How long is required to undertake step variability analysis during running? A pilot study

Abstract

BACKGROUND:

The accurate assessment of step variability remains problematic.

OBJECTIVE:

To determine the minimum time required for assessing spatiotemporal variability during continuous running.

METHODS:

Seventeen endurance runners performed a running protocol on a treadmill, with a 3-min recording period at 12 km/h. Spatiotemporal parameters (contact and flight times, step length and step frequency) were measured using the OptoGait system and step variability was considered for each parameter, in terms of within-participants standard deviation (SD) and coefficient of variation (CV%). Step variability was calculated over 6 different durations: 0–10 s, 0–20 s, 0–30 s, 0–60 s, 0–120 s and 0–180 s.

RESULTS:

The repeated measures ANOVA revealed no significant differences between measurements in mean spatiotemporal gait parameters ( $p\geqslant$ 0.396, ICC $\geqslant$ 0.90 in all parameters). The post-hoc analysis confirmed no significant differences in step variability (of each spatiotemporal parameter) between measurements. The Bland-Altman limits of agreement method showed that longer recording intervals yield smaller systematic bias, random errors, and narrower limits of agreement.

CONCLUSIONS:

The duration of the recording interval plays an important role in the accuracy of the measurement (i.e. variability in spatiotemporal gait parameters), with longer intervals (180 s) showing smaller systematic bias and narrower limits of agreement than shorter intervals (10 s, 20 s, 30 s, 60 s or 120 s).

Keywords

Biomechanics endurance runners gait variability movement variability

1. Introduction

Step variability seems to be related to both injuries [1, 2] and endurance performance [3]. Nevertheless, the accurate assessment of step variability remains problematic. In 1995, Belli et al. [4] indicated that step variability during running was difficult to estimate due to the lack of measurement devices. Today, many devices provide real-time feedback on spatiotemporal parameters while running (e.g., OptoGait™, Stryd™ or Myotest™). Therefore, the limitation is not how to collect the data but how long should last the data collection to obtain accurate assessments of step variability.

Belli et al. [4] suggested that 32–64 consecutive steps are required to assess step variability, which represents approximately 15–20 s when running at submaximal velocities. To the best of the authors’ knowledge, the evidence available about how many steps or how long should last the data collection to obtain accurate assessments of step variability is limited, and no more studies have reconsidered this topic adapted to the new devices. However, some studies have addressed this analysis during walking [5, 6, 7]. A previous work examined the minimum number of steps required to accurately estimate spatial and temporal step kinematic variability of subjects walking on a treadmill [5], concluding that at least 400 steps are required. To further examine the step variability during running, the aim of this study was to determine the minimum time required for assessing spatiotemporal variability during continuous running on an instrumented treadmill. The authors hypothesised that the variability in spatiotemporal gait parameters during running would be very similar in short recording intervals (e.g. 10 s, 20 s, 30 s, 60 s or 120 s) compared to a longer recording interval (180 s).

2. Methods

2.1 Participants

Seventeen trained male endurance runners (age: 34 $\pm$ 7 years; height: 1.74 $\pm$ 0.04 m; body mass: 71.2 $\pm$ 4.3 kg) participated in this study. Participants met the inclusion criteria: (i) older than 18 years old, (ii) able to run 10 km in $<$ 40 min (36.1 $\pm$ 1.9 min), (iii) training on a treadmill at least once per week, (iv) free from injury (points 3 and 4 refer to the 6 months preceding the study). After receiving information on the objectives and procedures of the study, participants signed an informed consent form, which complied with the ethical standards of the World Medical Association’s Declaration of Helsinki (2013). The study was approved by the local ethics committee (San Jorge University, Zaragoza, Spain).

2.2 Procedures

Participants were tested on a motorized treadmill (HP cosmos Pulsar 4P, HP cosmos Sports and Medical, Gmbh, Germany). A standardized 10-min warm-up (running at 10 km/h) was performed since previous studies on human locomotion have shown that accommodation to a new condition occurs in $\sim$ 6–8 min [8]. After warming-up, running velocity was increased 1 km/h every min until a speed of 12 km/h was reached. The participants ran at 12 km/h for 3 min with a complete recording period. Note that 12 km/h is a normal pace for these athletes and is consistent with previous studies [9]. All participants verbally reported feeling comfortable running at the set speed.

2.3 Measures

Spatiotemporal parameters were measured using the OptoGait™ system (Optogait; Microgate, Bolzano, Italy), which was previously validated for the assessment of spatiotemporal parameters of the gait of young adults [10]. The OptoGait™ system is able to measure both contact time (CT) and flight time (FT) at 1000 Hz. The two parallel bars of the OptoGait™ system were placed on the side edges of the treadmill at the same level of the contact surface. CT, FT, step length (SL) and step frequency (SF) were measured for every step [11].

Step variability was assessed for each spatiotemporal parameter through the within-participant standard deviation (SD) and the coefficient of variation (CV%). Since previous studies have used indistinctly the SD [12] or CV% [3], we incorporated both measures to make comparisons easier. Step variability was examined over 6 recording intervals within the 3-min recording period: 0–10 s, 0–20 s, 0–30 s, 0–60 s, 0–120 s and 0–180 s.

Table 1
Descriptive values and association between the magnitude of the spatiotemporal parameters obtained from six time intervals

	Mean (standard deviation)						ICC (0–10 s	ICC (0–20 s	ICC (0–30 s	ICC (0–60 s	ICC (0–120 s
	0–10 s	0–20 s	0–30 s	0–60 s	0–120 s	0–180 s	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)
							(95% CI)	(95% CI)	(95% CI)	(95% CI)	(95% CI)
CT (s)	0.26 (0.02)	0.26 (0.02)	0.26 (0.02)	0.27 (0.02)	0.27 (0.02)	0.27 (0.02)	0.981 (0.951–0.993)	0.987 (0.965–0.995)	0.991 (0.976–0.997)	0.997 (0.992–0.999)	0.998 (0.997–0.999)
FT (s)	0.09 (0.03)	0.09 (0.02)	0.09 (0.02)	0.09 (0.02)	0.09 (0.02)	0.09 (0.02)	0.961 (0.900–0.985)	0.973 (0.931–0.990)	0.980 (0.950–0.952)	0.993 (0.981–0.997)	0.999 (0.998–1.00)
SL (cm)	118 (5)	118 (5)	118 (5)	118 (5)	118 (5)	118 (5)	0.996 (0.913–0.988)	0.980 (0.947–0.992)	0.984 (0.960–0.994)	0.994 (0.984–0.998)	0.999 (0.998–0.999)
SF (step/min)	170 (8)	170 (8)	170 (8)	170 (7)	170 (7)	170 (7)	0.968 (0.917–0.988)	0.979 (0.942–0.996)	0.985 (0.961–0.994)	0.994 (0.984–0.998)	0.999 (0.998–1.00)

Table 2

Descriptive values and association observed for the within-participants standard deviation of the spatiotemporal parameters obtained from six time intervals

	Mean (standard deviation)						ICC (0–10 s	ICC (0–20 s	ICC (0–30 s	ICC (0–60 s	ICC (0–120 s
	0–10 s	0–20 s	0–30 s	0–60 s	0–120 s	0–180 s	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)
							(95% CI)	(95% CI)	(95% CI)	(95% CI)	(95% CI)
CT (s)	0.006 (0.002)	0.006 (0.002)	0.007 (0.002)	0.007 (0.002)	0.007 (0.002)	0.007 (0.002)	0.905 (0.845–0.937)	0.907 (0.853–0.962)	0.962 (0.903–0.985)	0.985 (0.962–0.994)	0.996 (0.990–0.998)
FT (s)	0.008 (0.002)	0.008 (0.002)	0.009 (0.002)	0.009 (0.002)	0.009 (0.001)	0.009 (0.002)	0.904 (0.808–0.907)	0.901 (0.766–0.922)	0.909 (0.873–0.950)	0.936 (0.833–0.975)	0.991 (0.976–0.996)
SL (cm)	3.52 (1.18)	3.49 (0.98)	3.66 (0.96)	3.63 (0.95)	3.56 (0.96)	3.56 (0.91)	0.909 (0.850–0.929)	0.940 (0.847–0.977)	0.957 (0.890–0.983)	0.963 (0.907–0.986)	0.988 (0.969–0.995)
SF (step/min)	4.65 (1.54)	4.59 (1.44)	4.88 (1.38)	4.77 (1.29)	4.76 (1.30)	4.72 (1.27)	0.909 (0.861–0.941)	0.933 (0.829–0.974)	0.967 (0.915–0.987)	0.980 (0.948–0.992)	0.995 (0.987–0.998)

Table 3

Descriptive values and association observed for the coefficient of variation (%) of the spatiotemporal parameters obtained from six time intervals

	Mean (standard deviation)						ICC (0–10 s	ICC (0–20 s	ICC (0–30 s	ICC (0–60 s	ICC (0–120 s
	0–10 s	0–20 s	0–30 s	0–60 s	0–120 s	0–180 s	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)	vs 0–180 s)
							(95% CI)	(95% CI)	(95% CI)	(95% CI)	(95% CI)
CT (%)	2.4 (0.8)	2.3 (0.8)	2.5 (0.8)	2.6 (0.8)	2.6 (0.8)	2.7 (0.8)	0.901 (0.681–0.946)	0.904 (0.697–0.949)	0.964 (0.907–0.986)	0.987 (0.966–0.995)	0.996 (0.990–0.998)
FT (%)	10.3 (5.0)	10.1 (4.2)	10.9 (3.8)	10.2 (3.6)	11.4 (4.1)	11.4 (4.0)	0.905 (0.689–0.944)	0.901 (0.685–0.964)	0.959 (0.894–0.984)	0.982 (0.953–0.993)	0.998 (0.995–0.999)
SL (%)	3.0 (1.1)	3.0 (0.9)	3.1 (0.9)	3.1 (0.9)	3.0 (0.9)	3.0 (0.8)	0.901 (0.657–0.943)	0.948 (0.866–0.980)	0.961 (0.901–0.985)	0.968 (0.919–0.988)	0.990 (0.973–0.996)
SF (%)	2.7 (0.8)	2.7 (0.8)	2.9 (0.8)	2.8 (0.7)	2.8 (0.7)	2.8 (0.7)	0.901 (0.732–0.951)	0.927 (0.815–0.972)	0.964 (0.906–0.986)	0.977 (0.940–0.991)	0.994 (0.986–0.998)

2.4 Statistical analysis

Descriptive statistics are represented as mean (SD). Tests of normal distribution and homogeneity (Kolmogorov-Smirnov and Levene’s test, respectively) were conducted on all data before analysis. One-way repeated measures ANOVA with Bonferroni post-hoc corrections were conducted on the magnitude of each spatiotemporal parameter as well as on variability outcomes (i.e., SD and CV%) to examine possible differences between the recording intervals (0–10 s, 0–20 s, 0–30 s, 0–60 s, 0–120 s, 0–180 s). Effect sizes were calculated using partial eta squared (Eta ${}^{2}$ ) [13]. The association of the magnitude and variability of the spatiotemporal parameters between the recording intervals was quantified through the intraclass correlation coefficient (ICC). Based on the characteristics of this experimental design and following the guidelines reported by Koo and Li [14], the authors decided to conduct a ‘two-way random-effects’ model (ICC [2, k]), ‘mean of measurements’ type, and ‘absolute’ definition for the ICC measurement. The interpretation of the ICC was based on the benchmarks reported by a previous study [15]: ICC $<$ 0 reflects ‘poor’ reliability, 0–0.20 ‘slight’, 0.21–0.40 ‘fair’, 0.41–0.60 ‘moderate’, 0.61–0.80 ‘substantial’, and $>$ 0.81 ‘almost perfect’ reliability. The Bland-Altman [16] limits of agreement method (mean difference $\pm$ 1.96 SD) was used to examine differences in step variability (i.e., CV%) between the shorter recording intervals (0–10 s, 0–20 s, 0–30 s, 0–60 s and 0–120 s) and the longest interval (0–180 s) for each spatiotemporal parameter (CT, FT, SL and SF). Heteroscedasticity of error was defined as an $r^{2}>$ 0.1 [17]. The level of significance used was $p<$ 0.05. Data analysis was performed using SPSS (version 21, SPSS Inc., Chicago, IL, USA).

3. Results

The ANOVAs showed no significant differences in the magnitude of the spatiotemporal parameters between the recording intervals (CT: $p=$ 0.454, $F=$ 0.998, partial Eta ${}^{2}$ $=$ 0.263; FT: $p=$ 0.396, $F=$ 1.117, partial Eta ${}^{2}$ $=$ 0.285; SL: $p=$ 0.433, $F=$ 1.039, partial Eta ${}^{2}$ $=$ 0.271; SF: $p=$ 0.671, $F=$ 0.643, partial Eta ${}^{2}$ $=$ 0.187). An almost perfect association was also observed in the magnitude of the four spatiotemporal parameters between the recording intervals (ICC $>$ 0.90) (Table 3).

The ANOVAs conducted on the within-participants SD revealed significant differences for FT ( $p=$ 0.007, $F=$ 5.153, partial Eta ${}^{2}$ $=$ 0.648) and SF ( $p=$ 0.035, $F=$ 3.320, partial Eta ${}^{2}$ $=$ 0.543), while no significant differences were observed for CT ( $p=$ 0.172, $F=$ 1.825, partial Eta ${}^{2}$ $=$ 0.395) and SL ( $p$ = 0.167, $F=$ 1.851, partial Eta ${}^{2}$ $=$ 0.398). Regarding the ANOVAs conducted on CV% values, significant differences were observed for FT ( $p=$ 0.025, $F=$ 3.680, partial Eta ${}^{2}$ $=$ 0.568) and SF ( $p=$ 0.036, $F=$ 3.283, partial Eta ${}^{2}$ $=$ 0.540), but not for CT ( $p=$ 0.155, $F=$ 1.916, partial Eta ${}^{2}$ $=$ 0.406) and SL ( $p=$ 0.172, $F=$ 1.825, partial Eta ${}^{2}$ $=$ 0.395). Bonferroni pairwise comparisons never reached statistical significance. An almost perfect association between the recording intervals was observed for both the within-participants SD (ICC $>$ 0.90; Table 3) and the CV% (ICC $>$ 0.90; Table 3).

Bland-Altman plots revealed heteroscedasticity of error for the FT (0–10 s vs. 0–180 s, $r^{2}=$ 0.222; and 0–60 s vs. 0–180 s, $r^{2}=$ 0.199) and for the SL (0–10 s vs. 0–180 s, $r^{2}=$ 0.115), while heteroscedasticity of error was not observed for the rest of the recording intervals of those spatiotemporal parameters, and for CT and SF ( $r^{2}<$ 0.1). Longer recording intervals yield smaller systematic bias, random errors, and narrower limits of agreement for each spatiotemporal parameter.

4. Discussion

The results demonstrated that the duration of the recording interval plays an important role in the accuracy of the measurement (i.e. variability in spatiotemporal gait parameters), with longer intervals (180 s) showing smaller systematic bias and narrower limits of agreement than shorter intervals (10 s, 20 s, 30 s, 60 s or 120 s). However, based on the analysis of the magnitude of the differences (i.e. ICC) between shorter and longer recording intervals, the authors suggest that in some contexts where the accuracy requirements are not maximum and time-efficient methods are needed (e.g. clinical setting or big groups of athletes), shorter intervals allow sufficient accuracy to assess step variability. Additionally, to correctly interpret these results it is important to note that might be restricted to trained endurance runners [3] and to a running protocol performed on a treadmill at a fixed submaximal velocity [11, 18].

Despite the importance of step variability, very few studies have focused on determining how many steps are required to accurately estimate spatial and temporal step variability during running. Indeed, the authors found just one study from 1995 [4], which examined step variability in terms of step duration and vertical body displacement. Belli et al. [4] concluded that 32–64 consecutive steps ( $\sim$ 20–25 s) are required to accurately estimate step variability during running at submaximal velocities. Since our findings are expressed in time intervals (s) and in order to make comparisons easier, it is worth noting that 10 s of recording interval equals 27–30 step/min, 20 s equals 54–60 step/min, 30 s equals 81–90 step/min, 60 s equals 163–177 step/min, 120 s equals 326–354 step/min and 180 s equals 489–531 step/min (i.e. on average, participants run at 12 km/h with a SF of 170 $\pm$ 7 step/min).

Despite methodological differences, the current results are partially in line with those reported by Belli et al. [4], highlighting that mean spatiotemporal gait parameters during running, and variability in those parameters, might be assessed through the data collected over a time period of only 10 s (i.e., $\sim$ 30 steps) or, as long as the accuracy requirements are not maximum. Nevertheless, the Bland-Altman method conducted in the current study showed that longer recording intervals yield smaller systematic bias, random errors, and narrower limits of agreement so, if maximum accuracy is required (e.g., scientific approach), longer recording periods (e.g. 180 s), with a greater number of steps (e.g. $\sim$ 500), must be used. These results seem to be consistent with the results reported by previous studies focused on the analysis of step variability during walking. Owings and Grabiner [5] indicated that at least 400 steps are required to accurately estimate step variability, König et al. [6] proposed the collection of at least 50 step cycles, whereas Bruijn et al. [7] concluded that longer data series ( $>$ 300 steps) led to more precise estimates of variables referred to step variability.

Finally, some limitations need to be considered. First, the footwear was not standardized, but all runners wore their own footwear to increase the ecological validity of the study. Second, the protocol itself, with participants running on a treadmill at a fixed submaximal velocity. Notwithstanding these limitations, the authors consider that this study provides an answer to the initial question and it leaves some unanswered questions about the influence of the running velocity or level of exhaustion on step variability and the time required to accurately measures it.

In summary, the data suggest that the duration of the recording interval plays an important role in the accuracy of the measurement (i.e. variability in spatiotemporal gait parameters during running at 12 km/h), with longer intervals (180 s) showing smaller systematic bias and narrower limits of agreement than shorter intervals (10 s, 20 s, 30 s, 60 s or 120 s).

Therefore, from a practical standpoint, if maximum accuracy is required (e.g., scientific approach) longer recording periods must be used, but shorter recording intervals might be a time-efficient option for clinicians or coaches working with big groups of athletes, or where logistical issues difficult long-lasting assessment protocols (e.g., athletes with pain during running).

Footnotes

Conflict of interest

The authors declare no conflict of interest.

References

Meardon

Hamill

Derrick

TRTR

. Running injury and stride time variability over a prolonged run. Gait Posture 2011 Jan 1; 33(1): 36-40.

Mann

Malisoux

Nührenbörger

Urhausen

Meijer

Theisen

. Association of previous injury and speed with running style and stride-to-stride fluctuations. Scand J Med Sci Sports 2015 Dec; 25(6): e638-45.

Nakayama

Kudo

Ohtsuki

. Variability and fluctuation in running gait cycle of trained runners and non-runners. Gait Posture 2010; 31(3): 331-5.

Belli

Lacour

Komi

Candau

Denis

. Mechanical step variability during treadmill running. Eur J Appl Physiol Occup Physiol 1995; 70(6): 510-7.

Owings

Grabiner

. Measuring step kinematic variability on an instrumented treadmill: How many steps are enough? J Biomech 2003; 36(8): 1215-8.

König

Singh

von Beckerath

Janke

Taylor

. Is gait variability reliable? An assessment of spatio-temporal parameters of gait variability during continuous overground walking. Gait Posture 2014 Jan; 39(1): 615-7.

Bruijn

van Dieën

Meijer

Beek

. Statistical precision and sensitivity of measures of dynamic gait stability. J Neurosci Methods 2009; 178(2): 327-33.

Schieb

. Kinematic accommodation of novice treadmill runners. Res Q Exerc Sport 1986 Mar; 57(1): 1-7.

Morin

Samozino

Zameziati

Belli

. Effects of altered stride frequency and contact time on leg-spring behavior in human running. J Biomech 2007 Jan; 40(15): 3341-8.

10.

Lee

Song

Lee

Jung

Shin

. Concurrent validity and test-retest reliability of the OPTOGait photoelectric cell system for the assessment of spatio-temporal parameters of the gait of young adults. J Phys Ther Sci 2014; 26(1): 81-5.

11.

Roche-Seruendo

García-Pinillos

Haicaguerre

Bataller-Cervero

Soto-Hermoso

Latorre-Román

PÁ

. Lack of influence of muscular performance parameters on spatio-temporal adaptations with increased running velocity. J Strength Cond Res 2018 Feb; 32(2): 409-15.

12.

Paquette

Milner

Melcher

. Foot contact angle variability during a prolonged run with relation to injury history and habitual foot strike pattern. Scand J Med Sci Sports 2017; 27(2): 217-22.

13.

Levine

Hullett

. Eta squared, partial eta squared, and misreporting of effect size in communication research. Hum Commun Res 2002; 28(4): 612-25.

14.

Koo

. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med 2016 Jun; 15(2): 155-63.

15.

Landis

Koch

. The measurement of observer agreement for categorical data. Biometrics 1977; 33(1): 159-74.

16.

Bland

Altman

. Comparing methods of measurement: Why plotting difference against standard method is misleading. Lancet 1995; 346(8982): 1085-7.

17.

Atkinson

Nevill

. Statistical methods for assessing measurement error (reliability) in variables relevant to sports medicine. Sport Med 1998; 26(4): 217-38.

18.

Riley

Dicharry

Franz

Della Croce

Wilder

Kerrigan

. A kinematics and kinetic comparison of overground and treadmill running. Med Sci Sports Exerc [Internet] 2008 Jun [cited 2014 Jul 10]; 40(6): 1093-100. Available from: http://www.ncbi.nlm.nih.gov/pubmed/18460996.