Introduction

Match and training demands of athletes can be described by external and internal loads [1]. In particular, internal loads are important to monitor, for example, with respect to neuronal, cardiovascular, metabolic, and hormonal stimuli because they are involved in regulating the gene expression required for all regeneration and adaptation processes [2,3,4]. In team and racquet sports, which are characterized by an intermittent activity profile [5, 6], cardiovascular loads are well investigated [7]. However, less is known about metabolic loads as they are more difficult to examine from a methodological point of view [8]. Standard procedures to assess metabolic loads are based on body temperature, heart rate, oxygen uptake, and lactate measures [3]. More advanced procedures include the measurement of creatine phosphate concentration [9] or the use of chemical isotopes as doubly labeled water [10]. However, these procedures are partially invasive and poorly reproducible [11, 12], difficult to apply during matches or training [3, 13], do not allow real-time monitoring [14], and do not fully and continuously scope the metabolic loads in intermittent sports [14]. Therefore, new methodological approaches to assess metabolic loads in a valid and practical manner for intermittent sports are needed.

In 2005, di Prampero et al. [15] suggested a solution where it is possible to determine the instantaneous metabolic power of accelerated running. Metabolic power describes the amount of energy needed to maintain a constant ATP level [16]. The approach is based on the extrapolation from the external (mechanical) to the internal (metabolic) load, for which two assumptions must be considered: (1) accelerated running on a flat terrain is energetically equivalent to running up a slope at constant speed and (2) the relative energy cost for running is independent of the speed and amounts to approximately 3.6–4.0 J/kg/m [17, 18]. Based on these assumptions, the relative energy cost for accelerated running on a flat terrain can be estimated. The subsequent multiplication with the underlying speed leads to the instantaneous metabolic power in W/kg [19]. The original equations are as follows:

$$\mathrm{EC}=\left(155.4{\mathrm{ES}}^{5}-30.4{\mathrm{ES}}^{4}-43.3{\mathrm{ES}}^{3}+46.3{\mathrm{ES}}^{2}+19.5\mathrm{ES}+3.6\right)\mathrm{EM}$$
(1)
$$P=\mathrm{EC}v$$
(2)

where EC is the energy cost, ES the equivalent slope, 3.6 the relative energy cost for running at constant speed, EM the equivalent mass, P the metabolic power, and v the speed. Knowledge of the metabolic power, calculated from acceleration and speed data, can be of value when investigating energetic match demands and assessing training loads in intermittent sports [19].

Even though the approach was published in 2005 [15], it has recently gained greater interest in research and practice. This may be caused by the development of several player tracking technologies, such as global (GPS) and local positioning systems (LPS), allowing easy [20] and accurate [21] access to the required acceleration and speed data [20]. However, in order to apply such an innovative metabolic approach into research and practice, it is necessary to clarify its validity on both a physiological and methodological level for which appropriate gold standards are essential [19]. A key aspect for validation purposes of the metabolic power approach is an in-depth understanding of the underlying concept, which has been described by a few narrative reviews [8, 19]. To date, there is no systematic review of the metabolic power approach in intermittent sports that clarifies the validity as the most important quality criterion of quantitative research [22]. In general, its strength and direction are prerequisites for providing trustworthy, consistent, neutral, and practical-applicable evidence [22, 23]. Therefore, the aim of this review was to systematically investigate the validity and point out the evidence of the metabolic power approach for estimating metabolic loads in intermittent team and racquet sports.

Methods

Research Design and Search Strategy

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was applied [24]. The literature search was conducted in English on March 1, 2021, and included the following four databases: PubMed®, Cochrane Library, Web of Science™, and BISp-surf. The search keywords were divided into components using the PICO scheme (P = Population, I = Intervention, C = Comparisons, and O = Outcomes) [24]. The components were as follows: P = Intermittent sports players without diseases or injuries; I = Tracking; C = Metabolic power approach by di Prampero et al. [15]; and O = Metabolic loads. The component for Comparisons (C) was excluded from the search line as it would have resulted in studies authored by di Prampero only. The resulting final search line was applied to all fields of the database search and was as follows: (team sport OR field sport OR racquet sport OR soccer OR football OR hockey OR rugby OR handball OR volleyball OR basketball OR lacrosse OR futsal OR tennis OR table tennis OR badminton) AND (tracking technology OR global positioning system OR local positioning system OR video camera system OR speed OR acceleration OR deceleration) AND (metabolic power OR energy cost OR energy expenditure). The identified entries were downloaded to a citation manager (Clarivate Analytics, EndNote X9.2, London, UK) and duplicates were removed. The remaining studies were transferred to a spreadsheet (Microsoft Office, Excel 2016, Redmond, USA). First, titles and abstracts followed by full texts were screened for eligibility criteria and studies deemed unsuitable were removed. Additionally, a secondary search based on the reference lists of the studies deemed eligible was conducted. All methodological procedures were executed independently by two researchers (JB, AS) and in case of any disagreement a third (MWH) made the decision.

Eligibility Criteria

The eligibility criteria were set and agreed on by both authors. The criteria for screening titles and abstracts were as follows:

  1. (1)

    Published in 2005 or later (as the original metabolic power approach was introduced in 2005);

  2. (2)

    Written in English;

  3. (3)

    Not systematic review; and

  4. (4)

    Topic on intermittent sports, metabolic power, no animals, no diseases, and no injuries.

The criteria for full texts were as follows:

  1. (1)

    Written in English;

  2. (2)

    Topic on the metabolic power approach by di Prampero et al. [15]; and

  3. (3)

    Either a physiological or methodological validation or conceptual study.

Quality Assessment

The quality assessment was implemented using a modified version of the Downs and Black checklist [20]. Briefly, 12 of the original 27 criteria were used. Original questions 5, 8, 9, 13–15, 17, 19, and 21–27 were removed as they were not suitable regarding the purpose of the study. For questions 3, 7, 10–12, and 18 “not applicable” was used as a fourth scoring option. These modifications were conducted especially due to the conceptual studies often being based on a theoretical approach without the inclusion of subjects as well as the absence of statistical analyses. To account for the non-medical purpose of the present review, further modifications were made concerning the terms used as “patient” was replaced with “subject,” “intervention” with “condition” and “treatment” with “testing,” as previously done [25]. The final quality score for a study—after excluding the questions marked as “not applicable”—was expressed as a percentage. Hence, a higher percentage shows a higher quality of a study regarding the applied quality assessment procedure. The rating of the studies was as follows: low (≤ 33.3%); moderate (33.4–66.7%); and high (≥ 66.8%) quality, as previously recommended [26].

Data Extraction

Data of the physiological and methodological validation and conceptual studies were extracted based on the PICO scheme by one researcher (JB). Thereby, the following items were presented (if applicable): (1) P = Type of sport, number of participants, sex, age, playing level, and nationality; (2) I = Setting of the study, tests, and matches; (3) C = Aim or gold standard used for validation; and (4) O = Main results.

Synthesis of Results

To further clarify the results of the data extraction concerning the direction and strength of evidence for the validation studies, a best-evidence synthesis was performed for which previously defined criteria were used (Table 1) [27]. Minor modifications were made regarding the terms used to describe the study quality: “acceptable” was replaced with “moderate” and “borderline” with “low” to account for the quality ratings of the applied Downs and Black checklist.

Table 1 Criteria for the best-evidence synthesis according to Asker et al. [27]
Full size table

Results

Literature Search

Figure 1 shows the results of the literature search. In total, 947 studies were found. After the removal of 167 duplicates, the titles and abstracts of the remaining 780 studies were screened. Out of these, 670 studies were removed because they did not meet the eligibility criteria, leaving 110 studies for screening of the full text. By screening the full texts, another 79 studies were excluded based on the eligibility criteria. No further studies were eligible via the reference lists. Finally, 31 studies were included. Twenty of these studies were validation studies from which 7 and 13 used physiological [28,29,30,31,32,33,34] and methodological approaches [35,36,37,38,39,40,41,42,43,44,45,46,47], respectively. The remaining 11 studies were conceptual studies [3, 8, 16, 17, 19, 48,49,50,51,52,53].

Fig. 1
figure 1

Flowchart of the literature search including the study selection process according to the PRISMA guidelines

Full size image

Quality Assessment

Table 2 presents the results of the quality assessment. The mean quality score of all studies was 78.6%. The corresponding scores for physiological and methodological validation as well as conceptual studies were 77.4, 75.6, and 82.8%, respectively. Questions 3, 6, 7, 8, 9, and 11 were partially “not applicable” for 4 conceptual studies [16, 17, 19, 50]. Questions 1, 2, 4, 10, and 12 were scored “yes” for all studies. Additionally, questions 3 and 11 were scored “yes” for all validation studies and the latter for all applicable conceptual studies. For all validation and applicable conceptual studies, question 8 was rated “unable to determine” and 9 “no.”

Table 2 Quality assessment of the validation and conceptual studies using the Downs and Black checklist
Full size table

Characteristics of the Studies

Table 3 summarizes the characteristics of the 7 physiological validation studies. Soccer was the most investigated sport with 4 studies [28, 30, 32, 34] followed by rugby with 1 study [31]. One study included different team sports (rugby, soccer, hockey, and netball) [33] and 1 study investigated undefined team sports [29]. The sex of the players was not specified in 3 studies [28, 30, 34]. Males [31, 32] and both sexes [29, 33] were studied in 2 studies each. Regarding age, 6 studies examined adult [28, 29, 31,32,33,34] and 1 study investigated youth players [30]. Three studies investigated amateurs [28, 29, 34], whereas professional [30, 32] and university players [31, 33] were examined in 2 studies, respectively.

Table 3 Characteristics and synthesis of the physiological validation studies using the PICO scheme
Full size table

Table 4 presents the 13 methodological validation studies. Soccer was the primary investigated sport including 10 studies [35, 36, 38,39,40,41,42,43,44,45] followed by rugby with 2 studies [37, 47] and hockey with 1 study [46]. Concerning sex, 10 studies did not specify [36,37,38,39,40,41,42,43,44,45], whereas 3 studies investigated males [35, 46, 47]. Adults were examined in 8 studies [35, 37,38,39, 42, 45,46,47] and 5 studies were based on youth players [36, 40, 41, 43, 44]. Regarding the playing level, professional players were investigated by 10 studies [35,36,37,38,39, 41, 42, 45,46,47]. Two studies addressed sub-elite players [43, 44], while 1 study did not specify the playing level [40].

Table 4 Characteristics and synthesis of the methodological validation studies based on the PICO scheme
Full size table

Table 5 shows the characteristics of the 11 conceptual studies. Three studies were based on theoretical analyses and thus did not include any subjects [16, 17, 19]. Two further studies included data sets from other studies, of which 1 study focused on track [8] and the other on soccer [50] with no detailed subject descriptions. Of the remaining 6 studies, soccer was the most investigated sport including 5 studies [3, 48, 49, 52, 53] followed by 1 study on tennis [51]. Regarding sex, males and females were examined in 3 [48, 51, 52] and 2 studies [49, 53], respectively. Sex was not specified in 1 study [3]. Adults were studied in 4 studies [3, 48, 49, 52], while 1 study was on youth players [51]. The age of the subjects was not stated in 1 study [53]. All of these studies examined professional players.

Table 5 Characteristics and synthesis of the conceptual studies based on the PICO scheme
Full size table

Synthesis of Physiological Validation Studies

Table 3 summarizes the 7 physiological validation studies according to the PICO scheme.

Of these 7 studies, the gold standard for validating tracking-based metabolic power was the use of oxygen uptake determined by portable metabolic carts in 6 studies [29,30,31,32,33,34], while lactate parameters were applied in 1 study [28]. As tracking technologies, 4–15 Hz GPS [28,29,30,31, 33], 10 Hz LPS devices [34], and a 25 Hz video camera system [32] were used in 5, 1, and 1 studies, respectively.

Concerning the interventions, 4 studies used constant or shuttle running test protocols until exhaustion [28, 32,33,34]. Specific sports circuits were utilized in 5 studies [28,29,30,31, 33], while 1 study applied official match play data [32].

Comparison between energy expenditure measured by oxygen uptake and GPS derived metabolic power during sport circuits showed lower results for the metabolic power approach in 4 studies [29,30,31, 33]. Conversely, during walking, energy expenditure was higher when measured with the metabolic power approach via GPS compared to oxygen uptake as revealed in 1 study [29]. The energy cost by LPS derived metabolic power was higher in constant and lower in shuttle running compared to energy cost via oxygen uptake as shown in 1 study [34]. Concerning the correlation between GPS or video camera system derived metabolic power and oxygen uptake, moderate [30, 31] and large [32, 33] relations were found in 2 studies, respectively. Moreover, 1 study showed small to large relationships between GPS derived metabolic power and lactate parameters [28].

Synthesis of Methodological Validation Studies

Table 4 shows the summary of the 13 methodological validation studies based on the PICO scheme.

The most frequently used gold standard for validating the tracking-based metabolic power was the traditional running speed analysis including 8 studies [35,36,37,38,39,40,41, 45]. Further gold standards were the critical speed approach with 2 studies [43, 46] and the use of relative and absolute running speed based on first and second ventilatory thresholds with 1 study [47]. Furthermore, the metabolic power approach was examined regarding intraindividual [42] and match-to-match variability [44] in 1 study each. As tracking technologies, 10–15 Hz GPS devices [36,37,38,39,40,41,42,43,44,45,46,47] and a 25 Hz video camera system [35] were used in 12 and 1 studies, respectively.

Regarding the interventions, official matches were investigated in 10 studies [35,36,37, 41,42,43,44,45,46,47] and small-sided games were inspected in 3 studies [36, 38, 40]. Shuttle running efforts were tested in 2 studies [43, 46]. Training sessions [39], modified matches [40], straight-line running efforts [43], intermittent fitness tests [47], and time trials [46] were applied in 1 study each.

Concerning the validation of the metabolic power approach to the traditional running speed approach, 6 studies showed that the distance covered in high-intensity zones was significantly higher for GPS derived metabolic power than for running speed [36,37,38,39,40,41]. One study found no differences between the two approaches in different intensity zones regarding playing positions [45]. Furthermore, 1 study showed that relative high metabolic threshold was likely to almost certainly be lower than absolute metabolic threshold during rugby matches [47]. A nearly perfect correlation between high intensity running and GPS or video camera system derived high intensity metabolic power distances was found in 2 studies [35, 37]. Comparison to the critical speed approach revealed a very large correlation between critical speed and GPS derived critical metabolic power [46] and a very large correlation of GPS derived critical metabolic power between different field tests [43] in 1 study each. Two further studies discovered a moderate correlation of GPS derived metabolic power between different matches [42, 44].

Synthesis of Conceptual Studies

Table 5 shows the summary of the 11 conceptual studies based on the PICO scheme.

Three studies addressed the metabolic power approach by extending the original equation based on air resistance [17], grass [3], and sand [48]. Additionally, the first study presented a solution to differentiate between running and walking periods [17]. Another study developed a new equation to estimate metabolic power following the original concept of the metabolic power approach [52]. The effects of different terrains were investigated in 5 studies [3, 48, 49, 51, 53], while collisions were examined in 1 study [16]. Furthermore, 1 study focused on the identification of high and low intensity energy bouts [50]. Finally, practical conclusions for the application of the metabolic power approach were given in 2 studies [8, 19].

The studies executing practical approaches included match data with 3 studies [3, 51, 53] as well as sprint and shuttle tests [48], small-sided games [49], treadmill tests to exhaustion [52], aerobic steady-state runs [52], and soccer-specific runs [52] with 1 study each using GPS (5–15 Hz) or a video camera system (25 Hz).

The investigation of 1 study on the effects of air resistance on metabolic power showed that these effects were negligible [17]. However, the same study stated that with the inclusion of walking periods, the former energy expenditure was overestimated. The examination of different terrains, as conducted in 2 studies, showed that running on grass is 1.29 [3] and on sand an additional 1.45 [48] times more energetically demanding compared to running on a treadmill. One study converted the original equation into a new equation, which both showed similar moderate correlations with oxygen uptake. However, there was no significant difference between the new equation and oxygen uptake, while a significant difference was found between the old equation and oxygen uptake [52]. Concerning the influence of different terrains, GPS derived metabolic power was significantly higher on sand compared to grass and artificial turf [48] and on clay compared to hard court [51] as well as significantly lower on ground compared to grass and artificial turf [49] as shown by 1 study, respectively. Another study showed that GPS derived metabolic power distances were higher on artificial than natural turf [53]. To include the impact of collisions, 1 study applied a mechanical work approach combining external and internal workloads based on speed and/or acceleration data [16]. Moreover, 1 study established an equation on the kinetics of oxygen uptake to detect the phases of aerobic and anaerobic energy supply and thereafter, following a 5-step procedure to differentiate between high and low intensity energy bouts [50]. Regarding the practical conclusions of the metabolic power approach, 1 study showed that GPS (20 Hz) derived oxygen uptake was similar to that through portable metabolic carts [8]. However, another study indicated that oxygen uptake and metabolic power cannot simply be compared because oxygen uptake only represents aerobic, while metabolic power represents both aerobic and anaerobic contributions [19]. Finally, the same study emphasized that the metabolic power approach is incapable of estimating overall energy expenditure or mechanical workload.

Synthesis of Results

Tables 6 and 7 show the results of the evidence assessment of the physiological and methodological validation studies, respectively.

Table 6 Results of the physiological validation studies using a best-evidence synthesis
Full size table
Table 7 Results of the methodological validation studies using a best-evidence synthesis
Full size table

Concerning the physiological validity, there was strong evidence that energy expenditure is lower when determined via GPS or LPS derived metabolic power compared to oxygen uptake by portable metabolic carts during intermittent running activities [29,30,31, 33, 34]. The correlation of metabolic power with oxygen uptake showed conflicting evidence as it ranges from small to large [30,31,32,33]. Additionally, the correlation with lactate parameters, which are moderate to large, disclosed moderate evidence [28].

In terms of the methodological validity, strong evidence was shown that energy expenditure via GPS derived metabolic power is higher than that via high-speed running [36,37,38,39,40,41, 45]. Similarly, the correlation with high-speed running presented strong evidence as it is very large to near perfect [35, 37]. Moderate evidence was revealed in terms of the correlation of critical metabolic power with critical speed [46] as well as with relative and absolute running speed [47]. There was moderate evidence concerning the correlation of the metabolic power within critical metabolic power [43]. Finally, the correlation within metabolic power, being moderate, showed strong evidence [42, 44].

Discussion

To the best of our knowledge, this is the first systematic review aimed to present the validity and evidence of the metabolic power approach that was first introduced by di Prampero et al. [15] in 2005 for estimating metabolic loads in intermittent team and racquet sports. Based on the best-evidence synthesis, the main findings were that (1) conflicting to strong evidence was shown concerning the physiological validity and (2) moderate to strong evidence was revealed regarding the methodological validity. Additionally, the conceptual studies showed that (3) the distinction between walking and running episodes, different terrains, as well as aerobic and anaerobic energy supply should be considered when analyzing metabolic power in intermittent sports.

Concerning the characteristics of the 7 physiological validation, 13 methodological validation, and 11 conceptual studies, the most investigated sport was soccer. The subjects were predominantly male adult professional players. For validation studies, the gold standards most commonly used for tracking-based metabolic power were oxygen uptake via portable metabolic carts and traditional running speed analysis. The tracking technology predominately used in these studies was GPS operating between 4 and 15 Hz. While physiological validation mainly focused on sport specific tests, match data were primarily examined in methodological validation studies. Updating the original concept was mostly based on the distinction between running and walking episodes, different terrains as well as aerobic and anaerobic energy supply (Tables 3, 4, and 5). These characteristics show a lack of research concerning children, females, and intermittent sports besides soccer and the application of more profound physiological approaches for the validation and assessment of tracking technology-based metabolic power is needed. This should be considered when planning and conducting future studies.

The first main finding of this study was that the physiological validity of the metabolic power approach showed conflicting to strong evidence (Table 6). However, all studies were high quality (Table 2). Strong evidence was shown that supposedly GPS or LPS derived metabolic power underestimates energy expenditure during team sport specific test protocols compared to that derived via oxygen uptake, especially concerning multidirectional activities [29,30,31, 33, 34]. In contrast, Brown et al. [29] disclosed an overestimation of GPS derived energy expenditure via metabolic power during walking episodes. However, the original and modified approaches by di Prampero et al. [15] and Osgnach et al. [3], respectively, were simply incapable of distinguishing between walking and running episodes. Regarding the update in 2018 [17], this limitation has been solved. Furthermore, certain studies tried validating activities that could not be registered by tracking devices (e.g., collisions, running with the ball, running backward or sideways) [30, 31]. Therefore, the resulting lower energy expenditure of GPS or LPS derived metabolic power is reasonable. When considering the correlation between the metabolic power approach and oxygen uptake, conflicting evidence was shown varying from small to large correlations. This indicates the inconsistency of the results regarding the validation of the metabolic power approach against oxygen uptake, especially considering that, in intermittent running, oxygen uptake only discloses aerobic energy supply, whereas metabolic power contains both aerobic and anaerobic energy supply [19]. Additionally, the inclusion of energy expenditure during passive resting periods when comparing oxygen uptake to the metabolic power approach is inappropriate [50, 54]. As the estimation of energy expenditure via tracking-based metabolic power is based on acceleration and speed, no data can be received during passive resting periods, where oxygen uptake is still elevated to balance the oxygen uptake deficit [50]. A further physiological validation study included the use of lactate parameters [28]. The correlation with metabolic power was moderate to large showing moderate evidence. However, as there was only one study including lactate parameters, oxygen uptake is still most often used when physiologically validating the metabolic power approach, regardless of its limitations when used during intermittent running. Overall, the studies showed that previous physiological validations are outdated as there have been adaptations concerning the metabolic power approach. Additionally, the implementation of the validation was often inappropriate due to the inclusion of, e.g., collisions or resting periods.

The second main finding was that moderate to strong evidence was revealed regarding the methodological validation of the metabolic power approach (Table 7). Three studies were of moderate quality, whereas the remaining studies were of high quality (Table 2). Strong evidence was shown that distances covered at high metabolic power are greater compared to high-speed running during matches or training. However, there were inconsistencies regarding the calculated threshold of high-speed running when comparing with an energetic equivalent metabolic power threshold. While the actual equivalent to 20 W/kg for constant speed running is 15.5 km/h [19], most studies used 14.4 km/h as a threshold [37,38,39, 41]. Consequently, the discrepancy between the two approaches regarding distances covered above these thresholds is likely greater than reported. This indicates that the identification of high-speed running alone does not show the full extent of the contribution to energy expenditure or rather the intensity of matches or training. Indeed, activities such as accelerations during lower speed largely contribute to high metabolic power and are important to monitor [19]. This shows the advantage of the metabolic power over the traditional running speed approach. Additionally, the correlation between high-speed running and high metabolic power was reported as very large to near perfect [35, 37] and thus showed strong evidence. This result was also shown in the correlation between critical speed and critical metabolic power as the relationship was very large [46]. Because there was only one study utilizing these parameters, the evidence was moderate. Concerning the reliability of the metabolic power approach, there was moderate evidence that the approach is very largely correlated within critical metabolic power [43] and good to moderately correlated within metabolic power itself [42, 44], which is a prerequisite for the validity of the approach. Collectively, there is strong evidence that the metabolic power approach is valid from this perspective and, regarding the traditional running speed approach, a superior method to monitor metabolic loads during matches and training in intermittent sports.

The last main finding was that the differences between walking and running episodes, different terrains, as well as between aerobic and anaerobic energy supply should be noted in terms of validation purposes as shown by the conceptual studies (Table 5). As revealed by a physiological validation study discussed above [29], the original metabolic power approach seemed to overestimate energy expenditure in terms of walking episodes compared to energy expenditure via oxygen uptake. However, contrary to walking, in running due to the flight phase between steps, part of the kinetic energy from each step is absorbed by active muscles and tendons and retained as mechanical energy for the next step [55]. Hence, the relative energy cost of running but not walking is independent of speed [18]. Therefore, an additional equation was established to account for walking episodes [17]. Another factor addressed in the conceptual studies is the influence of the underlying terrain. Depending on the surface, the estimated energy expenditure differs [48, 49, 51, 53]. To increase comparability between studies on different terrains, the original metabolic power equation was extended by individual factors regarding the different surfaces. However, only specific correction factors for grass and sand have been developed to date [3, 48]. In this context, the consideration of real surface adjustments based on Clegg hammer measurements to obtain more precision about surface rigidity is logical [56]. Lastly, an adaptation of the metabolic power approach was conducted regarding the separation of aerobic and anaerobic energy supply [50]. When maximum oxygen uptake is known, net oxygen uptake can be simulated as there is a known time delay of approximately 20 s between the oxygen uptake kinetics at the muscle and upper airway. Then, the metabolic origin can be distinguished as either aerobic or anaerobic when metabolic power is lower or higher than the simulated actual oxygen uptake, respectively [50]. In intermittent sports, different ways of energy supply are implicated [57, 58]. Therefore, the knowledge of energy expenditure derived from aerobic or anaerobic supply can provide a more relevant overview of the metabolic load. In addition to these conceptual aspects, the validity of the tracking technology as well as the impact of the sampling rate and filtering techniques to assess and process the required acceleration data should be considered when discussing the validity of the metabolic power approach. In fact, an important prerequisite is to assess valid acceleration data and reduce the noise without losing information for which no established procedures exist as yet [42, 54].

Overall, the metabolic power approach has recently evolved. In particular, the differentiation between walking and running episodes needs to be considered when using the metabolic power approach to, e.g., track matches where characteristically, walking episodes are present between running bouts. Furthermore, to guarantee objectivity and comparability, correction factors for more surfaces, such as clay or various indoor floorings, must be determined. Moreover, it is rational that different footwear [59] and physical capacities [60] have an impact on metabolic power that has not yet been investigated.

This systematic review has a few potential limitations. In line with all systematic reviews, selection bias regarding included studies cannot be completely precluded even though objectivity was improved as all studies were independently rated by two researchers. Additionally, because of the heterogeneity of the included studies, meaningful quantitative analyses such as a meta-analysis could not be implemented. However, as a compromise and strength of this systematic review, a best-evidence synthesis was conducted.

Conclusions

In conclusion, this review shows that several validation studies for the metabolic power approach have been conducted over the last few years. However, especially the physiological validation studies were often partially implemented incorrectly as shown by the in-depth analysis of the underlying concept. Therefore, the described evidence levels should be treated with caution. Nevertheless, the approach is valid from a methodological point of view. Based on these findings and the modification of the concept during recent years, there is a need for further physiological validation studies. Therefore, it must be considered what the metabolic power approach can and cannot actually display. Moreover, there is a need to differentiate the approach in a sport, sub-group, and terrain specific manner as there are different metabolic demands and capacities.