Dynamic postural control and physical stress: an approach to determining injury risk in real sporting conditions

In the past few years, substantial work has been done on injury risk assessment and injury prevention. Different test batteries using strength, jump, or balance tests have been evaluated to assess an athlete’s potential injury risk, and different prevention programs have been developed. Nevertheless, the incidence of injury remains high, especially in sports games (Ekstrand, Spreco, Bengtsson, & Bahr, 2021). Therefore, there is still a need for appropriate diagnostic measures and empirically justified prevention programs to reduce the number of non-contact injuries. One of the main challenges in the currently used diagnostics is their restricted prognostic validity. Diagnostic methods need to be improved because the currently used diagnostic tests are not sufficient or fail to detect the potential deficits of an athlete and the deficits still occur despite (regular) diagnostic testing and preventive training (Bolt, Heuvelmans, Benjaminse, Robinson, & Gokeler, 2021; Verschueren et al., 2020).

In this context, in order to gain reliable information about the potential injury risk of an athlete, it is helpful to analyze the injury mechanisms and to identify possible risk factors. Most sports injuries occur without an opponent’s contact, mainly ligament ruptures (anterior cruciate ligament [ACL], ankle) or muscle injuries (Hewett, Ford, Hoogenboom, & Myer, 2010). Such non-contact injures mostly occur during dynamic actions, e.g., landing, cutting, or stopping, when the athlete must suddenly accelerate or decelerate or when the athlete must maintain stability in situations with the high(est) and often unilateral demands. Besides strength to compensate for the high demands that impact on the body, balance is a necessity in these situations (Güler et al., 2020). The athlete must maintain balance while transitioning from a dynamic to a static state, denoted as an athlete’s dynamic postural control (Johnston, Dolan, Reid, Coughlan, & Caulfield, 2018). “Dynamic balance involves maintaining an upright posture while (a) the center of gravity and base of support are moving and (b) the center of gravity is moving outside the base of support (for example, in walking)” (Yim-Chiplis & Talbot, 2000, p. 322). Hence, dynamic postural control seems to be important in sports, especially in regard to the large number of dynamic actions during competitions or matches (Whyte, Burke, White, & Moran, 2015). Additionally, impairment of dynamic postural control is associated with a higher risk for non-contact injuries (Butler, Lehr, Fink, Kiesel, & Plisky, 2013; Whyte et al., 2015; Wright, Lyons, & Navalta, 2013). Therefore, it might help assess an athlete’s dynamic postural control to obtain more insights into the potential injury risk (Plisky, Rauh, Kaminski, & Underwood, 2006). To this end, methods such as the Star Excursion Balance Test (SEBT) or the Y Balance Test (YBT) should be included in injury risk assessment (Plisky et al., 2009; Plisky et al., 2006).

Risk factors for non-contact injuries, such a dynamic postural control, are mainly tested when the athlete is in a recovered state. Nevertheless, most non-contact injuries occur under loaded conditions during matches, competitions, or sports training, when the athlete is physically stressed (Ekstrand et al., 2021). Attention should be paid that an applied external load, i.e., “the organization, quality, and quantity of exercise” (p. 270), leads to internal load, i.e., “the psychophysiological responses occurring during the execution of the exercise” (p. 270), (Impellizzeri, Marcora, & Coutts, 2019), causing, for example, an alteration of muscle activation patterns or a reduced level of voluntary muscle activation (Barber-Westin & Noyes, 2017; Santamaria & Webster, 2010). This, in turn, possibly affects risk factors and potentially increases the risk of injury. Therefore, investigating potential risk factors and injury mechanisms not only in recovered but also under loaded conditions is advisable so as to increase the prognostic validity of the diagnostic tests and to obtain more insight into an athlete’s potential injury risk under real sporting conditions (Bolt et al., 2021; Verschueren et al., 2020).

Several studies examined the influence of physical load¹ on potential risk factors. These studies show an influence of physical load on isometric strength of the legs (Verschueren et al., 2020) or hip and knee flexion angles (Benjaminse et al., 2008). No influence was shown on ground reaction forces (Barber-Westin & Noyes, 2017) or kinetic values (Santamaria & Webster, 2010). Regarding dynamic postural control, several studies showed a negative effect of different types of load—e.g., cycling (Heil, Schulte, & Büsch, 2020b; Johnston et al., 2018), running (Wright et al., 2013), a high-intensity, intermittent exercise protocol (HIIP) (Whyte et al., 2015), functional protocols (Sarshin, Mohammadi, Shahrabad, & Sedighi, 2011), or local loading protocols (Gribble, Robinson, Hertel, & Denegar, 2009)—on dynamic postural control. This might be due to the former prescribed load-induced physiological alterations that also affect systems involved in the maintenance of dynamic postural control, e.g., proprioception, vision, and the vestibular system (Güler et al., 2020). This might lead to neuromuscular impairment and a loss of sensorimotor control, resulting in a decrease of dynamic postural control (Steib, Zech, Hentschke, & Pfeifer, 2013; Taylor & Gandevia, 2008; Whyte et al., 2015; Zech, Steib, Hentschke, Eckhardt, & Pfeifer, 2012). Nevertheless, there are also differences between the results of the aforementioned studies, and diverging and conflicting results in other studies. Zech et al. (2012), e.g., found no effect of a running protocol and no effect of a local loading protocol on dynamic postural control of healthy handball players. Additionally, in a recent study by Verschueren, Tassignon, Verhagen, and Meeusen (2021), no effect of a Wingate anaerobic protocol on dynamic postural control measured with the YBT was found in contrast to the results obtained by Johnston et al. (2018).

The aforementioned results show that physical load can change the potential injury risk of an athlete, but they also show that physical load is not equal to physical load and that the results are possibly influenced by the type of load and the applied protocol. The type of load, meaning the type of exercise implemented to physically stress the participants, varies a lot in the aforementioned studies and ranges from cycling or running to local loading protocols. Moreover, also in the same load type, different protocols with different durations or distances are used. Such differences in the load types and protocols possibly explain the ensuing differences in the results. Different load types and protocols probably lead to different resulting internal loads because different body systems and structures (vestibular, visual, proprioception, sensorimotor, muscular) are not stressed equally due to the varying external loads or load types (Impellizzeri et al., 2019).

This assumption is also supported in studies directly comparing the influence of two load types on the same parameter or risk factor. The results mainly show differences dependent on the load type and protocol, e.g., a greater decrement of dynamic postural control after running compared to cycling (James, Scheuermann, & Smith, 2010a; James, Scheuermann, & Smith, 2010b; Lepers, Bigard, Diard, Gouteyron, & Guezennec, 1997; Wright et al., 2013). Looking at this possible load-dependence of the changes, it seems logical to choose load types and protocols according to the demands of the sport that should be investigated to draw conclusions about the changes of potential injury risk under real sporting conditions (Benjaminse, Webster, Kimp, Meijer, & Gokeler, 2019; Bolt et al., 2021). In this context, cycling protocols seem to be easier to apply and to control in a laboratory setting. Nevertheless, besides cycling itself, most sports require running as a central element, e.g., in basketball, soccer, or track and field sports. Therefore, running protocols seem more suitable for reflecting the demands of sports and it might be better to include running protocols in injury risk assessment. Additionally, comparisons between different protocols might help detect differences in the resulting internal load and the use of different protocols can help to depict different aspects and situations of a sport that might occur during training, matches, or competitions.

In this context it must also be concerned that the resulting internal load is not only influenced by the load type or protocol but also by other factors, e.g., the task or procedures to measure a certain risk factor, sports-related factors, or individual characteristics that should be addressed and/or controlled (Fig. 1). Therefore, studies and injury risk assessment should gradually approximate real, i.e., representative, sports conditions by systematically varying and adapting factors shown in Fig. 1, e.g., the loading profiles, i.e., load types, duration etc., and tasks.

It was shown that in the context of injury prevention there is a need for appropriate diagnostic measures to assess an athlete’s potential injury risk. Thereby, the assessment of dynamic postural control should be one of the essential factors. Moreover, assessing an athlete’s potential injury risk under loaded conditions might provide additional insight and help to improve the prognostic validity of the diagnostics. In this context, the use of representative designs depicting the approximate real sport conditions is advisable. Nevertheless, although most sports require running as a central element, many studies still use cycling protocols to physically stress the participants, and studies using running in a systematic way are scarce. Therefore, the aim of the current study was to examine the influence of a running protocol on dynamic postural control and to compare it with a comparable widely used cycling protocol.

We hypothesized that there would be a decline of dynamic postural control after both types of load (H1). Moreover, we hypothesized that the impairments would be greater after running than after cycling (H2). Additionally, the recovery period after the load was regarded to examine how long the changes persist after the loading protocol. Accordingly, a successive increase of dynamic postural control during the recovery period was hypothesized (H3).

A sample size of n = 126 was determined a priori with a power estimation (F test: η²_p = 0.20, α = 0.01, 1 − β = 0.99) for a multivariate three-way mixed analysis of variance (MANOVA) using G*power software (vers. 3.1.9.7; Faul, Erdfelder, Lang, & Buchner, 2007). In total, 128 active and healthy people (64 males and 64 females, age: 23.64 ± 2.44, height: 176.54 ± 8.96 cm, weight: 68.85 ± 10.98 kg) with no history of musculoskeletal injury in the previous 6 months participated in the study (Table 1). The participants were randomly divided into two examination groups. Each group completed a comparable loading protocol. Group 1 completed the protocol on a bicycle ergometer and Group 2 on a treadmill.

Before testing, participants had to complete the Physical Activity Readiness Questionnaire (PAR‑Q; Warburton et al., 2011), and those answering “yes” to any question were excluded from participating. Additionally, participants were excluded if they were taking medication for a balance disorder or suffering from one of the following: balance disorder, vestibular or visual impairment, cardiovascular disease, previous reports of chest pain, neurological disease, or chronic ankle instability.

Testing was performed in a laboratory setting. Each participant was tested in one 90-min session, including administration, anthropometric measurements, familiarization, and testing procedures. At first, participants were informed about the procedures and provided written informed consent to the following experiment. Additionally, personal data, sporting background, injury history, and the questions from the PAR‑Q (Warburton et al., 2011) were recorded in a questionnaire. The kicking leg and standing leg of a person were prompted. Individuals not fulfilling eligibility criteria were excluded from the study.

Anthropometric measurements (weight, height, leg length) were recorded using the Inbody270 (InBody Co., Seoul, South Korea), a stadiometer (Seca GmbH & CO. KG, Hamburg, Germany), and measuring tape. Leg length was measured while the participant was standing in front of a wall. Leg length was defined as the distance between the participant’s anterior–superior iliac spine and the most distal part of the medial malleolus (Gribble & Hertel, 2003; Plisky et al., 2009). Since both legs are involved in the sporting movements, and injuries can occur in both legs (Paterno et al., 2010), measurements were taken for both legs.

To get familiar with the YBT and to minimize potential learning effects (Gribble, Hertel, & Plisky, 2012), each participant performed four practice rounds of the YBT before testing. Each round lasted about 60 s and consisted of one trial on each leg. The starting leg varied between the participants. In each group, one half of the subjects started the round standing on their preferred kicking leg, while the other half started the measurements standing on their preferred standing leg. After a short resting period, testing started with three YBT rounds at baseline, with a time of 10 min rest between the rounds, 20 min pre-load (pre01), 10 min pre-load (pre02), and immediately pre-load (pre03). Afterward, participants completed one of the two protocols. Directly after the loading protocol, participants went back on the YBT. Post-load measurements were identical to the pre-load measurements. All participants performed three rounds of the YBT again, with 10 min rest in between, immediately post-load (post01), 10 min post-load (post02), and 20 min post-load (post03). The design and procedures were adapted and expanded from the study by Johnston et al. (2018).

The reach distance in each direction was measured and normalized to leg length using the following equation (Plisky et al., 2009):

Moreover, a normalized composite score (CS) was computed for each leg (Plisky et al., 2009):

The data were analyzed using SPSS (version 28.0, IBM Corporation, Armonk, NY, USA). First, the normal distribution of the data was checked using the Shapiro–Wilk test. Reliability of the data was assessed using different values: The intraclass correlation coefficient (ICC 3, 1) with absolute agreement (Shrout & Fleiss, 1979) was calculated across the three baseline measurements of the YBT to determine the reliability of the normalized reach distances. The values were interpreted according to Koo and Li (2016) as > 0.9 = excellent, 0.75–0.9 = good, 0.5–0.75 = moderate, and < 0.5 = poor. Within-session reliability was assessed using the coefficient of variation (CV) calculated as CV = (SD / M) × 100. All CV values < 10% were acceptable according to Cormack, Newton, McGuigan, and Doyle (2008). Moreover, the standard error of measurement (SEM) was calculated as SD × √1 − ICC to assess the degree of variation between the repeated measures. Additionally, the minimal detectable change (MDC₉₀) was calculated according to Schmitt and Di Fabio (2004) representing the value that can be considered as real change beyond measurement error with 90% confidence.

Due to given multicollinearity between the three reach directions and the composite score (Tabachnick & Fidell, 2013), instead of a multivariate three-way mixed ANOVA, a univariate three-way mixed ANOVA (load type × leg × time) with time and leg as within-subject and load type as between-subject factor, as well as with a Greenhouse–Geisser correction, if necessary, was conducted to compare the normalized reach distances in each direction and the composite score between the two groups and between both legs. The significance level was set at p < 0.01. Additional contrast analyses for repeated measures were conducted to specify the differences between adjacent points of time. For the mixed ANOVA conducted, η²_p is stated as effect size. Additionally, 90% confidence intervals (CIs) for the η²_p were calculated using an SPSS syntax by Wuensch (2016). Additionally, effect sizes for repeated measures (Cohen’s d_z) and 95% CIs for the adjacent points of time were also calculated with the syntax by Wuensch (2012).