Exergaming (XBOX Kinect™) versus traditional gym-based exercise for postural control, flow and technology acceptance in healthy adults: a randomised controlled trial

A prospective, randomized controlled two-arm trial design: Group 1) exergaming with XBOX Kinect™ and Group 2) traditional gym-based exercise (TGB). All testing was carried out by the first author who was not blind to participant allocation.

Ethical approval was granted by Teesside University Research Governance and Ethics Committee. All testing took place at Teesside University physiotherapy laboratory. Convenience sampling was used to recruit active, XBOX Kinect™ naive adult participants. Inclusion criteria: male or female, aged 18–50 years, physically active (three or more moderate-vigorous physical activity sessions per week [22]), free from injury (no musculoskeletal injuries or neurological conditions) and able to take part in four weeks of exercise. Exclusion criteria: unable to give informed consent and/or to comprehend and write English, current (or history of) any medical condition or injury which would contraindicate participation, allergy to alcohol wipes and/or adhesive tape and previous experience of using the XBOX Kinect™.

After written informed consent, demographic information and baseline outcome data had been collected prior to participants getting randomly allocated to a group by blind card randomization. Both intervention groups were then introduced to the allocated programme, undertaking each of the programme-specific exercises, following which they were asked to complete both the UTAUT and FSS questionnaires. In both groups participants undertook four weeks of allocated exercise, three times per week for 30 min each session. All exercises were completed on a one-to-one basis with the first author supervising the sessions and exercising with the TGB group.

TGB exercises matched the movement patterns and physiological demands of exergaming. The XBOX Kinect™ (Redmond, WA, USA), group played six games (see Table 1 for full description) from Kinect™ Adventures™ (Reflex Ridge and River Rush) and Kinect Sports™ (Boxing). Those in the TGB group performed exercises that were matched for sequence, intensity, duration and mode of exercise by adopting open and closed kinetic chain movements, in the same range and loading as required in the Kinect™ group. For both groups intensity was increased each week staring from the second week by adding 1 kg weights leg weights, increasing to 2 kg leg weights in week 3 and 2 kg leg weight plus 1 kg wrist weights in week 4. The weight loads are relatively light and used to increase energy expenditure, the reason for low weights <50 % of body weight is to familiarise those who are not used to exercising with weights and to allow full body movements [23]. Progression was made in each exercise group depending on proficiency of movement and ability to match the skill level of the game (Kinect group only) and by the first author exercising with the TGB group. Balance data were collected at baseline and after week four (12th session).

The outcome measures were postural sway (Kistler™ force plate), heart rate (HR), technology acceptance (UTAUT questionnaire [19]) and flow experience (FSS questionnaire [20]) recorded at baseline and after the four week intervention.

Postural sway was measured using a portable Kistler™ force platform (Model 9286AA, W 40 × L 60 × H 3.5 cm) with a sampling rate of 1000 Hz [24]. Participants stood barefoot on the Kistler™ force plate, and looked directly ahead at a visual target (black circle) positioned 3 m from the centre of the force plate at eye height [25‐27]. Participants were instructed to stand as still as possible with their arms by their side and eyes open [28, 29], on their dominant leg (preferred kicking) for five periods of 30 s. Between trials, participants stepped off the force plate to allow calibration of the equipment which gave a 30 s rest.

Heart rate (HR) was recorded using a Polar™ Heart Rate Monitor™ (FS2C), recording watch and T31 coded chest strap (Polar Electro, Oy, Finland). Mean HR was collected at the end of every exercise session and calculated as a percentage of predicted HR max (220 - age). For a subjective measure of physiological cost the BORG Rate of Perceived Exertion (RPE) scale was used [21]. Mean HR and RPE data were recorded in each exercise session. RPE was defined as how hard participants felt their body was working in general based on the physical sensations they may experience during the activity, including increases in HR, respiration, breathing rate, sweating, and muscle fatigue. The 14 point numerical scale is supported by verbal descriptors, where 6 was defined as “no exertion at all”, 11 “light”, 13 “somewhat hard”, 15 “hard (heavy)”, 17 “very hard” and 20 “maximum exertion”. RPE values between 12 and 13 and 14 and 17 equate to “moderate” and “vigorous” intensity exercise respectively (ACSM, [22]). HR and RPE were taken at 3 separate time intervals during each 30 min session (10, 20 and 30 min) and a mean HR and RPE were calculated of each session.

Technology acceptance was measured using UTAUT which comprised a 7-point Likert scale, with response options on a Likert scale from 1 (strongly disagree) to 7 (strongly agree). The questionnaire has six main domains, performance expectancy (PE; system will help performance), Effort Expectancy (EE; ease of using system), Social Influence (SI; degree in which others believe they should use system), Facilitating Conditions (FC; support in using the system), Self-efficacy (SE; confidence in using the system) and Behavioural Intention (BI; intention to use the system again). The questionnaire was adapted for exergaming and TGB, as previously used in technology settings and validated [14].

Flow state scale (FSS) developed by Jackson and Marsh [20] assessed participants’ level of flow experience into the exercise using exergaming and TGB. Previously this questionnaire has been used in art and science [30, 31], music [32], sport [33, 34], exercise performance [35, 36], gaming [37] and human-computer interaction [38, 39]. The questionnaire consist of a 36-item questionnaire with nine subscales and response options on a Likert scale from 1 (strongly disagree) to 5 (strongly agree). Dimensions of flow include challenge-skill balance (CB; skills match the task and will be successful); clear goals (CG; experience of having a pre-set goal which on is aiming to achieve); unambiguous feedback (UF; feedback on performance); concentration of task (CT; focused on task); paradox of control (PC; performs task with ease); action-awareness-merging (AM; automatic response to task); transformation of time (TT; time speeds up or slows down during activity); loss of self-consciousness (LS; immersed in task) and autotelic experience (AE; activity intrinsically rewarding).

After completing the four-week exercise intervention all participants repeated baseline measures (postural sway, UTAUT and FSS). During all exercise mean HR and RPE were recorded in order to assess physiological cost. All questionnaires were completed unsupervised by the researcher to avoid bias.

Exergaming was performed using the Microsoft XBOX Kinect™ (Redmond, WA, USA), this consists of a Kinect sensor (Kinect head - the rectangular part - W110 × D25 × H15 mm) and the base (W30 × D30 × H15 mm) the system does not require any hand-held controller as it relies infrared sensors to detect and track participants’ movements which are used to generate an avatar which is projected in real time into the gaming environment which is displayed on a widescreen plasma screen (37, Hannspree, Type T73B, Greyenstraat 65, Venlo, Netherlands).

Range and standard deviation of the Centre of Pressure (CoP) displacements in the anterior-posterior (_AP) and medio-lateral (_ML) directions (CoP_AP SD, CoP_AP range, CoP_ML SD, CoP_ML all mm) and the resultant CoP velocity (mm.sec⁻¹) [25, 40] were extracted during unipedal standing assessments using Bioware software (Kistler™) CoP velocity was calculated using previous methods Raymakers [40] after low-pass filtering of the raw data at 10 Hz.

Data were analysed with Statistical Package for the Social Sciences Version 20 for Windows (SPSS™, Chicago, IL, USA). Analysis of covariance (ANCOVA) was conducted on each of the outcome measures, comparing the post-test differences between the groups, with baseline values comprising the covariate. Mixed analysis of variance (ANOVA) was performed to analyse the effect of treatment over time between groups. All analysis used a significance level of 0.05.

An ANCOVA revealed a significant main effect of exercise type for ML range, (F [1, 42] = 8.63, p = 0.005, ε² = 0.17), in favor of exergaming with better postural control. In an analysis of the effect of treatment over time between groups, a mixed ANOVA showed statistically significant differences in favor of exergaming, in the magnitude of change observed (pre-post intervention) for ML SD (F [1, 44] = 5.77, p = 0.02, ε² = 0.12), ML Range (F [1, 44] = 6.15, p = 0.02, ε² = 0.13) and CoP velocity (F [1, 44] = 10.47, p = 0.002, ε² = 0.20) (see also Table 3). A significant interaction effect between time and exercise was found for ML SD (F [1, 44] = 4.62, p = 0.04, ε² = 0.10) in favour of exergaming, as mean (SD) values for ML SD decreased more with a reduction from 6.12 (1.52) mm to 5.33 (1.14) mm than for TGB, from 5.55 (1.40) mm to 5.51 (0.78). A significant interaction effect between time and exercise was also established for ML Range (F [1, 44] = 4.75, p = 0.04, ε² = 0.10) also in favour of exergaming, with a larger reduction in ML range over time from 33.67 (9.11) mm to 28.23 (5.74) mm than with TGB 31.85 (10.02) mm to 31.48 (4.44) mm. No other significant differences were established for any of the other postural-control variables.

In order to assess success of matching exercise intensity, across groups, mean HR and RPE was compared. ANCOVA showed no statistically significant post-intervention differences for HR, indicating that intensity of the exercise was matched between groups. Despite no difference in mean HR, there was a statistically significant difference between the groups for mean RPE (F [1, 44] = 12.30, p = 0.001, ε² = 0.23). The Kinect™ group perceived less physical exertion than the TGB group. Mixed ANOVA showed that over time there was a significant increase in mean HR (F [1, 44] = 126.97, p < 0.001, ε² = 0.75) in both exercise groups (Kinect and TGB). This would be expected owing to increase in intensity of the exercise and therefore elicit physiological responses to exercise intensity. The same occurred in relation to RPE: over time this significantly increased, (F [1, 44] = 452.9, p < 0.001, ε² = 0.91) in both exercise groups (See Table 3).

We observed significant reductions in sway in CoP_ML SD, CoP_ML range, and CoPv, all in favour of the Kinect™, indicating better postural control. In relation to between-group differences, there was a significant difference between groups for CoP_ML range, again with greater reductions for the Kinect™ exercise group. Moreover, the statistically significant differences between the exercise groups were substantial (effect size: 0.10 < ε² < 0.20). While these differences were statistically significant, it is unclear to the authors what level of improvement would be clinically meaningful, and so inferences about clinical relevance should be made with caution. This support earlier work from Vernadakis et al. [16] who found in previosuly injured young males that Kinect™ improved postural stability over a 10-week training program. It should be noted, however, that their physiotherapy group also improved over time and the only improvement between groups occurred between both the physiotherapy group and Kinect™ group, and the control group with no exercise. These results though are encourgaing for the use of the Kinect™ in balance-training and would indicate the potential for use of the Kinect™ for balance-training in young healthy and older adults.

In relation to young healthy adults our results also reflect those of Brumels et al. [14] who found significant improvements in postural control in favour of exergaming (Wii™ and Dance Dance Revolution) compared to traditional exercise. Again, it should be noted that their traditional exercise group did significantly improve on the star excursion balance test (SEBT) compared to their exergaming groups; this, however, may have been a learning effect as their traditional exercise group practiced the SEBT during their exercise intervention. It is encouraging that our results are comparable with other literature that found a greater improvement in the exergaiming group over traditional or TGB exercise as a promising new method of training postural control using interactive technology. A possible reason why we found improvements in the ML direction between exercise groups could be the immersive and purposeful nature of the exergaming compared to TGB exercise: while we aimed to match the movements in both groups, it may be that the exergaming group used more rapid movements and movements outwith the base of support to control the avatar moving in a medial-lateral direction.

To our knowledge this is the first study to report a comparison of exergaming with a demonstrably matched exercise programme and hence, isolated, with more confidence, any effects of exergaming.

Both exercise groups produced moderate-vigorous levels of physical activity in accordance with the American College of Sports Medicine Guidelines (ACSM [22]) - moderate intensity exercise HR should be 64–77 % of max HR, and vigorous (hard) intensity 77–94 % max HR. Mean HR for our Kinect™ exercise group ranged from 61 to 82 % across the four weeks of exercise and the TGB ranged from 61 to 78 % of HR max. These results broadly concur with O’Donovan and Hussey [41] who found that playing the Nintendo Wii™ in young healthy adults that Wii™ fit Jogging elicited 71 % of max HR, yet Wii™ boxing, baseball and tennis achieved less than 60 % max HR. These results suggest that games requiring more dynamic movement (jogging and Kinect games) demand greater physiological response and energy expenditure. This also reflects Barry et al. [42] who found that young males achieved moderate intensity exercise [64–72 % HRmax] for exergaming using the XBOX Kinect™. The results indicate that moderate intensity exercise is achievable using the XBOX Kinect in young healthy adults.

No significant difference was found between groups for mean HR, yet, RPE level was significantly and substantially lower in the exergaming group (effect size ε² = 0.23). This reduced perception of effort could well be attributed to the immersive nature of exergaming. Vernadakis et al. [16] supported the notion that people playing the Kinect™ experienced a greater level of enjoyment than traditional physiotherapy, which would support that explanation. Similarly, in a study of the Nintendo Wii™, Brumles et al. [14] reported that those in the Wii™ and DDR group perceived the exercise to be less streneous and more enjoyable than traditional exercise.

This was the first study to apply a modified version of the UTAUT accross exergaming and TGB. The results showed that, in comparison to TGB, social influence, performance expectancy and behavioural intention were significantly higher in the exergaming group. In essence, particpants exergaming believed that the exercises they were doing would improve their balance (performance expectancy) and they would continue to do those exercises (behavioural intention) and that their peers would encourage this method of exercise (social influence) more so than those doing TGB. These results are encouraging as there is a distinct lack of research regarding acceptance of exergames as a means of exercise. Only one other study to date, Robinson et al. [7] applied the UTAUT in a group of people with MS using the Nintendo Wii™ and standard balance exercise. Their results showed that measures of performance expectancy and effort expectancy were higher in the Nintendo™ group.

In the current study we found significantly higher scores in the exergaming group compared to the TGB in six of the nine subscales post-intervention. This indicates that those in the exergaming group were more focused on what they had to achieve (concentrated on task), felt at ease in playing the game (paradox of control), had clear feedback during the game (unambiguous feedback), movements appeared to be automatic (action-awareness merging), and altered their perception of time (transformation of time) and felt immersed in the activity (exergaming, (loss of self-consciousness). Robinson et al. [7] found similar results when appling FSS to exergaming in a group of people with MS using the Nintendo Wii and traditional balance exercise. The results concurred with our study findings, in which five of the nine subscales were significantly higher in the Wii group. A potential reason for these similar findings could be the immersive nature of exergaming when skill sets of the player match the game [37] and the distraction element that is associated with gaming [43].

Although this study adopted a classic RCT design (which is unusual within the published literature on exergaming) there were limitations to the study. The sample size is relatively small and a larger study employing multi-centre recruitment is merited. For logistical reasons the primary researcher had to both deliver the interventions and collect the outcome data. We acknowledge this opens the possibility of bias but the use of objective outcome measures offsets that. Interestingly, the control findings showed no improvement in balance in any of the parameters. This highlights a limitation of the transferability of the findings to clinical situations. The control group exercises had a degree of clinical artificiality - they would not be used directly as a clinical programme to improve balance. They were designed specifically, and served that purpose, as an experimental control for the exergaming exercises to provide insight into the effects of the immersive aspects of exergaming. A challenge for future work is to design a study that provides a fair comparison between clinically valid balance exercise and a similar programme carried out within an exergaming environment. We would also propose that care be taken translating the results of supervised lab based exercise to clinical practice where a community, self-directed approach is preferred. As a result we would recommend future work examining home-based exergaming and more RCT designs with larger sample sizes.