Veterinary and Animal Science 9 (2020) 100120

Contents lists available at ScienceDirect

Veterinary and Animal Science

journal homepage: www.elsevier.com/locate/vas

Comparison between rectal and body surface temperature in dogs by the T

calibrated infrared thermometer
Blaž Cugmasa, , Primož Šušteričb, Nina Ružić Gorenjecc, Tanja Plavecd,e
⁎

a
Biophotonics Laboratory, Institute of Atomic Physics and Spectroscopy, University of Latvia, 19 Raina Blvd., LV-1586 Riga, Latvia
b
Primavet Veterinary Clinic, 38 Ptujska Rd, SI-2327 Rače, Slovenia
c
Institute for Biostatistics and Medical Informatics, Faculty of Medicine, University of Ljubljana, 2 Vrazov Sq., SI-1000 Ljubljana, Slovenia
d
Small Animal Clinic, Veterinary Faculty, University of Ljubljana, 60 Gerbiceva Str., SI-1000 Ljubljana, Slovenia
e
Small Animal Veterinary Hospital Hofheim, 7 Katharina-Kemmpler St., DE-65719 Hofheim, Germany

ARTICLE INFO ABSTRACT

Keywords: Because dogs tolerate conventional rectal temperature measurements poorly, a calibrated infrared thermometer
Body surface temperature was tested for assessing canine body surface temperature. Body surface temperature of 204 dogs was estimated
Dog on various sites (digit, snout, axilla, eye, gum, inguinal region, and anal verge). Having rectal temperature as the
Infrared thermometer gold standard, temperature difference, Spearman's correlation coefficient, hyperthermia and hypothermia de-
Health status
tection sensitivity and specificity, and stress response score was calculated for each measurement site. Although
Rectal temperature
the canine body surface temperature was considerably lower than the rectal temperature, there was a moderate
correlation between both temperatures. Spearman's coefficients were 0.60 (p < 0.001) for the inguinal region
with a single operator and 0.50 (p < 0.001) for the gum with multiple operators. Measurement site on the gum
additionally guaranteed hyperthermia detection sensitivity and specificity up to 90.0% (95% CI: [66.7 100]) and
78.6% (95% CI: [71.6 85.2]), respectively. Measurements with the infrared thermometer provoked a statistically
significant lower stress response (mean stress scores between 1.89 and 2.48/5) compared to the contact rectal
measurements (stress score of 3.06/5). To conclude, the correct body surface temperature measurement should
include a calibrated thermometer, reliable sampling, and the control of external factors such as ambient tem-
perature influence. The transformation of body surface temperature to the recognized rectal temperature in-
terval allows more straightforward data interpretation. The gum temperature exhibited the best clinical potential
since the differences to rectal temperatures were below 1°C, and hyperthermia was detected with the sensitivity
of up to 90%.

Introduction Carter, 2017; Kreissl and Neiger, 2015). Gomart et al. (Gomart et al.,
2014) showed that the RT compared to axillar or auricular temperature
Body temperature measurement is an essential part of the clinical measurement provoked more extensive animal defensive behaviour,
examination in dogs since deviations from the normal temperature (due which can affect various physiological variables like blood pressure and
to infection or shock for example) significantly influence clinical deci- pulse rate (Bragg et al., 2015). Furthermore, the rhythmicity of body
sions (Gomart et al., 2014; Smith et al., 2015). Because invasive core temperature is an important physiological process, which depends on
body temperature measurements cannot be performed in clinical set- body size (Piccione et al., 2009) and location (Giannetto et al., 2015).
tings (Greer et al., 2007; Hayes et al., 1996), the clinicians opt for Ideally, an alternative temperature measurement approach should be
measuring rectal temperature (RT) with a digital contact thermometer accurate but less stressful (Sousa, 2016).
(DCT). Although RT is in close agreement with the core body tem- Because of the RT measurement disadvantages mentioned above,
perature, RT is slightly lower and often lags after body temperature the correlation and agreement between rectal and better-tolerated body
changes due to intestinal air, faeces, and masses (Greer et al., 2007; surface temperature measurements are actively investigated
Kreissl and Neiger, 2015; Sousa, 2016). Importantly, RT measurements (Kunkle et al., 2004). Taking auricular temperature by a non-contact
are often poorly tolerated, especially by fractious patients and patients infrared thermometer (IRT) is common due to the tympanic membrane
with recto-anal and pelvic conditions (Gomart et al., 2014; Hall and potential to reflect core body temperature (Lamb and McBrearty, 2013;

⁎
Corresponding author.
E-mail address: [email protected] (B. Cugmas).

https://doi.org/10.1016/j.vas.2020.100120
Received 15 April 2020; Received in revised form 12 May 2020; Accepted 14 May 2020
Available online 16 May 2020
2451-943X/ © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
B. Cugmas, et al. Veterinary and Animal Science 9 (2020) 100120

Smith et al., 2015). However, the correlation between auricular and RT Table 1
ranged from high (González et al., 2002; Hall and Carter, 2017; Description of measurement sites with the corresponding thermometer.
Southward et al., 2006; Zanghi, 2016) to poor (Cichocki et al., 2017; Short name Precise location Thermometer
Greer et al., 2007; Konietschke et al., 2014; Sousa et al., 2011). It seems
that the correct thermometer placement is crucial to obtain real body Digit* the second digit, lateral side, left forelimb infrared (IRT)
Snout* rostral plane of the snout
temperature. Inappropriate IRT positioning probably samples cooler ear
Axilla* left side, fold between forelimb and body
canal wall, resulting in anomalous tympanic membrane temperature Eye left eye, the pupil was aimed
reading. Similarly, conflicting results were reported when ocular and Gum left side, gum above the canine tooth
axillar temperature was studied (Gomart et al., 2014; E. Hall et al., Inguinal region left side, fold between hind limb and body
2019; Kreissl and Neiger, 2015; Rizzo et al., 2017; Sousa, 2016; Anus* anal verge
Rectum (RT) rectal contact (DCT)
Zanghi, 2016).
Certainly, conflicting results cannot be attributed only to the ther- ⁎
- investigated only in the hospital stage
mometer operators. Pušnik and Drnovšek (Pušnik and Drnovšek, 2005)
showed that uncalibrated IRTs could cause faulty temperature readings strictly followed (Rizzo et al., 2017). First, infrared measurements in
with the error up to 3°C. The existing studies on comparing canine body random order were conducted on eight body sites. Finally, the rectal
surface and RT employed various, even human, contact (DCT) and in- temperature was estimated by DCT (DT-10, rigid rectal thermometer,
frared thermometers (IRT), which did not offer the possibility to be Advanced Monitors Corporation, San Diego, USA). For each measure-
clinically calibrated. Therefore, the retrieved thermometric data was ment, the ambient temperature was recorded.
evaluated in the incomparable temperature ranges, leading to con- (2) Clinical stage (realized with multiple operators). Temperature
flicting conclusions. measurements were taken from 153 dogs (84 males, 69 females) – 71 at
In this study, due to the poor canine tolerance towards conventional private clinics, 82 at animal shelters. Ninety-three dogs were mongrels;
rectal temperature measurements, we wanted to test a calibrated in- the rest were purebred. The average age was 3.2 years (full range from
frared thermometer for assessing body surface temperature on several 1.5 months to 13 years). The number of young dogs (below 1 y) was 41.
body sites, which could be easily used in the clinic. First, we had built a For each measurement site separately, three dogs were excluded from
custom-made IRT, which was calibrated in an expected canine tem- further analysis because either the infrared or rectal measurement was
perature range. Measurements were performed on several easily ac- missing. Multiple veterinarians performed measurements during a
cessible and well-tolerated body sites, which have potential as an al- regular clinical examination right after the animals were admitted to
ternative temperature measurement for clinicians and animal owners the facility. Therefore, acclimatization to the room temperature was not
(Hall and Carter, 2017). To access the clinical value of our results, RT guaranteed. Infrared measurements on the eye, the gum, and the in-
additionally served for the calculation of hyperthermia and hy- guinal region were taken randomly. Finally, RT was acquired. For each
pothermia detection sensitivity and specificity. Finally, the animal measurement, animal hairiness, body condition, body mass, and am-
stress response to the IRT and rectal measurement was estimated. bient temperature were recorded.
For non-contact measurements, we built a custom-made IRT
Materials and methods (Fig. 2). IRT was based on a miniature infrared sensor (MLX90615,
Melexis, Ypern, Belgium), which ensured the medical accuracy of 0.2°C
Animals and measurements in a limited temperature range. To guarantee actual temperature va-
lues, we calibrated IRT in the range between 28 and 41°C (9 evenly
We conducted a prospective multicenter and multi-operator study, distributed measurements) with the black body calibrator
which was approved by the Administration of the Republic of Slovenia (Miklavec et al., 2013). The thermometer additionally included a tiny
for Food Safety, Veterinary Sector and Plant Protection. Additionally, aluminium tube (diameter = 1.5 cm, length = 2.1 cm), which was
the owner's written permission was collected. The study evolved in two placed above the sensor in order to reduce the sensor's field of view
stages: (FOV) and to remove possible ambient IR sources. To minimize material
(1) Hospital stage (realized at the University's Small animal clinic emissivity, the tube was additionally polished. During the measure-
with a single operator) served to identify normal temperature ranges ments, the IRT thermometer was held approximately 0.5 cm away from
and correlations to RT of several measurement sites (Fig. 1, Table 1). the body surface, which resulted in 2.7 cm2 of the circular sampling
The leading surgeon took temperature readings of 51 dogs (21 males, area. IRT ran on the open-source Arduino platform. Rectal temperature
30 females) before, during, or after the anesthesia for the scheduled
surgeries. There were 46 purebred dogs. Out of 29 breeds, golden re-
trievers (n = 5), German shepherds (4), and Border collies (3) were the
most common. Dogs’ age ranged from 3 months to 15 years, with a
median of 7.25 years. The number of young dogs (below 1 y) was 8. At
least one hour-long acclimatization to the room temperature was

Figure 2. We used a custom-made infrared thermometer (IRT) calibrated in the

Figure 1. Measurement sites (*investigated only in the hospital stage). temperature range between 28 and 41°C.

2
B. Cugmas, et al. Veterinary and Animal Science 9 (2020) 100120

Table 2
Stress score criteria.
Scores Criteria

1 Dog is unconcerned by the measurement

2 Dog is attentive to the measurement but does not show any anxiety and remains still
3 Dog exhibits nervousness but does not try to slip away
4 Dog tries to slip away, and in order to perform a measurement, assistance is needed
5 Dog growls or is aggressive with attempts of biting or is desperate to escape, restraint is needed to measure temperature

Table 3 100,000 repetitions. We analyzed stress scores separately for dogs from
Hospital stage. Temperature mean value, standard deviation (STD), median, 5th the shelters and the private clinic. When comparing all measurement
and 95th percentile (in square brackets), and Spearman's correlation coefficient sites, we used repeated-measures ANOVA since the distributions of
(with 95% confidence intervals in square brackets) against rectal measurements differences between separate measurement sites were symmetric. For
(Corr. with RT). posthoc comparisons between two measurement sites, paired t-tests and
Measurement site Mean (STD) Median Corr. with RT adjusted p values with Holm's method were applied. The difference was
considered statistically significant if p < 0.05. The post-hoc power
Digit 30.43 (2.79) 29.85 [26.19, 34.80] -0.28 [-0.51, -0.01]
analyses were performed for t-tests at α=0.05. In addition, we reported
Snout 31.20 (3.02) 31.46 [26.54, 35.71] -0.34 [-0.55, -0.09]
Axilla 33.66 (1.96) 33.98 [30.37, 36.17] 0.10 [-0.15, 0.35]
the expected sample size at 80% power for each variable, assuming that
Eye 34.54 (0.81) 34.61 [33.07, 35.88] 0.46 [0.22, 0.67] the means and standard deviations from our sample reflect the true
Gum 35.06 (0.97) 35.09 [33.64, 36.32] 0.38 [0.15, 0.56] population. We processed the results in the Matlab environment
Inguinal region 35.40 (1.14) 35.52 [33.29, 36.98] 0.60 [0.41, 0.74] (R2016a, MathWorks, Natick, USA).
Anal verge 35.63 (1.18) 35.57 [34.09, 37.61] 0.43 [0.16, 0.67]
Rectum (RT) 38.1 (1.0) 38.30 [36.06, 39.30] /
Results

was acquired by a conventional electronic DCT. First, the thermometer Hospital stage (single operator)
was lubricated and then inserted approximately 1.5 cm into the rectum.
During the measurement, the thermometer was gently pressed against We measured the body surface and rectal temperature (RT) of 51
the rectal wall. In the end, animal body condition and measurement site dogs (Table 3 and Fig. 3). Rectal temperature (RT) had the highest
hairiness were scored with grades from 1 to 5. Age, body mass, and mean value (38.0°C). Oppositely, the mean temperature of the digit and
ambient temperature were obtained additionally. the snout was just 30.50°C and 31.41°C, respectively. Standard devia-
tion below 1.3°C appeared on the eye, on the gum, in the inguinal re-
Stress assessment gion, on the anal verge, and in the rectum. Greater variability was seen
when temperature was measured on the digit and snout. Inguinal
According to animal defensive behaviour, the leading clinician es- temperature exhibited the best correlation with rectal measurements;
timated each measurement acceptance with so-called stress score the correlation coefficient was 0.60. Finally, we did not discover any
(Gomart et al., 2014; Kreissl and Neiger, 2015). Scores were given in significant correlation between the body temperature and age or am-
steps of 0.5 in the range between 1 and 5 (Table 2). Stress scores were bient temperature (all correlation coefficients were below 0.29, all p >
assessed in the clinical stage for all dogs from shelters (82) and 44 dogs 0.09). Additionally, there was no gender-related impact on correlation
from the private clinic. For the rest of the animals, the clinicians did not coefficients between body and rectal temperature.
provide data. Fig. 4 shows the relationship between inguinal (InT) and rectal
temperature (RT). Fig. 4(a) additionally includes a fitted linear function

Statistical analysis

The results were summarized with means, standard deviations

(STD), medians, and percentiles. First, the monotonic relationship be-
tween RT and the infrared temperature was assessed with Spearman's
correlation coefficient. We additionally tested a possible impact of
gender on the correlation between body and rectal temperature, and
between body temperature and external factors (hairiness, body con-
dition, ambient temperature, age, and body mass). To assess the
agreement between rectal and IR temperature, Bland-Altman (BA) plots
were used (Martin Bland and Altman, 1986). In order to operate with
common temperature values and offer comparable results, the infrared
temperatures were additionally transformed (based on linear regres-
sion) to the RT temperature range. For all BA plots, we plotted 95%
limits of agreement, i.e., mean ± 1.95•STD.
Three different rectal temperature thresholds were used for hy-
perthermia (39.0, 39.3, and 39.6°C) and two for hypothermia (37.7 and
38.0°C) (Couto, 2009; Lorenz, 1995; Refinetti and Piccione, 2003). For
each threshold, we calculated cut-off temperature in the specific in- Figure 3. Temperature ranges as boxplots for all measurement sites from the
frared range (optimizing the geometric mean between sensitivity and hospital stage. The central mark (red) represents median, the bottom and top
specificity), detection sensitivity and specificity, and area under re- box edges indicate the 25th and 75th percentiles (blue), respectively. Whiskers
ceiver operating characteristic curve (AUC, ROC). 95% confidence in- extend to the most extreme data point, which is no more than 1.5 times the
tervals (CI) were estimated based on a bootstrap approach with interquartile range from the box. Outliers are presented as red crosses.

3
B. Cugmas, et al. Veterinary and Animal Science 9 (2020) 100120

Figure 4. Hospital stage with a single operator. (a) Rectal (RT) and corresponding inguinal temperature (InT) with a fitted linear regression line (black line) and 95%
prediction intervals (red lines). Bland-Altman plots (Δ = RT – InT) (b) before and (c) after the transformation of InT to RT range based on the fitted linear function
(Fig. 4 (a)). Mean difference (mean) and 95% limits of agreement are marked as black and red lines, respectively.

Table 4
Clinical stage. Temperature mean value, standard deviation (STD), median, 5th and 95th percentile (in squared brackets) and Spearman's correlation coefficient and
p value (with 95% confidence intervals in square brackets) against rectal measurements (Corr. with RT) and external factors (Corr. with EF).
Measurement site Mean ± STD Median Corr. with RT Corr. with EF*

Eye 32.30 (1.52) 32.39 [29.62, 34.53] 0.17, p = 0.04 [0.00, 0.32] /
Gum 33.20 (2.12) 33.70 [29.39, 35.97] 0.50, p < 0.001 [0.37, 0.61] /
Inguinal region 34.58 (1.91) 35.02 [31.19, 37.08] 0.13, p = 0.11 [-0.04, 0.29] -0.45a
-0.51b
-0.38c
Rectum (RT) 38.62 (0.65) 38.60 [37.60, 39.70] / 0.35d

External factors are denoted with the following letters.

a
– hairiness (p < 0.001)
b
– body condition (p < 0.001)
c
– ambient temperature (p < 0.001)
d
– body mass (p = 0.96).
⁎
Correlation coefficients against external factors with values between -0.30 and 0.30 are not listed.

between InT and RT. The original data is presented as a Bland-Altman

(BA) plot in Fig. 4(b). The mean difference between InT and RT was
2.68°C. In the end, we transformed the inguinal temperature into the RT
range (Fig. 4(c)) based on the linear function from Fig. 4(a). The
transformation lowered the average difference to zero (95% limits of
agreement [-1.67, 1.67]), and most of the values were well under 1°C.
Additionally, 40% of differences were below 0.5°C.
In the clinical stage, only the best-performing hospital measurement
sites were kept: eye, gum, and inguinal region. The choice was based on
the temperature variability (i.e., STD) and the correlation with RT
(Table 3).

Clinical stage (multiple operators)

Table 4 and Fig. 5 list the temperatures from the 153 dogs involved.
Compared to the hospital stage, mean temperatures of the eye, gum,
and inguinal region were lower (Table 4). Conversely, the mean RT was
Figure 5. Temperature ranges as boxplots from the clinical stage. The central slightly higher. Furthermore, correlation coefficients were lower for the
mark (red) represents median, the bottom and top box edges indicate the 25th eye and inguinal region (0.46/0.60 vs. 0.17/0.13), whereas the coef-
and 75th percentiles (blue), respectively. Whiskers extend to the most extreme ficient increased for the gum from 0.38 to 0.50, showing a moderate
data point. correlation. The negative correlation was additionally found between
inguinal temperature (InT) and hairiness, body condition, and ambient
temperature. On the other hand, RT correlated weakly with body mass

4
B. Cugmas, et al. Veterinary and Animal Science 9 (2020) 100120

Figure 6. Clinical stage with multiple operators. (a) Rectal (RT) and corresponding eye/ocular temperature (OT) with a fitted linear regression line (black line) and
95% prediction intervals (red lines). Bland-Altman plots (Δ = RT – OT) (b) before and (c) after the transformation of OT to RT range based on the fitted linear
function (Fig. 6 (a)). Mean difference (mean) and 95% limits of agreement are marked as black and red lines, respectively.

(Table 4). Animal age did not exhibit any correlation with the body or shelters). The mean score and p values of posthoc paired t-tests (i.e.,
rectal temperature. comparing two measurement sites), adjusted for multiple comparisons,
Figs. 6-8 show the relationship between infrared (eye, gum, and are listed in Table 6. The inguinal measurements were tolerated best,
inguinal region) and rectal temperature (RT). Fig. 6-8(a) include scatter mean stress score was 2.45 for the shelter dogs and 1.32 for the clinic
plots and fitted linear functions (black line), which served for the dogs, respectively. On the other hand, rectal measurements provoked
transformation of infrared temperatures to the RT range. The agree- the most stress – the mean score was around 3. The difference analysis
ment between infrared and RT is shown with Bland-Altman plots before is presented in Table 7.
(Figs. 6-8(b)) and after (Figs. 6-8(c)) temperature range transformation.
The purpose of the transformation is in 1) presenting data in the range Discussion
veterinarians are used to; 2) setting mean difference close to zero; and
3) removing trends of dependence between the infrared in RT (e.g., The mean rectal temperature (RT) of hospital animals was lower for
Fig. 7(b)). In Fig. 7(b), the mean difference was more than 5.4°C, with 0.62°C compared to the clinical stage. The reason for the difference
the 95 % limits of agreement between 1.7 and 9.1°C. After the trans- could lie in the fact that temperature measurement in approximately
formation (Fig. 7(c)), the mean was zero, and most of the differences one-third of hospital dogs took place under general anaesthesia, which
(91.3%, 137/150) between gum (GT) and RT were up to 1°C. can lower the body temperature (Rigotti et al., 2015). The rest of the
Table 5 lists the thresholds for clinical hyperthermia and hy- measurement sites in the hospital stage exhibited higher mean body
pothermia detection together with the corresponding sensitivities and surface temperature, probably due to the strict acclimatization proce-
specificities, all with 95% confidence intervals. The gum exhibited the dure followed. Conversely, winter months during which the clinical
best potential to correctly detect hyperthermia since the sensitivity was stage with mostly walk-in patients was executed, hindered a proper
up to 90.0 %. On the other hand, we could not detect hypothermia very temperature acclimatization process, typically taking up to one hour
well. (Rizzo et al., 2017). Our assumption was supported by the increased
The stress scores were statistically significantly different for mea- correlation between inguinal and ambient temperature in the clinical
surement sites (repeated measures ANOVA, p < 0.001 for the clinic and stage (Table 4).

Figure 7. Clinical stage with multiple operators. (a) Rectal (RT) and corresponding gum temperature (GT) with a fitted linear regression line (black line) and 95%
prediction intervals (red lines). Bland-Altman plots (Δ = RT – GT) (b) before and (c) after the transformation of GT to RT range based on the fitted linear function
(Fig. 7 (a)). Mean difference (mean) and 95% limits of agreement are marked as black and red lines, respectively.

5
B. Cugmas, et al. Veterinary and Animal Science 9 (2020) 100120

Table 5
Hyperthermia and hypothermia detection characteristics at three measurement sites from the clinical stage. The predetermined thresholds of rectal temperature:
39.6, 39.3, 39.0, 38.0, and 37.7°C, corresponding infrared cut-off temperature, sensitivity and specificity, and AUC with 95% confidence intervals (in squared
brackets) are listed.
Measurement Hyperth.(↑) Rectal (°C) Infrared (°C) Sensitivity (%) Specificity (%) AUC
site Hypoth. (↓)

Eye ↑ 39.6 33.3 [31.8, 34.4] 54.6 [22.2, 85.7] 74.8 [67.4, 81.9] 0.63 [0.43, 0.81]
39.3 32.6 [31.8, 34.2] 59.1 [37.5, 80.0] 57.0 [48.4, 65.6] 0.57 [0.43, 0.71]
39.0 32.6 [31.8, 33.0] 66.7 [53.0, 80.0] 64.7 [55.2, 73.9] 0.66 [0.56, 0.75]
↓ 38.0 32.0 [31.4, 33.5] 61.3 [43.5, 78.3] 41.2 [32.4, 50.0] 0.48 [0.37, 0.60]
37.7 32.6 [32.0, 33.6] 50.0 [20.0, 80.0] 52.2 [43.8, 60.6] 0.44 [0.24, 0.63]
Gum ↑ 39.6 34.7 [34.7, 35.1] 90.0 [66.7, 100] 78.6 [71.6, 85.2] 0.84 [0.70, 0.93]
39.3 34.7 [32.7, 35.1] 61.9 [40.8, 82.6] 79.8 [72.7, 86.5] 0.76 [0.65, 0.86]
39.0 34.0 [33.0, 34.7] 72.3 [59.0, 84.8] 68.0 [58.8, 76.9] 0.75 [0.66, 0.83]
↓ 38.0 33.3 [30.5, 35.2] 31.3 [15.6, 48.2] 33.9 [25.4, 42.5] 0.24 [0.15, 0.33]
37.7 32.8 [30.3, 33.7] 38.5 [11.8, 66.7] 32.9 [24.8, 40.6] 0.24 [0.15, 0.35]
Inguinal ↑ 39.6 34.6 [33.9, 36.4] 90.0 [66.7, 100] 42.9 [34.7, 51.1] 0.64 [0.47, 0.79]
39.3 34.6 [34.6, 36.0] 76.2 [56.3, 93.8] 43.4 [34.9, 51.9] 0.53 [0.40, 0.66]
39.0 34.7 [34.5, 35.9] 71.7 [58.1, 84.4] 50.0 [39.6, 59.1] 0.60 [0.50, 0.70]
↓ 38.0 35.7 [33.8, 35.8] 40.6 [23.5, 58.1] 67.8 [59.1, 76.0] 0.52 [0.41, 0.64]
37.7 35.3 [34.1, 37.1] 61.5 [33.3, 88.9] 62.0 [53.0, 69.6] 0.59 [0.40, 0.77]

Table 6
Stress scores from the clinical stage, separately for the shelters and clinic: mean (with standard deviation - STD) and p values from posthoc tests for comparisons of
stress scores evaluated on two different measurement sites.
M. site Shelters (n = 82) Clinic (n = 44)
Mean (STD) adjusted p values Mean (STD) adjusted p values

1: Eye 3.20 (0.72) p1-4 = 0.07 1.75 (0.92) p1-2 = 0.11

2: Gum 2.89 (0.67) p1-2, p1-3 < 0.01 1.95 (1.08) p1-3, p1-4 < 0.01
3: Inguinal 2.45 (0.59) p2-3, p2-4 < 0.01 1.32 (0.60) p2-3, p2-4 < 0.01
4: Rectum 3.37 (0.90) p3-4 < 0.01 2.75 (1.06) p3-4 < 0.01

The existing studies reported many discrepancies in differences and (Gomart et al., 2014). However, our study and the study of Mathis and
correlations between the body surface and RT. Since some measure- Campbell (2015) reported almost no correlation (ρ = 0.16 and 0.24).
ment sites are well vascularized and isolated (e.g., tympanic membrane, Regarding eye temperature, the reported correlations were moderate
axillar region), many authors expectedly reported small temperature (0.38-0.59) (Kreissl and Neiger, 2015; Rizzo et al., 2017; Yanmaz et al.,
differences to RT of up to 1.3°C (Cichocki et al., 2017; Goic et al., 2014; 2015), similar to the hospital stage of our study (0.50). However, the
Gomart et al., 2014; Greer et al., 2007; Hall and Carter, 2017; correlation in the clinical stage, which included multiple operators, was
Konietschke et al., 2014; Lamb and McBrearty, 2013; Mathis and weaker.
Campbell, 2015; Piccione et al., 2011; Sousa et al., 2011; According to the available data, we believe that conflicting results
Southward et al., 2006; Wiedemann et al., 2006; Yanmaz et al., 2015). can be attributed to the following factors:
On the other hand, our study recorded a higher difference between
axillar and RT of 4.32°C, similar to the study of Rizzo et al (2017). 1) Thermometers. The existing studies applied various contact (DCT) or
Significant temperature differences to RT were also reported on the eye infrared thermometers (IRT). DCT operates typically in the pre-
(our study: 3.41-6.30°C, Rizzo et al.: 6°C (Rizzo et al., 2017)), and in the dictive technique, which does not measure but only estimates the
digital (our study: 7.50°C) and inguinal region (our study: 2.68°C). The final temperature. On the other hand, uncalibrated IRT can exhibit
differences above are in the same range as the digit and torso tem- errors up to 3°C (Pušnik and Drnovšek, 2005). Therefore, many
perature of the human, where the values were 7.5°C and 4.9°C, re- commercial thermometers are specified only for a specific species
spectively (Taylor et al., 2014). Many discrepancies can also be found in and measurement site (e.g., the axillar temperature in humans). As
the reported correlation. The correlation between auricular and RT existing and our data show, thermometers should be calibrated for a
ranged from weak (0.3) to strong (0.89) (Gomart et al., 2014; specific temperature range and measurement site.
Konietschke et al., 2014; Wiedemann et al., 2006; Yanmaz et al., 2015). 2) Sampling. Since IRT samples a relatively small area, the operator can
A strong correlation (0.7) was found between axillar and RT easily acquire the temperature of (cooler) surrounding tissues.

Table 7
Mean and standard deviation (STD) and 95 % confidence intervals (CI) of stress score differences between rectal and infrared measurements. The post-hoc power
(α=0.05) and the expected sample size with a power of 80% are included.
Measurement site Shelters (n = 82) Clinic (n=44)
Mean (STD) 95% CI Power / Size Mean (STD) 95% CI Power / Size

Rectum – Eye 0.16 (0.80) -0.01 – 0.34 43% / 199 1.00 (1.01) 0.69 – 1.31 100% / 11
Rectum – Gum 0.48 (0.85) 0.29 – 0.66 100% / 27 0.80 (1.00) 0.49 – 1.10 100% / 15
Rectum – Inguinal 0.92 (0.91) 0.72 – 1.12 100% / 10 1.43 (1.11) 1.09 – 1.77 100% / 7

6
B. Cugmas, et al. Veterinary and Animal Science 9 (2020) 100120

Figure 8. Clinical stage with multiple operators. (a) Rectal (RT) and corresponding inguinal temperature (InT) with a fitted linear regression line (black line) and
95% prediction intervals (red lines). Bland-Altman plots (Δ = RT – InT) (b) before and (c) after the transformation of InT to RT range based on the fitted linear
function (Fig. 8 (a)). Mean difference (mean) and 95% limits of agreement are marked as black and red lines, respectively.

Fig. 3(a) displays two measurements on the anal verge with sig- hyperthermia as well; sensitivity and specificity were 61.9% and
nificantly lower temperatures. Both measurements probably include 79.8%, respectively. This fact can limit the usability of IRT to detect
the surrounding skin and hair. Furthermore, ocular measurements in fever, especially with the marginal cases, which should be followed by
conscious dogs can be challenging since animal movements can the standard method (RT). When it comes to hypothermia, IRT did not
increase the probability of false sampling. Stable and standardized prove very accurate. However, we should add that the recent work in
sampling is even more crucial if multiple operators are included in cats, horses, and rabbits (Gallego, 2017; Hall et al., 2019; Levy et al.,
the study. 2015) has suggested the lower limit for temperature ranges in healthy
3) External and body factors. Factors as ambient temperature, body animals. Therefore, 37.3°C would probably be a more reasonable hy-
mass, hairiness, gender, and age can crucially affect the difference pothermic limit, as applied in the study by Konietschke et al. (2014).
or correlation between rectal (RT) and body surface temperature However, in this case, the number of hypothermic samples would be
(Gomart et al., 2014). In dogs, the statistically significant correlation only four, preventing of doing any reliable detection analysis.
was found only between body surface temperature and mass, Our work additionally confirmed that RT measurements by DCT are
hairiness, and body condition (Lamb and McBrearty, 2013). The significantly more stressful than IRT. Lamb McBrearty showed that
same factors proved influential in our study with the addition of 45.5% of dogs needed additional restraint during rectal measurements
ambient temperature (Table 4). In both studies, increased body (Lamb and McBrearty, 2013). With measurements on the eye, this
condition score had a negative correlation coefficient. Surprisingly, percentage was only 20 %. In our study, RT measurements achieved the
the hairiness correlation was not consistent (positive vs. negative), average stress score of 3, statistically significantly more than IRT on the
which could be a consequence of different coat types (Kwon and inguinal (average score 1.32 for the clinic and 2.45 for the shelters),
Brundage, 2019) or ambient temperatures affecting the insulation gum (average score 1.95 for the clinic and 2.89 for the shelters), and
function of the coat. 1.75 for ocular measurements (only in the clinical setting, where pets
were better accustomed to human handling).
Comparing temperature differences directly does not contribute to
the analysis of agreement between IRT and DCT. If original data is
shown in Bland-Altman plots (Giavarina, 2015), substantial mean dif- Conclusions
ferences and apparent trends between temperatures can be noted (ex-
amples can be found in the previous studies (Greer et al., 2007; In this study, we investigated canine surface body temperature with
Kreissl and Neiger, 2015; Lamb and McBrearty, 2013) or Fig. 4(b), 6- an infrared thermometer (IRT), and we compared the results to the
8(b)). As suggested in our study, body surface temperature can be rectal temperature (RT). The transformation with the estimated re-
transformed into RT value range by, e.g., linear regression analysis. The gression linear function kept most of the differences between the gum
approach presents data in the RT range that veterinary clinicians and and rectal temperature (RT) below 1°C. What is more, hyperthermia
researchers are accustomed to, and the data is suitable to be drawn in (threshold value of 39.6°C) was detected quite well, with the sensitivity
the Bland-Altman plot. However, the “opposite” trend can be seen also up to 90 % (95% CI: [66.7, 100]), which is comparable to the axillar
after the transformation (Figs. 6-8(c)). It should be noted that this ap- region in humans. We also showed that RT measurements caused more
pears due to the fitting process. If a body surface temperature replaces stress than the approach with IRT.
the mean temperature on the x-axis, these trends disappear. We additionally discovered that due to different temperature ranges
For clinical medicine, the detection of hyper- or hypothermia is the and uncalibrated thermometers applied in the previous studies, a direct
most relevant. In humans, the hyperthermia detection sensitivity under comparison of results proved challenging (Sousa, 2016). We think that
the armpit (axillar region) can reach up to 87.5 %. In dogs, the axillar future research on new thermometers or measurement sites should
sensitivity was lower, i.e., 57 % (specificity = 100 %). On the other guarantee: (1) a calibrated thermometer, (2) stable conditions (e.g.,
hand, hypothermia was better detected (sensitivity = 86 %, specifi- fixed ambient temperature, the same measuring approach for multiple
city = 87 %) (Goic et al., 2014). Our study showed that only the mouth operators), (3) strict animal acclimatization process, which can take up
(i.e., gum) could be used for hyperthermia detection. When threshold to one hour. However, ensuring acclimatization could be challenging if/
temperature value was set to 39.6°C, up to 90 % (95% CI: [66.7, 100]) when novel methods are applied in clinical settings. Additionally, we
cases were discovered (specificity = 78.6 %, 95% CI: [71.6, 85.2]). could introduce better-designed IRT thermometers adjusted for the
With the lower threshold value (39.3°C), we could not detect specific veterinary use.

7
B. Cugmas, et al. Veterinary and Animal Science 9 (2020) 100120

Ethical statement Hall, E., Fleming, A., & Carter (Pullen), A. (2019). Investigating the use of non-contact
infrared thermometers in cats and dogs. Vet. Nurse, 10, 109–115.
Hall, E. J., & Carter, A. J. (2017). Comparison of rectal and tympanic membrane tem-
The authors ensure that the work described has been carried out perature in healthy exercising dogs. Comp. Exerc. Physiol. 13, 37–44.
according to the NC3Rs ARRIVE Guidelines. Our study was approved by Hall, E. J., Carter, A. J., Stevenson, A. G., & Hall, C. (2019). Establishing a Yard-Specific
the Administration of the Republic of Slovenia for Food Safety, Normal Rectal Temperature Reference Range for Horses. J. Equine Vet. Sci. 74, 51–55.
Hayes, J. K., Collette, D. J., Peters, J. L., & Smith, K. W. (1996). Monitoring body-core
Veterinary Sector and Plant Protection. We additionally obtained the temperature from the trachea: comparison between pulmonary artery, tympanic,
owner's agreement. esophageal, and rectal temperatures. J. Clin. Monit. 12, 261–269.
Konietschke, U., Kruse, B. D., Müller, R., Stockhaus, C., Hartmann, K., & Wehner, A.
(2014). Comparison of auricular and rectal temperature measurement in nor-
Declaration of Competing Interests mothermic, hypothermic, and hyperthermic dogs. Tierärztl. Prax. Kleintiere, 42,
13–19.
Kreissl, H., & Neiger, R. (2015). Measurement of body temperature in 300 dogs with a
The authors declare that they have no known competing financial
novel noncontact infrared thermometer on the cornea in comparison to a standard
interests or personal relationships that could have appeared to influ- rectal digital thermometer. J. Vet. Emerg. Crit. Care, 25, 372–378.
ence the work reported in this paper. Kunkle, G. A., Nicklin, C. F., & Sullivan-Tamboe, D. L. (2004). Comparison of Body
Temperature in Cats Using a Veterinary Infrared Thermometer and a Digital Rectal
Thermometer. J. Am. Anim. Hosp. Assoc. 40, 42–46.
Acknowledgements Kwon, C. J., & Brundage, C. M. (2019). Quantifying body surface temperature differences
in canine coat types using infrared thermography. J. Therm. Biol. 82, 18–22.
This study has received funding from the European Union's Horizon Lamb, V., & McBrearty, A. R. (2013). Comparison of rectal, tympanic membrane and
axillary temperature measurement methods in dogs. Vet. Rec. 173 524–524.
2020 research and innovation programme under the Marie Skłodowska- Levy, J. K., Nutt, K. R., & Tucker, S. J. (2015). Reference interval for rectal temperature in
Curie grant agreement No 745396 (DogSPEC), and from the Latvian healthy confined adult cats. J. Feline Med. Surg. 17, 950–952.
State Education Development Agency under the grant agreement Lorenz, M. (1995). Pyrexia (Fever). In M. Lorenz, & L. Cornelius (Eds.). Small Animal
Medical Diagnosis (pp. 15–22). Wiley.
1.1.1.2/VIAA/3/19/455 (GoBVM). We would also like to express our Mathis, J. C., & Campbell, V. L. (2015). Comparison of axillary and rectal temperatures
gratitude to veterinarians and all other personnel from Small Animal for healthy Beagles in a temperature- and humidity-controlled environment. Am. J.
Clinic (University of Ljubljana), Veterinary clinic Primavet (Rače) and Vet. Res. 76, 632–636.
Miklavec, A., Pušnik, I., Batagelj, V., & Drnovšek, J. (2013). A large aperture blackbody
animal shelters Meli, Turk, Mala Hiša, Maribor, Obalno Zavetišče and
bath for calibration of thermal imagers. Meas. Sci. Technol. 24, 025001.
Brežice. We thank prof. Dr. Igor Pušnik from the Laboratory of Piccione, G., Fazio, F., Giudice, E., & Refinetti, R. (2009). Body size and the daily rhythm
Metrology and Quality, University of Ljubljana, for the use of their of body temperature in dogs. J. Therm. Biol. 34, 171–175.
Piccione, G., Giannetto, C., Fazio, F., & Giudice, E. (2011). Accuracy of auricular tem-
equipment. There are no other conflicts of interest to declare.
perature determination as body temperature index and its daily rhythmicity in
healthy dog. Biol. Rhythm Res. 42, 437–443.
References Pušnik, I., & Drnovšek, J. (2005). Infrared ear thermometers—parameters influencing
their reading and accuracy. Physiol. Meas. 26, 1075–1084.
Refinetti, R., & Piccione, G. (2003). Daily Rhythmicity of Body Temperature in the Dog. J.
Bland, M. J., & Altman, D. G. (1986). Statistical methods for assessing agreement between Vet. Med. Sci. 65, 935–937.
two methods of clinical measurement. Lancet, 327, 307–310. Rigotti, C. F., Jolliffe, C. T., & Leece, E. A. (2015). Effect of prewarming on the body
Bragg, R. F., Bennett, J. S., Cummings, A., & Quimby, J. M. (2015). Evaluation of the temperature of small dogs undergoing inhalation anesthesia. J. Am. Vet. Med. Assoc.
effects of hospital visit stress on physiologic variables in dogs. J. Am. Vet. Med. Assoc. 247, 765–770.
246, 212–215. Rizzo, M., Arfuso, F., Alberghina, D., Giudice, E., Gianesella, M., & Piccione, G. (2017).
Cichocki, B., Dugat, D., & Payton, M. (2017). Agreement of Axillary and Auricular Monitoring changes in body surface temperature associated with treadmill exercise in
Temperature with Rectal Temperature in Systemically Healthy Dogs Undergoing dogs by use of infrared methodology. J. Therm. Biol. 69, 64–68.
Surgery. J. Am. Anim. Hosp. Assoc. 53, 291–296. Smith, V. A., Lamb, V., & McBrearty, A. R. (2015). Comparison of axillary, tympanic
Couto, C. (2009). Fever of undetermined origin. In R. Nelson, & C. Couto (Eds.). Small membrane and rectal temperature measurement in cats. J. Feline Med. Surg. 17,
Animal Internal Medicine (pp. 1274–1277). Mosby/Elsevier, St.Louis. 1028–1034.
Gallego, M. (2017). Laboratory reference intervals for systolic blood pressure, rectal Sousa, M. G. (2016). Measuring body temperature: how do different sites compare? Vet.
temperature, haematology, biochemistry and venous blood gas and electrolytes in Rec, 178, 190–191.
healthy pet rabbits. Open Vet. J. 7, 203–207. Sousa, M. G., Carareto, R., Pereira-Junior, V. A., & Aquino, M. C. (2011). Comparison
Giannetto, C., Fazio, F., Panzera, M., Alberghina, D., & Piccione, G. (2015). Comparison of between auricular and standard rectal thermometers for the measurement of body
rectal and vaginal temperature daily rhythm in dogs (Canis familiaris) under different temperature in dogs. Can. Vet. J. 52, 403–406.
photoperiod. Biol. Rhythm Res. 46, 113–119. Southward, E. S., Mann, F. A., Dodam, J., & Wagner-Mann, C. C. (2006). A comparison of
Giavarina, D. (2015). Understanding Bland Altman analysis. Biochem. Medica, 25, auricular, rectal and pulmonary artery thermometry in dogs with anesthesia-induced
141–151. hypothermia. J. Vet. Emerg. Crit. Care, 16, 172–175.
Goic, J. B., Reineke, E. L., & Drobatz, K. J. (2014). Comparison of rectal and axillary Taylor, N. A. S., Tipton, M. J., & Kenny, G. P. (2014). Considerations for the measurement
temperatures in dogs and cats. J. Am. Vet. Med. Assoc. 244, 1170–1175. of core, skin and mean body temperatures. J. Therm. Biol. 46, 72–101.
Gomart, S. B., Allerton, F. J. W., & Gommeren, K. (2014). Accuracy of different tem- Wiedemann, G. G. S., Scalon, M. C., Paludo, G., Silva, I. O., & Boere, V. (2006).
perature reading techniques and associated stress response in hospitalized dogs. J. Comparison between tympanic and anal temperature with a clinical infrared ray
Vet. Emerg. Crit. Care, 24, 279–285. thermometer in dogs. Arq. Bras. Med. Veterinária E Zootec. 58, 503–505.
González, A. M., Mann, F. A., Preziosi, D. E., Meadows, R. L., & Wagner-Mann, C. C. Yanmaz, L. E., Dogan, E., Okumus, Z., Şenocak, M. G., & Yildrim, F. (2015). Comparison
(2002). Measurement of body temperature by use of auricular thermometers versus of Rectal, Eye and Ear Temperatures in Kangal Breed Dogs. Kangal Irkı Köpeklerde
rectal thermometers in dogs with otitis externa. J. Am. Vet. Med. Assoc. 221, 378–380. Kulak Göz Ve Rektal Isıların Karşıştırılması, 21, 615–617.
Greer, R. J., Cohn, L. A., Dodam, J. R., Wagner-Mann, C. C., & Mann, F. A. (2007). Zanghi, B. M. (2016). Eye and Ear Temperature Using Infrared Thermography Are Related
Comparison of three methods of temperature measurement in hypothermic, eu- to Rectal Temperature in Dogs at Rest or With Exercise. Front. Vet. Sci. 3, 111.
thermic, and hyperthermic dogs. J. Am. Vet. Med. Assoc. 230, 1841–1848.