Hormones and Behavior 55 (2009) 163–168

Contents lists available at ScienceDirect

Hormones and Behavior

j ourna l homepage: www.e lsev ie r.com/ locate /yhbeh

Methods of collection for salivary cortisol measurement in dogs☆

Nancy A. Dreschel a,⁎, Douglas A. Granger b

a Department of Dairy and Animal Science, Pennsylvania State University, University Park, PA 16802, USAb Behavioral Endocrinology Laboratory, Department of Biobehavioral Health, Pennsylvania State University, University Park, PA 16802, USA

☆ In the interest of full disclosure, Dr. Granger isSalimetrics LLC (State College, PA).⁎ Corresponding author.

E-mail address: nad5@psu.edu (N.A. Dreschel).

0018-506X/$ – see front matter © 2008 Elsevier Inc. Aldoi:10.1016/j.yhbeh.2008.09.010

a b s t r a c t

a r t i c l e i n f o

Article history:

Salivary cortisol has been in Received 19 August 2008Revised 22 September 2008Accepted 22 September 2008Available online 7 October 2008

Keywords:DogCanineSalivary cortisolMethodsMeasurementStress

creasingly used as a measure of stress response in studies of welfare, reaction tostress and human–animal interactions in dogs and other species. While it can be a very useful measure, thereare a number of saliva collection issues made evident through studies in the human and animal fields whichhave not been investigated in the canine species. Collection materials and the volume of saliva that iscollected; the use of salivary stimulants; and the effect of food contamination can all dramatically impactcortisol measurement, leading to spurious results. In order to further examine the limitations of thecollection method and the effects of collection material and salivary stimulant on salivary cortisol levels, aseries of clinical, in vitro and in vivo studies were performed. It was found that there is a large amount ofinter- and intra-individual variation in salivary cortisol measurement. Beef flavoring of collection materialsleads to unpredictable variability in salivary cortisol concentration. Using salivary stimulants such as citricacid also has the potential to affect cortisol concentration measurement in saliva. Hydrocellulose appears tobe a useful collection material for salivary cortisol determination. Recommendations for collection materialsand use of salivary stimulants are presented.

© 2008 Elsevier Inc. All rights reserved.

Introduction

Salivary cortisol has been increasingly used as a measure of stressresponse in dogs (Beerda et al., 1996, 1998, 1999; Vincent and Michell,1992) and has found particular favor in canine studies of welfare(Coppola et al., 2005), reaction to stress challenges (Bergeron et al.,2002; Dreschel and Granger, 2005; Horváth et al., 2007) and human–animal interactions (Jones and Josephs, 2006). While plasma cortisollevels are the most indicative of the activity of the hypothalamic-pituitary-adrenal (HPA) system, blood collection is an invasiveprocedure, requiring skilled technical capabilities; a compliantsubject; and sample collection, processing and storage capabilities.Handling and venipuncture have been shown to increase canine bloodcortisol levels 20 min later (Hennessy et al., 1998).

Because saliva sampling is non-invasive and salivary cortisol ishighly correlated with plasma cortisol (Beerda et al., 1996), salivarycortisol can be used as a measure of HPA activity at convenient andmeaningful times of the day in naturalistic settings (Kirschbaum andHellhammer, 1994). Although handling of the animal is required,saliva seems to be tolerated well by many domestic animals. Kobeltet al. (2003) showed that up to 4 min could be taken to collect a salivasample from a dog without the effect of handling being reflected incortisol concentrations.

the founder and president of

l rights reserved.

While salivary cortisol is a useful measure, a number of constraintsassociated with its collection and measurement have been madeevident through studies in the human and animal fields. In particular,collection materials, the volume of saliva collected, the use of salivarystimulants, and the effect of food contamination have all been shownto dramatically impact cortisol measurement in humans, leading tospurious results (Granger et al., 2007).

Although many of the new assays require a small volume of salivasample (as little as 25 μl) for cortisol measurement, there may bedifficulty in obtaining even this amount from small dogs or lesscompliant subjects. This limits the number of tests that can be run andthe ability to run tests in duplicate for validity. Recent research alsoindicates that low volume samples result in a disproportionatelylower percentage of cortisol retrieved from absorbent materials,resulting in the potential for considerable error variance in measure-ment (Harmon et al., 2007). This effect wasmost pronounced in cottonbraided rope.

In order to ensure adequate sample volume, previous studies indogs have used salivary stimulants, such as citric acid, to inducesalivation. Techniques described include sprinkling a “few pellets”(Beerda et al., 1998) or a “number of crystals” (Bergeron et al., 2002) ofcitric acid on a dog's tongue, or wiping the dog's tongue and palettewith a citric acid soaked cotton ball (Kobelt et al., 2003). However, theuse of citric acid has been shown to artificially increase the levels ofboth cortisol and testosterone measured in human saliva, likelyrelated to an increase in sample acidity (Schwartz et al., 1998; Grangeret al., 2004). Because it is known that canine saliva has bufferingcapabilities to protect the individual against ingestion of more acidic

164 N.A. Dreschel, D.A. Granger / Hormones and Behavior 55 (2009) 163–168

substances (Reece, 1994), the effect of citric acid stimulation onmeasurement of canine salivary cortisol levels is unknown.

The materials used to collect saliva also have the potential tointerfere with hormone testing. Typical collection of saliva in dogsinvolves the collector holding a cotton dental rope in the dog's mouthfor 1–2 min. The saliva is then extracted by compressing the saturatedrope in a 5 ml syringe (can be performed on-site) or centrifuging in aSalivette device (Sarstedt, NC). Although cotton-based sample collec-tion methods have been shown to interfere with a variety of salivarybiomarkers in humans, cortisol levels seem unaffected (Shirtcliff et al.,2001). As stated before, however, it can be difficult to obtain a goodpercentage yield of saliva from a cotton rope, introducing smallsample issues. Surgical ophthalmic spongesmade from hydrocelluloseare effective at absorbing saliva and give a good percentage yieldwhen centrifuged (Harmon et al., 2007). In humans, a high correlationbetween cortisol measurement using hydrocellulose and passive droolhas been noted (Granger et al., 2007). Hydrocellulose has not beenpreviously tested as a collection material for canine saliva for cortisoldetermination.

It has been suggested by participants in previous studies (Drescheland Granger, 2005) that flavoring the cotton rope used would aid insaliva collection from dogs. However, animal or plant origin foodparticles in the mouth may contain products that cross-react with theantibodies used in salivary immunoassays and lead to spurious resultsin the measurement of many hormones, including cortisol in humansamples (Magnano et al., 1989; Shirtcliff et al., 2001). Whether thesmall amount of product used in flavoring a cotton rope for salivacollection would affect cortisol measurement is unknown.

While the effects of salivary stimulants, food contamination anddifferent collection media have been investigated in human saliva, theeffects of these factors on saliva collection for cortisol measurement indogs have not been investigated. This paper seeks to close these gapsin the literature by describing a series of clinical, in vitro and in vivostudies of collection methods and materials for salivary cortisolmeasurement in dogs. In all the studies, saliva was collected fromadult (over the age of 1 year), healthy dogs weighing at least 9 kg.Female dogs were excluded if they were in estrus or pregnant. All ofthe dogs used were privately owned animals with often unknownbirthdates and breed backgrounds. In each study, owners were askednot to have fed their dogs in the 20 min prior to sample collection. Allstudies were performed with the approval of the Pennsylvania StateUniversity Institutional Animal Care and Use Committee.

Study 1: Collection of saliva in a clinical setting

The first study reported here had a threefold purpose. First, aquantity of canine saliva was desired for further in vitro testing ofcollection methods. Second, the ease with which skilled professionalswere able to obtain saliva samples in a clinical setting was examined.Third, the effect of perceived subject “stress” or “excitement” on theease of collection and volume of sample collection was studied. Theclinical veterinary setting was chosen because of the large number ofpatients presented, the already established handling and technicalskills of the veterinarians and technicians, and the range of expectedtemperaments of animals presenting to the hospital.

Methods

Three veterinarians and three veterinary technician/assistantsworking at an active small animal hospital collected saliva samplesfrom healthy adult dogs presenting for wellness and vaccinationappointments. The primary investigator (ND), a veterinarian workingat the clinic, instructed one other veterinarian, a technician and aveterinary assistant on collecting samples, and provided writteninstructions and explanations that were posted near the samplecollection materials which were distributed throughout the hospital.

The trained individuals then trained the other veterinarian andassistant who also participated.

Saliva samples were obtained from 23 adult dogs of pure andmixed-breed backgrounds ranging in age from 1.5 to 12 years. Aftergaining owner permission, samples were obtained by holding a 7.5 cmcotton dental rope in the subject's mouth for 1–2min (with themouthheld closed if necessary) and the rope was placed in a salivette. Thesample collector recorded the time of collection, and rated each dogsubjectively on a scale of 1–5 on the following perceived character-istics: ease of handling (1— “very easy” to 5— “very hard”), excitementof dog (1— “verymellow”, 5— “very excited”), and “stress” of dog (1—

“not stressed”, 5— “majorly stressed”). The samples were immediatelyplaced in the clinic freezer at −18 °C and frozen until transport on ice tothe Behavioral Endocrinology Laboratory at Penn State Universitywhere they were stored at −40 °C until testing. At the time of testing,the samples were thawed and spun in a centrifuge at 3000 rpm for15 min. An approximate volume was estimated for each sample basedon comparison to a graduated tube of the same size. Sampleswith largevolumes were selected for use in a subsequent study.

Descriptive statistics were performed on the quantity of salivaobtained, ease of collection and characteristics of the subjects.Correlational analyses were performed to determine if there wererelationships between the volume of saliva obtained, the ease ofcollection and the relative excitement or stress of the dog. A 2×2(collector×characteristic) ANOVAwas performed to determine if therewere differences in amount collected or ease of collection if thesampler were a veterinarian or a veterinary assistant/technician.

Results of Study 1

The volume of saliva collected ranged from 0 to 1500 μl,(M=351.1 μl, SE= .48). The mean “ease” of collection recorded was1.6 (SE= .52), the mean “stress” of the dogs was 2.2 (SE= .52), and themean “excitement” of the subjects was 2.9 (SE= .54).

There were no correlations between perceived ease of sampling,“stress” of dog, “excitement” of dog and the amount of samplecollected. There were also no significant differences in amountscollected or ease of collection if the collector were a veterinarian orveterinary technician/assistant.

Conclusions

The mean volume of saliva collected was fairly high. All but onesample yielded enough to perform a salivary cortisol assay. There wasan inherent selection bias for good-natured, easily-handled dogs fromwhich to obtain samples, as the veterinary staff selected which dogs tosample. It is no surprise, then, that the mean “ease” of samplecollection was relatively low (“easy”). It is interesting to note thatthere was no correlation between perceived “stress” and perceived“excitement” in dogs, and that both of these characteristics fell in themiddle range. Many dogs presenting to veterinary clinics areperceived to be either “stressed” or “excited” and are often both.None of these factors influenced the volume of saliva collected.Veterinarians and veterinary assistants/technicians were equallyadept at collecting adequate saliva samples from this population.This study indicates that by themethod described here, with relativelylittle training, veterinarians, veterinary technicians and veterinaryassistants are able to collect adequate quantities of saliva to measuresalivary biomarkers in a clinical setting.

Study 2: In vitro determination of the effect of method collectionon salivary cortisol measurement

The purpose of the second study was to determine, in vitro, if citricacid stimulation interferes with cortisol measurement in canine saliva,and, if so, to determine at what concentration of citric acid this effect is

Fig. 1. The addition of citric acid to canine saliva causes a decrease in salivary pH and anincreased cortisol measurement in vitro (error bars represent SEM).

165N.A. Dreschel, D.A. Granger / Hormones and Behavior 55 (2009) 163–168

seen. Hydrocellulose and beef-flavored cotton ropes as collectionmaterials were also tested for interferencewith cortisol measurement.

Methods

Each sample from six dogs (1 neutered male and 5 spayed femalesincluding 2 Bassett hounds and 4 dogs of unknown mixed breeding)collected as part of the previously described study was aliquoted intosix microcentrifuge tubes and refrozen at −40 °C before cortisoltesting. The samples were thawed and assayed for salivary cortisolusing a highly-sensitive enzyme immunoassay kit (Salimetrics, StateCollege, PA). The test uses 25 μl of saliva (for singlet determinations),has a range of sensitivity from .007 to 1.8 μg/dl, and average intra- andinter-assay coefficients of variation of less than 10% and 15%respectively. Method accuracy, determined by spike recovery, andlinearity, determined by serial dilution are 105% and 95%. All sampleswere performed in duplicate and those with coefficients of variationgreater than 15% were repeated.

Control samples containing 80 μl of saliva from each subject weretested for cortisol and for pH using visual pH strips (colorpHastIndicator strips pH 0–14, EM Science, Gibbstown, NJ).

To determine the effect of citric acid on cortisol measurement, .01 gof food grade citric acid crystals was added to 100 μl of saliva from eachsubject to create a .1 g/ml solution. Serial dilution then yielded twoadditional concentrations of .01 g/ml and .001 g/ml citric acid/saliva.The pH of each concentration was measured using visual pH strips.Cortisol levels were measured at each concentration of citric acid.

To examine the effect of collection material on cortisol concentra-tion, 120 μl of saliva from each subject was absorbed with a BDhydrocellulose eye sponge. The handle of the sponge was cut shortand the spongewas placed handle down in amicrocentrifuge tube andre-spun. Saliva was harvested and tested for cortisol concentration.

Beef-flavored dental rope was prepared in advance by soaking acotton dental rope in a solution prepared by mixing one food gradebeef-bouillon cube with 1 cup of boiling water as per instructions onthe bouillon label. The cotton rope was then dried by baking in anoven at 60 °C for 2 h. The flavored dental rope was cut into 7.5 cmlengths and the ends were discarded. Two hundred μl of saliva fromeach subject was absorbed into the dental rope and re-extracted bypressing out of a 5 ml BD syringe. Cortisol was measured in eachsample.

Examination of the cortisol data revealed positively skeweddistributions. Natural log transformation was performed to establishnormal distributions prior to analysis. As appropriate, analyses wereconducted using the transformed values, but for ease of interpretation,the values in the text and figures are raw scores. Paired sample t-testswere performed to compare the control procedure with the experi-mental procedure. For ease of comparison with the follow-up studies,non-parametric Wilcoxon signed-rank tests were performed on theskewed data. The SPSS statistical software package (SPSS, Inc., Chicago,IL) was used for all analyses. A concordance correlation coefficient todetermine the reproducibility of scores using each method wascalculated for all pairs (Lin, 1989). The concordance correlationcoefficient evaluates agreement for two responses on a continuousscale and will have a value of 0 to 1.0 with 1.0 indicating perfectconcordance between the two variables.

Results of Study 2

The mean salivary cortisol for the control samples was .17 μg/dl(SE= .06). The mean cortisol with the beef-flavored rope was .20 μg/dl(SE= .06) and the mean cortisol from the samples passed through thehydrocellulose was .15 μg/dl (SE= .05). There was a significantdifference between the control (cotton) and the beef-flavored ropevalues (pb .01). There was no significant difference between the meanvalues of the control and the hydrocellulose. The concordance

coefficient for the beef-flavored rope collection method was .92 andthe concordance coefficient for the hydrocellulose method was 1.0.

The addition of citric acid decreased salivary pH from amean of 8.4at control to 1.8 at .1 g/ml, 3.2 at .01 g/ml and5.5 at .001 g/ml (see Fig.1).

Cortisol at the .1 g/ml level could not be measured, as it was offthe scale (N1.8 μg/dl). The mean cortisol at the .01 g/ml dilution was.85 μg/dl (SE= .37) and at the .001 g/ml citric acid dilution the meanwas .18 μg/dl (SE= .05). The difference between cortisol measure-ment at .01 g/ml and at control was significant (pb .01), but therewas not a significant difference between the .001 g/ml level and thecontrol (Fig. 1). The concordance coefficient for the .01 g/ml citricacid concentration was .16 and the concordance coefficient for the.001 g/ml citric acid concentration was 1.0.

Conclusions

This study clearly shows the potential limitations of using citric acidas a stimulant for saliva collection for cortisol measurement. As little as.01 g/ml of citric acid caused a significant drop in pH and acorresponding significant increase in measured salivary cortisol con-centration. Thirty individual crystals of citric acid weigh approximately.01 g. Because the quantity of saliva typically collected from a dog'smouth is in the .1–.5ml range, there is easily the possibility of achievinga .01 g/ml concentration with a “pinch” of citric acid placed in a dog'smouth. Lower concentrations (.001 g/ml, approximately 1–4 crystals ofcitric acid per ml) did not significantly lower the pH concentration anddid not have an effect on salivary cortisol measurement in vitro.

This study did show a high concordance between measurementstaken before and after passing the saliva through a beef-flavoredcotton rope. However, there was a significant difference between themeasurements with a higher cortisol concentration measured in thebeef-flavoring condition than in the control condition.

Hydrocellulose appeared in vitro to be a useful collection medium.There was not a significant difference between the salivary cortisolconcentration measured before and after the saliva was passedthrough a hydrocellulose sponge. The concordance coefficient wasalso very high for this measure.

Study 3: In vivo determination of the effect of collection methodson salivary cortisol measurement

Based on the findings of the previous study, an in vivo challengewas designed to determine if salivary stimulation with citric acid or

Fig. 2. There is no statistically significant difference in cortisol measurement in caninesaliva collected with beef-flavored ropes or hydrocellulose swabs or when salivation isstimulated with citric acid in vivo.

166 N.A. Dreschel, D.A. Granger / Hormones and Behavior 55 (2009) 163–168

sampling with beef-flavored cotton ropes or hydrocellulose swabswould have an effect on cortisol measurement in the live animalsetting. In particular, this study sought to determine if there aremechanisms within the live animal system to counteract the negativeconsequences of using citric acid stimulation. In addition, empiricdescriptions of the dogs' acceptance of different methods of salivacollection were obtained.

Methods

Saliva samples were collected from 24 purebred dogs (18 femalesand 6 males representing six sporting-dog breeds — 6 Germanshorthaired pointers, 5 English pointers, 3 English setters, 2 Springerspaniels, 2 Brittany spaniels, and 6 Labrador retrievers) by theresearcher at a breeding and training kennel. Eight subjects (2 malesand 6 females per group, with the breeds distributed as evenly aspossible over the groups) were enrolled in each experimentalcondition (citric acid, hydrocellulose, beef flavoring). Control salivasamples were collected from all dogs using a cotton dental rope aspreviously described. For each dog the control sample was collectedfirst, followed by the experimental sample, to avoid contaminationbetween the experimental procedures and the control. All sampleswere immediately put on ice and frozen. One subject in the beef-flavoring condition did not have an adequate volume of sample tomeasure control cortisol.

To determine the effect of citric acid on cortisolmeasurement in thefield, the inside of the dog's mouth and gums was swabbed using acitric acid coated cotton ball as described by Kobelt et al. (2003). Thecotton ballswere prepared byadding 1.25ml of a 5% citric acid solutionto one half of a cottonwool ball and drying in an oven at 60 °C for 3 h.Immediately following the swabbing of the dog's mouth, saliva wascollected using a plain cotton rope, placed in a salivette, and frozen.

To examine the effect of collection material on cortisol measure-ment, saliva was collected using a hydrocellulose eye sponge (BD)(DeWeerth et al., 2007). The sponge was held by the handle and heldin the dog's mouth for approximately 30–60 s which was adequate forfull saturation of the sponge. The handle of the spongewas then cut sothat it could be placed handle side down in amicrocentrifuge tube andfrozen. Beef-flavored dental ropes prepared as described above wereheld in the subject's mouth for 1–2 min to allow for saturation of thecollection material. They were then placed in salivettes and frozen.

Salivary cortisol was measured in all samples and the pH wasmeasured in the control and citric acid samples as describedpreviously in “in vitro determination of the effect of collectionmethods on salivary cortisol measurement”. Concordance correlationcoefficients were calculated for each pair of measurements. Becausethe data was skewed and could not be normalized throughtransformation, differences between the control and experimentalsamples were examined with the non-parametric Wilcoxon signedrank test.

Results of Study 3

The mean salivary cortisol values in the control and experimentalsamples are shown in Fig. 2. There were no significant differencesbetween the controls and any of the experimental conditions in theWilcoxon signed rank test. The concordance coefficient for hydro-cellulose and control in the kennel dogs was .53. The concordancecoefficient for the measurements following swabbing with citric acidand the control was .63. The concordance coefficient for themeasurement from the beef-flavored rope and the control was .3.Graphs of the mean salivary cortisol in the experimental and controlsamples are seen in Fig. 2. There was not a significant differencebetween the control pH (M=8.5, SD= .53) and the experimental pH(M=8.7, SD= .92) for those dogs whose mouths were swabbed withthe citric acid cotton ball.

Empirically, the dogs resisted having their mouths swabbed withthe citric acid-flavored cotton ball before saliva collection, but this didnot interfere with the researcher's ability to collect saliva. The dogsreadily chewed on the beef-flavored cotton ropes and also chewed onthe unflavored cotton ropes, although without as much enthusiasm.The hydrocellulose swabs were particularly easy to use, as the handlesfit within the interdental space in the dogs' mouths and they wereable to close their mouths without chewing on the swab or stick.

Conclusions

This experiment sheds light on the use of citric acid for salivastimulation in dogs. Using the technique of wiping the inside of thedog's mouth with a citric acid soaked cotton ball did not cause adecrease in pH of the dog's saliva. In fact, in several cases, the pHincreased following the procedure, indicating a buffering response ofthe dogs' saliva. However, despite the absence of a significant pHchange, there was a very low concordance in measurement betweenthe control condition and the experimental condition. Dogs resistedhaving their mouths wiped with the cotton swab but it did increasesaliva flowand allow for a second sample to be collected. Therewas nodifference in the volumes collected before and after swabbing thedogs' mouths.

Non-parametric testing did not show significant differences incortisol concentration measured in unflavored cotton rope vs. beef-flavored rope or hydrocellulose, but there was little correlationbetween the measurements taken with the various methods. Therewas also a very low concordance coefficient and a large standard errorin the measurements between the control and the beef-flavored ropeconditions as well as between the hydrocellulose and cotton ropeconditions. Qualitatively, some of the dogs did seem to enjoy chewingon the beef-flavored rope more than the plain cotton rope and thehydrocellulose sorbettes were quite easy to manipulate in the dogs'mouths.

Study 4: Consistency in cortisol measurement across time usingvarious methods of saliva collection

Because of the lack of concordance between different measureswhich was not reflected by significant differences between themeasures in the previous study, a fourth study was performed. Itwas postulated that because the samples were taken within minutesof each other in the in vivo study, the collection of the first control

167N.A. Dreschel, D.A. Granger / Hormones and Behavior 55 (2009) 163–168

sample may have interfered with the collection of the subsequentsamples. The final in vivo study looked specifically at the reliability oftest measures at two collection times spaced at a 20 minute intervalwhen no “stressor” was taking place.

Methods

Saliva was collected from six privately owned dogs (5 neuteredmales and one spayed female, age range approximately 1.8–9 years,M=4.4 years) by their owners in their own homes. Samples werecollected using three different methods at six distinct time points.Each method was repeated two times, with a period of 20 minoccurring between collection times. The hydrocellulose sorbette, beef-flavored rope and citric acid stimulation methods were used asdescribed in the previous study. Owners were instructed to choose atime of day when their animal was at rest and not stimulated by otheractivity. At least 1 h elapsed between collections using differentmethods.

Samples were frozen and stored at −40 °C until testing. Sampleswere assayed for salivary cortisol using an ELISA test (Salimetrics,State college, PA) as previously described. Data from samplespreviously collected using cotton rope at 20 minute intervals on acontrol day for a previous studywere also used (Dreschel and Granger,2005). This data was analyzed and compared to the other collectiontechniques.

Concordance coefficients were calculated for each pair of measure-ments. The Wilcoxon signed rank non-parametric test was used todetect differences in cortisol concentration between the time pointswithin each experimental condition.

Results of Study 4

There were no significant differences using the Wilcoxon signedrank test between cortisol concentrations at the 20minute time pointsfor any of the samples. The concordance coefficient over time forhydrocellulose was .88, for citric acid it was .49 and for the cotton ropealone, it was .36. The concordance coefficient for the beef-flavoredrope was 0.

Conclusion

From previous research (Dreschel and Granger, 2005), it wasknown that there was no significant difference in samples taken withcotton rope at 20 minute intervals on a control day in a pet's ownhome. In this study, we also found no significant differences inmeasurements taken at 20minute intervals using any of the collectiontechniques. We did, however, find a very low concordance for beef-flavored rope using this technique. The concordance for hydrocellu-lose was quite high.

General conclusions and recommendations for salivary collectionmaterials and techniques for cortisol determination

Previous research has shown that there are potential limitations inthe use of various collection methods and media for salivary cortisoldetermination. The evidence from the studies described in this articlesupport specific recommendations for the collection of canine salivafor cortisol measurement. It is also clear that there is a great deal ofinter- and intra-individual variation in salivary cortisol levels. Some ofthis variation is likely inherent in individual animals. In addition,variation exists in these studies because different individuals collectedthe samples, different breeds were involved and animals of different(and often unknown) ages were included. With the exception of thefirst study, in which the ability to collect saliva in a non-homeenvironment was part of the study, saliva collections took place at thedogs' own places of residence, often by their owners. As the dogs

served primarily as their own controls and therewas not a comparisonof cortisol levels between subjects, it is hoped that inter-individualvariation between breeds and environments would not affect themethod-focused results.

Beef-flavored ropes, though well accepted by canine subjects,introduce unpredictable variability to cortisol concentration measure-ment. This is consistent with previous research findings that foodparticles can contain products which interfere with cortisol measure-ment. Ropes flavored by this method are not reliable for salivacollection for cortisol determination. Whether different concentra-tions of beef flavoring or different types of flavoring would fail tointerfere with the immunoassay is not known.

Citric acid in small amounts does not appear to affect salivarycortisol measurement to a significant degree in the dog. Although thein vitro study illustrates the potential effects of high levels of citricacid, the dog's mouth in vivo appears to have a buffering capabilitythat counteracts the acidity of this salivary stimulant. Anecdotally,dogs resisted having their mouths swabbed with the citric acid cottonball and significantly more saliva was not obtained with the citric acidthan with cotton rope alone. However, in cases where larger volumeswould be useful, citric acid stimulation is unlikely to affect cortisolconcentration variability.

Hydrocellulose as a collection material appears to be very effectivefor canine saliva collection. The concordance with cotton rope washigh, particularly in the controlled in vitro study. One concern withthis technique is obtaining adequate sample volume with which torun the assays, as the swabs themselves are small. It is recommendedthat researchers use two ormore swabs at the same time and reinforcethe importance to those collecting the sample to leave the swabs inthe dog's mouth for an adequate length of time to allow for fullsaturation.

While there are limitations to the use of salivary cortisol, many ofthese limitations may be overcome by planning, careful instruction tothose collecting samples and research design. Because saliva samplesmay be collected in the absence of the researcher, adherence tocollectionmethodsand times takes onanew importance. The researcheris dependent on those collecting samples to adhere to procedures thathave been specified. It is important for the researchers to be aware ofthese limitations so that their results can be properly interpreted.

Acknowledgments

Data in this paper are from a thesis titled “The biobehavioral effectsof stress related to fear and anxiety in domestic canines” that wassubmitted in partial fulfillment of the requirements for the Degree ofDoctor of Philosophy in the Department of Biobehavioral Health, TheGraduate School of The Pennsylvania State University. We thankSalimetrics LLC (State College, PA) for the immunoassay reagents andmaterials and expertise. We thank the doctors, staff and clients ofCentre Animal Hospital in State College, PA for their assistance incollecting saliva. We also thank Eric and Bernadette Gilliland, ownersof the Warrior's Mark Wingshooting Lodge and other friends andcolleagues for access to their dogs' saliva.

References

Beerda, B., Schilder, M.B.H., Janssen, N.S.C.R.M., Mol, J.A., 1996. The use of saliva cortisol,urinary cortisol, and catecholamine measurements for a noninvasive assessment ofstress responses in dogs. Horm. Behav. 30, 272–279.

Beerda, B., Schilder, M.B.H., VanHooff, J.A.R.A.M., DeVries, H.W., Mol, J.A., 1998.Behavioural, saliva cortisol and heart rate responses to different types of stimuliin dogs. Appl. Anim. Behav. Sci. 58, 365–381.

Beerda, B., Schilder, M.B.H., VanHooff, J.A.R.A.M., DeVries, H.W., Mol, J.A., 1999. Chronicstress in dogs subjected to social and spatial restriction. I. Behavioral responses.Physiol. Behav. 66, 233–242.

Bergeron, R., Scott, A.L., Emond, J.-P., Mercier, F., Cook, N.J., Schaefer, A.L., 2002.Physiology and behaviors of dogs during air transport. Can. J. Vet. Res. 66, 211–216.

Coppola, C.L., Grandin, T., Enns, R.M., 2005. Human interaction and cortisol: can humancontact reduce stress for shelter dogs? Physiol. Behav. 87, 537–541.

168 N.A. Dreschel, D.A. Granger / Hormones and Behavior 55 (2009) 163–168

DeWeerth, C., Jansen, J., Vos, M.H., Maitimu, I., Lentjes, E.G.W.M., 2007. A new device forcollecting saliva for cortisol determination. Psychoneuroendocrinology 32,1144–1148.

Dreschel, N.A., Granger, D.A., 2005. Biobehavioral reactivity to stress in thunderstorm-anxious canines and their caregivers. Appl. Anim. Behav. Sci. 95, 153–168.

Granger, D.A., Shirtcliff, E.A., Booth, A., Kivlighan, K.T., Schwartz, E.B., 2004. The“trouble” with salivary testosterone. Psychoneuroendocrinology 29, 1229–1240.

Granger, D.A., Kivlighan, K.T., Fortunato, C., Harmon, A.G., Hibel, L.D., Schwartz, E.B.,Whembolua, G.-L., 2007. Integration of salivary biomarkers into developmental andbehaviorally-oriented research: problems and solutions for collecting specimens.Physiol. Behav. 92, 583–590.

Harmon, A., Hibel, L.C., Rumyantseva, O., Granger, D.A., 2007. Measuring salivary cortisolin studies of child development: watch out—what goes in may not come out ofcommonly used saliva collection devices. Dev. Psychobiol. 49, 495–500.

Hennessy, M.B., Williams, M.T., Miller, D.D., Douglas, C.W., Voith, V.L., 1998. Influence ofmale and female petters on plasma cortisol and behaviour: can human interactionreduce the stress of dogs in a public animal shelter? Appl. Anim. Behav. Sci. 61,63–77.

Horváth, Z., Igyártó, B.Z., Magyar, A., Miklósi, A., 2007. Three different coping styles inpolice dogs exposed to a short-term challenge. Horm. Behav. 52, 621–630.

Jones, A.C., Josephs, R.A., 2006. Interspecies hormonal interactions between man andthe domestic dog (Canis familiaris). Horm. Behav. 50, 393–400.

Kirschbaum, C., Hellhammer, D.H., 1994. Salivary cortisol in psychoneuroendocrineresearch: recent developments and applications. Psychoneuroendocrinology 19,313–333.

Kobelt, A., Hemsworth, P.H., Barnett, J.L., Butler, K.L., 2003. Sources of sampling variationin saliva cortisol in dogs. Res. Vet. Sci. 75, 157–161.

Lin, L.I., 1989. A concordance correlation coeeficient to evaluate reproducibility.Biometrics 45, 255–268.

Magnano, C.L., Diamond, E.J., Gardner, J.M., 1989. Use of salivary cortisol measurementsin young infants: a note of caution. Child Dev. 60, 1099–1101.

Reece, W.O., 1994. Dukes' Physiology of Domestic Animals. Cornell University Press,Ithaca, New York.

Schwartz, E.B., Granger, D.A., Susman, E.J., Gunnar, M., Laird, B., 1998. Assessing salivarycortisol in studies of child development. Child Dev. 69, 1503–1513.

Shirtcliff, E.A., Granger, D.A., Schwartz, E., Curran, M., 2001. Use of salivary biomarkers inbiobehavioral research: cotton based sample collection methods can interfere withsalivary immunoassay results. Psychoneuroendocrinology 26, 165–173.

Vincent, L.C., Michell, A.R., 1992. Comparison of cortisol concentrations in saliva andplasma of dogs. Res. Vet. Sci. 53, 342–345.