Limitation of Maximal Heart Rate in Hypoxia: Mechanisms and Clinical Importance

The use of exercise intervention in hypoxia has grown in popularity amongst patients, with encouraging results compared to similar intervention in normoxia. The prescription of exercise for patients largely rely on heart rate recordings (percentage of maximal heart rate (HRmax) or heart rate reserve). It is known that HRmax decreases with high altitude and the duration of the stay (acclimatization). At an altitude typically chosen for training (2,000-3,500 m) conflicting results have been found. Whether or not this decrease exists or not is of importance since the results of previous studies assessing hypoxic training based on HR may be biased due to improper intensity. By pooling the results of 86 studies, this literature review emphasizes that HRmax decreases progressively with increasing hypoxia. The dose–response is roughly linear and starts at a low altitude, but with large inter-study variabilities. Sex or age does not seem to be a major contributor in the HRmax decline with altitude. Rather, it seems that the greater the reduction in arterial oxygen saturation, the greater the reduction in HRmax, due to an over activity of the parasympathetic nervous system. Only a few studies reported HRmax at sea/low level and altitude with patients. Altogether, due to very different experimental design, it is difficult to draw firm conclusions in these different clinical categories of people. Hence, forthcoming studies in specific groups of patients are required to properly evaluate (1) the HRmax change during acute hypoxia and the contributing factors, and (2) the physiological and clinical effects of exercise training in hypoxia with adequate prescription of exercise training intensity if based on heart rate.

On the other hand, epidemiologic studies have shown that living at altitude (1,500-3,500 m elevation) is associated with lower prevalence of diabetes, obesity and hypertension than living at lower altitude, even after adjusting for multiple risk factors and potential confounders (Sharma, 1990;Woolcott et al., 2014Woolcott et al., , 2016Díaz-Gutiérrez et al., 2016). Intermittent exposure to hypoxia at rest is thus considered as an alternative nonpharmacological therapy (Serebrovskaya and Xi, 2016;Leone and Lalande, 2017), even though a thorough evaluation of its effectiveness has yet to be performed (Wilson et al., 2018). Based on these observations, several authors have suggested that exercise intervention in a hypoxic environment should be used with patients (Chapman et al., 1999;Millet et al., 2016b;Hobbins et al., 2017). Accordingly, the number of studies dedicated to the evaluation of exercise training in hypoxia has dramatically increased in the past few years (Figure 1).
Even if the discussion is still open and an important role seems to be played by the chosen intensity (Noordhof et al., 2013), submaximal oxygen uptake (VO 2 ) at a specific power output is similar at sea level and altitude (Clark et al., 2007). As VO 2max progressively declines with an increase in elevation (Wehrlin and Hallen, 2006), the relative difficulty of exercising at a specific absolute power output will progressively increase as elevation increases. To induce specific adaptations through exercise training in hypoxia, various exercise training programs can be used  where, in some cases, hypoxicinduced adaptive responses would result from reduced absolute training intensity (Hobbins et al., 2017). Hence, exercise training in hypoxic, or "living low-training high, " appears to be a promising method of optimizing training such that participants (e.g., patients with cardiovascular disease, obese people or elderly) receive the maximal metabolic and cardiovascular benefit whilst minimizing injury risk (Bailey et al., 2000;Haufe et al., 2008).
Most studies evaluating the effects of exercise training in hypoxia have used a control group training in normoxia. In healthy subjects, the total work performed (Faiss et al., 2015), or percentage of peak power output achieved in the corresponding environment (Wang et al., 2010) have been used to match the exercise intensity between the experimental and control groups. However, in studies involving patients, the intensity of the training program is often based on the percentage of maximal heart rate (HR max ; Netzer et al., 2008;Kong et al., 2014;Park and Lim, 2017) or heart rate reserve (HRR; Greie et al., 2006;Mao et al., 2011;Schreuder et al., 2014;Gutwenger et al., 2015). Heart rate (HR) monitoring is practically useful and very common with athletes and patients (Achten and Jeukendrup, 2003;Casillas et al., 2017). But, as for the debate on the correct metric to evaluate the hypoxic dose (Garvican-Lewis et al., 2016;Millet et al., 2016a), no consensus yet exists on the best way to tailor exercise training intensity in hypoxia. In a recent review dealing with exercise training in hypoxia for overweight/obese subjects, the authors suggested setting the intensity at 60-70% of HR max (Hobbins et al., 2017). With healthy people, an intensity target between 75 and 95% of HR max was proposed (Vogt and Hoppeler, 2010).
However, the authors did not mention if HR max used in the calculation of the target intensity is the HR max obtained at sea level or at the considered altitude. This is of importance as HR max may be altered in hypoxia. It is well established that HR max is decreased during prolonged (several days or weeks) exposure to a high altitude such as encountered during mountaineering (Christensen and Forbes, 1937;Richalet et al., 1992;Lundby et al., 2001b). On the contrary, the change under acute hypoxic exposure is still debated, particularly at moderate levels of hypoxia. Studies have shown that HR max is (e.g., Grataloup et al., 2007) or is not (e.g., Gallagher et al., 2015) significantly reduced in acute hypobaric and normobaric hypoxia (see Supplementary Table S1). Some researchers suggest that a decreased HR max with altitude is specific, with a threshold around 2,000-3,500 m (Dill et al., 1966;Martin and O'Kroy, 1993;Benoit et al., 2003;Calbet et al., 2003a;Fukuda et al., 2010;Gallagher et al., 2015), i.e., typically the altitude chosen for training/rehabilitation purpose (see Supplementary Table S2). However, the exact altitude at which HR max is decreased is not well known and it is likely that, similar to VO 2max , the decrement in HR max starts at low altitude (below 1,000 m; Gore et al., 1996Gore et al., , 1997Robergs et al., 1997). Several factors may account for this lack of a consistently observed HR max decline within increasing altitude such as exercise-induced hypoxemia (Benoit et al., 2003;Grataloup et al., 2007), training status (Kjaer et al., 1988;Benoit et al., 2003;Dufour et al., 2006;Roels et al., 2007;Mollard et al., 2007b) or sex (Shephard et al., 1988;Robergs et al., 1997) as well as the robustness of statistical analyses (usually small numbers of participants and high inter-individual variability in the HR response). Also, most of the previous data were obtained with healthy young participants, and thus extrapolation of this to older people or patients with limited capacities is unknown.
As shown in Figure 1, the number of publications related to exercise training in hypoxia amongst both healthy subjects and patients is increasing, highlighting the scientific and clinical interest of such exercise training. Hence, this literature review presents an examination of the available literature on HR max decrement with increasing hypoxia in both healthy subjects and patients, with the main aim of providing a general understanding of the use of HR max as a benchmark for exercise training prescription in hypoxia.

METHOD
A literature search was made up until March 2018 using the following databases: Pubmed, ScienceDirect, Scopus, SportsDiscus and Web of Science. Search terms were used to restrict studies to those involving humans under hypoxic conditions, including a combination of either "altitude" or "hypoxia" or "oxygen fraction" or "oxygen delivery" and each of the following: "exercise, " "exercise training, " "Live Low-Train High, " "intermittent hypoxia, " "incremental, " "GXT, " "graded, " "heart rate, " "exhaustion, " "patient, " "disease." Each title, abstract and full text were assessed for relevance to the topic and selected if they met the following inclusion criteria: involving humans; normoxia and hypobaric or normobaric hypoxia conditions; hypoxic level expressed as altitude (m) and/or inspired oxygen fraction (%). A classification was first completed to select articles reporting HR max in both hypoxia and normoxia. To do so, the common criteria to assess that maximal intensity was reached at the end of the incremental test to exhaustion were considered (Howley et al., 1995) except for older ones within which the authors ′ statement of "maximal intensity" was deemed adequate. Also, only acute exposure to hypoxia, i.e., the maximal test should have been performed within 8h of exposure to the hypoxic environment was taken into consideration. A second classification was made including studies using HR to monitor the training intensity. References of articles that fulfilled the criteria were also scanned for further relevant studies that were included if they met the inclusion criteria.
The initial search yielded a total of 2,077 publications. From these, 1,991 publications were excluded. Finally HR max data were collated from 81 publications involving healthy subjects and 5 publications involving patients (Supplementary Table S1). In five of the reviewed articles, the data were not reported in the text/Tables but in figures. In these cases, the data were extracted using the Digitizelt software (v2.2, BormiSoft2, Braunschweig, Germany). Thirteen publications with healthy subjects and 9 with patients compared exercise training in hypoxia vs. exercise training in normoxia matching the intensity with HR (Supplementary Table S2).
Due to the lack of standardization of study designs (the most important of which being the hypoxic stimulus, but also the type of disease, the age or gender of participants, the type of incremental test), a systematic assessment of hypoxic intervention was not done.

DECREASE IN HR MAX IN ACUTE HYPOXIA: A UNIVERSAL RESPONSE?
In high altitude (as defined elsewhere Bärtsch and Saltin, 2008), previous studies have shown that the decrement in VO 2max could not be fully explained by the reduction in arterial oxygen content (CaO 2 ) (Calbet et al., 2003a). It has been calculated that at 5,300 m, the VO 2max decrease could be explained by a two-thirds reduction in CaO 2 and one-third reduction in peak muscle blood flow and cardiac output (Calbet et al., 2003a). When subjects are exposed to acute hypoxia, the lower arterial oxygen saturation (SaO 2 ) is usually associated with lower HR max (Benoit et al., 2003). Reduction in HR max with increasing altitude and with time spent in altitude has been observed for a long time (Christensen and Forbes, 1937), with a HR max of 127 bpm reported at an altitude equivalent to 8,848 m (Cymerman et al., 1989). During acute exposure, HR max has long been believed to remain stable below simulated altitudes of 4,500 m (Cerretelli et al., 1967;Saltin et al., 1968;Fagraeus et al., 1973;Knuttgen and Saltin, 1973;Young et al., 1982;Escourrou et al., 1984;Bouissou et al., 1986;Kjaer et al., 1988;Lawler et al., 1988;Shephard et al., 1988;Hughson et al., 1995). In contrast, other studies have strongly argued in favor of decreased HR max during acute hypoxia (Dill et al., 1966;Drinkwater et al., 1979;Roach et al., 1996;Lundby and van Hall, 2001) even at lower altitudes. Lundby and van Hall suggested that HR max remains stable up to 3,100-3,300 m with a linear decrease thereafter, but this result was obtained in only 5 healthy subjects without a thorough evaluation below that threshold (Lundby and van Hall, 2001). The current literature research involving humans performing maximal incremental tests to exhaustion showed that HR max reduces with acute normobaric or hypobaric hypoxia (Supplementary Table S1 and Figure 2, showing group average data). When the data are pooled together, the overall tendency is that HR max starts to decrease as soon as altitude increase. This is accordance with the observation made at as low as 600-700 m above sea level (Gore et al., 1996). No clear threshold in the decrement is observed. Instead, a roughly linear trend seems to exist, but there is large inter-study variation. Together with the large inter-individual variation within studies, this likely explains why the reported differences in the literature are not always statistically significant (Supplementary Table S1). It should be acknowledged that likely more than one mathematical model could fit the data, as observed in other circumstances (Garvican-Lewis et al., 2016). Instead of absolute altitude per se, the altitude gain between the two incremental tests is likely to be more meaningful ( Figure 2B). In that case, the relationship between the two variables could be modeled by the following FIGURE 2 | Change in maximal heart rate with altitude (A) or altitude gain (B). Each data point represents the average value from the relevant group, as presented in previously published studies (references in Supplementary Table S1). equation: HR max (bpm) = −0.0024 × altitude (m) + 0.73. Hence, a 1,000 m gain will reduce HR max by 1.7 bpm. This is in agreement with, e.g., the 1.9 bpm per 1,000 m found in young subjects by Wehrlin et al. and with the 9 bpm decrease of HR max observed at 4,300 m (Lundby et al., 2001a). It implies that if participants usually train at 200 m and performed a hypoxic training session at 2,500 m, the expected decrement in HR max will be 5 bpm. Richalet (1990) proposed an equation that describes the decrease in the chronotropic drive during hypoxia to predict HR max : y (% of sea-level HR max ) = 116-0.0057x (x = altitude in meters). Compared to our findings, this model tends to underestimate the decrease at low altitudes and overestimate the decrease at high altitudes (Figure 3). But the original data used to obtain the model are somewhat different, and includes chronic hypoxic exposure, which is known to decrease HR max (Favret and Richalet, 2007).

Effect of Age
The large variability in the decrement in HR max with altitude could be explained by several factors. Most of the studies involved healthy young adults (between 20 and 39 years old as shown in Figure 4). Obviously, the general trends show a progressive decrease in HR max with altitude whatever the age groups, without marked differences (Puthon et al., 2017). A trend toward a slightly greater decrease is observed in 30-39 years vs. 20-29 years. In Figure 4, 14-19 years and 50-59 years age groups show different behaviors than 20-29 years and 30-39 years age groups, but likely due to small sample sizes and to the fact that for these ages groups, participants were patients for whom only few data are available (Supplementary Table S1). Hence, if HR max decreases with age (Lhuissier et al., 2012), age per se does not seem to alter the decrement in HR max with acute hypoxia (Puthon et al., 2017), but this needs to be confirmed by specifically designed studies. FIGURE 3 | Maximal heart rate at altitude: comparison between the measured data (see Supplementary Table S1) and data modelised according to Richalet (1990). Each data point represents the average value from the relevant group, as presented in previously published studies (references in Supplementary Table S1).
It also suggests that the hypoxia effects will be relatively greater for an older subject compared to a younger one. For example at 2,500m, using the equation presented in Figure 2, the decrement in HR max is 5 bpm (absolute value). For a 20 years old subject, this represents a decrease of 2.5% (assuming HR max = 200 bpm in normoxia). For a 70 years old subject, this represents 3.3% (assuming HR max = 150 bpm in normoxia). As older subjects already have cardiac limitations, this further decrease may result in an insufficient HR to maintain performance and a tendency to reach exercise intolerance at the suggested exercise target HR.

Effect of Sex
Most of the time, both male and female are mixed in the studies and only a few studies specifically focused on possible sex differences. Both females and males experience a decrease in HR max with altitude (Horvath et al., 1988;Shephard et al., 1988;Gore et al., 1997;Mollard et al., 2007b) without marked difference between sexes.

Effect of Fitness/Training Status
The magnitude of the HR max decrease is more likely dependent on the training status. When subjects were grouped based on their training characteristics (i.e., healthy untrained vs. trained subjects as characterized by the authors of the studies), the decrement in HR max appears a little more pronounced for untrained than trained subjects ( Figure 5A); HR max (bpm) = −0.0026 × altitude (m) + 1.2486 and HR max (bpm) = −0.0025 × altitude (m) + 1.4263, respectively. However, this observation is based on studies using different testing and hypoxic conditions and should be viewed with caution. Also, the definition of the "untrained" and "trained" states by the authors has not been harmonized as it should be (De Pauw et al., 2013), so that the same VO 2max was considered "trained" in some studies and "untrained" in others when VO 2max was reported. When the decrease in HR max with altitude takes VO 2max into account (Figure 5B), the higher the VO 2max , the greater the decrease in HR max . This is consistent with the results in the few studies that directly compared trained and untrained subjects. They usually report a larger decrease in HR max (Kjaer et al., 1988;Martin and O'Kroy, 1993;Benoit et al., 2003;Marconi et al., 2004;Mollard et al., 2007b,c) even if not universally observed (Lawler et al., 1988;Gore et al., 1996;Ferretti et al., 1997;Ofner et al., 2014). Hence, HR max reduction with acute hypoxia appears to depend on the training status, even if the reason for this decrement remains unclear (Mollard et al., 2007b). It should be noted that athletes involved in studies reporting results from incremental tests to exhaustion in hypoxia are usually aerobically-trained athletes. It is currently unknown whether another type of training background could alter the HR max response to hypoxia, while exercise training in hypoxia is more and more commonly used to improve intermittent and strength performance (Millet et al., 2010;Saeed et al., 2012;Brocherie et al., 2017;Feriche et al., 2017;Girard et al., 2017), even if for these activities HR max is less meaningful.

Effect of Pathology
In patients/individual with a low VO 2max (14-39 ml/min/kg), the trend is completely different with a paradoxical grseater decrease in HR max at the lowest reported altitude (Supplementary Table S1; Figure 5). However, HR max in normoxia and hypoxia were reported in only 6 studies (9 different groups/altitude) limiting the relevance of this finding. Interestingly, in one study with heart failure patients, an unexpected increase in HR max of 8 bpm was reported at 3454 m compared to 540 m (Schmid et al., 2015). The patients may likely have "learned" the incremental test procedure, as all tests in normoxia were performed before the hypoxic exposure. Also, a placebo effect could not be ruled FIGURE 4 | Change in maximal heart rate with altitude: Effect of age. Each data point represents the average value from the relevant group, as presented in previously published studies (references in Supplementary Table S1).
out because of the natural hypobaric hypoxia exposure studied (Jungfraujoch Mountain). Finally, peak HR, and not HR max , is usually recorded in patients due to the difficulty of obtaining the "true" HR max . Such misinterpretation can explain these paradoxical observations. In Figure 5, patients with different diseases were merged (e.g., chronic heart failure, Fontan patients, coronary artery disease) due to the small number of studies available. To confound results further, it is unknown if the specific pathology of these patients and the pharmacological agents they take modulate the HR max response with altitude. It has been recently highlighted that hypoxic exposure is safe for patients with cardiovascular diseases if specific recommendations are followed (for more details see Parati et al., 2018). However, the authors also highlight that healthy subjects taking carvedilol, a non-selective beta-blocker, display significant reductions in blood pressure responses to altitude, which was associated with reduced SaO 2 and exercise tolerance. Whenever possible, these authors proposed ß1 selective should be used in place of non-selective beta-blockers in chronic heart failure patients. Nevertheless, experiments that involve the discontinuation of drugs that have chronotropic effects on the heart, need to be performed at altitude to adequately address this question.

Effect of Genetics
The effects of hypoxia on arterial oxygenation, VO 2max and HR max in lowlanders depends on genetic factors (Masschelein et al., 2015). In this latter study, the authors found extremely high similarity amongst twins. The drop in HR max ranged 3 and 27 bpm, but the variance within twins was approximately sevenfold less than between twins. This suggests that HR responses during exercise in acute hypoxia are dependent on genotype, but here again more research is needed.

Cardiac Function
The decrease in HR max can be considered as a limiting factor of performance at altitude. However, a lower HR max associated with a decreased cardiac output, could alleviate the limitation in diffusion, and thus the decrement in SaO 2 and VO 2max . Consistent with this is the greater decrease in HR max in acute hypoxia in endurance athletes, who have the most important diffusion limitation (Dempsey et al., 1982). This mechanism is also protective for the heart, an organ that depends on oxygen supply too. During maximal exercise in healthy subjects, the coronary flow reserve has been shown to be limited to onethird above what is prevailing at sea level (Kaijser et al., 1990). Therefore, the compensation of decreased CaO 2 by increasing coronary blood flow is not possible above a certain altitude. The only option to preserve cardiac integrity is thus to decrease myocardial oxygen demand and therefore HR max . All these changes in cardiac chronotropic activity do not seem to be associated with alterations in the inotropic function (Boussuges et al., 2000;Boushel et al., 2001). In healthy individuals, left ventricular function and myocardial oxygen supply was found to be maintained during maximal exercise at an altitude of 7,625 m, at which HR max was reduced by 20% (Sutton et al., 1988).
Thus, evidence strongly suggests that in moderate acute hypoxia, HR max is reduced. It should be acknowledged that large inter-study (reporting themselves large inter-subject) variability exists. But this decrement should be clearly specified, as it has been previously reported in a review paper, which focused on the use of exercise training in hypoxia that hypoxia trigger FIGURE 5 | Decrement of maximal heart rate with altitude: Effect of physical fitness (A) groups based on subjects' training characteristics. (B) groups based on maximal oxygen uptake). Each data point represents the average value from the relevant group, as presented in previously published studies (references in Supplementary Table S1).
an increase in both basal and maximal HR max , which is contradictory to the current observation (Urdampilleta et al., 2012). However, the mechanisms leading to this HR max decline remain to be identified.

Arterial Oxygen Saturation
As reported (Supplementary Table S1, Figure 5), untrained subjects usually have a lower decrease in HR max with altitude than trained subjects. The lower reduction in SaO 2 generally observed in these subjects could be an explanation (Lawler et al., 1988;Martin and O'Kroy, 1993;Robergs et al., 1998;Mucci et al., 2004). During training, the repetition of long duration exercises by subjects who exhibit arterial oxygen desaturation could induce a situation similar to a chronic exposure to hypoxia. As with chronic exposure to hypoxia, chronic exercise reduces the number of cardiac ß-receptors (Werle et al., 1990). The repeated arterial oxygen desaturation with training could interfere with the effect of acute hypoxia exposure. It has been suggested that in normoxia, subjects with exercise-induced arterial hypoxemia (EIH) could present a lower HR max due to the repeated exposure to exercise hypoxemia (Zavorsky, 2000). Even if not always reported (Chapman et al., 1999), subjects with EIH usually have a greater reduction in HR max with a lower SaO 2 than subjects without EIH (non-EIH) (Benoit et al., 2003;Grataloup et al., 2007). Also, a correlation exists between SaO 2 in hypoxia and the reduction in HR max (Grataloup et al., 2007). SaO 2 could thus be considered as a determining factor in the HR max decline with acute hypoxia even if evidence suggests that only low values of SaO 2 could alter HR max (Grataloup et al., 2007). In the current literature search, we found 322 groups for whom both SaO 2 and HR max at the end of an incremental exercise to exhaustion in normoxia and hypoxia were reported. When the change in HR max was plotted vs. the percentage of change in SaO 2 (Figure 6), the hypothesized trend seems to be confirmed: overall the greater the desaturation, the greater the decline in HR max . However, the correlation remains low (Grataloup et al., 2007), with large variability in the decrease in HR max for a given decrease in SaO 2 , highlighting that other contributors alter HR max in hypoxia.

Myocardial Electrophysiology
Potential explanations for the reduced HR max include myocardial dysfunction, as suggested by Benoit et al. (2003). A direct effect of hypoxia on the cardiac electrophysiological properties including repolarization length and transmission time across the auriculo-ventricular node, occurs and relates to a reduction in HR max (Roche et al., 2003;Siebenmann and Lundby, 2015). The greater hypoxemia observed in EIH subjects in the hypoxic condition could induce greater modifications of electrophysiological cardiac properties (Grataloup et al., 2007).

Autonomic Nervous System
Another contributor, likely the major one, seems to be the change in the activity of the autonomic nervous system. The reduction in HR max in hypoxia has been linked to an alteration in sympathetic and parasympathetic activity mediated chronotropic function, with a densitization of the adrenergic pathway and/or an upregulation of the parasympathetic system (Hartley et al., 1974;Richalet et al., 1988Richalet et al., , 1992Kacimi et al., 1993;Boushel et al., 2001;Favret and Richalet, 2007). But the effect of acute hypoxia on the autonomic nervous system is not as clear (Rostrup, 1998;Naeije, 2010). Sympathetic traffic may be slightly decreased (Bouissou et al., 1986;Seals et al., 1991), unchanged (Lundby et al., 2001a) or may increase (Hughson et al., 1995) during maximal exercise in acute hypoxia. Using domperidone, a D2-receptors blocker, Lundby et al. (2001b) demonstrated that hypoxic exercise in humans activates D2-receptors with a subsequent decrease in circulating noradrenaline. However, HR max was unaffected by domperidone, indicating that dopamine D2-receptors are not involved in the HR max decrement (Lundby et al., 2001b). In that study, marked changes in plasma noradrenaline levels had no effect on HR max . This points toward a postsynaptic located mechanism. Oxygen breathing completely reversed the decrease in HR max to values not different from those at sea level. This led the authors to conclude that uncoupling, and not downregulation, of cardiac adrenoreceptors is responsible for the early decrease in HR max during hypoxic exercise. However, these results were obtained after 24 h (Lundby et al., 2001b) or a few days of acclimatization . During acute hypoxia, the involvement of the autonomic nervous system requires further study. Although the previous changes in receptor function could be rapidly triggered by exposure to hypoxia and high adrenergic activity, it is unlikely that these mechanisms are involved in a short exposure of only a few minutes/hours. It has been hypothesized that an increase in parasympathetic activity, rather than decreased sympathetic activity, might be responsible for the blunting of the cardiac chronotropic function during acute hypoxia (Lundby et al., 2001a). Indeed, HR max can be partly restored toward sea level values by administration of atropine, a FIGURE 6 | Change in maximal heart rate with altitude as a function of arterial oxygen saturation: Effect of physical fitness. Each data point represents the average value from the relevant group, as presented in previously published studies (references in Supplementary Table S1). parasympathetic antagonist (Hartley et al., 1974;Bouschel et al., 1999;Bogaard et al., 2002). Moreover, previous studies have concluded that plasma norepinephrine tends to underestimate sympathetic traffic during acute hypoxia (Escourrou et al., 1984), due to increased reuptake of norepinephrine compared to normoxia (Leuenberger et al., 1991). Therefore, it seems unlikely that the decreased HR max in acute hypoxia is related to a decrease in sympathetic activity (Naeije, 2010). Mechanistically, cardiac output can be blunted by direct hypoxic excitation of medullary neurons, which project to cardiovagal neurons, causing bradycardia (Golanov et al., 2000). Taken together, these results suggest that both sympathetic and parasympathetic activity is increased during maximal aerobic exercise in acute hypoxia. When the two efferent arms of the autonomic nervous system are activated, the parasympathetic activity generally dominates at the heart level (Levy, 1971;Mendelowitz, 1999), in accordance with the decrease in resting HR observed during cold water immersion despite a large increase in catecholamines in healthy subjects (Mourot et al., 2007). This could be a valuable explanation of the HR max decline with increasing acute hypoxia levels. Another explanation of the reduced HR max could be that hypoxia reduced myocardial contractility and/or the rate of diastolic filling (Shephard et al., 1988), but this remains to be confirmed (Boussuges et al., 2000;Boushel et al., 2001).

Central Factors
Since the peak power output achieved is also reduced in hypoxia, an alternative hypothesis could be that HR max is lowered due to a decrease in exercise effort. However, maximal plasma lactate and catecholamine concentrations (Stenberg et al., 1966;Bouissou et al., 1986;Lundby et al., 2000Lundby et al., , 2001a as well as perceptual responses (Horstman et al., 1979) at the termination of exercise are often similar at altitude and sea level. Maximal skeletal muscle VO 2 could also dictate maximal cardiac output in hypoxia (Wagner, 2000). Arterial hypoxemia may induce a decrease in muscle VO 2max , which in turn determines a decrease in cardiac output. This is in agreement with the association between the change in HR max and the change in VO 2max found in many studies . A reduced central drive to the cardioacceleratory centers could also be suggested. Chronic (Kayser et al., 1994) and acute hypoxia (Peltonen et al., 1995) do not alter neuromuscular fatigue, and hypoxia-induced systemic adaptations are not triggered by neural feedback from working muscles (Kjaer et al., 1999). This eliminates a potential peripheral effect. Most probably, a decrease in oxygen content is perceived by the central nervous system. Otherwise, the increase in cerebrovascular blood flow (Harik et al., 1996) would be a unsuitable adaptation. Two studies support the notion of a central nervous system limitation to exercise performance in hypoxia equivalent to 5,300 m (Calbet et al., 2003a,b). During an incremental test in normoxia, task failure is most likely caused by peripheral mechanisms while central mechanisms seem to prevail in severe hypoxia (Calbet et al., 2003a;Subudhi et al., 2007). These explanations are in accordance with the hypothesis of a "central governor" that regulates the mass of skeletal muscle recruited during exercise and protects the heart from ischemia (Noakes et al., 2001). With, as a consequence, a decreased in HR max . This is agreement with the fact that the changes in maximal power and maximal cardiac output are very similar . Noakes' hypothesis is that the oxygen tension in the coronary vascular bed is the monitored variable by the "central governor" to prevent a progressive myocardial ischemia (Noakes et al., 2001). However, this hypothesis has certain limitations. The results of previous studies suggest that the signals that inhibit the increase of HR in acute hypoxia may arise somewhere else than in the myocardium itself, likely the central nervous system (Kayser et al., 1994;Harik et al., 1996;Peltonen et al., 1997;Kjaer et al., 1999;Calbet et al., 2003a,b). Also, the fact that the drop in HR max appears at low altitude (below or at 1,000 m; Gore et al., 1996Gore et al., , 1997Robergs et al., 1997) indicates that it is improbable that such a low altitude would constitute a significant challenge for the central nervous system.
The inter-individual variability in the decrease in HR max with the hypoxic stimulus is important and the underlying mechanisms still needs to be specified. Nevertheless, this decrease has practical effects that should be taken into consideration, especially when prescribing hypoxic exercise for patients.

CONSEQUENCES IN CLINICAL PRACTICE
As seen in Figure 1, an increasing number of research studies focuses on exercise training in hypoxia for patients, with authors strongly arguing about its potential superior effects compared to exercise in normoxia (Wee and Climstein, 2015;Millet et al., 2016b;Hobbins et al., 2017;Thiersch et al., 2017). In hypoxia, the prescription of exercise intensity requires considerable attention due to the reduction in aerobic capacity (Friedmann et al., 2005). Hence, a proper evaluation of the superiority of a hypoxic training program should be done to decide whether or not a patient should be involved in this type of exercise environment. While dealing with hypoxic training to increase an athlete's performance, it is highly recommended to strictly control potential bias such as the placebo effect and to compare similar training programs to ensure that results obtained are due solely to hypoxia (Lundby et al., 2012).

Use of HR for Exercise Prescription
In studies which compare the hypoxia and normoxia results in patients, prescription of the intensity of an aerobic-based exercise training program is often based on percentage of HR max (Netzer et al., 2008;Kong et al., 2014;Park and Lim, 2017) or HRR (Greie et al., 2006;Mao et al., 2011;Schreuder et al., 2014;Gutwenger et al., 2015) (Supplementary Table S2), in agreement with guidelines (Pescatello et al., 2004(Pescatello et al., , 2015. Using this approach, studies involving patients and healthy subjects that used a percentage of theoretical or normoxic HR max /HRR have found additional effects of hypoxic vs. normoxic training (Supplementary Table S2). Hendriksen and Meeuwsen, 2003) compared training at 60-70% of HRR at sea level and 2,500 m and found that intermittent hypobaric training can improve the aerobic and anaerobic energy supplying system. Kong et al. (2014) showed that hypoxic training between 2,000 and 3,000 m at 60-70% of HR max caused more weight loss than normoxic training in obese young adults. Wiesner et al. (2010) reported a better response in physical fitness, metabolic risk markers, and body composition in obese subjects after comparing training completed at the same absolute HR (HR at 65% of VO 2max in normoxia). The absolute HR were identical in both normoxic and hypoxic groups in the study of Bailey et al. (2000) who reported an additive cardioprotective effect of normobaric hypoxia over normoxic training. This apparent superiority has led authors of reviews dealing with exercise training in hypoxia to suggest using specific percentages of HR max (e.g., 55-85% HR max for Wee and Climstein, 2015, and 60-70% of HR max for Hobbins et al., 2017). Unfortunately, they did not specify whether the HR max was the normoxic or hypoxic one. Since HR max decreases with altitude, the distinction between normoxic and hypoxic HR max is important and has several practical implications.
The altitude commonly used for exercise training with patients is often 2,000-3,500 m; (Haufe et al., 2008;Netzer et al., 2008;Wiesner et al., 2010;Kong et al., 2014;Schreuder et al., 2014;Gutwenger et al., 2015;Park and Lim, 2017;Pramsohler et al., 2017). As highlighted elsewhere (Schreuder et al., 2014), for this altitude range VO 2max under hypoxia is approximately 91% of the value achieved under normoxia. It means that HR max under hypoxia is reached at 91% of maximal VO 2 -uptake at sea level. Therefore, training at 70-75% of the HRR under hypoxia corresponds with an absolute level of oxygen uptake that is already achieved at 64-68% of the HRR when exercise is performed under hypoxia. Hypoxia during exercise shifts ventilation, HR and blood lactate concentration vs. intensity curves to the left compared with sea-level values (Friedmann et al., 2004). In addition, the blood lactate concentration vs. HR curve shifts to the left, especially at high HR and, consequently, training HR may be decreased during training in hypoxia (Levine and Stray-Gundersen, 1997;Peltonen et al., 2001).

Magnitude of the Effect of Hypoxia
Based on Figure 2, the decrease of HR max could be expected to be 5 and 7.5 bpm at 2,500 and 3,500 m, respectively. If we consider the author of this manuscript (40 years-old and a HR max at sea level of 181 bpm), an overestimation of 3 bpm (2.8%) and 5 bpm (2.8%) of target HR for moderate (40% HR max ) and vigorous (90% HR max ) intensity exercise, respectively, would have been observed for an exercise training session at 2,500 m. At 3,000 m, it would have been 4 bpm (3.5%) and 6 bpm (3.5%), respectively. Focusing only on the altitude gain ranging from 2,500 to 3,500 m (Supplementary  Table S1), the average HR max decrement is 4.2 ± 3.9 bpm. In that case, the overestimation is 3 bpm (2.3%) and 4 bpm (2.3%) for moderate and vigorous exercise, respectively. Also, resting HR increases under hypoxia due to increase sympathetic and decrease parasympathetic activity (Windsor et al., 2010;Siebenmann and Lundby, 2015). This increase could reach 4-5 bpm (Friedmann et al., 2004;Subudhi et al., 2011;Horiuchi et al., 2017), even at low altitude (equivalent to only 1,650 m for (Subudhi et al., 2011)). Knowing the resting HR of the author (55 bpm), the overestimation could have been 2 bpm (1.9%) for moderate (40% HRR) and 4 bpm (2.6%) for vigorous (85% HRR) intensity exercise at 2,500 m. The corresponding values would have been 2 bpm (2.4%) and 5 bpm (3.3%) at 3,500 m.
Finally, by using the average decrement of 4.2 ± 3.9 bpm of HR max , the overestimation would have been 1 bpm (1.6%) and 3 bpm (2.2%) for moderate and vigorous intensity calculated from HRR. Using the data presented in Figure 5 for VO2 max group 50-59 ml/min/kg, the overestimation is roughly the same for each given example. However, using the data presented in Figure 4 for 30-39 years age group, the overestimation increases by 1% in each example that has been provided. The overestimation calculated in relation to this particular case is consistent with those calculated from the data obtained in healthy subjects in previous studies (Friedmann et al., 2004;Horiuchi et al., 2017;Subudhi et al., 2011). Hence, depending on the calculation, a systematic overestimation of the intensity by 1-4% would be obtained in hypoxia. Fortunately, in this specific case, HR max is very close to the predicted HR max indirectly derived from the age. But such derived HR max could lead to large error in the prediction (Robergs, and Landwher, 2002;Gallagher et al., 2015). At the group level with the data obtained in Supplementary Table S1, the calculated 220-age HR max was 7.8 ± 9.1 bpm higher than the measured HR max in normoxia, and 14.3 ± 10 bpm higher in hypoxia. Using hypobaric or normobaric hypoxia could also be of importance, since a significantly higher resting HR in hypobaric hypoxia compared to normobaric hypoxia have been reported (Savourey et al., 2003). The importance of such difference as well as the underlying mechanisms has to be specified (Beidleman et al., 2014;Richard et al., 2014;Saugy et al., 2016;Bourdillon et al., 2017;Woods et al., 2017).

Practical Consequences of the Overestimation
In normoxia, the intraclass correlation coefficients (ICC) for submaximal and maximal HR could be considered as high, since they have been reported to be greater than 0.9, the higher the HR, the lower the variation (Lamberts et al., 2004;Tulumen et al., 2011). But expressed in bpm, the day-to-day variations in submaximal HR could be estimated to around 5-8 bpm (Lamberts et al., 2004). It means that the difference in HR max between normoxia and hypoxia that could be expected at 2,500 m (Figure 2 and specific example given above) could be hidden by the spontaneous fluctuations of HR for a given training load. Hence, this difference could be considered as being of low importance. But because there is actually no reason to consider that the day-to-day fluctuation of HR in hypoxia would be different from the variation in normoxia, it remains that hypoxia leads to a decrease in HR max . Such overestimation, even if of moderate magnitude, could lead to the prescription of exercise intensities difficult to reach or even unachievable by the participant. It may result in decreased adherence to exercise training sessions (Perri et al., 2002). Alternatively, in these studies the hypoxic groups would have exercised with higher intensities than the control group and that studies comparing the effect of exercise training in hypoxia to normoxia may not be simply comparing a difference in the inspired oxygen fraction, but also a difference in (relative) exercise intensity. Since this favors a systematic higher exercise intensity in hypoxia, it casts doubt on the validity of the previously reported results. Training at higher intensities than sea-level could trigger a state of fatigue. Even without the overestimation, the risk of overloading exists in hypoxia. As an example, Debevec et al. (2010) matched 20 training sessions at an HR level corresponding to that elicited at 50% of peak power output attained in normoxia or hypoxia, respectively. The group who trained in hypoxia experienced a significant decrease in HR max during the normoxic incremental test to exhaustion post-training. They suspected a possible state of fatigue for this group and highlighted the need pay attention to the sum of exercise and the hypoxic stimulus (Ventura et al., 2003;Friedmann et al., 2005). Using HR max at sea level or predicted from age during physiological assessment could also affect the conclusions of physiological studies. For example, HR relative to 220-age as an index of the intensity of exercise was used to understand the effect of altitude on microvascular parameters in both environments, thus overestimating the exercise intensity in hypoxia (Bauer et al., 2006). On the other hand, as no real adverse effects have been reported for this range of exercise intensity/altitude, this overestimation underlined that hypoxic exercise training seems well tolerated by patients (Agostoni, 2013;Levine, 2015;Parati et al., 2018) and provides a certain level of flexibility for the execution of the training program.

CONCLUSIONS AND FUTURE PERSPECTIVES
A body of scientific literature exists on the evaluation of passive exposure and/or exercise training in altitude/hypoxic environment in healthy subjects. Despite this, debates still exists on the benefits and physiological effects of hypoxic training. Nevertheless, the use of exercise training in hypoxia has grown in popularity with patients. Studies have provided encouraging results, but the effectiveness and potential superiority of hypoxic training over normoxic training remains to be established. To achieve this, the exercise training programs performed by the experimental and control groups should be comparable to allow a rigorous evaluation of the isolated hypoxic stimulus. The best metric to quantify the exercise training load needs to be specified, but the prescription of exercise for patients largely rely on HR recordings. This review underlines that HR max decreases progressively with increasing hypoxia. The dose-response is roughly linear and starts at a low altitude. However, the literature reports large inter-study and interindividual variabilities. Since the studies are usually conducted with a limited number of participants, this hampers findings of significant difference compared to sea-level especially at low altitudes where the change is limited. Sex and age do not seem to be major contributors to the HR max decrement with altitude. Rather, it seems that the greater the reduction in SaO 2 , the greater the reduction in HR max , likely due to an over activity of the parasympathetic nervous system. Only a few studies reported HR max at sea/low level and altitude with patients. Together with very different study designs, it is difficult to draw firm conclusions amongst these specific populations. The magnitude of the decrease in HR max seems to be augmented after days at high altitude (acclimatization) and therefore also depends on the duration of hypoxic exposure. It is not known if over the course of a training program HR max decreases in the same way, even if unlikely (a huge difference between daily exposure on one side and 3-5 sessions of 30-90 min per week exists in terms of hypoxic exposure). Hence, much remains to be done in this field and forthcoming studies with appropriate designs need to be conducted to answer questions in both the physiological and clinical area. In the first category, the effects of hypoxia on HR max (but also on resting HR which appears of importance for prescription based on HRR) amongst specific populations have to be better addressed, with a specific focus on sex and age. Also, it is of importance to better assess these effects in specific pathologies (in particular cardiovascular and metabolic diseases), as well as determining potential interaction with drugs. The difference between normobaric and hypobaric hypoxia should also be addressed. In all of these situations, the specific impact of SaO 2 and the autonomic nervous system should be taken into consideration. In the second category, a thorough evaluation of the potential superiority of exercise training in hypoxia over normoxia on relevant outcomes need to be done with a wellcontrolled design that involves accounting for the placebo effect and an adequate prescription of exercise training intensity based on heart rate. Under conditions of hypoxic exercise training, the long-term effect of hypoxia exposure on HR max and the potential need to adjust the relative HR target should also be addressed as well as the need to account for chronotropic effects of drugs. In some years, a review similar to the present one would be very instructive.

AUTHOR CONTRIBUTIONS
LM: Lead author, study writing. The manuscript presents clearly results, honestly, and without fabrication, falsification, or inappropriate data manipulation. Results of the present study do not constitute endorsement by ACSM.

ACKNOWLEDGMENTS
I thank Mark Rakobowchuk for correcting and improving the English and for his insightful and helpful comments.