Falls are a common mishap among older adults and impose a major burden on older adults and the community at large; it is therefore important to gather evidence on the effectiveness of intervention programs and their design features (e.g., adaptive versus repetitive stepping, high versus low practice intensity). The current study was conceived to examine the efficacy of gait adaptability exercises in intervention programs aimed at improving walking ability and reducing fall incidence. Recently, encouraging first steps in this direction have been taken (e.g., [24, 27, 62, 63]). Weerdesteyn et al. , for example, integrated overground gait adaptability exercises in a 5-week Nijmegen Falls Prevention Program for older adults, and found improved obstacle avoidance success rates, improved balance confidence and reduced fall incidence during one year follow-up. Given that the Nijmegen Falls Prevention Program consisted of overground walking, and hence practice intensity was presumably fairly low (i.e., number of steps taken per session), the observed beneficial effects are encouraging as they may be enhanced even further by increasing the practice intensity of gait adaptability exercises, such as with our C-Mill based gait adaptability training protocol.
In another encouraging study, Mirelman et al.  exploited the potential of virtual reality (VR) to practice step adjustments while walking on a treadmill. Participants’ feet positions were fed back in a virtual environment (i.e., a path with obstacles and targets) presented on a monitor in front of them. After 6 weeks of VR treadmill training, improved obstacle avoidance behavior, higher walking speed while performing a dual task and improved executive function (as assessed with the TMT) were observed for people with Parkinson’s disease. Also for people in the chronic phase after stroke, beneficial effects of VR treadmill training, using a head-mounted display  or a large screen in front of the treadmill , have been reported. All of these studies indicated that gait adaptability training with VR may help improve walking ability and reduce fall incidence in populations prone to falling. Note that C-Mill gait adaptability training is distinct from VR gait adaptability training in that C-Mill training elicits gait adjustments relative to visual context presented in the real environment, whereas with virtual reality training gait adjustments in the real world are detached from the context presented in the virtual environment. Given that gait adjustment is tied to task-relevant visual context in the VR environment only, VR gait adaptability training fails to utilize the direct visuomotor coupling of walking (e.g., [64–66]), where point of gaze is coupled to future foot placement locations, particularly in the presence of surface irregularities. Moreover, visuomotor control of targeted stepping deteriorates with aging, resulting in less accurate foot placement relative to stepping targets, especially in elderly with a high fall risk [67, 68], but proved to ameliorate with training . These examples illustrate the intimate relation between point of gaze and stepping accuracy relative to environmental context. The direct coupling between gaze and gait is fully exploited in C-Mill gait adaptability treadmill training in which step adjustments are elicited relative to visual context in the real environment (viz. augmented reality) rather than visual context in the virtual environment (viz. virtual reality). C-Mill gait adaptability training may thereby tentatively be even more effective than the abovementioned VR applications to practice step adjustments [27, 62, 63], which lack such a direct coupling.
The expected superior outcome of C-Mill gait adaptability treadmill training relative to conventional treadmill training is further supported by current insights into motor learning in the context of neurorehabilitation , which suggest that a considerable variability during practice enhances long-term training effects and transfer to new tasks and contexts compared to constant and repetitive practice (e.g., conventional treadmill training; [71–73]). Several theories of motor learning stress the importance of variability in training, including Schmidt’s schema theory , and theories of contextual interference  and differential learning . These theoretical approaches have in common that one should exploit rather than eliminate variation in task performance to yield optimal learning effects in terms of retention and transfer. This insight is highly relevant for rehabilitation, which aims at long-lasting and general improvements. Exercises during C-Mill gait adaptability treadmill training, such as obstacle avoidance, visually guided stepping, speeding-up and slowing-down, and the interactive gait adaptability games, promote variable practice, which is further reinforced by varying obstacle size, moment of obstacle presentation, and the amount of variability in visually guided stepping (see Methods, Intervention program; see Additional file 1). This variable practice environment affords deep motor learning in that each step needs to be planned and executed from scratch in order to position it successfully relative to the visual context projected on the belt’s surface (e.g., ).
Gait adjustments to environmental context strongly rely on executive function [16, 17], which comprises multiple cognitive processes including visual scanning, problem solving, planning, and task shifting (cf. the deeper form of motor learning referred to above). Interestingly, previous studies found indications for improved executive function after VR treadmill training with an emphasis on step adjustments , as measured with the Trail Making Test [43–45]. Such training induced changes in executive function are important because reduced executive function has been associated with gait impairments, reduced obstacle avoidance ability and falling [14, 15, 17]. Moreover, executive function is known to decline with age . Considering that visual scanning, problem solving, planning, and task-shifting are integral elements in C-Mill gait adaptability treadmill training (i.e., to secure adequate foot placement relative to the projected environmental context), similar training induced improvements in executive function –which will be quantified by the Trail Making Test– may be observed in the present study for AT compared to CT and UC groups.
The inclusion and exclusion criteria for this study are deliberately quite broadly defined, such that the included patients are likely to be representative of the older population seen in daily clinical practice. This implies that the included patients likely exhibit various co-morbidities, resulting in heterogeneous study groups akin to the general population of older adults. On the one hand this may hamper between-group comparisons, but on the other hand findings may generalize well to mixed populations of older adults, which likely facilitates future implementation of the study’s results.
The use of vitamin D will not be intervened in the current study, implying that participants will not be supplemented with vitamin D as a standard. This resembles daily clinical practice, and may therefore also facilitate future implementation of the study’s results. However, since vitamin D supplementation has been associated with reduced fall rate and risk [5, 7], supplementation will be reported in order to be able to control for vitamin D supplementation afterwards. Likewise, smoking history will be registered, since smoking is likely to adversely affect bone healing, bone mineral density, wound healing and the incidence of hip fractures [77–80].
Some outcome measures of walking ability (i.e., TUG, 10MWT, 10MWTcognitive and 10MWTobstacle) will not be assessed at T0, mainly because pain and exertion will limit the number and content of conductible tests at T0. Comparing outcome measures not assessed at T0 between groups entails the risky assumption that groups do not differ at T0. To test this assumption, POMA, FAC and EMS, which are good indicators of walking ability, will be examined at T0. Results based on outcome measures not assessed at T0 should be interpreted with care. Other points of consideration include the different assessors at T0, T1, T2 and T3, and the different setting of T3, which is administered during a home visit. Although the performed tests are well standardized and have good validity and reliability [35, 39, 42–44, 47, 50, 51, 57], differences in test setting should be kept in mind when interpreting the results. The limitations in the current study mostly concern logistic choices based on clinical constraints, which are handled as well as possible to minimize bias.