Introduction

Temperature assessment is a critical part of the physical examination in veterinary patients. Rectal temperature has long remained the gold standard for assessment; however, different techniques, such as axillary and infrared auricular probe, have been substituted for animals that are unable to have their rectal temperature obtained due to rectal disease or temperament. Furthermore, auricular temperature measurement has the additional benefits of being rapid and easily performed. As veterinarians explore alternatives, it is important that the temperature measured by alternative methods is accurate and truly reflects core body temperature.

Several recent studies have been completed comparing various temperature methods in dogs and cats.^1–6 Agreement between rectal and axillary and rectal and auricular temperature has been controversial. One study completed in patients with anesthesia-induced hypothermia revealed a strong correlation between auricular and pulmonary artery thermometry (bivariate correlation coefficient [r] > 0.846).¹ Alternatively, a study performed in awake animals reported a moderate correlation between auricular and rectal temperature (r = 0.609).² When comparing rectal and axillary temperature, several studies have consistently shown a poor correlation between the two.^2–4 As such, recommendations to define a reference range have been made.^4,5 The majority of completed, recent studies enrolled patients regardless of health status, and some patients were anesthetized or sedated.^3–5

To the authors' knowledge, there has not been a study completed in which multiple measurements were obtained in patients who did not have potentially life-threatening diseases. Assessment of human study designs revealed that temperatures were obtained at least three times on the same patient over the course of their hospitalization.^7–10 By including more temperature measurements on the same patient, a more accurate assessment for each technique can be made.

The objective of this study was to investigate the use of axillary and auricular temperatures in a population of systemically healthy dogs admitted to the hospital for an elective surgical procedure. Weight and body condition score (BCS) were also analyzed to determine if an agreement between temperature readings existed. Our hypothesis was that auricular temperature would have more agreement with rectal temperature than axillary temperature.

Materials and Methods

Systemically healthy adult dogs admitted to Dallas Veterinary Surgical Center and Oklahoma State University from August 2013 to July 2014 were enrolled into the study. The study design was approved by Oklahoma State University Animal Care and Use Committee. Medical records were reviewed for the following data: breed, sex, weight, age, BCS, temperature, and surgical procedure. Adult dogs had a physical examination, preoperative complete blood count, and serum biochemistry performed. Dogs were excluded from the study if they had signs of systemic illness (e.g., inflammatory leukogram, metabolic derangements, depression, etc.). Dogs were excluded from the study if rectal disease was present.

Sequential rectal and axillary temperatures were obtained with the same digital thermometer^a. Although this thermometer is reported to read out a temperature in 10 sec, the manufacturer package insert states that for axillary temperature, it generally takes 20 sec to read. Immediately following rectal and axillary thermometry, an infrared auricular thermometer^b was utilized. Temperatures of the axilla and ear were obtained on the patient’s left side. Rectal temperature was obtained by completely inserting the lubricated tip of the thermometer into the rectum. Axillary temperatures were obtained by placing the thermometer midway between the cranial and caudal aspects of the axilla and as far dorsally as possible. Auricular temperature was obtained by gently pulling the pinna lateral and ventral to straighten the external auditory canal. The probe was inserted as deep as possible into the ear canal. Per the manufacturer, disposable covers were placed on the infrared thermometer each time a temperature was obtained. All temperatures were recorded after the device beeped, signaling completion. All probes were disinfected after patient use. Temperatures were obtained in the aforementioned sequence without disruption of treatments or diagnostics. For accuracy and consistency, temperatures were obtained by one investigator.

Temperatures were obtained three times during the hospital stay at the time of physical examination. For patients at Oklahoma State University, the first set of temperatures was obtained on the day of presentation. The second set of temperatures was obtained the day of surgery before premedication, and the last set of temperatures was obtained after the patient had fully recovered from anesthesia. “Recovery” from anesthesia was defined as the patient being able to move in the kennel and no longer requiring heat support from a forced-air warming device. A minimum temperature of 36.6°C (97.8°F) was used, as it is considered mild hypothermia.^9,10 Due to the outpatient nature at Dallas Veterinary Surgical Center and the tendency for patients to be admitted the day of surgery, the first set of temperatures from patients in this population were obtained within 10–15 min of admission. The second set was obtained after the patient had recovered from anesthesia and was fully awake. The last set was obtained the morning of discharge. In this hospital, healthy patients generally had surgery one day and were discharged the next if no complications had occurred.

Pearson’s linear correlation coefficient was used to determine if a linear relationship existed amongst the temperature methods. Analysis of variance techniques were used to assess the difference in the temperatures among the methods. A Bland-Altman Limits of Agreement analysis was used to determine whether the axillary or auricular temperature was an underestimation or an overestimation of temperature. Pearson’s correlation coefficient was also utilized to determine the correlation between weight and BCS. A value of P < .05 was considered to be statistically significant for all analyses. Statistical analysis software^c was used for statistical analysis.

Results

Fifty dogs (26 females, 24 males) were enrolled into the study. The mean age of the dogs was 6.5 yr (range 1–11 yr), and median weight was 16.3 kg (35.8 kg; range 4–64 kg [8.8–140.8 lb]). One hundred and fifty temperature readings were performed during the course of the study. Breeds enrolled into the study included 14 dachshunds, 9 mixed-breed dogs, 7 Labrador retrievers, 2 golden retrievers, 2 Boston terriers, and 1 each of the following breeds: Chinese shar pei, Australian cattle dog, Akita, Pomeranian, German shepherd dog, boxer, French bulldog, bichon frise, Australian shepherd, Dalmatian, Pembroke Welsh corgi, English springer spaniel, Yorkshire terrier, beagle, and Great Pyrenees. Average BCS was 5.8/9 (median 6, range 4–8). One patient did not have a BCS recorded. The majority of the patients (n = 18) presented for elective surgery for a cranial cruciate ligament rupture. Within that category, 16 patients had a tibial plateau leveling osteotomy, 1 had a tibial tuberosity advancement, and 1 had a tightrope procedure performed. The second most common elective surgery was percutaneous laser disk ablation (10 patients). Hemilaminectomy was performed in nine patients. Three patients had a ventral slot procedure, and three patients had a surgical correction of a medial patellar luxation. One patient each had the following surgical procedures: femoral head ostectomy, tibial plateau leveling osteotomy explant, total hip arthroplasty, arthrotomy, laparoscopic-assisted gastropexy and orchidectomy, oronasal fistula repair, and calcaneal fracture repair.

Axillary and rectal temperatures had a poor agreement (r = 0.65, P < .001; Figure 1A). Mean axillary temperature (37.0°C, standard deviation [SD] 1.1°C [98.6°F, SD 2.0°F]) was statistically different than mean rectal temperature (38.1°C, SD 0.8°C [100.5°F, SD 1.4°F]) (P < .001). Bland-Altman analysis revealed that overall axillary temperature underestimated rectal temperature (Figure 2A). Three percent (5/150) of axillary temperatures were higher than rectal temperature, and 96% (145/150) of axillary temperatures were lower than rectal temperature. The number of axillary temperatures with a difference greater than 0.5°C [0.9°F] of rectal temperature was 138 (92%). The average discrepancy between the readings was 1.2°C [2.1°F]. The highest difference between axillary and rectal temperature was 4.0°C [7.3°F]. Axillary temperatures were never considered to be hyperthermic (>39.1°C [102.5°F]). Forty-six percent (69/150) of axillary temperatures were considered to be hypothermic (<36.9°C [98.5°F]). One axillary temperature reading was the same as the rectal temperature.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Auricular and rectal temperatures had a stronger agreement than axillary and rectal temperatures (r = 0.80; P < .001; Figure 1B). Bland-Altman analysis revealed that auricular temperature underestimated rectal temperature (Figure 2B). Mean auricular temperature (38.1°F, SD 0.9°C [100.7°F, SD 1.7°F]) was statistically different from mean rectal temperature (38.1°C, SD 0.8°C [100.5°F, SD 1.4°F]) (P < .001). Fifty-five percent (83/150) of auricular temperatures obtained were higher than the corresponding rectal temperature, 40.6% of auricular temperatures (61/150) were lower than rectal temperatures, and 2.6% (4/150) of auricular temperatures were the same as rectal temperatures. The percentage of auricular temperatures considered to hyperthermic (>39.1°C [102.5°F]) was 13.6% (20/150). The percentage of auricular temperatures considered to hypothermic (<36.9°C [98.5°F]) was 10.6% (16/150). The number of auricular temperatures with a difference greater than 0.5°C (0.9°F) of rectal temperatures was 84 (56%). The average discrepancy between the readings was 0.6°C [1.2°F]. The highest difference between auricular and rectal temperatures was 2.2°C [4.1°F]. Six auricular temperatures (4%) were the same as rectal temperatures.

Agreement was weakest between axillary and auricular temperatures (r = 0.55; P < .001; Figure 1C). Bland-Altman analysis revealed that auricular temperature overestimated axillary temperature (Figure 2C). Mean auricular temperature (38.1°F, SD 0.9°C [100.7°F, SD 1.7°F]) was significantly different from mean axillary temperature (37.0°C, SD 1.1°C [98.6°F, SD 2.0°F]). Ninety-four percent (141/150) of auricular temperatures were higher than axillary temperatures, and 4.6% were lower than axillary temperatures. The number of axillary temperatures with a difference greater than 0.5°C (0.9°F) of auricular temperature was 139 (92.6%). The average discrepancy between the readings was 1.3°C [2.4°F]. The highest difference between axillary and auricular temperature was 4.6°C [8.3°F]. Two auricular temperatures (1.3%) were the same as axillary temperatures.

Body condition score was not correlated with either auricular or axillary temperature measurement (r = 0.53, P = .09 [auricular]; r = -0.13, P = .36 [axillary]). Weight had a negative correlation with axillary temperature (r = -0.48, P = .003). That is, as weight increased, the axillary temperature decreased. No correlation was found between weight and auricular temperature (r = -0.03, P = .84).

As discussed previously, temperatures were obtained three times during the patient’s hospitalization. Agreement for each temperature set was determined. For each set obtained and analyzed individually, all sets had a significant difference between rectal and axillary temperature and auricular and rectal temperature. Agreement continued to be strongest for auricular temperature measurement. For the first data set obtained, the r value for axillary and rectal temperature was 0.44 (P < .01), and the r value for auricular and rectal temperature was 0.79 (P < .01). Correlation coefficients for the second set of temperatures obtained were 0.79 (axillary, P < .01) and 0.84 (auricular, P < .01). Correlation coefficients for the third set of temperatures obtained were as follows: axillary 0.64 (P < .01) and auricular 0.67 (P < .01).

Temperature agreement amongst the methods was also determined for hyperthermic and hypothermic animals. Of the patients with hyperthermia (four), no significant difference was found between rectal temperature and axillary temperature (P = .25) and rectal temperature and auricular temperature (P = .63). For patients with hypothermia (12), no significant difference was found amongst methods (rectal and axillary, P = .06; rectal and auricular, P = .21).

Sensitivity and specificity were determined for each temperature method. Sensitivity and specificity for axillary temperature were 68% (95% confidence interval [CI], 46–96) and 58% (95% CI, 49–66), respectively. Auricular temperature had a sensitivity of 75% (95% CI, 53–96) and a specificity of 76% (95% CI, 69–84).

Discussion

Although results of the current study show positive correlation between axillary and auricular temperatures with rectal temperature, overall predictability of rectal temperature was poor for each technique. Pearson’s linear correlation coefficients suggest a weak agreement with axillary and rectal temperatures and a stronger agreement between auricular and rectal temperatures. This is an important finding, as alternative temperature methods are becoming more widespread.

This study specifically evaluated animals that were undergoing nonemergent surgical procedures. One temperature set was obtained after the patient recovered from surgery. For Dallas Veterinary Surgical Center hospital, this was the second temperature set obtained, and for Oklahoma State University hospital, it was the final set of temperatures obtained. “Recovered” was defined as freely moving within the kennel and no longer needing heat support. A recent study revealed that auricular temperature had a strong correlation with rectal temperature (r > 0.846) during postoperative warming in hypothermic and normothermic postanesthetic patients.¹ During the postanesthetic period, the core body temperature can change rapidly.¹¹ Peripheral vasoconstriction to warm the body core may have occurred in our postanesthetic patients, leading to a decreased axillary temperature. This may explain the weaker agreement between axillary and rectal temperature in this study compared to other studies, in which correlation has been reported to be as high as 75%.^3,4

Statistical analysis by way of Bland-Altman and Limits of Agreement were performed, as it provides the most meaningful information in this context. Limits of agreement allow one to look quickly at a graph (Figure 1A–C) and make an accurate assessment of agreement.^14,15 Bland-Altman and Limits of Agreement are not tests of significance but are best used for estimation and agreement. This is particularly true in this study since it is unlikely that rectal, axillary, and auricular temperature would ever be the same.

Weight was found to be negatively correlated with axillary temperature. As the patient’s weight increased, the temperature recorded was cooler. This may be related to the amount of fat present within the axillary region, as it could displace the thermometer probe further from axillary vessels. Body condition score was not found to be correlated with either auricular or axillary temperature. The correlation of weight to a cooler axillary temperature is in contrast to a previous study in which BCS and weight were found to have no correlation in dogs.³

Environmental factors may influence the accuracy of these methods. Recently, environmental temperature has been shown to have a positive correlation with rectal and auricular temperature.⁶ Environmental temperature was unlikely to influence the results of the current study, as the hospital temperatures were well controlled. Other influences on temperature include time of day and bedding available. The time of day during temperature measurement was not recorded in the present study. It would be interesting to investigate how much of an influence time has on temperature in the future and if animals’ temperatures vary similarly with regard to time.

It is also important to note that auricular temperature measurement has been extrapolated from human medicine. The shape of the ear canal is straight in humans, allowing the infrared thermometer to obtain its reading from a measurement of the tympanic membrane.⁵ In dogs, the L-shape of the canal prevents direct measurement of the tympanic membrane. This may inadvertently lead to less predictable measurements than those found in human medicine since the canal wall or otic debris temperature may be measured inadvertently. However, one recent study suggested this was not a factor, as a strong correlation between auricular and rectal temperature occurred even in the presence of inflammation and debris.⁷

There are several limitations to the current study. Temperature measurement methods in two different facilities affected when temperatures were obtained. Although there was a difference, the authors believe that this did not affect the outcome of the study. All temperatures after surgery were consistently obtained once the animals were awake. In addition, a larger population of animals may also have allowed a more accurate assessment of temperature variation between the methods. However, a significant difference between rectal temperature and the alternative methods was found utilizing 50 dogs with 150 temperature measurements. Therefore, further patient enrollment into the study was deemed unnecessary, as it was unlikely to change the outcome of the study.

Conclusion

As auricular and axillary temperature methods increase in popularity, it is important that practitioners understand that the temperature obtained may not be a true estimation of core body temperature. If an alternative method must be used in dogs, the authors advocate the use of an auricular temperature measurement. However, the study illustrates that neither axillary nor auricular temperature can be accurately interchanged for rectal temperature in adult healthy dogs. Future studies should be aimed at establishing a reference range for auricular and axillary temperature in dogs rather than comparing agreement of temperature methods.