Effects of short-term training on thermoregulatory and sweat responses during exercise in hot conditions



Reasons for performing study: There is evidence that extensive training in cool conditions results in improvements to heat dissipation that contribute to successful acclimatisation. In horses, the effects of a less extensive training regimen have not been determined.

Objective: This study investigated whether 10 consecutive days of moderate intensity treadmill training in cool conditions would improve thermoregulatory and sweating responses of horses to exercise in the heat.

Methods: Six unfit Thoroughbred horses completed a standardised treadmill exercise test (SET) in hot, dry conditions (32–34°C, 45–55% RH) before (SET1) and after (SET2) 10 consecutive days of running at 55% VO2max for 60 min in cool conditions (19–21°C, 45–55% RH). Each SET consisted of a 5 min warm-up and cool down at a walk, 40 min of trotting (50% VO2max), 7 min at 75% VO2max and a 30 min standing recovery. Bodyweight was determined pre- and post SET. Heart rate, rectal, skin, pulmonary artery and muscle temperatures were measured throughout the SETs and sweating rate (SR) and sweat ion losses determined for each 5 min interval.

Results: Following training, mean VO2max increased by 8.9% (P<0.05). In SET2, PCV was lower during the last 30 min of exercise and end-exercise rectal, muscle and pulmonary artery temperatures were decreased by 1.5 ± 0.2, 0.8 ± 0.1 and 1.0 ± 0.2°C, respectively (P<0.05). Peak SR and the pattern of sweat ion losses during exercise was unchanged post training whereas SR and sweat losses during recovery were decreased (P<0.05).

Conclusions: Similar SRs for a given core temperature during exercise but a more rapid decrease in recovery resulted in an overall reduction in sweat fluid losses with no change in sweat ion losses after training.

Relevance: The results provide insight into the extent to which short-term training can improve the capacity of horses to exercise in hot conditions.


Athletic performance can be impaired in hot environmental conditions due to hyperthermia and dehydration. Appropriate preparation, including acclimatisation can reduce the effects of such stress factors. The equine athlete, like its human counterpart, is at risk when exercising in hot conditions, with greater limitations imposed by the more extensive proportion of its body mass used in locomotion and higher metabolic rate at a given workload (Lindinger 1999). Despite having higher sweating rate (SR) than well-trained human athletes, a lower surface area-to-body-mass ratio (m2/kg bwt) and high rate of metabolic heat production in horses limits the capacity to maintain core body temperature within a range considered safe for continued exercise. Previous studies have determined that the full benefits of heat acclimation may require 14–21 days of exercise in hot conditions in well-conditioned equine athletes (McCutcheon et al. 1995a, 1999a). Therefore, the untrained horse is at increased risk of hyperthermia at a given workload and this risk is exacerbated in hot ambient conditions. However, it has been argued that exercise training, even in a cool environment, can confer improvements in acclimatisation state. In human subjects, reductions in end-exercise core temperature have been demonstrated following physical training in a cool environment (Armstrong and Maresh 1991). When horses completed 8 weeks of exercise training in cool conditions, altered sweating responses contributed to improved heat dissipation during exercise and lower end-exercise core temperature (McCutcheon and Geor 2000). Few studies have investigated whether improvements to heat dissipation during exercise in a hot environment can be attained through a shorter period of moderate intensity training in cool conditions. The overall aim of this study was to determine whether 10 consecutive days of moderate intensity training in cool conditions would improve thermoregulatory and sweating responses to exercise in the heat. Specifically, we examined the effects of 10 consecutive days of moderate intensity training on: 1) measurements of body temperature and of SR during and following exercise in hot conditions; and 2) the concentration of sweat sodium ([Na+]), potassium ([K+]) and chloride ([Cl-]) measured in sweat obtained during and following exercise in the heat. We hypothesised that 10 days of moderate intensity exercise training would result in: 1) a reduction in end-exercise core body temperature; 2) an increase in SR during exercise and 3) a reduction in the sweat [Na+], and [Cl-] and an increase in sweat [K+] during a standardised exercise test (SET) in hot, dry conditions.

Materials and methods

The care and use of animals followed the Guide to the Care and Use of Experimental Animals (Canadian Council on Animal Care, Ottawa, Canada). All animal experiments were conducted after approval by the Animal Care Committee of the University of Guelph and were performed in compliance with their guidelines and recommendations.

Six unfit Thoroughbred horses (4 mares and 2 geldings, aged 4–8 years, mean ± s.e. weight 464 ± 8.7 kg) were used. All horses were paddock-rested for at least 8 months prior to the start of the study, but had previous treadmill exercise experience. Throughout the study period, horses were housed indoors in box stalls maintained at 18–20°C, during which time the diet for all horses consisted of meadow hay and approximately 5 kg/day of a mixed grain ration1. Water and a trace mineral/salt block were available ad libitum.

Experimental design

All treadmill exercise was completed in an environmentally controlled room. In the week prior to the study, horses completed an incremental exercise test in cool conditions for determination of VO2max. In preparation for this test, each horse received 3–5 short (2–3 min/day) daily sessions of slow trotting while wearing a respiratory gas collection mask. Horses completed a SET in hot conditions one day prior to (SET1), and one day following (SET2) 10 consecutive days of exercise training in cool conditions. All exercise tests and training were performed with the treadmill (Sato)2 set at a 6° incline.

Training regimen

For each horse, the results of the incremental exercise test were used assess an oxygen consumption (VO2) vs. speed regression equation over the linear portion of the data. From these equations, a speed equivalent to 55% of the pretraining VO2max was determined. For each day of training horses completed a 5 min walking warm-up (1.6 m/s) followed by 60 min of exercise at the calculated running speed (range 4.3–4.8 m/s) on each of 10 consecutive days. For the first 3 or 4 days of training, horses were allowed a 5 min rest stop after 20 and 40 min of exercise. Thereafter, all horses completed the training session without rest. The total distance covered was 13–14 km per day. During the training sessions, ambient temperature ranged from 19–21°C and relative humidity (RH) from 45–55%.

Exercise tests

Incremental exercise test: Prior to the each exercise test, horses were weighed (± 0.5 kg; KSL Animal Scale)3. The incremental exercise test, completed under cool conditions (18–20°C, 45–55% RH) consisted of running for 5 min at 4 m/s, 90 s at 6 and 8 m/s, and 1 min at 9 m/s. The speed was then increased by 1 m/s every 60 s until fatigue, evinced by inability to maintain treadmill speed. An open circuit calorimeter was used for collection of expired gases throughout the exercise protocol as described previously (Geor and McCutcheon 1998). Flow through the system was approximately 8000 l/min. Expired oxygen and carbon dioxide fractions were measured during the last 15 s of each speed increment. VO2 was calculated from standard equations and VO2max was assumed to have been attained when an increase in running speed resulted in little or no change in VO2. The gas analysers were calibrated daily with precision gases. The nitrogen dilution technique was used to verify the overall accuracy of the system (Fedak et al. 1981). The coefficient of variation for repeat determinations of VO2 with this system is approximately 3%.

Standardised exercise test: The SETs were completed after an overnight fast before and after the training period with bodyweight determined pre- and post SET. A thermohygrometer (Model 3309-60, Cole-Palmer)4 was used to monitor ambient conditions with the room maintained at 32–34°C and 45–55% RH throughout each SET. The test consisted of a 5 min warm-up at walk (1.6 m/s), followed by 20 min of trotting at 50% VO2max, 7 min canter at 75% VO2max, a further 20 min of trotting at 50% VO2max, 5 min of walking (1.6 m/s) and a 30 min standing recovery. A fan, mounted above and 0.5 m in front of the treadmill was used to maintain an air velocity of 3.5–4 m/s over the anterior and dorsal aspects of the horse. Heart rate (HR) (Equistat Model HR-8A)5, rectal, skin, pulmonary artery and muscle temperatures were measured throughout each SET, and SR and sweat ion losses were determined for each 5 min interval during exercise and recovery. Blood samples were collected prior to each SET for measurement of plasma volume as previously described (McCutcheon et al. 1999b), and were also collected at 5 min intervals during the SET for measurement of packed cell volume (PCV). Briefly, plasma volume at rest was determined after spectrophotometric measurement of the change in plasma absorbance at 620 nm of Evans blue dye. The calculation is based on dilution of a known quantity of Evans blue dye (Lindinger et al. 1995) administered i.v. and because the distribution volume changes with time, the analysis was based on at least 3 plasma samples obtained 10 min apart. After correction for absorbance changes at 740 nm due to interfering substances, absorbances were fitted to a linear regression against time and the initial plasma volume calculated by extrapolation to time 0 (Foldager and Blomqvist 1991).

Measurement of body temperature

A thermocouple (T-260)6 for measurement of pulmonary artery temperature (Tpa) was introduced into the pulmonary artery via a catheter (PE-240)7 placed in the left jugular vein after aseptic preparation and local anaesthesia of the overlying skin. Rectal temperature (Tre) was measured by inserting a thermocouple (T-180)6 20–30 cm proximal to the anal sphincter. Skin temperature (Tsk) was measured on the lateral thorax using a thermocouple (SST-20)5 affixed to the skin. Middle gluteal muscle temperature (Tmu) was measured at rest, within 30 s of the end of exercise and at the end of the recovery period in each SET by inserting a needle thermocouple (MT-23)6 coupled to a digital thermometer (BAT-10)6. Thermocouples had response times of ∼1°C/s when calibrated in a heated water bath with a NIST-traceable thermometer (± 0.1°C)8.

Measurement of sweating responses and sweat ion concentrations

Local SR was based on measurement of volume of sweat samples obtained from an area of the left lateral thorax by use of a direct sweat collection method, as described elsewhere (McCutcheon et al. 1995b). In brief, a sealed polyethylene pouch enclosing a 150 cm2 area of skin was attached to an area of shaved skin with a dermal adhesive. The edges of the pouch were further sealed with dermal tape that covered the pouch/skin margin. A ventral reservoir, formed by a deep fold in the polyethylene, separated accumulated sweat from the skin surface and facilitated the removal of collected sweat through polyethylene tubing (1.67 mm internal diameter)7 incorporated into the lateral margin of the pouch. Sweat samples were collected every 5 min throughout each phase of exercise and recovery. For the 2 SETs, placement of the pouch was alternated between left and right thorax.

Local SR, expressed as ml/m2/min, was calculated on the basis of the volume of sweat collected from the measured skin area within the pouch at the end of the 5 min interval. Therefore, the measured SR represents the average rate of sweat production over a 5 min period. Extrapolation of the local SR at each time point during exercise and recovery to the horse's total body surface area was used to calculate a mean whole body SR. Total body surface area (SA) was calculated by using the formula (Hodgson et al. 1993):


Peak SR was used to describe the highest rate of sweat production attained during the pre- and post training SET. Total sweat fluid loss for the duration of exercise and recovery was estimated from the total body water losses after correction for faecal and estimated respiratory water losses; this was assumed to represent a constant percentage (∼15%) of the overall water losses (Hodgson et al. 1993). No horse voided urine during the SETs.

The effect of training on the SR-body temperature relationship was evaluated by determining the slope (expressed as ml/m2/min1/°C) and the x intercept of the regression line representing the mean SR and temperature (Tpa or Tre) for each 5 min interval. The x intercept (the temperature at which the regression line of the sweating vs. temperature relationship extends to the zero value of the x axis) was used as an estimate of the sweating threshold and the slope was used as an estimate of sweating sensitivity (Roberts et al. 1977).

Concentration of sodium [Na+], potassium [K+] and chloride [Cl-] in sweat were determined with an ion-selective analyser (Nova Statprofile 9)9. For each SET, linear regression was used to examine the relationship between SR and sweat [Na+] for every 5 min interval during exercise.

Statistical analysis

Data were analysed by repeated measures ANOVA to compare measures for the 2 SETs with a statistical analysis program10. When a significant F ratio was obtained, the Bonferrroni post hoc test was used to test for differences among means. The slope and threshold of each subject's SR vs. Tpa and Tre relationship were determined by least squares linear regression and a one-way repeated measures ANOVA was used to determine whether differences existed in slope or intercept data. Data are reported as means ± s.e. Unless otherwise stated, significance was accepted at P<0.05.


Effects of training

Following training, mean VO2max increased by 8.9% (P = 0.004) (pretraining: 142 ± 4 ml/kg bwt/min vs. post training 155 ± 4 ml/kg bwt/min). Pre-exercise bodyweight was not different following training whereas the total change in bodyweight during exercise and recovery in the SET was lower (SET1 = 15.1 ± 0.8 vs. SET2 = 12.3 ± 0.7 kg, P<0.05).

Heart rate during exercise was not different between SETs but was significantly lower in the first 15 min of recovery following training. In each SET, HR was still significantly higher than pre-exercise values (P<0.05) after 30 min of recovery. No difference in resting PCV was detected following training (36 ± 1%). In SET2, PCV was significantly lower after the first 20 min of trotting when compared to SET1 and was not different from pre-exercise values at the end of 30 min of recovery. In contrast, PCV remained significantly elevated at the end of recovery in SET1 (End SET1 50 ± 4%; End SET2 37 ± 1%). Training resulted in a 13.8% increase in resting plasma volume (SET1: 20.9 ± 0.8 l; SET2: 23.8 ± 0.9 l; P = 0.03).

Pulmonary artery and rectal temperature measurements were not different at rest or early exercise in the SETs (Fig 1a,b). The SET resulted in a mean rise in Tpa and Tre of 5.1 ± 0.2°C and 4.3 ± 0.3°C, respectively, in SET1. In SET2, temperatures were lower than SET 1 during the last 15 min of higher intensity exercise such that end-exercise Tpa and Tre were lower by 1.0 ± 0.2°C and 1.5 ± 0.2°C (P<0.05), respectively. In SET2, Tpa and Tre following 30 min of recovery was not different from SET1. Skin temperature measured throughout exercise and recovery was not significantly different between SETs (Fig 1c). Tmu at rest was the same in each SET (37.1 ± 0.2°C). In SET1, Tmu increased to 42.8 ± 0.2°C after exercise and had declined to 40.2 ± 0.3°C at the end of 30 min of recovery. In SET2 Tmu was significantly lower at the end of exercise and of recovery (41.8 ± 0.3 and 39.2 ± 0.2, respectively; P<0.05).

Figure 1.

Mean ± s.e. temperature measured in the pulmonary artery (T pa ; a), rectum (T re ; b) and on the skin of lateral thorax (T sk ; c) in 6 horses during standardised exercise tests in hot, dry conditions (32–34°C and 45–55% RH) completed prior to (SET1) and following (SET2) 10 consecutive days of exercise training in cool, dry conditions (18–20°C and 45–55% RH). *Significantly different from SET1, P<0.05.

In each SET, SR increased continuously during the first 20 min of trotting and canter exercise, after which it was unchanged until declining at the end of exercise. Sweat losses, measured at 5 min intervals, were higher during the first 15 min of recovery in SET1 (Fig 2). Peak SR and the pattern of sweat ion losses during exercise were unchanged post training whereas SR and sweat losses during recovery were decreased in SET2 (P<0.05). Mean sweat [Na+] was decreased and mean sweat [K+] increased (P<0.05) throughout exercise in SET2 (Fig 3a,b). Sweat [Cl-] was significantly (P<0.05) lower during the latter half of exercise in SET2 but was not different from pretraining values during recovery (Fig 3c). Individual exercise-associated sweating threshold and sweating sensitivity values for SET1 and SET2 based on Tpa and Tre are presented in Tables 1 and 2, respectively. No change in sweating threshold was evident following 10 days of training whereas an increase in sweating sensitivity was detected post training (P<0.05). Individual correlations between SR and Tpa and Tre during exercise ranged from r = 0.568–0.953 with a mean value of r = 0.788 for Tpa and from r = 0.564–0.944 with a mean value of r = 0.794 for Tre.

Figure 2.

Mean ± s.e. sweating rate for 6 horses averaged for 5 min intervals during standardised exercise tests in hot, dry conditions (32–34°C and 45–55% RH) completed prior to (SET1) and following (SET2) 10 consecutive days of exercise training in cool, dry conditions (18–20°C and 45–55% RH). * Significantly different from SET1, P<0.05.

Figure 3.

Mean ± s.e. concentrations of sodium ([Na + ], a), potassium ([K + ], b) and chloride ([Cl - ], c), in sweat collected every 5 min from horses (n = 6) during standardised exercise tests completed in hot, dry conditions (32–34°C, 45–55% RH) prior to (SET1) and following (SET2) 10 consecutive days of exercise training in cool, dry conditions (18–20°C and 45–55% RH). In (a) and (b), all values in SET2 are significantly different from SET1. In (c), * indicates significantly different from SET1, P<0.05.

Table 1. Mean ± s.e. slope and intercept of the sweating rate-temperature relationship for pulmonary artery temperature during exercise in hot, dry conditions (32–34°C, 45–55% RH) in an SET completed before (SET1) and after (SET2) 10 days of training in cool, dry conditions
Subject No.SET1SET2
  1. n = 6. Slope = ml/m2/min1/°C; intercept = estimated value of pulmonary artery temperature (°C) at sweating rate = 0. *Significantly different from SET1, P<0.05.

1 7.7 ± ± 2.436.5
2 13.3 ± 2.537.216.9 ± 2.536.5
3 5.9 ± 3.236.813.5 ± 1.137.3
4 8.4 ± 2.936.616.9 ± 3.936.6
5 6.9 ± 2.337.68.3 ± 1.336.7
6 8.7 ± ± 3.136.0
Mean ± s.e.8.5 ± 1.137 ± 0.212.6 ± 2.0*36.6 ± 0.2
Table 2. Mean ± s.e. slope and intercept of the sweating rate-temperature relationship for rectal temperature during exercise in hot, dry conditions (32–34°C, 45–55% RH) in an SET completed before (SET1) and after (SET2) 10 days of training in cool, dry conditions
Subject No.SET1SET2
  1. Values are means ± s.e.; n = 6. Slope = ml/m2/min1/°C; intercept = estimated value of rectal temperature (°C) at sweating rate = 0. *Significantly different from SET1, P<0.05.

1 9.7 ± 1.936.511.6 ± 2.936.1
2 9.0 ± 3.636.615.7 ± 3.233.6
3 10.8 ± 2.137.413.6 ± 1.336.9
4 8.9 ± 2.937.610.8 ± 3.437.4
5 9.8 ± 3.336.213.2 ± 1.536.6
6 10.7 ± 2.736.114.5 ± 2.736.5
Mean ± s.e.9.9 ± 1.436.7 ± 0.313.2 ± 2.0*36.2 ± 0.6


The main finding of this study was that 10 consecutive days of moderate intensity training (50–55% VO2max) in cool conditions (19–21°C, 45–55% RH) resulted in a significant improvement in heat dissipation during a standardised submaximal exercise test in hot conditions. Lower pulmonary, rectal and muscle temperatures during the latter portion of exercise were associated with a more rapid return to pre-exercise core temperature. Similar SRs for a given core temperature during exercise, but a more rapid decrease in SR in recovery, resulted in an overall reduction in sweat fluid losses after training. Additionally, the reduction noted in sweat [Na+] at a similar SR during exercise and lower SR during recovery is consistent with an early adaptation in the sweat gland tubule towards lowering Na+ loss to assist in protection of plasma volume (Vrijens and Rehrer 1999). However, no change in the sweating threshold was detected and the increase in sweating sensitivity noted post training was substantially less than reported previously in trained horses following heat acclimation (Marlin et al. 1999; McCutcheon et al. 1999a).

Previous studies of thermoregulatory responses have utilised physically-conditioned horses subjected to additional training in hot conditions or have included a more prolonged period of moderate intensity training in cool conditions (8 weeks or more). By comparison, the current study demonstrated that as little as 10 days of training in cool conditions resulted in a significant increase in plasma volume and improved vascular volume regulation manifest as lower core temperature during submaximal exercise in the heat (32–34°C, 45–55% RH). McKeever et al. (1987) also demonstrated very early increases in plasma volume with relatively low-level training. These investigators noted no increase in daily water intake and a 24.5% decrease in 24 h urine output. Their findings supported an increase in water reabsorption via renal tubular reabsorption of urea and osmotically active substances other than sodium, emphasising the importance of renal control mechanisms in this initial increase in plasma volume.

Although Tpa, Tre and Tmu measured in the last 30 min of exercise in SET2 were decreased, Tsk, measured on the lateral thorax, was similar in the extent and pattern of rise. Training-associated changes in skin blood flow, subject to thermoregulatory and nonthermoregulatory (baroreceptors) controls will reflect competing demands for cooling and for maintenance of blood pressure during hyperthermia (Johnson 1998). With habitual exercise, there is enhanced thermoregulatory control of skin blood flow leading to lower threshold temperature for increased skin blood flow and greater cutaneous circulation at a given core temperature (Tankersley et al. 1991; Pergola et al. 1996; Ho et al. 1997). In the current study, similar Tsk despite lower core temperature during exercise in the heat suggests improved control of cutaneous circulation allowing increased skin blood flow at a given core temperature. Similarly, Tsk in recovery during SET2 was similar to that measured in SET1 despite a lower SR also suggesting improved thermoregulatory control of skin blood flow.

Several studies in human subjects have noted a reduction in sweat [Na+] during exercise following heat acclimation (Allan and Wilson 1971; Smiles and Robinson 1971; Nielsen et al. 1997). However, these studies have also demonstrated significant (14–26%) increases in SR. With increasing SR, sweat [Na+] appears to increase linearly, its retention reportedly limited by the capacity of the sweat duct for Na+ ion reabsorption (Inoue et al. 1998; Shamsuddin et al. 2005). More recently, a study by Buono et al. (2008) provided evidence that in human subjects, the rate of Na+ secretion increases more rapidly than the rate of Na+ reabsorption in the distal sweat duct, resulting in a more extensive Na+ ion loss in sweat with increasing SR. Few studies have determined sweat [Na+] in relationship to SR when assessing improvements in heat dissipation during exercise. Following training in the current study, SR was not significantly different during exercise, potentially reflecting a lower core body temperature. Despite a similar SR throughout exercise, sweat [Na+] was decreased by 12–20 mmol/l in the post training SET. The findings of a reduction in sweat [Na+] at a given SR are consistent with recent findings of increased ion reabsorption for a given SR following heat acclimation (Buono et al. 2007). The precise mechanism that allows for improved Na+ ion reabsorption following heat acclimation have not been determined although it is likely that enhancement of the number or capacity of Na+-K+-ATPase and sodium ion channels in the reabsorptive duct can be induced by aldosterone (Sato and Dobson 1970).

Recently in the human literature, measurement of sweat ion concentrations has generated renewed debate concerning the relative merits of various collection methodologies (Armstrong 2008; Weschler 2008; Baker et al. 2009). In question is whether an occlusive method of sweat collection, such as a capsule, arm bag or an absorbent pad significantly alters sweat composition (due to prolonged interaction with the stratum corneum) when compared to whole body washdown. Fluid remaining on the skin could leach electrolytes from the stratum corneum, resulting in artifactual elevations of sweat ion concentrations and detection of urocanic acid (a constituent of the stratum corneum) in sweat samples has been used to support this theory. In the present study, an area of skin was enclosed by a loose plastic covering, with sweat collecting by gravity in a ventral pouch that was not in contact with the skin surface. Although electrolyte leaching could occur with this collection technique, the relatively similar SR and Tsk in the 2 SETs allows viable comparison of sweat ion losses. Increases in sweat [K+] but decreases in sweat [Na+] in SET2 argues against an notable degree of electrolyte leaching associated with the collection technique utilised for these horses.

Studies in human subjects have demonstrated that at least 1 h of exercise at rates exceeding 50% VO2max can confer a degree of heat acclimation (Piwonka et al. 1963; Armstrong and Maresh 1991; Pandolf 1998). The current findings are consistent with previous evidence that heat production during moderately intense submaximal exercise, combined with constraints on heat loss inherent in equine athletes, contributes to training-associated enhancement of heat loss mechanisms. Whereas more efficient sweating coupled with improvements in the control of skin blood flow may delay the rise in body temperature and sweat losses associated with exercise, the fluid losses and accompanying dehydration associated with more prolonged submaximal exercise can ultimately limit performance.


The authors thank Hua Shen for her excellent technical assistance. The research was supported by the Equine Research Program of the Ontario Ministry of Agriculture, Food, and Rural Affairs.

Conflicts of interests

The authors have declared no potential conflicts.

Manufacturers' addresses

1 Purina Mills, Mississauga, Ontario, Canada.

2 Sato, Uppsala, Sweden.

3 KSL Scales, Kitchener, Ontario, Canada.

4 Cole-Palmer Instruments, Chicago, Illinois, USA.

5 EQB Inc., Unionville, Pennsylvania, USA.

6 Physitemp Instruments Inc., Clifton, New Jersey, USA.

7 Becton Dickinson, Parsippany, New Jersey, USA.

8 Fisher Scientific, Ottawa, Ontario, Canada.

9 Nova Biomedical Canada Ltd, Mississauga, Ontario, Canada.

10 SAS Institute, Cary, North Carolina, USA.