Performance parameters and post exercise heart rate recovery in Warmblood sports horses of different performance levels




Reasons for performing study: Standardised exercise tests are used for fitness evaluation of sports horses. Standards are described for Thoroughbreds and Standardbreds; however, limited information is available for Warmbloods.

Objectives: To establish normative standards of performance parameters and heart rate recovery (HRR) in Warmblood riding horses of different levels of fitness using a submaximal incremental exercise test (SIET) performed on a treadmill.

Methods: A SIET was carried out with 29 healthy and treadmill-accustomed Warmbloods: eleven 3-day event horses (TDE) and 18 horses from the National Equestrian Centre (NEC) competing in amateur jumping and/or dressage events. After a warm-up phase, horses performed 2 stages at trot and 3–5 stages at gallop at 6% incline. The first stage lasted 120 s, all others 90 s. Velocity (V) and heart rate (HR) were measured continuously and blood lactate concentration (LAC) at the end of each exercise stage. V at HR 150 and 200 beats/min (V150, V200), V and HR at 2 and 4 mmol/l LAC (V2, V4 and HR2, HR4, respectively) were calculated and compared between discipline groups. For reference values, horses were divided on the basis of the V4-results in good (GP) and average performers (AP) (performance groups). Five minute passive HRR was compared between performance groups. Fifteen NEC horses were retested within 1–3 months. Groups were compared with t tests and P<0.05 considered significant.

Results: Three-day event horses had higher V150, V2 and V4 values than NEC. GP had higher values in all performance parameters compared to AP. No differences were found between test and retest. GP mean recovery HR was different from that of AP from 120 s of recovery onwards.

Conclusion: Treadmill SIETs are suitable to objectify aerobic capacity in Warmblood riding horses. Normative standards were assessed for well and averagely-trained horses. The results can be referred to when diagnosing patients with exercise intolerance.


Evaluation of fitness is important in all equestrian sporting disciplines. Standardised exercise tests are widely recognised to be valuable for monitoring training progress and as a useful additional examination for patients with exercise intolerance.

However, standardisation of exercise tests under clinical and field conditions is often difficult. Besides the routinely controlled factors such as speed, incline and load of the rider, environmental variables such as temperature, humidity, surface conditions and airflow for efficient convective and evaporative heat loss (Marlin and Nankervis 2002) may influence the test results. All these arguments contribute to the difficulty of drawing comparisons between different exercise tests; without standardisation of these conditions, results cannot be compared without certain reservations. Moreover, previous studies were carried out using different test protocols, either on the treadmill or in the field. Some protocols and reference values have been previously described for Thoroughbred and Standardbred racehorses (Foreman et al. 1990; Rose et al. 1990; Seeherman and Morris 1990; Courouce 1999; Stahel 2004; Vermeulen and Evans 2006). Limited information is available for Warmblood horses (Sloet van Oldruitenborgh-Oosterbaan et al. 1987) and, in particular, there are as yet no performance data of horses competing in different sporting disciplines.

The assessment of exercise capacity is mainly based on heart rate (HR) response, blood lactate concentration (LAC) and oxygen uptake in relation to exercise intensity expressed as velocity (V) (Persson 1983; Evans 2007). The measurement of oxygen uptake is rather complicated, still confined to the laboratory and is, therefore, mainly reserved for research purposes. Consequently, HR and LAC measurements have become widespread in equestrian sports for evaluating fitness and controlling training intensity in the field. The response of HR and LAC to exercise is not only dependent on the aerobic capacity but may also be affected by the state of health and inherited parameters such as breed and body conformation (Marlin and Nankervis 2002).

Heart rate recovery (HRR) is monitored to assess different levels of fitness. In man, a marked interdependence of endurance exercise training on HRR was reported by Hagberg et al. (1980), where HR recovered faster in highly trained individuals. Further, HRR is a prognostic indicator for cardiovascular disease and all-cause mortality in healthy men (Cole et al. 1999). While changes in HR are a measure of physical stress during exercise, HRR recorded under standardised conditions may be interpreted as a practical measure of the human body's current capacity to respond to exercise stress (Borresen and Lambert 2007). This might also apply to horses. Foreman et al. (1990) and Hada et al. (2006) found in Thoroughbred racehorses that post exercise HRR was improved with training.

In order to evaluate patients with exercise intolerance and compare their exercise test results with reference values, exercise testing has to be sensitive enough to distinguish between different levels of fitness of horses. In the present study, exercise tests were performed on a treadmill to benefit from the fact that the tests could be conducted under highly standardised conditions. Only standardised data allows the differentiation between well and averagely-trained horses of the same breed and later comparison with data from clinical cases presenting with exercise intolerance. The aim of this study was to establish normative standards of performance parameters and HRR in Warmblood horses of different levels of fitness using a standardised submaximal incremental exercise test (SIET) on a treadmill. It was assumed that well trained three-day event horses would have higher values of performance parameters and faster HRR than horses trained for showjumping and dressage at amateur levels.

Materials and methods


In this prospective study, 29 Warmblood horses of 2 different levels of training depending on discipline were recruited: 11 three-day event horses (TDE; age 10 ± 3 years [mean ± s.d.]; bwt 552 ± 46 kg; withers height 1.69 ± 0.04 m) competing at international events up to CCI** level and 18 horses from the National Equestrian Centre in Berne, Switzerland (NEC; 10 ± 4 years, 586 ± 48 kg; 1.71 ± 0.05 m) competing in jumping and/or dressage events at amateur level. All horses were in regular training and judged to be sound based on a thorough clinical and orthopaedic examination. The horses were accustomed to the treadmill (Mustang 2200)1 over a period of 2 days comprising at least 4 training sessions.

Submaximal incremental exercise test

After a warm-up phase of approximately 30 min on the treadmill at walk, trot and a brief canter, the horses performed the SIET comprising 2 stages at trot (3.5 and 4.0 m/s) and 3–5 stages at canter and gallop (6.0–10.0 m/s with increments of 1 m/s) at 6% incline. The first trotting stage lasted 120 s, all following stages 90 s. The SIET was terminated when the horses reached a workload at which LAC exceeded clearly the 4 mmol/l threshold. Subsequently, the treadmill was stopped for 5 min to record the recovery HR; after that the horses were walked again to actively cool-down. In 15 NEC horses, a second treadmill SIET was performed within 1–3 months. In between the tests, the training regime of these horses was maintained at the same level.

Heart rate monitoring

Heart rate was continuously recorded at 5 s intervals using a commercially available heart rate monitor (Polar Equine CS600)2. Electrodes (Polar Equine Electrodes)2 were fixed under a girth on the right side of the chest, in the area of the withers and the sternum. Heart rate data were downloaded on a computer and analysed using the standard software (Polar Pro Trainer Equine Edition)2. The HR of each exercise stage was calculated by averaging HR from the last 30 s of each interval. Passive recovery HR was continuously determined during the first 5 min post exercise.

Blood sampling and analysis

For blood sampling an indwelling catheter (Vygon Intranule (PP) 13G)3 was placed in the jugular vein. Blood samples were withdrawn at rest before starting the exercise test and during the last seconds of each exercise stage. Blood lactate concentration was immediately determined using a hand-held lactate analyser (Lactate Pro LT-1710)4 which was evaluated previously for horses by Kobayashi (2007). Results were available within 65–75 s following withdrawal.

Data processing and statistical analysis

For each horse 6 performance parameters were calculated from the variables V (m/s), HR (beats/min) and LAC (mmol/l). The velocity-dependency of HR was approximated by linear regression HR = a1 V + a0 to determine V at a HR of 150 and 200 beats/min (V150, V200, respectively), whereby a1 is the gradient and a0 the y-intercept. The relationship between LAC and V as well as between LAC and HR were approximated by exponential regression LAC = a1 exp(a2 x) + a0 to derive discrete values at LAC of 2 and 4 mmol/l (V2, V4 and HR2, HR4, respectively), whereby a1 is the amplitude, a2 is the exponent factor and a0 the y-offset. Group mean regression functions were calculated from all measurements of all horses. Regression analyses were carried out in Microsoft Excel 20035, using a self-programmed macro.

Performance parameters were statistically analysed with SigmaStat 3.56. The level of significance was set at P<0.05 in all statistical analyses. Mean ± s.d. were computed for TDE and NEC horses separately and differences between discipline groups were tested for significance with a t test or a Mann-Whitney rank sum test dependent upon the results of the normality test (Kolmogorov-Smirnov).

To establish normative performance values, all horses were assigned to either a good performer group (GP) or an average performer group (AP) depending on whether the individual's V4 value was above or below the median V4 value of all horses, respectively. Data of performance groups were processed and compared as described before for the discipline groups. Additionally, the 95% tolerance limits (95% T.L.) were calculated from the s.d. of the respective parameters.

Repeatability between the 2 SIET of the 15 NEC horses was tested with a paired t test. From the s.d. of residuals of the one way repeated measures ANOVA, the 95% T.L. were calculated to characterise reproducibility.

For HRR analyses, horses HRs were expressed as a percentage of the maximal HR (HRpeak) reached at the end of the SIET. The relationship between HR decrease and recovery time (t) was approximated using the bi-exponential function HR (%) = afast exp(−t/τfast) + aslow exp(−t/τslow) + a0, which is a combination of 2 exponential decays with amplitude factors (afast, aslow) as well as time constants (τfast, τslow) and a basic offset a0 which equals resting HR. The differences of group mean HR between GP and AP were tested during the recovery phase for each 5 s sampling interval using a t test.


In the SIET horses reached a HRpeak of 195 ± 10 beats/min and maximal LAC of 7.3 ± 2.3 mmol/l. Three-day event horses were able to perform 4–5 stages (maximal V: 9.3 ± 0.6 m/s) at canter and gallop, while NEC horses managed 3–4 (maximal V: 8.2 ± 0.3 m/s). Regression analyses of test data from individual horses to calculate performance parameters revealed coefficients of determination (r2) between 0.90 and 1.00 (mean: 0.99).

Performance parameters

Three-day event horses had higher mean values in the performance parameters V150, V2 and V4 compared with NEC horses (Table 1). To establish normative standards at the different performance levels, more than 80% of the TDE horses and one third of the NEC horses were assigned into the GP group. With this classification all performance parameters were consequentially higher in the GP group. The quantitative description of normal ranges of both performance groups are listed in Table 1.

Table 1. Performance parameters of the standardised incremental exercise test of 29 Warmblood horses
ParameterA: discipline groupsB: performance groupsC: repeatability
Mean ± s.d.Mean ± s.d.Mean ± s.d.Mean ± s.d.Mean ± s.d.Mean ± s.d.Retest
  1. A: Three-day event (TDE) horses compared to horses of the National Equestrian Centre (NEC). B: Good (GP) compared to average performers (AP) grouped according to V4. These values represent normative standards for good and average trained Warmblood riding horses. 95% tolerance limits (95% T.L.) are given in parentheses. C: Test vs. retest of 15 NEC horses. 95% T.L. as percentage of mean of both tests are given in parentheses. The 95% T.L. for retest denote the range within which a repeated measure can be expected with a probability of 95%. V, velocity (m/s); HR, heart rate (beats/min); LAC, blood lactate concentration (mmol/l). V150, V200, V at HR 150 and 200 beats/min; V2, V4 and HR2, HR4, V and HR at 2 and 4 mmol/l LAC, respectively. *indicates significant differences between groups (P<0.05).

TDE (%)  81.8%18.2%   
NEC (%)  33.3%66.7%   
V150 (m/s)5.8 ± 0.95.0 ± 0.9*5.8 ± 0.84.8 ± 0.8*5.0 ± 0.95.0 ± 0.8± 0.5
95% T.L.  (3.7–8.0)(2.6–6.9)  (± 10%)
V200 (m/s)9.6 ± 0.88.9 ± 1.09.5 ± 0.88.7 ± 1.0*8.7 ± 1.18.9 ± 1.0± 0.6
95% T.L.  (7.5–11.6)(5.9–11.5)  (± 7%)
V2 (m/s)7.1 ± 0.95.9 ± 1.1*7.3 ± 0.65.3 ± 0.8*5.9 ± 1.05.7 ± 1.1± 0.8
95% T.L.  (5.5–8.3)(4.3–6.6)  (± 14%)
V4 (m/s)8.5 ± 0.97.1 ± 0.9*8.5 ± 0.66.7 ± 0.6*7.1 ± 0.87.1 ± 0.9± 1.0
95% T.L.  (6.8–10.6)(5.5–8.0)  (± 14%)
HR2 (beats/min)167 ± 12161 ± 14169 ± 11157 ± 13*163 ± 16160 ± 15± 9
95% T.L.  (138–191)(133–187)  (± 6%)
HR4 (beats/min)185 ± 10178 ± 12186 ± 9175 ± 12*180 ± 14177 ± 14± 8
95% T.L.  (160–218)(147–205)  (± 5%)

Mean linear regressions with HR as dependent variable showed no differences in the gradient (a1), either between TDE and NEC or between GP and AP. In contrast, the y-intercept (a0) was significantly decreased in the better trained horses (Fig 1a, Table 2). The regression analyses with LAC as dependent variable are shown in Figures 1b and c and Table 2. Curves of fitter horses appeared shifted to the right.

Figure 1.

Raw data and mean regression lines for good (GP: dots and black line) and average performers (AP: circles and grey line). The respective regression functions are given inTable 2. (a) Linear regression of heart rate (HR) vs. velocity (V), (b) Exponential regression of blood lactate concentration (LAC) vs. V and (c) LAC vs. HR. The results indicate a shift to the right, with better aerobic capacity.

Table 2. Regression equations and group mean regression coefficients
EquationA: HR vs. V (linear)B: LAC vs. V (exponential)C: LAC vs. HR (exponential)
HR = a1 V + a0LAC = a1 exp(a2 V) + a0LAC = a1 exp(a2 HR) + a0
  1. Results of 29 Warmblood horses arranged either according to disciplines (TDE and NEC) or performance (GP and AP), respectively. HR, heart rate (beats/min); V, velocity (m/s); LAC, blood lactate concentration (mmol/l). a0, a1, a2: calculated regression coefficients of the indicated regression functions in A, B, C. r2: coefficient of determination.


Test repeatability

No significant differences were found in any performance parameter between repeated tests of the 15 NEC horses. Retest revealed a considerable 95% T.L. with ± 7–10% for HR-based and ± 14% for LAC-based velocity parameters. LAC-based heart rate parameters showed the greatest reproducibility i.e. the lowest 95% T.L. with ± 5–6% (Table 1).

Heart rate recovery

At the beginning of recovery, HRpeak (= 100%) was 195 ± 9 beats/min for GP and 196 ± 11 beats/min for AP. Approximation of HRR by bi-exponential regression showed a very good fit with r2>0.99 (Table 3). With exception of the slow time constant (τslow), which was 46% longer in AP compared to GP, the coefficients of the regression equation were not different. From 120 s of recovery onwards the mean recovery HRs of GP were significantly lower from those of the AP; however, the s.d. were large (Fig 2).

Table 3. Bi-exponential regression equation of percentage heart rate recovery (HRR)
HRRHR vs. t (bi-exponential)
EquationHR (%) = afast exp(-t/ τfast) + aslow exp(-t/ τslow) + a0
  1. Regression coefficients of good (GP) and average performers (AP): The functional relationship is a combination of a fast and a slow exponential decay, each having an own amplitude factor (afast, aslow) as well as an own time constant (τfast, τslow). The maximally reached HR (HRpeak= 100%) at the end of SIET is the sum of afast, aslow and a0 (resting HR); a0 was set to 36 beats/min (corresponds to 18% of HRpeak in both groups).

Figure 2.

Percentaged post exercise heart rate recovery (HRR). Raw data (dots) and mean regression lines for good (GP: black) and average (AP: red) performers. Group mean HRs are significantly different from 120 s of recovery time onwards. Note that horses with the fastest and slowest HR decrease (thin lines) are the second-best and -worst performer, respectively, judged on the V4 value of all horses.


The aim of the present study was to establish reference values in Warmblood horses of good and average performance level. Therefore, horses with different training regimes depending on their competing discipline were recruited. Three-day event horses were expected to have higher aerobic capacity than NEC horses because of the larger amount of canter and gallop training required to compete successfully and safely in a cross country event. Although TDE horses had higher values than NEC horses (significant for V150, V2 and V4), the results of both discipline groups showed a wide distribution with overlapping s.d. in each of the performance parameters. Obviously, certain horses in each discipline groups performed either below or above expectation and the objectives were only partly achieved. This may be explained by differences in the individual training protocols (the details of which were not well known), by the horses' form on the day or simply by different levels of athletic ability or aptitude. Furthermore, the individual ability of the horse to cope with the testing procedure has to be taken into consideration.

In order to obtain representative normal values of Warmblood horses of different physical conditions, all horses of both discipline groups were ranked based on their V4 results and reallocated to 2 performance groups. The parameter V4 was selected, because it has been described as a good indicator of fitness (Persson 1983). Consequently, the spread around the means decreased in both performance groups and GP had significantly higher mean values in all performance parameters compared to AP. Nine of 11 TDE horses were assigned to the group of GP; one of the remaining 2 horses was the best performer in the AP group and the other was an inexperienced horse with limited ability according to the judgement of its trainer. Six NEC horses performed above expectation and were ‘upgraded’ to the GP group; however, 5 of them were found at the bottom of the table.

A submaximal exercise test was chosen because some of the horses involved in the study were young and testing took place in between competitions. Therefore, horses were not run to exhaustion and the test was terminated when LAC exceeded the threshold of 4 mmol/l. A disadvantage compared to maximal exercise testing might be the need to relate some submaximal values to maximal values (Rose and Christley 1995). In the present study, this concerned the parameter V200, which sometimes required extrapolation from lower values. This appeared to be unproblematic because of the linear relationship between heart rate and workload up to a certain level (Ehrlein et al. 1973) and the high r2 of the individual linear regressions observed.

The lower y-intercept (a0) of GP compared to AP, together with an equal gradient (a1) in both groups implied that the linear regression line of HR vs. V was shifted towards higher velocities; i.e. GP had lower HR for the same velocity than AP. This observation is consistent with reported changes of the HR-V relationship in response to training (Kobayashi et al. 1999; Vermeulen and Evans 2006). Between untrained Andalusian and Anglo-Arabian horses Munoz et al. (1999) described significant interbreed differences in both regression coefficients. In the 2 Warmblood groups of the present study, no differences were found between the mean gradients; however, in contrast to Munoz et al. (1999) a considerable intrabreed variation of a1 (range: 9.5–15.4) was observed within the groups indicating no breed specificity. A dependency of the gradients on performance levels was not observed.

Cikrytova et al. (1991) reported on breed-specific differences in V170. Thoroughbreds had significantly higher mean V170 than Warmbloods, but no differences were observed between Warmbloods of different origins and no correlation was found between V170 and the cross country performance in 339 Warmbloods with similar training level. This confirms that a horse's performance in competition is not only dependent on fitness but also on other parameters such as the general state of health, inherited traits, surface conditions and ‘will to win’.

The exponential approximation of V vs. LAC, as well as of HR vs. LAC revealed clear differences between AP and GP (Figs 1b,c). Again, the right shift of the GP curve indicated higher aerobic capacity (Figs 1b,c).

In the context of training monitoring and control, results of exercise testing are often transferred into recommendations for daily training. However, it has to be considered that there are differences in metabolic response to track and treadmill exercise (Sloet van Oldruitenborgh-Oosterbaan and Clayton 1999). Furthermore, publications in human literature described sport specific differences and the need for performance tests to be as specific as possible for a certain discipline (Coen et al. 2003; Roecker et al. 2003). The transfer of results from treadmill testing to recommendations for actual training is not sufficiently established in horses.

Horses' recovery HRs have been described as decreasing in a bi-exponential manner, with a faster initial and a slower secondary decay (Rugh et al. 1992). This would support the theory of a coordinated interaction of parasympathetic reactivation and sympathetic withdrawal during exercise recovery. Hada et al. (2006) demonstrated that post exercise HRR is affected by the activity of the autonomic nervous system (ANS). The question, to which extent the 2 fractions of the ANS would interact and influence HRR following exercise, could not be answered. The exact regulation mechanism still remains unclear in both horses and man. Results of Pierpont and Voth (2004) indicate that in man, parasympathetic reactivation has a faster response time than the sympathetic withdrawal, which would agree with the faster pharmacokinetics of Acetylcholine. In contrast, Savin et al. (1982) reported that in man sympathetic withdrawal affected HR mainly initially, while parasympathetic activation was predominant later in recovery.

In the present study, differences in HRR between groups were observed. The longer slow time constant in AP vs. GP resulted in significant differences in recovery HR after 120 s. By this time, HR had dropped to 50% of the maximal post exercise HR. The substantial variability within each group may be at least partly an expression of the different performance levels of the respective horses, as the ones with fastest and slowest HR decrease were also the ones with the most extreme V4 values (Fig 2). Therefore, the measurement of HRR may be a very helpful parameter for making a preliminary assessment of exercise intolerant horses, as the method is easy to perform.

In conclusion, a treadmill SIET is suitable to objectify aerobic capacity in Warmblood riding horses competing at amateur and higher levels. The reference values of the different performance parameters are helpful to judge patients with exercise intolerance but also sound athletes for estimation of their aerobic capacity. Differences in HRR of healthy sports horses in this study provide an indication of more aerobic fitness in GP than AP. Heart rate recovery may, therefore, be a useful qualifier of fitness level, ability and soundness. Further research is required to investigate the differences expected in sports horses with exercise intolerance of defined aetiologies.


The study has been supported by the Foundation ‘Forschung für das Pferd’ (Zurich, Switzerland). We would also like to thank the National Equestrian Centre (Berne, Switzerland) and the owners of the 3-day event horses, who kindly left their horses at the study's disposal, Katja von Peinen, Carole Braun, Selma Latif and the staff of the Equine Hospital of the University of Zurich, who assisted in carrying out the exercise tests, Nina Waldern and Isabel Imboden for their valuable advice with the manuscript.

Conflicts of interest

The authors declare no potential conflicts.

Manufacturers' addresses

1 Graber AG, Fahrwangen, Switzerland.

2 Polar Electro Oy, Kempele, Finland.

3 Laboratoires Pharmaceutiques Vygon, Ecouen, France.

4 Arkray Inc., Kyoto, Japan.

5 Microsoft, Redmond, Washington, USA.

6 Systat Software Inc., Cranes Software Int. Ltd, Bangalore, India.