Short Trail Running Race: Beyond the Classic Model for Endurance Running Performance : Medicine & Science in Sports & Exercise

Journal Logo


Short Trail Running Race

Beyond the Classic Model for Endurance Running Performance


Author Information
Medicine & Science in Sports & Exercise 50(3):p 580-588, March 2018. | DOI: 10.1249/MSS.0000000000001467
  • Free



This study aimed to examine the extent to which the classical physiological variables of endurance running performance (maximal oxygen uptake (V˙O2max), %V˙O2max at ventilatory threshold (VT), and running economy (RE)) but also muscle strength factors contribute to short trail running (TR) performance.


A homogeneous group of nine highly trained trail runners performed an official TR race (27 km) and laboratory-based sessions to determine V˙O2max, %V˙O2max at VT, level RE (RE0%) and RE on a +10% slope, maximal voluntary concentric and eccentric knee extension torques, local endurance assessed by a fatigue index (FI), and a time to exhaustion at 87.5% of the velocity associated with V˙O2max. A simple regression method and commonality analysis identifying unique and common coefficients of each independent variable were used to determine the best predictors for the TR race time (dependent variable).


Pearson correlations showed that FI and V˙O2max had the highest correlations (r = 0.91 and r = −0.76, respectively) with TR performance. The other selected variables were not significantly correlated with TR performance. The analysis of unique and common coefficients of relative V˙O2max, %V˙O2max at VT, and RE0% provides a low prediction of TR performance (R2 = 0.48). However, adding FI and RE on a +10% slope (instead of RE0%) markedly improved the predictive power of the model (R2 = 0.98). FI and V˙O2max showed the highest unique (49.8% and 20.4% of total effect, respectively) and common (26.9% of total effect) contributions to the regression equation.


The classic endurance running model does not allow for meaningful prediction of short TR performance. Incorporating more specific factors into TR such as local endurance and gradient-specific RE testing procedures should be considered to better characterize short TR performance.

The International Trail Running Association defines trail running (TR) as “a pedestrian and off-road race conducted in a natural environment (e.g. mountain) with minimal possible paved or asphalt road (<20% of the total duration race).” Classically, TR races are performed on mountain single tracks including positive and negative elevation with repeated technical sections on rocky and root-covered paths. TR profiles may extend from short (<42 km) to ultralong (>100 km) distances. In TR races, in which popularity has markedly increased during the last decade (1,2), the performance (race time) among the best runners usually ranges between 1.5 and 4 h for a short distance race of 20 to 42 km.

The main difference between short TR races and more conventional, on-road, running events (i.e., level road profile) such as the marathon is that TR races are characterized by successive uphill and downhill off-road sections, leading to major changes in physiological and mechanical responses (for review, see Ref. [1]). In such cases, prolonged and intense concentric and eccentric actions occur in lower limb muscle–tendon units during uphill and downhill sections of TR events, respectively (3). The modality of muscle action and the contraction time are specific to TR sections and differ from level road running, which is mainly characterized by repeated and continuous stretch-shortening cycles for lower limb extensors (4). Dewolf et al. (5) recently showed that the classic mechanical model of level running clearly differs from incline conditions. Specifically, during level running, the upward and downward movements of the center of mass are overall equal, as are the positive and negative external work within each step. In contrast, during incline running, the “bouncing” mechanism gradually disappears as speed and slope increase (5). On positive slopes, the step period and the downward movements of the body are reduced, whereas on negative slopes, the step period increases and the upward movement decreases. Major changes in ground reaction forces are also apparent from a steep downhill to a steep uphill: the normal impact force peaks and the parallel braking force peaks decrease, whereas the parallel propulsive force peaks increase (5,6). Therefore, repeated changes in slope and associated mechanical responses during TR likely influence the modality of muscular contraction and metabolic demand (for review, see Refs. [1,7]).

In a recent review, Giandolini et al. (1) reported that central and peripheral mechanisms of muscle fatigue as well as mechanical muscle damage largely contributed to the decline of TR performance. During uphill sections, predominantly concentric muscle contractions induce less mechanical stress and thereby less potential muscle damage. Excitation–contraction failures reported after uphill running (UHR) seem mainly due to the high exercise intensity required (1). In contrast, a marked decline in maximal voluntary contraction (MVC) torque (>−15%) for plantar flexor and/or leg extensor muscles has been reported after treadmill and outdoor downhill running (DHR) exercises (8,9). Although central fatigue might play an important role in the MVC decline, especially in prolonged exercises (1), studies investigating short distance TR or DHR modalities report clear decreases in both central and peripheral fatigue-related parameters (3,8–10). For instance, Vercruyssen et al. (10) recently reported a ~4.5% to 6.5% decrease in quadriceps voluntary activation (i.e., central component) associated with a significant reduction in the low- to high-frequency doublet ratio (i.e., peripheral component) after various 18.6-km TR sessions performed close to race intensity. Thus, the severe muscular alterations induced by isolated downhill sections or simulated TR events contribute to intense and prolonged fatigue and suggest that muscle strength and/or resistance to fatigue is an important factor in the understanding of short TR performance.

Although muscular actions differ between TR and level road running, the duration of short TR races reported in trained runners is comparable with that observed during half-marathon or marathon events (<4 h). For these endurance events, it is classically accepted that the key physiological determinants of performance include maximal oxygen uptake (V˙O2max), the percentage of V˙O2max (%V˙O2max) that can be sustained during running—which has been shown to be closely linked to the lactate threshold (LT)—and running economy (RE; expressed as energy cost), that is, the metabolic energy spent per unit of distance covered (11–16). The relationship between each one of these physiological variables and level running performance has been widely studied. For instance, Costill et al. (14) reported a strong negative correlation (r = −0.91) between V˙O2max and 10-mile race time in runners who varied greatly in V˙O2max. Similarly, LT was also highly correlated (r ≥ 0.91) with performance on running distances ranging from 3.2 to 42.2 km (15). Finally, McLaughlin et al. (16) recently reported a high correlation (r = 0.81) between RE and 16-km time in well-trained distance runners. As a result, these three variables (V˙O2max, %V˙O2max at LT, and RE) have been often used in a classical physiological model of endurance running performance (12,17). Given the specific muscle actions in TR, it would be interesting to know whether these physiological determinants of level running are also related to short TR performance.

As described previously, it is well established that differences in RE explain a great part of the interindividual variability in running performance among athletes with similar V˙O2max values (18). In this regard, improved RE is associated with better marathon performance in world-class marathon athletes, independent of changes in V˙O2max (19). However, to elucidate whether RE plays a similar role in short TR, dissimilar to level road running in terms of terrain and pacing, a specific investigation of the relationship between RE and TR performance is necessary. When focusing on RE responses to slope conditions on a treadmill, Balducci et al. (20) recently reported that level RE was not correlated with RE measured on positive slopes (12.5% and 25%). Conversely, RE values at 12.5% and 25% were well correlated (r = 0.75), suggesting that specific mechanisms are active in determining RE on inclines and preserved throughout different gradients. Interestingly, a recent study focusing on physiological and biomechanical determinants of uphill mountain marathon performance (21) showed that RE was significantly correlated with overall race time and that athletes with smaller changes in RE during the race also had greater maximal lower limb power. This is in line with previous studies that emphasize the importance of lower limb muscle strength and specific strength training (e.g., uphill bouts, resistance) to improve RE and, in turn, running performance (11,21). On the basis of these reports and considering the impairment in muscle strength consistently observed after simulated TR events or races (3,22,23), the implication of factors associated with muscular strength might be expected in short TR performance.

The objective of this study was therefore to identify the physiological determinants of short TR performance using the classic endurance performance model (16) and including specific factors to TR such as local endurance or uphill RE. Given the differences in muscle contraction modalities and race profiles between TR and level road running, we hypothesized that the predictive power of a commonality regression analysis using the classical model of endurance running would increase through inclusion of TR-specific factors (e.g., local endurance, uphill RE) in a homogeneous group of highly trained trail runners.



Nine experienced, high-level male trail runners (age, 39 ± 8 yr; height, 1.73 ± 0.06 m; body mass, 68.4 ± 5.8 kg (mean ± SD)) volunteered to participate in this study. Recruitment was based on the performance level within national and regional short distance TR races, with subjects consistently ranking in the first 20 finishers. Participants had a mean of 8.5 ± 2.1 yr of TR experience and a mean weekly running mileage of 75 ± 6 km completed on 3–5 d·wk−1. Furthermore, on the basis of their training log, subjects also performed a minimum of five cycling sessions per month (~250 to 350 km) including specific uphill bouts (~2000 to 3000 m of cumulated positive elevation). Subjects gave their informed written consent to participate in this study, which was approved by the local ethics committee for the protection of individuals and conducted according to the Declaration of Helsinki.

Experimental design

Experiments were conducted within a 21-d period including laboratory sessions (separated by at least 72 h) and the TR race (Fig. 1). On five separate occasions, each participant completed (i) a maximal test performed on a motorized treadmill with +10% slope (Gymrol; HEF Tecmachine, Andrézieux-Boutheon, France) to determine V˙O2max, velocity associated with V˙O2max (vV˙O2max), maximal heart rate (HRmax), and ventilatory threshold (VT); (ii) a submaximal treadmill running test to measure RE at various velocities and slopes; (iii) muscle performance tests to determine MVC in concentric and eccentric muscle modes and local endurance of the knee extensors; (iv) a treadmill run time to exhaustion; and (v) an official TR race to determine running performance as the total racing time. For all running conditions conducted in laboratory, V˙O2, carbon dioxide production and minute ventilation (E) were analyzed breath-by-breath and every 10 s by using an Oxycon Alpha metabolic measurement cart (Jaeger®, Wuppertal, Germany).

Graphic representation of experimental conditions. TTE, run time to exhaustion.

Maximal running test

Subjects started running on a treadmill at a velocity of 7 km·h−1, which increased by 1 km·h−1 every 2 min until exhaustion. During the maximal test, the slope was fixed at +10% to better represent the characteristics of TR training and racing (10), although it was recently reported that treadmill slope has no effect on the determination of relative V˙O2max in endurance mountain runners (20). According to ACSM’s Guidelines for Exercise Testing and Prescription, V˙O2max was considered to have been achieved if there was no increase (<100 mL·min−1) in V˙O2 with an increase in treadmill speed or if the following criteria occurred at the end of exercise: respiratory exchange ratio of >1.15 and subjects reached their age-predicted HRmax (220 − age). All tests were terminated at volitional exhaustion, and all subjects achieved V˙O2max by these criteria. V˙O2max was averaged using the three highest consecutive values (i.e., during a 30-s interval) reached during the last stage of the maximal running test. For each subject, the 30-s interval enabling the V˙O2max determination was used to identify vV˙O2max. Finally, VT was determined as the point at which an increase in E/V˙O2 was observed with no concomitant increase in E/carbon dioxide production (24).


A standardized 10-min warm-up (5-min level and 5-min UHR at 2.77 and 2.08 m·s−1, respectively) was performed before the RE test. Subsequently, each subject completed, in random order, two 5-min running stages at two different speeds and slopes, with a 5-min rest (in a seated position) between stages: 3.88 m·s−1 (0% grade; RE0%) and 2.5 m·s−1 (+10% grade; RE+10%). V˙O2 values were measured continuously with the metabolic cart during the 5-min running tests and averaged over the final 2 min to calculate RE values. RE at each speed was expressed in units of energy (J·kg−1·m−1) (25). The level speed was determined from pilot testing conducted in three subjects of the present study and fixed at 3.88 m·s−1 to be close to the relative intensity (~80% V˙O2max) reported by McLaughlin et al. (16) in their model of level running performance. Similarly, uphill speed was chosen to reproduce a metabolic intensity close to that expected at level running, but also to reflect the metabolic demand induced during the uphill sections of short distance TR races. Thus, the %V˙O2max reached by subjects during level RE and uphill RE tests was 81.3% ± 6.9% and 82.6% ± 7.5%, respectively.

Muscle strength factors

Subjects were familiarized with all procedures concerning muscle force testing on their first visit to the laboratory. Muscle force characteristics of the right knee extensors were evaluated using (i) maximal voluntary concentric and eccentric torques (MVCCon and MVCEcc, respectively) and (ii) a local endurance test. During these tests, participants were securely strapped into an isokinetic dynamometer (Biodex System 3, Shirley, NY) with a knee joint angle of 90° (full leg extension, 0°) for the assessed (right) leg. The axis of the knee joint was carefully aligned with the rotational axis of the dynamometer, and all settings were kept constant throughout the experiment. Before each MVCCon and MVCEcc, participants warmed up on the isokinetic dynamometer by repeating ten 1-s submaximal concentric or eccentric contractions (1-s rest between contractions). After 2-min rest, two MVCCon or MVCEcc (angular velocity, 60°·s−1) were performed for the full range of motion, each lasting around 4–5 s (55-s rest between attempts). Strong verbal encouragement was given and torque was visually displayed. MVCCon and MVCEcc tests were conducted in random order and a 10-min rest was granted between MVC modalities. The highest MVCCon and MVCEcc achieved during the two attempts were retained for analysis.

The local endurance test was conducted 15 min after MVCCon and MVCEcc testing. After a standardized warm-up consisted of submaximal concentric contractions, subjects performed 40 consecutive maximal concentric contractions (angular velocity, 60°·s−1) of the knee extensors over their full range of motion (i.e., from full knee flexion to full knee extension) (26). After full extension, subjects were instructed to relax during the flexion phase of the cycle, while the isokinetic machine arm returned to full flexion position (60°·s−1). Local endurance was assessed through a fatigue index (FI) expressed in percent: FI = 100 − [(last 5 repetitions/first 5 repetitions) × 100] (26). FI was therefore determined by averaging maximal concentric torque values recorded at the start and at the end of local endurance test only in the knee extension phase (Fig. 2).

Data from the local endurance test: the left panel (A) shows the individual decrements of peak concentric torque during the muscular endurance test. The gray bars represent each individual’s mean of two maximal voluntary concentric torques (AverageMVCCon). The black dots represent the first three concentric torque peaks recorded at the beginning of the local endurance test, and the white dots show the last three concentric torque peaks recorded at the end of the test. Each individual’s torque loss throughout the test is illustrated with a black dotted line. The right panel (B) shows a representative individual’s torque trace with a magnified extract in which flexion (recovery relaxation phase) and extension (active contraction phase) are indicated along with the position of the Biodex arm (black dotted line). The bottom right window shows the entire local endurance test (40 full cycles inducing a decrease in peak torque) for reference. All values used in this graph are absolute torque values.

Run time to exhaustion

Subjects were requested to perform the same warm-up routine as for the RE tests. Immediately after the warm-up, a relative velocity corresponding to 87.5% V˙O2max was set on the treadmill at +10% slope and the subjects ran until they could no longer maintain the required velocity. On the basis of pilot testing, this velocity was selected to obtain running times between 10 and 15 min. This running intensity was chosen to induce fatigue in less than 15 min with major aerobic contribution to running performance. The run time to exhaustion was measured using a manual stopwatch to the nearest second from the moment the participant released the handrail until he pushed on the security button fixed on the handrail. All subjects received strong verbal encouragement to continue as long as possible.

TR race

Running performance was determined from an official short distance TR race (December 2012; southeast of France; total number of participants, 120) with a medium elevation to distance ratio of 51.9 (1400-m positive elevation for 27-km total distance). The TR race was exclusively run on mountain single tracks with repeated technical rocky sections. All subjects wore a cardio-GPS watch (RS800CX; Polar, Kempele, Finland) during the TR race for continuous HR and speed monitoring. Although different refueling points were available during the TR race, subjects were free to carry light backpacks or drinking belts containing fluids or carbohydrates (e.g., drinks, gels, bars).

Statistical analysis

All data are expressed as mean ± SD. Pearson’s correlation test was used to determine which of the variables measured during the laboratory sessions was the best predictor for running performance defined as TR race time. Only for this analysis, a P value of ≤0.05 was considered statistically significant. Given the limitations related to the stepwise regression method, especially the overlapping variance attributed to the individual predictors, Nathans et al. (27) have suggested alternative multiple regression methods to assess each variable’s importance in a prediction model. The commonality regression analysis was implemented here to identify unique and common effects (i.e., commonality coefficients) of each predictor (independent variable) on the dependent variable (27,28), which was, in this case, TR race time as a measure of performance. Briefly, unique effects identify how much variance is unique to an observed variable or total effect (i.e., no shared variance with other independent variables), and common effects identify how much variance is common to groups of variables (i.e., shared variance or “overlap” in independent variables) (27). Finally, negative commonality coefficients occur in the presence of suppressor effects when some of the independent variables affect each other in the opposite direction (28).


The subjects’ average race time was 2 h 58 min 49 s ± 10 min 35 s, which corresponds to an average running speed of 9.42 ± 0.55 km·h−1. Their relatively high level of performance was shown by their final ranking between the 2nd and the 16th place but also V˙O2max responses ranging from 61.1 to 69.7 mL·min−1·kg−1. On average, the relative metabolic intensity sustained during the TR race represented 89.8% ± 2.8% HRmax. Mean run time to exhaustion was 773 ± 266 s (range, 552–1403 s). RE and muscle strength parameters are presented in Table 1.

RE and muscular characteristics measured during laboratory-based sessions.

The simple Pearson product–moment correlations of physiological variables with TR performance are listed in Table 2. On the basis of this statistical analysis, FI and relative V˙O2max showed the highest correlations with the TR race time, whereas the %V˙O2max at VT, RE0%, and RE+10% were not correlated. Mean values in vV˙O2max (r = −0.75; P = 0.03) were also associated with the TR race time. Conversely, no significant association was found between the TR race time and MVCCon (r = 0.27; P = 0.52), MVCEcc (r = −0.22; P = 0.60), or run time to exhaustion (r = −0.39; P = 0.29). Additional simple correlations indicated that after controlling for body mass, V˙O2max was not associated with RE0% (r = 0.61; P = 0.06) or RE+10% (r = 0.38; P = 0.28). Finally, run time to exhaustion was significantly correlated with RE0% (r = −0.85; P = 0.01) or RE+10% (r = −0.73; P = 0.02).

Pearson product–moment correlations with TR performance and commonality matrix with unique and common effects for each independent variable.

Unique and common effects of each independent variable were assessed using the commonality analysis (Table 2). A first statistical analysis was applied to the classic endurance running model, which produced a model summary with a total coefficient of R2 = 0.48 (Fig. 3)—a low predictive power. In this analysis, relative V˙O2max alone accounted for 90.2% of the total regression effect. However, a second commonality analysis was applied and included TR-specific factors such as FI and RE+10% (Fig. 3). In this case, the commonality matrix identified the best predictive model with total R2 = 0.98 from the use of relative V˙O2max, FI, and RE+10%, as independent variables. In total, these three predictors uniquely accounted for 74.8% of the total R2. The remaining 25.2% was due to the variance that the sets of predictors shared in common with TR race time. The most noticeable common effect observed was between FI and V˙O2max, which accounted for 26.9% of the regression effect. The percentage of unique and common contributions of each independent variable to the total R2 is detailed in Figure 3.

Graphic representation of commonality regression models for predicting TR performance. A, Commonality analysis for the classic endurance running model. B, Commonality analysis for the adapted and specific model to TR. Single arrows, dashed lines (double arrows), and external dotted lines represent, respectively, the contribution-percentage of unique, common to two factors, and common to all factors for each independent variable in the total regression effect (i.e., R 2). Negative values indicate a suppressor effect between independent variables. The sum of common and unique effects for each model corresponds to the total regression effect (i.e., total R 2).


The objective of this study was to identify the physiological determinants of short TR performance based on the classic model of level running performance. Using a commonality regression analysis, the major result of this study is that the classic endurance performance model does not explain short TR performance in a homogeneous group of trained trail runners. The novelty of this study is that the inclusion of factors more representative of TR such as local endurance and RE+10% improved the predictive power of the model and herewith provides new insights into the analysis of short TR performance.

In this investigation, the commonality regression analysis showed that laboratory-based physiological measures from the classic endurance running model explained only 48.1% of the total variance in TR performance (Fig. 3). As a consequence, various suppressor effects were identified through combinations between the three predictors selected in the model (Table 2), which tends to indicate that the amount of variance in the regression effect (total R2) is confounded by a set of independent variables (28). Although it is acknowledged that relative V˙O2max, %V˙O2max at LT, and RE are the three primary physiological factors in well-trained distance runners (13,16–18), the importance of these variables in determining running performance may depend on the degree of homogeneity in the population studied (e.g., characterized by relative V˙O2max responses and/or performance level). For instance, using a stepwise regression model, McLaughlin et al. (16) reported a strong prediction of a 16-km time trial performance (97.3%) from relative V˙O2max and RE variables in well-trained male and female distance runners. These results are in agreement with those previously reported over a longer distance running event (17). In this specific context, the flatter the running surface (i.e., road), the higher the relevance of laboratory-based physiological measures of relative V˙O2max, %V˙O2max, and RE. Many findings focusing on predictors of running performance have been observed within heterogeneous cohorts of well-trained runners (e.g., Refs. [16,18]). However, the physiological variables used in the classic endurance running model might have lower predictive value when the recruited population is highly trained and relatively homogeneous in terms of performance level (29,30). For homogeneous groups, it has been reported that other predictive variables were more appropriate to better describe running performance. In addition, the specificity of the running course profile (i.e., rough terrain) may have also influenced the predictive power of the classic model. In the present study, the profile of the TR race featured positive and negative changes in elevation over rocky and uneven terrain, and contrary to more conventional running events, flat sections were scarce (Fig. 1). In TR, prolonged concentric and eccentric muscle actions during uphill and downhill sections are known to induce specific mechanical and metabolic alterations (for review, see Ref. [1]), which could partly explain why the traditional model of endurance running is less appropriate for predicting TR performance based only on the classic physiological determinants.

Within this framework, incorporating local endurance (as assessed with the FI variable) into the model significantly improved the predictive power, uniquely accounting for not only 49.8% but also 26.9% in shared variance with relative V˙O2max of the total (R2 = 0.98; Fig. 3). The common effects identified between these two predictors indicate how particular sets of variables operate in combination in predicting TR performance, possibly generating recommendations regarding how to jointly target these two variables to produce desired effects. To our knowledge, this is the first study to highlight the importance of local endurance in short TR performance, even within a homogeneous group of trained athletes. Our results showed that trail runners with the highest local endurance (as assessed by a lower FI) had better TR final race times. On the basis of a previous study (31), it is likely that greater local endurance could limit the extent of any change in muscle recruitment and/or coordination and might have a potential regulatory role in fatigue development and, in turn, on TR performance. Interestingly, additional results also showed a strong and positive correlation between FI and the cumulated UHR times over TR sections (Fig. 4). Collectively, these results suggest that local endurance is a key physiological determinant of TR race time and performances especially in UHR sections.

Relationship between cumulated UHR or cumulated DHR times and TR performance (A), but also between UHR times and FI (%; B).

In the present work, local endurance was assessed through an FI variable calculated from repeated maximal concentric contractions of the knee extensors, that is, the modality of muscle action that is dominant not only in UHR sections but also during cycling or cross-country mountain bike exercises. As detailed in the “Methods” section, trail runners frequently use the latter activities in their training programs during specific uphill sessions. We suggest that the greater local endurance observed in the trained trail runners studied here may result in chronic muscular adaptations induced by specific uphill training, which in turn probably contributes to improved UHR and total TR race times.

The relationship between local endurance and TR race time provides a new practical insight into the comprehensive approach of short TR performance. Nevertheless, FI was the only significant strength variable correlated with the TR race, whereas maximal strength capabilities including MVCCon and MVCEcc showed a poor correlation. Only few studies about the relationship between muscle performance (e.g., muscle strength and power) and off-road running performance are available (21,32). In these recent investigations focusing on uphill marathons, the major findings indicated that runners with greater maximal mechanical power of lower limbs demonstrated smaller changes in running mechanics or lower fatigue-induced alterations in RE. Consequently, it has been suggested that specific power training of the lower limbs may contribute to the improvement of uphill marathon performance. Collectively, these results support the importance of muscle strength capacity and endurance in determining off-road running performance. In terms of training content, there is an interest for trail runners to include uphill cycling sections and local endurance training (11).

Furthermore, the simple Pearson product–moment correlations showed that relative V˙O2max and vV˙O2max were related to TR performance (r = −0.76 and r = −0.75, respectively; Table 1). The finding regarding vV˙O2max is in agreement with previous investigations about factors affecting running performance (e.g., Refs. [16,33]). Although no specific test was used to determine vV˙O2max in the current study, it would be interesting to determine this variable from a treadmill running protocol, which has been previously used (16), to better characterize the aerobic profile of trail runners. Indeed, it is well known that reaching a high vV˙O2max can be accomplished by having either a high V˙O2max or an improved RE (e.g., Ref. [16]). In addition, a high correlation between these two variables (r > 0.8) has consistently been observed among distance runners (12,14,34) over many years. In the present study, the correlation obtained between relative V˙O2max and TR performance was lower than those previously reported in distance runners who were greatly heterogeneous in terms of relative V˙O2max responses (e.g., Refs. [12,14]). This difference could be due, in great part, to the higher homogeneity in our experimental group (V˙O2max values ranging from 61.1 to 69.7 mL·min−1·kg−1). Consequently, these results support the idea that relative V˙O2max is an important factor in determining running performance (road or TR events), whether the studied population is heterogeneous or not.

In contrast, our results showed a poor correlation between %V˙O2max at VT or RE0% and TR race time (Table 1). The %V˙O2max at VT or LT is not systematically a reliable predictor of running performance. Indeed, McLaughlin et al. (16) found a nonsignificant correlation (r = 0.13) between %V˙O2max at LT and 16-km running performance in trained distance runners, probably because of the low variability of this physiological variable observed in this group of distance runners. Similarly, the low variability of runners’ %V˙O2max at VT reported in this study might explain the lack of a clear relationship with TR race time. Moreover, the finding that RE0% was not related to TR performance is in line with earlier studies conducted in Kenyan or nonelite European runners (29,35). We assume that, in this homogeneous group of trained runners, the relative importance of RE is lowered by other factors such as FI and V˙O2max to maintain high performance levels. It has been postulated in a large cohort of highly trained distance runners that a slight positive relationship (r = 0.25) exists between RE and V˙O2max when body mass is appropriately accounted for (36). In the present study, however, correlation analysis controlling for body mass revealed no relationship between V˙O2max and RE0%, suggesting that RE and V˙O2max are primarily determined independently.

The major drawback in translating the classic endurance running model to off-road races was the traditional evaluation of RE based on a level treadmill protocol. The determination of RE specific to TR incline ground constraints might provide an improved predictive power of the model in off-road runners. This is supported by the better total R2 when including RE+10% in the model (Fig. 3). Because the expression of RE as a caloric unit cost (e.g., kcal·kg-1·km−1) has been suggested to be more sensitive to changes in relative speed than RE expressed as O2 unit cost (25), it would be interesting to calculate an average caloric unit cost specific to TR including at least two running velocities or relative intensities set at level ground, downhill, and uphill. Moreover, our findings also indicated that better treadmill time to exhaustion was associated with improved RE0% (r = −0.85) and RE+10% (r = −0.73), suggesting the importance of determining RE on a surface/slope on which runners can reproduce a running pattern representative of that produced in field conditions. Using a telemetric system, Jensen et al. (37) reported that orienteering athletes who included uneven terrain sessions in their daily training demonstrated superior RE on this type of surface compared with track runners who trained on flat roads. Recently, in a group of elite and amateur orienteering athletes, Hébert-Losier et al. (38) reported a lower correlation between laboratory-based RE measures and 2-km time trial performance when the field test was performed in a forest-path compared with road condition. Collectively, these results suggest that using steep inclines and/or uneven terrains during field-based RE testing procedures may lead to a more realistic/relevant assessment of physiological factors. Another limitation of this study is the small sample size of the population, explained by the selective recruitment of highly trained athletes with extensive TR experience. This resulted in homogeneous relative V˙O2max values and TR race times. These criteria are relatively unique in the analysis of short TR races and may be compared with off-road studies using small groups of elite athletes (e.g., Refs. [20,38]).

In conclusion, the findings of this study indicate that the classic physiological model of endurance running does not allow for the successful identification of physiological predictors for short TR performance within a homogeneous group of trained trail runners. However, the predictive power of the model was markedly improved when incorporating more specific factors to TR such as local endurance or RE measured in a positive slope condition. Although this study provides a new insight into the comprehensive approach of short TR performance, future studies should include field-based RE testing procedures and various local endurance tests specifically under concentric and eccentric conditions.

The authors thank the subjects for their willingness to participate in the present work. They also warmly thank Prof. Kim Nimon from the University of North Texas for her precious help in the commonality regression analysis. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.

No funding was received for this study from National Institutes of Health, Welcome Trust, Howard Hughes Medical Institute, or others. The authors report no conflict of interest. The results of the present study do not constitute endorsement by the American College of Sports Medicine.


1. Giandolini M, Vernillo G, Samozino P, et al. Fatigue associated with prolonged graded running. Eur J Appl Physiol. 2016;116(10):1859–73.
2. The Outdoor Foundation. Outdoor Participation Report 2016 [Internet]. 2016;63 [cited 2017 Oct 5]. Available from:
3. Millet GY, Martin V, Lattier G, Ballay Y. Mechanisms contributing to knee extensor strength loss after prolonged running exercise. J Appl Physiol (1985). 2003;94(1):193–8.
4. Nicol C, Avela J, Komi PV. The stretch-shortening cycle: a model to study naturally occurring neuromuscular fatigue. Sports Med. 2006;36:977–99.
5. Dewolf AH, Peñailillo LE, Willems PA. The rebound of the body during uphill and downhill running at different speeds. J Exp Biol. 2016;219(16):2276–88.
6. Gottschall JS, Kram R. Ground reaction forces during downhill and uphill running. J Biomech. 2005;38(3):445–52.
7. Vernillo G, Giandolini M, Edwards WB, et al. Biomechanics and physiology of uphill and downhill running. Sports Med. 2016;9:1–15.
8. Martin V, Millet GY, Lattier G, Perrod L. Why does knee extensor muscles torque decrease after eccentric-type exercise? J Sports Med Phys Fitness. 2005;45(2):143–51.
9. Giandolini M, Horvais N, Rossi J, Millet GY, Morin JB, Samozino P. Acute and delayed peripheral and central neuromuscular alterations induced by a short and intense downhill trail run. Scand J Med Sci Sports. 2016;26(11):1321–33.
10. Vercruyssen F, Gruet M, Colson SS, Ehrstrom S, Brisswalter J. Compression garments, muscle contractile function, and economy in trail runners. Int J Sports Physiol Perform. 2017;12(1):62–8.
11. Barnes KR, Kilding AE. Strategies to improve running economy. Sports Med. 2014;45:37–56.
12. Bassett DR Jr, Howley ET. Limiting factors for maximum oxygen uptake and determinants of endurance performance. Med Sci Sports Exerc. 2000;32(1):70–84.
13. Daniels J, Daniels N. Running economy of elite male and elite female runners. Med Sci Sports Exerc. 1992;24:483–9.
14. Costill DL, Thomason H, Roberts E. Fractional utilization of the aerobic capacity during distance running. Med Sci Sports. 1973;5(4):248–52.
15. Farrell PA, Wilmore JH, Coyle EF, Billing JE, Costill DL. Plasma lactate accumulation and distance running performance. Med Sci Sports. 1979;11(4):338–44.
16. McLaughlin JE, Howley ET, Bassett DR, Thompson DL, Fitzhugh EC. Test of the classic model for predicting endurance running performance. Med Sci Sports Exerc. 2010;42(5):991–7.
17. Joyner MJ, Coyle EF. Endurance exercise performance: the physiology of champions. J Physiol. 2008;586(1):35–44.
18. Foster C, Lucia A. Running economy: the forgotten factor in elite performance. Sports Med. 2007;37(4–5):316–9.
19. Jones AM. The physiology of the world record holder for the women’s marathon. Int J Sport Sci Coach. 2006;1(2):101–16.
20. Balducci P, Clémençon M, Morel B, Quiniou G, Saboul D, Hautier CA. Comparison of level and graded treadmill tests to evaluate endurance mountain runners. J Sports Sci Med. 2016;15(2):239–46.
21. Lazzer S, Salvadego D, Taboga P, Rejc E, Giovanelli N, Di Prampero PE. Effects of the Etna uphill ultramarathon on energy cost and mechanics of running. Int J Sports Physiol Perform. 2015;10(2):238–47.
22. Millet GY, Tomazin K, Verges S, et al. Neuromuscular consequences of an extreme mountain ultra-marathon. PLoS One. 2011;6(2):e17059.
23. Easthope CS, Nosaka K, Caillaud C, Vercruyssen F, Louis J, Brisswalter J. Reproducibility of performance and fatigue in trail running. J Sci Med Sport. 2014;17(2):207–11.
24. Wasserman K, Whipp BJ, Koyl SN, Beaver WL. Anaerobic threshold and respiratory gas exchange during exercise. J Appl Physiol. 1973;35(2):236–43.
25. Fletcher JR, Esau SP, MacIntosh BR. Economy of running: beyond the measurement of oxygen uptake. J Appl Physiol. 2009;107(6):1918–22.
26. Louis J, Hausswirth C, Easthope C, Brisswalter J. Strength training improves cycling efficiency in master endurance athletes. Eur J Appl Physiol. 2012;112:631–40.
27. Nathans LL, Oswald FL, Nimon K. Interpreting multiple linear regression: a guidebook of variable importance. Pract Assess Res Eval. 2012;17(9):1–19.
28. Nimon K, Gavrilova M. Regression commonality analysis: demonstration of an SPSS solution. Mult Linear Regres Viewpoints. 2010;36(1):10–7.
29. Mooses M, Mooses K, Haile DW, Durussel J, Kaasik P, Pitsiladis YP. Dissociation between running economy and running performance in elite Kenyan distance runners. J Sports Sci. 2015;33(2):136–44.
30. Abe D, Yanagawa K, Yamanobe K, Tamura K. Assessment of middle-distance running performance in sub-elite young runners using energy cost of running. Eur J Appl Physiol. 1998;77(4):320–5.
31. Hayes PR, French DN, Thomas K. The effect of muscular endurance on running economy. J Strength Cond Res. 2011;25(9):2464–9.
32. Giovanelli N, Taboga P, Rejc E, Lazzer S. Effects of strength, explosive and plyometric training on energy cost of running in ultra-endurance athletes. Eur J Sport Sci. 2017;17(7):805–13.
33. Morgan DW, Baldini FD, Martin PE, Kohrt WM. Ten kilometer performance and predicted velocity at VO2max among well-trained male runners. Med Sci Sports Exerc. 1989;21(1):78–83.
34. Davies CT, Thompson MW. Aerobic performance of female marathon and male ultramarathon athletes. Eur J Appl Physiol Occup Physiol. 1979;41(4):233–45.
35. Williams KR, Cavanagh PR. Relationship between distance running mechanics, running economy, and performance. J Appl Physiol. 1987;63(3):1236–45.
36. Shaw AJ, Ingham SA, Atkinson G, Folland JP. The correlation between running economy and maximal oxygen uptake: cross-sectional and longitudinal relationships in highly trained distance runners. PLoS One. 2015;10(4):1–10.
37. Jensen K, Johansen L, Kärkkäinen O-P. Economy in track runners and orienteers during path and terrain running. J Sports Sci. 1999;17(12):945–50.
38. Hébert-Losier K, Jensen K, Mourot L, Holmberg HC. The influence of surface on the running velocities of elite and amateur orienteer athletes. Scand J Med Sci Sport. 2014;24(6):e448–55.


© 2018 American College of Sports Medicine