Individual training adaptations in response to chronic exposure to physical exercise are related to athletes’ individual physical fitness level and the magnitude (i.e., intensity and duration) of the prescribed training load (TL) (Impellizzeri et al., 2005). As a result, the quantification of the individual responses to a given external TL is important to optimize physical fitness performance. However, in a team sport such as soccer, the extensive use of group exercises where inter-player TL differences vary depending on the type of ball-drill performed (Los Arcos et al., 2014) make the prescription of training difficult in relation to individual characteristics (Impellizzeri et al., 2005). For instance, those players with the highest VO_2max showed the lowest percentage of VO_2max during the small group play (Hoff et al., 2002). Therefore, the remarkable differences in TL accumulation between soccer players (Akubat et al., 2012; Impellizzeri et al., 2005; Manzi et al., 2013) and the difficulty to ensure an appropriate level of training stimulus for each soccer player, when a team-based training approach is undertaken, suggest how important the quantification of the individual’s TL is (Alexiou and Coutts, 2008) and the relevance of this information that makes possible the identification of the players with insufficient or excessive TL and individual training strategies.

The quantification of TL is generally based on both external (e.g., distance, power output, number of repetitions) and internal (e.g., oxygen uptake, heart rate, blood lactate, rate of perceived exertion) indicators of effort intensity (Buchheit, 2014). In order to know the stimulus for exercise-induced adaptations, Foster et al. (2001) proposed the session-rating of perceived exertion (sRPE) as a practical tool for evaluating internal TL in endurance and team sport athletes (Foster et al., 1996, 2001, 1995; Foster, 1998). This method requires from the players to subjectively rate the intensity of the entire training session using a modified version of category ratio scale (CR-10 scale) developed by Borg et al. (1998), the Foster’s 0–10 scale (Foster et al., 2001). This intensity value is then multiplied by the total duration (minutes) of the training session to create a single measure of internal TL (sRPE-TL) in arbitrary units (AUs) (Foster et al., 2001). Previous investigations have shown sRPE-TL to compare favourably with more sophisticated methods of quantifying TL such as heart rate (HR) in endurance (Foster et al., 2001), team sport (Alexiou and Coutts, 2008; Casamichana et al., 2013; Foster et al., 2001; Impellizzeri et al., 2004; Scott et al., 2013), and resistance-trained athletes (Day et al., 2004). Specifically for soccer training, several studies have described substantial positive correlations between sRPE-TL and HR-based TL (Alexiou and Coutts, 2008; Casamichana et al., 2013; Impellizzeri et al., 2004; Scott et al., 2013). Thus, the sRPE-TL appears to be a good indicator of the global internal load for soccer training (Impellizzeri et al., 2004). Furthermore, RPE method is easy, versatile and cheap (Borg, 1982), it does not require technical expertise and the TL is measured quickly without chance for technical errors (Alexiou and Coutts, 2008).

Despite the potential limitations of HR measures to quantify TL in team sports (Borresen and Lambert, 2009), previous training studies have generally reported substantial and positive associations between the changes in HR-based TL and physical fitness parameters in soccer players (Akubat et al., 2012; Castagna et al., 2011; 2013; Manzi et al., 2013). For example, after 6 to 8 weeks of pre-season training, large to very-large correlations were reported between training time spent at high-intensity (> HR at 4 mmol·l⁻¹) and changes in running speed at 2 and 4 mmol·l⁻¹ of blood lactate (Castagna et al., 2011; 2013), maximal oxygen uptake (VO_2max) (Castagna et al., 2013) and performance in the Yo-Yo IR1 (Castagna et al., 2013). Moreover, after 8 weeks of pre-season training, Manzi et al. (2013) reported large to very-large correlations between HR-derived individualised training impulse (iTRIMP) TL and changes in the running speed at 4 mmol·l⁻¹ of blood lactate and performance in the Yo-Yo IR1. Similarly, Akubat et al. (2012) reported a large association between the HR-derived iTRIMP and changes in the running speed at 2 mmol·l⁻¹ of blood lactate in young soccer players after six week of in-season training. Interestingly, in this last study authors did not find any substantial association between changes in several aerobic fitness variables (i.e., running speed at 2 mmol·l⁻¹ and 4 mmol·l⁻¹ of blood lactate) and overall sRPE-TL (Akubat et al., 2012). Thus, despite the reported validity and advantages of the RPE derived TL (Alexiou and Coutts, 2008; Impellizzeri et al., 2004), it is still unclear how useful this measure is for monitoring changes in physical fitness performance in soccer players.

To our knowledge, all previous soccer studies that have assessed the relationship between longitudinal changes in TL and physical fitness performance have only paid attention to aerobic fitness variables (Akubat et al., 2012; Castagna et al., 2011, 2013; Manzi et al., 2013). In addition to aerobic fitness, neuromuscular factors (i.e., strength, power, speed) are important performance determinants of soccer match-play physical performance (Stølen et al., 2005) and are heavily taxed during matches and training sessions (Ascensão et al., 2008; Krustrup et al., 2006; Mohr et al., 2004; Rampinini et al., 2011; Robineau et al., 2012; Thorlund et al., 2009). Despite neuromuscular factors being perceived as crucial for soccer physical performance, to the best of our knowledge, no previous study has assessed the relationship between changes in TL and neuromuscular fitness parameters in soccer. In order to evaluate these associations between RPE-derived TL and both aerobic and neuromuscular fitness components, the differentiated use of respiratory sRPE-TL (sRPEres-TL) and muscular sRPE-TL (sRPEmus-TL) (Aliverti et al., 2011; Borg et al., 2010; Green et al., 2009; Mahon et al., 1998) may offer, also in team sports (Arcos et al., 2014; Yanci et al., 2014; Weston et al., 2014), an interesting alternative to the global sRPE.

The purpose of this study was, therefore, to examine the usefulness of sRPEres-TL and sRPEmus-TL for monitoring changes in several aerobic fitness and neuromuscular performance variables during 9 weeks of soccer training in young professional players.

No substantial differences were found between the sRPE magnitudes declared 10 or 30 min after the training sessions (n = 207) for sRPEres (2.8 ± 1.1 vs 2.8 ± 1.1; p = 0.29; ES = -0.02, 90% confidence limits [-0.05;0.03]; most likely trivial) nor sRPEmus (2.8 ± 1.2 vs 2.8 ± 1.2; p = 0.13; ES = -0.03, 90% confidence limits [-0.06;0.00]; most likely trivial). ICC and TEM were 0.96 (90% confidence limits [0.96-0.97]) and 0.21 (90% confidence limits [0.19;0.23]) for the sRPEres, respectively, while ICC was 0.97 (90% confidence limits [0.96;0.97]) and TEM 0.21 (90% confidence limits [0.20;0.23]) for sRPEmus magnitude.

Team mean total training volume, sRPEres-TL, sRPEmus-TL, sumRPEres and sumRPEmus during the study is presented in Table 1.

Looking into pre-season (5 weeks) and in-season (4 weeks) periods separately, the weekly team mean sRPEres-TL and sRPEmus-TL were greater in the pre-season (Figure 1). However, session team mean sRPEres-TL and sRPEmus-TL was similar almost for all weeks (Figure 2).

As displayed in Table 2, substantial improvements were found from Test 1 to Test 2 in sprint times (5 m: ES = -0.85 (0.41), very likely; 15 m: ES = -0.42 (0.28), likely l) and in aerobic fitness (V₃: ES = 0.34 (0.49), possibly Lac12: ES = -0.34 (0.46), possibly; Lac13: ES = -0.35 (0.37), likely).

Table 3 shows the correlations between different methods of quantification of the TL and changes in physical fitness performance.

Large and negative correlations were found between both sRPEmus-TL and sumRPEmus and change in jump performance in CMJD and CMJnD. Figure 3 shows the correlation between sumRPEmus and changes in jump performance in CMJnD. Similarly, large negative correlation was described between sRPEmus-TL and changes in performance at 13 km·h⁻¹ (Lac13) (Figure 4) and between activity volume and change in sprint performance (i.e., 5 m and 15 m) and CMJAS jump.

The purpose of this study was to examine the usefulness of the perceived exertion-derived TL for monitoring changes in several aerobic fitness and neuromuscular parameters during 9 weeks of soccer training in young professional players. The main finding of the present study was that the accumulated perceived (leg) muscular exertion, and the associated training load, and the accumulated training volume were negatively correlated with the changes in most physical fitness attributes after 9 weeks of training. That is, higher training volume and feelings of strain in the leg muscles were typically associated with impairments in physical fitness performance.

Looking into both periods (i.e., pre-season and in-season) separately, the mean weekly sRPEres-TL and sRPEmus-TL were substantially higher for the first one: the mean sRPE-TL was 2164 ± 395 vs. 1449 ± 274 for sRPEres while 2270 ± 365 vs. 1434 ± 300 AU for sRPEmus (Figure 1). These differences between pre-season and in-season weekly TL are in agreement with those reported in previous studies (Algrøy et al., 2011; Impellizzeri et al., 2006; Jeong et al., 2011) where overall sRPE-TL was used. Despite these differences in mean weekly TL, average session (daily) TL were very similar between all weeks (Figure 2). Therefore, the greater weekly mean sRPEres-TL and sRPEmus-TL during pre-season can be attributed to the higher frequency of training sessions and matches (i.e., 5-8 trainings and 1-2 friendly matches a week).

Several studies have previously investigated the association between HR-based TL and changes in different aerobic fitness parameters in soccer players (Akubat et al., 2012; Castagna et al., 2011, 2013; Manzi et al., 2013). All those aforementioned studies found linear dose-response relationships. That is, the more training accumulated at higher intensities the greater the improvements in aerobic fitness. For example, Castagna et al. (2011) reported very large to large associations between the training time spend at high intensity (i.e., > HR at 4 mmol.L⁻¹ lactate threshold) and the changes in speed at 2 (r = 0.84, p = 0.001) and 4 mmol.L⁻¹ (r = 0.65, p = 0.001) of blood lactate after 6 weeks of pre-season training in elite soccer players. Similarly, the same authors (Castagna et al., 2013) found that training spent at high intensity (i.e., > HR at 4 mmol.L⁻¹ lactate threshold) positively related to changes in speed at 2 (r = 0.78, p = 0.002) and 4 mmol.L⁻¹ (r = 0.60, p = 0.03) of blood lactate, maximal oxygen uptake (r = 0.65; p = 0.02) and Yo-Yo IR1 (r = 0.66; p = 0.01) after 8 weeks of pre-season training in elite-standard male soccer players. Despite the large to very large positive associations previously reported between HR- and RPE-based TL in soccer training (Alexiou and Coutts, 2008; Casamichana et al., 2013; Impellizzeri et al., 2004; Scott et al., 2013), only negative, most of them unclear, associations were obtained between most of the RPE-derived TL indicators and observed changes in aerobic fitness (Table III). Somehow surprisingly, the only substantial association was a large negative correlation between the muscular RPE-derived TL and dLac13 (Table III). That is, the higher the perceived TL the smaller the improvement in Lac13. Previous studies reported no significant associations between RPE-based TL and training-induced changes in markers of aerobic fitness in youth soccer players (Akubat et al., 2012; Brink et al., 2010) and rugby players (Coutts et al., 2003; Gabbett and Domrow, 2007). Possible reasons for the different findings are the alternative RPE scales and population investigated in each study. For example, all previous studies (Akubat et al., 2012; Brink et al., 2010; Coutts et al., 2003; Gabbett and Domrow, 2007) employed the global Borg RPE scale, while the present study used the differentiated muscular and respiratory RPE (Arcos et al., 2014). In addition, the current study dealt with professional players, while previous studies focused on young and/or sub-elite populations.

While a large negative association between sRPEmus-TL and changes in Lac13 was reported, the magnitude of this association (r = -0.57) indicates that only 0 33% of the variance in Lac13 changes can be explained by the sRPEmus-TL. While such magnitude can be certainly relevant in a practical, “real-life” setting, it also suggests the presence of substantial individual responses. This might prevent practitioners to use this measure as a definitive monitoring tool with all the players. Nevertheless, taken together, present and previous results seem to suggest that, unlike HR-derived TL methods, RPE accumulated TL does not appear to fit positive, dose-response relationships with aerobic fitness in soccer players.

A new finding of the present study was the negative correlation between the changes in neuromuscular fitness parameters (i.e., jumping and sprinting) and accumulated TL and training time (Table III). Specifically, single leg CMJ displayed large (r = -0.52 to -0.61) correlations with sRPEmus and sRPEmus-TL, while total accumulated training time showed a large negative correlation with both bilateral jumping (CMJAS: r = -0.51) and sprinting (5 m: r = -0.54; 15 m: r = -0.64). That is, the players who perceived the prescribed training load as harder were more likely not to improve single leg CMJ performance as much at the players with lower perceived load. Similarly, players that accumulated higher training volume were most likely to show an impaired sprinting performance from T1 to T2. To the authors’ knowledge, no previous studies have examined the relationship between TL and neuromuscular fitness parameters in soccer. However, Gabbett and Domrow (2007) reported that increases in RPE-derived training load during the early-competition training phase decrease agility performance in sub-elite rugby players.

Present results suggest that a high perception of leg muscular effort associated with training and matches, as well as an excessive accumulation of training volume (time), can impair several physical fitness factors believed to be relevant for on-field soccer performance. Therefore, the independent assessment of muscular effort to quantify TL can be an interesting additional monitoring measure in soccer training. The fact that the magnitude of most correlations, despite being meaningful, was less that 0.70 (indicating less than 50% of shared variance between the two variables) appears to indicate the presence of substantial individual (between-player) responses in the reported associations. This warrants caution when using leg perceived exertion and/or training volume as sole indicators of load tolerance in soccer players. Moreover, taken together, present and previous results seem to suggest that, unlike HR-derived TL methods, RPE accumulated TL does not appear to fit positive, dose-response relationships with aerobic fitness in soccer players. Thus, despite that several cross-sectional have validated the use of the sessions RPE as a proxy of HR-derived physiological stress arising from soccer practice, the present longitudinal data indicate that RPE- or HR-derived measures of exercise intensity/load in soccer training could not be used interchangeably.