Recurrence Quantification Analysis as a Form of Postural Control Assessment: A Systematic Review

Abstract

Human postural control is commonly assessed by center of pressure (CoP) displacement analysis. However, traditional linear parameters do not provide a complete picture of postural control, so a number of nonlinear analyses have been introduced. One of these is recurrence quantification analysis (RQA), which is used to determine the number and duration of repeated states in a dynamic system. This review aimed to show how the RQA measures look in different groups of subjects when assessing postural stability during quiet standing and how the authors interpret them. Therefore, a systematic review was conducted of papers published from 2000 to February 2023. Searched databases were PubMed, ScienceDirect, and EBSCO. Twenty-eight publications were included in this review. The RQA parameters most commonly found in papers are %DET (determinism), ENT (Shannon entropy), %REC (recurrence), and %LAM (laminarity). %LAM is the most sensitive factor in determining differences in CoP position between different age groups, as it describes motion fluidity. The vision affects the deterministic structure of CoP motions. When the sensory organization test conditions became difficult, CoP variability increased, while the %DET decreased. It was concluded that traditional and nonlinear methods provide complementary and not redundant information for assessing age- and health-related changes in standing balance.

1. Introduction

The phenomenon of postural control requires a discussion of two crucial elements. The first refers to defining balance while maintaining an upright standing position, and the second refers to evaluation methods. Gerbino et al. [1] define standing balance as the capacity to stand with as little sway as possible. The general tendency for postural sway to increase with age, pathology, or when standing tasks come with a sensory challenge has been repeatedly highlighted and proven in the literature [2,3,4]. Such results appear to be influenced by many factors. Thus, the primary systems involved in postural and balance control are visual, vestibular, and somatosensory. The central nervous system (CNS) regulates sensory information from all these systems to produce adequate motor power to maintain a controlled, upright posture. However, how the nervous system handles this information describes postural control [3].

Assessment of postural control involves the platform’s data captured during various balance tasks. Linear and nonlinear parameters are calculated from time series describing the displacement of the foot’s center of pressure separately for the mediolateral and anterior-posterior directions. The linear analysis describes the quality of motion, magnitude, and/or variance of CoP displacements (CoP path length, CoP velocity, sway range, and area of the ellipse). However, this is not enough information, given how complex the process has to be described. Therefore, the classical linear methods have been recently increasingly supplemented by nonlinear methods. Nonlinear measures provide information about the temporal organization of the variation in CoP displacement regarding motor behavior over time. Nonlinear measures include the entropy family, the fractal dimension, the largest Lyapunov exponent, the Hurst index, and recurrence quantification analysis (RQA). In their review, Kędziorek and Błażkiewicz [3] showed the use of entropy, fractal dimension, and the Lyapunov exponent to evaluate postural control. It seems that the next step should be to demonstrate the application of RQA analysis to assess the complexity of this phenomenon.

1.1. Recurrence Quantification Analysis Parameters

Recurrence quantification analysis (RQA) is a nonlinear method of describing signal dynamics by determining the duration and number of repetitions of a dynamic system represented by its trajectory in phase space [5,6]. The concept of phase space and recurrence was developed in the late 19th century by Henri Poincaré [7]. Eckmann et al. [8] included these concepts in a qualitative tool—the recurrence plot (Figure 1). A few years later, Zbilut and Webber [9,10] quantified the recurrence plot by defining five variables: % of recurrence (%REC), % of determinism (%DET), the length of the longest diagonal line segment in the plot (LMAX), entropy (ENT), and trend of recurrences (TREND) [11]. Later, Marwan et al. [12] added three new variables: % of laminarity (%LAM), trapping time (TT), and divergence (DIV). The listed measures are the results of the density of the recurrence points and the structure of the diagonal and vertical lines of the recurrence plot (Figure 1). An important argument in favor of using RQA is that this technique does not require normal distribution or data stationarity. RQA involves the construction of a recurrence plot of the analyzed signal (Figure 1), from which the quantitative measures listed above and described in detail below are extracted.

1.1.1. Recurrence Rate (REC)

REC is a nonlinear autocorrelation measure (% recurrence) indicating the degree to which data points repeat themselves in the time series, with repetition defined in terms of proximity in reconstructed multidimensional phase space. Thus, REC describes the percentage of recurrent points lying within a specified radius r(i). This variable can range from 0% (no recurrent points) to 100% (all points are recurrent) [11,13]. REC is also known as the density of recurrence points in the recurrence plot.

$$\mathrm{REC}=\frac{1}{N^2}{\sum }_{i,j=1}^{N}R\left(i,j\right)$$

where $R\left(i,j\right)=\Theta \left(r\left(i\right)-\Vert x\left(i\right)-x\left(j\right)\Vert \right)$; i, j = 1, 2, …, N; N—number of points on the phase space trajectory; $r\left(i\right)$—specified threshold, $\Vert \Vert$ —the norm or metric (in this study, the Euclidean norm was calculated), and $\Theta \left(\right)$—is the Heaviside function, defined as:

$$\Theta \left(x\right)=\{\begin{array}{c}1, x\ge 0\\ 0, x<0\end{array}$$

The 0 and 1 values of the Heaviside function are represented in white and black, respectively. According to Zbilut et al. [14], the r(i) value was 1% of the maximum phase space diameter.

Selection of the radius threshold is necessary because two points will never lie exactly together in the reconstructed phase space, and thus the threshold permits two locally similar points to be considered recurrent, and any points outside the threshold will not be summed into the percent recurrence of the signal. The radius threshold can have a great impact on the output variables of RQA, so guidelines have been proposed such that the appropriate radius dimension maintains the total percentage of recurrent points under 5% [11,15]. Moreover, a large radius threshold that saturates determinism or recurrence should be avoided.

1.1.2. Determinism (DET)

DET is the degree to which recurrent data points generate sequences of recurrent data, demonstrating the deterministic structure of the time series. DET determines the proportion of recurrent points lying along diagonal line structures, except those within the main diagonal [16]. The diagonal line segments must have a certain minimum length in order not to be excluded. Usually, this is set as two adjacent recurrence points without intervening white space (one can choose larger, more conservative values).

$$\mathrm{DET}=\frac{{\sum }_{l=l_{min}}^{N}lP\left(l\right)}{{\sum }_{i=1}^{N}lP\left(l\right)}$$

where $P\left(l\right)$—histogram of the lengths $l$ of the diagonal lines; $N$—number of points on the phase space trajectory.

%DET reflects the degree of determinism observed because, as stated above, upward diagonal line segments indicate that the system is repeatedly revisiting the same region of the attractor (or of the reconstructed space). A higher DET value indicates a more predictable, more deterministic, and less random CoP motion compatible with better equilibrium performance. A lower DET value suggests a less predictable, less deterministic, and more random CoP movement, consistent with lower balance efficiency [17].

1.1.3. Average and Maximal Diagonal Line Length (LMEAN/LMAX)

LMAX is the length of the longest diagonal line in the plot, excluding the main diagonal. According to Schmit et al. [18], LMAX is a measure of mathematical stability, not equivalent to postural stability. Mathematical stability refers to the response of a dynamic system to a change in initial conditions. Therefore, the lower the value of LMAX, the more chaotic the analyzed signal is [11,13].

$$\mathrm{LMAX}=max\left(\left\{l_i;i=1,\dots N_l\right\}\right);$$

where $l_i$—diagonal structures; $N_l$—number of diagonal lines in the recurrence plot.

In addition, Webber et al. [19] also report the LMEAN parameter. LMEAN is the average length of all diagonal lines, interpreted as the average prediction time. According to van den Hoorn et al. [20], a higher value of LMEAN means less impact of small perturbations on equilibrium maintenance, resulting in more similar temporal dynamic CoP patterns. A lower LMEAN value denotes reduced equilibrium performance, with a larger influence of small perturbations, resulting in less similar (more random) CoP dynamic patterns.

1.1.4. Divergence (DIV)

DIV is the inverse of LMAX and is related to the sum of positive Lyapunov exponents.

1.1.5. Shannon Entropy (ENT)

ENT is the Shannon entropy of the probability $p\left(l\right)=\frac{P\left(l\right)}{N_l}$ to find a diagonal line of length $l$ in the recurrence plot among the total number of diagonal lines $N_l$ [11,13]. ENT demonstrates the complexity of the recurrent trajectories versus the diagonal lines. Thus, ENT is a measure of the complexity or irregularity of the deterministic structure of the data.

$$\mathrm{ENT}=-{\sum }_{l=l_{min}}^{N}p\left(l\right)lnp\left(l\right)$$

According to Bonnette et al. [21], higher values of Shannon entropy indicate that the distribution of the lengths of diagonal lines or the lengths of time periods of repeated postural sway has become less predictable. In this case, the postural system repeats many different subsets of the length of its behavior over time.

1.1.6. Laminarity (LAM)

LAM is similar to DET. LAM measures the percentage of repeated points containing vertical line structures, not diagonal lines such as in the case of DET.

$$\mathrm{LAM}=\frac{{\sum }_{v=v_{min}}^{N}vP\left(v\right)}{{\sum }_{v=1}^{N}vP\left(v\right)}$$

where $P\left(v\right)$—is the histogram of the lengths of the vertical lines; $N$—number of points on the phase space trajectory.

LAM quantifies the ratio of recurrence points belonging to laminar structures against the total frequency of recurrence points and is therefore considered an overall measure of signal stability. According to Bonnette et al. [21], a higher value of laminarity indicates that a higher proportion of recurrent points form periods of unchanging postural sway behavior. Higher LAM values indicate more interrupted CoP motion with more periods of minimal CoP fluctuations. Low LAM values indicate less discontinuous CoP motion with fewer periods of minimal CoP fluctuations [20].

1.1.7. Trapping Time (TT)

Trapping time (TT) is the average length of recurrent points that form a vertical line [6,11,13].

$$\mathrm{TT}=\frac{{\sum }_{v=v_{min}}^{N}vP\left(v\right)}{{\sum }_{v=v_{min}}^{N}P\left(v\right)}$$

TT describes the average length of laminar (vertical/horizontal) structures and is analogous to how the mean diagonal line length captures periodic durations. Higher trapping time indicates that postural sway displays, on average, longer-lasting sequences of invariant behavior [21], while lower values denote shorter periods of CoP fluctuations.

1.1.8. Trend of Recurrences (TREND)

TREND quantifies the degree of deviation of the recurrence plots from the main diagonal and assesses the system’s non-stationarity [18]. If the recurring points are homogeneously distributed in the recurrence graph, TREND values will be near zero units. TREND is calculated as the slope of the line of best fit drawn for %REC as a function of distance from the main diagonal [11].

TREND is expressed in units of %REC per 1000 data points. TREND values will usually be negatively signed (if %REC decreases with increasing distance from the diagonal, the regression line will have a negative slope).

$$\mathrm{TREND}=1000\left(slope of \%local recurrence vs. displacement\right)$$

At this point, it is worth noting that the RQA method, due to so many parameters, allows the assessment of such features as determinism, dimensionality, the complexity of the dynamics of the analyzed system, and the amount of patterning it holds [22]. Moreover, it is remarkably robust for short, noisy, and non-stationary signals [11,23,24,25].

1.2. Embedding Parameters for RQA Analysis

RQA requires reconstructing the state space of a dynamic system from a single scalar time series through an embedding process [26,27]. The embedding method uses time-delayed versions of the original time series to reconstruct the state space, which involves the choice of time delay (tau) and embedding dimension (D). The mathematics behind this technique was presented in detail by Takens [26]. The embedding dimension (D) is the minimum number of variables required to create a new state space from a given time series. The vectors in this new embedding space are built from the time-delayed values of the scalar measurement [27,28]. Many methods have been developed for estimating embedding dimensions [29,30,31,32]. Of all the possible techniques, the most prominent are the correlation dimension [33] and the false nearest neighbors (FNN) [34,35]. The average mutual information function (AMIF), time series autocorrelation (ACF), or average displacement (AVD) method is a common and classical approach to time lag selection [35,36,37,38]. However, with both the choice of delay time and the embedding dimension, it is unclear which method is the best. The methods mentioned are briefly described below.

1.2.1. Time Lag Determination Methods

Average Mutual Information Function

Fraser and Swinney [36] developed a method for finding optimal time delays based on time-delayed coordinates that are as independent as possible of each other. They defined this relationship as the mutual information AMIF(x(t), x(t + tau)) between the original time series x(t) and the tau-shifted time series x(t + tau). Since mutual information is calculated for a time series and a time-shifted version of the same time series, it is called mutual auto-information or average mutual information (AMIF). To measure AMIF, a data histogram is created using bins, where pi is the probability that the signal has a value inside the i-th bin, p(tau) is the probability that x(t) is in the i-th bin and x(t + tau) is in bin j.

$$\mathrm{AMIF}\left(tau\right)={\displaystyle \sum }_{i,j}{p}_{ij}\left(tau\right)log\left(\frac{p_{ij}\left(tau\right)}{p_i{p}_j}\right)$$

Fraser and Swinney [36] proposed using the position of the first minimum of AMIF(x(t), x(t + tau)) as the optimal value of tau to obtain the most independent coordinates for time-delayed phase space embedding. However, AMIF may not have a local minimum. Therefore, other criteria were developed, such as the smallest value of tau for which the AMIF function falls below 1/e [35].

Time Series Autocorrelation Method

Autocorrelation (ACF) is the correlation of a signal with a delayed copy of itself as a function of delay. The lag is chosen as the first zero-crossing of the autocorrelation function for the data or, if there is no zero-crossing, at which the first local minimum of the ACF occurs [15].

Average Displacement Method

The AVD, like the autocorrelation method, involves comparing a time series to its lagged version at different time lags to determine the optimal time lag for analysis. It can be calculated by taking the absolute value of the difference between the corresponding values in the two series and then taking the average of those differences [38]. The time lag corresponding to the minimum average displacement can be considered the optimal time lag for analysis. Otherwise, the tau value at which the slope of the AVD curve decreases by 40% of the initial slope is selected, as recommended by Rosenstein et al. [38]. In summary, this method quantifies the extension of the reconstructed trajectory from the line of identity as a function of tau and seeks to achieve a balance between the error associated with redundancy and irrelevance.

1.2.2. Embedding Dimension Determination Methods

False Nearest Neighbors

This section introduces the false nearest neighbor algorithm according to the idea proposed by Kennel et al. [32]. Let us assume that two data points in a one-dimensional time series are close to each other—then they are neighbors. Their difference in magnitude gives us the distance between these neighbors. If we embed this time series (first in two dimensions) with a certain time delay, tau, then we can use the coordinates of these data points to examine whether the distance between them has changed significantly. If this embedding has changed the distance between the neighbors significantly, they are false neighbors, indicating the need for further data embedding in subsequent dimensions. In case the distance between them does not change significantly, they are called true neighbors. At that point, the embedding leaves the attractor shape unchanged, which means that the current embedding dimension is sufficient. The procedure repeats successively by increasing the embedding dimension D. The value of D is chosen at the point where the number of FNN falls to 0 (Figure 2).

Correlation Dimension

Grassberger and Procaccia [30,39] in 1983 introduced a fractal dimension measurement called the correlation dimension (D₂). D₂ is created by covering the set with boxes of a given size (r) and then calculating the probability p_i(r) of having a point of the set in the i-th box. The correlation dimension is defined as:

$$D_2=\underset{r\to 0}{\lim }{\displaystyle \sum }_{i=1}^N\frac{log\left({\sum }_i{p}_i{\left(r\right)}^2\right)}{logr}$$

The ${\sum }_i{p}_i{\left(r\right)}^2$ is the probability of finding a pair of points in a box of size r. The authors established that for small values of the separation distance (r) and sufficiently large N, the correlation sum grows like a power:

$$C\left(r\right)~r^{D_2}\Rightarrow D_2~\frac{log\left(C\left(r\right)\right)}{log\left(r\right)}$$

Here, time series are embedded at increasing embedding dimensions (D) until the measured D₂ reaches a plateau (plot of D₂ versus D). One of the most significant limitations of these applications is that the time series must be large enough. Eckmann and Ruelle [40] suggested that the upper limit of a reliable D₂ measurement is $D_2=2lo{g}_{10}N$. Thus, any plateau in the D₂ vs. D plot at D₂ values higher than this limit should be discounted as artifacts of finite data size.

After mentioning the general principles required for the RQA method and describing the individual parameters and features assessed by them, it is time to specify the aim of this paper. The purpose of this review was to show how the values of the above-discussed RQA measures look in various groups of subjects during the assessment of postural stability in typical free-standing tasks and how the authors interpret them.

2. Materials and Methods

2.1. Search Strategy

One author (RB) searched the three electronic databases PubMed, EBSCO, and Science-Direct to find relevant papers. The databases were searched between the 4th and 6th of February 2023. Only full-text articles in English written since 2000 were included. Letters to the editor, scientific reports, and post-conference papers were excluded from the review. The review also excluded papers that used variations of the RQA method (RQAEn) instead of the RQA method. Detailed search terms are in

Table 1. The search terms combinations applied within each of the three electronic databases.

PubMed	ScienceDirect	EBSCO
(Upright standing) AND (RQA) AND (postural control)—1 paper	(upright standing) AND (RQA)—24 papers
(RQA) AND (CoP)—16 papers	(RQA) AND (CoP)—43 papers
(Recurrence quantification analysis) AND (CoP)—27 papers	(Recurrence quantification analysis) AND (CoP)—98 papers	(Recurrence quantification analysis) AND (CoP)—18 papers
(Recurrence quantification analysis) AND (center of pressure)—45 papers	(Recurrence quantification analysis) AND (center of pressure)—390 papers	(Recurrence quantification analysis) AND (center of pressure)—34 papers
(Recurrence quantification analysis) AND (postural control)—53 papers		(Recurrence quantification analysis) AND (stability control) AND (center of pressure)—1 paper
(Recurrence quantification analysis) AND (postural stability)—24 papers		(Recurrence quantification analysis) AND (postural control) AND (center of pressure)—16 papers

. The differences in the search terms used for the listed databases are due to the need to limit the search accordingly to each database’s search strategy.

2.2. Review Process

The review process conducted by the authors (RB and MB) consisted of several steps. First, the duplicates were removed. Then, it was checked whether the title met the inclusion criteria. If so, the abstract was read to confirm inclusion. In cases where the abstract did not provide sufficient information, it was sought in the full text. Next, another author (MB) checked the relevance of the publications to the review topic and sorted them into three thematic groups based on the field of research. These groups were (1) children/young/elderly; (2) disability/disease; and (3) sport/influence of motion/impact of visual stimuli.

Excluded were papers in which RQA was conducted based on data other than CoP displacement and those in which CoP displacement was assessed in individuals in positions other than quiet standing. One reviewer (MB) compiled all articles using reference manager software (EndNote X7.7, Clarivate Analytics, Philadelphia, PA, USA).

The included publications were then searched by one of the authors (RB) and checked by (MB and AH) for six factors, which included: (1) Study group characteristics; (2) the aim of the study; (3) used equipment and conducted trials; (4) use of additional linear and nonlinear measures; (5) RQA parameters, measures, and toolbox; and (6) results of each study. The characteristics of the study groups involved the number of participants, their gender, the mean and standard deviation values of age, body weight, and height. The equipment’s description included: the type of equipment, name, and sampling frequency. The table also included the time delay (tau), embedding dimension (D), and threshold (R) needed to calculate RQA measures. The aforementioned items from each paper are listed in Table 2. Subsequently, within the three subject groups described above, the values of the RQA parameters were given in Table 3, Table 4 and Table 5. Finally, the maximum and minimum values of the RQA parameters were extracted and summarized in Table 6.

2.3. Quality Assessment

The included publications were assessed for methodological quality using a checklist for randomized and non-randomized studies [41]. The checklist consisted of five sub-scales: (1) Reporting (10 items)—where the information contained in the paper was checked to see if the study’s results were sufficient for unbiased evaluation; (2) external validity (3 items)—to evaluate the degree to which the retained results can be generalized to the entire population; (3) bias (7 items)—concerned bias in the measurement of intervention and outcome; (4) confounding (6 items)—to determine bias in the selection of study participants; and (5) power (1 item)—where an attempt was made to assess whether negative study results could be due to chance. A total of 32 points were available for each article. The scores for each paper are included in Table 2.

3. Results

The initial process of screening the electronic database yielded 790 papers. Screening of titles and abstracts led to the rejection of 714 articles. The approval of 28 studies was obtained after eliminating duplicates and publications with irrelevant measurement methods. The steps of the reviewing process are shown in Figure 3.

Due to the large number of included publications, there was a need to divide them according to the field of research into three groups. Eight publications were allocated to a group called children/young/elderly. In these papers, the authors investigated RQA measurements according to the age of the participants. Another eight studies were classified in the group disability/disease, where the authors used RQA to examine the impact of pathological conditions on the subjects. The last twelve papers were in the sports/impact of motion/impact of visual stimuli group. These articles covered the topics of movement and sport influence on RQA measures.

The summary results of the publication quality assessment are presented in Table 2. Of all publications, the paper with the lowest score was Hasson et al. [42] with 11 points. The reasons for this score were primarily the low number of participants, the subjects were not representative of the entire population from which they were recruited, and the actual probability values were not used. The highest score was 19 points. A total of four articles [20,37,43,44] reached this value. All of these publications scored a maximum of 5 for the size of the study group, and the participants were representative of the entire population from which they were recruited. Moreover, each publication, with the exception of Ferrufino et al. [43], reported full p-values. However, only Ferrufino et al. [43] blinded the subjects among the four publications listed.

In the children/young/elderly group, the lowest score was 11 [42], and the highest was 19 [37,44]. In the disability/disease group, the lowest score was 15 [45], and the highest was 18 [46,47]. In the sports/influence of motion/influence of visual stimuli group, the lowest score was 15 [48,49], and the highest was 19 [20,43]. The low evaluation values of all publications were influenced mainly by questions from the internal validity-confounding section.

Table 2. Data extracted from reviewed articles that used the RQA method to assess postural stability.

Study/ Quality	Study Group Age [Years] Body Mass [kg] Body Height [cm] BMI [kg/m2]	Aim	Equipment/ Trials	Additional Linear and Nonlinear Measures Time Delay (tau) Embedded Dimension (D) Threshold (R) RQA Measures and Toolbox	Results
Children/Young/Elderly
Hasson et al. [42] 13/32	4 young, healthy M: Age: 30 ± 4.8; Weight: 82.9 ± 9.0; Height: 177 ± 5.6. 1 healthy, older M: Age: 52; Weight: 86.5; Height: 194.	To evaluate and recommend methods for selecting embedding parameters that give clear and consistent results for quiet stance CoP data.	AMTI force platform (100 Hz). 1 trial (30 s) standing on hard surface with EO. Additionally: older subject—1 trial (30 s) standing on foam surface with EC.	ACF, AMIF, AVD: tau = 1 to 30 samples; FNN: D = 2 to 20; R = 20% of the mean Euclidean distance between all embedded vectors. %REC, %DET, ENT (entropy quantified by the distribution of the lengths of diagonal line segments parallel to the main diagonal), TREND, LMAX. Calculations performed in MATLAB according to Webber and Zbilut [10].	For all subjects, the RQA variables were sensitive to the embedding parameter values and the level of noise in the CoP data. Both FNN and average displacement algorithms gave clear and consistent results for all subjects with either raw or noisy data.
Seigle et al. [50] 18/32	11 young adults: Age: 24 ± 3; Weight: nd; Height: nd. 12 elderly adults: Age: 73 ± 8; Weight: nd; Height: nd	To quantify the effect of aging and visual feedback on postural stability.	Medicapteurs force platform (40 Hz). 2 trials (51,2 s) standing with EO and EC. 30 s rest between EO and EC.	Linear measurements: CoP path length, area of the ellipse. ACF: tau = 6; FNN: D = 8; R = 20% of the mean distance of all data points; RQA performed by Cross Recurrence Plot Toolbox 5.13 (R25) [12]. %DET (calculated for diagonal length Lmin = 4), ENT (Shannon entropy).	For the elderly group, removing visual feedback had a significant effect on ENT, in both directions, and on %DET in the ML direction (%DET was lower in EC than EO). In the young adults group, there was no difference between standing with EC and EO for the RQA results. Elderly people showed a decrease in the complexity (ENT) of the deterministic structure of CoP dynamics and a decrease in the predictability (%DET) of these postural oscillations.
King et al. [51] 15/32	12 young adults (6 M, 6 F)	To investigate the deterministic and stochastic properties of the recurrent dynamics of the CoP trajectories of each leg and whole-body (CoPnet) as a function of different foot positions and the availability of visual information.	2 synchronized AMTI force plate (100 Hz). 5 trials (60 s) standing: side-by-side stance; staggered-R and -L; tandem-R and -L. Each of the trials involved EO and EC. 15 s of rest between trials.	For the linear measurement, the kinetic data were down-sampled to 50 Hz and low-pass filtered by a 4th order double-pass Butterworth filter with 10 Hz cutoff frequency. Linear measures: CoP left, right, and net path length. Nonlinear measures: sample entropy with distance m = 3 and threshold r = 0.30—calculation with PhysioToolkit-PhysioNet software [52]. AMIF: tau = 10 data points; FNN: D = 6; R = 22% of the mean distance of all data points; %REC, %DET, ENT.	The availability of visual information increased CoP path length but had no effect on its recurrent dynamics. RQA showed that REC, DET, and ENT were dependent on the direction (AP/ML) of CoP motion and foot position. The CoP of the less loaded leg had a lower presence of stochastic processes (higher %REC) than the more loaded leg.
Tallon et al. [53] 15/32	101 older women: Age: 68.4 ± 4.2; Weight: 67.8 ± 12.2; Height: 162.1 ± 6.5.	To investigate the complementarities of several measures extracted from CoP during quiet standing.	Medicapteurs force plate (40 Hz) 1 trial (51.2 s) standing: EO	Linear measures: CoP path length, area of the ellipse. Nonlinear measures: sample entropy. ACF: tau = 6; FNN: D = 8; R = 25% of the mean of all distances, and LMIN = 4 [50]. %DET, ENT (Shannon entropy). The Cross Recurrence Plot Toolbox 5.13 (R25) [12].	The two families of measures (linear and dynamical) are not redundant but complementary.
Ramdani et al. [37] 19/32	14 non-fallers (3 M, 11 F): Age: 85.7 ± 6.7; Weight: 73.2 ± 12.6; Height: 160.0 ± 6.4. 14 fallers (5 M, 9 F): Age: 81.1 ± 9.1; Weight: 67.2 ± 13.6; Height: 161.2 ± 8.9	To investigate whether RQA outputs could reveal differences between dynamical features of CoP recordings of non-fallers vs. fallers older adults.	Medicapteurs force plate (40 Hz) 1 trial (51.2 s) standing: EO	Linear measures: CoP path length, range of CoP displacement in AP and ML direction. AMIF: tau = 6; FNN: D = 8; R = 30% (to assess the robustness of results regarding the radius selection, computations were also performed for 25% and 35%) of the mean of all distances. %DET (LMIN = 4), ENT. The Cross Recurrence Plot Toolbox [12].	CoP path length in ML direction, DET_ML and ENT_ML were significantly higher for the fallers group. RQA showed significantly increased complexity and predictability for fallers in the ML direction.
van den Hoorn et al. [17] 18/32	267 older adults (143 M, 124 F): Age: 75 ± 6; Weight: 74.9 ± 14.7; Height: 168 ± 9.	To determine how the method of threshold determination, below which a recurrence is defined, affects the within-session reliability of RQA in an elderly population.	AMTI force plate (1000 Hz) and Vicon Mx. 4 trials (30 s) with EO and EC; firm and rubber surface 30 s of rest between trials	CoP data were filtered with a bi-directional, 2nd low-pass Butterworth filter. The cut-off frequency was set at 20 Hz; bi-directional filtering increased the filter order to 4. CoP data were decimated to 100 samples/s. AVD: tau = separately for each 30 s CoP signal [38]; FNN: D = 5; The recurrence threshold was determined in two ways: R1 = 27% of the mean distance between all points in phase space; R2 = set so that the recurrence rate was fixed at 5%. %DET, %LAM, TT, LMAX, LMEAN Recurrence Plot Toolbox 5.21 (31) [6].	Fixing the recurrence rate does improve the reliability of the RQA measures in both AP and ML directions more than using a threshold based on the amplitude of the signal.
Hourova et al. [54] 17/32	103 healthy women: N = 54 aged 60–69; N = 49 aged 70–79. Weight: both groups divided into a subgroup of normal and overweight subjects based on BMI. Height: nd.	To discuss the applicability of the RQA method to assessing changes in postural stability in people over the age of 60.	AMTI force plate (nd Hz) 3 trials (30 s): standing on firm surface with EC; standing on foam surface with EO and EC.	tau = nd., D = nd., R = nd., %DET, %LAM, TT.	%LAM is the most sensitive factor to determine differences in CoP position between different age groups as it is fluidity of motion. For ML direction of CoP motion while standing on foam surface with EO, a statistically significant difference was found for all indicators (%LAM, %DET and TT) in subjects with a normal BMI.
Hao et al. [44] 19/32	3 groups of preschool children: I group: 3 y.o. (N = 36; 19 M, 17 F): Weight 17.35 ± 3.16 Height: 103.44 ± 4.9 II group: 4 y.o. (N = 38; 17 M, 21 F): Weight 19.19 ± 3.3 Height: 111.5 ± 4.03 III group: 5 y.o. (N = 39; 21 M, 18 F): Weight 23.38 ± 5.47 Height: 119.71 ± 6.6	To investigate how age and sensory perturbation affect the control of standing balance on a firm surface.	AMTI force plate (1000 Hz) 3 trials (15 s) standing with EO, EC, and ECHB (eyes closed and head extended backward). 30 s of rest between trials.	The CoP were filtered using a 20 Hz low-pass, 2nd order, zero-lag Butterworth filter. Linear measures: CoP path length, area of the ellipse, range, mean velocity. Nonlinear measures: DFA II group under the condition of ECHB: AMIF: tau = 28 samples; FNN: D = 3; Recurrence rate: R = 5%. Minimal length of both diagonal and vertical line features was set as 0.1 s. %DET, %LAM Cross-Recurrence Plot Toolbox [6].	Higher %DET and %LAM in the AP direction were found for 5 y.o. children compared to 3 and 4 y.o. children. As sensory conditions became more challenging, all traditional measures increased, while %DET and %LAM significantly declined in the AP direction.
Disability/Disease
Schmit et al. [55] 17/32	6 Parkinson’s patients (2 M, 4 F): Age: 70.83 ± 15.89; Weight: nd; Height: nd. 6 elderly, healthy adults (2 M, 4 F): Age: 70.17 ± 4.71; Weight: nd; Height: nd.	To investigate whether greater determinism in the time structure of CoP time series would occur in participants with Parkinson’s disease.	Bertec 4060-NC force platform (100 Hz). Combination of visual control (EO, EC) and cognitive demand (no-task, visual-spatial cognitive task). Each condition was repeated 4 times. Each trial lasted 30 s.	Linear measure: CoP path length. tau = 9 samples; D = 8; R = 32% of the mean Euclidean distance separating points in the reconstructed phase space; and number of successive points defining a line segment = 2. %REC, %DET, ENT, LMAX, TREND.	The CoP path length was greater for PD patients than for controls. %REC, %DET, LMAX, ENT_AP were greater for PD patients.
Negahban et al. [46] 18/32	ACLD group (27 M with anterior cruciate ligament deficient): Age: 26.74 ± 5.84; Weight: nd; Height: 177 ± 64; BMI: 23.69 ± 2.42. H group (27 M): Age: 26.29 ± 5.07; Weight: nd; Height: 179 ± 55; BMI: 22.84 ± 2.38.	To investigate the non-linear dynamical structure of the CoP time series in a population with ACLD knees. To investigate the effects of cognitive loading on non-linear dynamical features of postural sway in ACLD patients compared to healthy controls.	Bertec 9090-15 force platform (200 Hz). 6 trials both leg standing (30 s): EO; cognitive task with EO; EC, cognitive task with EC; foam surface with EC; foam surface with cognitive task and EC. 4 trials single leg standing (20 s): on injured leg; on injured leg with cognitive task; on non-injured leg; on non-injured leg with cognitive task.	AMIF: tau = nd; FNN: D = 5; R = 10% of the maximum diameter of the reconstructed attractor. A minimum number of three points was employed to detect diagonal lines. With these input parameters, distance matrices and recurrence plots were computed, from which the four RQA measures were extracted. %REC, %DET, ENT (Shannon), TREND.	For both leg standing: %DET and ENT in the AP and ML directions were higher for ACLD patients compared to the H participants. When postural difficulty was increased—%DET and ENT significantly increased. When cognitive difficulty was increased (single to dual-task conditions), %DET and ENT significantly decreased. For the single leg standing: %DET in the AP and ML directions and ENT in the ML direction were significantly higher for both the injured and non-injured leg in the ACLD group when compared to the H participants. When cognitive difficulty was increased (single to dual-task conditions), %DET and ENT significantly decreased.
Mazaheri et al. [56] 17/32	22 patients with low back pain (LBP) (13 M, 9 F): Age: 26.1 ± 6.2; Weight: 67.1 ± 11.2; Height: 172 ± 10. 22 healthy adults (13 M, 9 F): Age: 25.0 ± 5.5; Weight: 66.5 ± 12.1; Height: 173 ± 10.	To determine whether people with LBP show atypical dynamic patterns of postural sway when performing attention-demanding tasks compared with their matched control participants.	Bertec 4060-10 force platform (200 Hz)—down-sampled without filtering to 100 Hz. 9 trials (30 s) with combination of: 3 postural tasks: both leg standing on rigid force platform with EO; with EC; on the foam with EC. 3 levels of cognitive difficulty (no-task, easy, and difficult).	tau—was calculated for each case using the AVD; D = 5; R = 10% of the maximum diameter of the reconstructed attractor. %REC, %DET, ENT, TREND	As postural conditions became more difficult, the postural sway in both AP and ML directions became more regular (higher %REC and %DET), more complex (higher ENT), and more stationary (lower TREND). The only exception was for %REC, which decreased in the foam and EC condition. Increasing cognitive difficulty was associated with less regularity (lower %REC and %DET), less complexity (lower ENT), and more stationary (lower TREND) in both directions.
Negahban et al. [47] 17/32	23 patients with multiple sclerosis (MS) (8 M, 15 F): Age: 32.7 ± 7.9; Weight: nd; Height: 160 ± 14; BMI = 24.1 ± 3.8. 23 healthy adults (8 M, 15 F): Age: 31.4 ± 7.9; Weight: nd; Height: 160 ± 96; BMI = 24.6 ± 3.8.	The characterization of the nonlinear dynamical structure of postural sway obtained from the RQA in MS patients as compared to healthy controls. Investigating the effect of a cognitive task on postural control.	Bertec 4060-10 force platform (100 Hz). 3 trials (30 s) of combinations of three levels of postural difficulty (both leg standing with EO, EC, and on foam surface with EC) with two levels of cognitive difficulty (single and dual-tasks). 5 min rest.	AMIF: tau = nd; FNN: D = 5; R = 10% of the maximum phase space diameter using Euclidean norm. This value kept the recurrence rate below 5% in most cases. %REC, %DET, ENT, TREND.	For both groups: as the postural conditions became more difficult, the CoP time series were characterized by lower %REC, ENT, and TREND. As cognitive conditions became more difficult, CoP time series were characterized by lower %REC, ENT and TREND in AP direction and lower %DET in both directions.
Cao et al. [57] 16/32	89 patients with multiple sclerosis (MS) (20 M, 69 F): Age: 46.11 ± 10.26 Weight: nd; Height: nd. 29 healthy adults (8 M, 21 F): Age: 34.79 ± 12.31; Weight: nd; Height: nd.	To estimate Expanded Disability Status Scale (EDSS) scores by using values extracted from the posturographic data of MS patients, in order to select the most pertinent parameters.	Satel force plate (40 Hz). 1 trial (51.2 s) both leg standing with EC.	Linear measures: CoP path length, sway velocity. tau = 1/40 s; D = 1; R = 15 mm for maximum correlation between EDSS and each recurrence estimate. %REC, ENT (Shannon), LMAX, TT. RQA parameters were calculated for position, instantaneous velocity and acceleration of the CoP.	This study emphasizes the possibility of distinguishing EDSS scores using postural sway and RQA parameters. %REC and LMAX were in best agreement with clinical assessments.
Cao et al. [58] 17/32	117 patients with multiple sclerosis (MS) (26 M, 91 F): Age: 46.53 ± 11.31; Weight: nd; Height: nd. 16 healthy adults (4 M, 12 F): Age: 44.59 ± 6.12; Weight: nd; Height: nd.	To evaluate strategy in both mono- and multi-dimensional analyses in order to determine the most discriminative RQA measure for estimating EDSS.	Satel force plate (40 Hz). 1 trial (51.2 s) both leg standing with EC.	Linear measures: CoP path length, sway velocity. Mono-dimensional analysis: AMIF: tau = 1 FNN: D = 1 Multi-dimensional analysis: AMIF: tau = 1 to 30 samples; FNN: D = between 2 and 20; R1 = 20% of the maximal distance of one participant; R2 = equal to the mean of the series calculated in R1. REC, ENT, LMAX, TT, TREND.	The unidimensional RQA of the CoP signal is a more accurate method for quantifying disability in MS patients than multidimensional RQA.
Negahban et al. [59] 18/32	27 patients with knee osteoarthritis (OA) (8 M, 19 F): Age: 54.8 ± 6.4; Weight: nd; Height: 162.4 ± 8.4; BMI = 27.9 ± 2.9. 27 healthy adults (8 M, 19 F): Age: 54.4 ± 6.2; Weight: nd; Height: 162.8 ± 8.7; BMI = 28 ± 3.2.	To investigate differences in the complexity and variability of the CoP dynamics between OA patients and healthy controls.	Bertec 4060-10 force platform (100 Hz). 3 trials for 8 conditions (60 s): both leg standing on rigid and foam surface. Each trail with EO; with EO and cognitive task; with EC; with EC and cognitive task. 5 min break	Second-order difference (SOD) plots quantified with central tendency measure (CTM). AMIF: tau = 3; D = 5; R = 10% of the maximum phase space diameter using Euclidean norm. No filtering data. %DET	Knee OA patients had a greater %DET in ML direction than healthy subjects. In both groups, the %DET increased with increasing postural difficulty, while %DET decreased when moving from single to dual-task.
Shokouhyan et al. [45] 15/32	20 males with non-specific low back pain (NSLBP): Age: 24.5 ± 0.9; Weight: 62 ± 7.5; Height: 172 ± 7.5. 20 healthy adults (M): Age: 25.5 ± 0.7; Weight: 64 ± 8.6; Height: 174 ± 6.5	Evaluation and comparison of proprioceptive parameters (body sway and stability) between patients with NSLBP and healthy controls.	Bertec force plate (100 Hz) and Vicon motion capture system (100 Hz). In-house vibrator apparatus VR (70 Hz) to alter proprioception of the soleus and lumbar muscles. 8 trails (30 s) with EC standing on a rigid and foam surface: without VR; with the VR of the triceps; with VR of the multifidus; with the VR of both muscles.	Linear measures: CoP path length, velocity Nonlinear measures: Lyapunov exponents. AMIF: tau = 0.35–0.6 s; FNN: D = 3 or 4; R = 2.5% of the mean distance. RQA software developed by Webber [60] %REC, %DET, ENT, TREND.	An increase of the standard deviation of amplitude and velocity among the NSLBP participants. %DET and ENT were greater in the NSLBP group. RQA parameters for the CoP on both sides and for the trunk sagittal angle indicated more repeated patterns of movement among the NSLBP group.
Sport/Influence of motion/Impact of visual stimuli
Schmit et al. [18] 16/32	10 undergraduate dance majors (5 M, 5 F): Age: 20; Weight: nd; Height: nd. 10 runners (5 M, 5 F): Age: 19.5; Weight: nd; Height: nd.	To compare postural stability and postural sway dynamics in dancers and a control group. To determine if dancers merely exhibit a different amount of postural sway variability or if, in addition (or instead), they exhibit qualitatively different dynamic patterns of postural sway.	Bertec 4060-NC force platform (100 Hz). 4 trials (30 s) both leg standing with EO/EC on rigid/foam surface. Breaks as needed.	Linear measures: CoP path length. For AP sway: tau = 9 samples; D = 11; R = 24% of the mean distance separating points in reconstructed phase space. For ML sway: tau = 10 samples; D = 10; R = 20% of the mean distance separating points in reconstructed phase space. %REC, %DET, ENT, LMAX, TREND.	RQA revealed that the postural sway of dancers was less regular (lower %REC), less stable (lower LMAX), less complex (lower ENT), and more stationary (lower TREND) than that of track athletes.
Clark and Riley [49] 15/32	12 college students (4 M, 8 F): Age: 20 (18–24); Weight: nd; Height: 169.8 (154.9–190.5)	To investigate the multisensory control of posture by altering sensory information across the visual and somatosensory systems.	Smart Balance Master® force plate (100 Hz). 6 trials (20 s) for 3 gain settings for 4 conditions (total 72 trials): (1) EO with sway-referenced visual surround (SV); (2) EO with support surface (SS); (3) EC/SS; (4) EO/SV-SS. 2 min rest.	Linear measures: CoP path length, the coefficient of variation of CoP displacements. AMIF: tau = 0.07 s; FNN: D = 9; R = 11% of the mean Euclidean distance separating data points. %DET, TREND.	The CoP was less random (higher %DET) the higher the gain condition was. The CoP became increasingly more deterministic across more challenging sensory organization test conditions and with increasing gain, and more nonstationary across more challenging conditions and when the support surface was sway-referenced using a 1.8 gain setting.
Ferrufino et al. [43] 19/32	16 contemporary dancers (CD) (16 F): Age: 73.7 ± 5.5; Weight: nd; Height: nd; BMI: 26.9 ± 4.8. 25 fall prevention (FP) (2 M, 23 F): Age: 72.9 ± 7.2; Weight: nd; Height: nd; BMI: 24.6 ± 3.3.	To study the effects of two physical activities (contemporary dance and fall prevention training) on postural control in older adults.	Techno-Concept platform (40 Hz). 4 conditions (51.2 s): EC, EO with vision fixed on a point at a distance of 600, 150, or 40 cm. Recordings before and after intervention.	Linear measures: CoP path length, area of ellipse, the mean velocity of the CoP displacements. Convex hull area including 100% of the positions of the CoP, the Romberg quotient, and the fractal dimension ratio. tau = 50 ms (corresponding to 2 samples); D = 12; R = 2–3% of the mean distance between data points. %REC, %DET, LMAX.	CD reduced both %REC and mathematical stability (%LAM) suggesting that CoP displacements were less likely to repeat themselves over time and that their dynamics were more flexible, whereas FP yielded the opposite tendency.
Markovic et al. [61] 16/32	40 young adults (22 M, 18 F): Age: 27 ± 2.6; Weight: 71 ± 13; Height: 176 ± 7. 34 older persons (15 M, 19 F): Age: 79 ± 2.8; Weight: 71 ± 14; Height: 165 ± 9	To evaluate the intra-session reliability of traditional and nonlinear measures of postural sway during upright semi-tandem quiet stance in young vs. older subjects.	AMTI force plate (1000 Hz). 3 trials (30 s) of semi-tandem stance with EO. 3 min break	Linear measures: 13 parameters, nonlinear measures: 20 parameters tau = nd; D = nd; R = nd. %REC, %DET, ENT, %LAM, TT, LMAX. Removal of the potential noise from the CoP signal (2nd order Butterworth, 0.1–20 Hz band-pass, bidirectional filter).	The RQA (%DET, ENT) and sample entropy analyses, but not diffusion plot and sway density plot analyses, can be reliably used in young and elderly age groups.
Apthorp et al. [48] 15/32	13 young adults (3 M, 10 F): Age: 20.9 ± 0.8; Weight: 75 ± 17; Height: 171.7 ± 9.1	To investigate if strength of visually induced illusions of self-motion (vection) depends on the extent to which people rely on vision to maintain their postural stability.	Bertec force plate (1000 Hz). 3 trials (60 s) both leg standing with EO and EC. 2 sessions (30 s) blank period, followed by 30 s of the optic flow stimulus, 30 s of a blank screen, and 30 s of a simple fixation screen. Then seated trials were performed.	Linear measures: CoP path length, area of ellipse, standard deviation, and sway magnitude. AMIF: tau = 15; FNN: D = 8; R = 30. %REC, %DET, %LAM, TT, ENT, LMAX	The ratio between eyes-open and eyes-closed CoP excursions during quiet stance (using the area of postural sway) significantly predicted seated vection for both measures. RQA was a much stronger predictor of seated vection for both expanding and contracting stimuli.
Coubard et al. [62] 17/32	19 contemporary dancers (CD) (19 F): Age: 70.6 ± 7.3; Weight: nd; Height: nd; BMI = 26.5 ± 3.8. 19 no dancers (ND) (1 M, 18 F): Age: 72.6 ± 8.6; Weight: nd; Height: nd; BMI = 27.0 ± 3.9.	To investigate the effects of a dance practice (CD), on the quality of posture in older adults.	Techno-Concept platform (40 Hz). 1 trial (51.2 s) both leg standing with EO, EC. Recordings before and after training intervention.	Linear measures: CoP path length, ellipse area, the convex hull area, which include 90% and 100% of CoP positions, respectively. Romberg quotient and Signal Diffusion Analysis (SDA). Nonlinear measures: Detrended Fluctuation Analysis (DFA). tau = 50 ms; D = 12; R = 2–3% of the mean distance between data points. %REC, %DET, LMAX	The ND group showed an increase in length and mean velocity of postural signal. While both legs standing with EO, LMAX decreases in AP direction. In ML direction, %DET and LMAX showed higher values in post-test period. In AP direction, LMAX showed higher values in post-test period.
Decker et al. [63] 16/32	126 healthy women: Age: 65.7 ± 4.1; Weight: 66.4 ± 12.2; Height: 161.9 ± 6.3.	To determine whether postural sway complexity could discriminate asymptomatic sedentary postmenopausal women with normal or subnormal physical function from those with lower physical function.	Medicapteurs force platform (40 Hz). 1 trial (51.2 s) both leg standing with EO and EC. 30 s break between trials.	Six-Minute Walking Distance test. Linear measures: CoP path length, ellipse area, standard deviations in both directions. Nonlinear measures: sample entropy (SampEn), complexity index of multiscale entropy (CIMSE). tau = 6; D = 8; R = 0.25 of the mean of all distances. %REC, %DET, ENT Cross Recurrence Plot Toolbox 5.13 (R25) [6]	Traditional measures did not differentiate women with lower, subnormal, or normal physical function. Women with lower physical function showed lower SampEn values and higher %DET values in ML direction during standing with EC. No significant difference in the CIMSE values was found between the two groups.
Bernard et al. [64] 18/32	49 Sedentary community-dwelling women: Age: 64.98 ± 3.87; Weight: 65.54 ± 11.37; Height: 160.6 ± 5.35	To investigate whether aerobic exercise at submaximal intensity has detrimental effects on balance in older sedentary adults.	Mdicapteurs force platform (40 Hz). 1 trial (51.2 s) both leg standing with EO and EC. 30 s break between trials. Than 2 trials of 6 min walk tests (6 MWT). After each test: 1 trial (51.2 s) both leg standing with EC. 20 min break between 6 MWT trials.	Linear measures: CoP path length, ellipse area, the mean along the AP and ML axes, central tendency measure (CTM). tau = 6; D = 8; R = 25% of mean of all the distances. %DET, ENT. Cross Recurrence Plot Toolbox 5.13 (R25) [6].	In older, sedentary women, two consecutive cycles of 6 min aerobic exercise separated by a 20 min rest period did not significantly change kinematic measures, only measures of complexity. This study showed the effects of walking exercise on balance parameters: CTM_AP, %DET_ML and ENT_ML. The significant increase in the %DET_ML represents an increase in the predictability of the time series, a marker of reduced postural complexity and less adaptive postural capabilities in the ML direction, according to the loss of complexity of aging and disease theory. A possible interpretation of this finding is a less adaptive behavior related to a hip muscle strategy adversely affected by the exercise of walking.
Palmisano et al. [65] 18/32	23 adults (7 M, 16 F): Age: 22.9 ± 6.0; Weight: nd; Height: nd.	To compare the predictive ability of traditional and RQA measures of postural activity.	Bertec force plate (1000 Hz). 1 trial (60 s) both leg standing with: EO and EC. After these trials, subjects were seated and exposed to 5 smooth and 5 oscillating radial flow displays (random order: 30 s). 60 s break between vection trials. Display apparatus (60 Hz).	Linear measures: CoP path length, ellipse area. Tau = 15; D = 8; R = 0.6 Minimum line length of 2. %REC, %DET, ENT, LMAX Recurrence Quantification Toolbox for MATLAB [66].	%REC in AP direction significantly predicted the strength of oscillating vection. %REC during trial with EO significantly predicted the strength of the vection induced by oscillating radial flow. %REC, %DET, and ENT along the AP axis correlated significantly with the average reported strength of the smooth vection. %REC, ENT, and LMAX along the AP axis also correlated significantly with the average reported strength of the oscillating vection. Lower %REC, ENT, and shorter LMAX were associated with stronger vection (both smooth and oscillating). Lower %DET was also associated with stronger smooth vection.
Bernard et al. [67] 18/32	Trained (61 F): Age: 65.46 ± 4.37; Weight: 64.14 ± 12.06; Height: 161.25 ± 6.29. Controls (60 F): Age: 65.54 ± 4.04; Weight: 70.06 ± 11.57; Height: 162.53 ± 6.6.	To evaluate the influence of an active 6-month walking program of three 60 min sessions per week at 60–80% of maximum heart rate on sedentary elderly women’s kinematic and dynamic parameters of the CoP.	Medicapteurs force platform (40 Hz). 1 trial (51.2 s) both leg standing with EO and EC. 30 s break between trials.	Linear measures: CoP path length, ellipse area, the mean along the AP and ML axes. Nonlinear measures: sample entropy (SampEn) and multiscale entropy (MSE). tau = nd; D = nd; R = nd. % DET Cross Recurrence Plot Toolbox 5.13 (R25) [6].	In older sedentary women, this study indicated a systemic lack of influence of 6 months’ walk-training on flat ground on kinematic postural responses and on dynamical measures.
van den Hoorn et al. [20] 19/32	106 elderly (64 M, 42 F): Age: 75 ± 6; Weight: 78 ± 15; Height: 169 ± 9. 23 young healthy adults (9 M, 14 F): Age: 21 ± 2; Weight: 65 ± 10; Height: 172 ± 5.	To compare CoP motion between young and older individuals before, during, and after removal of bilateral calf vibration, and to compare measures between older individuals who subsequently do or do not go on to fall in the following 12 months.	Kistler force plate (2000 Hz). 1 trial (135 s) both leg standing with EC. Vibrators (60 Hz) attached bilaterally over the triceps surae muscles were activated twice for 15 s; after 15 and 75 s (45 s for recovery).	Linear measures: CoP path length, CoP position relative to the mean CoP position at the baseline. tau = 180 ms; D = 5; R = 5%. %DET, LMEAN, %LAM, TT. RQA toolbox [6].	At baseline (before vibration), compared to fallers, young people had lower %DET (more predictable), had longer LMEAN (less sensitive to small perturbations), and had higher %LAM with longer TT (were more intermittent, with longer static episodes). Compared to non-fallers, young people also had longer LMEAN, TT, and were less anti-persistent. %DET and %LAM were not significantly different between young and non-fallers.
Chowdhury et al. [68] 16/32	15 young adults (11 M, 4 F): Age: 24.6 ± 5.8; Weight: nd; Height: nd.	To examine the effects of simulating self-motion via a head-mounted display (HMD) on standing postural sway and spatial presence.	Bertec force plate (1000 Hz). Oculus Rift CV1 HMD. 1 trial both leg standing: 30 s stationary screen followed by 60 s of screen motion.	Linear measures: CoP path length, CoP sway variability. The spectrum function and Fourier analysis. tau = 15; D = 8; R = 0.6 and line minimum of 2. %REC.	%REC declined as the display oscillation amplitude increased.

Abbreviations: nd—no data, M—males, F—females, N—number of people in the group, y.o.—years old, CoP—center of pressure, AP—anterior-posterior, ML—medial-lateral direction, EO—eyes open, EC—eyes closed, L—left, R—right, FNN—false nearest neighbors, AMIF—average mutual information function, ACF—autocorrelation technique, AVD—the average displacement method, DFA—detrended fluctuation analysis, SDA—signal diffusion analysis, CTM—central tendency measure, PD—Parkinson’s disease, ACLD—patients with anterior cruciate ligament deficiency, LBP—low back pain, MS—multiple sclerosis, EDSS—Expanded Disability Status Scale, OA—knee osteoarthritis, NSLBP—non-specific low back pain, SV—visual surround, SS—support surface, CD—contemporary dancers, ND—no dancers, HMD—head-mounted display.

Table 2. Data extracted from reviewed articles that used the RQA method to assess postural stability.

Study/ Quality	Study Group Age [Years] Body Mass [kg] Body Height [cm] BMI [kg/m2]	Aim	Equipment/ Trials	Additional Linear and Nonlinear Measures Time Delay (tau) Embedded Dimension (D) Threshold (R) RQA Measures and Toolbox	Results
Children/Young/Elderly
Hasson et al. [42] 13/32	4 young, healthy M: Age: 30 ± 4.8; Weight: 82.9 ± 9.0; Height: 177 ± 5.6. 1 healthy, older M: Age: 52; Weight: 86.5; Height: 194.	To evaluate and recommend methods for selecting embedding parameters that give clear and consistent results for quiet stance CoP data.	AMTI force platform (100 Hz). 1 trial (30 s) standing on hard surface with EO. Additionally: older subject—1 trial (30 s) standing on foam surface with EC.	ACF, AMIF, AVD: tau = 1 to 30 samples; FNN: D = 2 to 20; R = 20% of the mean Euclidean distance between all embedded vectors. %REC, %DET, ENT (entropy quantified by the distribution of the lengths of diagonal line segments parallel to the main diagonal), TREND, LMAX. Calculations performed in MATLAB according to Webber and Zbilut [10].	For all subjects, the RQA variables were sensitive to the embedding parameter values and the level of noise in the CoP data. Both FNN and average displacement algorithms gave clear and consistent results for all subjects with either raw or noisy data.
Seigle et al. [50] 18/32	11 young adults: Age: 24 ± 3; Weight: nd; Height: nd. 12 elderly adults: Age: 73 ± 8; Weight: nd; Height: nd	To quantify the effect of aging and visual feedback on postural stability.	Medicapteurs force platform (40 Hz). 2 trials (51,2 s) standing with EO and EC. 30 s rest between EO and EC.	Linear measurements: CoP path length, area of the ellipse. ACF: tau = 6; FNN: D = 8; R = 20% of the mean distance of all data points; RQA performed by Cross Recurrence Plot Toolbox 5.13 (R25) [12]. %DET (calculated for diagonal length Lmin = 4), ENT (Shannon entropy).	For the elderly group, removing visual feedback had a significant effect on ENT, in both directions, and on %DET in the ML direction (%DET was lower in EC than EO). In the young adults group, there was no difference between standing with EC and EO for the RQA results. Elderly people showed a decrease in the complexity (ENT) of the deterministic structure of CoP dynamics and a decrease in the predictability (%DET) of these postural oscillations.
King et al. [51] 15/32	12 young adults (6 M, 6 F)	To investigate the deterministic and stochastic properties of the recurrent dynamics of the CoP trajectories of each leg and whole-body (CoPnet) as a function of different foot positions and the availability of visual information.	2 synchronized AMTI force plate (100 Hz). 5 trials (60 s) standing: side-by-side stance; staggered-R and -L; tandem-R and -L. Each of the trials involved EO and EC. 15 s of rest between trials.	For the linear measurement, the kinetic data were down-sampled to 50 Hz and low-pass filtered by a 4th order double-pass Butterworth filter with 10 Hz cutoff frequency. Linear measures: CoP left, right, and net path length. Nonlinear measures: sample entropy with distance m = 3 and threshold r = 0.30—calculation with PhysioToolkit-PhysioNet software [52]. AMIF: tau = 10 data points; FNN: D = 6; R = 22% of the mean distance of all data points; %REC, %DET, ENT.	The availability of visual information increased CoP path length but had no effect on its recurrent dynamics. RQA showed that REC, DET, and ENT were dependent on the direction (AP/ML) of CoP motion and foot position. The CoP of the less loaded leg had a lower presence of stochastic processes (higher %REC) than the more loaded leg.
Tallon et al. [53] 15/32	101 older women: Age: 68.4 ± 4.2; Weight: 67.8 ± 12.2; Height: 162.1 ± 6.5.	To investigate the complementarities of several measures extracted from CoP during quiet standing.	Medicapteurs force plate (40 Hz) 1 trial (51.2 s) standing: EO	Linear measures: CoP path length, area of the ellipse. Nonlinear measures: sample entropy. ACF: tau = 6; FNN: D = 8; R = 25% of the mean of all distances, and LMIN = 4 [50]. %DET, ENT (Shannon entropy). The Cross Recurrence Plot Toolbox 5.13 (R25) [12].	The two families of measures (linear and dynamical) are not redundant but complementary.
Ramdani et al. [37] 19/32	14 non-fallers (3 M, 11 F): Age: 85.7 ± 6.7; Weight: 73.2 ± 12.6; Height: 160.0 ± 6.4. 14 fallers (5 M, 9 F): Age: 81.1 ± 9.1; Weight: 67.2 ± 13.6; Height: 161.2 ± 8.9	To investigate whether RQA outputs could reveal differences between dynamical features of CoP recordings of non-fallers vs. fallers older adults.	Medicapteurs force plate (40 Hz) 1 trial (51.2 s) standing: EO	Linear measures: CoP path length, range of CoP displacement in AP and ML direction. AMIF: tau = 6; FNN: D = 8; R = 30% (to assess the robustness of results regarding the radius selection, computations were also performed for 25% and 35%) of the mean of all distances. %DET (LMIN = 4), ENT. The Cross Recurrence Plot Toolbox [12].	CoP path length in ML direction, DET_ML and ENT_ML were significantly higher for the fallers group. RQA showed significantly increased complexity and predictability for fallers in the ML direction.
van den Hoorn et al. [17] 18/32	267 older adults (143 M, 124 F): Age: 75 ± 6; Weight: 74.9 ± 14.7; Height: 168 ± 9.	To determine how the method of threshold determination, below which a recurrence is defined, affects the within-session reliability of RQA in an elderly population.	AMTI force plate (1000 Hz) and Vicon Mx. 4 trials (30 s) with EO and EC; firm and rubber surface 30 s of rest between trials	CoP data were filtered with a bi-directional, 2nd low-pass Butterworth filter. The cut-off frequency was set at 20 Hz; bi-directional filtering increased the filter order to 4. CoP data were decimated to 100 samples/s. AVD: tau = separately for each 30 s CoP signal [38]; FNN: D = 5; The recurrence threshold was determined in two ways: R1 = 27% of the mean distance between all points in phase space; R2 = set so that the recurrence rate was fixed at 5%. %DET, %LAM, TT, LMAX, LMEAN Recurrence Plot Toolbox 5.21 (31) [6].	Fixing the recurrence rate does improve the reliability of the RQA measures in both AP and ML directions more than using a threshold based on the amplitude of the signal.
Hourova et al. [54] 17/32	103 healthy women: N = 54 aged 60–69; N = 49 aged 70–79. Weight: both groups divided into a subgroup of normal and overweight subjects based on BMI. Height: nd.	To discuss the applicability of the RQA method to assessing changes in postural stability in people over the age of 60.	AMTI force plate (nd Hz) 3 trials (30 s): standing on firm surface with EC; standing on foam surface with EO and EC.	tau = nd., D = nd., R = nd., %DET, %LAM, TT.	%LAM is the most sensitive factor to determine differences in CoP position between different age groups as it is fluidity of motion. For ML direction of CoP motion while standing on foam surface with EO, a statistically significant difference was found for all indicators (%LAM, %DET and TT) in subjects with a normal BMI.
Hao et al. [44] 19/32	3 groups of preschool children: I group: 3 y.o. (N = 36; 19 M, 17 F): Weight 17.35 ± 3.16 Height: 103.44 ± 4.9 II group: 4 y.o. (N = 38; 17 M, 21 F): Weight 19.19 ± 3.3 Height: 111.5 ± 4.03 III group: 5 y.o. (N = 39; 21 M, 18 F): Weight 23.38 ± 5.47 Height: 119.71 ± 6.6	To investigate how age and sensory perturbation affect the control of standing balance on a firm surface.	AMTI force plate (1000 Hz) 3 trials (15 s) standing with EO, EC, and ECHB (eyes closed and head extended backward). 30 s of rest between trials.	The CoP were filtered using a 20 Hz low-pass, 2nd order, zero-lag Butterworth filter. Linear measures: CoP path length, area of the ellipse, range, mean velocity. Nonlinear measures: DFA II group under the condition of ECHB: AMIF: tau = 28 samples; FNN: D = 3; Recurrence rate: R = 5%. Minimal length of both diagonal and vertical line features was set as 0.1 s. %DET, %LAM Cross-Recurrence Plot Toolbox [6].	Higher %DET and %LAM in the AP direction were found for 5 y.o. children compared to 3 and 4 y.o. children. As sensory conditions became more challenging, all traditional measures increased, while %DET and %LAM significantly declined in the AP direction.
Disability/Disease
Schmit et al. [55] 17/32	6 Parkinson’s patients (2 M, 4 F): Age: 70.83 ± 15.89; Weight: nd; Height: nd. 6 elderly, healthy adults (2 M, 4 F): Age: 70.17 ± 4.71; Weight: nd; Height: nd.	To investigate whether greater determinism in the time structure of CoP time series would occur in participants with Parkinson’s disease.	Bertec 4060-NC force platform (100 Hz). Combination of visual control (EO, EC) and cognitive demand (no-task, visual-spatial cognitive task). Each condition was repeated 4 times. Each trial lasted 30 s.	Linear measure: CoP path length. tau = 9 samples; D = 8; R = 32% of the mean Euclidean distance separating points in the reconstructed phase space; and number of successive points defining a line segment = 2. %REC, %DET, ENT, LMAX, TREND.	The CoP path length was greater for PD patients than for controls. %REC, %DET, LMAX, ENT_AP were greater for PD patients.
Negahban et al. [46] 18/32	ACLD group (27 M with anterior cruciate ligament deficient): Age: 26.74 ± 5.84; Weight: nd; Height: 177 ± 64; BMI: 23.69 ± 2.42. H group (27 M): Age: 26.29 ± 5.07; Weight: nd; Height: 179 ± 55; BMI: 22.84 ± 2.38.	To investigate the non-linear dynamical structure of the CoP time series in a population with ACLD knees. To investigate the effects of cognitive loading on non-linear dynamical features of postural sway in ACLD patients compared to healthy controls.	Bertec 9090-15 force platform (200 Hz). 6 trials both leg standing (30 s): EO; cognitive task with EO; EC, cognitive task with EC; foam surface with EC; foam surface with cognitive task and EC. 4 trials single leg standing (20 s): on injured leg; on injured leg with cognitive task; on non-injured leg; on non-injured leg with cognitive task.	AMIF: tau = nd; FNN: D = 5; R = 10% of the maximum diameter of the reconstructed attractor. A minimum number of three points was employed to detect diagonal lines. With these input parameters, distance matrices and recurrence plots were computed, from which the four RQA measures were extracted. %REC, %DET, ENT (Shannon), TREND.	For both leg standing: %DET and ENT in the AP and ML directions were higher for ACLD patients compared to the H participants. When postural difficulty was increased—%DET and ENT significantly increased. When cognitive difficulty was increased (single to dual-task conditions), %DET and ENT significantly decreased. For the single leg standing: %DET in the AP and ML directions and ENT in the ML direction were significantly higher for both the injured and non-injured leg in the ACLD group when compared to the H participants. When cognitive difficulty was increased (single to dual-task conditions), %DET and ENT significantly decreased.
Mazaheri et al. [56] 17/32	22 patients with low back pain (LBP) (13 M, 9 F): Age: 26.1 ± 6.2; Weight: 67.1 ± 11.2; Height: 172 ± 10. 22 healthy adults (13 M, 9 F): Age: 25.0 ± 5.5; Weight: 66.5 ± 12.1; Height: 173 ± 10.	To determine whether people with LBP show atypical dynamic patterns of postural sway when performing attention-demanding tasks compared with their matched control participants.	Bertec 4060-10 force platform (200 Hz)—down-sampled without filtering to 100 Hz. 9 trials (30 s) with combination of: 3 postural tasks: both leg standing on rigid force platform with EO; with EC; on the foam with EC. 3 levels of cognitive difficulty (no-task, easy, and difficult).	tau—was calculated for each case using the AVD; D = 5; R = 10% of the maximum diameter of the reconstructed attractor. %REC, %DET, ENT, TREND	As postural conditions became more difficult, the postural sway in both AP and ML directions became more regular (higher %REC and %DET), more complex (higher ENT), and more stationary (lower TREND). The only exception was for %REC, which decreased in the foam and EC condition. Increasing cognitive difficulty was associated with less regularity (lower %REC and %DET), less complexity (lower ENT), and more stationary (lower TREND) in both directions.
Negahban et al. [47] 17/32	23 patients with multiple sclerosis (MS) (8 M, 15 F): Age: 32.7 ± 7.9; Weight: nd; Height: 160 ± 14; BMI = 24.1 ± 3.8. 23 healthy adults (8 M, 15 F): Age: 31.4 ± 7.9; Weight: nd; Height: 160 ± 96; BMI = 24.6 ± 3.8.	The characterization of the nonlinear dynamical structure of postural sway obtained from the RQA in MS patients as compared to healthy controls. Investigating the effect of a cognitive task on postural control.	Bertec 4060-10 force platform (100 Hz). 3 trials (30 s) of combinations of three levels of postural difficulty (both leg standing with EO, EC, and on foam surface with EC) with two levels of cognitive difficulty (single and dual-tasks). 5 min rest.	AMIF: tau = nd; FNN: D = 5; R = 10% of the maximum phase space diameter using Euclidean norm. This value kept the recurrence rate below 5% in most cases. %REC, %DET, ENT, TREND.	For both groups: as the postural conditions became more difficult, the CoP time series were characterized by lower %REC, ENT, and TREND. As cognitive conditions became more difficult, CoP time series were characterized by lower %REC, ENT and TREND in AP direction and lower %DET in both directions.
Cao et al. [57] 16/32	89 patients with multiple sclerosis (MS) (20 M, 69 F): Age: 46.11 ± 10.26 Weight: nd; Height: nd. 29 healthy adults (8 M, 21 F): Age: 34.79 ± 12.31; Weight: nd; Height: nd.	To estimate Expanded Disability Status Scale (EDSS) scores by using values extracted from the posturographic data of MS patients, in order to select the most pertinent parameters.	Satel force plate (40 Hz). 1 trial (51.2 s) both leg standing with EC.	Linear measures: CoP path length, sway velocity. tau = 1/40 s; D = 1; R = 15 mm for maximum correlation between EDSS and each recurrence estimate. %REC, ENT (Shannon), LMAX, TT. RQA parameters were calculated for position, instantaneous velocity and acceleration of the CoP.	This study emphasizes the possibility of distinguishing EDSS scores using postural sway and RQA parameters. %REC and LMAX were in best agreement with clinical assessments.
Cao et al. [58] 17/32	117 patients with multiple sclerosis (MS) (26 M, 91 F): Age: 46.53 ± 11.31; Weight: nd; Height: nd. 16 healthy adults (4 M, 12 F): Age: 44.59 ± 6.12; Weight: nd; Height: nd.	To evaluate strategy in both mono- and multi-dimensional analyses in order to determine the most discriminative RQA measure for estimating EDSS.	Satel force plate (40 Hz). 1 trial (51.2 s) both leg standing with EC.	Linear measures: CoP path length, sway velocity. Mono-dimensional analysis: AMIF: tau = 1 FNN: D = 1 Multi-dimensional analysis: AMIF: tau = 1 to 30 samples; FNN: D = between 2 and 20; R1 = 20% of the maximal distance of one participant; R2 = equal to the mean of the series calculated in R1. REC, ENT, LMAX, TT, TREND.	The unidimensional RQA of the CoP signal is a more accurate method for quantifying disability in MS patients than multidimensional RQA.
Negahban et al. [59] 18/32	27 patients with knee osteoarthritis (OA) (8 M, 19 F): Age: 54.8 ± 6.4; Weight: nd; Height: 162.4 ± 8.4; BMI = 27.9 ± 2.9. 27 healthy adults (8 M, 19 F): Age: 54.4 ± 6.2; Weight: nd; Height: 162.8 ± 8.7; BMI = 28 ± 3.2.	To investigate differences in the complexity and variability of the CoP dynamics between OA patients and healthy controls.	Bertec 4060-10 force platform (100 Hz). 3 trials for 8 conditions (60 s): both leg standing on rigid and foam surface. Each trail with EO; with EO and cognitive task; with EC; with EC and cognitive task. 5 min break	Second-order difference (SOD) plots quantified with central tendency measure (CTM). AMIF: tau = 3; D = 5; R = 10% of the maximum phase space diameter using Euclidean norm. No filtering data. %DET	Knee OA patients had a greater %DET in ML direction than healthy subjects. In both groups, the %DET increased with increasing postural difficulty, while %DET decreased when moving from single to dual-task.
Shokouhyan et al. [45] 15/32	20 males with non-specific low back pain (NSLBP): Age: 24.5 ± 0.9; Weight: 62 ± 7.5; Height: 172 ± 7.5. 20 healthy adults (M): Age: 25.5 ± 0.7; Weight: 64 ± 8.6; Height: 174 ± 6.5	Evaluation and comparison of proprioceptive parameters (body sway and stability) between patients with NSLBP and healthy controls.	Bertec force plate (100 Hz) and Vicon motion capture system (100 Hz). In-house vibrator apparatus VR (70 Hz) to alter proprioception of the soleus and lumbar muscles. 8 trails (30 s) with EC standing on a rigid and foam surface: without VR; with the VR of the triceps; with VR of the multifidus; with the VR of both muscles.	Linear measures: CoP path length, velocity Nonlinear measures: Lyapunov exponents. AMIF: tau = 0.35–0.6 s; FNN: D = 3 or 4; R = 2.5% of the mean distance. RQA software developed by Webber [60] %REC, %DET, ENT, TREND.	An increase of the standard deviation of amplitude and velocity among the NSLBP participants. %DET and ENT were greater in the NSLBP group. RQA parameters for the CoP on both sides and for the trunk sagittal angle indicated more repeated patterns of movement among the NSLBP group.
Sport/Influence of motion/Impact of visual stimuli
Schmit et al. [18] 16/32	10 undergraduate dance majors (5 M, 5 F): Age: 20; Weight: nd; Height: nd. 10 runners (5 M, 5 F): Age: 19.5; Weight: nd; Height: nd.	To compare postural stability and postural sway dynamics in dancers and a control group. To determine if dancers merely exhibit a different amount of postural sway variability or if, in addition (or instead), they exhibit qualitatively different dynamic patterns of postural sway.	Bertec 4060-NC force platform (100 Hz). 4 trials (30 s) both leg standing with EO/EC on rigid/foam surface. Breaks as needed.	Linear measures: CoP path length. For AP sway: tau = 9 samples; D = 11; R = 24% of the mean distance separating points in reconstructed phase space. For ML sway: tau = 10 samples; D = 10; R = 20% of the mean distance separating points in reconstructed phase space. %REC, %DET, ENT, LMAX, TREND.	RQA revealed that the postural sway of dancers was less regular (lower %REC), less stable (lower LMAX), less complex (lower ENT), and more stationary (lower TREND) than that of track athletes.
Clark and Riley [49] 15/32	12 college students (4 M, 8 F): Age: 20 (18–24); Weight: nd; Height: 169.8 (154.9–190.5)	To investigate the multisensory control of posture by altering sensory information across the visual and somatosensory systems.	Smart Balance Master® force plate (100 Hz). 6 trials (20 s) for 3 gain settings for 4 conditions (total 72 trials): (1) EO with sway-referenced visual surround (SV); (2) EO with support surface (SS); (3) EC/SS; (4) EO/SV-SS. 2 min rest.	Linear measures: CoP path length, the coefficient of variation of CoP displacements. AMIF: tau = 0.07 s; FNN: D = 9; R = 11% of the mean Euclidean distance separating data points. %DET, TREND.	The CoP was less random (higher %DET) the higher the gain condition was. The CoP became increasingly more deterministic across more challenging sensory organization test conditions and with increasing gain, and more nonstationary across more challenging conditions and when the support surface was sway-referenced using a 1.8 gain setting.
Ferrufino et al. [43] 19/32	16 contemporary dancers (CD) (16 F): Age: 73.7 ± 5.5; Weight: nd; Height: nd; BMI: 26.9 ± 4.8. 25 fall prevention (FP) (2 M, 23 F): Age: 72.9 ± 7.2; Weight: nd; Height: nd; BMI: 24.6 ± 3.3.	To study the effects of two physical activities (contemporary dance and fall prevention training) on postural control in older adults.	Techno-Concept platform (40 Hz). 4 conditions (51.2 s): EC, EO with vision fixed on a point at a distance of 600, 150, or 40 cm. Recordings before and after intervention.	Linear measures: CoP path length, area of ellipse, the mean velocity of the CoP displacements. Convex hull area including 100% of the positions of the CoP, the Romberg quotient, and the fractal dimension ratio. tau = 50 ms (corresponding to 2 samples); D = 12; R = 2–3% of the mean distance between data points. %REC, %DET, LMAX.	CD reduced both %REC and mathematical stability (%LAM) suggesting that CoP displacements were less likely to repeat themselves over time and that their dynamics were more flexible, whereas FP yielded the opposite tendency.
Markovic et al. [61] 16/32	40 young adults (22 M, 18 F): Age: 27 ± 2.6; Weight: 71 ± 13; Height: 176 ± 7. 34 older persons (15 M, 19 F): Age: 79 ± 2.8; Weight: 71 ± 14; Height: 165 ± 9	To evaluate the intra-session reliability of traditional and nonlinear measures of postural sway during upright semi-tandem quiet stance in young vs. older subjects.	AMTI force plate (1000 Hz). 3 trials (30 s) of semi-tandem stance with EO. 3 min break	Linear measures: 13 parameters, nonlinear measures: 20 parameters tau = nd; D = nd; R = nd. %REC, %DET, ENT, %LAM, TT, LMAX. Removal of the potential noise from the CoP signal (2nd order Butterworth, 0.1–20 Hz band-pass, bidirectional filter).	The RQA (%DET, ENT) and sample entropy analyses, but not diffusion plot and sway density plot analyses, can be reliably used in young and elderly age groups.
Apthorp et al. [48] 15/32	13 young adults (3 M, 10 F): Age: 20.9 ± 0.8; Weight: 75 ± 17; Height: 171.7 ± 9.1	To investigate if strength of visually induced illusions of self-motion (vection) depends on the extent to which people rely on vision to maintain their postural stability.	Bertec force plate (1000 Hz). 3 trials (60 s) both leg standing with EO and EC. 2 sessions (30 s) blank period, followed by 30 s of the optic flow stimulus, 30 s of a blank screen, and 30 s of a simple fixation screen. Then seated trials were performed.	Linear measures: CoP path length, area of ellipse, standard deviation, and sway magnitude. AMIF: tau = 15; FNN: D = 8; R = 30. %REC, %DET, %LAM, TT, ENT, LMAX	The ratio between eyes-open and eyes-closed CoP excursions during quiet stance (using the area of postural sway) significantly predicted seated vection for both measures. RQA was a much stronger predictor of seated vection for both expanding and contracting stimuli.
Coubard et al. [62] 17/32	19 contemporary dancers (CD) (19 F): Age: 70.6 ± 7.3; Weight: nd; Height: nd; BMI = 26.5 ± 3.8. 19 no dancers (ND) (1 M, 18 F): Age: 72.6 ± 8.6; Weight: nd; Height: nd; BMI = 27.0 ± 3.9.	To investigate the effects of a dance practice (CD), on the quality of posture in older adults.	Techno-Concept platform (40 Hz). 1 trial (51.2 s) both leg standing with EO, EC. Recordings before and after training intervention.	Linear measures: CoP path length, ellipse area, the convex hull area, which include 90% and 100% of CoP positions, respectively. Romberg quotient and Signal Diffusion Analysis (SDA). Nonlinear measures: Detrended Fluctuation Analysis (DFA). tau = 50 ms; D = 12; R = 2–3% of the mean distance between data points. %REC, %DET, LMAX	The ND group showed an increase in length and mean velocity of postural signal. While both legs standing with EO, LMAX decreases in AP direction. In ML direction, %DET and LMAX showed higher values in post-test period. In AP direction, LMAX showed higher values in post-test period.
Decker et al. [63] 16/32	126 healthy women: Age: 65.7 ± 4.1; Weight: 66.4 ± 12.2; Height: 161.9 ± 6.3.	To determine whether postural sway complexity could discriminate asymptomatic sedentary postmenopausal women with normal or subnormal physical function from those with lower physical function.	Medicapteurs force platform (40 Hz). 1 trial (51.2 s) both leg standing with EO and EC. 30 s break between trials.	Six-Minute Walking Distance test. Linear measures: CoP path length, ellipse area, standard deviations in both directions. Nonlinear measures: sample entropy (SampEn), complexity index of multiscale entropy (CIMSE). tau = 6; D = 8; R = 0.25 of the mean of all distances. %REC, %DET, ENT Cross Recurrence Plot Toolbox 5.13 (R25) [6]	Traditional measures did not differentiate women with lower, subnormal, or normal physical function. Women with lower physical function showed lower SampEn values and higher %DET values in ML direction during standing with EC. No significant difference in the CIMSE values was found between the two groups.
Bernard et al. [64] 18/32	49 Sedentary community-dwelling women: Age: 64.98 ± 3.87; Weight: 65.54 ± 11.37; Height: 160.6 ± 5.35	To investigate whether aerobic exercise at submaximal intensity has detrimental effects on balance in older sedentary adults.	Mdicapteurs force platform (40 Hz). 1 trial (51.2 s) both leg standing with EO and EC. 30 s break between trials. Than 2 trials of 6 min walk tests (6 MWT). After each test: 1 trial (51.2 s) both leg standing with EC. 20 min break between 6 MWT trials.	Linear measures: CoP path length, ellipse area, the mean along the AP and ML axes, central tendency measure (CTM). tau = 6; D = 8; R = 25% of mean of all the distances. %DET, ENT. Cross Recurrence Plot Toolbox 5.13 (R25) [6].	In older, sedentary women, two consecutive cycles of 6 min aerobic exercise separated by a 20 min rest period did not significantly change kinematic measures, only measures of complexity. This study showed the effects of walking exercise on balance parameters: CTM_AP, %DET_ML and ENT_ML. The significant increase in the %DET_ML represents an increase in the predictability of the time series, a marker of reduced postural complexity and less adaptive postural capabilities in the ML direction, according to the loss of complexity of aging and disease theory. A possible interpretation of this finding is a less adaptive behavior related to a hip muscle strategy adversely affected by the exercise of walking.
Palmisano et al. [65] 18/32	23 adults (7 M, 16 F): Age: 22.9 ± 6.0; Weight: nd; Height: nd.	To compare the predictive ability of traditional and RQA measures of postural activity.	Bertec force plate (1000 Hz). 1 trial (60 s) both leg standing with: EO and EC. After these trials, subjects were seated and exposed to 5 smooth and 5 oscillating radial flow displays (random order: 30 s). 60 s break between vection trials. Display apparatus (60 Hz).	Linear measures: CoP path length, ellipse area. Tau = 15; D = 8; R = 0.6 Minimum line length of 2. %REC, %DET, ENT, LMAX Recurrence Quantification Toolbox for MATLAB [66].	%REC in AP direction significantly predicted the strength of oscillating vection. %REC during trial with EO significantly predicted the strength of the vection induced by oscillating radial flow. %REC, %DET, and ENT along the AP axis correlated significantly with the average reported strength of the smooth vection. %REC, ENT, and LMAX along the AP axis also correlated significantly with the average reported strength of the oscillating vection. Lower %REC, ENT, and shorter LMAX were associated with stronger vection (both smooth and oscillating). Lower %DET was also associated with stronger smooth vection.
Bernard et al. [67] 18/32	Trained (61 F): Age: 65.46 ± 4.37; Weight: 64.14 ± 12.06; Height: 161.25 ± 6.29. Controls (60 F): Age: 65.54 ± 4.04; Weight: 70.06 ± 11.57; Height: 162.53 ± 6.6.	To evaluate the influence of an active 6-month walking program of three 60 min sessions per week at 60–80% of maximum heart rate on sedentary elderly women’s kinematic and dynamic parameters of the CoP.	Medicapteurs force platform (40 Hz). 1 trial (51.2 s) both leg standing with EO and EC. 30 s break between trials.	Linear measures: CoP path length, ellipse area, the mean along the AP and ML axes. Nonlinear measures: sample entropy (SampEn) and multiscale entropy (MSE). tau = nd; D = nd; R = nd. % DET Cross Recurrence Plot Toolbox 5.13 (R25) [6].	In older sedentary women, this study indicated a systemic lack of influence of 6 months’ walk-training on flat ground on kinematic postural responses and on dynamical measures.
van den Hoorn et al. [20] 19/32	106 elderly (64 M, 42 F): Age: 75 ± 6; Weight: 78 ± 15; Height: 169 ± 9. 23 young healthy adults (9 M, 14 F): Age: 21 ± 2; Weight: 65 ± 10; Height: 172 ± 5.	To compare CoP motion between young and older individuals before, during, and after removal of bilateral calf vibration, and to compare measures between older individuals who subsequently do or do not go on to fall in the following 12 months.	Kistler force plate (2000 Hz). 1 trial (135 s) both leg standing with EC. Vibrators (60 Hz) attached bilaterally over the triceps surae muscles were activated twice for 15 s; after 15 and 75 s (45 s for recovery).	Linear measures: CoP path length, CoP position relative to the mean CoP position at the baseline. tau = 180 ms; D = 5; R = 5%. %DET, LMEAN, %LAM, TT. RQA toolbox [6].	At baseline (before vibration), compared to fallers, young people had lower %DET (more predictable), had longer LMEAN (less sensitive to small perturbations), and had higher %LAM with longer TT (were more intermittent, with longer static episodes). Compared to non-fallers, young people also had longer LMEAN, TT, and were less anti-persistent. %DET and %LAM were not significantly different between young and non-fallers.
Chowdhury et al. [68] 16/32	15 young adults (11 M, 4 F): Age: 24.6 ± 5.8; Weight: nd; Height: nd.	To examine the effects of simulating self-motion via a head-mounted display (HMD) on standing postural sway and spatial presence.	Bertec force plate (1000 Hz). Oculus Rift CV1 HMD. 1 trial both leg standing: 30 s stationary screen followed by 60 s of screen motion.	Linear measures: CoP path length, CoP sway variability. The spectrum function and Fourier analysis. tau = 15; D = 8; R = 0.6 and line minimum of 2. %REC.	%REC declined as the display oscillation amplitude increased.

Abbreviations: nd—no data, M—males, F—females, N—number of people in the group, y.o.—years old, CoP—center of pressure, AP—anterior-posterior, ML—medial-lateral direction, EO—eyes open, EC—eyes closed, L—left, R—right, FNN—false nearest neighbors, AMIF—average mutual information function, ACF—autocorrelation technique, AVD—the average displacement method, DFA—detrended fluctuation analysis, SDA—signal diffusion analysis, CTM—central tendency measure, PD—Parkinson’s disease, ACLD—patients with anterior cruciate ligament deficiency, LBP—low back pain, MS—multiple sclerosis, EDSS—Expanded Disability Status Scale, OA—knee osteoarthritis, NSLBP—non-specific low back pain, SV—visual surround, SS—support surface, CD—contemporary dancers, ND—no dancers, HMD—head-mounted display.

3.1. Children/Young/Elderly

Eight publications were included in this group. A total of 652 participants were analyzed. Their mean age ranged from 3 [44] to 85.7 ± 6.7 years [37]. Their mean body weight and height values ranged from 17.35 ± 3.16 kg [44] to 86.5 kg [42] and from 103.44 ± 4.9 cm [44] to 194 cm [42], respectively. Two types of force platforms were used to measure CoP displacement: AMTI in five papers [20,42,44,51,54] and Medicapteurs in three [37,50,53]. Their sampling frequency ranged from 40 Hz [37,50,53] to 1000 Hz [17,44]. In the remaining papers, the sample rate was set at 100 Hz. Only Hourova et al. [54] did not specify its value. Describing the conditions of the measurements, the three authors van den Hoorn et al. [17], Hasson et al. [42], and Hourova et al. [54] evaluated trials lasting 30 s both leg standing with and without visual control. Ramdani et al. [37], Seigle et al. [50], and Tallon et al. [53] evaluated 51.2 s both leg standing attempts with eyes open. King et al. [51] evaluated trials lasting 60 s in various foot placement configurations. The shortest measurements of 15 s were made by Hao et al. [44]. Three studies [17,42,54] additionally evaluated CoP fluctuation on the foam surface with a visual inspection.

It is worth noting in three papers [17,37,53], the Cross Recurrence Plot Toolbox [12] was used to calculate RQA parameters. Hasson et al. [42] made calculations in MATLAB according to Webber and Zbilut [10]. Hao et al. [44], on the other hand, used Marwan’s RQA toolbox [6]. The other authors did not specify which programs they used.

To perform RQA analysis, time delay, embedding dimension, and threshold were used. Time lag values ranged from 1 to 30, but tau = 6 was the most common [37,50,53]. The embedding dimension ranged from 2 to 20, but D = 8 was the most frequent value [37,50,53]. The threshold was defined in three ways. (1) As a percent of mean distance; (2) as a number of data points; or (3) as a percent of recurrence rate. In the first case, values were between 20% and 35%. A number of data points were used only in one study [51], where R = 2. A percent of the recurrent rate was utilized in two publications [17,44]. In both cases, R = 5% of the recurrent rate. Hourova et al. [54] did not provide information about tau, D, or R values. To improve the within-session reliability of RQA, van den Hoorn et al. [17] suggest fixing the recurrence rate.

Table 3 includes RQA parameter values specific to the children/young/elderly group. The authors of the two papers did not provide any values. The %REC values were provided only by King et al. [51]. They were on the plots; therefore, it was possible to read numerically only the averages. However, it is noteworthy that the highest %REC values were for fluctuations in the AP direction. %DET was present in every paper except the two mentioned. The %DET was higher for standing with eyes open compared to trials with eyes closed [17,44,50]. In addition, %DET values were higher in the ML direction in four papers [37,44,50,53]. ENT values were noted only in four studies [37,50,51,53]. Ramdani et al. [37] and Tallon et al. [53] showed that its higher values were in the ML direction. Seigle et al. [50] reported similarly, but only for trials with closed eyes. This trend continued in tests with visual control included, but only in the elderly group. In only two studies [17,44], authors included %LAM, and in one—%LMAX and TT. All authors showed that %LAM remained higher while standing with eyes open compared to values in trials without visual control. Hao et al. [44] showed that %LAM increases with age, but only for the AP direction when standing with eyes open and closed. Only van den Hoorn et al. [17] included %LMAX and TT. Both coefficients had higher values in trials involving visual control.

Table 3. Data extracted from reviewed articles RQA parameters in the children/young/elderly group.

Study	%REC	%DET	ENT	%LAM, LMAX, TREND, TT
Hasson et al. [42]	No data
Seigle et al. [50]	No data	Young group (EO/EC): ML: 96.2 ± 1.4/96.4 ± 2 AP: 95.1 ± 3.6/94.8 ± 4 Elderly group (EO/EC): ML: 94.8 ± 6.5/92.8 ± 6.5 AP: 88.7 ± 11.1/86.6 ± 8.6	Young group (EO/EC): ML: 3.7 ± 0.2/3.7 ± 0.3 AP: 3.8 ± 0.5/3.6 ± 0.5. Elderly group (EO/EC): ML: 3.77 ± 0.53/3.44 ± 0.55 AP: 3.25 ± 0.55/3.00 ± 0.46	No data
King et al. [51]	AP/ML: Side-by-side: 4.39/2 Staggered-R: 3.21/1.87 Staggered-L: 2.42/1.64 Tandem-R: 1.59/0.64 Tandem-L: 2.04/0.92	ML: Side-by-side: 34.84 Staggered-R: 73.1 Staggered-L: 71.32 Tandem-R: 92.55 Tandem-L: 90.89	AP/ML_EO: Side-by-side: 4.23/1.3 Staggered-R: 3.94/nd Staggered-L: 3.81/nd Tandem-R: 3.36/nd Tandem-L: 3.45/nd	No data
Tallon et al. [53]	No data	AP/ML: 0.934 ± 0.061/0.965 ± 0.035	AP/ML: 3.721 ± 0.426/3.902 ± 0.356	No data
Ramdani et al. [37]	No data	Non-fallers (AP/ML): 76.2 ± 22.6/90.6 ± 11.5 Fallers (AP/ML): 85.5 ± 15.9/96.8 ± 3.0	Non-fallers (AP/ML): 3.0 ± 0.8/3.6 ± 0.7 Fallers (AP/ML): 3.5 ± 0.8/4.1 ± 0.4	No data
van den Hoorn et al. [17]	No data	EO/EC firm surface (first repetition and fixed recurrence threshold): 89.4 ± 5.7/88.9 ± 6.0 EO/EC foam surface (first repetition and fixed recurrence threshold): 93.4 ± 3.1/92.9 ± 3.4	No data	%LAM: EO/EC firm surface (first repetition and fixed threshold): 93.6 ± 3.8/92.9 ± 4.1 EO/EC foam surface (first repetition and fixed threshold): 95.7 ± 2.2/95.0 ± 2.7 %LMAX: EO/EC firm surface (first repetition and fixed threshold): 2.274 ± 0.714/1.985 ± 0.574 EO/EC foam surface (first repetition and fixed threshold): 2.030 ± 0.750/1.888 ± 0.650 TT: EO/EC firm surface (first repetition and fixed threshold): 0.366 ± 0.099/0.323 ± 0.083 EO/EC foam surface (first repetition and fixed threshold): 0.333 ± 0.123/0.298 ± 0.104
Hourova et al. [54]	No data
Hao et al. [44]	No data	AP direction (I/II/III groups) EO: 72.88 ± 8.51/75.06 ± 7.03/79.07 ± 6.62 ML direction (I/II/III groups) EO: 80.66 ± 6.46/80.10 ± 5.78/79.50 ± 6.55 AP direction (I/II/III groups) EC: 70.60 ± 7.81/73.82 ± 7.19/77.38 ± 7.04 ML direction (I/II/III groups) EC: 81.47 ± 4.85/78.95 ± 6.53/79.94 ± 5.44	No data	%LAM: AP (I/II/III groups) EO: 77.46 ± 6.66/78.94 ± 5.60/82.95 ± 5.27 ML (I/II/III groups) EO: 83.36 ± 5.03/82.89 ± 5.30/81.73 ± 4.94 AP (I/II/III groups) EC: 75.76 ± 6.30/77.93 ± 5.60/81.19 ± 5.25 ML (I/II/III groups) EC: 82.78 ± 3.87/80.64 ± 5.94/82.26 ± 4.39

Abbreviations: AP—anterior-posterior direction, ML—medial-lateral direction, EO—eyes open, EC—eyes closed.

Table 3. Data extracted from reviewed articles RQA parameters in the children/young/elderly group.

Study	%REC	%DET	ENT	%LAM, LMAX, TREND, TT
Hasson et al. [42]	No data
Seigle et al. [50]	No data	Young group (EO/EC): ML: 96.2 ± 1.4/96.4 ± 2 AP: 95.1 ± 3.6/94.8 ± 4 Elderly group (EO/EC): ML: 94.8 ± 6.5/92.8 ± 6.5 AP: 88.7 ± 11.1/86.6 ± 8.6	Young group (EO/EC): ML: 3.7 ± 0.2/3.7 ± 0.3 AP: 3.8 ± 0.5/3.6 ± 0.5. Elderly group (EO/EC): ML: 3.77 ± 0.53/3.44 ± 0.55 AP: 3.25 ± 0.55/3.00 ± 0.46	No data
King et al. [51]	AP/ML: Side-by-side: 4.39/2 Staggered-R: 3.21/1.87 Staggered-L: 2.42/1.64 Tandem-R: 1.59/0.64 Tandem-L: 2.04/0.92	ML: Side-by-side: 34.84 Staggered-R: 73.1 Staggered-L: 71.32 Tandem-R: 92.55 Tandem-L: 90.89	AP/ML_EO: Side-by-side: 4.23/1.3 Staggered-R: 3.94/nd Staggered-L: 3.81/nd Tandem-R: 3.36/nd Tandem-L: 3.45/nd	No data
Tallon et al. [53]	No data	AP/ML: 0.934 ± 0.061/0.965 ± 0.035	AP/ML: 3.721 ± 0.426/3.902 ± 0.356	No data
Ramdani et al. [37]	No data	Non-fallers (AP/ML): 76.2 ± 22.6/90.6 ± 11.5 Fallers (AP/ML): 85.5 ± 15.9/96.8 ± 3.0	Non-fallers (AP/ML): 3.0 ± 0.8/3.6 ± 0.7 Fallers (AP/ML): 3.5 ± 0.8/4.1 ± 0.4	No data
van den Hoorn et al. [17]	No data	EO/EC firm surface (first repetition and fixed recurrence threshold): 89.4 ± 5.7/88.9 ± 6.0 EO/EC foam surface (first repetition and fixed recurrence threshold): 93.4 ± 3.1/92.9 ± 3.4	No data	%LAM: EO/EC firm surface (first repetition and fixed threshold): 93.6 ± 3.8/92.9 ± 4.1 EO/EC foam surface (first repetition and fixed threshold): 95.7 ± 2.2/95.0 ± 2.7 %LMAX: EO/EC firm surface (first repetition and fixed threshold): 2.274 ± 0.714/1.985 ± 0.574 EO/EC foam surface (first repetition and fixed threshold): 2.030 ± 0.750/1.888 ± 0.650 TT: EO/EC firm surface (first repetition and fixed threshold): 0.366 ± 0.099/0.323 ± 0.083 EO/EC foam surface (first repetition and fixed threshold): 0.333 ± 0.123/0.298 ± 0.104
Hourova et al. [54]	No data
Hao et al. [44]	No data	AP direction (I/II/III groups) EO: 72.88 ± 8.51/75.06 ± 7.03/79.07 ± 6.62 ML direction (I/II/III groups) EO: 80.66 ± 6.46/80.10 ± 5.78/79.50 ± 6.55 AP direction (I/II/III groups) EC: 70.60 ± 7.81/73.82 ± 7.19/77.38 ± 7.04 ML direction (I/II/III groups) EC: 81.47 ± 4.85/78.95 ± 6.53/79.94 ± 5.44	No data	%LAM: AP (I/II/III groups) EO: 77.46 ± 6.66/78.94 ± 5.60/82.95 ± 5.27 ML (I/II/III groups) EO: 83.36 ± 5.03/82.89 ± 5.30/81.73 ± 4.94 AP (I/II/III groups) EC: 75.76 ± 6.30/77.93 ± 5.60/81.19 ± 5.25 ML (I/II/III groups) EC: 82.78 ± 3.87/80.64 ± 5.94/82.26 ± 4.39

Abbreviations: AP—anterior-posterior direction, ML—medial-lateral direction, EO—eyes open, EC—eyes closed.

3.2. Disability/Disease

Eight publications were included in this group. A total of 501 participants were analyzed. Their mean age ranged from 24.5 ± 0.9 [45] to 70.83 ± 15.89 years [55]. Their mean body weight and height values ranged from 62 ± 7.5 kg [45] to 67.1 ± 11.2 kg [56], and from 160 ± 14 cm [47] to 179 ± 55 cm [46], respectively. At this point, it is worth noting that most of the authors did not provide a complete characterization of the group. In particular, height and weight were missing. In this group, Bertec platforms with a sampling rate of 100 Hz were used to evaluate fluctuations. In two papers [57,58], Satel platforms with a much lower sampling frequency of 40 Hz were used.

In this group, most studies [45,46,47,55,56] evaluated CoP fluctuations in trials lasting 30 s with and without visual control. Negahban et al. [59] evaluated trials lasting 60 s with both legs standing on rigid and foam surfaces. Cao et al. [57] and Cao et al. [58] evaluated the trials lasting 51.2 s with both feet standing only with eyes closed.

In this group, only Shokouhyan et al. [45] stated with what tool the RQA parameters were calculated. As in the previous section, the lag time, embedding dimension, and threshold had to be determined to perform RQA. Negahban et al. [46,47,59] did not specify a lag time in any of their papers. In all the papers belonging to the disability/disease group, the embedding dimension was counted using FNN. In four papers, it was 5, and in two papers: D = 1. The radius for recurrence was considered 10% of the maximum phase space diameter using the Euclidean norm in studies by Negahban et al. [46,47,59] and Mazaheri et al. [56]. Other authors used different values: 32% and 2.5% of the mean Euclidean distance separating points in the reconstructed phase space [45,55]. The two authors [57,58] used a completely different approach, resulting from a more complicated methodology and purpose of the study.

Table 4 includes RQA parameter values specific to the disability/disease group. Only Cao et al. [57] did not provide any values. It is remarkable that in this group, almost all RQA measures were used in four papers [45,47,55,56]. The least information was provided by Negahban and co-authors. They used only %DET in the study [59] and %DET and ENT in the paper [46]. In [47,55], the %REC value was higher for standing with visual control than standing with eyes closed in the ML direction. Mazaheri et al. [56] obtained an inverse relationship in the AP direction while standing on a rigid surface in both study groups. In several studies [45,47,56], healthy subjects had higher %REC values. However, this is not a rule. %DET and ENT were reported in every paper except the one mentioned. In most of the studies, the values of both parameters were higher for healthy people.

The last group of RQA parameters in Table 4 is the most variable. Schmit et al. [55] reported only LMAX values, showing that these values were higher in the group of Parkinson’s patients. Cao et al. [58] reported LMEAN values as a function of embedding dimension and time lag. They were also the only ones to report TT values. Three papers [45,47,56] included TREND data. In each study, the values of this parameter were negative and lower in the healthy group when standing with eyes open on a rigid surface. The one exception to this rule was the result of Negahban et al. [47].

Table 4. Data extracted from reviewed articles RQA parameters in the disability/disease group, where: * denotes statistically significant differences.

Study	%REC	%DET	ENT	%LAM, LMAX (Maximum Diagonal Line Length) LMEAN (Average Diagonal Line Length) TREND, TT (Trapping Time)
Schmit et al. [55]	PD/Healthy AP: 9.09 ± 4.3/5.37 ± 3.41 * EO/EC: ML: 10.11 ± 2.85/7.42 ± 3.64 *	PD/Healthy AP: 80.86 ± 28.29/52.68 ± 30.56 *	PD/Healthy AP: 3.57 ± 1.70/1.98 ± 1.24 *	LMAX: PD/Healthy AP: 2138 ± 1140/908 ± 1165 *
Negahban et al. [46]	No data	ACLD/Healthy Rigid surface, no task EO: AP: 92.06 ± 2.99/90.40 ± 5.45 ML: 93.65 ± 2.21/91.52 ± 2.60 Rigid surface, no task EC: AP: 94.63 ± 2.13/92.80 ± 2.39 ML: 95.44 ± 1.88/94.82 ± 2.32 Foam surface, no task EC: AP: 97.86 ± 0.83/97.19 ± 0.95 ML: 97.62 ± 0.67/97.30 ± 0.75	ACLD/Healthy Rigid surface, no task EO: AP: 4.72 ± 0.40/4.55 ± 0.55 ML: 4.98 ± 0.42/4.65 ± 0.34 Rigid surface, no task EC: AP: 5.22 ± 0.41/4.83 ± 0.39 ML: 5.46 ± 0.46/5.35 ± 0.50 Foam surface, no task EC: AP: 6.06 ± 0.38/5.81 ± 0.38 ML: 5.90 ± 0.26/5.86 ± 0.32	No data
Mazaheri et al. [56]	LBP/Healthy Rigid surface, no task EO: AP: 8.36 ± 2.5/8.75 ± 3.73 ML: 6.62 ± 2.33/9.09 ± 2.8 Rigid surface, no task EC: AP: 8.77 ± 2.97/9.92 ± 2.78 ML: 8.27 ± 2.49/8.21 ± 1.7 Foam surface, no task EC: AP: 7.63 ± 2.12/8.34 ± 3.52 ML: 6.60 ± 1.82/6.98 ± 2.57	LBP/Healthy Rigid surface, no task EO: AP: 96.73 ± 1.6/97.24 ± 1.79 ML: 96.19 ± 1.27/96.45 ± 1.37 Rigid surface, no task EC: AP: 97.18 ± 1.41/97.59 ± 1.66 ML: 96.49 ± 1.48/96.62 ± 1.57 Foam surface, no task EC: AP: 98.89 ± 0.40/98.84 ± 0.59 ML: 98.17 ± 0.69/98.40 ± 0.71	LBP/Healthy Rigid surface, no task EO: AP: 5.71 ± 0.52/6.04 ± 0.66 ML: 5.47 ± 0.38/5.71 ± 0.51 Rigid surface, no task EC: AP: 5.96 ± 0.57/6.91 ± 0.59 ML: 5.71 ± 0.45/5.76 ± 0.53 Foam surface, no task EC: AP: 6.33 ± 0.30/6.27 ± 0.35 ML: 6.05 ± 0.30/6.07 ± 0.32	TREND: LBP/Healthy Rigid surface, no task EO: AP: −6.13 ± 2.37/−6.77 ± 4.0 ML: −4.11 ± 2.17/−4.58 ± 2.34 Rigid surface, no task EC: AP: −5.49 ± 3.07/−6.04 ± 2.73 ML: −3.73 ± 2.69/−3.71 ± 1.74 Foam surface, no task EC: AP: −3.39 ± 1.71/−3.76 ± 2.28 ML: −2.65 ± 1.12/−3.41 ± 1.39
Negahban et al. [47]	MS/Healthy Rigid surface, single task EO: AP: 2.66 ± 1.06/2.44 ± 1.02 ML: 3.34 ± 1.32/3 ± 1.4 Rigid surface, single task EC: AP: 2.5 ± 0.8/2.67 ± 0.99 ML: 2.75 ± 1.14/2.33 ± 1.11 Foam surface, single task EC: AP: 1.91 ± 0.52/2.36 ± 0.96 ML: 2.15 ± 0.87/2.3 ± 1.1 Rigid surface, dual task EO: AP: 1.99 ± 0.54/2.42 ± 0.96 ML: 2.9 ± 1.52/2.75 ± 1.08 Rigid surface, dual task EC: AP: 2.2 ± 0.79/2.44 ± 0.94 ML: 2.73 ± 0.98/2.37 ± 1.15 Foam surface, dual task EC: AP: 1.7 ± 0.69/2.07 ± 0.98 ML: 2.13 ± 1.05/2.36 ± 1.29	MS/Healthy Rigid surface, single task EO: AP: 99.79 ± 0.18/99.83 ± 0.15 ML: 99.9 ± 0.1/99.91 ± 0.12 Rigid surface, single task EC: AP: 99.81 ± 0.12/99.78 ± 0.32 ML: 99.91 ± 0.04/99.85 ± 0.12 Foam surface, single task EC: AP: 99.76 ± 0.23/99.81 ± 0.1 ML: 99.86 ± 0.07/99.85 ± 0.13 Rigid surface, dual task EO: AP: 99.71 ± 0.22/99.76 ± 0.24 ML: 99.88 ± 0.1/99.89 ± 0.11 Rigid surface, dual task EC: AP: 99.65 ± 0.26/99.79 ± 0.23 ML: 99.88 ± 0.08/99.84 ± 0.14 Foam surface, dual task EC: AP: 99.7 ± 0.19/99.74 ± 0.17 ML: 99.83 ± 0.1/99.84 ± 0.13	MS/Healthy Rigid surface, single task EO: AP: 5.85 ± 0.46/5.75 ± 0.49 ML: 6.16 ± 0.43/6.05 ± 0.52 Rigid surface, single task EC: AP: 5.73 ± 0.45/5.72 ± 0.38 ML: 5.89 ± 0.49/5.74 ± 0.59 Foam surface, single task EC: AP: 5.48 ± 0.42/5.54 ± 0.4 ML: 5.54 ± 0.43/5.63 ± 0.52 Rigid surface, dual task EO: AP: 5.57 ± 0.3/5.65 ± 0.53 ML: 5.9 ± 0.43/5.99 ± 0.5 Rigid surface, dual task EC: AP: 5.46 ± 0.48/5.68 ± 0.53 ML: 5.83 ± 0.49/5.72 ± 0.59 Foam surface, dual task EC: AP: 5.25 ± 0.43/5.31 ± 0.42 ML: 5.52 ± 0.46/5.63 ± 0.47	TREND: MS/Healthy Rigid surface, single task EO: AP: −3.01 ± 0.99/−2.8 ± 1.18 ML: −3.72 ± 1.45/−3.49 ± 1.41 Rigid surface, single task EC: AP: −2.93 ± 0.7/−2.89 ± 1.13 ML: −2.79 ± 1.05/−2.87 ± 1.35 Foam surface, single task EC: AP: −2.22 ± 0.71/−2.31 ± 0.84 ML: −2.47 ± 0.98/−2.58 ± 1.14 Rigid surface, dual task EO: AP: −2.41 ± 0.57/−2.46 ± 0.95 ML: −3.29 ± 1.23/−3.38 ± 1.61 Rigid surface, dual task EC: AP: −2.5 ± 0.92/−2.77 ± 1.28 ML: −3.08 ± 1.34/−2.87 ± 1.53 Foam surface, dual task EC: AP: −1.89 ± 0.88/−2.05 ± 0.89 ML: −2.25 ± 1.06/−2.74 ± 1.48
Cao et al. [57]	No data
Cao et al. [58]	EDSS = 2, raw data: A. D = 1, tau = 1: 0.22 B. D = 3, tau = 16: 0.09 EDSS = 2, noisy data: C. D = 1, tau = 1: 0.22 D. D = 3, tau = 16: 0.08	No data	EDSS = 2, raw data: A. D = 1, tau = 1: 3.97 B. D = 3, tau = 16: 3.97 EDSS = 2, noisy data: C. D = 1, tau = 1: 2.87 D. D = 3, tau = 16: 2.62	LMEAN: A.22.63; B. 26.41; C. 9.85; D. 9.44 TT: EDSS = 2, raw data: A. D = 1, tau = 1: 33.48 B. D = 3, tau = 16: 35.24 TT: EDSS = 2, noisy data: C. D = 1, tau = 1: 14.34 D. D = 3, tau = 16: 13.35
Negahban et al. [59]	No data	OA/Healthy Rigid surface and EO: AP_single task: 98.06 ± 1.03/97.61 ± 1.13 ML_single task: 97.03 ± 1.71/99.64 ± 344 AP_dual task: 97.65 ± 1.14/97.06 ± 1.74 ML_dual task: 96.38 ± 4.12/93.68 ± 7.24 Rigid surface and EC: AP_single task: 97.64 ± 1.40/97.61 ± 1.44 ML_single task: 97.05 ± 2.49/94.2 ± 11.66 AP_dual task: 97.66 ± 1.46/97.21 ± 1.59 ML_dual task: 96.56 ± 2.57/94.37 ± 3.69 Foam surface and EO: AP_single task: 98.78 ± 0.96/98.60 ± 0.59 ML_single task: 98.09 ± 0.81/97.50 ± 1.14 AP_dual task: 98.67 ± 0.67/98.14 ± 0.97 ML_dual task: 97.65 ± 1.48/96.72 ± 2.13 Foam surface and EC: AP_single task: 98.98 ± 0.49/98.74 ± 0.54 ML_single task: 98.52 ± 0.65/97.96 ± 0.96 AP_dual task: 98.21 ± 2.02/98.07 ± 1.01 ML_dual task: 97.64 ± 1.70/96.78 ± 1.90	No data	No data
Shokouhyan et al. [45]	NSLBP/Healthy Without vibration: AP_rigid surface: 0.24 ± 0.1/0.32 ± 0.17 ML_rigid surface: 0.13 ± 0.08/0.22 ± 0.14 AP_foam surface: 0.1 ± 0.05/0.13 ± 0.06 ML_foam surface: 0.19 ± 0.12/0.16 ± 0.09 Ankle vibration: AP_rigid surface: 0.28 ± 0.15/0.14 ± 0.08 ML_rigid surface: 0.23 ± 0.16/0.12 ± 0.06 AP_foam surface: 0.45 ± 0.28/0.11 ± 0.07 ML_foam surface: 0.4 ± 0.25/0.14 ± 0.08 Back vibration: AP_rigid surface: 0.23 ± 0.13/0.17 ± 0.09 ML_rigid surface: 0.27 ± 0.17/0.12 ± 0.06 AP_foam surface: 0.19 ± 0.09/0.15 ± 0.09 ML_foam surface: 0.33 ± 0.17/0.21 ± 0.11 Both vibration: AP_rigid surface: 0.18 ± 0.1/0.24 ± 0.12 ML_rigid surface: 0.21 ± 0.12/0.17 ± 0.07 AP_foam surface: 0.18 ± 0.13/0.09 ± 0.05 ML_foam surface: 0.18 ± 0.1/0.15 ± 0.05	NSLBP/Healthy Without vibration: AP_rigid surface: 99.81 ± 1.56/99.84 ± 1.25 ML_rigid surface: 99.42 ± 1.56/99.65 ± 1.37 AP_foam surface: 99.1 ± 1.68/99.5 ± 1.36 ML_foam surface: 99.64 ± 1.65/99.41 ± 1.89 Ankle vibration: AP_rigid surface: 99.06 ± 1.16/98.81 ± 2.16 ML_rigid surface: 99.33 ± 1.77/98.28 ± 1.41 AP_foam surface: 99.52 ± 1.13/96.44 ± 1.48 ML_foam surface: 98.83 ± 1.33/98.53 ± 1.8 Back vibration: AP_rigid surface: 99.6 ± 1.53/99.53 ± 2.08 ML_rigid surface: 99.71 ± 1.3/98.38 ± 1.2 AP_foam surface: 99.49 ± 1.81/97.67 ± 1.11 ML_foam surface: 98.63 ± 1.59/99.74 ± 1.8 Both vibration: AP_rigid surface: 99.79 ± 1.25/99.81 ± 1.46 ML_rigid surface: 99.5 ± 1.42/99.76 ± 1.52 AP_foam surface: 99.54 ± 1.77/98.26 ± 1.67 ML_foam surface: 98.93 ± 1.14/98.63 ± 1.59	NSLBP/Healthy Without vibration: AP_rigid surface: 4.6 ± 0.44/4.85 ± 0.56 ML_rigid surface: 4.28 ± 0.44/4.2 ± 0.56 AP_foam surface: 4.07 ± 0.5/4.2 ± 0.5 ML_foam surface: 4.7 ± 0.36/4.61 ± 0.59 Ankle vibration: AP_rigid surface: 3.85 ± 0.4/4.17 ± 0.53 ML_rigid surface: 4.2 ± 0.48/4 ± 0.38 AP_foam surface: 4.69 ± 0.57/3.9 ± 0.42 ML_foam surface: 4.03 ± 0.38/4.11 ± 0.54 Back vibration: AP_rigid surface: 4.31 ± 0.51/4.47 ± 0.41 ML_rigid surface: 4.05 ± 0.51/3.9 ± 0.36 AP_foam surface: 4.39 ± 0.44/3.9 ± 0.47 ML_foam surface: 4.63 ± 0.48/4.25 ± 0.5 Both vibration: AP_rigid surface: 4.46 ± 0.5/4.13 ± 0.44 ML_rigid surface: 4.57 ± 0.46/4.13 ± 0.55 AP_foam surface: 4.49 ± 0.52/4 ± 0.43 ML_foam surface: 3.95 ± 0.34/4.26 ± 0.42	TREND: NSLBP/Healthy Without vibration: AP_rigid surface: −0.42 ± 0.17/−0.62 ± 0.23 ML_rigid surface: −0.25 ± 0.08/−0.44 ± 0.2 AP_foam surface: −0.18 ± 0.07/−0.26 ± 0.1 ML_foam surface: −0.38 ± 0.18/−0.31 ± 0.12 Ankle vibration: AP_rigid surface: −0.54 ± 0.18/−0.28 ± 0.1 ML_rigid surface: −0.46 ± 0.21/−0.23 ± 0.08 AP_foam surface: −0.89 ± 0.37/−0.2 ± 0.07 ML_foam surface: −0.76 ± 0.32/−0.26 ± 0.1 Back vibration: AP_rigid surface: −0.46 ± 0.14/−0.33 ± 0.13 ML_rigid surface: −0.53 ± 0.17/−0.23 ± 0.07 AP_foam surface: −0.35 ± 0.13/−0.3 ± 0.09 ML_foam surface: −0.66 ± 0.25/−0.41 ± 0.15 Both vibration: AP_rigid surface: −0.35 ± 0.13/−0.48 ± 0.18 ML_rigid surface: −0.43 ± 0.16/−0.33 ± 0.13 AP_foam surface: −0.35 ± 0.09/−0.17 ± 0.07 ML_foam surface: −0.34 ± 0.08/−0.3 ± 0.12

Abbreviations: AP—anterior-posterior, ML—medial-lateral direction, EO—eyes open, EC—eyes closed, LBP—people with nonspecific low back pain, ACLD—patients with anterior cruciate ligament deficiency, MS—patients with multiple sclerosis, PD—Parkinson’s patients, OA—patients with knee osteoarthritis, EDSS—Expanded Disability Status Scale, NSLBP—non-specific low back pain.

Table 4. Data extracted from reviewed articles RQA parameters in the disability/disease group, where: * denotes statistically significant differences.

Study	%REC	%DET	ENT	%LAM, LMAX (Maximum Diagonal Line Length) LMEAN (Average Diagonal Line Length) TREND, TT (Trapping Time)
Schmit et al. [55]	PD/Healthy AP: 9.09 ± 4.3/5.37 ± 3.41 * EO/EC: ML: 10.11 ± 2.85/7.42 ± 3.64 *	PD/Healthy AP: 80.86 ± 28.29/52.68 ± 30.56 *	PD/Healthy AP: 3.57 ± 1.70/1.98 ± 1.24 *	LMAX: PD/Healthy AP: 2138 ± 1140/908 ± 1165 *
Negahban et al. [46]	No data	ACLD/Healthy Rigid surface, no task EO: AP: 92.06 ± 2.99/90.40 ± 5.45 ML: 93.65 ± 2.21/91.52 ± 2.60 Rigid surface, no task EC: AP: 94.63 ± 2.13/92.80 ± 2.39 ML: 95.44 ± 1.88/94.82 ± 2.32 Foam surface, no task EC: AP: 97.86 ± 0.83/97.19 ± 0.95 ML: 97.62 ± 0.67/97.30 ± 0.75	ACLD/Healthy Rigid surface, no task EO: AP: 4.72 ± 0.40/4.55 ± 0.55 ML: 4.98 ± 0.42/4.65 ± 0.34 Rigid surface, no task EC: AP: 5.22 ± 0.41/4.83 ± 0.39 ML: 5.46 ± 0.46/5.35 ± 0.50 Foam surface, no task EC: AP: 6.06 ± 0.38/5.81 ± 0.38 ML: 5.90 ± 0.26/5.86 ± 0.32	No data
Mazaheri et al. [56]	LBP/Healthy Rigid surface, no task EO: AP: 8.36 ± 2.5/8.75 ± 3.73 ML: 6.62 ± 2.33/9.09 ± 2.8 Rigid surface, no task EC: AP: 8.77 ± 2.97/9.92 ± 2.78 ML: 8.27 ± 2.49/8.21 ± 1.7 Foam surface, no task EC: AP: 7.63 ± 2.12/8.34 ± 3.52 ML: 6.60 ± 1.82/6.98 ± 2.57	LBP/Healthy Rigid surface, no task EO: AP: 96.73 ± 1.6/97.24 ± 1.79 ML: 96.19 ± 1.27/96.45 ± 1.37 Rigid surface, no task EC: AP: 97.18 ± 1.41/97.59 ± 1.66 ML: 96.49 ± 1.48/96.62 ± 1.57 Foam surface, no task EC: AP: 98.89 ± 0.40/98.84 ± 0.59 ML: 98.17 ± 0.69/98.40 ± 0.71	LBP/Healthy Rigid surface, no task EO: AP: 5.71 ± 0.52/6.04 ± 0.66 ML: 5.47 ± 0.38/5.71 ± 0.51 Rigid surface, no task EC: AP: 5.96 ± 0.57/6.91 ± 0.59 ML: 5.71 ± 0.45/5.76 ± 0.53 Foam surface, no task EC: AP: 6.33 ± 0.30/6.27 ± 0.35 ML: 6.05 ± 0.30/6.07 ± 0.32	TREND: LBP/Healthy Rigid surface, no task EO: AP: −6.13 ± 2.37/−6.77 ± 4.0 ML: −4.11 ± 2.17/−4.58 ± 2.34 Rigid surface, no task EC: AP: −5.49 ± 3.07/−6.04 ± 2.73 ML: −3.73 ± 2.69/−3.71 ± 1.74 Foam surface, no task EC: AP: −3.39 ± 1.71/−3.76 ± 2.28 ML: −2.65 ± 1.12/−3.41 ± 1.39
Negahban et al. [47]	MS/Healthy Rigid surface, single task EO: AP: 2.66 ± 1.06/2.44 ± 1.02 ML: 3.34 ± 1.32/3 ± 1.4 Rigid surface, single task EC: AP: 2.5 ± 0.8/2.67 ± 0.99 ML: 2.75 ± 1.14/2.33 ± 1.11 Foam surface, single task EC: AP: 1.91 ± 0.52/2.36 ± 0.96 ML: 2.15 ± 0.87/2.3 ± 1.1 Rigid surface, dual task EO: AP: 1.99 ± 0.54/2.42 ± 0.96 ML: 2.9 ± 1.52/2.75 ± 1.08 Rigid surface, dual task EC: AP: 2.2 ± 0.79/2.44 ± 0.94 ML: 2.73 ± 0.98/2.37 ± 1.15 Foam surface, dual task EC: AP: 1.7 ± 0.69/2.07 ± 0.98 ML: 2.13 ± 1.05/2.36 ± 1.29	MS/Healthy Rigid surface, single task EO: AP: 99.79 ± 0.18/99.83 ± 0.15 ML: 99.9 ± 0.1/99.91 ± 0.12 Rigid surface, single task EC: AP: 99.81 ± 0.12/99.78 ± 0.32 ML: 99.91 ± 0.04/99.85 ± 0.12 Foam surface, single task EC: AP: 99.76 ± 0.23/99.81 ± 0.1 ML: 99.86 ± 0.07/99.85 ± 0.13 Rigid surface, dual task EO: AP: 99.71 ± 0.22/99.76 ± 0.24 ML: 99.88 ± 0.1/99.89 ± 0.11 Rigid surface, dual task EC: AP: 99.65 ± 0.26/99.79 ± 0.23 ML: 99.88 ± 0.08/99.84 ± 0.14 Foam surface, dual task EC: AP: 99.7 ± 0.19/99.74 ± 0.17 ML: 99.83 ± 0.1/99.84 ± 0.13	MS/Healthy Rigid surface, single task EO: AP: 5.85 ± 0.46/5.75 ± 0.49 ML: 6.16 ± 0.43/6.05 ± 0.52 Rigid surface, single task EC: AP: 5.73 ± 0.45/5.72 ± 0.38 ML: 5.89 ± 0.49/5.74 ± 0.59 Foam surface, single task EC: AP: 5.48 ± 0.42/5.54 ± 0.4 ML: 5.54 ± 0.43/5.63 ± 0.52 Rigid surface, dual task EO: AP: 5.57 ± 0.3/5.65 ± 0.53 ML: 5.9 ± 0.43/5.99 ± 0.5 Rigid surface, dual task EC: AP: 5.46 ± 0.48/5.68 ± 0.53 ML: 5.83 ± 0.49/5.72 ± 0.59 Foam surface, dual task EC: AP: 5.25 ± 0.43/5.31 ± 0.42 ML: 5.52 ± 0.46/5.63 ± 0.47	TREND: MS/Healthy Rigid surface, single task EO: AP: −3.01 ± 0.99/−2.8 ± 1.18 ML: −3.72 ± 1.45/−3.49 ± 1.41 Rigid surface, single task EC: AP: −2.93 ± 0.7/−2.89 ± 1.13 ML: −2.79 ± 1.05/−2.87 ± 1.35 Foam surface, single task EC: AP: −2.22 ± 0.71/−2.31 ± 0.84 ML: −2.47 ± 0.98/−2.58 ± 1.14 Rigid surface, dual task EO: AP: −2.41 ± 0.57/−2.46 ± 0.95 ML: −3.29 ± 1.23/−3.38 ± 1.61 Rigid surface, dual task EC: AP: −2.5 ± 0.92/−2.77 ± 1.28 ML: −3.08 ± 1.34/−2.87 ± 1.53 Foam surface, dual task EC: AP: −1.89 ± 0.88/−2.05 ± 0.89 ML: −2.25 ± 1.06/−2.74 ± 1.48
Cao et al. [57]	No data
Cao et al. [58]	EDSS = 2, raw data: A. D = 1, tau = 1: 0.22 B. D = 3, tau = 16: 0.09 EDSS = 2, noisy data: C. D = 1, tau = 1: 0.22 D. D = 3, tau = 16: 0.08	No data	EDSS = 2, raw data: A. D = 1, tau = 1: 3.97 B. D = 3, tau = 16: 3.97 EDSS = 2, noisy data: C. D = 1, tau = 1: 2.87 D. D = 3, tau = 16: 2.62	LMEAN: A.22.63; B. 26.41; C. 9.85; D. 9.44 TT: EDSS = 2, raw data: A. D = 1, tau = 1: 33.48 B. D = 3, tau = 16: 35.24 TT: EDSS = 2, noisy data: C. D = 1, tau = 1: 14.34 D. D = 3, tau = 16: 13.35
Negahban et al. [59]	No data	OA/Healthy Rigid surface and EO: AP_single task: 98.06 ± 1.03/97.61 ± 1.13 ML_single task: 97.03 ± 1.71/99.64 ± 344 AP_dual task: 97.65 ± 1.14/97.06 ± 1.74 ML_dual task: 96.38 ± 4.12/93.68 ± 7.24 Rigid surface and EC: AP_single task: 97.64 ± 1.40/97.61 ± 1.44 ML_single task: 97.05 ± 2.49/94.2 ± 11.66 AP_dual task: 97.66 ± 1.46/97.21 ± 1.59 ML_dual task: 96.56 ± 2.57/94.37 ± 3.69 Foam surface and EO: AP_single task: 98.78 ± 0.96/98.60 ± 0.59 ML_single task: 98.09 ± 0.81/97.50 ± 1.14 AP_dual task: 98.67 ± 0.67/98.14 ± 0.97 ML_dual task: 97.65 ± 1.48/96.72 ± 2.13 Foam surface and EC: AP_single task: 98.98 ± 0.49/98.74 ± 0.54 ML_single task: 98.52 ± 0.65/97.96 ± 0.96 AP_dual task: 98.21 ± 2.02/98.07 ± 1.01 ML_dual task: 97.64 ± 1.70/96.78 ± 1.90	No data	No data
Shokouhyan et al. [45]	NSLBP/Healthy Without vibration: AP_rigid surface: 0.24 ± 0.1/0.32 ± 0.17 ML_rigid surface: 0.13 ± 0.08/0.22 ± 0.14 AP_foam surface: 0.1 ± 0.05/0.13 ± 0.06 ML_foam surface: 0.19 ± 0.12/0.16 ± 0.09 Ankle vibration: AP_rigid surface: 0.28 ± 0.15/0.14 ± 0.08 ML_rigid surface: 0.23 ± 0.16/0.12 ± 0.06 AP_foam surface: 0.45 ± 0.28/0.11 ± 0.07 ML_foam surface: 0.4 ± 0.25/0.14 ± 0.08 Back vibration: AP_rigid surface: 0.23 ± 0.13/0.17 ± 0.09 ML_rigid surface: 0.27 ± 0.17/0.12 ± 0.06 AP_foam surface: 0.19 ± 0.09/0.15 ± 0.09 ML_foam surface: 0.33 ± 0.17/0.21 ± 0.11 Both vibration: AP_rigid surface: 0.18 ± 0.1/0.24 ± 0.12 ML_rigid surface: 0.21 ± 0.12/0.17 ± 0.07 AP_foam surface: 0.18 ± 0.13/0.09 ± 0.05 ML_foam surface: 0.18 ± 0.1/0.15 ± 0.05	NSLBP/Healthy Without vibration: AP_rigid surface: 99.81 ± 1.56/99.84 ± 1.25 ML_rigid surface: 99.42 ± 1.56/99.65 ± 1.37 AP_foam surface: 99.1 ± 1.68/99.5 ± 1.36 ML_foam surface: 99.64 ± 1.65/99.41 ± 1.89 Ankle vibration: AP_rigid surface: 99.06 ± 1.16/98.81 ± 2.16 ML_rigid surface: 99.33 ± 1.77/98.28 ± 1.41 AP_foam surface: 99.52 ± 1.13/96.44 ± 1.48 ML_foam surface: 98.83 ± 1.33/98.53 ± 1.8 Back vibration: AP_rigid surface: 99.6 ± 1.53/99.53 ± 2.08 ML_rigid surface: 99.71 ± 1.3/98.38 ± 1.2 AP_foam surface: 99.49 ± 1.81/97.67 ± 1.11 ML_foam surface: 98.63 ± 1.59/99.74 ± 1.8 Both vibration: AP_rigid surface: 99.79 ± 1.25/99.81 ± 1.46 ML_rigid surface: 99.5 ± 1.42/99.76 ± 1.52 AP_foam surface: 99.54 ± 1.77/98.26 ± 1.67 ML_foam surface: 98.93 ± 1.14/98.63 ± 1.59	NSLBP/Healthy Without vibration: AP_rigid surface: 4.6 ± 0.44/4.85 ± 0.56 ML_rigid surface: 4.28 ± 0.44/4.2 ± 0.56 AP_foam surface: 4.07 ± 0.5/4.2 ± 0.5 ML_foam surface: 4.7 ± 0.36/4.61 ± 0.59 Ankle vibration: AP_rigid surface: 3.85 ± 0.4/4.17 ± 0.53 ML_rigid surface: 4.2 ± 0.48/4 ± 0.38 AP_foam surface: 4.69 ± 0.57/3.9 ± 0.42 ML_foam surface: 4.03 ± 0.38/4.11 ± 0.54 Back vibration: AP_rigid surface: 4.31 ± 0.51/4.47 ± 0.41 ML_rigid surface: 4.05 ± 0.51/3.9 ± 0.36 AP_foam surface: 4.39 ± 0.44/3.9 ± 0.47 ML_foam surface: 4.63 ± 0.48/4.25 ± 0.5 Both vibration: AP_rigid surface: 4.46 ± 0.5/4.13 ± 0.44 ML_rigid surface: 4.57 ± 0.46/4.13 ± 0.55 AP_foam surface: 4.49 ± 0.52/4 ± 0.43 ML_foam surface: 3.95 ± 0.34/4.26 ± 0.42	TREND: NSLBP/Healthy Without vibration: AP_rigid surface: −0.42 ± 0.17/−0.62 ± 0.23 ML_rigid surface: −0.25 ± 0.08/−0.44 ± 0.2 AP_foam surface: −0.18 ± 0.07/−0.26 ± 0.1 ML_foam surface: −0.38 ± 0.18/−0.31 ± 0.12 Ankle vibration: AP_rigid surface: −0.54 ± 0.18/−0.28 ± 0.1 ML_rigid surface: −0.46 ± 0.21/−0.23 ± 0.08 AP_foam surface: −0.89 ± 0.37/−0.2 ± 0.07 ML_foam surface: −0.76 ± 0.32/−0.26 ± 0.1 Back vibration: AP_rigid surface: −0.46 ± 0.14/−0.33 ± 0.13 ML_rigid surface: −0.53 ± 0.17/−0.23 ± 0.07 AP_foam surface: −0.35 ± 0.13/−0.3 ± 0.09 ML_foam surface: −0.66 ± 0.25/−0.41 ± 0.15 Both vibration: AP_rigid surface: −0.35 ± 0.13/−0.48 ± 0.18 ML_rigid surface: −0.43 ± 0.16/−0.33 ± 0.13 AP_foam surface: −0.35 ± 0.09/−0.17 ± 0.07 ML_foam surface: −0.34 ± 0.08/−0.3 ± 0.12

Abbreviations: AP—anterior-posterior, ML—medial-lateral direction, EO—eyes open, EC—eyes closed, LBP—people with nonspecific low back pain, ACLD—patients with anterior cruciate ligament deficiency, MS—patients with multiple sclerosis, PD—Parkinson’s patients, OA—patients with knee osteoarthritis, EDSS—Expanded Disability Status Scale, NSLBP—non-specific low back pain.

3.3. Sport/Influence of Motion/Impact of Visual Stimuli

The highest number—twelve papers—were in the sport/influence of motion/impact of visual stimuli group. A total of 661 participants were analyzed. Their mean age ranged from 20 [18,49] to 79 ± 2.8 years [61]. Their mean body weight and height values ranged from 66.4 ± 12.2 kg [63] to 78 ± 15 kg [20] and from 160.6 ± 5.35 cm [64] to 176 ± 7 cm [61], respectively. Six types of force platforms measured CoP fluctuations with different sampling rates. In two papers, authors used Techno Concept platforms (40 Hz) [43,62]. Three studies [63,64,67] used Medicapteur force plates (40 Hz). Moreover, in three works [48,65,68], Bertec platforms (1000 Hz) were used. In a single study, data were sampled at 2000 Hz using a Kistler platform [20] and at 1000 Hz [61] using an AMTI force platform. The remaining papers sampled data at 100 Hz and used Smart Balance [49] and Bertec [18] force plates.

In this group, a very large discrepancy in measurement times was noted. The traditional 30 s trials with eyes open and closed were performed in one study [18]. Markovic et al. [61] performed the same long trial, however, only with eyes open and while standing in semi-tandem. Chowdhury et al. [68] took into account only standing on two legs with eyes open. This group was dominated by tests lasting 51.2 s, which can be found in 5 papers [43,62,63,64,67]. van den Hoorn et al. [20] performed tests lasting 135 s; Apthorp et al. [48] and Palmisano et al. [65]—60 s. In Clark and Riley’s [49] paper, there were trials lasting only 20 s. Only in five papers [20,63,64,65,67], the authors indicated what tool they used to calculate RQA measures. The Cross Recurrence Plot Toolbox by Marwan et al. [6] was used four times. The Recurrence Quantification Toolbox for MATLAB of Li et al. [66] was used by Palmisano et al. [65].

To perform RQA analysis, time delay, embedding dimension, and threshold were used. Bernard et al. [67] and Markovic et al. [61] did not provide their values. In five works [48,63,64,65,68], the embedding dimension was 8. In the remaining works, it was 5 [20], 9 [49], 10 [18], 11 [18], or 12 [43,62]. It is worth noting that there were significant discrepancies in both the time lag values and the recurrence thresholds. The time delay ranged from tau = 50 ms (corresponding to 2 samples) [43,62] to tau = 180 ms [20]. The radius for recurrence was 2–3% [43,62], 11% [49], 20% [18], 24% [18], and 25% [64] of the mean distance between data points, among others.

Table 5 includes RQA parameter values specific for the sport/influence of motion/impact of visual stimuli group. In four papers [48,49,65,68], the authors did not provide RQA parameter values. In four [20,63,64,67] of the remaining eight papers, the authors did not provide %REC values. Schmit et al. [18] showed that %REC was higher for balance tests in runners relative to the dancer’s group. In both groups, %REC was higher for trials with visual control included for AP direction. Ferrufino et al. [43] showed that %REC for open-eye trials was higher in the dancer group relative to the fall prevention group, not dependent on the direction or recurrence radius. Markovic et al. [61] showed that %REC was higher in the elderly group relative to the young, mainly for the AP direction when standing in semi-tandem. Coubard et al. [62] showed that dancers in most trials had higher %REC values relative to those recorded for the control group.

%DET was in each of the eight papers. Schmit et al. [18] showed that %DET values were higher for standing with eyes closed in both study groups. However, the result of Decker et al. [63] was the opposite. In Ferrufino et al.’s [43] paper, %DET maintained the trend set by %REC. Markovic et al. [61] showed that %DET had higher values in the young group. van den Hoorn et al. [20] showed that %DET was lowest in the young group and highest in the group of non-fallers.

ENT values were mentioned in only three papers [18,61,64]. Schmit et al. [18] and Markovic et al. [61] maintained the trend indicated by %DET. Bernard et al. [64] showed that ENT values were higher toward ML in each measurement.

The last column in Table 5 indicates high variability. Three authors [18,43,62] analyzed LMAX values, but only Schmit et al. [18] reported additional—TREND values. %LAM, LMEAN, and TT were considered in two papers [20,61].

In Schmit et al.’s [18] work, LMAX and TREND values supported the trend set by ENT and %DET. The situation was similar in Ferrufino et al.’s [43] work, where LMAX values were higher in dancers compared to the faller prevention group. Markovic et al. [61] showed that all parameters (%LAM, %LMEAN, and TT) were higher when standing in semi-tandem in the young group. The same relationships were found in the work of van den Hoorn et al. [20].

Table 5. Data extracted from reviewed articles RQA parameters in the sport/influence of motion/impact of visual stimuli group.

Study	%REC	%DET	ENT	%LAM, LMAX (Maximum Diagonal Line Length) LMEAN (Average Diagonal Line Length) TREND, TT (Trapping Time)
Schmit et al. [18]	Dancers/runners AP: 5.38 < 7.01 ML: 1.62 < 2.46 In both groups: EO > EC AP: 6.70/5.68 Foam > rigid AP: 7.38/5	In both groups: EC > EO AP: 76.15/62.30 ML: 67.19/64.63 Foam > rigid AP: 84.76/44.75 ML: 83.05/57.73	Dancers/runners ML: 3.20/3.72 In both groups: EC > EO AP: 3.68/3.25 ML: 3.80/3.13 Foam > rigid AP: 4.44/2.49 ML: 4.01/2.92	LMAX Dancers/runners AP: 1480.21 < 1909.01 ML: 1730.98 < 2217.21 In both groups: EC > EO AP: 1807.80/1581.42 ML: 2254.07/1694.12 Foam > rigid AP: 2709.87/779.35 ML: 2562.26/1385.91 TREND Dancers/runners AP: −3.24 > −5.12 ML: −1.44 > −2.25
Clark and Riley [49]	No data
Ferrufino et al. [43]	CD/FP AP_EO radius 2%: Pre-test: 0.690 ± 0.218/0.308 ± 0.054 Post-test: 0.538 ± 0.124/0.389 ± 0.056 ML_EO radius 2%: Pre-test: 1.08 ± 0.227/0.269 ± 0.040 Post-test: 0.589 ± 0.122/0.425 ± 0.089 AP_EO radius 3%: Pre-test: 1.72 ± 0.483/0.941 ± 0.128 Post-test: 1.43 ± 0.284/1.16 ± 0.130 ML_EO radius 3%: Pre-test: 2.71 ± 0.463/0.955 ± 0.104 Post-test: 1.62 ± 0.253/1.34 ± 0.201 AP_EC radius 2%: Pre-test: 0.420 ± 0.120/0.332 ± 0.069 Post-test: 0.353 ± 0.080/0.389 ± 0.076 ML_EC radius 2%: Pre-test: 0.392 ± 0.116/0.418 ± 0.070 Post-test: 0.658 ± 0.183/0.253 ± 0.063 AP_EC radius 3% Pre-test: 1.22 ± 0.268/1.01 ± 0.170 Post-test: 1.02 ± 0.181/1.15 ± 0.177 ML_EC radius 3% Pre-test: 1.16 ± 0.257/1.39 ± 0.187 Post-test: 1.80 ± 0.415/0.880 ± 0.152	CD/FP AP_EO radius 2%: Pre-test: 73.2 ± 2.64/72.1 ± 1.12 Post-test: 75.1 ± 2.06/72.8 ± 1.38 ML_EO radius 2%: Pre-test: 78.1 ± 2.18/65.5 ± 2.49 Post-test: 74.1 ± 2.35/70.2 ± 2.14 AP_EO radius 3%: Pre-test: 77.4 ± 2.41/75.5 ± 1.13 Post-test: 78.9 ± 1.89/77.3 ± 1.02 ML_EO radius 3%: Pre-test: 83.5 ± 1.76/74.5 ± 1.52 Post-test: 80.4 ± 1.69/77.5 ± 1.57 AP_EC radius 2%: Pre-test: 73.9 ± 1.65/72.8 ± 1.12 Post-test: 70.1 ± 3.34/73.6 ± 1.47 ML_EC radius 2%: Pre-test: 69.0 ± 3.25/69.8 ± 2.64 Post-test: 74.1 ± 2.58/68.6 ± 1.65 AP_EC radius 3% Pre-test: 76.7 ± 2.13/76.6 ± 1.12 Post-test: 76.2 ± 2.13/77.3 ± 1.27 ML_EC radius 3% Pre-test: 75.8 ± 2.59/77.4 ± 1.57 Post-test: 79.5 ± 2.20/75.2 ± 1.28		LMAX CD/FP AP_EO radius 2%: Pre-test: 23.7 ± 5.90/12.0 ± 1.85 Post-test: 22.0 ± 4.98/15.6 ± 2.32 ML_EO radius 2%: Pre-test: 34.4 ± 5.39/9.88 ± 1.77 Post-test: 23.1 ± 4.67/15.0 ± 2.89 AP_EO radius 3%: Pre-test: 45.1 ± 8.45/33.7 ± 4.35 Post-test: 49.5 ± 7.01/36.1 ± 4.10 ML_EO radius 3%: Pre-test: 66.0 ± 6.96/32.8 ± 4.00 Post-test: 55.3 ± 8.01/39.8 ± 4.61 AP_EC radius 2%: Pre-test: 15.7 ± 3.70/12.5 ± 2.39 Post-test: 14.5 ± 3.75/18.2 ± 3.31 ML_EC radius 2%: Pre-test: 16.8 ± 6.19/12.9 ± 2.45 Post-test: 19.6 ± 4.64/9.45 ± 2.05 AP_EC radius 3% Pre-test: 32.8 ± 6.43/28.3 ± 4.27 Post-test: 29.6 ± 4.49/35.5 ± 4.33 ML_EC radius 3% Pre-test: 33.4 ± 7.46/41.1 ± 5.59 Post-test: 40.9 ± 7.08/26.3 ± 4.02
Markovic et al. [61]	Semi-tandem quiet stance young/older: AP: 0.08 ± 0.02/0.09 ± 0.11 ML: 0.10 ± 0.09/0.10 ± 0.07	Semi-tandem quiet stance young/older: AP: 0.92 ± 0.02/0.90 ± 0.04 ML: 0.95 ± 0.02/0.93 ± 0.03	Semi-tandem quiet stance Young/Older: AP: 2.16 ± 0.16/2.03 ± 0.17 ML: 2.33 ± 0.15/2.26 ± 0.17	%LAM Semi-tandem quiet stance young/older: AP: 0.65 ± 0.09/0.56 ± 0.12 ML: 0.71 ± 0.07/0.65 ± 0.10 %LMEAN: AP: 4.97 ± 0.57/4.57 ± 0.68 ML: 5.66 ± 0.64/5.43 ± 0.79 TT Semi-tandem quiet stance young/older: AP: 3.83 ± 0.61/3.18 ± 1.02 ML: 3.99 ± 0.69/3.49 ± 0.65
Apthorp et al. [48]	No data
Coubard et al. [62]	CD/C AP_EO radius 2%: Pre-test: 0.428 ± 0.095/0.480 ± 0.117 Post-test: 0.408 ± 0.88/0.248 ± 0.057 ML_EO radius 2%: Pre-test: 0.428 ± 0.095/0.480 ± 0.117 Post-test: 0.607 ± 0.081/0.524 ± 0.104 AP_EO radius 3%: Pre-test: 1.12 ± 0.14/1.09 ± 0.21 Post-test: 1.14 ± 0.19/0.79 ± 0.14 ML_EO radius 3%: Pre-test: 1.28 ± 0.20/1.44 ± 0.26 Post-test: 1.70 ± 0.18/1.47 ± 0.23 AP_EC radius 2%: Pre-test: 0.389 ± 0.085/0.263 ± 0.043 Post-test: 0.381 ± 0.073/0.311 ± 0.062 ML_EC radius 2%: Pre-test: 0.464 ± 0.091/0.617 ± 0.192 Post-test: 0.734 ± 0.159/0.838 ± 0.206 AP_EC radius 3% Pre-test: 1.13 ± 0.21/0.82 ± 0.11 Post-test: 1.10 ± 0.18/0.95 ± 0.16 ML_EC radius 3% Pre-test: 1.36 ± 0.20/1.77 ± 0.47 Post-test: 1.89 ± 0.32/2.22 ± 0.45	CD/C AP_EO radius 2%: Pre-test: 74.5 ± 1.52/72.7 ± 1.18 Post-test: 74.5 ± 1.30/71.8 ± 0.96 ML_EO radius 2%: Pre-test: 73.8 ± 1.82 74.9/± 1.40 Post-test: 77.2 ± 1.40/75.2 ± 2.10 AP_EO radius 3%: Pre-test: 78.0 ± 1.42/76.4 ± 1.32 Post-test: 78.0 ± 1.29/75.0 ± 0.99 ML_EO radius 3%: Pre-test: 79.5 ± 1.39/80.2 ± 1.28 Post-test: 82.1 ± 1.23/80.3 ± 1.47 AP_EC radius 2%: Pre-test: 73.1 ± 1.51/72.9 ± 1.14 Post-test: 74.5 ± 1.12/71.1 ± 1.76 ML_EC radius 2%: Pre-test: 73.6 ± 2.42/74.7 ± 1.72 Post-test: 76.7 ± 2.19/76.9 ± 1.51 AP_EC radius 3% Pre-test: 75.8 ± 1.83/75.8 ± 1.12 Post-test: 77.6 ± 1.24/75.0 ± 1.41 ML_EC radius 3% Pre-test: 79.2 ± 1.66/76.9 ± 1.52 Post-test: 81.8 ± 1.64/81.7 ± 1.34		LMAX: CD/C AP_EO radius 2%: Pre-test: 17.7 ± 2.61/18.7 ± 4.88 Post-test: 16.6 ± 2.50/8.34 ± 1.56 ML_EO radius 2%: Pre-test: 15.0 ± 3.11/20.1 ± 6.28 Post-test: 23.4 ± 3.52/20.9 ± 5.47 AP_EO radius 3%: Pre-test: 40.8 ± 5.40/38.2 ± 6.61 Post-test: 36.2 ± 5.49/23.4 ± 3.40 ML_EO radius 3%: Pre-test: 48.0 ± 7.09/49.2 ± 7.51 Post-test: 54.5 ± 6.83/49.8 ± 7.63 AP_EC radius 2%: Pre-test: 17.6 ± 3.40/11.4 ± 2.35 Post-test: 18.3 ± 3.78/14.4 ± 3.46 ML_EC radius 2%: Pre-test: 18.0 ± 4.31/18.0 ± 4.20 Post-test: 30.6 ± 5.83/25.9 ± 5.45 AP_EC radius 3% Pre-test: 38.2 ± 5.69/31.9 ± 4.51 Post-test: 40.9 ± 6.53/35.2 ± 6.60 ML_EC radius 3% Pre-test: 43.5 ± 6.88/45.4 ± 6.71 Post-test: 60.6 ± 8.51/55.3 ± 5.41
Decker et al. [63]		6MWD EO/EC AP: 0.93 ± 0.09/0.88 ± 0.15 ML: 0.96 ± 0.04/0.94 ± 0.07
Bernard et al. [64]		EC_Beginning/6 MWD_1/6MWD_2 AP: 0.86 ± 0.18/0.86 ± 0.19/0.89 ± 0.12 ML: 0.92 ± 0.09/0.93 ± 0.08/0.95 ± 0.04	EC_Beginning/6MWD_1/6MWD_2 AP: 3.38 ± 0.71/3.38 ± 0.71/3.49 ± 0.58 ML: 3.65 ± 0.57/3.72 ± 0.53/3.86 ± 0.44
Palmisano et al. [65]	No data
Bernard et al. [67]		Trained group Beginning/after 6 months: EO_AP: 0.93 ± 0.07/0.91 ± 0.08 EO_ML: 0.96 ± 0.04/0.96 ± 0.03 EC_AP: 0.88 ± 0.14/0.86 ± 0.14 EC_ML: 0.95 ± 0.06/0.95 ± 0.05 Control group Beginning/after 6 months: EO_AP: 0.94 ± 0.07/0.94 ± 0.05 EO_ML: 0.97 ± 0.02/0.97 ± 0.03 EC_AP: 0.9 ± 0.12/0.86 ± 0.16 EC_ML: 0.95 ± 0.04/0.94 ± 0.08
van den Hoorn et al. [20]		Young/fallers/non-fallers: 0.91 ± 0.09/0.86 ± 0.15/0.88 ± 0.11		Young/fallers/non-fallers: %LAM: 0.95 ± 0.06/0.88 ± 0.17/0.91 ± 0.12 LMEAN: 39.49 ± 16.01/30.28 ± 12.02/32.64 ± 14.03 Young/fallers/non-fallers: TT: 35.91 ± 17.57/24.88 ± 14.05/27.55 ± 16.68
Chowdhury et al. [68]	No data

Abbreviations: AP—anterior-posterior, ML—medial-lateral direction, EO—eyes open, EC—eyes closed, CD—contemporary dance group, C—controls, FP—fall prevention group, 6MWD—Six-Minute Walking Distance test.

Table 5. Data extracted from reviewed articles RQA parameters in the sport/influence of motion/impact of visual stimuli group.

Study	%REC	%DET	ENT	%LAM, LMAX (Maximum Diagonal Line Length) LMEAN (Average Diagonal Line Length) TREND, TT (Trapping Time)
Schmit et al. [18]	Dancers/runners AP: 5.38 < 7.01 ML: 1.62 < 2.46 In both groups: EO > EC AP: 6.70/5.68 Foam > rigid AP: 7.38/5	In both groups: EC > EO AP: 76.15/62.30 ML: 67.19/64.63 Foam > rigid AP: 84.76/44.75 ML: 83.05/57.73	Dancers/runners ML: 3.20/3.72 In both groups: EC > EO AP: 3.68/3.25 ML: 3.80/3.13 Foam > rigid AP: 4.44/2.49 ML: 4.01/2.92	LMAX Dancers/runners AP: 1480.21 < 1909.01 ML: 1730.98 < 2217.21 In both groups: EC > EO AP: 1807.80/1581.42 ML: 2254.07/1694.12 Foam > rigid AP: 2709.87/779.35 ML: 2562.26/1385.91 TREND Dancers/runners AP: −3.24 > −5.12 ML: −1.44 > −2.25
Clark and Riley [49]	No data
Ferrufino et al. [43]	CD/FP AP_EO radius 2%: Pre-test: 0.690 ± 0.218/0.308 ± 0.054 Post-test: 0.538 ± 0.124/0.389 ± 0.056 ML_EO radius 2%: Pre-test: 1.08 ± 0.227/0.269 ± 0.040 Post-test: 0.589 ± 0.122/0.425 ± 0.089 AP_EO radius 3%: Pre-test: 1.72 ± 0.483/0.941 ± 0.128 Post-test: 1.43 ± 0.284/1.16 ± 0.130 ML_EO radius 3%: Pre-test: 2.71 ± 0.463/0.955 ± 0.104 Post-test: 1.62 ± 0.253/1.34 ± 0.201 AP_EC radius 2%: Pre-test: 0.420 ± 0.120/0.332 ± 0.069 Post-test: 0.353 ± 0.080/0.389 ± 0.076 ML_EC radius 2%: Pre-test: 0.392 ± 0.116/0.418 ± 0.070 Post-test: 0.658 ± 0.183/0.253 ± 0.063 AP_EC radius 3% Pre-test: 1.22 ± 0.268/1.01 ± 0.170 Post-test: 1.02 ± 0.181/1.15 ± 0.177 ML_EC radius 3% Pre-test: 1.16 ± 0.257/1.39 ± 0.187 Post-test: 1.80 ± 0.415/0.880 ± 0.152	CD/FP AP_EO radius 2%: Pre-test: 73.2 ± 2.64/72.1 ± 1.12 Post-test: 75.1 ± 2.06/72.8 ± 1.38 ML_EO radius 2%: Pre-test: 78.1 ± 2.18/65.5 ± 2.49 Post-test: 74.1 ± 2.35/70.2 ± 2.14 AP_EO radius 3%: Pre-test: 77.4 ± 2.41/75.5 ± 1.13 Post-test: 78.9 ± 1.89/77.3 ± 1.02 ML_EO radius 3%: Pre-test: 83.5 ± 1.76/74.5 ± 1.52 Post-test: 80.4 ± 1.69/77.5 ± 1.57 AP_EC radius 2%: Pre-test: 73.9 ± 1.65/72.8 ± 1.12 Post-test: 70.1 ± 3.34/73.6 ± 1.47 ML_EC radius 2%: Pre-test: 69.0 ± 3.25/69.8 ± 2.64 Post-test: 74.1 ± 2.58/68.6 ± 1.65 AP_EC radius 3% Pre-test: 76.7 ± 2.13/76.6 ± 1.12 Post-test: 76.2 ± 2.13/77.3 ± 1.27 ML_EC radius 3% Pre-test: 75.8 ± 2.59/77.4 ± 1.57 Post-test: 79.5 ± 2.20/75.2 ± 1.28		LMAX CD/FP AP_EO radius 2%: Pre-test: 23.7 ± 5.90/12.0 ± 1.85 Post-test: 22.0 ± 4.98/15.6 ± 2.32 ML_EO radius 2%: Pre-test: 34.4 ± 5.39/9.88 ± 1.77 Post-test: 23.1 ± 4.67/15.0 ± 2.89 AP_EO radius 3%: Pre-test: 45.1 ± 8.45/33.7 ± 4.35 Post-test: 49.5 ± 7.01/36.1 ± 4.10 ML_EO radius 3%: Pre-test: 66.0 ± 6.96/32.8 ± 4.00 Post-test: 55.3 ± 8.01/39.8 ± 4.61 AP_EC radius 2%: Pre-test: 15.7 ± 3.70/12.5 ± 2.39 Post-test: 14.5 ± 3.75/18.2 ± 3.31 ML_EC radius 2%: Pre-test: 16.8 ± 6.19/12.9 ± 2.45 Post-test: 19.6 ± 4.64/9.45 ± 2.05 AP_EC radius 3% Pre-test: 32.8 ± 6.43/28.3 ± 4.27 Post-test: 29.6 ± 4.49/35.5 ± 4.33 ML_EC radius 3% Pre-test: 33.4 ± 7.46/41.1 ± 5.59 Post-test: 40.9 ± 7.08/26.3 ± 4.02
Markovic et al. [61]	Semi-tandem quiet stance young/older: AP: 0.08 ± 0.02/0.09 ± 0.11 ML: 0.10 ± 0.09/0.10 ± 0.07	Semi-tandem quiet stance young/older: AP: 0.92 ± 0.02/0.90 ± 0.04 ML: 0.95 ± 0.02/0.93 ± 0.03	Semi-tandem quiet stance Young/Older: AP: 2.16 ± 0.16/2.03 ± 0.17 ML: 2.33 ± 0.15/2.26 ± 0.17	%LAM Semi-tandem quiet stance young/older: AP: 0.65 ± 0.09/0.56 ± 0.12 ML: 0.71 ± 0.07/0.65 ± 0.10 %LMEAN: AP: 4.97 ± 0.57/4.57 ± 0.68 ML: 5.66 ± 0.64/5.43 ± 0.79 TT Semi-tandem quiet stance young/older: AP: 3.83 ± 0.61/3.18 ± 1.02 ML: 3.99 ± 0.69/3.49 ± 0.65
Apthorp et al. [48]	No data
Coubard et al. [62]	CD/C AP_EO radius 2%: Pre-test: 0.428 ± 0.095/0.480 ± 0.117 Post-test: 0.408 ± 0.88/0.248 ± 0.057 ML_EO radius 2%: Pre-test: 0.428 ± 0.095/0.480 ± 0.117 Post-test: 0.607 ± 0.081/0.524 ± 0.104 AP_EO radius 3%: Pre-test: 1.12 ± 0.14/1.09 ± 0.21 Post-test: 1.14 ± 0.19/0.79 ± 0.14 ML_EO radius 3%: Pre-test: 1.28 ± 0.20/1.44 ± 0.26 Post-test: 1.70 ± 0.18/1.47 ± 0.23 AP_EC radius 2%: Pre-test: 0.389 ± 0.085/0.263 ± 0.043 Post-test: 0.381 ± 0.073/0.311 ± 0.062 ML_EC radius 2%: Pre-test: 0.464 ± 0.091/0.617 ± 0.192 Post-test: 0.734 ± 0.159/0.838 ± 0.206 AP_EC radius 3% Pre-test: 1.13 ± 0.21/0.82 ± 0.11 Post-test: 1.10 ± 0.18/0.95 ± 0.16 ML_EC radius 3% Pre-test: 1.36 ± 0.20/1.77 ± 0.47 Post-test: 1.89 ± 0.32/2.22 ± 0.45	CD/C AP_EO radius 2%: Pre-test: 74.5 ± 1.52/72.7 ± 1.18 Post-test: 74.5 ± 1.30/71.8 ± 0.96 ML_EO radius 2%: Pre-test: 73.8 ± 1.82 74.9/± 1.40 Post-test: 77.2 ± 1.40/75.2 ± 2.10 AP_EO radius 3%: Pre-test: 78.0 ± 1.42/76.4 ± 1.32 Post-test: 78.0 ± 1.29/75.0 ± 0.99 ML_EO radius 3%: Pre-test: 79.5 ± 1.39/80.2 ± 1.28 Post-test: 82.1 ± 1.23/80.3 ± 1.47 AP_EC radius 2%: Pre-test: 73.1 ± 1.51/72.9 ± 1.14 Post-test: 74.5 ± 1.12/71.1 ± 1.76 ML_EC radius 2%: Pre-test: 73.6 ± 2.42/74.7 ± 1.72 Post-test: 76.7 ± 2.19/76.9 ± 1.51 AP_EC radius 3% Pre-test: 75.8 ± 1.83/75.8 ± 1.12 Post-test: 77.6 ± 1.24/75.0 ± 1.41 ML_EC radius 3% Pre-test: 79.2 ± 1.66/76.9 ± 1.52 Post-test: 81.8 ± 1.64/81.7 ± 1.34		LMAX: CD/C AP_EO radius 2%: Pre-test: 17.7 ± 2.61/18.7 ± 4.88 Post-test: 16.6 ± 2.50/8.34 ± 1.56 ML_EO radius 2%: Pre-test: 15.0 ± 3.11/20.1 ± 6.28 Post-test: 23.4 ± 3.52/20.9 ± 5.47 AP_EO radius 3%: Pre-test: 40.8 ± 5.40/38.2 ± 6.61 Post-test: 36.2 ± 5.49/23.4 ± 3.40 ML_EO radius 3%: Pre-test: 48.0 ± 7.09/49.2 ± 7.51 Post-test: 54.5 ± 6.83/49.8 ± 7.63 AP_EC radius 2%: Pre-test: 17.6 ± 3.40/11.4 ± 2.35 Post-test: 18.3 ± 3.78/14.4 ± 3.46 ML_EC radius 2%: Pre-test: 18.0 ± 4.31/18.0 ± 4.20 Post-test: 30.6 ± 5.83/25.9 ± 5.45 AP_EC radius 3% Pre-test: 38.2 ± 5.69/31.9 ± 4.51 Post-test: 40.9 ± 6.53/35.2 ± 6.60 ML_EC radius 3% Pre-test: 43.5 ± 6.88/45.4 ± 6.71 Post-test: 60.6 ± 8.51/55.3 ± 5.41
Decker et al. [63]		6MWD EO/EC AP: 0.93 ± 0.09/0.88 ± 0.15 ML: 0.96 ± 0.04/0.94 ± 0.07
Bernard et al. [64]		EC_Beginning/6 MWD_1/6MWD_2 AP: 0.86 ± 0.18/0.86 ± 0.19/0.89 ± 0.12 ML: 0.92 ± 0.09/0.93 ± 0.08/0.95 ± 0.04	EC_Beginning/6MWD_1/6MWD_2 AP: 3.38 ± 0.71/3.38 ± 0.71/3.49 ± 0.58 ML: 3.65 ± 0.57/3.72 ± 0.53/3.86 ± 0.44
Palmisano et al. [65]	No data
Bernard et al. [67]		Trained group Beginning/after 6 months: EO_AP: 0.93 ± 0.07/0.91 ± 0.08 EO_ML: 0.96 ± 0.04/0.96 ± 0.03 EC_AP: 0.88 ± 0.14/0.86 ± 0.14 EC_ML: 0.95 ± 0.06/0.95 ± 0.05 Control group Beginning/after 6 months: EO_AP: 0.94 ± 0.07/0.94 ± 0.05 EO_ML: 0.97 ± 0.02/0.97 ± 0.03 EC_AP: 0.9 ± 0.12/0.86 ± 0.16 EC_ML: 0.95 ± 0.04/0.94 ± 0.08
van den Hoorn et al. [20]		Young/fallers/non-fallers: 0.91 ± 0.09/0.86 ± 0.15/0.88 ± 0.11		Young/fallers/non-fallers: %LAM: 0.95 ± 0.06/0.88 ± 0.17/0.91 ± 0.12 LMEAN: 39.49 ± 16.01/30.28 ± 12.02/32.64 ± 14.03 Young/fallers/non-fallers: TT: 35.91 ± 17.57/24.88 ± 14.05/27.55 ± 16.68
Chowdhury et al. [68]	No data

Abbreviations: AP—anterior-posterior, ML—medial-lateral direction, EO—eyes open, EC—eyes closed, CD—contemporary dance group, C—controls, FP—fall prevention group, 6MWD—Six-Minute Walking Distance test.

3.4. Extreme Values for RQA Parameters

Table 6 shows the minimum and maximum values of the parameters derived from the RQA analysis, with a link to the group of papers from which the value came. It is noteworthy that seven extreme values came from group number two (disability/disease), four from the first group (children/young/elderly), and three from the third group (sport/influence of motion/impact of visual stimuli). The DIV parameter did not appear at all.

Table 6. The minimal and maximal values of RQA parameters obtained in the reviewed papers, where 1st group—children/young/elderly; 2nd group—disability/disease; 3rd group—sport/influence of motion/impact of visual stimuli.

RQA Parameters	Minimal Values	Maximal Values
Recurrence (%REC)	0.08 ± 0.02 Markovic et al. [61] (3rd group)	10.11 ± 2.85 Schmit et al. [55] (2nd group)
Determinism (%DET)	65.5 ± 2.49 Ferrufino et al. [43] (3rd group)	99.9 ± 0.1 Negahban et al. [47] (2nd group)
Entropy (ENT)	1.3 King et al. [51] (1st group)	6.91 ± 0.59 Mazaheri et al. [56] (2nd group)
Trend of recurrences (TREND)	−6.77 ± 4.0 Mazaheri et al. [56] (2nd group)	−0.17 ± 0.07 Shokouhyan et al. [45] (2nd group)
Laminarity (%LAM)	0.56 ± 0.12 Markovic et al. [61] (3rd group)	95.7 ± 2.00 van den Hoorn et al. [17] (1st group)
Trapping time (TT)	0.298 ± 0.104 van den Hoorn et al. [17] (1st group)	3.99 ± 0.69 Markovic et al. [61] (2nd group)
Length of the longest diagonal line (LMAX)	1.888 ± 0.650 van den Hoorn et al. [17] (1st group)	66.0 ± 6.96 Ferrufino et al. [43] (2nd group)
Length of the mean diagonal line (LMEAN)	9.44 van den Hoorn et al. [20] (3rd group)	39.49 ± 16.01 van den Hoorn et al. [20] (3rd group)
Divergence (DIV)	Did not appear in publications.	Did not appear in publications.

Table 6. The minimal and maximal values of RQA parameters obtained in the reviewed papers, where 1st group—children/young/elderly; 2nd group—disability/disease; 3rd group—sport/influence of motion/impact of visual stimuli.

RQA Parameters	Minimal Values	Maximal Values
Recurrence (%REC)	0.08 ± 0.02 Markovic et al. [61] (3rd group)	10.11 ± 2.85 Schmit et al. [55] (2nd group)
Determinism (%DET)	65.5 ± 2.49 Ferrufino et al. [43] (3rd group)	99.9 ± 0.1 Negahban et al. [47] (2nd group)
Entropy (ENT)	1.3 King et al. [51] (1st group)	6.91 ± 0.59 Mazaheri et al. [56] (2nd group)
Trend of recurrences (TREND)	−6.77 ± 4.0 Mazaheri et al. [56] (2nd group)	−0.17 ± 0.07 Shokouhyan et al. [45] (2nd group)
Laminarity (%LAM)	0.56 ± 0.12 Markovic et al. [61] (3rd group)	95.7 ± 2.00 van den Hoorn et al. [17] (1st group)
Trapping time (TT)	0.298 ± 0.104 van den Hoorn et al. [17] (1st group)	3.99 ± 0.69 Markovic et al. [61] (2nd group)
Length of the longest diagonal line (LMAX)	1.888 ± 0.650 van den Hoorn et al. [17] (1st group)	66.0 ± 6.96 Ferrufino et al. [43] (2nd group)
Length of the mean diagonal line (LMEAN)	9.44 van den Hoorn et al. [20] (3rd group)	39.49 ± 16.01 van den Hoorn et al. [20] (3rd group)
Divergence (DIV)	Did not appear in publications.	Did not appear in publications.

The lowest recurrence value appeared in the study of Markovic et al. [61] in the AP direction during a 30 s quiet stance in a semi-tandem for young subjects. The highest value was recorded by Schmit et al. [55] in the ML direction during 30 s of both legs standing with eyes open.

The lowest %DET value was found in the paper of Ferrufino et al. [43] in the ML direction during 51.2 s of both legs standing with eyes open for the group called fall prevention before training intervention. The highest %DET was in the ML direction during 30 s of both legs standing on a rigid surface with eyes open [47].

The ENT value was the lowest in the ML direction during both legs standing (side-by-side) trials with eyes open lasting 60 s [51]. The highest ENT value was noted by Mazaheri et al. [56] during both leg standing trials lasting 30 s with eyes closed in the AP direction for healthy persons.

TREND achieved all the time negative values. In both cases, the extreme values came from the disability/disease group. Mazaheri et al. [56] noted the lowest TREND value during 30 s of both legs standing with eyes open in the healthy group. Shokouhyan et al. [45] found the highest TREND value during 30 s of both legs standing on a foam surface with vibrations of the ankle and back muscles.

The lowest %LAM was reported by Markovic et al. [61] in the direction of AP during a 30 s semi-tandem quiet stance with eyes open in a group of elderly. The highest %LAM was during 30 s of both legs standing on a foam surface with eyes open for the first repeated trial in the group of elderly individuals [17]. In the same study, during standing with both legs on a foam surface with eyes closed, TT and LMAX values appeared the lowest. The highest TT was 3.99 ± 0.69 in the previously discussed paper [61] during a semi-tandem quiet stance for young subjects. Ferrufino et al. [43] showed the highest LMAX value in the ML direction before the dance training intervention.

van den Hoorn et al. [20] found maximal LMEAN values in the young person’s group during 135 s of both legs standing before vibration training. Cao et al. [58] showed the lowest value of LMEAN for the noisy data of one participant.

4. Discussion

This review aimed to summarize and update information on the current published research explicitly linked to the application of recurrence quantification analysis (RQA) to assess postural stability during upright standing. Following the research on biological signals in various scientific fields, it is remarkable that over the past few years, the trend to analyze the postural stability system using nonlinear measures and, consequently, the CoP signal has emerged [3]. The RQA method provides information on the temporal structure of postural sway, including their level of predictability, randomness, or potential chaos. This review analyzed 28 studies published in the last 23 years, which fell into three thematic groups: the first—children/young/elderly (8 papers); the second—disability/disease (8 studies); and the third—sports/impact of motion/impact of visual stimuli (12 papers). The discussion was divided into several sections, covering the issue of embedding parameter selection and discussing results related to the application of RQA analysis.

4.1. Impact of Embedding Parameters Selection

The papers in this review are very different, even though they deal with the same issues. Unfortunately, many authors do not provide information on how the parameters needed for RQA analysis, that is, embedding dimension, time delay, and recurrence threshold, are calculated, which makes the procedures hardly reproducible. A similar problem applies to providing information on the program with which they performed the calculations. Among all 28 papers, only in 12 of them was it possible to find precise information on this issue. Most often, in nine studies [17,20,37,44,50,53,63,64,67], authors chose different versions of the Cross Recurrence Plot Toolbox 5.13 (R25, R13) developed by Marwan et al. [6,12]. Both Hasson et al. [42] and Shokouhyan et al. [45,69] based their calculations on Webber and Zbilut’s [10] paper. Palmisano et al. [65] chose the Recurrence Quantification Toolbox for MATLAB (http://au.mathworks.com/matlabcentral/fileexchange/46765-recurrence-quantification-analysis-rqa, accessed on 29 April 2023) [66].

Referring to the embedding parameters. In the first group, all authors provided relevant information. In the second group, there was no information in the two papers on determining embedding dimension and delay time. In the third group, this information was not present in 10 of the 12 studies. The embedding dimension in each paper was calculated using FNN. The lag time was calculated using the autocorrelation function in three studies [42,50,53]. Similarly, the AVD method was used in three papers [17,42,56]. In contrast, eleven papers used the AMIF method. Only Hasson et al. [42] used three (AVD, AMFI, and ACF) methods to determine the lag time. Hasson et al. [42] found similarly to Riley et al. [15] that the autocorrelation and mutual information techniques used to select time lag values were inconsistent and sometimes yielded values that were too high. They showed that the average displacement method (AVD) developed for noisy datasets by Rosenstein et al. [38] was a preferred technique. van den Hoorn et al. [17] addressed the determination of the recurrence threshold as one of the crucial parameters of RQA [70], as below its level, the recurrence is defined. Various methods have been used to determine the recurrence threshold. The most popular ones are the percentage of the maximum diameter of a geometric shape in phase space [37] or the percentage of the average distance between all data points in phase space [15]. Generally, this percentage is adjusted so that the %REC values are acceptable (around 1% up to 10%) [50]. Finally, van den Hoorn et al. [17] demonstrated that fixing the recurrence rate threshold (5%) improves the reliability of RQA within sessions and can increase sensitivity in differentiating study groups. In summary, in any of the previously published papers, it is not specified which method is best for determining the time lag or recurrence threshold. Therefore, it is recommended to use several of them to check the reproducibility of the results. In contrast, the embedding dimension was always determined using the FNN method.

4.2. RQA Results across Age Groups

Several studies [3,71] report that the ability to control standing balance develops during childhood into early adulthood and deteriorates from age 40 to 59. Of course, this is related, among other things, to the fact that postural control is closely linked to the functionality of the central nervous system and consequently to cognitive and perceptual processes [72]. van den Hoorn et al. [20] and Seigle et al. [50] showed that RQA measures in the anterior-posterior direction are sufficient to distinguish the young and elderly groups and even distinguish non-fallers from fallers. Hao et al. [44] addressed the description of changes in standing balance in preschoolers. They found that 5-year-old children showed increased CoP sway’s regularity (higher %DET) and intermittency (higher %LAM) in the AP direction than 3- and 4-year-old children. The intermittent behavior of CoP movement reflects the interrupted control mechanism of standing balance, which manifests as changes in CoP dynamics from fluctuating to quasi-stationary [20]. van den Hoorn et al. [20] observed a decrease in %LAM with age and for older fallers compared to older non-fallers. These results indicate that intermittent CoP movement increases with age in preschoolers and decreases in older adults, suggesting that this may be related to physiological changes in development or aging. Similar conclusions came from Hourova et al. [54]. They concluded that %LAM is the most sensitive factor in determining differences in CoP position between different age groups, as it describes motion fluidity.

4.3. RQA Results across the Sensory Conditions (Visual Control, Surface)

Many of the papers included in this review followed similar testing conditions. These consisted of combinations of both legs standing on firm or foam surfaces with eyes open and closed [44]. Hao et al. [44] showed that as the sensory conditions became more severe (eye closing and/or head extension), the amount and variability of postural sway increased while its intermittency decreased. The same conclusions appear in the study by Riley et al. [15], who were the first to apply RQA analysis to CoP data to investigate the effects of head position and visual feedback on postural sway dynamics. They indicated that vision affects the deterministic structure (in degree and complexity) of CoP motions and that spontaneous swaying in different head orientations affects CoP non-stationarity.

Seigle et al. [50] showed that for the elderly group, removing visual feedback decreased ENT values significantly in both directions and %DET in the ML direction. Mazaheri et al. [56] showed that as postural conditions became more challenging (changing from both legs standing on a rigid surface with visual control to standing on a foam surface without visual inspection), the postural sway in both AP and ML directions became more regular (higher %REC and %DET), more complex (higher ENT), and more stationary (lower TREND) in healthy adults and people with nonspecific low back pain (LBP). Moreover, they showed that the increase in cognitive difficulty was associated with less regularity (lower %REC and %DET), less complexity (lower ENT), and more stationary (lower TREND) in both directions. Additionally, LBP patients did not reduce their %REC, %DET, and TREND in the AP direction with increasing cognitive difficulty to the same extent as healthy people. For the corresponding measurement conditions, similar results were obtained by Negahban et al. [59]. Unfortunately, of the RQA measures, they only considered %DET, which was higher in the medial-lateral direction in patients with knee osteoarthritis compared to healthy subjects. Moreover, in both groups, %DET increased with increasing postural difficulty, while it decreased when moving from single- to dual-task conditions. Negahban et al. [46] also investigated the effect of a prior anterior cruciate ligament injury on the deterministic pattern of postural sway under the above-mentioned conditions. They showed that as postural difficulty increased from open-eyes to closed-eyes and rigid-surface to foam-surface, the center of pressure regularity (%determinism) and ENT increased as well. As previously mentioned, the performance of a secondary cognitive task (a backward digit span task) caused less center of pressure regularity than the single postural task. In all studies, the following relationship appears: an increase in variability is associated with a decrease in randomness (i.e., an increase in determinism), and vice versa. Schmit et al. [55] found that higher levels of postural variability in people with Parkinson’s disease compared to a matched control group accompanied greater determinism, % recurrence, and max line. Riley and Clark [73] found that when the sensory organization test conditions became difficult, CoP variability increased, while the %DET decreased. Those findings indicate that the CoP time series in the previously mentioned groups are not noisier than the postural sway of controls but exhibit strongly deterministic dynamic patterns.

4.4. Impact of Training on RQA Parameter Values

This section includes papers where the effect of training (brisk walking program, aerobic exercises, and practice of contemporary dance) on postural responses was evaluated. Bernard et al. [67] indicated a lack of influence of a 6-month brisk walking program with three sessions of 60 min per week at moderate intensity on postural responses and on dynamical measures obtained by time series analysis. In subsequent studies, Bernard et al. [64] showed that in older sedentary women, two consecutive bouts of 6 min of aerobic exercise separated by a 20 min rest period did not alter linear postural kinematic measures but significantly increased the %DET in the ML direction. This result indicates reduced postural complexity and less adaptability in the ML direction [74]. The authors interpret it as a less adaptive behavior related to hip muscle work strategies negatively affected by walking exercise. Ferrufino et al. [43] examined the impact of contemporary dance (CD) and fall prevention (FP) programs on the postural control of older adults. The authors showed that the practice of CD in older adults reduced recurrence and mathematical stability in RQA (LMAX) as compared to the practice of FP, which tended to induce reverse patterns. Similar studies were conducted by Coubard et al. [62]. The authors examined whether massed CD training (4.5 h a week for one month in duration) might induce similar effects on postural control in older adults. They showed that LMAX (mathematical stability) decreased in the non-dancers (ND) group in the eyes open condition between pre- and post-test periods, whereas it remained stationary in the CD group. The authors report that such an effect may have resulted from the test-retest. Participants may have adopted a different strategy of standing upright in the post-retest assessment without intervention.

van den Hoorn et al. [20] evaluated the effect of added and removed calf vibration on balance control (without visual control) in elderly individuals who reported falling or not and in a group of young subjects. Vibrators attached bilaterally over the triceps surae muscles were activated twice for 15 s, after 15 and 75 s. CoP movement was compared in 15 s epochs before, during, and after calf vibration removal between groups. Before vibration, compared to fallers, young people had a lower %DET (were more predictable), a longer LMEAN (less sensitive to small perturbations, i.e., better balance performance), and a higher %LAM with a longer TT (were more intermittent). Compared to non-fallers, the young also had longer LMEAN and TT, but %DET and %LAM were not significantly different between young and non-fallers. In the young group, %DET and LMEAN (balance performance), %LAM, and TT (intermittent control) were higher in all epochs as compared to fallers and non-fallers. No differences were observed between fallers and non-fallers at baseline. During ankle vibration, fallers and non-fallers were not significantly different in any linear or non-linear outcome variables. In conclusion, compared to older subjects, CoP movement in the young was more predictable and less sensitive to small perturbations. Balance measurements did not differ between falling and non-falling subjects before and during vibration.

4.5. RQA, Linear Parameters, and Future Research

This section includes papers in which the authors combined RQA results with traditional linear measures such as CoP path length, ellipse area, and mean sway velocity. Within the first group, linear and RQA measures were used together in five [37,44,50,51,53] of the eight studies. In the second group, such a connection was noted in four manuscripts [45,55,57,58]. In the last group, all 12 studies included both sets of parameters. It is worth highlighting that in all the mentioned studies, linear and non-linear methods were described separately as they evaluate different features of postural control.

Seigle et al. [50] showed that removing vision during quiet standing results in a significant increase in CoP path length and surface area in a group of young and elderly subjects. Such findings were in line with results from the papers [18,43,44,51,61,63] included in this review. Moreover, Seigle et al. [50] observed three typical postural behaviors associated with a lack of visual control. The first was characterized by a rigid posture associated with a reduction in the magnitude of postural oscillations. The second behavior was characterized by a functional exploratory strategy [75] associated with increased CoP magnitude. The third behavior was characterized by increased postural fluctuations and an inability to adapt to the conditions. As mentioned in the introduction, such behaviors should be described using both linear and nonlinear measures [3]. Linear measures provide crucial information about the magnitude and direction of postural sway. On the other hand, nonlinear measures, and RQA in particular, are relatively new and provide a different perspective on postural control. They consider the complexity and dynamics of the human movement system, which cannot be easily captured by linear measures. Overall, as discussed in seven papers of this review [17,20,44,53,61,65], the use of both linear and nonlinear measures provides a comprehensive understanding of postural control and helps researchers and clinicians identify the underlying mechanisms of postural instability and develop appropriate interventions.

According to some authors [37,51,53], future studies should include additional EMG-type equipment, functional tests, or more challenging sensory conditions to facilitate data interpretation. Hao et al. [44] recommended future studies for preschool and primary school-aged children to assess the developmental trend of standing balance in children. It also seems that the inclusion of RQA analysis for patients with various disabilities or diseases is an open space, given that there were only eight papers in this section.

5. Conclusions

Recurrence quantification analysis seems to be a well-accepted assessment method in postural control studies during quiet standing.

The importance of selecting appropriate parameter values for RQA is undeniable. No single optimal value for time delay or recurrence threshold exists. A good indicator of well-chosen parameters is the high repeatability of the results. The embedding dimension should be determined using the FNN method.

Using RQA, it is possible to see that the randomness of postural responses, represented by the %REC, %DET, ENT, and TREND values, decreases as the difficulty of the measurement conditions increases. Many authors have shown that the RQA is a good tool for determining differences between the young and elderly, and the %LAM seems to be the most sensitive indicator. It is worth highlighting that most authors concluded that traditional and nonlinear methods provide complementary, rather than redundant, information for assessing age- and health-related changes in standing balance. They also argued that the interpretation of these measures needs to be combined.