Certification testing as an acute naturalistic stressor for disaster dog handlers

Abstract

USA Federal Disaster Canine Teams, consisting of a handler and a dog, are essential for locating survivors following a disaster. Certification, required by the Federal Emergency Management Agency Urban Search and Rescue organization, requires two successful mock searches. Confirmation of the certification testing process as an acute stressor might offer further opportunities to consider stress effects on handlers and dogs in a controlled environment. This study used a pretest–posttest design to evaluate relationships between salivary hormone concentrations (cortisol and testosterone) and subjective stress ratings in handlers and controls, handler assessments of stress in their dogs, and posttest temperature and pulse rate in dogs. Posttest, both subjective stress ratings and salivary cortisol concentration were higher in handlers than controls with both correlated to handlers' assessment of stress in their dogs. Handlers' posttest salivary cortisol concentration was associated with posttest dog pulse and temperature. Posttest cortisol concentration was lower in handlers who were successfully certified compared with those who failed, and was also lower in handlers whose primary occupation was “firefighter”. Salivary testosterone concentrations increased from pretest to posttest in handlers but decreased in controls, and higher posttest handler testosterone concentration was negatively associated with posttest dog pulse rate. These findings confirm certification testing as an acute stressor, suggest a relationship between stress and performance moderated by occupation, and demonstrate an interaction between handler stress and dog physiological responses. This certification testing offers a controlled environment for targeted evaluation of effects of an acute naturalistic stressor on disaster dog handlers and dogs.

Keywords::

• Cortisol
• firefighter
• naturalistic stressor
• search dog
• stress
• salivary testosterone

Introduction

Canine search and rescue (SAR) teams each consist of a highly trained dog and the human who works with that dog, referred to as the handler. SAR dogs used to locate surviving humans in the event of a disaster are termed “disaster dogs”. Although variables affecting performance of disaster dogs have not been well studied, research has demonstrated that factors other than a dog's ability to locate scent might affect performance. For example, training dogs to find both live and cadaver scent impair performance when dogs are required to find live human scent in the presence of previously reinforced cadaver scent (Lit and Crawford 2006). Another factor that might affect performance but has not been investigated is acute stress experienced by handlers during a deployment. Although there has been research examining hormonal responses to stress in working dogs (Slensky et al. 2004; Wakshlag et al. 2004; Haverbeke et al. 2008; Horvath et al. 2008; Rovira et al. 2008), acute stress responses in disaster dog handlers have not been evaluated. Moreover, while interactions between handler and dog hormonal levels have been demonstrated in agility dogs (a recreational competition where dogs are timed while running through obstacle courses; Jones and Josephs 2006; Mehta et al. 2008), it is likely that stressors experienced by disaster dog handlers might be more extreme than those of handlers participating in recreational dog-related activities, and thereby affect the disaster dogs. Because the unpredictable and often catastrophic nature of a disaster situation makes it less suitable for in-depth study of such stress effects on performance, one means to evaluate effects of acute stress in a controlled environment might be the certification testing that teams must undergo in order to be deployed.

Because handlers have received extensive training prior to certification testing, it is unclear whether disaster dog certification, or anticipation of the certification, would generate significant changes in cortisol secretion or perception of stress. Moreover, increased cortisol secretion might arise due to physical exertion during a search exercise (Mastorakos et al. 2005) or simply from the social stress involved in being observed while performing the search exercise, particularly when those observing are aware of the locations of the hidden victims (Kirschbaum et al. 1993). In contrast, psychological stress induced by testing may be exacerbated by the realization that team deployment is contingent on maintenance of a minimum number of certified dog/handler teams.

However, because effects of stress might also be important under real-world disaster conditions, where impaired performance might result in failure to detect live victims, certification testing may offer an opportunity to evaluate stress effects in a controlled working environment. Evaluating how stress affects a disaster handler/dog team during certification could offer insight into responses during the stress of an actual deployment. To evaluate certification testing as an acute naturalistic stressor, the present study investigated whether certification testing modified salivary cortisol and testosterone concentrations differently in handlers compared with controls that completed the same exercises but were not attempting certification. Specifically, salivary cortisol and testosterone concentrations, subjective stress ratings by handlers for themselves and their dogs, and relationships between handler and dog physiological responses were examined.

We hypothesized that (1) salivary cortisol and testosterone concentrations would be higher in handlers than controls both prior to testing and immediately following testing; (2) handler subjective stress ratings would reflect cortisol concentrations; and (3) handler cortisol and testosterone concentrations would be associated with physiological responses in their dogs. Because handlers would not know upon completion whether they had certified successfully, we did not anticipate differences in cortisol concentrations between handlers who passed and those who failed. Additionally, we predicted lower cortisol responses in handlers whose primary occupation was firefighter when compared with non-firefighters, demonstrating that occupational exposure to stress might moderate responses to stress.

Materials and methods

Participants

This was a convenience sample, obtained during a Federal Emergency Management Agency (FEMA) certification evaluation (CE) (FEMA 2008). Permission to conduct this study was obtained from the chairperson for the FEMA Canine Subgroup and the Task Force hosting the certification. Handler participation was requested at the evening briefings prior to certification. The Institutional Review Board and Animal Care and Use Committee at the University of California at Davis approved this study, and all participants provided written consent.

Handlers

Handlers (n = 16; 10 males, 6 females; age group, years, 30–39: n = 3, 40–49: n = 8, 50–59: n = 3, over 60: n = 1, declined to state: n = 1) were attempting to certify their disaster dogs for deployment. Certification testing occurred over two consecutive days, with eight handlers tested per day; order was assigned by testing coordinators.

Controls

Certification evaluators and one member of the hosting team not participating in the certification testing served as controls (n = 6 controls; four males, two females; age group, years, 30–39: n = 2, 40–49: n = 2, 50–59: n = 2). Evaluators were FEMA-certified individuals assigned to observe handlers and complete CEs on handler performance.

Certification evaluation (CE)

The CE required the dog to search two separate simulated disaster scenes (“piles”) between 930 and 1400 m² (Figure 1A). Each test lasted for a total of 75 min. Handlers were tested in groups of two, so that testing was ongoing at each pile at the same time. Two groups were scheduled in the morning, starting at 8:00 and 9:40 h, followed by two groups in the afternoon, starting at 12:00 and 13:40 h. The handler was permitted to fully access one scene, which might have had zero to four hidden victims (Figure 1B). The second scene was a “limited access search”, with 1–4 hidden victims, in which the handler had to deploy the dog from outside the scene. The handler was not permitted to enter the scene until the dog had indicated location of at least one victim with a bark alert. At least one food and one clothing distraction were present, although there might have been additional scent distractions; noise distractions were permitted as well. If the dog alerted by a distraction or any place where there was no victim present (“false alert”), the team failed. Inability to locate all but one hidden victim would also result in a team failure. The handler was also evaluated on ability to successfully demonstrate defined “Canine Search Specialist” skills, which included controlling the dog, assessing the site and developing a search strategy, mapping location of victims, and appropriately debriefing prior to leaving each scene. Teams were allowed access to the search scenes only at their pre-assigned times, and were not told whether they had passed or failed upon completion of their searches; results were announced at the end of each day. The CE was the same for both days of testing, although participants did not know whether this was the case.

Figure 1  (A) Rubble pile simulating disaster scene. (B) Sample hiding location for victim, shown by arrow.

Saliva collection

For all saliva sampling, subjects chewed sugar-free gum for approximately 15 s, and then drooled approximately 3 ml of saliva into a sterile polypropylene microtubule. We could not control for variables that might have affected findings such as food or caffeine consumption prior to saliva collection. Samples were stored on ice upon collection and subsequently frozen and maintained at − 20°C until analysis.

Handlers

Pre-certification saliva samples were taken from handlers at their briefing on the evening prior to their assigned certification day. Post-competition salivary samples were collected within 15 min of completion of each handler's certification testing (range = 3–15 min following test completion). Time of pre-competition salivary sample collection was between 19:30 and 20:30 h and post-competition salivary sample collection ranged from 09:30 h to 15:00 h.

Controls

Controls arrived at the testing site on the day prior to certification testing. Upon arrival, controls ran their dogs on the same testing scenarios that would be used for certification testing. Posttest saliva samples for (evaluator) controls were collected upon completion of the exercise (range = 3–15 min following test completion). However, because all pretest saliva samples for handlers were collected at the evening briefing, in order to optimally control for differences in diurnal cortisol and testosterone variation, saliva samples were also collected from evaluators at that time. The evening saliva samples were then analyzed as “pretest” samples. Thus, for controls, “pretest” saliva sample collection was between 19:30 and 20:30 h at the first evening briefing (with the first group of handlers) and posttest salivary sample collection ranged from 11:00 to 15:00 h.

Immunoassay

Prior to assay, samples were centrifuged at 1880g for 20 min to separate the aqueous component from mucins and other suspended particles. Salivary concentrations of testosterone were estimated in duplicate using a commercial radioimmunoassay kit (Beckman Coulter, Inc., formerly Diagnostics Systems Laboratories, Webster, TX, USA). The assay procedures were those outlined in Granger et al. (1999). The assay has a least detectable concentration of 1.37 pg/ml and an intra-assay coefficient of variation 0.55%. Salivary concentrations of cortisol were estimated in duplicate using a commercial radioimmunoassay kit (Siemens Medical Solutions Diagnostics, Los Angeles, CA, USA). Assay procedures were modified to accommodate overall lower concentrations of cortisol in human saliva relative to plasma: (1) standards were diluted to concentrations ranging from 2.76 to 345 nmol/L; (2) sample volume was increased to 200 μl; and (3) incubation times were extended to 3 h. Serial dilution of samples indicates that the modified assay displays a linearity of 0.98 and a least detectable concentration of 1.38 nmol/L. Intra-assay coefficient of variation was 3.0%.

Additional data collection

Handler occupation and pass/fail data were also collected. Subjective stress assessments using a Likert scale (range: 1–5) were obtained from handlers for themselves and their dogs. The questions, “On a scale of 1–5, 1 being not at all stressed and 5 being so stressed you can barely cope, how stressed do YOU feel right now?” and “On a scale of 1–5, 1 being not at all stressed and 5 being so stressed you can barely cope, how stressed do you think your DOG is now?”, were completed by handlers both at the evening briefing, prior to testing, and immediately following testing. Posttest dog temperature and pulse rate data were collected for each dog by a veterinarian at the same time as posttest saliva samples were obtained from that dog's handler (range = 3–15 min following completion of certification testing; data for one dog were missing). Temperature was obtained with a rectal probe quick-read digital thermometer. While auscultating the heart, pulse was obtained using digital pressure over the femoral artery. Pulses over a 15-s period were multiplied by four to obtain pulse rate per minute.

Statistical analyses

Data were analyzed using SPSS Version 17.0.1. All analyses used a significance threshold of α < 0.05 (two tailed). For comparisons between handlers and controls, the dependent variables cortisol and testosterone concentrations were analyzed using general linear model repeated measures, with testing status (two levels: handler or control) as the between-subjects factor, and sampling time (two levels: pretest or posttest) as the within-subject factor. Planned pairwise comparisons were also performed to evaluate pretest and posttest differences separately for handlers and controls. All contrasts were adjusted for multiple comparisons using a Bonferroni adjustment.

The dependent variables subjective stress-handler and subjective stress-dog were analyzed using nonparametric tests. To evaluate pretest and posttest differences, separate Wilcoxon Signed-Rank tests were run for handlers and controls. Differences between handlers and controls were evaluated with separate Mann–Whitney U-tests for each sampling time.

Correlations were conducted within the handler group to examine associations between cortisol concentrations, testosterone concentrations, subjective stress ratings for themselves, and subjective stress ratings for their dogs. Pearson's r was used for correlations between cortisol and testosterone concentrations, Spearman's ρ was used for correlations between cortisol/testosterone concentrations and subjective stress ratings, and Kendall's τ-b was used for correlations between subjective stress ratings. Pearson's r was used for correlations between dog posttest temperature and pulse measures and between handler cortisol and testosterone concentrations.

An exploratory analysis was conducted within the handler group to examine the effect of certification success on cortisol concentrations, using general linear model repeated measures, with certification success (two levels: pass or fail) as the between-subjects factor, and sampling time (two levels: pretest or posttest) as the within-subject factor. Out of 16 handlers in total, 12 successfully certified and 4 failed. Planned pairwise comparisons were also performed to evaluate pretest and posttest differences separately for handlers who passed and those who failed.

Handler occupations were assigned to the dichotomous variable firefighter (Yes, n = 7; No, n = 9). A separate analysis using a student's t-test within the handler group considered whether posttest cortisol concentrations differed between handlers employed as firefighters and those in other occupations.

Results

There was no difference in gender representation between handler and control groups, X²(1, 22) = 0.03, p = 0.86, or age group, X²(6, 21) = 8.633, p = 0.20. There was no effect of day of testing (Saturday and Sunday) on pretest salivary cortisol concentrations, F_(1,14) = 1.07, p = 0.3, posttest salivary cortisol concentrations, F_(1,14) = 2.16, p = 0.2, pretest salivary testosterone concentrations, F_(1,14) = 0.43, p = 0.5, or posttest salivary testosterone concentrations, F_(1,14) = 0.23, p = 0.6. Also, there was no effect of day of testing on subjective stress ratings by handlers for themselves, pretest U(Z = − 0.967, p = 0.3), posttest U(Z = − 1.739, p = 0.08) or for their dogs, pretest U(Z = − 1.245, p>0.2), posttest U(Z = − 0.876, p = 0.38). Data for both days were therefore combined for further analyses, and gender was not included as a covariate in any analyses.

To evaluate the effect of diurnal rhythms on posttest cortisol and testosterone concentrations, times of saliva collection for subjects and controls combined were defined as either before noon or after noon. There were no significant effects of time of day comparing cortisol concentrations, t(20) = 0.668, p = 0.51, or testosterone values, t(20) = 1.24, p = 0.23, collected before noon and after noon. In addition, there was no correlation between time of collection and posttest cortisol or testosterone concentration, p>0.05.

Within the handler group, posttest cortisol and testosterone concentrations displayed a non-normal distribution (Shapiro–Wilk < 0.05). These variables were transformed for all subjects using a log₁₀ transformation. In order to assure appropriate comparisons, the pretest cortisol and testosterone concentrations were also transformed.

Cortisol analyses

To evaluate differences in cortisol levels between handlers and controls, the analysis “cortisol = Sampling Time (pretest, posttest) × Testing Status (handler, control)” showed a main effect of sampling time, F_(1,20) = 28.68, p < 0.001, η² = 0.59, and testing status, F_(1,20) = 8.53, p = 0.008, η² = 0.30, on cortisol concentrations (Figure 2A). Both main effects were in the predicted direction, with posttest cortisol concentrations higher than pretest, and handlers having higher cortisol concentrations than controls. There was no interaction between cortisol sampling time and testing status. Planned contrasts revealed higher posttest cortisol concentrations compared with pretest concentrations for both handlers, F_(1,20) = 33.35, p < 0.001, η² = 0.63, and controls, F_(1,20) = 7.53, p = 0.013, η² = 0.27 (Figure 2A). Importantly, handlers had higher posttest cortisol concentrations than controls, F_(1,20) = 7.56, p = 0.012, η² = 0.27 (all contrasts Bonferroni adjusted), with no difference for pretest cortisol concentrations between handlers and controls (Figure 2A).

Figure 2  Pretest and posttest salivary (A) cortisol concentrations (nmol/L) and (B) testosterone concentrations (pg/L) of handlers (black bars; N = 16) and controls (gray bars; N = 6). Data represent means ± SEM (SE of mean). Asterisks represent statistically significant differences between groups as shown by Bonferroni-adjusted post hoc tests following ANOVA; ***p ≤ 0.001; *p < 0.05.

Testosterone analyses

For differences in testosterone levels between handlers and controls, the analysis “testosterone = Sampling Time (pretest, posttest) × Testing Status (handler, control)” showed no main effects of sampling time or testing status on testosterone concentrations. There was an interaction between sampling time and testing status, F_(1,20) = 13.06, p = 0.002, η² = 0.40. When compared with pretest testosterone concentrations, posttest testosterone concentrations were increased in handlers (Figure 2B). Interaction contrasts revealed a difference between pretest and posttest testosterone concentrations for handlers only, F_(1,20) = 14.68, p = 0.001, η² = 0.42 (Figure 2B), and differences between handler and control posttest testosterone concentrations approaching significance, F_(1,20) = 4.274, p = 0.052, η² = 0.18 (all contrasts Bonferroni adjusted; Figure 2B).

Men have higher testosterone levels than women, and because men are more likely to show larger fluctuations in testosterone than women, we analyzed men as a separate group. As there were only two women in the control group, statistics comparing testosterone differences between women handlers and controls were not performed. When considering men only, findings remained the same (data not shown) except that the difference between handler and control posttest testosterone concentrations was significant, F_(1,12) = 9.228, p = 0.01, η² = 0.44 (Bonferroni adjusted).

Subjective stress

When considering subjective stress ratings within the handler and control groups, handlers rated their own pretest stress as greater than their dogs' pretest stress, W(Z = − 2.673, p = 0.008) (Table I), and considered their dogs to be more stressed after certification than before, W(Z = − 2.411, p = 0.01) (Table I). Comparing across handlers and controls, subjective posttest stress ratings by handlers of their own stress were higher than controls, U(Z = − 2.195, p = 0.04) (Table I). Overall, handlers considered stress to be increased posttest compared with pretest for both themselves and their dogs, while the reverse was true for controls, although not significantly (Table I).

Table I.  Pretest and posttest subjective stress ratings by handlers and controls for themselves and their dogs.

					Percentiles
	Mean	Std. deviation	Minimum	Maximum	25th	50th (Median)	75th
Controls (n = 6)
Pretest control	1.67	0.516	1	2	1.00	2.00	2.00
Posttest control*	1.17	0.408	1	2	1.00	1.00	1.25
Pretest dog	1.50	0.837	1	3	1.00	1.00	2.25
Posttest dog	1.33	0.516	1	2	1.00	1.00	2.00
Handlers (n = 16)
Pretest handler**	2.06	0.680	1	4	2.00	2.00	2.00
Posttest handler*	2.31	1.250	1	5	1.00	2.00	3.00
Pretest dog***	1.38	0.719	1	3	1.00	1.00	1.75
Posttest dog**	2.19	1.223	1	5	1.00	2.00	3.00

Note. Significant comparisons: *p = 0.04; **p = 0.008; ***p = 0.01.

Correlations

Exploratory analyses within the handler group considered whether pretest and posttest cortisol and testosterone concentrations correlated with each other or with subjective stress ratings by handlers for themselves and their dogs. Within the handler group, there were no correlations between pretest and posttest cortisol concentrations. However, handler pretest and posttest testosterone concentrations were positively correlated, Pearson's r(16) = 0.74, p = 0.001 (Table II). When considering testosterone concentrations for male and female handlers separately, only the correlation for males remained significant, Pearson's r(10) = 0.73, p = 0.02. Higher posttest cortisol concentrations were also associated with higher posttest dog-subjective stress ratings, Spearman's ρ(16) = 0.55, p = 0.03 (Table III). Similarly, higher pretest–posttest cortisol concentration changes were also associated with higher pretest–posttest dog-subjective stress rating changes, Spearman's ρ(16) = 0.72, p = 0.002. Using Kendall's τ-b to examine correlations in handler- and dog-subjective stress ratings, handler posttest subjective stress ratings were positively associated with dog posttest subjective stress ratings, r(16) = 0.56, p = 0.009 (Table IV).

Table II.  Pearson correlations between cortisol and testosterone values within the Handler group (n = 16).

	Pretest cortisol	Posttest cortisol	Pretest testosterone	Posttest testosterone
Pretest cortisol	1	0.190	− 0.094	0.021
Posttest cortisol		1	− 0.263	0.021
Pretest testosterone			1	0.799*
Posttest testosterone				1

*Correlation is significant at the 0.01 level (two tailed).

Table III.  Spearman correlations between cortisol/testosterone concentrations and subjective stress ratings for handlers and dogs within the handler group (n = 16).

Pretest/Posttest subjective stress ratings	Pretest cortisol	Posttest cortisol	Pretest testosterone	Posttest testosterone
Pretest subjective stress ratings: Handlers	0.389	0.408	0.151	0.420
Pretest subjective stress ratings: Dogs	− 0.356	− 0.217	0.151	0.306
Posttest subjective stress ratings: Handlers	− 0.444	0.282	− 0.207	0.170
Posttest subjective stress ratings: Dogs	− 0.138	0.548*	− 0.063	0.127

*Correlation is significant at the 0.05 level (two tailed).

Table IV.  Kendall τ-b correlations between subjective stress ratings for handlers and dogs within the handler group (n = 16).

Subjective stress ratings	Pretest handlers	Pretest dogs	Posttest handlers	Posttest dogs
Pretest handlers	1	0.286	0.140	0.299
Pretest dogs		1	0.297	0.345
Posttest handlers			1	0.561*
Posttest dogs				1

*Correlation is significant at the 0.01 level (two tailed).

Handler and dog posttest responses were also related. Handler posttest testosterone concentrations were negatively correlated with dog posttest pulse rates (n = 15, mean = 139 beats per minute (bpm), median = 140 bpm, range = 110–170 bpm), Pearson's r(15) = − 0.53, p = 0.04; higher posttest testosterone was associated with lower posttest dog pulse rates. There was no correlation between posttest testosterone and dog posttest pulse rate when controlling for gender, and there were no mean differences in dog posttest pulse rates for male compared with female handlers. In contrast, handler posttest cortisol concentrations were positively correlated with dog posttest pulse rates, r(15) = 0.55, p = 0.04 and dog posttest temperatures (n = 15, mean = 39.4°C, median = 39.4°C, range = 38.3–40.4°C), r(15) = 0.53, p = 0.04. There were no significant correlations between dog posttest temperature/pulse measures and subjective stress ratings.

Handler cortisol: Pass vs. fail; occupation

When considering whether there were differences in posttest salivary cortisol concentrations between handlers who passed or failed the certification, handlers who failed had higher posttest cortisol concentrations compared with handlers who passed, t(14) = − 2.621, p = 0.02 (Figure 3A). There was no difference in pretest cortisol concentrations between handlers who passed and those who failed. There was no effect of passing or failing on dog posttest temperature or pulse measures. Finally, posttest cortisol concentrations in handlers whose occupation was listed as “firefighter” were lower than those in handlers whose occupation was listed as anything other than “firefighter”, t(14) = − 2.517, p = 0.03 (Figure 3B). There was no difference in pretest cortisol concentrations between these groups. In this sample, being a firefighter was not statistically related to passing or failing, X² = 0.76, p = 0.4.

Figure 3  Posttest salivary cortisol concentrations (nmol/L) of handlers who (A) passed (black bars; N = 12) or failed (gray bars; N = 4) certification and (B) not firefighters (black bar; N = 9) or firefighters (gray bar; N = 7). Data represent means ± SEM (SE of mean). Asterisks represent statistically significant differences between groups as shown by student's t-tests; *p < 0.05.

Discussion

This study confirmed that a certification event for disaster dog handlers represents an acute naturalistic stressor, and supports our hypothesis that posttest cortisol concentrations in handlers were differentially increased compared with controls. Moreover, handlers were subjectively cognizant of their increased stress levels, and associated their own increased stress with subjective ratings of increased stress in their dogs. Indeed, higher posttest handler salivary cortisol concentrations were associated with higher posttest dog pulse and temperature measures. In addition, handlers who failed to certify demonstrated enhanced salivary cortisol responses. Cortisol responses in handlers listing their occupation as “firefighter” were attenuated regardless of pass/fail outcome.

Although more difficult to interpret because of higher levels and more fluctuation in testosterone levels in men than in women, higher posttest testosterone concentrations in handlers than controls might be associated with search performance and handler/dog interactions. Changes of testosterone have previously been identified in competitive environments. These changes have primarily been associated with winning or losing, and changes in handler testosterone level were shown to affect their dogs' performances (Mehta and Josephs 2006; Mehta et al. 2008). However, increases in testosterone in a competitive environment were considered important and decreases in testosterone when the competitive stressor was considered non-important have also been noted (Elloumi et al. 2008). Because certification can be perceived as a form of winning, its perceived importance might generate a testosterone response in addition to a cortisol response, and produce downstream effects on performance. The present study's testosterone findings resembled previous findings implicating effects of perceived importance of a stressor on direction of testosterone change (Chatterton et al. 1997). In that study, testosterone increased from basal concentrations in handlers attempting to certify but decreased in controls for whom the activity was essentially practice.

Handlers appeared aware of their increased stress compared with controls, as reflected in their subjective posttest stress ratings. However, it was intriguing that posttest cortisol level in handlers was correlated with subjective ratings of their dogs' stress, yet not with their own subjective stress ratings. The reasons for this are unclear, and may indicate that either this correlation represents a statistical anomaly, handlers' subjective ratings of stress do not accurately reflect a level of stress suggested by their cortisol levels, handlers' subjective stress immediately following completion of certification drops before a corresponding decrease in cortisol levels, or that handlers' cortisol levels do not represent an accurate measure of subjective stress under these conditions.

Handlers' higher posttest subjective stress ratings for their dogs, together with posttest handler hormonal concentration association with pulse rates and temperatures in dogs, suggest that additional investigation into possible interactions is warranted. Based on handlers' higher posttest subjective stress ratings for their dogs, further insight into interactive stress between handlers and dogs may be obtained using cortisol, urinary homovanillic acid, and blood glucose concentrations as well as core temperature testing, obtained concomitantly with handler saliva samples at multiple timepoints immediately prior to and subsequent to certification testing. Additionally, based on post-stress ratings by handlers, comparison of these parameters in training, certification, and deployment would be important to thoroughly investigate the interaction between handler and dog stress and effects on performance. It is impossible to determine from the current dataset if handler–dog interactions underlie these associations, or whether there might be additional factors responsible for coordinated physiological changes in both team members.

One unexpected finding was the elevated cortisol level in handlers who failed the certification. Because handlers are not told upon completion how many victims were successfully located or whether their dogs issued a false alert, it is unclear why their cortisol concentrations were significantly higher than those whose dogs successfully passed the certification. It is also unclear why, though dog temperature and pulse were correlated with handler cortisol levels, there was no effect of passing or failing on dog temperature and pulse. It is possible that either these findings are by chance in a small sample, the timing of temperature and pulse rate for dogs was too variable to identify an effect, or numbers of pass versus fail lacked sufficient power. It is also possible that overall physiological response to certification stress affected handlers' working interactions with their dogs, although it is not clear whether these interactions would be the same as previously described in agility dogs (Jones and Josephs 2006; Mehta et al. 2008). Alternatively, it is possible that these handlers suspected a failing performance based on intangible and/or subtle actions of their dogs.

In contrast, lower cortisol concentrations for firefighters compared with non-firefighters were not completely unexpected. We believe that this is the first published report of differential cortisol response to an acute naturalistic stressor in firefighters compared with non-firefighters. This prediction was reasonable based on reports of lower than expected cortisol response to an experimental stressor in firefighters (Roy 2004) and cortisol response to work stress (Maina et al. 2009). Previous exposure as a moderator of stress has also been noted in naturalistic stressors such as parachute jumping (Deinzer et al. 1997) and helicopter underwater evacuation training (Robinson et al. 2008), where stress responses are attenuated by experience with the stressor (Deinzer et al. 1997). Reduced subjective stress levels have been reported in FEMA-trained dog handlers following deployment compared with non-FEMA-trained handlers (Alvarez and Hunt 2005). However, approximately 40% of FEMA dog handlers are professional firefighters; therefore, it is possible that relative to the stresses involved in firefighting, deployment is not considered a stressor. Our finding, together with the substantial representation of firefighters within the FEMA dog handler rosters, suggests that previous work identifying lower self-reports of stress in FEMA-trained dog handlers might be a consequence of FEMA handlers also being professional firefighters.

There are a number of limitations to note for this study. First, because this was a convenience sample, we were limited in our opportunity to control for potential confounding variables. We acknowledge that there were fewer controls than participants, control participants were not explicitly matched to handlers on experience, and that the small number of handler–dog groups failing might limit generalization of our findings without future replication.

We could not control for food or caffeine consumption prior to saliva collection. While saliva collection would have been optimal immediately prior to each handler's certification attempt, we did not feel it prudent to interfere with handlers and their dogs at those timepoints. However, for both handlers and controls, pretest saliva was collected at the same times, and posttest samples reflected the same time range. Though a potential effect of day was not controlled, as performance data for controls were collected on the day prior to the certification testing, the lack of significant differences in cortisol concentrations for handlers across Saturday and Sunday suggests that day of collection did not represent a significant confound.

Because the differences for both handlers and controls between pretest and posttest cortisol concentrations may reflect diurnal fluctuation, our initial predictions and subsequent conclusions were not based on this difference. Our comparison of interest was centered on whether posttest concentrations would be different when comparing handlers and controls. Indeed, the difference in posttest concentrations between handlers and controls confirmed our prediction of increased cortisol level in handlers.

It is possible that physical factors such as differences in physical exertion and physical fitness might have accounted for differences in posttest cortisol and/or testosterone concentrations between handlers and controls. However, this is unlikely, as controls completed the same searches as the handlers. It is also possible that findings may reflect social evaluation associated with being observed. Further investigation would be required to rigorously examine this possibility.

Although we collected limited data on the dogs, and did not have temperature or pulse data for the control dogs, we speculate that the association of posttest dog pulse and temperature with handler cortisol level suggests that at least some variation in dog physiological response arises from dog–handler interaction. Although the small sample size and the large number of analyses created a risk of overfitting these data, and the small sample size precluded the development of regression models, the large effects obtained after Bonferroni adjustments help to support our findings and provide direction for future research efforts.

In summary, these findings demonstrate that certification testing affects both cortisol and testosterone responses in dog handlers. Cortisol increases were enhanced in subjects who were unsuccessful in their certification attempt. Finally, the cortisol response was lower in dog handlers listing “firefighter” as their primary occupation. These findings demonstrate that certification testing offers a controlled environment for targeted evaluation of effects of an acute naturalistic stressor on disaster dog handlers, and the real-world nature of this study may generally serve as a model paradigm for experimental studies within the field of stress and traumatic stress.

Acknowledgments

We would like to acknowledge the handlers and their dogs who agreed to participate in this research. We are grateful to the hosting team and the FEMA people who helped in the data collection process. L. Lit would like to thank John Dean of Arizona Search Dogs and Teresa MacPherson, Chairperson, FEMA Canine Subgroup, for assistance in manuscript preparation. We also thank the private donor whose contribution made this project possible. Finally, we appreciate the assistance and suggestions offered by reviewers to improve this manuscript.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.