Contribution of oxygen extraction fraction to maximal oxygen uptake in healthy young men

Abstract We analysed the importance of systemic and peripheral arteriovenous O2 difference ( a-v¯O2 difference and a‐vfO2 difference, respectively) and O2 extraction fraction for maximal oxygen uptake ( V˙O2max). Fick law of diffusion and the Piiper and Scheid model were applied to investigate whether diffusion versus perfusion limitations vary with V˙O2max. Articles (n = 17) publishing individual data (n = 154) on V˙O2max, maximal cardiac output ( Q˙max; indicator‐dilution or the Fick method), a-v¯O2 difference (catheters or the Fick equation) and systemic O2 extraction fraction were identified. For the peripheral responses, group‐mean data (articles: n = 27; subjects: n = 234) on leg blood flow (LBF; thermodilution), a‐vfO2 difference and O2 extraction fraction (arterial and femoral venous catheters) were obtained. Q˙max and two‐LBF increased linearly by 4.9‐6.0 L · min–1 per 1 L · min–1 increase in V˙O2max (R 2 = .73 and R 2 = .67, respectively; both P < .001). The a-v¯O2 difference increased from 118‐168 mL · L–1 from a V˙O2max of 2‐4.5 L · min–1 followed by a reduction (second‐order polynomial: R 2 = .27). After accounting for a hypoxemia‐induced decrease in arterial O2 content with increasing V˙O2max (R 2 = .17; P < .001), systemic O2 extraction fraction increased up to ~90% ( V˙O2max: 4.5 L · min–1) with no further change (exponential decay model: R 2 = .42). Likewise, leg O2 extraction fraction increased with V˙O2max to approach a maximal value of ~90‐95% (R 2 = .83). Muscle O2 diffusing capacity and the equilibration index Y increased linearly with V˙O2max (R 2 = .77 and R 2 = .31, respectively; both P < .01), reflecting decreasing O2 diffusional limitations and accentuating O2 delivery limitations. In conclusion, although O2 delivery is the main limiting factor to V˙O2max, enhanced O2 extraction fraction (≥90%) contributes to the remarkably high V˙O2max in endurance‐trained individuals.


| INTRODUCTION
Under resting conditions in humans, the O 2 uptake (VO 2 ) is 3-5 mL · kg -1 · min -1 , and only a small fraction is consumed within the skeletal muscles. 1 However, during incremental exercise, the pulmonary V O 2 increases gradually and can reach a maximum (VO 2max ) of ~90 mL · kg -1 · min -1 depending on gender, age, body weight, genetics, training status and health. [1][2][3] According to the Fick equation, V O 2max is determined by the product of the maximal cardiac output (Q max ) and the arterial to mixed venous O 2 difference (a-vO 2 difference). Q max multiplied by the arterial O 2 content (CaO 2 ) sets the upper limit of systemic O 2 delivery, which is the principal limitation to V O 2max during exercise recruiting a large muscle mass, at sea level. [4][5][6] Despite extensive research since the 1950s on the factors limiting V O 2max , it is still debated whether peripheral O 2 extraction capacity contributes to limiting V O 2max . 7,8 Several original studies 4,5,[9][10][11][12] and review articles 6,[13][14][15] have addressed this topic in recent decades, yet no study has aimed to statistically analyse all the existing data on the association between V O 2max and its limiting factors. This kind of analysis is warranted, as the original studies often used homogenous groups with a small number of subjects (<10) since they applied costly and invasive techniques involving catheterizations to determine Q max (indicator-dilution techniques or the direct Fick method), regional blood flows (thermodilution or indicator-dilution techniques) and O 2 extraction fraction (calculated by the Fick equation or directly measured through arterial and venous catheters). Consequently, the statistical power is often too low to detect small but meaningful differences between subjects, groups with different training status and before and after training, thus precluding a definite conclusion.
It is documented that the a-vO 2 difference at V O 2max is only slightly different between untrained and endurance-trained individuals, 16,17 suggesting that peripheral adaptations to endurance training have only a minor impact on V O 2max . However, the CvO 2 difference is determined not only by the peripheries' ability to extract O 2 , reflected in the mixed venous O 2 content (a-vO 2 ) but also by the CaO 2 , which sets the upper limit for the a-vO 2 difference during maximal exercise. The CaO 2 is set by the haemoglobin concentration ([Hb]) and the O 2 saturation of Hb (SO 2 ), which may change with training and is acutely modified during exercise. For instance, endurance training causes plasma volume expansion 18 that can lead to haemodilution and a lower O 2 carrying capacity of the blood. 16 A high Q max shortens the time for alveolar/capillary gas equilibration at the lung causing exercise-induced arterial hypoxemia that further reduces the CaO 2 . 19,20 Therefore, it may be that the a-vO 2 difference does not increase substantially after endurance training because of a concurrent training-induced lowering of CaO 2 , whereas the systemic O 2 extraction fraction (O 2 extraction: a-vO 2 difference/CaO 2 ) may improve.
Another aspect of this discussion is whether the measurement techniques are sensitive enough to detect meaningful changes in the a-vO 2 difference. Most studies have not measured a-vO 2 difference directly but calculated it using the Fick equation (VO 2max /Q max ). 16,17,[21][22][23][24] The reason why so few studies have measured a-vO 2 difference directly during maximal exercise is because of the need for right heart catheterization. Therefore, studies measuring the arterial to femoral venous O 2 difference (a-v f O 2 difference) and leg O 2 extraction fraction directly using peripheral catheters may be more sensitive in evaluating whether the O 2 extraction capacity changes with endurance training.
It is important to note that the factors limiting V O 2max may change over the course of training. For instance, the maximal mitochondrial respiratory capacity (OXPHOS) measured in permeabilized muscle fibres ex vivo and V O 2max is associated in untrained, but not in trained individuals. 25 These and other data 26 suggest that peripheral factors contribute to limit V O 2max in the untrained state, but their influence may diminish with increased V O 2max and training status.
In the present study, we critically reviewed and statistically analysed the previously published data on the association between V O 2max and O 2 extraction fraction, in men, by focusing on catheterization studies. Two approaches were used: Part 1) articles containing individual data on pulmonary V O 2max , Q max (indicator-dilution techniques or the Fick method), a-vO 2 difference (mostly calculated) and O 2 extraction fraction measured during whole-body maximal exercise (running, cycling) were included; Part 2) to investigate the relationship between limb V O 2 and peripheral O 2 extraction fraction, mean data from studies reporting leg blood flow (LBF), a-v f O 2 difference and leg O 2 extraction fraction (catheters) measured during whole-body maximal exercise (running, cycling, cross-country skiing) were included. To investigate whether the limiting factors vary with V O 2max , we employed the Fick law of diffusion to calculate the muscle O 2 diffusing capacity (D M O 2 ) and subsequently used the Piiper and Scheid model to calculate the relative roles of perfusion versus diffusion limitations to V O 2max . 27 Finally, we discuss the potential mechanisms behind the elevated O 2 extraction fraction observed after endurance training.

| ANALYSIS OF EXISTING DATA
The strategy to use individual and mean data to investigate the systemic and peripheral responses, respectively, was chosen since a large amount of individual data has been published on systemic responses, whereas we were unable to identify other than mean values in studies investigating peripheral haemodynamics and O 2 extraction fraction. The data were identified through searches conducted in the PubMed database using several combinations of the following search terms: circulation, circulatory, hemodynamic(s), cardiac output, leg blood flow, arteriovenous oxygen difference, oxygen extraction and exercise. Cross-reference checks were also conducted, in addition to separate searches on authors with articles already included in the database. Only exercise modes engaging a large muscle mass that could elicit V O 2max were included (cycling, running and cross-country skiing using the diagonal technique). Data from cross-sectional studies or before and after training interventions that were collected in normoxia on young (<40 years old) and healthy individuals were included. Data collected in hypoxia, after acclimatization to altitude, in altitude natives, in hyperthermia, with atrial pacing, after bed rest and after blood volume manipulations were excluded. The control condition was used when the above forms of manipulations of the cardiovascular system were conducted. Only catheterization studies that used invasive methods to measure Q max (indicator-dilution techniques or the Fick method) and LBF (bolus or continuous infusion thermodilution and indicator-dilution techniques) were included. Only individual data from men are used (Part 1). In Part 2, studies that had a sample with a majority of men were used (≥50%). When several papers reported data from the same data collection, only one of the articles was included. If an article used some of the same subjects as previously reported, but with supplementation with new subjects, the data were included. The included articles are presented in Tables 1 and  2 for Parts 1 and 2, respectively.

| Calculations
When the data were published in graphs and not in tables or text, ImageJ (v1.50b; National Institutes of Health, USA) was used for data extraction. If not all variables were reported in the articles, the reported data were used to derive the missing values via the following formulas or combination of formulas if possible: If no arterial partial pressure of O 2 (PO 2 ) was reported, 100 mmHg was assumed for the calculation of CaO 2 (ie, 3 mL O 2 freely dissolved in blood plasma per 1 L of blood). Central venous pressure (CVP) at V O 2max was taken as 5 mmHg 4 when calculating systemic vascular conductance.   The calculated a-vO 2 difference (VO 2max /Q max ) showed an inverse J-shaped curve, reaching the highest level between 4.5-5.0 L·min −1 before declining at higher V O 2max ( Figure 1C). After accounting for the decrease in CaO 2 with increasing V O 2max ( Figure 1E; P < .001), the calculated O 2 extraction fraction increased up to a V O 2max of ~4.5-5.0 L·min −1 and then approached a maximal value at ~90% ( Figure 1D) when restricting the exponential decay model to plausible physiological limits (VO 2max : 6-7 L·min −1 ). The linear decrease in CaO 2 was explained by arterial hypoxemia (decreased arterial SO 2 ; Figure 1F; P < .001) and a non-significant negative relationship between [Hb] and V O 2max ( Figure 1G; P = .232). The calculated CvO 2 gradually decayed and approached a minimum at ~10-15 mL · L -1 in the subjects with the highest V O 2max ( Figure 1H).
When controlling the regression between the individual data of V O 2max and the calculated O 2 extraction fraction with mean values from studies measuring O 2 extraction fraction directly using the Fick method (right heart catheterization), or indirectly using the Fick equation (Q max : indicator-dilution or transpulmonary thermodilution), most values fell close to the regression curve ( Figure 3).

| Part 2: Peripheral responses during maximal exercise (mean data)
LBF and two-LBF rose by 4.6 and 5.7 L·min −1 for each L·min −1 increase in leg and pulmonary V O 2max respectively (Figure 4A,D; both P < .001). Leg and pulmonary V O 2max displayed a linear relationship (y = 1.27x -2.01; R 2 = .85; n = 28; P < .001). The directly measured leg a-v f O 2 difference and leg O 2 extraction fraction were best explained by exponential decay models and increased gradually with the increase in leg and pulmonary V O 2max to approach a maximum at ~180-190 mL · L -1 and ~90-95% respectively ( Figure 4B,C,E,F). These relationships were equally strong when V O 2max was standardized to body weight (Supporting material Figure 1). Note that leg a-v f O 2 difference was not lower for the subjects with the highest V O 2max , as observed for the systemic a-vO 2 difference ( Figure 1C), possibly since only one subject group exceeded a V O 2max of 4.7 L·min −1 , where this occurred for the systemic responses (see Figure 1C). In connection, no association was evident between pulmonary V O 2max and CaO 2 for these data (y = 1.07 + 195; R 2 < .01; n = 30; P = .701).

| Oxygen delivery
To match O 2 delivery to O 2 consumption, Q max and two-LBF increased by ~5-6 L·min −1 per 1 L·min −1 increase in pulmonary V O 2max . These relationships were strong and complied F I G U R E 1 The relationship between individual values (from studies reported in Table 1) of pulmonary maximal oxygen uptake and cardiac output (A), stroke volume (B), arterial to mixed venous oxygen difference (a-vO 2 difference; C), systemic oxygen extraction fraction (D), arterial oxygen content (E), arterial oxygen saturation (F), haemoglobin concentration ([Hb]; G) and the calculated mixed venous oxygen content (H). All data were obtained during maximal exercise. Inserted in each graph are the formulas for the regression equations along with the goodness of fit (R 2 ) and the number of data pairs (n)

F I G U R E 2
The relationship between individual values (from studies reported in Table 1)  with previous research and the "classic" view that O 2 delivery is the primary determinant of whole-body V O 2max . 4,7,11 As maximal heart rate showed no apparent relationship with V O 2max , the high stroke volumes (>180 mL · beat -1 ) explained the large Q max in the athletes included in the present analysis (>35 L·min −1 ), in agreement with previous knowledge. 13,16,30 Despite increased Q max , MAP was unchanged with increasing V O 2max as a result of increased vascular conductance. Although untrained individuals typically display a rise in MAP from rest to maximal exercise, 31 well-trained athletes can display an unchanged MAP or even a small reduction owing to profound peripheral vasodilation. 32 Consequently, vasodilation of a well-developed peripheral vascular network likely contributed to the extremely high stroke volumes by minimizing afterload in the subjects with the highest V O 2max . To substantiate, endurance training of each leg separately, to evoke extensive peripheral adaptations without stimulating the central circulation substantially, has been shown to decrease MAP and the total peripheral resistance during two-legged maximal exercise that likely contributed to the elevated stroke volume and Q max after training. 30 The high stroke volumes are probably achieved through the combined effect of a large left ventricular mass, 33,34 compliant cardiac chambers 35,36 and an expanded blood volume 37,38 that facilitates a high end-diastolic volume and preload combined with the relatively low afterload.

| Oxygen extraction
The calculated systemic a-vO 2 difference showed a large variability for a given V O 2max and was, if anything, lower in those subjects displaying the highest V O 2max (>5 L·min −1 ) compared to those being moderately to well trained (VO 2max : 4-5 L·min −1 ). This agrees with previous studies showing only a small difference between non-endurance-trained and active individuals 16,17 and no apparent difference between welltrained individuals and elite athletes. 16 This has led previous investigators to argue that improved O 2 extraction does not contribute or only minimally contributes to the remarkably high V O 2max observed in elite athletes. 14,39 However, these papers may not have considered that endurance training causes plasma volume expansion, 18 which often leads to haemodilution and a lower O 2 carrying capacity of the arterial blood. 16 Combined with the below-average haemoconcentration from rest to maximal exercise that occurs in well-trained individuals 16 and the exercise-induced arterial hypoxemia that often accompanies a high Q max , 19,20 individuals with the highest V O 2max displayed a substantially lower CaO 2 (~10%) than those with a low V O 2max (<180 mL · L -1 vs >200 mL · L -1 ; Figure 1E). Therefore, the lower CaO 2 may explain why moderately and well-trained individuals can have a similar a-vO 2 difference, despite differing markedly in D M O 2 , mitochondrial mass and capillary density. 40,41 Actually, parts of this mechanism are demonstrated experimentally since acute plasma volume expansion increases Q max but lowers the CaO 2 and, hence, reduces the a-vO 2 difference during maximal exercise. 42,43 Opposite to the a-vO 2 difference, the systemic O 2 extraction fraction-ie, the fraction of O 2 that is taken up with respect to the amount available for utilization (a-vO 2 difference/CaO 2 )-increased with V O 2max until reaching ~90%. This pattern was confirmed in the leg when measured using catheters, with the O 2 extraction fraction increasing progressively with leg and pulmonary V O 2max until reaching ~90 to 95%. Therefore, the calculated systemic O 2 extraction fraction (Fick equation) is supported by direct measurements via arterial and femoral venous blood sampling and strongly F I G U R E 3 Mean values (±95% confidence limits, where available) of systemic oxygen extraction fraction versus maximal oxygen uptake from studies using the direct (pulmonary artery catheter) or the modified (right atrium catheter) Fick method, 4,9,28,32,51,55,61,[121][122][123][124][125][126][127] the indicator dilution method 5,11,16,17,21,22,24,31,42,[52][53][54]57,85,112,114,[128][129][130] and the transpulmonary thermodilution method. 56 Broken line is the regression equation obtained from Figure 1D  In most endurance training studies investigating the interplay between central and peripheral adaptations in improving V O 2max , Q max was measured by non-invasive methods (such as inert-gas rebreathing techniques, impedance cardiography and bioreactance) and the Fick equation was used to derive the a-vO 2 difference (for references, see the meta-analysis by Montero et al 44 ). The majority of these studies failed to detect a statistically significant change in the a-vO 2 difference. However, this finding does not necessarily mean that V O 2max was exclusively increased by elevated Q max for three F I G U R E 4 The relationship between one-leg or pulmonary maximal oxygen uptake and leg blood flow ( Figure 4A,D, respectively), arterial to femoral venous oxygen difference (a-v f O 2 difference; Figure 4B,E, respectively) and leg oxygen extraction fraction ( Figure 4C,F, respectively). Black circles and white squares denote cycling and diagonal cross-country skiing respectively. The skiers are excluded from the regression in Figure 4D owing to the combined leg and arm use for locomotion that distributed 6.6 L·min −1 blood flow to the exercising arms (see the discussion). Data are mean values (±95% confidence limits, where available) from studies reported in Table 2 0.6 0.9 1. SKATTEBO ET Al was measured directly during maximal exercise (arterial and venous catheters), the vast majority found an increased O 2 extraction fraction after training. 12,30,[45][46][47] A particular case, concerning the relationship between one-leg V O 2max and O 2 extraction fraction ( Figure 4C) and between pulmonary V O 2max and two-LBF ( Figure 4D) deserves some attention (the white squares). These data were collected during combined upper-and lower-body exercise (cross-country skiing using the diagonal technique) and 6.6 L·min −1 of Q max was distributed to the two arms. 32 Hence, when combining the locomotor blood flow (arms+legs), the data fall perfectly on the regression line between blood flow and pulmonary V O 2max . When redistributing LBF towards other exercising musculature, the erythrocyte capillary mean transit time (MTT) is increased. Therefore, the conditions for Hb-O 2 off-loading are improved, resulting in a slightly higher O 2 extraction fraction for a given leg V O 2 . The same phenomenon can be seen when adding arm cycling to ongoing leg cycling 48 or vice versa, 49 which increases the O 2 extraction fraction that compensates for some of the reduction in blood flow.

| Limitations to V O 2max by O 2 delivery and O 2 extraction varies with training status
The equilibration index Y was positively correlated with V O 2max . Therefore, endurance training leads to a situation where the muscles become gradually more O 2 -delivery limited. Thus, individuals with the highest V O 2max can only achieve a further substantial improvement in V O 2max by increasing O 2 delivery, a conclusion supported by the extremely low levels of Cv f O 2 and CvO 2 in these subjects. Therefore, the limiting factors to V O 2max change with training status and V O 2max : (a) untrained, but healthy individuals display mixed perfusion-diffusion limitations; and (b) this diffusional limitation reduces as V O 2max is increased. 26 These conclusions are similar to those of Gifford et al, 25 who found a clear relationship between OXPHOS measured in permeabilized muscle fibres ex vivo and V O 2max in untrained but not in trained individuals.

| Why is not all the O 2 extracted from the blood?
The entire Q max cannot be directed to the skeletal muscles during exercise. Other organs like the brain, heart, splanchnic organs and skin need perfusion and O 2 delivery to maintain homeostasis. Q max must also serve the O 2 demand of the respiratory muscles and the muscles in the trunk and the arms that stabilize the subject's position on the cycle ergometer, and these tissues are characterized by a substantially lower O 2 extraction than the legs during maximal exercise. 5,50 As a mean of those investigations measuring Q max and LBF simultaneously (Table 3), the non-leg blood flow was 6.4 L·min −1 and was unaffected by the level of Q max (y = 0.002x + 6.4; R 2 < .001; n = 12; P > .999). 4,5,9,11,31,[51][52][53][54][55][56][57] The O 2 extraction was calculated to be 68% on average for all non-leg tissues (head, trunk and arms), explaining why the O 2 extraction fraction of the central circulation was slightly lower than in the legs (79% vs 84%, respectively; Table 3). A mean difference of 5 percentage points might be a small underestimation since the studies using right heart catheterization 4,9,28,29,51,55 combined with arterial and femoral venous catheters indicated a mean difference of 8 percentage points. A difference of 5%-8% points fits well, since the O 2 extraction fraction of the arms, myocardium, brain and trunk range from 40% to 80% during exercise. 5,50,58-60 Therefore, the CvO 2 can never reach the same level as the Cv f O 2 during exercise involving T A B L E 3 Data from studies measuring pulmonary O 2 uptake, cardiac output (indicator-dilution, Fick method or transpulmonary thermodilution), leg blood flow (thermodilution) and leg arteriovenous O 2 difference (a-vO 2 difference; catheters) simultaneously during maximal exercise. From these measurements, O 2 extraction fraction was calculated for the central circulation and the non-leg tissue (combined trunk, arms and head) the legs and was calculated to reach a minimum of ~15 mL · L -1 in subjects having a V O 2max of 6 L·min −1 ( Figure 1H). To our knowledge, the lowest CvO 2 measured at sea level using right heart (atrium) catheterization is 20.1 mL · L -1 (group mean) in athletes with a V O 2max of 5.1 L·min −1 . 29 A slightly lower value was measured in one of these cross-country skiers (15.5 mL · L -1 ), and a mean value of 18.6 mL · L -1 has been measured in moderately trained individuals after acclimatizing to 6500 metres above sea level 61 ; indicating that 15 mL · L -1 or lower is approachable. The highest recorded leg O 2 extraction fraction was 93% (group mean) 29 and the regression models indicated a plateau at ~95% within physiological limits for pulmonary V O 2max . Hence, a minimum of ~10 mL O 2 remains in each litre of femoral venous blood associated with a PO 2 of ~10 mmHg, even for the best trained individuals. In this situation, a PO 2 gradient persists between the blood and myoglobin (myoglobin/intracellular PO 2 : ~1-2 mmHg), 62 where myoglobin-facilitated diffusion should proceed given the high myoglobin O 2 affinity (myoglobin P 50 O 2 : ~5 mmHg) and the low myoglobin SO 2 at maximal exercise. 62 However, according to the Fick law of diffusion, the diffusive flux is directly proportional to the PO 2 gradient and will, thus, gradually decrease along the capillary and be very small when approaching low capillary PO 2 values such as 10 mmHg. It has also been shown that the primary site of resistance to O 2 diffusion is between the capillaries and the sarcoplasm and it has been estimated that the "critical capillary PO 2 " needed to overcome this resistance may be as high as 10-20 mmHg. [62][63][64][65] The remaining O 2 may, therefore, represent diffusional limitations across the combined capillary wall, interstitium and sarcolemma barriers together with a MTT that is too short for complete Hb-O 2 off-loading. This is supported by the need for an infinitesimal PO 2 gradient for O 2 to diffuse from the sarcoplasm to cytochrome c oxidase 66 and the estimate that a mitochondrial PO 2 of ~1 mmHg may be sufficient to support maximal mitochondrial respiration. 67,68 The remaining O 2 may also represent muscle metabolism-perfusion mismatch 69,70 and an inevitable lower O 2 extraction from the blood perfusing the skin, connective tissue, fat and bone marrow of the leg causing venous admixture. In this context, the end-capillary PO 2 , assessed using video microscopy, was found to be lower than the PO 2 both in the venule (O 2 microelectrode) and vein (blood gas) draining the muscle region of interest. 71 Hence, the lowest femoral venous PO 2 values of ~10 mmHg indicates an even lower end-capillary PO 2 in the capillaries adjacent to the most metabolically active muscle regions during maximal exercise, possibly approaching ~5 mmHg. Therefore, no matter which kind of limitation prevails, it is highly unlikely that leg O 2 extraction fraction can improve much further, and that a theoretical threshold of ~95% exists because of the above diffusional and distributional limitations and barriers.

| THE MECHANISMS EXPLAINING THE IMPROVEMENTS OF O 2 EXTRACTION WITH TRAINING
The systemic O 2 extraction fraction may increase through two main mechanisms with training: (a) by directing a higher fraction of Q max to the exercising muscles and (b) by increasing the peripheral O 2 extraction fraction.
Both in trained and untrained subjects, during exercise with a large muscle mass (such as running and cycling), the muscle-specific blood flow (per unit of mass) is restrained as a result of sympathetically mediated vasoconstriction of peripheral vascular beds, caused by a limited Q max . 9,32,48 Even in "untrained" leg skeletal muscle, the reserve in vasodilatory capacity is very high and supports 2-3 times larger blood flow per unit of mass, as observed during dynamic one-legged knee extension. 72 Simply increasing Q max (for instance, by training), without any peripheral adaptations, may increase the systemic O 2 extraction fraction by two mechanisms. First, the recruitment of a larger portion of the already existing capillary network may reduce diffusion distances and thereby increase the O 2 extraction. This additional recruitment may also serve to maintain MTT despite increased LBF. Second, a larger fraction of Q max will flow through the exercising muscles (Figure 7) because the non-exercising tissue blood flow is independent of Q max in healthy young subjects (at ~6.4 L·min −1 , see section 3.4). 4,5,9,11,31,[51][52][53][54][55][56][57] Consequently, even without any peripheral adaptations, the systemic O 2 extraction fraction may increase when Q max and LBF are elevated with training.

F I G U R E 7
The fraction of maximal cardiac output (Q max ) that is directed to the legs during maximal exercise (cycling) as a function of V O 2max . The included studies measured Q max by using the indicatordilution method, Fick method or transpulmonary thermodilution, and leg blood flow was measured by thermodilution. 4,5,9,11,31,[51][52][53][54][55][56][57] Note that the uppermost data point (0.915; ie, only 2.2 L·min −1 in calculated non-leg blood flow) is supra-physiological, but the correlation was similar after its exclusion (R 2 = .42) P | 13 of 19

SKATTEBO ET Al
The peripheral O 2 extraction depends on the interplay between several factors: (a) the kinetics of O 2 off-loading from Hb; (b) the erythrocyte MTT, which is determined by the blood flow, the capillary density, the capillary recruitment and the degree of matching of blood flow distribution to the metabolic demand; (c) the diffusional O 2 conductance over the combined capillary wall, interstitium and sarcolemma barriers; and (d) the muscle oxidative capacity, the mitochondrial p50 and the mitochondrial activation. 10,29,73 A right-shifted O 2 -Hb dissociation curve (elevated P 50 O 2 ) increases the O 2 extraction fraction in pump-perfused dog muscle. 74 A close relationship has also been demonstrated between O 2 extraction fraction and in vivo P 50 O 2 in humans during exercise. 29 Very few of the studies included in the present analysis reported the in vivo P 50 O 2 , but it was possible to calculate it from the other blood gas parameters using Kelman's Equation 75 after assuming a femoral venous blood temperature of 39.0°C at maximal exercise. 9,55,76 Based on 15 of the studies presented in Table 2, the P 50 O 2 was linearly associated with leg O 2 extraction fraction (R 2 = .27; n = 15; P = .048). Despite this relationship, a high P 50 O 2 does not seem to be compulsory to achieve high O 2 extraction during whole-body maximal exercise, as demonstrated in experiments using a small dose of carbon monoxide (carboxyhaemoglobin at 6%-7%), which left-shifts the ODC without a negative impact on O 2 extraction fraction. 56 Increased MTT has the potential to increase O 2 extraction, but whether this occurs after endurance training is determined by the balance between the changes in blood flow and the capillary blood volume. Capillary density typically improves by 10-30% after 4-24 weeks of endurance training, [77][78][79] which is similar to the changes in V O 2max for this training duration. [78][79][80] Moreover, cross-sectional data indicate a similar difference in capillary density to that of V O 2max between untrained and endurance trained men. 41 Therefore, the capillary growth probably maintains the MTT despite elevated Q max and peripheral blood flow after training. In support, similar improvements in arm blood flow and capillary density have been observed after a period of arm training, causing no change in the calculated MTT. 47 The arm O 2 extraction fraction was increased in the same study, suggesting that elevated MTT is not the primary mechanism by which O 2 extraction is improved after training. However, this may differ between arms and legs (ie, small vs large muscle mass exercise). Moreover, in the calculation of MTT in the study mentioned above, full capillary recruitment was assumed. Therefore, even though the changes in capillary density and muscle blood flow share magnitudes after endurance training, the MTT may still be increased if the capillary recruitment is altered.
An increased capillary-to-fibre ratio after endurance training increases the number of contact points between the capillary and the muscle fibre. This increases the diffusional surface area that, according to the Fick law of diffusion, increases the diffusive flux in a directly proportional manner. Therefore, the capillary-to-fibre ratio is regarded as a critical determinant of O 2 diffusion from the erythrocytes to the cytoplasm. 81,82 As an example, a larger diffusional area and shorter diffusional distance are proposed to contribute to the higher O 2 extraction fraction in the legs than in the arms during exercise. 29 Moreover, if the capillary recruitment is changed with training, this may also affect the effective diffusional surface area similarly to de novo capillarization.
During whole-body maximal exercise, the oxidative capacity of skeletal muscle exceeds the O 2 delivery, as illustrated by the twofold higher V O 2 per unit of muscle mass during dynamic one-legged knee extension compared to cycling exercise (approximately 2.5 vs 20 kg active muscle mass, respectively). 10,72 Therefore, the leg muscles possess an oxidative reserve capacity at V O 2max during whole-body exercise, which has frequently been used as an argument to indicate that the large improvements in mitochondrial and capillary networks after endurance training are likely only crucial for improvements in endurance performance and do not affect the limiting factors to V O 2max . 83 In support of this view, the calculated O 2 extraction fraction is maintained or increases after prolonged bed rest (3-6 weeks), although a substantial reduction in mitochondrial volume density occurs. 84,85 However, the O 2 extraction fraction depends on the interactions between several factors. For instance, by acutely decreasing Q max and LBF using β-adrenergic blockade, a-vO 2 difference and a-v f O 2 difference increase during submaximal and maximal exercise, facilitated by increased erythrocyte MTT. 86,87 This is substantiated by the positive relationship between the ratio of OXPHOS/O 2 delivery and the leg O 2 extraction fraction, 10 meaning that the balance between muscle oxidative capacity and blood flow (ie, oxidative capacity and MTT) is more critical for O 2 extraction than any of these factors alone. Therefore, as bed rest reduces Q max dramatically but causes only a minor change in capillary density, 84,85 the MTT is elevated, and the ratio of OXPHOS/ O 2 delivery is probably the same, in favour of increased or maintained O 2 extraction fraction. In contrast, by changing the exercise mode from upright to supine cycling after bed rest, which preserves Q max at the pre-bed rest level, the calculated a-vO 2 difference is decreased (154 to 120 mL · L -1 ). 88 Similarly, after a dog gastrocnemius muscle was immobilized for 3 weeks, followed by electrical stimulation to V O 2max while being pump perfused to receive a similar O 2 delivery as a control muscle, the O 2 extraction fraction was dramatically reduced. 82 Therefore, muscle oxidative capacity seems to play a role in determining O 2 extraction, and the bed rest studies need to be evaluated carefully because of the consequences for peripheral MTT.
If O 2 extraction fraction improves after endurance training, is probably affected by the balance between central and peripheral adaptations. For instance, after 2 weeks of high-intensity interval training that elevated the cytochrome c oxidase activity by 20% but caused no change in Q max , V O 2max was increased by 8% and was entirely attributed to the improved systemic (calculated a-vO 2 difference) and leg (increased deoxyhaemoglobin and decreased tissue oxygenation index in Vastus Lateralis, assessed using NIRS) O 2 extraction. 89 However, after 3-8 weeks of endurance training, improvements in Q max explain almost the entire increase in V O 2max , as indicated by meta-regression. 44 If the training lasts longer (>8 weeks), enhancements of Q max decelerate and improvements in a-vO 2 difference are again evident. 44,90 Therefore, the peripheral adaptations are probably just sufficient to counteract the "negative influence" of elevated Q max and LBF on MTT in periods with large central adaptations, and improvements in O 2 extraction fraction is likely only evident when the peripheral adaptations largely surpass those of the central circulation. This can be substantiated by findings from one-legged endurance training that induces robust peripheral adaptations without stimulating the central circulation substantially and commonly improves leg a-v f O 2 difference by 5-10 mL · L -1 . 30,45 The mitochondrial volume density can differ by as much as 150% between untrained and well-trained individuals in extreme cases (eg, ~4 vs ~10 vol. %) 91,92 and can improve by as much as ~40%-55% after 6 weeks of endurance training in previously sedentary individuals. 38,93,94 Why does this disproportionate adaptation occur when the muscle already possesses an oxidative reserve capacity? Does it have any physiological meaning for V O 2max or is it only important for improvements in, for example, fat oxidation 95 and the lactate threshold, 96 thus improving endurance?
Although an impressive increase in leg O 2 extraction fraction from 72% to 82% has been reported after only 9 weeks of intense endurance training in previously sedentary subjects, 12 we propose that remarkable increases in muscle oxidative capacity are needed to achieve the outstanding leg O 2 extraction fraction observed in elite athletes (close to 95%). 29,97 By analogy, the oxidative reserve capacity may act as a "bottomless pit", keeping the myoglobin SO 2 and intracellular PO 2 low. This, in turn, maintains the PO 2 gradient between the capillary and the muscle cell, promoting O 2 diffusion and O 2 extraction even at a very low capillary PO 2 .
Emerging evidence suggests that the mitochondrial volume density is increased while their intrinsic OXPHOS (OXPHOS divided by mitochondrial volume density or citrate synthase activity) is unchanged 89,98,99 and sometimes even reduced 94,100 after training. Since the mitochondrial respiratory rate and the ex vivo mitochondrial p50 increase in parallel, 10,73 the unchanged or reduced intrinsic OXPHOS after training may permit an increased OXPHOS per unit of muscle mass while preserving (or increasing) the mitochondrial O 2 affinity (ie, by keeping the mitochondrial p50 low). 73 Thus, a large pool of mitochondria with high O 2 affinity may preserve mitochondrial activation at low O 2 availability (low capillary PO 2 ) and promote peripheral O 2 extraction, but is yet to be experimentally tested. Moreover, the subsarcolemmal mitochondrial population increases relatively more than the intermyofibrillar population after endurance training. 93,94 These mitochondrial clusters in close proximity to the capillaries may, speculatively, amplify the O 2 concentration gradient, shorten the diffusional distance and, thus, promote O 2 diffusion across the sarcolemma 101 and enable further O 2 extraction at the end of the capillaries.
As shown in Figure 6C, a subject's V O 2max becomes gradually less sensitive to adaptations improving diffusion when V O 2max is already high. Therefore, to raise the O 2 extraction fraction even slightly (eg, 2%), it is likely that more substantial improvement in peripheral adaptations is needed. However, a change in leg O 2 extraction fraction from, for example, 93% to 95% would only have a small impact on whole-body V O 2max : for an athlete with a V O 2max of 5 L·min −1 , a two-LBF of 24 L·min −1 (Q max : ~31 L·min −1 ) and an CaO 2 of 190 mL · L -1 , the V O 2max would only increase by ~90 mL · min -1 (1.8%). In comparison, an increase of 1 L·min −1 in two-LBF would increase V O 2max by ~170 mL · min -1 (3.4%) if all other factors remained the same.

| STUDY CONSIDERATIONS
The data were collected from several research groups and published over six decades (1958-2017) using a variety of gas analysers, flow sensors, methods to determine blood O 2 content and PO 2 , and several procedures to analyse the indicator-dilution and blood temperature curves for Q max and LBF measurements respectively. Therefore, for a given V O 2max , the between-subject variability presented here may be overestimated. Moreover, several different averaging strategies for V O 2 and the associated variables have likely been applied (rarely stated in the manuscripts). Despite these potential sources of noise, in general, the studies' mean values converged to similar values. The fact that, despite the combination of several measurements with distinct methods (such as pulmonary gas exchange, thermodilution and blood gas analyses), the integrations of the obtained values fitted into the physiological range and agreed between studies, demonstrates the quality of these studies and the robustness of the analysis presented here.

| CONCLUSION AND PERSPECTIVE
In conclusion, measurements of Q max and LBF show that O 2 delivery is the primary determinant of whole-body and limb V O 2max . However, we also show that a very high O 2 extraction fraction contributes to the remarkably high V O 2max in welltrained individuals and elite endurance athletes. To reinforce this conclusion we can, using the regression lines established in the present investigation, compare a typically sedentary subject and an elite endurance athlete with a large difference in V O 2max (3.0 vs 5.5 L·min −1 ): the elite athlete has a 1.83fold higher V O 2max , a 1.60-fold higher Q max and a 1.26-fold higher O 2 extraction fraction ( Figure 8). However, because of the lower CaO 2 , the a-vO 2 difference is only 1.13-fold higher in the elite athlete. This also stresses that a-vO 2 difference and O 2 extraction fraction cannot be used interchangeably when evaluating central versus peripheral limitations to V O 2max . Finally, the limitations for whole-body V O 2max change with training status, with an accentuated O 2 delivery limitation and conversely a decreasing O 2 diffusional limitation with increasing V O 2max .