Peak muscle perfusion and oxygen uptake in humans: importance of precise estimates of muscle mass

G. Rådegran, E. Blomstrand, B. Saltin


The knee extensor exercise model was specifically developed to enable in vivo estimates of peak muscle blood flow and O2 uptake in humans. The original finding, using thermodilution measurements to measure blood flow in relation to muscle mass [P. Andersen and B. Saltin.J. Physiol. (Lond.) 366: 233–249, 1985], was questioned, however, as the measurements were two- to threefold higher than those previously obtained with the133Xe clearance and the plethysmography technique. As thermodilution measurements have now been confirmed by other methods and independent research groups, we aimed to address the impact of muscle mass estimates on the peak values of muscle perfusion and O2 uptake. In the present study, knee extensor volume was determined from multiple measurements with computer tomography along the full length of the muscle. In nine healthy humans, quadriceps muscle volume was 2.36 ± 0.17 (range 1.31–3.27) liters, corresponding to 2.48 ± 0.18 (range 1.37–3.43) kg. Anthropometry overestimated the muscle volume by ∼21–46%, depending on whether quadriceps muscle length was estimated from the patella to either the pubic bone, inguinal ligament, or spina iliaca anterior superior. One-legged, dynamic knee extensor exercise up to peak effort of 67 ± 7 (range 55–100) W rendered peak values for leg blood flow (thermodilution) of 5.99 ± 0.66 (range 4.15–9.52) l/min and leg O2 uptake of 856 ± 109 (range 590–1,521) ml/min. Muscle perfusion and O2 uptake reached peak values of 246 ± 24 (range 149–373) and 35.2 ± 3.7 (range 22.6–59.6) ml ⋅ min−1 ⋅ 100 g muscle−1, respectively. These peak values are ∼19–33% larger than those attained by applying anthropometric muscle mass estimates. In conclusion, the present findings emphasize that peak perfusion and O2 uptake in human skeletal muscle may be up to ∼30% higher than previous anthropometric-based estimates that use equivalent techniques for blood flow measurements.

  • blood flow
  • computer tomography
  • exercise
  • muscle volume
  • thermodilution

in vivo measurements of the arterial inflow to or venous outflow from a contracting muscle group can now be obtained quite precisely in humans (2, 16). Knee extensor exercise has commonly been used because the muscle activity has been verified to be confined to the quadriceps muscle, as found by applying T2-weighted magnetic resonance imaging, electromyography, and intramuscular temperature measurements, as well as by comparing voluntary with electrically induced knee extensor contractions (1, 11, 15, 18). Moreover, during intense dynamic contractions the various portions of the quadriceps muscle appear to be close to or fully engaged (15, 18).

As the first thermodilution blood flow measurements were published (2), they were questioned as being too high. Subsequent studies by others have however confirmed these values (Table1), and with even higher magnitudes during more extreme interventions (12, 13, 19-22, 24). The initial criticism was mainly based on traditional findings measuring blood flow with 133Xe clearance and plethysmography. 133Xe clearance has, however, been shown to be nonlinear and to underestimate blood flow (3, 5, 6, 23). In addition, plethysmography is subject to inherent limitations such as motion and occlusion artifacts confined to application at rest (4, 7, 28). Furthermore, the plethysmography measurements performed after or in between contractions only reflect an immediate postexercise value and not exercise per se. Moreover, as the magnitude of the thermodilution measurements has now been verified by other independent methods such as ultrasound Doppler and indocyanine green dye infusion (2, 10, 16, 26), and as these three techniques enable equivalent measurements during both the contraction and relaxation phases, they seem today to offer the best regional measurements of blood flow during exercise in humans (17).

View this table:
Table 1.

Average values of peak work rate, Q˙, andV˙o 2

In the previous estimates of muscle perfusion, the size of the knee extensor muscle engaged has often been determined from anthropometric estimates (2). The mathematical formulas (2, 8) applied for the anthropometric estimates are based on the original model derived by Jones and Pearson (9). The quadriceps muscle mass is determined from a size relationship between the quadriceps and hamstring muscle in relation to the total limb volume, on the basis of a small number of autopsy studies (2). There are comparisons between such anthropometric estimates and imaging techniques, but the number of image scans have generally been too few (8). Recently, Ray and Dudley (15) also used MRI to determine the muscle volume engaged in contractions, but instead of measuring blood flow, they used data in the literature to estimate muscle perfusion.

Thus, to obtain accurate values of muscle perfusion, not only blood flow but also the muscle volume engaged in the exercise needs to be accurately determined in the same subjects. As the thermodilution measurements today have been verified as accurate by other methods, we aimed to address the impact of the muscle mass estimates. In the present study, muscle blood flow was determined with thermodilution in the femoral vein. The size of the knee extensor muscle was determined by multiple computed tomography (CT) image cross sections along the full length of the muscle. Moreover, to evaluate their validity, three different estimates of the quadriceps muscle length were applied to anthropometrically estimated knee extensor muscle volume. The present study, therefore, provides a guideline for the accuracy of muscle mass estimates and an improved knowledge of peak perfusion and oxygen uptake in human skeletal muscle.



Nine healthy human subjects, eight men and one woman, with a mean ± SE age of 26 ± 1 (range 22–35) yr, height of 178 ± 3 (range 158–186) cm, and weight of 78 ± 4 (range 49–98) kg, volunteered to participate in this study. The subjects' engagement in exercise varied from daily activities to regular endurance training. Before the experiments, the subjects were informed about the experimental procedures and potential risks and discomfort and were advised that they could withdraw at any time without consequences. They participated after first giving signed informed consent. The experiments were carried out with the approval of the Ethical Committees of Copenhagen and Fredriksberg (KF-01–403/95).

Experimental procedures.

Before the experiments all subjects were familiarized with the one-legged, dynamic, knee extensor exercise model (1) by training at 60 rpm until they were comfortable and could fully relax the hamstring muscles, so that the work was done solely by the knee extensors. The isolation of the quadriceps muscle was supported by the kicking-force tracings collected during exercise. The peak power output that the subjects could sustain at 60 rpm was determined before the experiments by an incremental test starting at 10 W and increasing by 5–10 W every third minute until they could no longer sustain the rhythm.

The subjects reported to the laboratory on the morning of the experimental day. The femoral artery and vein of one leg were cannulated under local anesthesia (Lidokain, 20 mg/ml, Sygehus Apotekerne, Copenhagen, Denmark). The arterial (Ohmeda, Wiltshire, UK) and the venous catheter (Cook) were inserted proximally ∼2–5 cm below the inguinal ligament (IL). A thermistor (model 94–030–2.5F, TD probe, Edwards Edslab, Baxter, Irvine, CA) was inserted in the venous catheter and connected to a cardiac output computer (model 9520A, Harvard Apparatus, American Edwards Laboratories, Irvine, CA) for thermodilution blood flow measurements (2). A Harvard mechanical syringe pump (model 44, Harvard Apparatus) was used for constant infusion of cold saline (2). Arterial and venous blood samples were taken and analyzed for hemoglobin, oxygen saturation, and PO2 (AVL 912 CO-oxylite, AVL Medical Instruments, Schaffhausen, Switzerland), as well as hematocrit. Limb oxygen uptake was calculated by multiplying the blood flow measurements in the femoral vein (outflow) with the arterial and venous difference in oxygen content (Fick principle). Heart rate (electrocardiogram) and intra-arterial blood pressure were recorded via a Kone patient data monitor (model 565A, Medicoline, Valby, Denmark). The knee extensor force was measured with a strain gauge attached to the ergometer lever arm. All variables were recorded on a Gould recorder (model TA200, Gould) or in a personal computer (IBM compatible, Pentium based). The pulmonary oxygen uptake was determined on-line by a CPX Medical Graphics instrument (Spiropharma, Klampenborg, Denmark).

In the experiments, the subjects exercised for ∼10 min at an absolute workload of 30 W and then for 5–7 min at a relative workload corresponding to 75–80% of their individual peak power output. The relative workload was included as a comparison to the peak power output to ensure that all subjects reached their peak intensity. The subjects were then rested for ∼30–40 min before performing an experimental ramp protocol starting at 20–30 W for 3 min and increasing by 5 W every 30 s up to their peak power output. The peak power output was defined as the highest workload at which the subjects could maintain the pace for at least 1 min. Measurements of blood flow and pulmonary oxygen uptake as well as sampling of the blood were performed at rest and during steady-state exercise and during the last minute at peak intensity. To minimize blood flow contributions from the lower leg, a cuff was placed just below the knee of the working leg and temporarily inflated to a suprasystolic blood pressure (≥240 mmHg) during the measurements.

Tissue volume measurements.

A computer tomograph (model Prospeed VX, General Electric, Paris, France) was used to obtain precise estimates of muscle volume. It operated at a voltage of 120 kV, a current of 200 mA, and a scan-to-scan cycle time of 3.0 s. A standardized reconstruction algorithm and a 512 × 512 reconstruction matrix with a field of view of 35–45 cm were used, giving a spatial resolution in the range of 0.59–0.87 mm. In total, 21–25 serial segments, with a 10-mm thickness, were obtained every 20 mm along the thigh, starting distally from the patella and going in a proximal direction toward the spina iliaca anterior superior (SIAS). All segments taken of the first subject were analyzed by two experienced persons who were independent of the study. After comparison, verification, and agreement as to the definition of the different tissue structures, one of the analysts continued to analzye data from the remaining subjects.

The cross-sectional tissue areas, obtained from computer measurements on consecutive CT images along the thigh, are shown in Fig.1. The figure emphasizes the fact that, because the muscle configuration markedly varies along its length, a series of frequent cross-sectional images has to be obtained to make true volume estimates. The tissue volume was calculated from the mean area of two neighboring sections multiplied by the section distance, summed over the full length of the tissue. Moreover, the muscle mass of the CT measurements were calculated by assuming a muscle tissue density of 1.049 kg/l (14).

Fig. 1.

Multiple cross sections (2 cm apart) of the thigh from proximal edge of patella toward spina iliaca anterior superior measured with computer tomography in 9 subjects. Cross sections of total thigh over the length of the quadriceps femoris muscle are shown. Values are means ± SE. CT, computed tomography.

The mean quadriceps muscle volume measured by CT was specifically compared with traditional anthropometric measurements (2, 8, 9). The anthropmetric thigh volume (V) was calculated from the formulaV=L(12π)1(O12+O22+O32) (S0.4)21L(O1+O2+O3)31 whereL is the estimated muscle length (see below); O1, O2, and O3 are the circumferences 10 cm above the middle, at the middle, and 10 cm below the middle of the segment of the muscle length, respectively; and S is the correction for subcutaneous fat measured by skinfold callipers at each circumference (2, 8). The anthropometric quadriceps femoris muscle mass (M qf) was calculated as previously described, whereM qf = 0.307 ⋅ V (l) + 0.353 kg (2). Because the muscle length, for the volume and mass calculation, has been estimated previously with muscle insertion points from patella to either the pubic bone (PB), inguinal ligament (IL), or spina iliaca anterior superior (SIAS), all three lengths were measured and separate calculations were performed.

Statistics and data analysis.

Parametric statistics were used for data analysis. Analysis of variance for repeated measures was used when more than two data groups were compared, and Tukey's honestly significant difference post hoc tests were used to distinguish the differences. Linear regression and Pearson's correlation were employed.P < 0.05 was considered as statistically significant. A nonstatistically significant comparison is indicated by P = NS. The values given in the text are means ± SE.


Estimates of muscle volume and mass.

The CT quadriceps femoris muscle volume averaged 2.36 ± 0.17 (range 1.31–3.27) liters, which corresponded to a mass (M qf.CT) of 2.48 ± 0.18 (range 1.37–3.43) kg. The anthropometric muscle mass estimates overestimated (P < 0.02) the CT measurement by ∼21% (M qf.PB, 3.00 ± 0.18 kg, range 2.60–3.99), 29% (M qf.IL, 3.21 ± 0.19 kg, range 2.15–4.20), and 46% (M qf.SIAS, 3.61 ± 0.22 kg, range 2.26–4.64), respectively (Fig.2 A). However, they all correlated with and were linearly related to each other (0.93 ≤ r ≤ 0.98,P < 0.001)Mqf=Mqf.CT0.924Mqf.PB0.292kg Mqf=Mqf.CT0.882Mqf.IL0.352kg Mqf=Mqf.CT0.792Mqf.SIAS0.382kg

Fig. 2.

A:quadriceps femoris muscle mass. B andC: peak muscle perfusion (Q˙peak mus) and peak muscle oxygen uptake (V˙o 2 peak mus), respectively, during 1-legged, dynamic, knee extensor exercise. Values are means ± SE. Quadriceps femoris muscle volume was measured with CT as well as anthropometrically with different muscle lengths estimated from patella to pubic bone (PB), inguinal ligament (IL), and spina iliaca anterior superior (SIAS). Quadriceps femoris muscle mass values measured with CT and anthropometry were all significantly different from each other (P < 0.02). As a consequence of different muscle mass estimates using anthropometric procedure,Q˙peak musandV˙o 2 peak muswere underestimated compared with CT-scan measurements by ∼18–33%. Significant differences (P < 0.05): * from CT, § from CT and PB, and † from CT, PB, and IL estimates.

Knee extensor exercise.

Leg blood flow (Q˙leg) and leg oxygen uptake (V˙o 2 leg) increased (P < 0.05) during submaximal exercise at 30 W from their resting baseline values of 0.24 ± 0.02 (range 0.17–0.35) l/min and 18.7 ± 2.6 (range 9.4–34.2) ml/min, respectively, to 3.57 ± 0.25 (range 2.69–5.01) l/min and 434 ± 24 (range 361–528) ml/min, respectively. A further increase (P < 0.05) was observed at peak effort (67 ± 7, range 55–100 W) to 5.99 ± 0.66 (range 4.15–9.52) l/min and 856 ± 109 (range 590–1,521) ml/min for Q˙leg and V˙o 2 leg, respectively.

As a consequence of the different muscle mass estimates using the anthropometric procedure, peak muscle perfusion (Q˙peak.qf) was underestimated compared with the CT scan measurement (Q˙peak.qf.CT, 246.2 ± 24.2 ml ⋅ min−l ⋅ 100 g−l, range 149.2–373.0) by ∼19% (Q˙peak.qf.PB, 199.5 ± 16.6 ml ⋅ min−l ⋅ 100 g−l, range 136.6–298.4); 25% (Q˙ peak.qf.IL , 185.7 ± 14.4 ml ⋅ min−l ⋅ 100 g−l, range 129.2–266.7); and 33% (Q˙peak.qf.SIAS, 166.0 ± 13.9 ml ⋅ min−l ⋅ 100 g−l, range 115.0–241.0), respectively (Fig.2 B). Peak muscle oxygen uptake (V˙o 2 peak.qf) was underestimated with anthropometry compared with the CT-scan measurement (V˙o 2 peak.qf.CT, 35.2 ± 3.7 ml ⋅ min−l ⋅ 100 g−l, range 22.6–59.6) by ∼18% (V˙o 2 peak.qf.PB, 28.7 ± 2.8 ml ⋅ min−l ⋅ 100 g−l, range 19.8–47.7); 24% (V˙o 2  peak.qf.IL , 26.6 ± 2.4 ml ⋅ min−l ⋅ 100 g−l, range 19.6–42.6); and ∼32% (V˙o 2 peak.qf.SIAS, 23.8 ± 2.2 ml ⋅ min−l ⋅ 100 g−l, range 16.8–38.5), respectively (Fig. 2 C).


This study demonstrates that anthropometric muscle volume estimates markedly overestimate the knee extensor muscle volume, which in turn means that previous anthropometrically based estimates of peak muscle perfusion and oxygen uptake in humans, using various types of blood flow measurements, are underestimated by up to ∼30% (2, 21, 24). In light of the fact that the initial thermodilution blood flow measurements were criticized for being too high, as peak muscle perfusion and oxygen uptake were studied in a single active muscle group (2), these findings are of major importance because they emphasize that the peak values are even higher than previously found. As the magnitude of the thermodilution measurements also have been verified by ultrasound Doppler and indocyanine green dye infusion (2,10, 16, 26), the low peak values found by133Xe clearance and plethysmography seem to be even less believable.

In this study multiple CT image scans were taken along the full length of the quadriceps muscle (Fig. 1) of subjects with a large variation in the length (L) of the quadriceps muscle (L CT 42.4 ± 0.73 cm, range 38–46) and its points of insertion. Furthermore, the large variation in muscle length with the three different anthropometric approaches (L PB 38.5 ± 0.62 cm, range 36–41;L IL 41.0 ± 0.54 cm, range 39–43.5;L SIAS 45.3 ± 0.65 cm, range 40.5–47) induces a great source of error in the muscle volume and mass calculations. Moreover, the configuration of the muscle varies markedly along its length (Fig. 1). Thus only three circumference measurements or CT image scans along the thigh may not give an accurate estimate of muscle volume (2, 8, 27). In addition, the mathematical formulas used to estimate the muscle volume on the basis of anthropometry do not account for the variation in muscle configuration, because they are based on the fact that the muscle is constructed from two symmetrical “cones” that narrow toward the ends of the muscle. More importantly, the anthropometric formulas do not account for either the relative area that the quadriceps muscle group occupies in the thigh region or the fact that the quadriceps muscle volume may vary. In the present study, this variation in quadriceps muscle volume ranged from ∼1.31 to 3.27 liters. The present study thus provides a set of formulas that relate the anthropometric estimates to the CT measurements, which potentially can be used to correct for the anthropometric overestimations. However, one should be careful with such a generalization because the formulas are based on results from nine subjects only and the relationship is likely to be affected by different types and levels of activity, inactivity, gender, and age. If anthropometry is used to estimate muscle mass, muscle length estimated from the patella to the PB induces the smallest error in the muscle volume and mass estimate.

Besides providing a relationship between anthropmetric and CT estimates of muscle volume, knowing the number of CT or magnetic resonance image (MRI) sections that need to be performed is also of interest to allow a reasonable measure of accuracy. In the present data, the maximum width of the quadriceps muscle was located approximately one-half the distance between the greater trochanter and the top of patella. Thus, if muscle volume is estimated by using an area from the section of maximum width and from two other sections, imaged at one-half the distance between the maximum width and the greater trochanter as well as at one-half the distance between the maximum width and the patella, respectively, the three sections would give a muscle volume of ∼2.20 liters, i.e., an underestimation of ∼7%. If an estimate of the missing muscle volume proximally and distally, respectively, to the two outer sections is added, assuming a decrease in muscle size toward its end points, as in the case of two narrowing cones, then the muscle volume would be ∼2.64 liters, i.e., an overestimation of ∼12%. If the estimate is based on five sections, the first section halfway between the greater trochanter and patella and then four additional sections measured every 8 cm, i.e., two on each side of the first section, then the muscle volume would be ∼2.37 liters, i.e., an overestimation of ∼0.5%. By adding the missing outer cone-shaped muscle volumes, the volume would be ∼2.46 liters, i.e., an overestimation of ∼4.2%. Similarly, 13 sections taken every 4 cm, starting from the same midpoint between the greater trochanter and patella, would render an overestimation of only ∼0.4%. Thus, to carefully describe muscle volume, as many imaged sections as are in the range of every every 2–4 cm are needed, because the shape of the muscle varies individually along its full length. However, five sections may give a reasonable estimate, but one should be aware that our sample of subjects is fairly small.

There is ample evidence for the whole quadriceps muscle to be engaged in very intense knee extensor exercise (11, 15, 18). However, at submaximal efforts the recruitment pattern has been less clear. Electromyography tracings from the various portions of the quadriceps muscle suggest a gradual and equal involvement. Ray and Dudley's (15) recent findings using T2-weighted MRI scans support this notion but more importantly add a value for the muscle mass being recruited at different intensities. On the basis of their findings, with a muscle volume “engagement” of ∼16% at rest, ∼54% at 20 W, and ∼94% at 40 W, and assuming ∼100% at peak effort, a linear relationship can be expressed as the recruited muscle volume (%) = 25.187 + 1.285 ⋅ load (W). Interpolation gives the notion that in the present study ∼64% of the muscle volume could have been engaged at 30 W, rendering values for muscle perfusion and oxygen uptake of ∼235 ± 24 (range 154–383) and 28 ± 2.3 (range 21–41) ml ⋅ min−l ⋅ 100 g−l, respectively, on the basis of blood flow and oxygen uptake in the present study (Fig.3). This can be compared with ∼151 ± 15 (range 99–245) and 18 ± 1.4 (range 14–26) ml ⋅ min−l ⋅ 100 g−l, respectively, values that are found if, as in the past (2), blood flow and oxygen uptake in the present study are related to the whole quadriceps muscle mass. Thus, as pointed out by Ray and Dudley, peak muscle perfusion may be achieved in certain regions at already submaximal intensities. However, muscle oxygen uptake appears to be further elevated at peak effort, possibly due to a larger oxygen extraction (Fig. 3). A tentative explanation for higher oxygen uptake per unit of muscle at increasing work intensities despite recruitment of more muscle to perform the work could be that first the frequency of activation is increased and then energy turnover.

Fig. 3.

Muscle perfusion (Q˙mus;A) and muscle oxygen uptake (V˙o 2 mus;B) during 1-legged, dynamic, knee extensor exercise, at 30 W and at peak effort (∼ 67 ± 7 W), on the basis of Q˙ andV˙o 2 of the present study. Values are means ± SE. Values are estimated, as if the whole quadriceps femoris (qf) muscle volume was recruited (qf 100%), or only a portion of the muscle (qf 64%) on the basis of findings of Ray and Dudley (15). ThusQ˙peak musmay be achieved in portions of muscle already at intensities lower than peak effort. Significant differences (P < 0.05): * from exercise at 30 W related to recruitment of 100% of qf muscle mass, and ** from exercise at 30 W related to recruitment of 100 and 64% of qf muscle mass.

In conclusion, anthropometric overestimates of muscle volume in the past have markedly underestimated peak muscle perfusion and oxygen uptake. Thus peak perfusion and oxygen uptake in human skeletal muscle may be up to ∼30% higher than previous estimates on the basis of blood flow measurements with techniques equivalent to themodilution (2, 16, 19-22, 26), making values obtained by133Xe clearance and plethysmography even less believable. Moreover, to obtain an estimate of muscle mass recruited when muscle perfusion and muscle oxygen uptake are estimated at intensities lower than at peak effort, T2-weighted MRI is a necessity, in addition to either CT or MRI muscle volume determinations.


The present work was supported by a grant from the Danish National Research Foundation (504–14).


  • Address for reprint requests and other correspondence: G. Rådegran, Copenhagen Muscle Research Centre, Rigshospitalet, Section 7652, Blegdamsvej 9, DK-2100 Copenhagen 0, Denmark (E-mail:goran{at}

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.


View Abstract