Characterization of trial duration in traditional and emerging postural control measures

Abstract

Researchers may select from varied technological and practical options when evaluating balance. Methodological choices inform the quantitative outcomes observed and allow practitioners to diagnose balance abnormalities. Past investigations have differed widely on sampling duration, and these discrepancies hinder comparisons among studies and confidence in outcomes where trials were excessively short. This study aimed to identify necessary trial lengths for common and emerging center of pressure-based measures. We hypothesized that dependent variables would fluctuate over time but eventually reach a stable magnitude. Ninety-seven apparently healthy adults performed quiet standing for 180-seconds (s) with eyes (A) open and (B) closed on a force platform. Anterior-posterior and medial–lateral elements of the center of pressure were used to calculate velocity, time-to-boundary, and Hurst exponents using 15, 30, 90, 120, 150, and 180 s of data. Two-way repeated measures ANOVAs were used to differentiate postural measures over time and between vision conditions. Outcomes were considered stable when significant changes in the measure were no longer observed in the time factor. Dependent measures stabilized for velocity between 60 and 120 s, time-to-boundary between 120 and 150 s, and the Hurst exponent between 30 and 120 s. Velocity measures stabilized quicker with eyes open, whereas vision had no effect or the eyes closed condition was faster to stabilize in time-to-boundary and detrended fluctuation analysis measures. We conclude that 150 s of standing data is sufficient to capture a broad range of postural stability outcomes regardless of vision condition.

Introduction

Maintaining stable posture or balance is a fundamental task that underlies the activities of daily living, largely contributing to the well-being, independence, healthspan, and lifespan of an adult human being. Static posturography has become an integral part of the balance assessment process, allowing researchers and clinicians to detect balance abnormalities and differentiate between populations (Blaszczyk, 2016, Prieto et al., 1996, Roman-Liu, 2018). Traditionally, quiet (i.e., static) standing is completed by tracking the center of mass or, more commonly, the center of pressure (CoP) over time. Force plates, force platforms, or balance plates derive the CoP coordinates, enabling the extraction of descriptive measures (Richmond et al., 2021a). These derived measures may be confounded by several methodological practices, including sample frequency (Rhea et al., 2015), stance (i.e., foot) positioning (Cobb et al., 2014, Kirby et al., 1987, Uimonen et al., 1992), noise (Park et al., 2011), and visual feedback (D'Anna et al., 2017). Experimental data and tradition have produced standard technological (e.g., sample frequency, duration, filtering) and participant-focused (vision, stance) static balance protocols for research and clinical practice. However, optimal trial duration is far from a settled debate.

The length of a standing trial is one suggested key regulatory mechanism of postural control (van der Kooij et al., 2011), influencing the reliability and validity of descriptive measures during quiet standing (Carpenter et al., 2001, Le Clair and Riach, 1996). For example, CoP velocity may require 60-seconds (s) of standing data for reliable outcomes (Carpenter et al., 2001, van der Kooij et al., 2011). When sample durations are extended beyond 60 s, the probability of incorporating both high and low-frequency components of the CoP signal is enhanced, as opposed to the increased prominence of higher frequencies during shorter sampling durations, indicating that an extension of the sample duration to at least 180-seconds is required for a reliable assessment of visual implications on quiet standing (van der Kooij et al., 2011) However, this recommended sample duration was based on outcomes from a limited sample size (n = 10) (van der Kooij et al., 2011) and further work seems warranted to investigate this question.

Prior examinations regarding trial duration in standing tasks have also prioritized comprehensive (e.g., position, velocity, or frequency) descriptive measures and do not advise on sample durations for advanced discrete (i.e., time-to-boundary (TtB)) static postural measures. TtB incorporates the instantaneous position and movement of the CoP relative to the base of support (Richmond et al., 2021a), representing the time window available to make corrective postural changes in the prevention of a fall (Hertel et al., 2006, Richmond et al., 2020, Slobounov et al., 1998, van Wegen et al., 2002), offering an alternative conceptual view of postural control and higher sensitivity compared to comprehensive measures (Whittier et al., 2020). Currently, there is little consensus on the approach to capture TtB. Prior TtB investigations have evaluated 10 s (Cobb et al., 2014, Hertel et al., 2006, Hertel and Olmsted-Kramer, 2007, McKeon et al., 2010, McKeon and Hertel, 2007), between 10 and 20 s (at least 10 s of data cropped) (Yamanaka et al., 2012), 20 s (Richmond et al., 2020), 30 s (D'Anna et al., 2015, DiLiberto et al., 2021, Richmond et al., 2021b, Wikstrom et al., 2010), and 60 s (D'Anna et al., 2017) of data. Several of these evaluations utilize multiple trials and produce aggregate outcomes from the repeated trials (Cobb et al., 2014, D'Anna et al., 2015, Hertel et al., 2006, Hertel and Olmsted-Kramer, 2007, Wikstrom et al., 2010, Yamanaka et al., 2012). However, averaging across multiple shorter trials (e.g., ten trials of 60-seconds) to produce aggregate outcomes do not overcome the limitations of measuring shorter compared to longer trial durations (van der Kooij et al., 2011), and to date, no work has established a minimum trial length for obtaining reliable TtB outcomes.

Finally, detrended fluctuation analysis (DFA) can be used to estimate the Hurst exponent of the increment process (velocity) for fractional Brownian motion (position) (Peng et al., 1995). DFA allows the sample to be broken down into segments (e.g., 1–5 s) to identify a scaling relation between the residual errors of the trendline and the time window of the sample (Delignières et al., 2011). Prior work utilizing this popular approach has recorded standing data for at least 20 s (Amoud et al., 2007), 30 s (Hansen et al., 2017), 65 s (Wang and Yang, 2012), or as long as 75 s (Kim et al., 2008). (van der Kooij et al., 2011) concluded that 60 s of data is appropriate to capture DFA measures. They further speculate that the DFA coefficients should be relatively insensitive to duration as long as the length of the longest time window used is still short relative to the trial duration (van der Kooij et al., 2011). The variance in these measures is inversely related to sample duration, as the underlying process is sampled over a greater number of data segments. (van der Kooij et al., 2011) suggests that the increased variance for shorter duration samples, specifically those under 180 s, may cause finding statistically significant effects of sample duration more difficult. There is room for additional exploration on this family of outcomes given that study’s limited sample size, as noted above.

Common and emerging CoP-based measures either lack experimental support for trial duration or, where recommendations are given, limited sample size restricts their generalizability. The overarching aim of this work was to advance the understanding of postural control by evaluating trial duration’s effects on CoP-based postural deficiency metrics in neurotypical adults. We hypothesized that dependent measures would either increase or decrease over time but trend toward a value that would remain stable for the rest of the trial. Implications of this research could affect the clinical assessment and diagnosis of balance impairments, in addition to the evaluation of rehabilitation approaches.