Comparative Responses to Squats Completed with Free Weights and an Exoskeleton

Long-term manned spaceflight imparts significant physiological impairments that compromise astronaut health and safety during and after exposure to μg. In-flight exercise countermeasures attempt to limit those impairments. Recent advances in exercise hardware for long-term manned spaceflights include the development of robotic exoskeletons. One such device, from the Institute for Human and Machine Cognition (IHMC), weighs less than 18 kg. The exoskeleton is lighter, smaller, and consumes less power than the Advanced Resistive Exercise Device (ARED) now aboard the International Space Station (ISS) (Carpinelli, 2014; Hargens et al., 2013). Thus the portability and design of the IHMC exoskeleton may be ideally suited to the in-flight exercise hardware requirements for small space vehicles, such as Orion, which is slated to be NASA's first spacecraft to transport humans to Mars (Carpinelli, 2014). The IHMC exoskeleton attaches to a user's torso like a backpack as they stand upright. Two exoskeleton actuators, each of which is oriented parallel to the user's knee joints as they stand upright, impart high-fidelity torque control. Per actuator, one of its ends connects to a frictionless joint at the exoskeleton's base plate, while the other attaches to another frictionless joint near the user's hip. Since the IHMC exoskeleton uses motors to provide resistance, it imposes loads independent of gravity and may serve as in-flight hardware to limit lower body muscle atrophy and strength losses (Bamman et al., 1998; Hargens et al., 2013). In operation, the actuators put a compressive load between a user's feet and torso, which they resist by exerting forces. During exercise repetitions on the exoskeleton, subjects first flex and then extend their knees to mimic a squatting (deep knee bend) motion. A computer controls the actuators in order to provide different levels of resistance. The images in Figure 1 depict squats done on the IHMC exoskeleton.

Figure 1

Recent versions of in-flight resistive exercise hardware, such as the ARED, are typically designed to enable performance of the squat. Prime movers for the squat are the lower body extensors, which are the muscles most prone to atrophy and strength loss from actual and simulated spaceflight (Bamman et al., 1998; Stauber, 1989). It is the lower body extensors that incur the greatest degree of pre-stretch as squat depth (descent) increases, which leads to more motor unit recruitment when the aforementioned muscles shorten during the ascent phase of each exercise repetition. Thus the squat may be particularly efficacious as an in-flight countermeasure to μg-induced atrophy and strength losses incurred by the lower body extensor musculature. The exoskeleton's intended operation is to function like free weight squats done with a traditional barbell, without the customary requirement of gravity to impart resistance. Exoskeleton features similar to those of free weights include its ability to impart concentric and eccentric loads to the knee extensors. This is a concern, since a lack of low level eccentric loading during manned spaceflight was deemed a stimulus for knee extensor atrophy and strength losses (Stauber, 1989). In-flight use of the exoskeleton may address this important concern, but first the degree to which its exercise data are like those for free weights must be established. Only then can the validity of the exoskeleton be assessed, and its future use as in-flight exercise hardware be determined.

For a new or novel device, validity is essential to establish before its use is accepted by the population at large (Keppel et al., 1992). Validity denotes how well new or novel devices elicit similar responses to those derived from equipment deemed a criterion or “gold standard” by industry experts (Keppel et al., 1992). As it pertains to weight training, standard free weight exercises are known to elicit muscle mass and strength gains under ambulatory conditions, as well as attenuate atrophy and strength losses in simulated μg models (Bamman et al., 1998; Fleck and Kraemer, 2014). Our study's purpose was to compare and assess physiological, perceptual, and exercise performance responses derived from squats done on the IHMC exoskeleton to those from free weights. Comparatively, similar responses to those from free weights suggest the exoskeleton exhibits acceptable levels of validity, which may improve the likelihood of its use as in-flight exercise hardware. We hypothesized squats done on the exoskeleton will evoke similar responses as to those produced when the same exercise is done with free weights.

All subjects successfully completed both workouts and none were injured through their project participation. Anthropometric dimensions (mean ± sd) were as follows: height 171 ± 5 cm, mass 70.2 ± 5.0 kg, body fat 22.8 ± 3.4%, hip width 29.3 ± 3.8 cm, hip circumference 83.6 ± 4.4 cm, torso length 40.6 ± 4.5 cm, upper leg length 41.8 ± 5.3 cm, and lower leg length 42.4 ± 2.5 cm. Z score analyses revealed approximately 10% of our total data set were outliers. Most outliers were peak power and peak EMG values produced from the exoskeleton workout; such instantaneous measurements are more prone to elicit outliers than dependent variables that exhibit more stability over time (Caruso et al., 2013; Davidson et al., 2013). EMG, in particular, is a very sensitive measure; despite the similarity of inter-workout EMG preparations to our subjects’ left legs, which included replication of electrode placement over a marked skin site, there is inherently high variability when such data is obtained from dynamic exercise (Davidson et al., 2013). Table 3 and Table 4 each include a column that displays the number of subjects who provided data (with outliers and corresponding values excluded) per dependent variable. Our raw data (mean ± sd; range) from each workout appear in Table 1 and Table 2.

Our results, whereby dependent variables from the exoskeleton workout were compared to the same indices obtained from the free weight exercise bout, appear in Table 3 and Table 4. Table 3 includes metabolic, HR, and RPE results. Table 3 Pearson correlation coefficient results show post-exercise [BLa⁻], all HR, as well as RPE dependent variables produced higher r values and thus the most inter-workout agreement. In contrast, pre-exercise [BLa⁻] and [C], post-exercise [C], and peak O₂ display more inter-workout variability. The variability in post-exercise [C] may be due to time differences at which saliva was obtained. Post-exercise saliva collection was delayed until the mask, worn to assess O₂ uptake, was removed so that a swab could be placed in subjects’ mouths. Since post-exercise collection times between workouts varied slightly, that is a potential source of variability. Table 3 t-test results exhibit non-significant inter-workout differences for [BLa⁻], [C], peak O₂, and some of our HR data. However, Table 3 t-test results also include significant inter-workout differences for post-set 2 HR and RPE, with higher HR data from free weight, yet greater RPE values from exoskeleton workouts. Table 3 Cohen's d results show pre- and post-exercise [BLa⁻] and [C] values produced the smallest standardized inter-workout differences. Yet most Table 3 Cohen's d results include far higher values, indicative of greater inter-workout variability and less similarity for responses obtained from both workouts.

Table 4 displays our accelerometry and EMG results. Table 4 Pearson correlation coefficient results show peak EMG set 1, peak power set 2, and peak forces for sets 3 and 4 each yielded r values of 0.50 or greater. Yet most Table 4 dependent variables exhibited far weaker inter-workout correlations. In contrast to Table 3, our Table 4 t-test results exhibit non-significant inter-workout differences for all its dependent variables. Table 4 Cohen's d results show low standardized differences for peak EMG for sets 3 and 4, as well as peak force for set 4. Yet the majority of our Table 4 Cohen's d results include far higher values, which exhibit greater inter-workout variability and less comparative similarity between data obtained from both workouts.

For our study, the similarity of results from both workouts was used to assess the validity of data provided by the exoskeleton. Validity refers to the extent data from new and established devices are similar (Keppel et al., 1992). Different forms of validity exist. Our study compared the similarity in responses derived from exoskeleton and free weight workouts, and is best described as an examination of convergent validity, which refers to the degree two data sets that in theory should be related, actually are (Keppel et al., 1992; Measurement Validity Types, 2015).

Research examined the ability of in-flight exercise hardware to mitigate muscle mass and strength losses produced by long-term stays aboard the ISS (Smith et al., 2012; Trappe et al., 2009). Despite use of in-flight aerobic hardware 5 days/week at moderate intensities, and concurrent strength training on an interim resistive exercise device (iRED) 3–6 days/week with several lower body resistive exercises, crewmembers experienced significant plantar flexor mass and strength losses after six months on the ISS (Trappe et al., 2009). In addition, there were in-flight changes to muscle fibers associated with a reduced resistance to fatigue (Trappe et al., 2009). It was concluded the ISS should be equipped with hardware that offers a greater mechanical loading stimulus to better attenuate muscle mass and strength losses (Trappe et al., 2009). That conclusion was affirmed by a bone study done on the ISS that compared the merits of different forms of in-flight hardware (Smith et al., 2012). With no crossover, crewmembers were assigned to resistive exercise done with either the iRED or ARED for their 4–6 month stays (Smith et al., 2012). Unlike the iRED, which employed elastic bands for resistance, the ARED uses pneumatic cylinders and flywheels to simulate the manner in which weights are lifted on Earth. Bone losses, measured before and after flights, were best abated in those who exercised on the ARED, which provided comparatively more resistance (Smith et al., 2012).

Ground-based studies also examined the utility of actual and potential in-flight hardware as prospective exercise countermeasures (Beck et al., 2014; Loehr et al., 2011; Rea et al., 2013). A training study compared physiological changes from strength training on the ARED to free weights (Loehr et al., 2011). Assigned to one of two groups with no crossover, subjects did identical workouts, which included the squat exercise, three days/week for 16 weeks. Results showed both groups incurred similar improvements over time. It was concluded physiological changes from ARED workouts were like those of free weights (Loehr et al., 2011). Yet it is important to interpret differences between the current and ground-based ARED results with caution. For instance, the ARED study examined chronic changes in physiology; over time, actual inter-group differences could be assessed. In contrast, the current trial quantified acute changes to dependent variables of a far more instantaneous nature, in which recorded values were attained for only a very brief time period. Thus our dependent variables inherently exhibit more data variability. In addition, the ARED study did not assess convergent validity, but rather compared changes over time in both groups (Loehr et al., 2011).

Little research exists on the convergent validity of exoskeleton data, with the goal of improved in-flight exercise hardware (Beck et al., 2014; Rea et al., 2013). Knee extension and flexion torque data from NASA's X1 exoskeleton were compared to those derived from an isokinetic dynamometer (Beck et al., 2014). Subjects performed one workout on each device. Results showed high levels of agreement for knee extension, but not knee flexion, torques obtained from each device (Beck et al., 2014). It was suggested the X1 could be used to assess lower body muscle strength (Beck et al., 2014). The validity of that data exceeds that of the current study's; yet it is important to note X1 and dynamometer torque values were derived from single joint exercises, unlike our study that entailed squats—a dynamic, multi-joint movement of far greater methodological rigor. Thus the validity of our data is expected to be less than that of the X1 paper based on the exercises examined (Beck et al., 2014).

The achievement of comparatively similar data from both of our workouts was made difficult by numerous factors inherent to the current study. They include performance of a multi-joint, multi-planar movement at relatively high velocities with a novel exoskeleton (Caruso et al., 2013; Caruso et al., 2012; Davidson et al., 2013). Yet the low degree of comparative similarity, as seen in some Table 3 and Table 4 results, is still disconcerting and the reason we did not address our study hypothesis. Due to the limited inter-workout similarity, our exoskeleton data does not exhibit an acceptable degree of convergent validity. This is most likely due to the brief amount of familiarization (a few practice repetitions) with the exoskeleton prior to actual data collection, which is a serious limitation. For our study, the exoskeleton was transported and stayed in the principal investigator's laboratory for a short (~4 days) period of time before it had to be returned to the IHMC. The brevity of the exoskeleton's stay, combined with our subjects’ busy schedules, did not afford them an ideal opportunity to familiarize themselves to squats done with the exoskeleton. Clearly, limited familiarization with a complex dynamic exercise performed on a novel device appears to have impacted our results.

Anecdotal claims from 13 subjects inferred free weight squats were easier. This claim is supported by our RPE data, despite the same number of sets and repetitions, rest periods, and loads used for each workout. Due to their considerable background squatting with a barbell, subjects could perhaps put more effort into repetitions for that exercise mode, which could in part explain the higher HR values seen with free weight squats. In contrast, subjects stated the exoskeleton distributed loads in a manner they were unaccustomed to as they performed repetitions. Differences in load distributions and the resultant kinesthetic and biomechanical changes, as well as limited familiarity with the exoskeleton, likely made those squats comparatively more difficult. This is supported by our post-exercise [BLa⁻] and [C] values, as well as peak EMG data, which show higher mean values from exoskeleton workouts. Higher peak EMG data from the exoskeleton exercise bout, despite generally greater performance-based values from free weight workouts, suggests subjects recruited more motor units for the exercise device they were less familiar with, which concurs with prior research (Jakobsen et al., 2013).

Familiarization requirements are based on the nature of the task and the length of inter-session time intervals (Donovan and Radosevich, 1999). The number of familiarization sessions done by human subjects prior to actual data collection certainly impacts convergent validity results. Less familiarization is required for tasks with a low methodological rigor (Sleivert and Wenger, 1994; Viitasalo et al., 1980); the opposite is true of squats, a complex dynamic motor skill that requires refined patterns of muscle activity executed in a specific sequential fashion (Donovan and Radosevich, 1999; Frost et al., 2012; Lee and Genovese, 1989). Recent work on the required number of familiarization sessions examined exercises less rigorous than the squat, and included dependent variables, such as subjects’ 1RM values, which tend to exhibit less variability than our study's performance-based dependent variables, which is in part due to the speed at which repetitions are performed. Current study repetitions occurred at faster velocities than are generally seen with 1RM attempts, which lead to higher rates of movement and more data variability (Caruso et al., 2013; Caruso et al., 2012; Davidson et al., 2013). In exercise studies with low rigor (e.g., vertical jump, isometric contractions, and elastic band exercise), familiarization occurred with a single session administered prior to actual test data collection (Calder and Gabriel, 2007; Colado et al., 2014; Frost et al., 2012), or no familiarization whatsoever for vertical jumps done by physically active men (Moir et al., 2004).

Familiarization requirements were compared among young (23 ± 4 years) and old (66 ± 5 years) women who each performed multiple knee extension 1RM tests (Ploutz-Snyder and Giamis, 2001). They engaged in at least two test sessions; if their 1RM values exceeded the prior sessions’ by more than 1 kg, they were required to perform an additional trial. Older women required more familiarization (8–9 sessions) than younger subjects (3–4 sessions) to achieve consistent 1RM values (Ploutz-Snyder and Giamis, 2001). While more familiarization improves validity, too many sessions may induce a training effect and thus not reflect subjects initial performance capabilities. Two such studies examined the number of familiarization sessions needed to achieve valid 1RM values for three exercises (bench press, squat, and arm curl) in women (Soares-Caldeira et al., 2009) and men (Dias et al., 2005). To derive 1RM values, subjects performed four (Dias et al., 2005) or five (Soares-Caldeira et al., 2009) sessions spaced 2–3 days apart. There was a consistent rise in 1RM values for each exercise over successive trials, which infers multiple tests induced a training stimulus and caused the required number of sessions recommended to be inflated (Dias et al., 2005; Soares-Caldeira et al., 2009). Multiple tests may evoke a 5–10% strength gain, which can be avoided by spacing sessions farther (7–10 days) apart (Schroeder et al., 2007).

Hardware development for in-flight exercise must address the adverse impacts long-term μg has upon human physiology, while conforming to the novel requirements of the spacecraft environment (Carpinelli, 2014; Hargens et al., 2013). Establishment of acceptable levels of convergent validity is essential before exercise hardware can be used as an in-flight countermeasure to muscle atrophy and strength losses. The IHMC hopes to have their exoskeleton flown on future spaceflights. An acceptable level of convergent validity was not established for the IHMC exoskeleton as it was used in the current study. Based on prior outcomes, and given our subjects and their background with the squat exercise, it appears one familiarization session to accustom them to the exoskeleton before actual data collection would have certainly increased our study's convergent validity and improved the likelihood of hypothesis affirmation. However, since crewmembers are generally less fit and do not have as extensive a resistance exercise background as the current subjects, two or more familiarization sessions under ground-based and simulated μg conditions may be warranted if the IHMC exoskeleton becomes in-flight hardware. Due to its novelty and its considerable promise as in-flight hardware, future research studies that involve the IHMC exoskeleton are warranted. Such research should first attempt to establish enough comparatively similarity between data obtained from exoskeleton and free weight workouts to establish acceptable levels of convergent validity.

With ~10% of our total data as outliers, which is usually higher than that seen in other trials, it appears to have foretold the rejection of our hypothesis. The lack of acceptable levels of convergent validity was made difficult by several limitations, which should be addressed in future exoskeleton studies. Some limitations are inherent to squats, which are complex motor skills with a high methodological rigor. A major limitation that prevented acceptable convergent validity, and could address the aforementioned concern with the squat exercise, is the brief amount of familiarization subjects had to the exoskeleton prior to data collection. Depending on the subjects employed and the methods by which data are obtained, future studies that involve exoskeleton squats should include one or more familiarization sessions with the device prior to actual data collection. Another current study limitation may include an order effect whereby due to its brief stay in the principal investigator's laboratory, we initially collected our exoskeleton squat data, followed by the free weight workout, 14 days later. Yet the occurrence of an order effect was likely blunted by the large disparities in the degree of familiarity subjects had with the two types of exercise devices examined. Future trials may wish to vary the sequence that subjects perform workouts in order to reduce the likelihood of an order effect. With respect to our post-exercise [C] results, the convergent validity of that data may have been higher if the time point at which those measurements were obtained was consistent. Since, for our study, we waited until O₂ uptake rates returned to pre-exercise levels in order to remove the mask and subsequently insert the oral swab for saliva collection, this is a potential source of variability. New exoskeleton trials may wish to employ different methods in order to increase the convergent validity of post-exercise [C] measurements. Finally, most of our outliers came from dependent variables that entailed instantaneous measurements, in which recorded values were attained for only a very brief time period. Future studies may wish to assess the convergent validity of the exoskeleton with dependent variables whose values show greater stability over time. Adoption of these recommendations in future exoskeleton trials should increase convergent validity and thus improve the likelihood of its use as in-flight exercise hardware.