623.86) with its lowest value of 3 and highest value of 99. Contrastingly, the lowest variance was observed for RPE (1.78) with most shoes being rated 11/20 (DEFGH), followed by 13/20 (BAI) and 9/20 (C). See Figure 6 for a graphical illustration of the results.
21
Figure 7. Self-rated A: Comfort (VAS, 100 mm), 0 = uncomfortable, 100 = most comfortable; B:
Perceived Exertion (Likert, 6-20), 6 = No exertion, 20 = Maximal exertion. C: Perception of Ride (Likert, 1-5), Yield (1 = Rigid, 5 = Flexible), Energy Return (1 = Responsive, 5 = Unresponsive), Weight (1 = light, 5 = heavy). D = Adidas Adizero Adios 5, C = Skechers GOrun Speed Elite, F = Xtep 160X, B = ASICS Gel-Foundation 13, F = Brooks Hyperion Elite, A = Kalenji Run 100, G = Nike Vaporfly 4%, H = Nike Next%, I = Kalenji Run 100 ‘Sham condition’.
3.5 Qualitative Assessment
After each shoe condition, the participant was given a shoe catalogue with pictures and names of running shoe models, and asked to select the model which the participant believed they were running in. This was as an attempt to see if the participant blinding could work. Out of 9 trials, the participant identified the correct shoe model during 2 trials in total, however it was also noted that the Kalenji Run 100 ‘sham condition’ (I) felt similar to Kalenji Run 100 ‘normal condition’ (A). The shoes correctly identified were the Nike Vaporfly 4% (F) and the Nike Next% (H). Trying to explore how the participant was able to correctly identify these shoes, without revealing to him that his guess was indeed correct, the participant was asked why he believed these were the models. Without much hesitation, the participant replied:
The Vaporfly 4%, the first of the two shoes to be tested.
(Paraphrased)“The upper material, with its compression-like garment, gave it
22
back part of the midsole had a flare extending further back which was noticeable when running. Therefore, I believe this to be the Vaporfly 4%.”
The Nike Next%, the second and last of the two shoes to be tested.
(Paraphrased)“The midsole material felt identical to the other shoe [referring to the Vaporfly 4%]” I previously tested. The upper material was different
however, more like regular shoes compared with the compression-like garment of the other shoe I previously tested. Therefore, I believe this to be the Next%.”
Further, after all tests were completed, a qualitative interview with the participant showed that he believed the blinding protocol was good, but that reflections in the glass walls in the room could potentially reveal the shoes the participant was being equipped with. Additionally, the participant suggested using a blindfold as an alternative strategy for blinding the participant while equipping the shoes and moving onto the treadmill.
4 Discussion
This is the first study to use a shoe-blinding protocol, descriptively examine a sham condition applied to a running shoe, compare several non-customized carbon-plated running shoes, assess perceptions of comfort and ride for a large heterogeneous set of shoe models, and compare carbon-fiber plated running shoes against a heavy motion-control-, and an extremely cheap shoe model. The descriptive data suggest that physiological and biomechanical measures could vary substantially between running shoe models, although statistical- and particularly causal inference was not possible for this pilot-study.
4.1 Blinding Protocol
The novel shoe-blinding protocol developed for this study may work well, with little risk of the participant visually detecting which models were being tested. This is a positive finding as previous studies on RE (Squadrone & , 2009; Mercer et al., 2018; Hoogkamer et al., 2018; Barnes & Kilding, 2018; Hunter et al., 2019) or biomechanical parameters (Squadrone & , 2009; Squadrone et al., 2015; Hollander et al, 2015; Au et al., 2018; Hunter et al, 2019; Day & Hahn, 2020) have not implemented strategies to achieve participant blinding, but future studies may now confidently do so. However, certain improvements can be made to the blinding protocol used in this study.
23
Firstly, the study-specific apparatus which was constructed in order to blind the participant while equipping the shoes, while working well, was at times awkward because the participant had to sit with an uncomfortable posture. Moreover, it was at times difficult for the participant to grab onto the laces of the shoes, possibly as a result of the unusual sitting posture. Overall, these minor flaws increased the time between trials and prolonged the total session duration. The alternative solution to use a blindfold instead, as suggested by the participant, would at first glance not guarantee proper blinding due to the risk of unintentionally missing a spot in the corner of their eyes; however, after further consideration, modifying a pair of ski-goggles could potentially reduce this risk.
Interestingly, the participant’s ability to correctly identify the Nike Vaporfly 4%, simply due to tactile sensation on top of the foot was not expected. In fact, in a yet unpublished study (preprint: Hébert-Losier et al., 2020) on running economy and performance, the researchers included the Nike Vaporfly 4%, spray-painted it black, and almost all participants (94%) were unable to identify the model despite its unusual upper-material construction being similar to a compression garment. Their sample of participants were novice runners, and this could suggest that to completely perfect participant-blinding when including shoes with unusual material constructions like the Nike Vaporfly 4%, one would have to exclude participants with knowledge of these models. If this is not practical, future studies comparing several running shoes of similar material design (e.g. Nike Vaporfly 4%, and Nike Next% which use the same cushioning material) should arguably separate their tests by a couple of days in order to have the participant be subject of recall-bias. Indeed, as was shown in this study, if the participant can correctly identify a model, and then equips another very similar model, then it is not surprising that the second model will be correctly identified as well. Further evidence that similar shoes will be recognized due to tactile sensation, despite visually blinding the participant to the models, was shown when the participant stated that the Kalenji Run 100 ‘sham’ and ‘normal’ condition felt similar.
4.2 Physiological and Biomechanical Parameters
Several unexpected findings were observed for RE. Firstly, the Next%, the newer version of the Vaporfly 4% did not return descriptively favorable data and was ranked 6th on RE, across all shoes. The Vaporfly 4% was ranked 4th on RE and was beaten by shoes with
TPU/EVA-24
midsoles, both with and without carbon-fiber plates. Based on material characteristics, it was expected to see the Vaporfly 4% and Next% ranked first. Indeed, if all things are equal, higher midsole energy return and increased LBS through the use of a carbon-fiber plate, should result in better performance. The descriptive difference in most parameters, between the ‘sham’ and ‘normal condition’ were negligible. For example, running economy were 0.61% lower in the ‘sham’, which falls within the range of measurement error (Foss & Hallén, 2005).
It was not unexpected to identify descriptive differences in the spatiotemporal parameters, as this has been found in previous studies (Squadrone & Gallozi, 2009; Squadrone et al., 2015; Hollander et al, 2015; Hunter et al, 2019; Day & Hahn, 2020). However, it was unexpected to see the shoe with highest cadence (Kalenji Run 100 ‘normal condition’) also had the highest stride length. This cannot be the case considering that the runner ran with the same velocity during all tries. Consequently, if all things are equal, an increase in cadence should result in a reduction in stride length (Schubert et al., 2014). A likely reason for this anomaly may be technical errors in the apparatus used, or limitations in its ability to obtain valid measurements, and thus interpreting these findings are difficult. Foot Strike Type findings were overall expected, as differences in ankle joint angles have been observed in previous research (Squadrone et al., 2015). However, it was surprising to see the habitual FFS participant use a RFS in the relatively minimalistic shoe (I), and use a FFS/MFS in the heaviest, thickest, motion control shoe (B). Normally, a more anterior FST (FFS/MFS) is expected the more minimalistic the shoe (Squadrone & Gallozi, 2009; Hollander et al, 2015), although at times individuals remain in their FST and only slightly change their ankle joint angle (Squadrone et al., 2015).
Although most of the descriptive differences were relatively small (e.g. a few millimeter or milliseconds in spatiotemporal variables) and the participant mostly remaining in his habitual FST, it can be speculated that these small changes could be of importance regarding RRI. For example, gait-retraining focusing on increasing cadence by 10% has shown to be associated with reductions in patellofemoral pain risk factors and self-rated pain scores, among runners with patellofemoral pain (Bramah et al., 2019). The most recent frameworks on RRI-etiology propose that tissue-load exceeding tissue-capacity is the causal mechanism behind RRI (Bertelsen et al., 2017; Edwards, 2018). It could therefore be speculated that slight changes in biomechanics (e.g. spatiotemporal variables and FST) may distribute load in different ways, enabling a runner to keep local tissue-load below its local tissue-capacity reducing the risk of RRI. While it is difficult to predict RRI based on crude biomechanics, running shoes or foot
25
type (Nigg et al., 2015; Napier & Willy, 2018), it was shown in a prospective cohort study that runners who alternated between a few shoes were associated with reduced risk of RRI, compared with runners who only used one model (Malisoux et al., 2015).
4.3 External Validity and Inter-Individual Variation
This study merely presents findings on the descriptive differences and similarities regarding physiological and biomechanical parameters across several shoe conditions for one individual. Furthermore, analytical and causal inference was not conducted, and it is not possible to extrapolate these findings onto other individuals. As previously mentioned, large inter-individual differences regarding one’s response to Nike Vaporfly 4% exist, ranging from 1.56 to 6.26% (Hoogkamer et al., 2018), 0.0 to 6.4% Hunter et al., 2019), and 1.72 to 7.15% (Barnes & Kilding, 2018). This variance may be explained by extrinsic and intrinsic factors, and likely exist for other modern performance shoes as well. Consequently, this limits our ability to generalize findings from one individual to another.
Foot Strike Type
For example, the only previous study on the Vaporfly 4% including a mix of FST found greater improvements in RE among RFS (4.63-4.78%) compared with MFS (3.50-3.67%), although this was not statistically significant (p = 0.0502) (Hoogkamer et al., 2018). It can be speculated that this is due to the new midsole material being mostly placed at the heel of the shoe, with improvements in its compliance (softness) and resilience being able to return to ~85% of the energy stored, as compared with TPU (~75%), and EVA (~65%) materials (Hoogkamer et al, 2018). Indeed, this would be especially plausible if the midsole material was solely responsible for the improvement in RE, however that is likely not the case.
Running Velocity, BMI, and other potential confounders
Aside from the cushioning material, another important component of the modern running shoes to improve RE is the carbon-fiber plate and its ability to increase the LBS of the shoe. Carbon-fiber plates are constructed in different shapes and thicknesses (Oh K & Park S, 2017), with varying effect on LBS. An individual’s optimal LBS may vary based on extrinsic or intrinsic variables such as running velocity (Day & Hahn, 2020, McLeod et al., 2020) or BMI (Roy & Stefanyshyn 2006), and it can therefore be speculated that the gain in RE from the modern performance shoes (e.g. Vaporfly 4%) will be different between runners with low and high
26
BMI, as well as those running in low and fast pace. Consequently, this may be a potential explanation to why the Vaporfly 4% and Next% were descriptively worse than the cheaper models with older material and lower LBS. It can be speculated that perhaps these cheaper shoes provided a lower but more favorable LBS for the participant. Since as previously stated, too high LBS can be detrimental (McLeod et al., 2020).
4.4 Novelty and Strengths
This is the first study to compare several purchasable shoe models using carbon-fiber plates without modifying them in any way, which increases the ecological validity of this study. Previous studies comparing several carbon-plated running shoes have used custom made models unavailable for purchase which limits the ecological validity of those studies. Furthermore, no previous published study has tested a sham-, and blinding procedure, which may be further improved and potentially used in future research.
4.5 Limitations
It must be recognized that there are several limitations in this study. Firstly, gas analyzer drift was not controlled for by re-calibrating the machine between the shoe tests, and it has been observed that gas analyzer drift could influence the obtained results, and thus negatively affect the internal validity of the study (Garcia-Tabar et al., 2015). Secondly, and similarly, the concurrent validity between the Garmin HRM-Run™ and more sophisticated motion capture systems (8-12 camera setups; e.g. Vicon, Qualisys) has been reported to be 0.931, 0.749, and 0.963 ICC for Cadence, GCT and VO respectively, with no reports regarding SL (Adams et al., 2016).
Thirdly, prolonged periods of exercise due to testing nine shoe conditions in one session could impose a cardiovascular drift (Coyle & González-Alonso, 2001) which has been found to influence VO2 in hot environments (Wingo et al, 2005), and could potentially have influenced the results although significant changes in VO2 have not been observed in temperate environments (Lafrenz et al, 2008; Wingo et al, 2012). Fourthly, and lastly, it cannot be disregarded that carry-over effects from the shoes may be present. Consequently, it cannot be concluded that these observed results would remain if it was tested on the same participant once more using a different sequence order. To control for this, while trying to assess the effects of the shoes, one must include a decent sample size and arrange the crossover trial’s experimental
27
sequences so that every shoe-condition precedes every other shoe-condition the same number of times (Williams, 1949).
In the example of a crossover trial with 4 shoe-conditions, the sequence order in which the participants should be randomized to should be ABDC, BCAD, CDBA, or DACB. The reason why the participant’s shoe-condition sequence order was DCEBFAGIH was because this project was initially planned to be a large-scale crossover study with participants randomized to one of many sequences (see Table VIII). Consequently, the participant in this study was randomized to one of these sequences and performed their tests before it was concluded that the project would be put on hold, and future bookings by other participants had to be cancelled due to Covid-19.
Table VIII. Shoe-condition sequences based on a Williams design. Shoe condition sequence (1) A B I C H D G E F Shoe condition sequence (2) B C A D I E H F G Shoe condition sequence (3) C D B E A F I G H Shoe condition sequence (4) D E C F B G A H I Shoe condition sequence (5) E F D G C H B I A Shoe condition sequence (6) F G E H D I C A B Shoe condition sequence (7) G H F I E A D B C Shoe condition sequence (8) H I G A F B E C D Shoe condition sequence (9) I A H B G C F D E Shoe condition sequence (10) F E G D H C I B A Shoe condition sequence (11) G F H E I D A C B Shoe condition sequence (12) H G E F A E B D C Shoe condition sequence (13) I H A G B F C E D Shoe condition sequence (14) A I B H C G D F E Shoe condition sequence (15) B A C I D H E G F Shoe condition sequence (16) C B D A E I F H G Shoe condition sequence (17) D C E B F A G I H Shoe condition sequence (18) E D F C G B H A I
4.6 Future Directions
Overall, the study-specific blinding procedure may work, but may also be improved in some regards. The field of footwear science is rapidly expanding with the new features in running shoes, which are not fully understood. Moreover, there are several research hypotheses which have not been adequately tested. This warrants further research in the field of footwear science,
