Benefits,Challenges,and Potential Utility of a Gait Database for Diabetes Patients

What Is Gait Analysis?

Gait analysis is the assessment of how people walk and negotiate their environment, providing insight to how an individual or group of people are functioning during a given activity. Gait analysis has applications in both identification of movement-related problems as well as enabling condition or individual specific comparisons to be drawn for tailored treatment to improve movement during a given task. The greatest strength of gait analysis clinically is the ability to identify movement impairments and the potential underlying physiological mechanisms typically using noninvasive techniques. However, this relies on understanding what is considered to be “normal” gait to identify any deviations from this “norm.” An individual’s gait is influenced by numerous physiological elements including the musculoskeletal, vestibular and nervous systems, as well as psychological implications on control of movement. Therefore, any condition/factor that impacts on these elements has implications for how an individual performs a given activity. Identification of gait strategies within both healthy individuals and patients with diabetes is therefore critical to the identification of differences between the populations, as well as individual differences within a single patient, which allows for personalized diagnoses and care. Gait can be highly individual with variance existing across any group/population, with exclusion criteria helping to more clearly define populations. This typical experimental design controls for the influence of potential confounding variables, but on the other hand limits the ability to assess multiple complications simultaneously.

What Is a Database?

Databases are common for collating large quantities of data. Most research studies will store patient information and research data within databases to clearly and robustly separate out variables and allow individual components of a large data set to be easily compared. Within modern culture and increasing global interconnectivity, large collaborative databases are becoming more common. A collaborative database allows multiple individuals or organizations to contribute data/information, which is stored in a consistent manner and allows members to then ask questions of that database. The aim of this article is to discuss the potential benefits, challenges, and utility of a gait database for diabetes patients.

The Value of Gait Analyses of Patients With Diabetes

Gait in patients with diabetes is a highly researched area of interest: a search of Elsevier at the time of writing yields over 12 000 articles relevant to diabetes and gait, with the annual publication rate rising steadily over the past 2 decades from 165 in 1998 to 924 new publications in 2015. Much of this research has focused on aspects of great importance most notably falls and foot ulceration. Both of these issues, which are of particular risk to patients with diabetes,^11-13 present challenges to health and quality of life and are associated with the mechanisms of walking. Falling is directly related to an individual’s ability to maintain an upright posture during standing and walking and foot ulceration is largely contributed to by the daily mechanical loading applied to the foot. Therefore, it is crucial to understand the mechanisms behind these issues and develop treatments to reduce their incidence.

Foot Ulceration: Prevention and Treatment

Diabetic foot ulceration is particularly common in patients with diabetic peripheral neuropathy (DPN) ¹⁴ due to the loss of the natural protective sensory responses. With a DPN present in up to 50% of patients, DPN and related foot ulceration place a huge burden on the population and the health care service. Gait analysis has provided insight to those mechanisms which increase the risk of foot ulceration, thereby facilitating preventative measures. Plantar pressure measurement performed during gait allows for identification of high-pressure areas indicative of high-risk sites for ulceration as can be seen by comparison of 3 examples in Figure 1; where increased pressure can be seen under the heel and metatarsal heads in the foot of an individual with diabetes, and under the midfoot of a patient with a rocker-bottom Charcot deformity. This allows interventional therapies to be employed to reduce pressures in the identified areas, often in the form of specialized footwear or orthoses.^15,16 Other methods shown to reduce plantar pressures include: surgery to increase dorsiflexion range ¹⁷ or replace depleted soft tissue; ¹⁸ as well walking with shorter step lengths.₁₉ There is some discussion on whether vertical plantar pressures alone are a strong indicator of ulceration. ²⁰ It is generally considered that additional influences, including loading time ²¹ and shear forces, ²² should also be considered.

OPEN IN VIEWER

Figure 1.

Example images of plantar pressures during walking for a healthy foot (left), a foot with heightened pressures underneath the heel and metatarsal heads (center), and a foot with a rocker-bottom Charcot deformity (right), characterized by heightened pressure underneath the midfoot underneath the deformity and reduced pressure at the heel. Levels of pressure are indicated by dark blue for low pressure, lightening shades indicate heightening pressure, with dark red indicating the highest levels of pressure.

Participant Populations in Typical Studies

While gait analyses from various isolated studies have contributed greatly to understanding gait impairment in diabetes and the implementation of appropriate interventions, there are limitations introduced by how these are utilized. The manner in which a patient with diabetes is affected by the disease is highly individual: sensory, vascular, and muscular complications are all common. However, a single patient may develop all or none of these complications; this means that when comparing this population to individuals without diabetes, care must be taken in interpreting the cause of any potential alterations. A typical gait study will assess somewhere between 20 and 100 individuals, divided into subgroups, 2 recent reviews of gait literature regarding diabetes found average study sizes of n = 48 ⁴³ and n = 52. ⁴⁴ These study populations are subdivided most commonly into people with and without diabetes (with or without additional complications), giving an average group size within the studies reported of n = 21-25.^43,44 Within the United Kingdom there are in excess of 3 million patients diagnosed with diabetes;^1,2 therefore a typical gait study assesses a sample <0.00003% of the population, a sample level that is common among studies that require voluntary human participation and interaction.

The Limitations of Restricted Sample Sizes

As a result of the typical sample sizes seen within gait analyses the scope of an individual study is often narrow by necessity; strict exclusion criteria are applied to control for potential confounding conditions, thus allowing focused conclusions to be made from each study regarding a specific aspect of how diabetes affects gait. What few gait studies have been able to do however, is simultaneously separate out the different effects of multiple complications present within the population. Meta analyses allow findings to be compared between studies, and trends to be identified across studies and attributed to individual complications of the disease. However, the utility of these reviews is limited by the variation of parameters reported within each study, the level of detail provided for each parameter, and the variation in methodologies between individual studies. This can cause differences to be noted between studies, that while similar may have resulted from populations with slightly different characteristics, due to variations in the inclusion criteria. For example, Sawacha et al ⁴⁵ showed increased anteroposterior ground reaction forces in patients with DPN; however Katoulis et al ⁴⁶ showed no difference in this variable for patients with DPN, unless the patients also had a history of previous ulcers. This subgrouping utilized by Katoulis et al ⁴⁶ demonstrated a difference in the fundamental mechanics of walking in patients who have previously had an ulcer. However, the findings appeared to disagree with those presented by Sawacha et al, ⁴⁵ who did not differentiate between patients with and without a history of ulcers. Including more subgroups, however, requires additional recruitment, data collection, and increased complexity of the comparisons to be performed. However, this example demonstrates the need to account for multiple variables to identify accurately the causes of any observed differences within a sample population.

Databases: Expanding Your Data Pool

A potential method to address the difficulty in collecting large quantities of data is that of a database. Current methodologies allow data to be collected for individual studies, recruiting a set of patients to assess a specific question, or series of questions. At the end of this process the data are discarded or stored according to the ethical requirements of the study. Following this traditional research approach, gait studies need to recruit new participants every time a new hypothesis requires testing. However, multiple studies may assess many of the same variables. Sawacha et al ⁴⁷ and Brown et al ³² each assessed walking along level ground; the foci of each study were different, and the exact methodologies differed; however, both studies recorded kinematic and kinetic data and the studies assessed 67 and 80 participants, respectively. The collaborative nature possible with databases, would share the burden of data collection between studies, allowing analyses to be performed with a far greater sample population forming a closer representation of the overall population. These future analyses could allow examination of an increased number of covariates (which would otherwise have required controlling), including those whose influences may be too subtle for statistical analyses to identify interactions within current study sample sizes. In addition, a long-term database system could create the potential for longitudinal studies to be performed with greater ease, allowing for the development of the many alterations to gait patterns in patients with diabetes to be further understood, or for alterations within a single patient to be identified and flagged for potential intervention.

Understanding of the subtleties of gait within the population may increase the diagnostic capabilities of gait analyses. Currently, plantar pressure analysis is one of the more clinically applied methods of gait analysis, used to identify specific areas on the plantar surface of the foot likely to ulcerate. Other techniques, such as a timed up-and-go test, ⁴⁸ a simple assessment that can be indicative of high fall risk, ⁴⁹ provide flags for specific risk factors within an individual patient. By increasing the number of complications considered during a single gait activity, there could be potential for using more in-depth gait analysis to identify the presence of multiple possible complications through noninvasive testing. Given the progression of technologies used for gait analysis, including the development of marker-free systems,^50-52 this could become a valuable diagnostic tool in the future, assuming adequate understanding of the influence of diabetes, to support its use.

Creating a Useful Gait Database

Ensuring Comparability

Opportunities for errors between data sets also exist: methodological, environmental, and instrumental. This is particularly pertinent if again considering a database with multiple contributing organizations. The technologies utilized for gait analysis vary greatly, and a single variable can be assessed by a variety of techniques. Gait kinematics, for instance, are generally measured by optoelectronic camera systems; these exist with active and passive marker technologies, are sold by multiple companies, or could also be measured by standard video cameras, accelerometers or goniometers. Figure 2 shows an example of a passive marker technology in use with a motion analysis system—reflective markers placed on the skin are detected by multiple cameras within the gait laboratory to calculate precise 3D coordinates of the marker positions. From this data a number of gait parameters can be determined. Each different variation of equipment will exist with a different range of instrumental error, as well as different limitations on its use (eg, video systems require a staged area, whereas accelerometers could be worn in an unprepared environment). This introduces issues when comparing data from different laboratories, collected using varying equipment. A useful database would require specific equipment to be utilized to maximize comparability, a potential for multiple variations of equipment could be designed into the database, but would have to be considered properly within any analysis.

OPEN IN VIEWER

Figure 2.

Example of a participant undertaking gait analysis stepping over a simulated obstacle, using an optoelectronic camera system and passive reflective markers. Also shown are wireless electromyography receivers (black boxes) and force platforms embedded in the walkway. Due to the highly reflective nature of the passive reflective markers they appear to glow brightly when lights such as a camera flash hits them.

Apart from the equipment variations between studies, gait analysis utilizes a huge range of specific methodologies and environments. One laboratory may have a 10 m level walkway, while another has a 50 m walkway or a treadmill. If the entire walkway was to be used during gait analysis, the results from each laboratory would be influenced differently by factors including fatigue, acceleration/deceleration times. For example, a treadmill may enable acceleration/deceleration times to be eliminated by not recording the start and finish, whereas a short walkway may not provide sufficient space. The short walkway however may have less impact on fatigue due to ease of taking breaks after each walking trial. A treadmill also provides easy standardization of walking speed across participant, however, if a study is interested in a person’s self-selected walking speed this becomes very difficult to implement on a treadmill.

Other variations of more challenging activities could include different types of “irregular surface,” different dimensioned staircases, and so on. Even when considering data from a single laboratory, methodologies may vary, for example: 2 different studies of gait, using the same equipment, looking at the same kinematic variables, could potentially utilize different kinematic marker sets (eg, a study utilizing a piece of equipment that would interfere with a specific marker set, it would alter the marker set to accommodate the need for both types of data as demonstrated by comparison of the 2 exemplar marker sets shown in Figure 3). Without a definitive “gold standard” kinematic marker sets are designed for specific purposes, and each have individual benefits and limitations based on marker positions on the body, calculations utilized to generate kinematics and kinetics, and so on. Guidelines would be required therefore for how all data were collected as well as considering what equipment was utilized for it to be suitable for inclusion in the database.

OPEN IN VIEWER

Figure 3.

Anterior views of 2 example marker sets that may be utilized by video-based 3D gait analyses. A Helen-Hayes marker set (left) is a commonly used clinical marker set with the minimal number of markers for 3D kinematics. An alternative marker set (right) shows how a marker set may be altered if the knee is obscured and no markers can be placed on them (eg, when a knee brace is in use). This alternative marker set employs different calculations to track the knee joint than the Helen-Hayes model, utilizing relative positions of the shank and thigh segments tracked independently by the 3 markers on each, rather than utilizing a shared marker placed on the joint as is found in the Helen-Hayes marker set. This difference in methodologies results in different levels of systematic error between marker sets for the same kinematic measure.

Conclusion

We have presented a brief overview of the importance of gait analyses within the population of patients with diabetes and some of the potential benefits, challenges and the utility of a gait database. Mobility is critical to quality of life, and people with diabetes face numerous challenges to the physiological systems responsible for the control and application of movement. Current gait studies have provided useful insights into how diabetes influences movement for a range of activities experienced during daily life, and has allowed the development of diagnostic techniques and interventional therapies to address ulceration and fall risk as well as improving mobility. However, due to the large variety of complications that may be present alongside diabetes, it can be difficult to determine the causal factor within small studies.

A gait database would allow for more extensive investigations to be performed by expanding the potential study population size and enabling multiple influencing factors to be simultaneously analyzed. We have discussed some of the numerous considerations that would require addressing to create a collaborative gait database and while these would require appropriate consideration, the possibility of creating a database is real. Within this article, however, we have not had the scope to discuss each consideration in full; what is clear however, is that a gait database within a single organization would provide immense simplification for the issues of managing and standardizing the data contributing to a database. Therefore, while a gait database would have considerable benefits, the complexity of creating such a database makes it more suitable for consideration within a single organization such as a hospital or university.