The work was done at the Clinic for Small Animal Internal Medicine, Vetsuisse Faculty, University of Zurich, Winterthurerstrasse 260, 8057 Zurich, Switzerland. Parts of the study were presented as an abstract at the 18th ECVIM-CA Congress, Ghent, Belgium, 4th–6th September 2008.
Corresponding author: Nadja S. Sieber-Ruckstuhl, Clinic for Small Animal Internal Medicine, Vetsuisse Faculty, University of Zurich, Winterthurerstrasse 260, 8057 Zurich, Switzerland; e-mail: email@example.com.
Background: Measurement of salivary cortisol is a useful diagnostic test for hypercortisolism (HC) in humans.
Objectives: To determine whether measurement of salivary cortisol concentration is a practical alternative to plasma cortisol to diagnose HC, to validate the use of salivary cortisol, and to examine the effect of time of day and sampling location on salivary cortisol.
Animals: Thirty healthy dogs and 6 dogs with HC.
Methods: Prospective, observational clinical trial including healthy volunteer dogs and dogs newly diagnosed with HC. Salivary and plasma cortisol concentrations were measured with an immunoassay analyzer. Intra- and interassay variability, linearity, and correlation between salivary and plasma cortisol concentrations were determined.
Results: The required 300 μL of saliva could not be obtained in 88/326 samples from healthy dogs and in 15/30 samples from dogs with HC. The intra-assay variability for measurement of salivary cortisol was 5–17.7%, the interassay variability 8.5 and 17.3%, and the observed to expected ratio 89–125%. The correlation (r) between salivary and plasma cortisol was 0.98. The time of day and location of collection did not affect salivary cortisol concentrations. Dogs with HC had significantly higher salivary cortisol values than healthy dogs (10.2 ± 7.3 nmol/L versus 1.54 ± 0.97 nmol/L; P < .001).
Conclusions and Clinical Importance: The ROCHE Elecsys immunoassay analyzer correctly measured salivary cortisol in dogs. However, a broad clinical application of the method seems limited, because of the large sample volume required.
Measurement of salivary cortisol concentration is a straightforward, minimally invasive procedure in humans. Sample storage is easy, which makes it an increasingly popular test.1 Determination of “late-night” and “morning” salivary cortisol concentrations are established screening tests for the diagnosis of hypercortisolism (HC) and hypocortisolism in people.2–4
Because cortisol in saliva is unbound, its concentration is approximately 12% of that of plasma; determination of salivary cortisol therefore requires either more sensitive or adapted laboratory methods than the ones used for plasma cortisol.5–7 The concentration of free cortisol in saliva and plasma establishes an equilibrium within 5 minutes.7 Blood contamination, hemolysis, macromolecules, pH, and the material used to collect saliva are factors that can affect measured salivary cortisol concentrations.8–12 Plasma proteins are only present in trace amount in saliva and therefore their influence on salivary cortisol levels is less important.
Salivary cortisol concentrations have been used mainly to assess stress in dogs.8,13–15 Similar to humans, the cortisol concentration in saliva and plasma is increased significantly after administration of adrenocorticotropic hormone (ACTH) in dogs.8,16–18 Measurement of salivary cortisol levels was considered a simple and noninvasive test in horses and guinea pigs suspected of having HC.16–19 A diagnosis of HC in dogs is currently based primarily on the results of a low-dose dexamethasone suppression test (LDDS test), characteristic clinical and laboratory findings, and/or an increase in the urine cortisol-to-creatinine ratio (UC/C). However, the LDDS test is time-consuming and invasive, and some owners find it difficult to collect a sample of urine. A noninvasive test requiring sample material that is easy to collect by the owner would be a distinct advantage in the diagnosis of HC in dogs.
The goals of this study were to investigate the feasibility of collecting salivary samples in dogs, the accuracy of an immunoassay for measurement of salivary cortisol, and the effects of location (home or clinic) and time of day on salivary cortisol levels. In addition, salivary cortisol concentrations in dogs with HC and healthy dogs were compared.
Materials and Methods
All dogs were privately owned. Informed consent was obtained from all owners and the study was approved by the Swiss Cantonal Veterinary Office.
Thirty clinically healthy dogs, which consisted of 20 females (17 spayed) and 10 males (7 neutered), were used. They ranged in age from 5 to 16 years (median, 7 years) and weighed 6.7–42.0 kg (median, 24.0 kg). Breeds included Borzoi (n = 5), Labrador Retriever (4), Australian Shepherd (2), Shetland Sheepdog (2), Jack Russell Terrier (1), Border Collie (1), American Staffordshire Bull Terrier (1), Alaskan Husky (1), Island Hound (1), and mixed-breed dogs (12). The dogs were considered healthy based on the medical history and results of physical examination.
Dogs with HC
Fourteen dogs with HC were intended to be used in the study. They consisted of 7 spayed females and 7 males (2 neutered), which ranged in age from 6 to 16 years (median, 10 years) and weighed 4.6–38.6 kg (median, 15.0 kg). Breeds included American Cocker Spaniel (n = 1), Bichon Frise (1), Cavalier King Charles Spaniel (1), Dachshund (1), French Bulldog (1), Galgo Espagnol (1), German Shepherd Dog (1), Griffon de Fauve de Bretagne (1), Keeshond (1), Miniature Pinscher (1), Parson Jack Russell Terrier (1), Yorkshire Terrier (1), and mixed-breed dogs (2). Hematological and biochemical analyses, an LDDS test, measurement of the UC/C (1st morning urine, obtained at home on 1, 2, or 3 consecutive days), endogenous ACTH, and ultrasonographic examination of the adrenal glands were carried out in all dogs. Dogs were included in the study when clinical and laboratory findings were consistent with HC, the LDDS test and/or the UC/C were positive for HC, treatment with trilostane resulted in an adequate response, and no other treatments for HC had been administered.
A diagnosis of pituitary-dependent or adrenal-dependent HC was based on the concentration of endogenous ACTH and ultrasonographic appearance of the adrenal glands. Pituitary-dependent HC was diagnosed in 12 dogs and adrenal-dependent HC in 2.
Dogs were fasted for 1 hour before salivary samples were obtained. Saliva was collected with a cotton pad,a which was held with a pair of clamps and placed between the upper and lower premolars for 2 minutes (Fig 1). Salivation was stimulated by placing an open can of dog food within sight of the dog. The pad was centrifuged at 4,500 ×g for 25 minutesb to recover the saliva.
Blood samples were collected from a jugular vein into plastic tubes containing heparin. The blood samples were immediately centrifuged at 1,500 ×g for 5 minutes and the plasma was harvested. Salivary and plasma samples were stored in plastic tubes at −80°C until further analysis within 3 months.
Cortisol in saliva and plasma was measured by use of a competitive electrochemiluminescence immunoassay (ECLIA).c The required sample volume was 300 μL saliva or plasma.
Linearity was assessed by measuring 3 salivary samples with different cortisol concentrations (3.2, 9 , and 17.5 nmol/L), before and after making 2-, 4-, 8-, and 16-fold dilutions with Elecsys universal diluent.d The expected values were plotted against the observed values.
The intra-assay variability was determined by evaluating 3 salivary samples with low (1.4 and 3.1 nmol/L) or high (145 nmol/L) cortisol concentrations 10 times within the same run of assays.
The interassay variability was determined by evaluating 2 salivary samples with low (2.4 nmol/L) or moderate (13.9 nmol/L) cortisol concentrations at 10 different days.
The correlation between salivary and plasma cortisol concentrations was determined by evaluating plasma and saliva samples taken within 4 minutes from the same dogs (n = 28).
To examine the effect of contamination of salivary samples with blood, 3 salivary samples (A, B, and C) were mixed in vitro with 3 blood samples (I, II, and III) at saliva-to-blood ratios of 98 : 2 (2% blood), 96 : 4 (4% blood), 94 : 6 (6% blood), and 92 : 8 (8% blood). Salivary sample A (cortisol concentration 5.2 nmol/L) was mixed with blood sample I (cortisol concentration 79 nmol/L), salivary sample B (cortisol concentration 19.9 nmol/L) with blood sample II (cortisol concentration 32 nmol/L), and salivary sample C (cortisol concentration 0.5 nmol/L) with blood sample III (cortisol concentration 491 nmol/L).
Effect of Time of Day and Sampling Location on Salivary Cortisol Concentration
To determine the effect of time of day and sampling location on salivary cortisol concentrations, salivary samples of healthy dogs were collected at 4 different time points (8:00 am, 11:00 am, 2:00 pm, and 5:00 pm) at home and in the clinic.
Results were analyzed by nonparametric statistics.e,f The ratio of observed to expected values was calculated to assess linearity. The intra- and interassay variability was determined by calculating the coefficient of variation (CV). Passing-Bablok linear regression analysis was used to calculate the correlation coefficient between salivary and plasma cortisol concentrations. Spearman's correlation calculations were used to assess correlation of salivary cortisol and blood contamination.
Differences in cortisol concentrations between time points and between samples collected at home and in the clinic were analyzed by the Friedmanns repeated measures test and Dunn's posttest. Differences between healthy dogs and dogs with HC were analyzed by the Mann-Whitney U-test. Differences were considered significant at P < .05.
Feasibility of Saliva Collection
A total of 326 salivary samples were collected from 30 healthy dogs, which belonged to 23 owners, 21 of whom were veterinarians, veterinary students, or technicians. Eighty-eight (27%) samples yielded an insufficient amount of saliva and could not be used for analysis. Reasons for these failures included defense movements of the dog, such as head shaking, backing off, and pawing at the clamp, failure to chew the cotton pad, or excessive chewing resulting in destruction of the cotton pad. It was extremely difficult for 7 owners to obtain salivary samples, and they considered the method unpractical.
Although 14 dogs with HC had initially been included, 8 had to be excluded after the start of the study because of fearful or aggressive behavior, which prevented collection of saliva. Thirty salivary samples were collected from the remaining 6 dogs; in 15 samples (50%), the amount of saliva was insufficient for hormone analysis because of the reasons listed for healthy dogs. Two owners were asked to collect saliva at home, which posed marked problems for both; they were successful in 1 of 4 and 2 of 4 attempts, respectively.
The slope of the regression equations and the r2 were close to 1 for each dilution (Fig 2). The ratio of the observed versus the expected values ranged from 100 to 125% for a cortisol concentration of 3.2 nmol/L, from 89 to 116% for a concentration of 9 nmol/ L, and from 90 to 100% for a concentration of 17.5 nmol/L.
The intra-assay CV was 17.7, 5.5, and 8.1% for the samples with 1.4, 3.1, and 145 nmol/L cortisol, respectively.
The interassay CV was 17.3 and 8.6% for the samples with 2.4 and 13.9 nmol/L cortisol, respectively.
The correlation coefficient for salivary and plasma cortisol was 0.984 for a range of 0.6–427.5 nmol/L cortisol in the saliva.
Mixing a salivary sample with moderate cortisol concentration with a blood sample with a 15 times higher cortisol concentration or a salivary sample with a very low cortisol concentration with a blood sample with a very high cortisol concentration (800 times higher) resulted in a positive, significant correlation (P= .0167 and .017) between blood content and cortisol concentration (Fig 3).
Effect of Time of Day and Sampling Location on Salivary Cortisol Concentration
Salivary samples for comparison of cortisol concentrations at different time points and sampling locations were available in 21 of 30 dogs (4 samples obtained in the clinic and 4 at home for each dog). There was no significant difference between concentrations at different time points or at different locations (Fig 4).
Dogs with HC
Twelve salivary samples were collected in the clinic and 3 at home. The low number of samples taken at home excluded a statistical comparison between the 2 locations.
In 12 of 15 salivary samples, the cortisol concentration was higher than the 95th percentile of the values in healthy dogs (Fig 5).
The mean salivary cortisol concentration of dogs with HC was significantly higher than that of healthy dogs (P < .001) (Fig 6).
The ROCHE Elecsys immunoassay correctly measured canine salivary cortisol. Salivary cortisol concentrations were not affected by location (home or clinic) or time of day of collection. Dogs with confirmed HC had significantly higher salivary cortisol concentrations than healthy dogs. However, because of problems encountered with collection of saliva and because of the high sample volume needed, the method described here is currently not an alternative to the determination of plasma cortisol levels in dogs. With the use of a different assay requiring less saliva volume and with further development of the saliva collection technique the method may promise useful in future.
Based on the linearity, the intra- and interassay variability, and the correlation between the cortisol concentration in saliva and blood, we concluded that the ROCHE Elecsys immunoassay was an accurate and reliable method for cortisol measurement in saliva. One drawback of the immunoassay is the large sample volume (300 μL) needed. Other assays require as little as 25 μL saliva for cortisol measurement.20 Because the sample volume was often a limitation in this study, the ROCHE Elecsys immunoassay might have limited clinical utility in dogs.
Another goal of this study was to determine the feasibility of collecting saliva in dogs. In human medicine, collection of saliva is a straightforward and noninvasive method, which can be carried out by the patient.2 In contrast, collection of saliva in dogs proved to be relatively difficult in our study; even experienced veterinarians and technicians often had problems collecting an adequate amount of saliva. Collection of saliva at home was only accomplished in good-natured dogs with motivated owners. It soon became clear that fearful or aggressive dogs could not be included.
Salivary secretion is under neural control. Stimulation of the sympathetic nerve supply causes vasoconstriction and a decreased rate of salivary flow.21 Therefore, production of inadequate amounts of saliva in the study reported here may have been attributable to stress and excitement. In a recent study, however, no association between stress or excitement of dogs and the volume of saliva collected was found.20
The use of citric acid to increase saliva production has been reported in dogs.13,22 When using chemicals or medications to increase salivary flow, however, one has to consider influences on the cortisol concentrations. In humans, stimulated saliva sample with increased acidity has been shown to artificially increase the level of cortisol in saliva.23 The same effect has been shown in vitro in saliva of dogs; in vivo, however, citric acid did not affect salivary cortisol measurement to a significant degree, possibly because of a buffering capability of the dogs, mouth.20 The amount of saliva gained, however, was not significantly higher, if citric acid was used, which certainly questioned its use, especially because dogs resisted having their mouths swabbed with citric acid.20
The collection pad material may have played a role in the amount of saliva collected. We used pads made of cotton,a which retain 54.7% (±2.3%) of the collected saliva. Pads made of polyesterg are a new commercial product that allows a return of 95.8% (±1.1%) of the saliva after centrifugation.24 However, in a preliminary trial in our clinic, there was no difference in the amount of saliva obtained with either the pads made of cotton or the ones made of polyester in 8 healthy dogs (data not shown). Thus, the pad material did not seem to play an important role in the amount of saliva collected.
Another aspect could be the comfort of the collection material. The cotton and the polyester pads are large and dogs may experience discomfort with their use. Newer devices to collect saliva, like eye sponges, have been described in humans and seem more comfortable.25 With these devices enough saliva could be collected in humans for a cortisol assay running with 50–100 μL of saliva.25
Factors that can affect the cortisol concentration in saliva include stress, macromolecules within the saliva, pH of the saliva, collection pad material, and blood contamination.8–12,23 A stress-induced increase in cortisol levels would be expected not until 30 minutes after the stressful incident.8 The saliva samples were collected within 4 minutes of blood collection, which was not long enough for elevation in cortisol levels caused by stress. To decrease the effects of macromolecules or pH on cortisol levels in humans, patients are not allowed to eat 1 hour before saliva collection and the mouth is cleaned and rinsed with cold water immediately before obtaining saliva.4 In our study, these measures were not followed except for removal of food 1 hour before the collection. Thus, food particles may have had an effect on the composition and pH of the saliva. However, anesthesia would have been necessary to thoroughly rinse the mouth, and this would have eliminated the apparent simplicity of the method.
Contamination of saliva samples with blood was often seen macroscopically in our study. The presence of blood in saliva seems an important factor because the cortisol concentration of saliva is only 4–12% of the plasma concentration and blood contamination occurs relatively easily.6 Our results showed that in a nearly physiologic situation where the blood cortisol concentration is about 15 times higher than the salivary cortisol concentration, a significant, positive correlation between blood content and cortisol concentration can occur. In a salivary sample with a blood content of 8%, which corresponded to distinct gross evidence of blood contamination, the measured cortisol value was 2 times that of the uncontaminated sample. Although in children the prevalence of blood contamination was shown to be infrequent, there was a positive correlation with the hormone concentration. However, <1% of the statistical outliers were attributable to blood contamination of the saliva sample.10 Notwithstanding these findings, blood contamination should always be noted when determining salivary cortisol levels, and measurements repeated if values are doubtful.
The time of day and location of collection (home versus clinic) had no significant effect on cortisol concentrations in healthy dogs. It is important to note that the measured cortisol concentration is affected by the method of analysis (ROCHE Elecsys) and the pad material. One study found that saliva collected with cotton pads had 38% lower cortisol values than saliva collected with polyester pads.24 Thus, our values are study-specific and only apply to saliva collected with Salivetten (Sarstedt) pads made of cotton.
Cortisol measurements in salivary samples collected at midnight have a sensitivity of 92–100% and a specificity of 93–100% for the diagnosis of HC in humans.3 Importantly, the high diagnostic accuracy is mainly due to the loss of circadian rhythm of cortisol secretion in humans with HC.3 Healthy dogs lack a circadian rhythm in cortisol secretion.26 Therefore, the diagnostic accuracy of basal salivary cortisol in dogs is not expected to reach that of humans with HC. Our results showed that the salivary cortisol concentration in dogs with confirmed HC was significantly higher than that of healthy dogs, although 3 of 15 saliva samples of dogs with HC had cortisol concentrations in the same range as healthy dogs. Determination of the sensitivity and specificity of salivary cortisol measurement would entail testing dogs with confirmed HC as well as dogs with other diseases that have signs similar to HC. The main limiting factors in this study, however, were the difficulties in collecting saliva and the high sample volume needed. Therefore, before the diagnostic accuracy of salivary cortisol measurement can be investigated, the sampling method has to be improved.