Comparison of net anaerobic energy utilisation estimated by plasma lactate accumulation rate and accumulated oxygen deficit in Thoroughbred horses


  • H. OHMURA,

    Corresponding author
    1. Equine Research Institute, Japan Racing Association, Tochigi, Utsunomiya, Tokami-cho, Japan; and Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California, Davis, USA.
    Search for more papers by this author
  • K. MUKAI,

    1. Equine Research Institute, Japan Racing Association, Tochigi, Utsunomiya, Tokami-cho, Japan; and Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California, Davis, USA.
    Search for more papers by this author

    1. Equine Research Institute, Japan Racing Association, Tochigi, Utsunomiya, Tokami-cho, Japan; and Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California, Davis, USA.
    Search for more papers by this author
  • A. MATSUI,

    1. Equine Research Institute, Japan Racing Association, Tochigi, Utsunomiya, Tokami-cho, Japan; and Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California, Davis, USA.
    Search for more papers by this author
  • A. HIRAGA,

    1. Equine Research Institute, Japan Racing Association, Tochigi, Utsunomiya, Tokami-cho, Japan; and Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California, Davis, USA.
    Search for more papers by this author
  • J. H. JONES

    1. Equine Research Institute, Japan Racing Association, Tochigi, Utsunomiya, Tokami-cho, Japan; and Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California, Davis, USA.
    Search for more papers by this author



Reasons for performing study: Accumulated O2 deficit (AOD) and plasma lactate accumulation rate (PLAR) are alternative methods for estimating net anaerobic energy utilisation (NAEU) in exercising horses. How they compare or their accuracy is unknown.

Objectives: We hypothesised net anaerobic energy utilisation calculated by PLAR (NAUEPLAR) is equivalent to NAUE estimated by AOD (NAUEAOD).

Methods: Six Thoroughbred horses ran at identical supramaximal speeds (118% aerobic capacity) until exhaustion for 2 runs while breathing normoxic (NO, 21% O2) or hyperoxic (HO, 26% O2) gas. Jugular blood was sampled at 15 s intervals to measure plasma lactate concentration. Horses also ran at incremental submaximal speeds from 1.7–11.0 m/s to determine the linear relationship between speed and O2 consumption to estimate O2 demand for AOD calculations.

Results: Maximum O2 consumption of horses increased 11.6 ± 2.3% in HO and NAEUPLAR and NAUEAOD decreased 38.5 ± 8.0% and 46.2 ± 17.7%, respectively. The NAEUPLAR in NO was 114.5 ± 27.4 mlO2 (STPD) equivalent/kg bwt contributing 23.5 ± 3.7% to total energy turnover and in HO was 70.9 ± 19.8 mlO2 (STPD) equivalent/kg bwt contributing 14.6 ± 3.8% to total energy turnover. The NAUEAOD in NO was 88.6 ± 24.3 mlO2 (STPD) equivalent/kg bwt contributing 19.9 ± 2.1% to total energy turnover and in HO was 56.2 ± 19.1 mlO2 (STPD) equivalent/kg bwt contributing 10.9 ± 4.3% to total energy turnover. Overall, NAEUAOD was systematically biased -23.5 ± 16.8 mlO2 (STPD) equivalent/kg bwt below NAEUPLAR. Total energy demand estimated by PLAR was 11.1 ± 5.4% greater than that estimated by AOD and was higher in every horse.

Conclusions: The NAUEPLAR estimates average 40.0 ± 29.6% higher than NAUEAOD and are highly correlated (r2= 0.734), indicating both indices are sensitive to similar changes in NAEU. Accuracy of the estimates remains to be determined. Multiple considerations suggest NAUEAOD may underestimate total energy cost during high-speed galloping, thus biasing low the AOD estimate of NAEU.


Accumulated O2 deficit


Hyperoxia (26% O2)


Heart rate


Maximal accumulated O2 deficit


Net anaerobic energy utilisation


Net anaerobic energy utilisation calculated by AOD method


Net anaerobic energy utilisation calculated by PLAR method


Net anaerobic power


Normoxia (21% O2)


Plasma lactate accumulation rate


Respiratory exchange ratio

inline image

Rate of metabolic CO2 production

inline image

Rate of metabolic O2 consumption

inline image

Maximum rate of O2 consumption


Horses increase energy demand and ATP turnover when they run. Replenishment of high-energy PO4 to i.m. ATP is primarily accomplished by oxidative phosphorylation if sufficient O2 is available to the cells. If not, anaerobic glycolysis with the formation of lactate as the end product is the major source of energy. Although lactate is continuously being generated in the body, it only accumulates and increases its concentration in the blood if its rate of production exceeds its rate of removal by organs that use it as a carbon substrate for the Krebs Cycle or gluconeogenesis, e.g. skeletal muscle or liver (Cori cycle) (Brooks 1986). Therefore, lactate only accumulates in blood and increases in concentration if net anaerobic power (NAP) is contributing to the animal's total power output. The accumulation of lactate in the blood during exercise indicates that not all metabolic power is derived from aerobic metabolism.

It is relatively simple to estimate the aerobic power of an animal from its rate of O2 consumption (inline image) and the biochemical stoichiometry that approximately 20.1 J of energy are liberated for every 1 mlO2 (STPD) consumed in aerobic metabolism (Schmidt-Nielsen 1997). However, anaerobic power is much more difficult to estimate. In man, the technique of maximal accumulated O2 deficit (MAOD) was developed to calculate net anaerobic energy utilisation (NAEU) as the deficit between measured aerobic energy supplied and the total metabolic energy required during maximal exercise to fatigue, extrapolated from measured submaximal inline image (Medbøet al. 1988). A similar approach has been used for estimating anaerobic capacity in horses (Eaton et al. 1995; Hinchcliff et al. 1996, 2002; Tyler et al. 1996; Lacombe et al. 1999, 2001; Geor et al. 2000; Prince et al. 2002). A problem with validating this technique in horses is that there is no ‘gold standard’ with which to measure NAEU.

Horses increase their aerobic capacity (inline image) when breathing hyperoxic (HO) gas compared with breathing normoxic (NO) gas (Jones et al. 1988; Jones 1994, 1998; Wagner et al. 1996; Ohmura et al. 2006). The inline image increases more than 10% when inspired O2 concentration is raised by only 4%, the higher inspired O2 partial pressure overcoming mechanisms, e.g. hypoventilation and diffusion limitation, that normally compromise O2 transport and cause severe hypoxaemia in heavily exercising horses (Wagner et al. 1989; Jones 1994, 1998). The increase in inline image and supply of aerobic energy to meet demand in HO should be accompanied by a stoichiometric decrease in NAP, an index of which is plasma lactate accumulation rate (PLAR) (Seeherman et al. 1981). Relating the decrease in PLAR to the increase in inline image yields a quantitative index of how much NAP a horse is utilising when lactate accumulates in the blood at a given rate (Ohmura et al. 2006).

It is unknown how these 2 methods for calculating NAEU compare with each other, considering they are based on different approaches and assumptions and are estimating different, but related variables: total aerobic energy deficit (= sum of NAEU, AOD) vs. NAP (= rate of NAEU, PLAR). However, PLAR can theoretically be used to generate an index of total NAEU by integrating the rate (NAP) with respect to time during a run; both indices would then have identical units of total net anaerobic energy equivalents utilised in the exercise bout. The purpose of this study was to evaluate the relationship between NAEU calculated by AOD (NAEUAOD) with NAEU calculated by integrating PLAR estimates of NAP (NAEUPLAR) during the same supramaximal runs. We hypothesised that NAEU estimates by PLAR and AOD should be related to each other and yield similar estimates of net anaerobic contribution to total energy turnover.

Materials and methods

A protocol for the study was reviewed and approved by the Animal Use and Care Committee and the Animal Welfare and Ethics Committee of the Equine Research Institute, Japan Racing Association, where the study was conducted.


Six Thoroughbreds (3 geldings, 2 males, 1 female, 4.5 ± 1.0 [s.d.] year, 477 ± 18 kg bwt) were studied. Horses were exercised 5 days/week on a treadmill (Säto I)1 at a 6% incline for a period of more than 6 week prior to measurements, and ran 3 times/week at speeds of 11–13 m/s. The horses were all clinically normal and had unremarkable medical histories.

Treadmill protocol

For each run, a 14 gauge catheter was percutaneously placed in a jugular vein and a surcingle with ECG electrodes placed around the horse's thorax to measure heart rate (HR) with a commercial HR monitor (S810)2. Horses ran for 3 data collection experiments. For each run, the horse warmed-up by walking for 3 min at 1.7 m/s then trotted for 2 min at 3.5 m/s. Following the warm-up, an open-flow mask was placed over the horse's muzzle. For the first run, horses ran up a 6% incline for 1 or 2 min each at incremental speeds of 1.7 or 1.8 m/s, 3.5 or 4.0 m/s, 6 m/s, 7 or 8 m/s, 10 and 11 m/s to determine the linear relationship between speed and O2 consumption (inline image) for estimation of total power output and O2 demand for AOD calculations. At 1 week intervals following, horses completed 2 protocols that were identical except for the gas they breathed. In the first run, they breathed normoxic gas (NO, 21% O2) while they walked for 2 min at 1.7 or 1.8 m/s, then trotted for 2 min at 3.5 or 4.0 m/s, cantered for 2 min at 7.0 m/s then galloped up a 6% incline until exhaustion (1.8–3.0 min) at a speed (12.7–14.0 m/s) estimated to require approximately 120% inline image based on the first run's data and previous measurements of the horses' inline image. Horses followed this a week later with an identical protocol except they breathed hyperoxic (HO, 26% O2) gas through a semi-open-flow mask. Horses ran for the same duration as the NOinline image run rather than running to exhaustion in HO. Blood was sampled from the jugular catheter at 15 s intervals for measurement of plasma lactate concentration.

Gas mixing, bias flow and sampling

The mask was made of a T-shaped piece of lightweight 20 cm diameter PVC connected to flexible 20 cm diameter PVC tubing on the 2 side arms of the T. On the centre tube of the mask a foam and rubber diaphragm and gasket was mounted that fitted around and over the horse's muzzle and, when taped to the muzzle, formed a gas-tight seal. Ropes that passed through overhead pulleys and elastic cords attached to the ceiling supported the weight of the mask. All joints in the flow system upstream of the gas analysers were sealed with duct tape.

A mixing chamber was connected to the upstream flexible tubing on the mask through which a flow of diluent O2 was blown into the upstream end of the bias-flow system and mixed with a bias flow of air of 100–150 ℓ (ATP)/S to create the desired inspired O2 concentration. Diluent gas flow was controlled with a mass flow controller3 with maximum flow rate of 33.3 ℓ (STPD)/S connected to compressed gas cylinders through a gas manifold. Diluent gas flow was adjusted as needed to maintain 26% O2 by monitoring the concentration of O2 in the upstream arm of the mask with an O2 analyser4 when horses ran with HO; no diluent gas was used for NO runs.


The inline image was measured with the N2-dilution technique (Fedak et al. 1981) as modified for use in semi-open-flow systems in which inspired O2 concentration is not 20.94% (Ohmura et al. 2006) and the rate of CO2 production (inline image) was measured by CO2 bleed (Birks et al. 1991). Gas samples were measured with O2 and CO2 analysers4 after gas was dried by passing through a 2 m long Nafion drying tube5 with countercurrent dry gas flow (Drierite)6 (CaSO4) to remove H2O, and Ascarite7 (NaOH) removed CO2 before the O2 analyser. An electronic mass flow controller4 measured N2 and CO2 calibration flows. Respiratory exchange ratio (RER) was calculated as inline image.

Blood sampling and lactate analysis

At the start of the gallop, blood was sampled from the jugular catheter into heparanised syringes at 15 s intervals for measurement of plasma lactate concentration. Samples were stored on ice, centrifuged to separate plasma, and lactate concentration was measured in duplicate on a lactate analyser8. The PLAR was calculated as the slope of the rate of plasma lactate concentration increase following the first 30 s of galloping, after which the concentration tends to increase linearly (Seeherman et al. 1981; Birks et al. 1991; Ohmura et al. 2006).

Calculation of net anaerobic energy utilisation and power with AOD and PLAR

For estimation of NAEU by the 2 techniques, we assumed that the energy cost of running at identical supramaximal speed in NO and HO was equal.

Accumulated O2 deficit was calculated by subtracting the measured inline image during the supramaximal runs from the estimated total energy demand. Total energy demand was estimated by extrapolating the linear relationship between running speed and inline image during the initial submaximal incremental exercise steps performed separately from the supramaximal runs to the running speed selected for the supramaximal runs. This speed was chosen to elicit approximately 120% inline image for each horse. The difference in the measured rate of O2 consumption during the supramaximal runs and the extrapolated cost estimate when integrated with respect to time yields a volume of O2 that is the AOD. This estimate of NAEU is then expressed in energetically-equivalent mass-specific units of mlO2 (STPD) equivalent/kg bwt.

The increase in inline image for a horse running at identical supramaximal speed in HO vs. NO represents an addition of aerobic power to meet its total energy demand that stoichiometrically reduces NAP. This reduced NAP, when integrated over the duration of the run, reduces NAEU. For each horse, the reduction in PLAR associated with increased inline image in HO was used to calculate the energetic equivalence of a unit of PLAR to a unit of inline image, i.e. its aerobic power equivalent. For each horse, this energy-equivalence factor was calculated for the decrease in its PLAR when inline image increased in HO. The measured PLAR for each horse was then converted by that factor into units of aerobic power equivalents and integrated with respect to time for the duration of that horse's runs to estimate its total energy demand. For both AOD and PLAR, the fractional contribution of NAEU to total energy demand during the supramaximal run was calculated as the ratio of NAEU to total O2 demand.

Statistical analysis

Data are shown as mean ± s.d. Linear regression evaluated the energy cost relationship for submaximal exercise for extrapolation to estimate total energy cost for the AOD technique as well as the association between the 2 NAEU estimates. Analysis of covariance (ANCOVA) tested for differences between slopes of regressions. The lower limits of the 95% confidence interval were calculated to determine if PLAR at individual speeds included 0 or not. A paired Student's t test was used to evaluate if variables differed between techniques. A Kolmogorov-Smirnov test determined normality of comparative method data evaluated with a Bland-Altman plot. SigmaPlot 11 software9 was used for statistical procedures and P≤0.05 was considered significant.


Horses ran at steady-state for estimation of energy cost of locomotion over ranges of speed that varied from 1.8–7.0 m/s to 1.7–11.0 m/s (Fig 1). Slopes of the aerobic energy-cost relationship varied from 0.220–0.265 mlO2 (STPD)/(kg m) for individual horses. The ANCOVA detected significant differences between the slopes (P<0.001). Estimates of the speed required to elicit inline image averaged 11.4 ± 0.8 m/s.

Figure 1.

Specific O2consumption of 6 horses walking, trotting and cantering/galloping up a 6% incline on a treadmill. For each horse, the measured aerobic power (solid line) is extrapolated (long dashed line) linearly to the speed (18% higher than that predicted to elicitinline image) at which the horse ran for its AOD measurement to estimate its total aerobic-equivalent energy cost at that speed. For each horse, the lower horizontal dashed line is the horse'sinline imagewhen breathing NO (21% O2) and the upper horizontal dashed line is itsinline imagewhen breathing HO (26% O2). Differences between theinline imagelines and the extrapolated aerobic power line (shaded areas) represent the magnitudes of the aerobic power shortfall at speeds above that eliciting NO or HOinline imagefor each horse. Letters indicate individual horses. For horse SH, the open circle represents the speed (14 m/s) extrapolated to be required for the horse to achieve theinline imagemeasured in HO, although the horse actually ran at 13.4 m/s for its high speed runs.

In order to determine if the submaximal estimates of energy expenditure were completely aerobic, we calculated PLAR at each of the submaximal speeds above the walk used for estimating total energy demand with the AOD technique (Fig 2). The lower limits of the 95% confidence intervals for all speeds above the trot (6–11 m/s) exceeded zero, although the values only exceeded 0.4 mmol/min for the 2 highest speeds (10 and 11 m/s).

Figure 2.

Mean specificinline image(circles, linear regression) and PLAR (triangles, curvilinear regression) for 6 horses at speeds used to estimate total energy cost for AOD calculations (individual horse data shown in Fig 1). Significant accumulation of lactate at the higher submaximal work rates indicates that total energy costs were higher than those predicted byinline imagealone. Mean values shown at each speed were based on 2–4 horses each, bars are ± 1 s.d. * indicates PLAR for which the lower limit of the 95% confidence interval (CL) was 0<CL<0.4 mmol/l/min; ** indicates PLAR values for which lower limit of the 95% CL>2.0 mmol/l/min.

Horses became exhausted and could not maintain position on the treadmill after 1.8–3.0 min of galloping at the supramaximal speeds selected (13.5 ± 0.5 m/s). Specific inline image of the horses in NO was 165 ± 9 mlO2 (STPD)/(min kg) (55.4 ± 2.9 W/kg) and in HO (25.7 ± 0.7% O2) was 185 ± 10 mlO2 (STPD)/(min kg) (61.8 ± 3.4 W/kg), 11.6% higher. There was no difference in HR in NO vs. HO, 211 ± 7.2 beat/min−1 vs. 214 ± 3.9 beat/min−1, respectively, however, RER was higher in NO than in HO, 1.21 ± 0.10 in NO vs. 1.12 ± 0.09 (P<0.05).

Technical problems with catheters prevented collection of samples for all time points for all runs; however, all estimates of PLAR were based on a minimum of 5 points. The r2 for all regressions of plasma lactate concentration vs. time (PLAR) exceeded 0.95 and for 7 of the estimates exceeded 0.99. The slopes of the PLAR regressions for every horse decreased from their NO values when the horses ran in HO and inline image increased, from 11.2 ± 4.3 mmol/min to 6.8 ± 2.4 mmol min (Fig 3). The slopes differed significantly between treatments for 4 of the 6 horses. These values yielded an estimate of energy equivalence for PLAR and inline image of 5.08 ± 1.8 mmol/min equalling 60 mlO2 (STPD)/(min kg). For every horse, the measured or extrapolated (from the PLAR regression) end-run lactate concentration in NO was higher than in HO.

Figure 3.

Time course of specificinline imageand plasma lactate accumulation in 6 horses galloping up a 6% incline on a treadmill at speeds predicted to elicit 118% NOinline image. Solid symbols (circles,inline image; triangles, plasma lactate concentration) are while breathing NO (21% O2) and open symbols are while breathing HO (26% O2). Solid and medium-dashed lines representinline imageand PLAR for NO and HO gas runs, respectively. Horizontal dotted line represents the calculated aerobic-equivalent power cost for running that speed based on the aerobic power extrapolation shown in Figure 1 by the AOD method. Letters indicate individual horses as in Figure 1.

The mean NAEUAOD in NO and HO were 88.6 ± 24.3 mlO2 (STPD) equivalent/kg and 56.2 ± 19.1 mlO2 (STPD) equivalent/kg, respectively, a 46.2 ± 17.7% decrease and were lower in every horse in HO (Fig 4). The mean NAUEPLAR in NO and HO were 114.5 ± 27.4 mlO2 (STPD) equivalent/kg and 70.9 ± 19.8 mlO2 (STPD) equivalent/kg, respectively, a 38.5 ± 8.0% decrease and were also lower in every horse in HO (Fig 4). The estimates between the techniques are related by the regression equation:

Figure 4.

Calculated NAEU for 6 horses during runs at speeds calculated to elicit 118% NOinline imagewhile breathing NO (21% O2, solid symbols and line) and while breathing HO (26% O2, open circles and dashed line). Ordinate represents NAEU estimated by AOD technique, abscissa is NAEU estimated by PLAR technique for the same runs. Letters indicate individual horses as in other figures.


The value of every estimate of NAUEAOD was lower than the corresponding value of NAEUPLAR (Fig 4).

Total energy turnover estimated by AOD and PLAR were 444 ± 113 mlO2 (STPD) equivalent/kg and 490 ± 108 mlO2 (STPD) equivalent/kg, respectively, an 11.1% difference. The AOD technique estimated NAEU contributed 19.9 ± 2.1% to total energy turnover in NO and 10.9 ± 4.3% in HO. The PLAR technique estimated NAEU contributed 23.5 ± 3.7% to total energy turnover in NO and 14.6 ± 3.8% in HO.

A Bland-Altman plot (Bland and Altman 1986) of the data (Fig 5) assessed bias and agreement of the 2 methods. The mean difference (bias) of the AOD vs. PLAR estimates of NAEU is −23.5 ± 16.8 mlO2 (STPD) equivalent/kg bwt with 95% confidence intervals of −13.7 to −33.2 mlO2 (STPD) equivalent/kg bwt (the differences are normally distributed [P = 0.121]). The limits of agreement between the methods are 9.5 to −56.5 mlO2 (STPD) equivalent/kg bwt.

Figure 5.

Bland-Altman diagram comparing estimates of NAEU during supramaximal runs in 6 horses while breathing NO (21% O2, solid circles) or HO (26% O2, open circles). Letters indicate individual horses as in other figures. Dotted horizontal lines represent ± 1 and 2 s.d. of the measured differences between the estimates (NAEUAOD– NAEUPLAR), abscissa is the average of the estimates.


Maximal accumulated O2 deficit has been, until recently, the only method available to assess anaerobic capacity in horses. Maximal accumulated O2 deficit is calculated by subtracting the measured inline image from the estimated O2 demand obtained by extrapolating the linear relationship between running speed and inline image during submaximal exercise (Medbøet al. 1988). Medbøet al. (1988) reported that AOD reached a maximum value for exhaustive bouts of running lasting 2–5 min in man. They defined this maximum value as the anaerobic capacity. This method has been applied to horses (Eaton et al. 1995; Hinchcliff et al. 1996, 2002; Tyler et al. 1996; Lacombe et al. 1999, 2001; Geor et al. 2000; Prince et al. 2002); however, some controversy has attended the interpretation of estimates of anaerobic capacity from AOD as there is no ‘gold standard’ against which to compare it, and its accuracy and reliability are unknown.

Specific inline image of highly-trained Thoroughbred horses is more than twice as high as that of the most highly-trained human athlete (Saltin and Astrand 1967; Seeherman and Morris 1991) and Thoroughbreds also tolerate peak plasma lactate concentrations 1.5–2 times higher than human athletes (Marlin et al. 1991; Brooks et al. 2005). However, MAOD of horses were reported to be 80–110 mlO2 (STPD) equivalent/kg bwt (Eaton et al. 1995; Hinchcliff et al. 1996, 2002; Lacombe et al. 1999, 2001; Geor et al. 2000; Prince et al. 2002), similar to values reported as MAOD for human subjects: 40–100 mlO2 (STPD) equivalent/kg bwt (Medbøet al. 1988; Medbø and Burgers 1990; Scott et al. 1991; Weber and Schneider 2001). It seemed surprising that anaerobic capacities should be so similar when aerobic capacities between these species differ so greatly. Given there is no ‘gold standard’ with which to assess the accuracy of the MAOD estimate, Ohmura et al. (2006) developed an alternative method for assessing net anaerobic contribution to total metabolic power using PLAR.

The PLAR technique estimates NAEU based on a different principle than the AOD technique. The PLAR technique calculates the aerobic power equivalent of a given PLAR by calculating the ratio of their inverse changes when inline image is changed by altering the inspired O2 fraction. For AOD, extrapolation of the submaximal energy cost line is used to predict total energy cost during supramaximal exercise, and the difference between that value and the measured aerobic energy contribution is used to calculate the net anaerobic energy contribution (areas between the inline imagecurves and horizontal dotted line in Figure 3). The 2 techniques are based on different assumptions. Both techniques assume, in the context of this experiment, that the total energy cost of running at identical supramaximal speed in HO is the same as NO. The AOD technique assumes the energy cost of submaximal locomotion is met entirely aerobically and that the cost increase remains linear at high galloping speeds. The PLAR technique assumes the rate at which lactate accumulates in the plasma is proportional to the rate at which it is being produced in working muscles in excess of its uptake by other tissues. Data supporting this assumption are shown in Figure 2 of Seeherman et al. (1981) - graphs of inline image and PLAR as functions of running speed for 9 species of mammals with specific inline image that vary by 440%. Although the species vary by 775% in their energy cost of locomotion, all species have similar linear slopes for their individual inline image vs. speed relationships below inline image and their PLAR vs. speed relationships above inline image. Indeed, the inline image energy stoichiometry ratio (11 mmol plasma lactate/min = 60 mlO2 (STPD)/min) that yields those equivalent slopes is the same that Ohmura et al. (2006) found best fit their HO vs. NO data in horses. In our experiment, the ratio was about half that value (5.1 ± 1.8 mmol/min). The PLAR technique also assumes there will be a stoichiometric decrease in PLAR with the increase in inline image in HO.

The values we calculated for AOD in NO in these experiments fall within the range previously reported for equine MAOD. It should be noted that while the value measured for NAEUAOD in these experiments is equivalent to MAOD in the previous reports, the values we measured for NAEUAOD in HO are not MAOD values. Our experiments were designed to run horses for equivalent duration in both gases, whereas, to measure MAOD in HO they presumably would need to run longer to reach exhaustion (and maximum AOD) while breathing HO. In this experiment, horses breathing HO were stopped after running the same duration required to elicit exhaustion in NO without showing signs of exhaustion in HO. Even as Eaton et al. (1995) found that MAOD was essentially constant when horses reached it with run durations inversely proportional to speed, we would expect horses to run longer in HO before fatiguing because their inline image is higher and PLAR is lower and it should take them longer to reach an equivalent AOD as in NO. The fact that none of the horses were fatigued after the same run duration in HO supports that interpretation.

For all horses, HR were near typical maximal rates at inline image (Birks et al. 1991) and did not differ between gases. The RER were greater than 1.0 during high-speed runs when breathing both NO and HO, indicating the horses were still exercising at inline image in HO. However, the fact that RER decreased for all horses in HO in conjunction with their higher inline image suggests they were more dependent on aerobic power and had lower NAEU in HO.

Several previous studies have exercised horses in HO (Jones et al. 1988; Pelletier and Leith 1995; Wagner et al. 1996; Jones 1998; Ohmura et al. 2006) and have shown that horses increase inline image from 12–20% above NOinline image when they run with inspired O2 concentration of 25–35%. In this study, horses increased inline image in HO by 12% over NO. Increased inline image in HO is a precondition for using PLAR to estimate NAP.

The NAEUPLAR is highly correlated (r2= 0.734, Fig 4) with the NAEUAOD calculated for the same runs, suggesting that both techniques are sensitive to the relative contribution of NAEU even though they are based on 2 very different approaches. The slope of the regression between the 2 techniques (m= 0.807) did not differ significantly from 1. However, despite the relative agreement in the estimates, Figure 4 shows that all estimates of NAEUAOD are lower than the corresponding NAEUPLAR.

To further examine agreement between the techniques, we plotted the data as a Bland-Altman diagram (Fig 5). This showed no systematic relation between the differences of the estimates and their means. This indicates the estimate of NAEU by AOD was consistently biased lower than PLAR by −23.5 ± 16.8 ml O2 (STPD) equivalent/kg bwt with 95% confidence intervals of −13.7 to −33.2 mlO2 (STPD) equivalent/kg and limits of agreement between the methods of 9.5 to −56.5 mlO2 (STPD) equivalent/kg bwt.

Although the 2 techniques showed good correlation as indices of relative changes in NAEU, there is no validated reference available for quantifying net anaerobic energy or power in horses. In man, it is possible to utilise a cycle ergometer to quantify the mechanical power output of an individual when estimating associated increased metabolic energy demand, but this can't be done directly in horses.

Why might AOD-technique estimates of NAEU consistently be biased below those of PLAR? It is possible that one of the assumptions inherent in the techniques is erroneous, e.g. that the PLAR slope does not change stoichiometrically with the change in inline image in HO. This assumption is difficult to validate or falsify. However, there are at least 5 lines of reasoning that seriously question the fundamental assumption of the AOD technique that the energy-cost curve extrapolates linearly from the submaximal inline image line during high-speed galloping:

  • 1) Significant accumulation of plasma lactate at submaximal speeds considered to be fully aerobic (Fig 2) suggests that total energy cost is higher than that calculated from only inline image, biasing low the extrapolated estimate of high-speed cost with the AOD technique.
  • 2) AOD estimates of total energy demand based on extrapolated submaximal power for 2 of the horses were less than or equal to the horses' measured inline image in HO and the horses still had significant PLAR; clearly the AOD estimates were too low.
  • 3) Extrapolation to higher speeds of the U-shaped energy-cost vs. speed curve for ponies cantering/galloping (Hoyt and Taylor 1981) predicts hyperbolically-increasing energy cost for high-speed galloping.
  • 4) Direct measurements of total energy cost (fully aerobic with no lactate accumulation) in 50 kg pronghorn antelope (Antilocapra americana) show a linear increase in inline image with speed to approximately 7 m/s followed by a hyperbolic increase in inline image up to 14–15 m/s (Jones et al. 2001). These animals can attain these speeds aerobically because of their high specific inline image (210–300 mlO2 (STPD)/kg bwt) (Lindstedt et al. 1991; Jones et al. 2001). Because both pronghorns and horses run with allometrically-predicted values for biomechanical variables, e.g. cost of transport, stride frequency and gait-transition speeds (Taylor et al. 1970; Heglund et al. 1974; Fedak and Seeherman 1979; Heglund and Taylor 1988; Taylor 1994; Jones et al. 2001), their locomotor energy costs would be expected to follow similar patterns. Alterations in the body's ability to store and recover elastic energy with extended high-speed galloping stride lengths might be a primary factor causing the energy cost to increase at higher speeds (Alexander 1988).
  • 5) Application of the inline image stoichiometry measured by Seeherman et al. (1981) and Ohmura et al. (2006) predicts hyperbolically-increasing energy cost at high galloping speeds in horses (Jones and Carlson 1995). The stoichiometric value for the inline image ratio measured in our experiments was lower than in those previous estimates and would result in even higher estimated energy costs.

More work is needed to better define the energy cost of high-speed locomotion to assist in determining the accuracy of these techniques. Both techniques appear to be sensitive to similar changes in NAEU in horses galloping at speeds above inline image, and the AOD technique consistently measures lower values than does the PLAR technique. It is difficult to assess independently the accuracy of the PLAR technique, but several considerations argue that AOD estimates are biased low.

Conflicts of interest

None declared.

Manufacturers' addresses

1 Säto AB, Knivsta, Sweden.

2 POLAR, Kempele, Finland.

3 Model DPM3, Kofloc, Tokyo, Japan.

4 METS-900, VISE Medical. Chiba, Japan.

5 PermaPure, Toms River, New Jersey, USA.

6 Drierite, W.A. Hammond, Xenia Ohio.

7 Ascarite II, Thomas Scientific, Swedesboro, New Jersey, USA.

8 2300 STAT Plus lactate analyser, YSI Instruments, Yellow Springs, Ohio, USA.

9 SigmaPlot 11, Systat Software, Inc., San Jose, California, USA.