Biomechanical parameters for gait analysis: a systematic review of healthy human gait

Background: Modern gait analysis offers a broad variety of biomechanical parameters through which to quantify gait. However, no consensus has yet been established with regards to which biomechanical parameters are most relevant to evaluate during gait analysis in the healthy population. Purpose: The purpose of the current systematic review was to determine the most relevant biomechanical parameters for gait analysis in the healthy adult population. Methods: PubMed, EMBASE and Web of Science databases were searched. Two independent reviewers participated in the article selection and attributed a Level of Evidence score to each article to account for quality based on participant selection, intervention and analysis. A score combining both frequency and number of articles was calculated. Correlations were carried out between the Level of Evidence score, Journal Impact Factor and the frequency of biomechanical parameters. Results: Spatio-temporal parameters were found to be the most often measured biomechanical parameters and reported by the greatest number of articles; walking velocity, cadence and step/stride length appearing to be the most relevant biomechanical parameters for gait analysis in the healthy adult population. No correlation was found between Level of Evidence score and Journal Impact Factor, nor between the frequency of parameters and Level of Evidence score. Conclusion: This systematic review provides recommendations for variables to assess in future gait evaluations in healthy adults.


Introduction
Walking is the most common form of locomotion and it is part of almost all activities of daily living [1,2]; therefore, the ability to walk is an indicator of overall health as it dictates autonomy [3]. Although walking is usually learned at a young age, the mechanics of walking are not as simple as they may appear [1].
From the first studies of human walking elaborated through a series of photographic images, by early Biomechanics enthusiasts Edweard Muybridge and Étienne-Jules Marey, gait analysis as it is known today has evolved significantly [4]. The walking pattern of individuals has become an area of broad interest and the focus of much research as seen by the numerous journals and articles published. The importance of gait analysis lies in its application; through years of research and experimentation, gait analysis has become widely used as a means to diagnose pathology, set a prognosis and establish and evaluate a treatment plan [5,6]. Today, a variety of different parameters of various types exist and are readily used to examine and explain human gait [7][8][9][10].
In clinical settings, gait analysis is often carried out solely through clinician observation [11]. Although clinicians have developed good expertise through many years of practice and training, these observations remain subjective [12]. Principal reason for main, and perhaps sole use of clinician observation as means of gait analysis, is ease of measurement [8,13,14].
In the research setting, numerous parameters have been used to quantitatively describe gait. Parameters of various types such as spatio-temporal parameters, ground reaction forces, Physical Therapy and Rehabilitation ISSN 2055-2386 | Volume 4 | Article 6 CrossMark ← Click for updates doi: 10.7243/2055-2386-4-6 joint kinematics and the energy expense are a few [1,15,16].
In accordance with evidence-based-medicine, the biomechanical parameters chosen are as important as rigorous gait analysis technique [17]. Because of the quasi-infinite number of parameters available, it seems reasonable that certain parameters would be best suited for gait analysis in the healthy population.
Systematic reviews have been realized in an attempt to organize and add understanding to the practice of gait analysis in various populations. For example, a systematic review carried out by Sagawa and colleagues [18], using an original methodological approach, was able to identify the most relevant biomechanical parameters for assessing gait in individuals with a lower limb amputation. The results obtained by Sagawa and colleagues [18] leads to question whether the same biomechanical parameters are most relevant for gait analysis in the healthy adult population.
The aim of this systematic review is to determine the most relevant biomechanical parameters used for gait analysis in a healthy adult population.The term relevant was defined as those biomechanical parameters being able to identify gait abnormalities in the healthy adult population and applicable to the clinical and rehabilitation setting. This definition is an adaptation of that used by Sagawa et al. [18].

Procedure for the identification of selected articles
We performed an online search in three databases: PubMed, EMBASE and Web of Science. These three databases were selected for search because of their broad inclusion of multidisciplinary topics within the Biomedical and Health Sciences domain. Each database was searched for all years included in the respective databases with the last search completed in May 2016.
The following search was inputted to all three databases: [abstract/title] (speed OR cadence OR (stride time) OR (swing time) OR (step time) OR (single support time) OR (double support time) OR (foot flat time) OR (stance time ratio) OR (swing time ratio) OR timing OR (stride length) OR (step length) OR (step width) OR angle OR moment OR power OR (center of mass) OR (ground reaction force) OR (ground reaction impulse) OR (center of pressure) OR rotation OR symmetry OR velocity OR (stance phase) OR (swing phase) OR (cycle time) OR (spati* temporal) OR hip Or knee OR ankle OR foot) OR (biomechanic*) AND ([MeSH] gait OR walking OR locomotion).
In databases where applicable, certain additional parameters were used to narrow the search. In PubMed, filters including human studies of adults aged 18 to 65 years old, published in French and English and with regards to the nature of the study (i.e., original articles, review articles, case study) were applied. In the EMBASE and Web of Science databases, filters were applied to include human studies, French and English language publications and specific nature of study (i.e., original articles, review articles, case study).

Inclusion and exclusion criteria
The inclusion and exclusion criteria were developed based upon the purpose of the systematic review, to examine the biomechanical parameters used to study healthy gait. Thus, studies including participants living with pathologies, disabilities, health concerns and/or neurological deficits were excluded. To be selected, articles had to evaluate adults aged 18 to 65 years old with no walking aids. Participants could have been evaluated barefoot, wearing socks, wearing shoes and/or any combination of these three situations. As well, no studies were included if they measured the effect of a treatment or equipment. Selected articles had to at least evaluate participants walking at their self-selected speed on an overground and flat surface.

Analysis of selected articles
A census of all biomechanical parameters measured was undertaken by two evaluators by carefully reading and analyzing the chosen articles. First, all methodological aspects of the selected articles were tabulated and briefly summarized. Second, the biomechanical parameters measured in all articles were tallied. For each parameter, all articles which measured this parameter were reported and counted. Third, because of the many various instruments, techniques, planes of measurement, etc. used to quantify parameters in the studies selected, the parameters measured were summarized under broader parameter names (i.e. sagittal, frontal and transverse plane knee power were combined under the broader name of knee power).
Lastly, after a summation of parameters, the number of different articles measuring a type of biomechanical parameter was counted; this was also done for single parameters. Indeed, it seems inevitable to consider not only the most frequently measured parameters, but as well the number of different articles which measure a parameter to observe any disparities between the number of times a parameter was measured versus the number of different articles which measured this parameter.
In an attempt to evaluate relevance of biomechanical parameters, both the frequency of measure and the number of different articles which measure the parameter were combined to produce a score using the summarized parameters. For the first factor, all frequency of measurement scores were divided by the parameter having been measured the most times (hip power: 66 times) and multiplied by 0.5. For the second factor, all number of articles were divided by the parameter having been measured by the most amount of different articles and multiplied by 0.5. Both values were then added to obtain a score weighting both factors. It was deemed that both factors were as important as the other, each contributing to 50% of the score. The following is an example of the calculation for walking velocity, which was measured 50 times by 50 articles: Walking velocity: ((50/66)*0.5) + ((50/50)*0.5)= 0.879.

Quality of selected articles
We evaluated quality of the selected articles by attributing doi: 10.7243/2055-2386-4-6 a Level of Evidence score for each selected article. Our Level of Evidence score was a modified version of that used by Sagawa and colleagues [18], since they were interested in gait analysis in a population with a lower limb amputation and the current systematic review addresses healthy adult gait analysis. The 14 criteria were subdivided between three main article elements: 1) selection of participants, 2) intervention and assessment, and 3) statistical validity. The maximum possible score was therefore 14, with each article receiving a score of 1 (if they met the requirements) or 0 (if they did not meet the requirements) for each criterion (score of 1 for a non-applicable criterion). Two independent evaluators assessed the score of all articles. For any disparities between scores, both evaluators determined the best suited scoring through discussion. If a consensus could not be reached by the two evaluators, a third evaluator intervened in order to break tie between both scores suggested.

Data/ Statistical analysis
A Spearman correlation was carried out in order to determine if higher Level of Evidence articles were published in higher Impact Factor journals. Also, a Spearman correlation was sought between the mean Level of Evidence attributed to all articles measuring a given parameter and its frequency of measurement. All statistical analyses were carried out using SPSS 22 (IBM Corp., NY). Level of significance was set at p<0.05.

Selection of articles
The preliminary database search, using the previously mentioned keyword combination, yielded 16 023 abstracts throughout all three databases. Upon reading the titles and applying the inclusion and exclusion criteria, 1388 articles were retained for further selection. After reading the abstract, 515 articles remained. Finally, after a careful reading, 65 articles fulfilled the inclusion and exclusion criteria and were selected for further analysis (Figure 1). Table 1 outlines the main methodological aspects of these selected articles.

Participant characteristics
The main participant characteristics of the 65 selected articles are outlined in Table 1.

Article data quality
The Level of Evidence score attributed to each article was in agreement between reviewers. The mean Level of Evidence for all articles was 11.8±1.8, with scores ranging from 6 to 14. The Level of Evidence scores attributed to the 65 articles are outlined in Table 1. Table 2 indicates that parameters of various types were measured and counted in the selected articles. Parameters from power, work and/or torque were recorded 269 times, Flowchart as per PRISMA guidelines [19] summarizing the procedure for the selection of articles after the interrogation of three databases. All articles were retained or dismissed for analysis by the application of the inclusion and exclusion criteria (see methods). First, the articles were retained or dismissed on the basis of the article titles. A second step consisted of the reading of the article abstracts. Finally, all retained articles were read and a final selection was made.
All measured biomechanical parameters in the selected articles are outlined in Table 2. The parameter most often measured and/or calculated was the walking velocity (50 times) followed by cadence (30 times), stride length (23 times) and step length (21 times).

Parameter summation
As stated, a summation of parameters was carried out (results outlined in Table 3) and the results show that the hip power is the most often measured biomechanical parameter (66 times) followed by the knee power (61 times), walking velocity (50 times) and the ankle angle (47 times).
Also outlined in Table 3 is the number of different articles measuring summarized single parameters. Spatio-temporal parameters were measured in 59 of the 65 articles, angles by 29 different articles and forces in 16 articles. When considering summarized single parameters, walking velocity was measured in 50 different articles and stride length and cadence were measured in 36 and 35 different articles, respectively.
The calculation to account for both frequency of measurement and number of articles was carried out with the highest To analyze foot and ankle kinematics from gait recordings of healthy subjects walking at comfortable and slower speeds.
11 [34] 10 (7M: 3F) 23 (2) To analyze the 3D angle between the joint moment and the joint angular velocity vectors at the ankle, knee and hip during the gait cycle and to investigate if these joints are predominantly driven or stabilized.
11 [35] 46 (32M: 14F) --Velocity, stride length and stride frequency were treated as independent variables in relation to each other in a graphic form to see how they interact in gait. To achieve this, a Velocity Field Diagram (VFD) was described.
12 To investigate the short-term relationships between footstep variables during steady state, straight-line, over-ground walking in healthy adults and to explore the extent to which the performance of a step or stride is dependent on the performance of an earlier step or stride in a sequence.
13 [54] 10 (6M: 4F) 23 (5) 1) Quantifying gait pseudo-periodicity using information concerning a single stride; 2) investigating the effects of walking pathway length on gait periodicity; 3) investigating separately the periodicity of the upper and lower body part movements; 4) verifying the validity of foot-floor contact events as markers of the gait cycle period. To examine whether there is an optimal walking speed with minimum intrasubject variability in step length and step width during free walk and whether there is an optimal step rate with minimum step length variability during walking with imposed step rates. To determine the kinematic variability of the lower extreity joints using methods from the mathematical chaos theory in a normal walking environment in conjunction with a large population of healthy young adults and to test the hypothesis that variability characteristics are different between joints and to further investigate differences between male and female and right and left subgroups.
13 [69] 10 (5M: 5F) 19-34 1) To introduce the knee moment arm length as a measure to evaluate knee pre-and postoperatively; (2)  To determine the familiarization period required to obtain consistent measurements of the angular movements of the lumbar spine and pelvis during treadmill walking. To study the effect of walking at a self-selected and at a slower speed on the angular movements of the pelvis and lumbar spine and how interpretation of speed effects on lumbar spine movements was influenced by frame of reference, either relative to the pelvis or relative to a global reference frame.  Step length 21 [20,21,26,33,41,43,44,48,52,53 Step time 4 [20,41,75,83] Stance time 12 [38,24,29,39,40,41,49,59 Biomechanical parameters measured in included studies. This chart tabulates each biomechanical parameter as it was measured in the designated study. The reference measuring each given parameter is given, as well as the total for single parameters. The following parameters are grouped according to their type and a total of number of parameters measured per type is given in parentheses. Only parameters measured more than once are shown here.  Summation of all biomechanical parameters measured in included studies. This chart tabulates each parameter under a broader theme of parameters as well as the number of different articles which measure this summarized parameter. The total number of parameters measured per type is shown in parentheses beside the parameter type; the total number of different articles measuring a type of parameter is given beside these parentheses. The breakdown of the summation is not shown here. Only parameters measured more than once are shown.
frequency of measurement being the hip power (66 times) and the greatest number of articles being walking velocity (50 articles). Walking velocity obtained the highest score (0.879), followed by stride length (0.686), cadence (0.630), hip power (0.590) and knee power (0.552). The results of this score are presented in Table 4.

Level of evidence score and Journal Impact Factor
It was sought whether a correlation existed between the article Journal Impact Factor (not shown) and the Level of Evidence score attributed to each article by means of a Spearman correlation. The result of this correlation is a very weak, negative and non-significant correlation (r s =-0.133, p=0.105). The Impact Factor scores of 4 articles [21,54,59,67] were unavailable and were therefore excluded.

Frequency of parameters and Level of Evidence score
When the frequency of the most often reported parameters was correlated with the mean Level of Evidence score of articles (not shown), via a Spearman correlation, a weak, negative and non-significant correlation was found (r s =-0.224, p=0.06).

Number of articles
The current review was based on 65 articles. This number may appear small knowing that the review of Sagawa and colleagues [18] included 89 articles of a clinical population. The present study reflects the restrictiveness of our inclusion and exclusion criteria.

Types of biomechanical parameters
Considering types of parameters, it was found that power, work and energy parameters were measured most often (269 times): spatio-temporal parameters followed closely being measured 256 times. Joint angle parameters were measured 177 times, joint moment parameters were measured 115 times and forces were also measured 115 times. In comparison, the systematic review of Sagawa and colleagues [18], revealed that parameters of spatio-temporal type were measured 153 times, joint angles 78 times, platform parameters (i.e. ground reaction forces and center of pressure) 72 times, powers 64 times and joint moments 58 times. Thus, in general, the number of times a type of parameter was measured was less in the review of Sagawa and colleagues [18] than in the present review despite the fact that fewer articles were included for analysis in the current review.
These larger numbers are explained by the fact that both studies did not group parameters in the same manner; therefore, the number of parameters in relation to the total number of articles included in each study is different. Also, Sagawa and colleagues [18] carried out a summation of parameters in which both time sub-parameters and amplitude sub-parameters were grouped separately. For the purpose of our systematic review, it was thought more appropriate to group parameters accordingly, since all are yielded from one measure.
Omitting these disparities, it is possible to note that spatiotemporal parameters are of high relevance in both systematic doi: 10.7243/2055-2386-4-6 reviews. As well, all most frequent types of parameters are the same, although they differ in number and order of relevance.

Single parameters
When looking at single parameters, the walking velocity (50 times), cadence (30 times), stride length (23 times) and step length (21 times) were those parameters most frequently measured. These results are in agreement with Sagawa and colleagues [18] who conclude the same parameters were most often measured: walking velocity, cadence, stride and step length. It is interesting to note that for these two different populations the same parameters would appear to be most relevant. This may be because parameters of spatio-temporal type have a certain ease of measurement in comparison to other parameters.
Despite the fact that in the current systematic review, power, work and energy parameters were the most frequently reported measures as a type of parameter, when considering single parameters, the most frequently measured were spatio-temporal. Interestingly, for power, work and energy type of parameters, no single parameter was reported more than 10 times and most parameters were measured only once. In fact, for these types of parameters, a given parameter can be measured at different instances of the gait cycle, in three different planes and for minima and maxima values, making the number of parameters somewhat inflated.
As well, more minima and maxima power values exist at the hip joint when compared to the ankle joint, for example. This may also explain some disparity in the frequency of measure-ment of some parameters, especially kinematic parameters of the lower limb joints.

Summation of parameters
After a summation of parameters, we observe that those parameters most frequently measured are hip power (66 times), knee power (61 times), walking velocity (50 times) and ankle angle (47 times). Following parameter summation, Sagawa and colleagues [18] concluded walking velocity (43 times), knee angle (31), knee moment (27 times) and hip power (26 times) were most often measured. These differences might reflect that the results are somewhat inflated and the angle, moment and power parameters need to be interpreted cautiously. Indeed, Sagawa and colleagues [18] did not group parameters in the same way as was done in the present review and the frequency obtained for angle, moment and power parameters are smaller. As well, for certain types of parameters (i.e. power, work and energy), the number of total parameters measured (i.e. 269 times) may also be inflated. Again, this may explain some disparity between the number of parameters measured with regards to the total number of articles.
Another explanation for these differences is the type of population studied. Indeed, their choice of clinical population implied the absence of the ankle joint which can explain the lack of ankle joint measures in their population with a transtibial amputation. In the healthy adult population, ankle joint measures were in the top four most relevant parameters after parameter summation.
Also interesting is that articles which measure hip mo- ments, also tend to measure joint moments at the knee and ankle, as they are necessary in inverse dynamic calculations. As well, it is interesting to note that forces are needed in the calculation of moments and angular kinematics are needed for power calculations. Therefore, articles measuring powers, would also measure kinematics, forces and moments and this plays an important role when looking at frequency and relevance of parameters.
In addition to the frequency of measurement, it is also important to consider the number of different articles measuring a given parameter. Out of the total 65 articles included in our systematic review, spatio-temporal parameters were reported by 59 different articles, joint angles were reported by 29 articles, followed by forces (16 articles), joint moments (13 articles) and power, work and energy (13 articles). So for power, work and energy parameters (measured 269 times), the type of parameters which appear to have been measured most often, only 13 out of 65 articles measured these types of parameters. In comparison, spatio-temporal parameters (measured 256 times), were evaluated in 59 of the 65 articles.
As for the type of parameters discussed above, using the summarized parameters, the walking velocity remained the most often measured (50 articles out of 65 total articles) followed by cadence (35 articles), stride length (36 articles), gait cycle parameters (23 articles) and stance time (19 articles). However, when comparing these results to those of Sagawa and colleagues [18], we observe that a higher number of articles reported the most common parameters in our present study: walking velocity was measured only 43 times in 89 articles, cadence 19 times and step and stride length 19 times and 15 times, respectively. An important note must be made here that parameters were summarized differently by both reviews. The differences in the number of articles can, in large part, be explained by the choice of inclusion and exclusion criteria.
As shown by our results, both frequency of measurement and the number of different articles measuring a parameter are of importance when investigating the most relevant biomechanical parameters for gait analysis. The results of the score combining both these factors show that walking velocity, stride length and cadence appear to be most relevant.

Level of Evidence score
The mean Level of Evidence score for all articles was 11.8±1.8 out of 14 points. This mean score is high; one can argue that it almost reaches a ceiling effect. It is perhaps because the Level of Evidence was not discriminatory enough in the limits for scoring. This can also be due to high quality and soundly based studies. It is perhaps simpler to carry out quality experimentation in a healthy population since there may be less physical restrictions and/or needs as compared to other clinical populations. This may also be due to the inclusion/ exclusion criteria weeding out the lower quality articles. A Level of Evidence score with a wider array of possible scores would be needed.

Relation between the Level of Evidence score and Journal Impact Factor
The Level of Evidence score of articles were correlatedwith the Journal Impact Factor. The weak, negative and non-significant Spearman correlation found is in agreement with that of Sagawa and colleagues [18] who carried out this same analysis but with the Journal Impact Factors of the year of publication of their systematic review. It is possible to conclude that both Level of Evidence score and the Journal Impact Factor are not related.

Relation between the frequency of parameters and their mean Level of Evidence score
The mean Level of Evidence of the articles was correlated with the frequency of the parameter measured. As stated in the results section, a weak, negative and non-significant Spearman correlation was found. It is therefore possible to conclude that the frequency of measurement of a parameter is not related to the mean Level of Evidence of the articles which measure this parameter.

Most relevant biomechanical parameters
Spatio-temporal parameters, namely walking velocity, cadence and step and stride length, appear to be the most relevant biomechanical parameters in both individuals with a transtibial amputation and healthy adults. In addition, walking velocity is of even greater relevance since it also measures, and has a direct effect on such parameters as cadence and stride length.
Additionally, these spatio-temporal parameters have a certain ease of measurement: measuring simple spatio-temporal parameters such as walking velocity would appear to be an effective and simple manner to add objectivity to clinical gait analysis which is primarily aimed at ease of measurement [8,13,14].
Future studies should aim to identify if the most relevant biomechanical parameters for gait analysis found in healthy adults are also relevant to other clinical populations. Individuals with a transtibial amputation and healthy adults yielded the same most relevant parameters, but perhaps the results obtained in other populations would be different, such as in populations with a neurological disorder (i.e.: Parkinson's, Stroke or Cerebral Palsy) or with a more severe mechanical impairment (i.e.: bilateral trans-femoral amputation).

Conclusion
A systematic review of the literature pertaining to healthy adult gait was performed and the most relevant biomechanical parameters were identified. Spatio-temporal parameters were those parameters most often measured and by the most amount of articles. Additionally, many specific spatio-temporal parameters were those most often measured (walking velocity, cadence and step/stride length), walking velocity being doi: 10.7243/2055-2386-4-6 measured most often, and by the greatest number of articles. Walking velocity, and other spatio-temporal parameters would therefore appear to be the most relevant biomechanical parameters to healthy adult gait analysis.
To our knowledge, this is a first systematic review of its kind in a healthy adult population and the implications of these findings are important for choosing the most relevant biomechanical parameters for gait analysis.