*4.5. Results*

The effects of the different dietary treatments on biopsy-derived muscle glycogen concentrations, MuscleSound® scores and estimated fuel level points, and the influence of placebo and creatine supplementation are summarised in Figure 2. These data represent measurements taken on one leg using the established protocol to site sequential muscle biopsies (B1–B4), with the MuscleSound® score being taken on the contra-lateral leg at the corresponding site (U1–U4). There was a significant main effect for the different dietary treatments on biopsy-derived muscle glycogen concentrations (F(3,29) = 61.2, *p* < 0.001). Values of biopsy-derived glycogen concentrations for glycogen depletion were lower than baseline, CHO loaded, and normal conditions (*p* < 0.001), while values for carbohydrate loading were significantly greater than normal (*p* = 0.013). Since there were no differences in muscle glycogen between the creatine and placebo groups (F(1,10) = 0.1; *p* = 0.760), a combined mean value for the results for each treatment was derived. Nevertheless, creatine dry weight was a significant variable within the model, indicating that higher creatine dry weight values were associated with increased muscle glycogen content (F(1,33) = 8.6; *p* = 0.006).

There were no differences in MuscleSound® scores between dietary treatments (F(3,27) = 1.1; *p* = 0.384) or between the placebo and creatine groups (F(1,10) = 0.3; *p* = 0.627). Furthermore, no statistically significant differences between dietary treatments (F(3,28) = 1.1; *p* = 0.352) or between the placebo and creatine groups (F(1,10) = 0.2; *p* = 0.701) were evident for the estimated fuel level points. Finally, creatine dry weight was not associated with either MuscleSound® (F(1,28) = 2.04; *p* = 0.165) or estimated fuel level points (F(1,37) = 3.07; *p* = 0.088). An ICC of −0.75 (95% CI −0.85, −0.59) was apparent in the relationship between biopsy-derived muscle glycogen content and MuscleSound® scores, with a similarly

unclear relationship between biopsy−derived muscle glycogen content and estimated fuel level points (ICC of −0.72 (95% CI −0.83, −0.55)). The estimated fuel level points were quantitatively higher than the MuscleSound® score, reflecting an amplification of the original values from an absolute value to a relative range.

**Figure 2.** Biopsy-derived muscle glycogen concentrations (**A**,**B**) MuscleSound® scores in arbitrary units (**C**,**D**) and estimated fuel level points (**E**,**F**) for the creatine and placebo conditions during each dietary treatment: baseline, glycogen depleted, normal and carbohydrate loaded. Values are mean ± SD in panels A, C and E with individual data shown in panels B, D and F. \* Indicates a significant difference to depleted. # Indicates a significant difference to loaded.

There were no differences between the MuscleSound® score values collected from the site on the contra-lateral leg corresponding to the biopsy site (U1–U4) and the measurements taken on a static site of the leg (U1) for each treatment. Indeed, there was a significant correlation (*r* = 0.87 (95% CI: 0.78–0.93); *p* < 0.001) between the values from the two different sites (Figure 3A). Likewise, there were no between-site differences in the values of estimated fuel level points for each treatment, but the correlation between these values was lower (Figure 3B; (*r* = 0.63 (0.42–0.78); *p* < 0.001)). A separate examination of the results of MuscleSound® scores taken at the same site (U1) showed a small but significant (*p* = 0.024) difference between depleted (53 ± 13 a.u.) and loaded treatments (57 ± 10 a.u.). However, these did not differ from baseline (56 ± 3 a.u.) or normal (57 ±12 a.u.) values. Furthermore, the numerical difference between the mean values was smaller than the gradation (bands of 5 a.u.) between sequential results. The estimated fuel level points mirrored these outcomes with an increased spread in both the mean values and SD. Differences were detected between depleted (59 ± 30) and loaded (91 ± 14) treatments (*p* = 0.026), but neither of these differed from the baseline (78 ± 19) or normal (77 ± 22) values.

**Figure 3.** Correlation between values for MuscleSound® score (**A**) and the estimated fuel level (**B**) measured at the same site (U1) and at the shifting site (U11–U4) corresponding to the muscle biopsy.

#### **5. The Literature Involving MuscleSound® Assessment of Muscle Glycogen**

A summary of the available literature in which MuscleSound® technology was used to assess changes in muscle glycogen content resulting from dietary and exercise interventions is provided in Table 1. This includes the validation studies, which originally introduced the use of MuscleSound® as a proxy for biopsy-derived measures of muscle glycogen [4,5], two other data sets in which muscle glycogen content and its changes were assessed by ultrasound and chemical protocols [9], the Bone data presented here, and a recently published study in which MuscleSound® alone was used to assess changes in muscle glycogen content over an exercise session [10]. A final paper, involving the use of the newer estimated fuel level metric, was not included in this table due to differences in its focus and methodology, but it is included in the discussions.



M, male; a.u., arbitrary units; Bx, biopsy; U, Ultrasound scan; ASIS, anterior superior iliac spine; @, at; CHO, carbohydrates; VO max, maximal oxygen capacity; NA, not available; PPO, 2 peak power output; NS, not significant; and EFL, estimated fuel level; \* All biopsy-derived glycogen values presented as mmol/kg dry weight (dw), with conversion from mmol/L wet weight (ww) involving multiplication by 4.28 [34].

Although laboratory-based cycling protocols represent the most frequently investigated mode of exercise, several studies have included real-world competition involving field-based team sports (see Table 1). Dietary manipulations include low, moderate, and high CHO intakes, as well as creatine loading. Muscle and body water content, although not directly measured in any of these studies, is likely to be altered by the acute effects of exercise as well as exercise-associated dehydration. Although vastus lateralis was the muscle investigated in the majority of studies, differences in study protocols around the MuscleSound® assessment included muscle tension (relaxed vs. contracted), whether the same or contra-lateral leg was used between or within glycogen-assessment protocols, whether the scan was meant to represent the same or a related muscle site, and how many scans were used to derive the MuscleSound® outcome.

The first two publications involving MuscleSound® were designed to directly validate its use for indirect assessment of muscle glycogen concentrations, measuring glycogen content before and after a 90 min steady-state [4] or ~158 min time-trial cycling protocol [5] at the same or a similar site in the chosen muscle. In both studies, the ultrasound scan and subsequent biopsy were undertaken at the same site, with the ultrasound being conducted first, followed by the collection of the biopsy, guided by the ultra-sound. In the first study [4], one leg was used for the pre-exercise assessment, while the contra-lateral leg was used in the same manner for the post-exercise assessment to avoid the effect of the muscle biopsy on subsequent glycogen storage at that muscle site [19]. In the second study, the same leg was used for both assessments, but the second biopsy was taken at a site 2 cm from the first; this is sufficient to avoid the effects of such muscle damage on glycogen content, at least by the biopsy technique [19]. With the longer cycling protocol, Nieman et al. reported significant correlations between the two measurement techniques for pre- (0.92, *p* < 0.001), post- (0.90, *p* < 0.001), and exercise-associated changes (0.92, *p* < 0.001) in glycogen concentrations in the vastus lateralis muscle [5]. Here, the chemical method showed a reduction in muscle glycogen content by 77 ± 17%, representing an absolute change of ~71 mmol/kg ww (~306 mmol/kg dw) glycogen; the absolute scores on the MuscleSound® 0–100 a.u. rating were not provided [5]. These data represent a more practical and representative examination of glycogen utilisation during a prolonged endurance sport than the earlier study of Hill and San Millan [4], which employed a 90 min cycling protocol and biopsy collection from the infrequently studied rectus femoris muscle. Indeed, in the earlier study, absolute glycogen values achieved by the dietary preparation protocol and their subsequent utilisation during exercise were lower, with muscle glycogen being reduced by 36% according to chemical analysis and a MuscleSound® change score of ~60 to ~40 a.u. (33% decrease). Nevertheless, correlations between the chemical and ultrasound-mediated assessments of muscle glycogen concentration had pre- (0.92, *p* < 0.001), post- (0.90, *p* < 0.001), and exercise-associated changes (0.92, *p* < 0.001) [4].

In contrast to these earlier reports, an investigation of two separate exercise scenarios by another research group failed to find consistency between the MuscleSound® scores and biopsy-derived assessments of muscle glycogen changes due to exercise and diet [9]. In these studies, which involved cycling and a rugby league match, measurements were made on the same leg, with the biopsy sites 2 cm apart [9]. Although the muscle biopsy protocol identified a ~40% reduction in glycogen content as a result of match play in a real-world rugby league competition (pre-game: 443 ± 65 and post-game: 271 ± 94 mmol/kg dry weight (dw), *p* < 0.001), there were no changes in the MuscleSound ® scores (47 ± 6 vs. 49 ± 7, *p* = 0.4).

A separate study, involving a cycling protocol, was undertaken to remove any potential confounding effects associated with the characteristics of rugby play (i.e., intermittent nature and the magnitude of the muscle contractile forces) that might interfere with the ultrasound image and explain the discrepant results. This second investigation involved an exercise-depletion protocol after which either a low carbohydrate diet or a carbohydrate loading regimen was implemented for 36 h [9]. Although biopsy-derived muscle glycogen concentrations after the carbohydrate loading diet were more than doubled in

comparison to 36 h of low carbohydrate recovery (~531 vs. 252 mmol/kg dw, Table 1), there were no differences (*p* = 0.9) in corresponding MuscleSound® scores (56 ± 7 vs. 54 ± 6 a.u.). In summary, two separate studies of different types of exercise failed to find significant correlations between changes in muscle glycogen concentration and changes in MuscleSound® scores, and, in both protocols, the ultrasound results failed to detect what could be considered predicable changes in glycogen stores.

The results of the Bone study, presented in this paper, are in agreemen<sup>t</sup> with the latter two datasets in finding that the MuscleSound® technique failed to provide meaningful information about muscle glycogen concentrations in athletes. The mean values for muscle glycogen derived from chemical analysis of mixed muscle samples showed larger ranges than reported in the comparative literature, with pre-exercise values after a glycogen loading technique of ~730 mmol/kg wet weight (ww) and a post-exercise reduction of ~364 mmol/kg ww. These values reflect the more aggressive CHO loading regimen and the demanding nature of the exercise protocol. Despite a greater opportunity to detect differences in muscle glycogen, we found that the original MuscleSound® technique generally failed to track the results achieved by chemical analysis of mixed muscle biopsy samples across a range of diet and exercise manipulations, and failed to show the expected significant changes in glycogen concentrations. Individual data showed a range of responses, both in magnitude and direction, in response to each treatment (Figure 2). The only MuscleSound® comparison that yielded a statistically significant difference involved measurements taken from the same site between the depleted and loaded treatments. However, in the case of the original scoring system, the difference was numerically small (53 ± 13 vs. 57 ± 10 a.u.) and was less than the band (5 a.u.) by which results were provided, rendering it of minimal clinical value. Furthermore, this analysis failed to detect differences between the normal glycogen stores and treatments that either increased or decreased these. The estimated fuel level, an updated MuscleSound® metric representing results relative to the lowest and highest scores for the individual athlete, mirrored these results. Although this metric amplified the numerical value of the original score results, and created a greater difference between the mean values, it also increased the range of the results. Therefore, it failed to change the ability of the protocol to detect differences between most treatments.

Two additional publications, which involved the use of MuscleSound® to investigate changes in muscle glycogen in scenarios of real-life sports without alternative confirmation of glycogen stores, are available. One study [10] involved an investigation of changes in muscle glycogen during a football (soccer) match in a professional American league (Table 1). Players followed their typical nutrition practices before and during the match, while the MuscleSound® technique was used to assess glycogen stores pre- and postgame. From the methodology described in the paper, we assumed this protocol involved the traditional MuscleSound® score technique, albeit with results presented as "points", rather than the new metrics described in the company's position stand [13]. There was no confirmation of these results with an independent chemical measurement of glycogen, nor was the hydration status of the players measured before or after the match. Nevertheless, the study reported a mean decline in MuscleSound® glycogen scores of 20% over the course of the match, with inter-individual ranges of 6% to 44%, and some variability in the size of the pre-game stores. As predicted, but not verified by information on individual workload characteristics of the specific game, the decline in muscle glycogen points was numerically greater in midfield and forward players than defence players, and was lowest in the goal keeper. Although these results appear unremarkable, the authors suggested that the protocol identified players who had not adequately fuelled prior to the game, as well as players who might undertake more aggressive fuelling strategies during the game. Here, we note that if within-game fuelling provides an additional exogenous fuel source as glycogen stores become depleted, rather than substantially changing patterns of glycogen depletion during the match, the pre- and post-measurement of glycogen by any technique may provide confusing results.

The final publication involved the use of MuscleSound® to monitor resting levels of glycogen in U.S. Division 1 collegiate female volleyball players on each morning of a 9 day pre-season training camp [35]. The MuscleSound® information was provided in the form of muscle fuel rating, which, as previously noted, remains unvalidated in a peer-reviewed published format. This investigation focused on bilateral asymmetries in the glycogen stores in the rectus femoris in these athletes prior to each morning's training session. The study reported an increase in muscle fuel ratings from the first to second day, with a sustained elevation over the rest of the camp and a 58% difference (higher level) between ratings for the dominant versus non-dominant leg. Although the temporal changes did not track with the training load over the camp (higher in the first days), the authors noted that no dietary control or assessment was implemented. The difference in fuel ratings between legs was attributed to faster rates of glycogen storage in "the more conditioned" dominant leg. Although endurance-trained muscle is known to have higher resting glycogen stores than non-trained muscle (e.g., 500 vs. 350 mmol/kg dw [36]), it is difficult to imagine that the magnitude of difference between legs within the same well-trained athlete would be as large as reported, albeit with a different assessment metric (muscle fuel rating of 52 vs. 33 points). The authors suggested that bilateral asymmetries in glycogen content in volleyball athletes might be used to assess for injury risk, noting that large asymmetries and bilateral deficits in muscle strength are sometimes linked to injuries in athletes [35]. Although this would be a potentially valuable application, there is presently no validation of either the muscle fuel rating score as a measure of muscle glycogen, whether glycogen utilisation patterns are sufficiently different between limbs across a range of symmetrical and asymmetrical exercise activities detected by any technique, nor whether this is associated with injury risk or patterns.

In summary, evidence supporting the use of ultrasound technology, and particularly the MuscleSound® proprietary technique, as a valid measure of muscle glycogen stores is equivocal. In terms of its use as a research tool, two data sets involving laboratory-based cycling protocols validated a correlation with measurements of the glycogen content of a biopsy-derived mixed muscle sample, providing a measure of muscle glycogen from 0–100 in arbitrary units under controlled conditions. Furthermore, the changes in muscle glycogen stores were in line with the expected outcomes of various diet and exercise protocols. Another data set collected in a field setting provided glycogen score results that were logical, but not independently verified. Three other data sets involving lab and field-based uses, however, conflict with these findings. Two collected in cycling models in controlled laboratory conditions, and another undertaken in a real-life team sport competition, failed to find correlations between the two sources of information on glycogen stores. Most importantly, none of these data sets were able to consistently detect differences in MuscleSound® scores despite supervised manipulations of diet and exercise that are known to achieved substantial changes. In one of these studies, a new technique to present MuscleSound® results, described in a company-issued position stand, and presumed to represent its current commercial application, also failed to detect outcomes that would be predicted by the study interventions. This occurred even when undertaken with standardised protocols (e.g., use a single trained tester, laboratory conditions, and the averaging of five separate scans) that might not be possible under the real-life conditions for which it is promoted. Two major issues around the validity and reliability of the MuscleSound® technique have been identified for discussion.

#### **6. Validity of the MuscleSound® Technique: The Glycogen: Water Ratio**
