Problem

A 5-year-old castrated male Labrador Retriever weighing 25 kg (55 lb) was brought to the University of Georgia Small Animal Hospital for emergency care after being hit by a car. The dog was physically examined and determined to have a fracture in the right humerus and the right tibia. The dog was stabilized and, days later, was scheduled to be anesthetized for radiographic examination and surgery to repair both fractures. When the dog was returned to the hospital, physical examination and diagnostic imaging revealed no cardiovascular or respiratory abnormalities, and no injuries were evident except for the fractures. Glycopyrrolate (0.005 mg/kg [0.0023 mg/lb]), hydromorphone (0.05 mg/kg [0.023 mg/lb]), and midazolam (0.2 mg/kg [0.09 mg/lb]) were administered IV. Anesthesia was subsequently induced by IV administration of ketamine (5 mg/kg [2.3 mg/lb]) and diazepam (0.25 mg/kg [0.11 mg/lb]) and was maintained with isoflurane in oxygen.

The dog was prepared for surgery and taken for preoperative radiographs. During this time, the arterial oxygen saturation of hemoglobin (SpO₂) was monitored with a transmittance pulse oximeter probe positioned on the tongue.^a The readings varied from 98% to 90% during diagnostic imaging, and heart rate, respiratory rate, and arterial blood pressure were stable. Heart rate was also measured by use of the audible signals from a Doppler ultrasonographic monitor and coincided with the pulse reading from the pulse oximeter probe. The dog was then taken to the surgical theater where connections to the mobile pulse oximeter unit were removed and the dog was connected to a gas analyzer with pulse oximeter function.^b Again the probe was positioned on the tongue. The pulse oximeter displayed an SpO₂ of ≥ 98% for the remainder of the anesthetic episode. In the surgical theater, the heart rate obtained with the pulse oximeter probe was compared with that recorded via ECG and the audible readings from the Doppler monitor and again coincided.

Formulation of the Clinical Question

Because there were 2 dramatically different readings from 2 different SpO₂ probes, whereas the dog's status remained the same, the question arose as to the accuracy of these devices.

Clinical Question

How accurate are noninvasive transmission pulse oximeters in dogs?

Evidentiary Search Strategy

A targeted search of the PubMed database was performed by use of the following search terms: pulse oximetry accuracy in dogs. The search yielded 9 scientific reports. One report¹ was omitted because review of the associated abstract indicated the report concerned a noninvasive reflectance pulse oximeter.

Review of the Evidence

Each of the studies in the 8 reports included was considered a cohort study, and in each, arterial hemoglobin saturation results from a co-oximeter were used as the reference criterion (gold standard) to which other measuring devices were compared. Three studies^2–4 revealed that pulse oximeters are unreliable at estimating arterial oxygen saturation (SaO₂) at values of SpO₂ ≤ 70%. Results of 1 study² involving 5 anesthetized mixed-breed research dogs indicated that pulse oximetry underestimated high SaO₂ values (≥ 70%) and overestimated low SaO₂ values (< 70%), compared with values measured by use of a multiwavelength spectrophotometer. The same study also revealed that high arterial CO₂ tension (≥ 60 mm Hg) was associated with greater overestimation of SaO₂, compared with measurements obtained during normocarbia (30 to 60 mm Hg) or hypocarbia (≤ 30 mm Hg). The magnitude of error also increased at SaO₂ values ≤ 70%, with 20% error evident at an SpO₂ < 30%.⁴ However, another study⁵ revealed that SpO₂ measurements in dogs were fairly accurate, with a mean ± SD bias of 5.5 ± 4.2% at SaO₂ values > 22%.

Location of the pulse oximeter probe plays an important part in its accuracy. In 1 study,² measurements obtained from a probe designed to be placed on ears and another designed to be placed on fingers were compared at 2 monitoring sites (the tongue and tail). Readings from the ear probe were less accurate when obtained from the tail at SaO₂ values ≥ 70%, compared with readings from the finger probe at either anatomic site or the ear probe placed on the tongue. Another report³ indicated that pulse oximeter readings varied according to anatomic site of probe placement, and coefficients of determination (R²) were as follows: tongue (0.95), lip (0.87), ear (0.69), toe (0.97), and prepuce or vulva (0.95).

Various models of pulse oximeters were also evaluated. In 1 study,³ 5 pulse oximeters were tested at SaO₂ values of 98%, 85%, and 72%. The results suggested that 4 of the 5 models overestimated SaO₂, whereas 1 model underestimated it.³ In another study,⁶ 2 models of pulse oximeters and their degree of accuracy were compared, revealing that both instruments recorded SaO₂ accurately until the SaO₂ was < 90%, after which both overestimated the SaO₂ value. A third study⁷ involved comparison of 2 models of pulse oximeters, revealing that at an SaO₂ value ≥ 80%, both pulse oximeters yielded accurate measurements but at an SaO₂ value < 80%, the devices became less accurate. In that study, pulse oximeter A had a coefficient of determination of 0.62 when compared with the co-oximeter and pulse oximeter B had a value of 0.86 at an SaO₂ > 80%. At an SaO₂ ≤ 60%, both oximeters were highly inaccurate with a coefficient of determination of 0.13 for pulse oximeter A and a value of 0.23 for pulse oximeter B when compared with the co-oximeter. Finally, a fourth study⁴ involved comparison of 3 models of pulse oximeters models tested at various SaO₂ values, revealing that at an SaO₂ > 70%, all the pulse oximeters performed well but at an SaO₂ value < 70%, results were inaccurate.

Certain disease conditions also reportedly affect the accuracy of pulse oximeters. A report⁸ on the reliability of an SpO₂ probe in anemic dogs indicated the overall mean ± SD measurement bias and precision was 0.2 ± 7.6% for Hct values from 40% to 10%, compared with the gold standard. At an Hct value < 10%, the bias and precision worsened to 5.4 ± 18.8%. Failure of the pulse oximeter to yield a signal also dramatically increased at Hct values < 10%, and a high signal failure rate was also evident at Hct values > 40%.

The effects of carbon monoxide inhalation on pulse oximeter accuracy were also evaluated.⁹ In that study, as the blood concentration of carboxyhemoglobin increased after carbon monoxide exposure and that of oxyhemoglobin decreased, the pulse oximeter continued to indicate an SaO₂ > 90%, whereas the actual arterial oxyhemoglobin concentration decreased to < 30%. In the presence of carboxyhemoglobin, the SpO₂ is the approximate sum of carboxyhemoglobin and arterial oxyhemoglobin concentration and thus may seriously overestimate arterial oxyhemoglobin concentration in patients with recent carbon monoxide exposure.⁹

The accuracy of the pulse oximeter was also evaluated during experimentally induced hypoxemia, hypotension, and hypertension in 7 healthy research dogs.⁴ The results indicated that acute blood loss or druginduced changes in peripheral arterial blood pressure do not affect SpO₂ readings. However, pulse pressure did affect whether values could be obtained because signal detection failure occurred during both vasodilation and hypertension, with hypertension causing the greatest percentage of signal detection failure.