Coffee husk consumption by the horses was spontaneous and caused temporary intoxication, thus confirming the suspicions of the owners and veterinarians, who reported similar conditions in horses that ingested coffee husks used as bedding in their stalls. The signs observed were similar to those described in animals that were spontaneously intoxicated.
Coffee husk intake was responsible for the clinical intoxication observed in horses, and the improvement of signs was associated with discontinuation of the coffee husk supply. According to Barcelos et al. , the caffeine concentration in the husks of Arabica coffee varies between 0.5% and 1.3%, thus supporting the amount of caffeine found in the coffee husks used in this experiment. Caffeine was the substance responsible for the clinical signs observed, given that the mycotoxin analysis and the absence of insecticides in coffee husks excluded other conditions due to intoxication that exhibit neurological signs in horses, such as leukoencephalomalacia , and those that include signs of hyperexcitability, such as intoxication by chlorinated insecticides . Furthermore, cattle have been shown to display signs of intoxication after one week of ingesting 3 kg, on average, of coffee husks used as bedding in their stalls . Nevertheless, the addition of up to 1 kg of coffee husks with 0.97% caffeine in the diet of cattle did not produce clinical signs of intoxication [3, 6].
The animals were not very interested in consuming coffee husks during the first hours after they were supplied; however, after ingesting the husks for the first time, the animals generally preferred them to hay. This behavior was also described by Nazário et al. . It is worth noting that in the present experiment, the animals were fully adapted, and the supply of hay, in its own trough, was not discontinued; furthermore, hay consumption by the animals was within the levels recommended by the National Research Council (NRC)  for the species.
The ELISA method, which was used to quantify plasma caffeine levels, has been compared with the gas chromatography method in humans  and in horses , and it has been proven effective in quantifying the substance in both species. ELISA has the advantage of being cheap and practical; however, if the individual ingests supplements that contain other methylxanthines, cross-reactions may occur and the concentration of caffeine may be overestimated . The ELISA kit used in this study had cross-reactivity of 24% with theobromine and 0.06% with theophylline. In the initial part of this experiment, animals were only given coast cross hay, which does not contain any substances belonging to the methylxanthine group, which can be demonstrated by the low concentration of caffeine detected in the plasma and urine at T0, thus excluding the likelihood of this interference. The high plasma caffeine levels detected at T56 was not influenced by cross-reactivity with theobromine and theophylline, since they were not detected in coffee husk by HPLC. It was previously described that caffeine is metabolized to many methylxanthine compounds, including theobromine and theophylline . These substances may influence urine caffeine levels found in this present study at T56, however the high magnitude concentration increase (compared to T0) was mainly due to the caffeine present in the ingested coffee husk.
The bioavailability of caffeine in horses after oral administration was 39% , and as the total average caffeine consumption by the animals in the present study was 78 mg per kilogram of body weight (mg/kg BW), therefore, on average, approximately 30 mg/kg BW of caffeine were absorbed by the animals over 56 h. Behavioral changes occur in horses when plasma caffeine concentrations are greater than 2,000 ng/ml . Similar results were also observed by Vickroy et al. , who reported an increase in motor activity in horses with plasma caffeine concentrations of 4,000 ng/ml. Therefore, the clinical signs observed in the horses participating in this study (compulsive walking, mydriasis, congested ocular mucosa and episcleral vessels and intense sweating) may reflect the high plasma concentrations of caffeine (51,564 ± 5,708 ng/ml) observed at T56. In humans, clinical signs of caffeine intoxication manifested after an intake of over 10 mg/kg BW , and a dose of 15 mg/kg BW of caffeine caused severe changes in the central nervous system (anxiety, delirium, vomiting and seizures) and circulatory system .
The clinical signs of intoxication ceased between 12 h (animals 3, 4 and 6) and 40 h (animals 1, 2 and 5) after access to coffee husks was discontinued, which is similar to the results found in cattle after spontaneous intake of coffee husks . These authors reported that most animals showed rapid and complete remission of clinical signs between 3 and 4 h, whereas in other animals, signs only ceased 24 h after restricting access to coffee husks. Similar signs to the involuntary movements (dyskinesia) of the mouth and tongue shown by animals 1, 2, 3 and 5 were also described in humans after excessive caffeine intake and has been called "bucco-linguo-masticatory syndrome" . The heart and respiratory rates and the rectal temperatures of animals increased after 36 h of supplying them with coffee husks, and unlike the neurological signs, these measurements only returned to normal values 64 h after discontinuing the coffee husk supply.
In our study, there were no residual sequelae or deaths; however, there are reports from professionals in the field that some intoxicated animals die when coffee husk intake is not discontinued. Nevertheless, it is emphasized that these reports do not constitute scientific proof; furthermore, the lethal dose of caffeine for horses was not found in the literature. In humans  and dogs , doses of 75 mg/kg BW and 140 mg/kg BW of caffeine, respectively, are considered lethal.
There was an increase in serum total protein, albumin and globulin at T56, which occurred as a result of the 5% dehydration that was clinically detected in four animals. This dehydration could be attributed to the excessive sweating observed at T56 and the lack of water consumption by horses exhibiting clinical signs. Hyperglobulinemia can also be observed during inflammation, but the serum iron concentration, an early indicator of inflammation in horses , was not affected and remained within normal parameters for horses.
Serum CK activity increased at T56 relative to T0, and in five animals, they were greater than those considered to be normal for the species (100 to 300 IU/l) . The activity of this enzyme increases rapidly after muscle damage . In the present study, this increase could have resulted from muscle tremors and increased motor activity, because the highest CK activities were observed in animals that exhibited this sign more intensely. The average serum concentration of AST, used to evaluate muscle and liver functions, was higher at T56 than at T0 (p = 0.03); however, it was still within the normal range for the species . These values could have been higher and exceeded this threshold, given that the serum concentration of this enzyme, unlike CK, increases gradually and only reaches its peak 24 to 36 h after muscle damage . Although AST concentrations also increase after liver damage, the average increase in AST concentration observed in the animals from this study suggested muscle damage, because the serum activity of GGT, another enzyme that assesses liver function, did not differ between T0 and T56 (p > 0.05). Furthermore, it is known that coffee intoxication can cause muscle damage; cases of rhabdomyolysis have been reported in humans as a result of caffeine overdose .
The serum urea concentration decreased at T56, but there was no significant difference between the two time periods. In contrast, serum creatinine significantly increased (p < 0.001) at T56 relative to T0. Creatinine is derived from muscle creatine, and intense muscle activity can explain the increased creatinine in this study .
The blood gas values at T56 revealed metabolic acidosis in animals 1, 2, 3 and 5, which had lower HCO3 values than the standard values for the species. However, caffeine stimulation of respiratory centers could have caused bronchodilation  and increased oxygen uptake , resulting in hyperpnea (an increase in respiratory rate and amplitude), which was observed between T48 and T60. This condition caused an increase in PO2 and SO2, as well as a reduction in pCO2 and tCO2, thus counteracting the metabolic acidosis with respiratory alkalosis. Although arterial blood gas is more representative for the variables of PO2, SO2, pCO2 and tCO2 , the blood gas changes described above were apparent, even in venous blood.
The levels of chloride and sodium electrolytes detected were not significantly different at T56 relative to T0; however, the levels of potassium and calcium were significantly reduced (p < 0.05). Similar results have been found by other authors who evaluated these ions after animals exercised and began sweating [41, 42].
The relative polycythemia observed was associated with hemoconcentration (confirmed by the increase in serum TP at T56) and splenic contraction, which releases a large number of erythrocytes into the circulation . In a study performed by Kurosawa et al. horses that were given caffeine and subjected to exercise had a greater packed cell volume than horses that were not given caffeine . Caffeine increases the release of catecholamines, especially epinephrine, which stimulates splenic contraction, thus increasing Htc [36, 43]. Furthermore, caffeine can induce diuresis in horses , although urine production was not measured in the present study, this effect might have contributed to the Htc increase.
The detection of caffeine in sport horses has been considered doping by the International Equestrian Federation; however, in addition to the intentional administration of caffeine with the intent of improving performance, caffeine can also be ingested as a result of feed contamination. Currently, caffeine is not on the list of prohibited substances, but it is on a monitoring list, and investigations are only conducted when high concentrations are detected . The high concentration of caffeine found in the plasma and urine in this study could be characterized as doping, even if intake was accidental. Compared to T0, the caffeine levels found at the onset of clinical signs (T56) were, on average, 6,366 times greater in the plasma and 4,981 times greater in the urine.