The comparison between different exercise modalities has the potential to improve our understanding of the physiology of endurance performance. Classical studies have compared leg cycling (Leg) with arm+leg cycling (Arm+leg) to unravel the mechanisms limiting maximal oxygen uptake (Astrand and Saltin, 1961; Secher et al., 1974; Bergh et al., 1976). Findings from these studies have contributed to outlining the important role of cardiocirculatory factors in setting the upper limit for maximal aerobic power. Indeed, the peak value of oxygen uptake ( V ˙ O₂peak) is not always proportional to the differences in the amount of muscle mass involved in various exercise modalities, as the addition of arm work to leg work generally does not increase V ˙ O₂peak more than about 5%–10% (Astrand and Saltin, 1961; Gleser et al., 1974; Secher et al., 1974; Secher and Volianitis, 2006). In fact, the comparison between arm+leg cycling and leg cycling is suitable for gaining insight into other physiological responses that have received less attention so far, including the variables associated with physical effort and changes in exercise tolerance. Exercise tolerance is here defined as the tolerated duration during a time-to-exhaustion (TTE) test performed at a constant work rate or the peak power output (PPO) achieved during an incremental test (Van De Walle and Vukovich, 2018). Some findings have shown that arm+leg cycling results in a greater exercise tolerance compared to leg cycling alone (arms hanging on the participant’s side) (Gleser et al., 1974; Secher et al., 1974; Nagle et al., 1984), but it is unclear if this difference is still evident when arm+leg cycling is compared to conventional leg cycling (Secher et al., 1974; Bergh et al., 1976; Nagle et al., 1984). When matched for the same absolute total power output, the greater exercise tolerance that might be expected for arm+leg cycling makes the comparison with (conventional) leg cycling valuable for testing the proposition that improvements in exercise tolerance are accompanied by consistent changes in the responses of respiratory frequency (f _R) and perceived exertion (Nicolò and Sacchetti, 2023).

Growing evidence suggests that f _R is a valid marker of physical effort (Nicolò et al., 2014; Nicolò et al., 2016; Nicolò et al., 2017a; Nicolò et al., 2017b; Nicolò et al., 2019; Nicolò and Sacchetti, 2023) and that its time course reflects changes in exercise tolerance (Nicolò and Sacchetti, 2023). In a variety of conditions where exercise tolerance is reduced (experimentally) or lowered (in a cross-sectional comparison), the rate of increase in f _R is higher, both during incremental and TTE tests. On the other hand, the rate of increase in f _R is lower when assessing exercise strategies, experimental interventions or other conditions leading to an improvement in exercise tolerance (Nicolò and Sacchetti, 2023). The sensitivity of f _R to changes in exercise tolerance and the close association between f _R and perceived exertion are among the factors suggesting that f _R is to a large extent modulated by central command (the activity of motor and premotor brain areas relating to voluntary locomotor muscle contraction) during high-intensity exercise (Nicolò and Sacchetti, 2023). This explains why f _R can be considered a marker of physical effort, which is defined as the degree of motor effort (i.e. the magnitude of central command) (Nicolò et al., 2017b). However, measuring the magnitude of central command during “real” exercise conditions is particularly challenging. Hence, it is important to use different approaches (including the comparison of different exercise modalities) to provide indirect evidence on the contribution of central command to f _R modulation (Nicolò and Sacchetti, 2023).

The comparison between arm+leg cycling and leg cycling may either challenge or reinforce the notion that the increase in f _R during high-intensity exercise reflects changes in exercise tolerance and is influenced by central command. f _R is also modulated by muscle afferent feedback from groups III and IV (hereinafter muscle afferent feedback) (Dempsey et al., 2014; Girardi et al., 2021; Nicolò and Sacchetti, 2023), and this drive to breathe may have a greater relative contribution to f _R modulation when arm muscles assist leg muscles during arm+leg cycling, in view of a potentially larger amount of muscle mass concomitantly involved in exercise (Dempsey et al., 2014; Nicolò and Sacchetti, 2023). Indeed, even the passive movement of the legs leads to a substantial increase in f _R that is at least partially mediated by muscle afferent feedback (Girardi et al., 2021), and this drive to breathe may increase further when adding the movement of the upper limbs. The contribution of muscle afferent feedback to f _R may partially confound the association between f _R and perceived exertion because the latter is supposed to be largely independent of muscle afferent feedback (Marcora, 2009; Bergevin et al., 2023). While arm+leg cycling may also result in a higher V ˙ O₂ compared to leg cycling at the same submaximal power output (Hoffman et al., 1996), f _R largely dissociates from metabolic rate and is not substantially modulated by metabolic inputs, unlike tidal volume (V_T) (Nicolò et al., 2017a; 2018; Nicolò and Sacchetti, 2019; Nicolò et al., 2020a; Nicolò and Sacchetti, 2023). On the other hand, some findings seem to support the association between f _R and perceived exertion during both arm+leg cycling and leg cycling. Robertson et al. (1986) found similar responses of f _R—but not V_T—and the ratings of perceived exertion (RPE) during arm cycling, leg cycling and arm+leg cycling for intensities ranging from 20% to 80% of V ˙ O₂peak. However, the authors neither reported the responses of f _R and RPE when exhaustion was approaching nor described if the two variables were sensitive to between-modality changes in exercise tolerance. Further studies are required to address this issue.

The purpose of the present study was to systematically assess whether exercise tolerance improves with arm+leg cycling vs. (conventional) leg cycling and whether f _R and perceived exertion are sensitive to the expected differences in exercise tolerance. To increase the robustness of our evaluation, we compared arm+leg cycling vs. leg cycling using two exercise paradigms (i.e. incremental test and TTE test) and three comparisons, as the TTE test was performed at two different intensities. We tested the hypotheses that i) arm+leg cycling improves exercise tolerance compared to leg cycling irrespective of the exercise paradigm; and ii) f _R is a good marker of physical effort sensitive to between-modality changes in exercise tolerance, unlike other physiological variables such as V ˙ O₂ and heart rate (HR).

This study aimed to systematically assess whether exercise tolerance improves with arm+leg cycling vs. leg cycling and whether f _R and perceived exertion are sensitive to the expected differences in exercise tolerance. This goal was achieved by comparing the two exercise modalities using three performance tests and two exercise paradigms (i.e. incremental test and TTE test). The main findings of the study are as follows: 1) exercise tolerance was substantially improved in all the Arm+leg tests; 2) perceived exertion, minute ventilation and respiratory frequency were particularly sensitive to the between-modality changes in exercise tolerance observed, unlike V ˙ O₂ and heart rate. These findings support the notion that respiratory frequency is a marker of physical effort during high-intensity exercise and that its time course reflects changes in exercise tolerance. This holds true even during arm+leg cycling, which is an exercise modality where the responses of V ˙ O₂ and heart rate do not reflect the reduction in physical effort and the improvement in exercise tolerance observed when comparing it with leg cycling.

Our findings provide convincing evidence that arm+leg cycling substantially increases exercise tolerance when compared to (conventional) leg cycling, hence expanding on the limited literature dealing with performance differences between the two exercise modalities (Secher et al., 1974; Bergh et al., 1976; Nagle et al., 1984). The apparent difference in the percentage improvement in exercise tolerance found between the incremental tests and the TTE tests is, in fact, in line with the different characteristics of the two performance paradigms. Indeed, a 1% improvement in power output in an incremental test, or in a time trial, results in a performance improvement that can exceed 10% in a TTE test (Hopkins et al., 1999). Hence, the average increase in power output of about 15% that we found in the incremental test is compatible with the average increase in TTE found in the TTE tests (i.e. 108% in the Arm+leg_TTE90 and 90% in the Arm+leg_TTE75). Such an improvement in exercise tolerance observed in the Arm+leg tests implies that the effort required to sustain a given power output is substantially lower at isotime compared to that of the Leg tests, and this premise is supported by our findings.

The notion that physical effort was lower in the Arm+leg tests than in the Leg tests is substantiated by the lower values of perceived exertion and f _R generally found at isotime. While the decrease in perceived exertion during incremental Arm+leg did not reach statistical significance (p = 0.057), it was substantial when comparing Arm+leg vs. Leg during the TTE tests. Indeed, perceived exertion was found to be among the most sensitive variables to variations in exercise tolerance during the TTE tests. These findings extend previous findings showing a lower RPE during arm+leg cycling vs. leg cycling at submaximal work rates (Hoffman et al., 1996; Hill et al., 2018). At isotime, f _R was significantly lower (p < 0.003; η _P ² > 0.40) in the three Arm+leg tests than in the Leg tests, although this difference generally reached statistical significance in the last 20% of the Leg tests. This is an important feature of the f _R response that might have been missed in previous studies that did not compare Arm+leg and Leg until exhaustion (Robertson et al., 1986). Likewise, the reduction in f _R that is observed in the last part of a TTE test when exercise tolerance improves may not appear if the variability in TTE is not addressed on an individual basis when analyzing TTE data (Nicolò et al., 2019). We have overcome this problem by using the previously described “individual isotime” analysis (Nicolò et al., 2019), which reduces extensively the data loss that occurs when using more traditional analyses. Our findings collectively suggest that the time course of f _R reflects changes in exercise tolerance both during incremental exercise and TTE exercise, thus supporting the study hypothesis.

The association found between f _R and RPE and the sensitivity of f _R to changes in exercise tolerance support a substantial modulation of f _R by central command (Nicolò and Sacchetti, 2023). However, a partial dissociation was found between f _R and RPE, which may suggest that also muscle afferent feedback contributed to the f _R response. Although evidence suggests that muscle afferent feedback has a greater relative contribution to f _R during moderate exercise than during high-intensity exercise (Amann et al., 2010; Dempsey et al., 2014; Girardi et al., 2021; Nicolò and Sacchetti, 2023), arm+leg cycling implies the simultaneous use of the muscle mass of both the upper and lower limbs, possibly resulting in a greater magnitude of muscle afferent feedback, and especially of its mechanosensitive component (i.e. mechanoreflex). This may explain the slightly higher f _R shown in the first part of the TTE tests in the Arm+leg vs. the Leg modality. Indeed, it has been suggested that the relative contribution of muscle afferent feedback to ventilation is higher when the muscle mass recruited is larger (Amann et al., 2011; Dempsey et al., 2014; Nicolò and Sacchetti, 2023). Conversely, it is conceivable that the magnitude of the metabosensitive component of muscle afferent feedback (i.e. metaboreflex) was reduced at isotime in the Arm+leg tests because of the lower intramuscular metabolic perturbation. Hence, the metaboreflex cannot be ruled out as an input contributing to the decrease in f _R observed in the Arm+leg tests. Nevertheless, the relative contribution of muscle afferent feedback may reduce over time during a TTE test because the contribution of other inputs increases substantially (e.g. central command) (Dempsey et al., 2014; Nicolò and Sacchetti, 2023). We cannot exclude that afferent feedback from pulmonary mechanoreceptors or alterations in chest wall mechanics might have contributed to the partial dissociation observed between the f _R and RPE responses. Arm movements may increase the contribution to ventilation of afferent feedback from pulmonary mechanoreceptors or alter the mechanics of breathing, although these propositions require further investigation. Furthermore, it cannot be excluded that the lower pedaling cadence observed in the Arm+leg tests might have contributed to the between-modality differences observed in f _R, although variations in pedaling cadence and f _R might not be proportional, especially during high-intensity exercise (Girardi et al., 2021). Conversely, it is less plausible that metabolic acidosis or other metabolic inputs might have provided a substantial direct contribution to the f _R modulation, and the reader is referred to previous studies where evidence supporting this proposition has been reviewed (Nicolò et al., 2020a; Nicolò and Sacchetti, 2023).

The ventilatory responses observed in this study can be interpreted in the light of a recently proposed model of ventilatory control during exercise, which suggests that f _R and V_T are modulated to a large extent by behavioral and metabolic inputs respectively (Nicolò and Sacchetti, 2023). While it was proposed that ventilation is differentially regulated during incremental exercise and TTE exercise (Syabbalo et al., 1994), our findings suggest that the breathing pattern is affected by the magnitude of the inputs modulating ventilation rather than by the type of exercise paradigm. Syabbalo et al. (1994) observed a more rapid and shallow breathing pattern during a TTE test at about 76% of the PPO reached in an incremental exercise than during this latter test. The lower V_T reported by Syabbalo et al. (1994) during the TTE test is in line with the V_T response observed during the TTE tests at 75% of the PPO_Leg in this study. However, the breathing pattern we observed in the TTE tests at 90% of the PPO_Leg was, conversely, more similar to that found during the incremental exercise than during the TTE tests at 75% of the PPO_Leg (see Figure 7). Hence, the rapid and shallow breathing pattern is not a feature of TTE exercise, which is in contrast with what Syabbalo et al. (1994) had suggested. Conversely, the observed findings can be explained by the differential control of f _R and V_T (Nicolò et al., 2017a; Nicolò et al., 2018; Nicolò and Sacchetti, 2019; Nicolò and Sacchetti, 2023). While f _R generally shows similar peak values during incremental exercise and constant work rate exercise (Syabbalo et al., 1994; Nicolò and Sacchetti, 2023), the V_T peak reached during exercise is largely influenced by the magnitude of metabolic inputs and is generally associated with the V ˙ CO₂ peak (Nicolò and Sacchetti, 2023). As such, we found considerably lower V_T and V ˙ CO₂ peak values in the Arm+leg_TTE75 and Leg_TTE75 tests than in the incremental tests. Conversely, when the difference in V ˙ CO₂ peak between the incremental tests and the TTE tests was greatly reduced (i.e. when comparing the incremental tests with the Arm+leg_TTE90 and Leg_TTE90), the difference in V_T peak values decreased accordingly. Although an association between V_T peak and V ˙ CO₂ peak is commonly found when considering different exercise conditions, populations and levels of exercise capacity (Nicolò and Sacchetti, 2023), the relationship between V_T and V ˙ CO₂ is not always proportional because the V_T response is to some extent influenced by the f _R response (see discussion below). Furthermore, we acknowledge that V_T is influenced by various metabolic inputs that have not been measured in this study, including metabolic acidosis (Nicolò and Sacchetti, 2023).

The association between V ˙ CO₂ and V_T helps explain why V ˙ _E resulted to be more sensitive than f _R to changes in exercise tolerance in this study. Indeed, both the magnitude of central command and that of metabolic inputs were probably higher in the Leg tests at isotime, thus increasing f _R and V_T respectively. However, the interpretation of the f _R and V_T responses observed requires careful consideration of the interdependence between the two components of V ˙ _E, which has been advocated to explain the close match between alveolar ventilation and metabolic requirements (Haouzi, 2014). Substantial evidence suggests that V_T is fine-tuned based on f _R levels and the magnitude of metabolic inputs (Nicolò and Sacchetti, 2023), and this notion is reinforced by the present findings. While Figure 7 generally shows a consistent increase in V_T with increases in V ˙ CO₂, the responses of the two variables diverge (i.e. V_T stabilizes or even decreases) when f _R starts to increase at a much steeper rate compared to V ˙ CO₂. Notably, the V_T plateau did not occur at specific values of V_T or V ˙ CO₂, and this is especially evident when considering the TTE tests at 75% of PPO_Leg, where the steeper increase in f _R occurred at relatively low V ˙ CO₂ levels. Although it has been proposed that the V_T plateau that occurs during high-intensity exercise depends on mechanical constraints (Jensen et al., 1980), evidence suggesting this proposition is scarce (Nicolò et al., 2020a; Nicolò and Sacchetti, 2023). Conversely, evidence suggesting that the stabilization of V_T depends to a large extent on the increase in f _R is substantial (Nicolò et al., 2020a; Nicolò and Sacchetti, 2023), and it is even more convincing during TTE tests performed at relatively low intensities, where pulmonary mechanical limitations in healthy individuals may not occur (Nicolò et al., 2020a; Nicolò and Sacchetti, 2023).

Individual responses further suggest that V_T may not change proportionally to V ˙ CO₂ values because it is affected by f _R values. At given V ˙ CO₂ levels, individuals with lower f _R values show higher V_T values and vice versa (see Figure 8). Different combinations of f _R and V_T may guarantee the match between alveolar ventilation and metabolic requirements, and the V ˙ _T fine-tuning feature is supposed to facilitate such a link (Nicolò and Sacchetti, 2023). Hence, individual responses reveal that V ˙ _E is more closely associated with V ˙ CO₂ than V_T, as also found in other exercise protocols (Nicolò et al., 2018; Girardi et al., 2021). The ability of the ventilatory control system to adjust V_T according to changes in f _R has nicely been shown both at rest and during exercise in studies replacing spontaneous breathing with different levels of voluntarily imposed f _R (Lamb et al., 1965; Kennard and Martin, 1984; Haouzi and Bell, 2009; Ohashi et al., 2013; Nicolò and Sacchetti, 2023). Conversely, the ventilatory control system appears not to match metabolic requirements effectively when V_T is imposed and f _R is free to vary (Ohashi et al., 2013). In this perspective, V_T may counteract interindividual differences in f _R and guarantee an appropriate match between alveolar ventilation and V ˙ CO₂ for any values of f _R (Nicolò and Sacchetti, 2023). Our findings reveal the potential of comparing different exercise modalities and paradigms to shed light on the f _R and V_T modulation during high-intensity exercise.

The cardiocirculatory adjustments that occur when exercising simultaneously with the upper and lower limbs may provide further mechanistic support to the improvement in exercise tolerance observed in the Arm+leg modality. Arm+leg cycling may result in a greater peak cardiac output than Leg cycling (Secher et al., 1974; Reybrouck et al., 1975), and the higher V ˙ O₂peak found in the Arm+leg_INC vs. Leg_INC is in line with this notion. Hence, it is conceivable that during Arm+leg_TTE90 and Arm+leg_TTE75 participants were exercising at a lower fraction of peak cardiac output compared to Leg_TTE90 and Leg_TTE75 tests respectively, especially when similar V ˙ O₂ values were found across conditions at isotime. This may have contributed to accommodating the blood flow requests of both arm and leg muscles, thus improving muscle perfusion. Indeed, the reduction in the leg power output observed in the Arm+leg TTE tests vs. the Leg TTE tests of about 20% may have reduced the leg blood flow demand in the Arm+leg modality. In turn, the lower demand of the legs may have delayed the development of leg muscle fatigue and the increase in the magnitude of central command, thus contributing to the improvement in exercise tolerance observed in the Arm+leg modality. While it has been shown that the addition of (intense) arm work to leg work reduces the leg blood flow observed at a given leg power output (Secher et al., 1977; Secher and Volianitis, 2006), the relatively low intensity sustained by the arms in our study may have not impaired leg blood flow substantially.

The fact that f _R and V ˙ _E are considerably more sensitive to changes in exercise tolerance than V ˙ O₂ and HR is particularly evident from the present study. Neither V ˙ O₂ nor HR showed lower values in the Arm+leg tests than in the Leg tests at isotime, despite the lower physical effort and the improved exercise tolerance found in the Arm+leg modality. This is not surprising considering that V ˙ O₂ is to a large extent associated with absolute power output during endurance cycling, although it also depends on other factors, including metabolic efficiency, which is lower for arm cycling compared to leg cycling (Cotes et al., 1969; Vokac et al., 1975; Louhevaara et al., 1990; Itoh et al., 2002). However, only a relatively small portion of the total power output is sustained by the arms during arm+leg cycling, and the oxygen uptake of arm+leg cycling has been reported to be minimally higher than that of leg cycling for a given power output (Hoffman et al., 1996). In our study, the addition of arm work to leg work resulted in a slightly higher or similar V ˙ O₂ in the Arm+leg tests compared to the Leg tests. Likewise, similar or slightly higher values were observed for HR in the Arm+leg modality at isotime, and a higher maximal HR was observed in all the Arm+leg tests. In line with our findings, previous studies had raised concerns about HR monitoring during arm+leg cycling because of the different values of maximal HR and HR relative to RPE/% V ˙ O₂peak that are observed when this modality is compared to leg cycling (Kitamura et al., 1981; Hoffman et al., 1996). Hence, the prescription and monitoring of arm+leg cycling should take this HR response into consideration and may benefit from the concomitant measurement of breathing variables (especially f _R), which is technically feasible even in applied settings (Massaroni et al., 2019; Nicolò et al., 2020b).

The between-modality comparison of exercise tolerance, perceived exertion and V ˙ O₂ shows important practical implications of exercising in the Arm+leg modality. Our findings suggest that this exercise modality allows individuals to nearly double the amount of time spent at a given V ˙ O₂ during constant work rate exercise, or, by extension, to exercise at a higher V ˙ O₂ for the same exercise duration and perceived exertion. This consideration is particularly relevant for exercising individuals interested in maximizing energy expenditure, for those willing to maximize the cardiometabolic stimulus of exercise, and for those interested in lowering effort for a given absolute cardiometabolic stimulus (Hill et al., 2018). Indeed, arm+leg cycling has the potential to increase exercise adherence because a high perceived exertion is commonly viewed as one of the main barriers to exercise participation (Cheval and Boisgontier, 2021), and a relatively low perceived exertion may be associated with a sufficient cardiometabolic stimulus in this exercise modality. Furthermore, considering that exercise tolerance is closely associated with morbidity and mortality (Kokkinos et al., 2010; Nesti et al., 2020), arm+leg cycling may have clinical implications. Arm+leg cycling involves the simultaneous use of arm and leg muscles and may result in a time-efficient training strategy for enhancing both health and performance (Zinner et al., 2017). However, further research is needed to test these propositions.

It is worth mentioning that the Arm+leg modality poses some measurement challenges that we had to face in this study. First, we did not attempt to compute the GET and the RCP because of the relatively high initial stage of the incremental tests imposed by the ergometer (i.e. 150 W), which has limited their detection. As such, the TTE tests were not prescribed based on exercise intensity domains. Hence, especially in the Arm+leg_TTE75 test, it is possible that some participants were exercising in the severe intensity domain while others were in the heavy domain. Second, the measurement of blood lactate is very challenging during the Arm+leg modality, and we decided not to collect it to avoid interfering with the performance tests. Third, perceived exertion can only be rated verbally during the Arm+leg tests, thus generating breathing artifacts and affecting gas exchange measures. However, we used a filtering technique (Lamarra et al., 1987) that addressed this limitation and helped preserve the integrity of the physiological responses. Even the other limitations outlined were partially counteracted by the experimental design and the method of analysis used. Indeed, we performed a detailed between-modality comparison of the responses of some of the main physiological variables commonly used to compute the GET and the RCP. This comparison reveals a greater metabolic perturbation at isotime in the Leg tests, especially when considering the time courses of V ˙ _E, V ˙ CO₂ and P_ETCO2. Hence, it is conceivable that participants were exercising at a lower relative exercise intensity during the Arm+leg tests, not only from an effort perspective but also from a metabolic perspective.

This study shows that exercise tolerance is substantially higher in the Arm+leg modality than in the leg cycling modality. The average improvement in exercise tolerance was 15% in the incremental test and 108% and 90% in the TTE tests at 90% and 75% of PPO_Leg respectively. Perceived exertion, minute ventilation and respiratory frequency were among the most sensitive variables to the improvement in exercise tolerance provided by the Arm+leg modality, hence suggesting that the common mechanism modulating these variables (i.e. central command) plays an important role in endurance performance. These findings reinforce the notion that respiratory frequency is a better marker of physical effort than oxygen uptake and heart rate. Our results have implications for devising exercise strategies to reduce perceived exertion while maintaining a high cardiometabolic demand, and for maximizing energy expenditure for the same level of effort exerted.