Materials and Methods

Criteria for case selection—The medical records database of the Matthew J. Ryan Veterinary Hospital of the University of Pennsylvania was searched to identify dogs that had undergone anesthesia for diagnosis or surgical treatment of IVDD between January 1, 1992, and December 31, 1996, or between January 1, 2002, and December 31, 2006. Dogs were included in the study if they were > 1 year old and had been anesthetized at the Matthew J. Ryan Veterinary Hospital or at the Veterinary Imaging Center in Ambler, Pa, an off-site referral facility for MRI.

Dogs were excluded from the study if the medical record was incomplete or missing; if they had radiographic or clinical evidence of parenchymal lung disease such as pneumonia prior to undergoing anesthesia, evidence of heart disease (eg, left atrial enlargement evident on thoracic radiographs or echocardiograms), or any other spinal cord disease in addition to IVDD; or if they had undergone general anesthesia at any other facility within 24 hours prior to examination at the Matthew J. Ryan Veterinary Hospital.

Medical records review—Information obtained from the medical records of dogs included in the study consisted of signalment, body weight, diagnostic and surgical procedures that were performed, location of IVDD (cervical, thoracolumbar [T3 through L3], or lumbosacral [L4 through S3]), number of anesthetic episodes, number of days hospitalized, and outcome (ie, survived to discharge from the hospital vs died or euthanatized). In addition, medical records were reviewed to determine whether there was any evidence of pneumonia that developed within the first 48 hours after anesthesia. Dogs were considered to have developed postoperative pneumonia if, within 48 hours after anesthesia, they developed radiographic evidence of alveolar infiltrates compatible with pneumonia in conjunction with at least one of the following signs: cough, auscultatory abnormalities, oxygen saturation as determined by means of pulse oximetry (SpO₂) < 94%, or arterial partial pressure of oxygen (PaO₂) < 85 mm Hg. Dogs were also considered to have developed pneumonia if the diagnosis was confirmed at necropsy.

Identification of risk factors—To identify risk factors for the development of postoperative pneumonia, findings for dogs examined between 2002 and 2006 that developed pneumonia were compared with findings for a randomly selected control group of unaffected dogs from the same population, with 3 control dogs identified for each case dog. Information obtained from the medical records of case and control dogs consisted of age, sex, body weight, body condition score, preexisting diseases, whether vomiting or regurgitation occurred before or after anesthesia, neurologic status prior to anesthesia, location of IVDD, number of anesthetic procedures, American Society of Anesthesiologists physical status score, duration of anesthesia, body temperature following anesthesia, surgical and anesthetic complications, and drugs administered before, during, and after anesthesia (ie, anesthetics, analgesics, anxiolytics, sedatives, and anticholinergics).

Statistical analysis—For continuous variables, the Shapiro-Wilks test was used to determine whether data were normally distributed. Data that were normally distributed were summarized as mean ± SD; data that were not normally distributed were summarized as median and range. The Student t test, for normally distributed data, or Wilcoxon rank sum test, for data that were not normally distributed, was used to compare continuous variables between groups. Categoric data were summarized as percentages, and the χ² or Fisher exact test was used to compare percentages between groups. For putative risk factors, odds ratios and their 95% CIs were calculated by means of the exact method. All calculations were performed with standard software.^a Values of P < 0.05 were considered significant.

Results

A total of 739 dogs underwent general anesthesia for diagnosis or surgical treatment of IVDD during the 2 study periods, including 347 dogs during the earlier study period (1992 through 1996) and 392 dogs during the later study period (2002 through 2006). Of these, 32 were excluded from the study because of missing or incomplete records (n = 12), evidence of other spinal cord diseases (10), preanesthetic evidence of cardiac disease (5), or preanesthetic evidence of parenchymal lung disease (4) or because the dog was < 1 year old (1). The remaining 338 dogs from the earlier study period and 369 dogs from the later study period were included in the study (Table 1).

Comparison between earlier and later study periods—Two of the 338 (0.6%) dogs anesthetized in the earlier study period developed postoperative pneumonia. Both developed auscultatory abnormalities, hypoxemia (SpO₂ < 94%), and radiographic evidence of alveolar infiltrates within 48 hours after undergoing anesthesia. Seventeen of the 369 (4.6%) dogs anesthetized in the later study period developed postoperative pneumonia. All 17 dogs developed radiographic evidence of alveolar infiltrates in combination with auscultatory abnormalities (n = 14), hypoxemia (SpO₂ < 94% or PaO₂ < 85 mm Hg; 12), cough (8), or fever (3) within 48 hours after undergoing anesthesia.

The incidence of postoperative pneumonia was significantly (P = 0.001) higher during the later study period than during the earlier study period. There were no significant differences between the 2 time periods with regard to age, breed, body weight, sex, location of IVDD, or survival rate (ie, percentage that survived to discharge). Alternative diagnostic imaging procedures such as MRI and CT were used significantly (P < 0.001) more frequently than myelography, and hemilaminectomy was performed significantly (P < 0.001) more frequently than dorsal laminectomy during the later study period, compared with the earlier study period. Number of anesthetic procedures per patient was significantly (P = 0.031) higher and duration of hospitalization was significantly (P < 0.001) longer during the later study period.

When all 707 dogs were considered together, there was a significantly (P < 0.001) higher incidence of pneumonia among dogs that were anesthetized more than once (10/68 [15%]) than among dogs that were anesthetized only once (9/639 [1.4%]). During the earlier study period, postoperative pneumonia was not diagnosed in any of the 314 dogs that were anesthetized only once but was diagnosed in the 2 of the 24 (8%) dogs that were anesthetized more than once (P = 0.004). During the later study period, postoperative pneumonia was diagnosed in 9 of 325 (3%) dogs that were anesthetized only once and in 8 of 44 (18%) dogs that were anesthetized more than once (P < 0.001). After adjustment for study period, dogs that were anesthetized more than once were 7.8 times as likely (95% CI, 2.4 to 24.2) to develop postoperative pneumonia as were dogs that were anesthetized only once.

The distribution of location of IVDD did not differ significantly between dogs in the earlier versus later study period (Table 1). When all 707 dogs were considered together, dogs that had cervical lesions had a significantly (P < 0.001) higher incidence of postoperative pneumonia (11/129 [9%]) than did dogs with thoracolumbar (6/494 [1%]) or lumbosacral lesions (2/84 [2%]). Dogs with cervical lesions were 6.6 times as likely (95% CI, 2.4 to 19.4) to develop postoperative pneumonia as were dogs with lesions in other locations. For all 707 dogs as a whole, those that had cervical lesions were significantly (P < 0.001) more likely to be anesthetized more than once (34/129 [26%]) than were dogs with thoracolumbar (24/494 [5%]) or lumbosacral (10/84 [12%]) lesions. Dogs with cervical lesions were significantly (P = 0.001) more likely to develop postoperative pneumonia if they had undergone anesthesia more than once (8/34 [24%]) than if they had undergone anesthesia only once (3/95 [3%]). Dogs with cervical lesions in the later study period were 6.8 times as likely (95% CI, 1.3 to 66.5; P = 0.01) to develop postoperative pneumonia (9/56 [16%]) as were dogs with cervical lesions in the earlier study period (2/73 [3%]).

For all 707 dogs as a whole, those that underwent MRI were significantly (P < 0.001) more likely to develop postoperative pneumonia (7/54 [13%]) than were dogs that underwent other diagnostic imaging techniques, such as myelography or CT (12/652 [2%]). Dogs that underwent MRI were 7.9 times as likely (95% CI, 2.5 to 23) to develop postoperative pneumonia as were dogs in which other diagnostic imaging modalities were used. Among the 639 dogs that were anesthetized only once, there was a significantly (P = 0.019) greater incidence of postoperative pneumonia if the dog had undergone MRI (2/16 [12.5%]) than if other diagnostic imaging modalities had been used (7/623 [1%]). Therefore, for dogs that had anesthesia only once, those that underwent MRI were 12.6 times as likely (95% CI, 1.2 to 74) to develop postoperative pneumonia as were dogs in which other diagnostic imaging modalities were used. However, for dogs that had anesthesia more than once, the incidence of postoperative pneumonia did not differ significantly (P = 0.736) between dogs that underwent MRI (5/38 [13%]) and dogs in which other diagnostic imaging modalities were used (5/29 [17%]).

Risk factors for development of postoperative pneumonia—To identify potential risk factors for development of postoperative pneumonia, data for the 17 dogs that developed postoperative pneumonia in the later study period were compared with data for 56 randomly selected, unaffected control dogs from the same period. There were no significant differences in age, sex, body weight, body condition score, American Society of Anesthesiologists physical status score, or postoperative body temperature between affected and control dogs.

Two of the affected dogs had a history of possible upper airway disease prior to anesthesia, including a Yorkshire Terrier with a history of collapsing trachea and a Pug with moderate upper airway noise. None of the control dogs reportedly had upper airway disease prior to anesthesia.

One affected dog but none of the control dogs had a history of vomiting or regurgitation prior to anesthesia. Six of the 17 (35%) affected dogs and 6 of the 56 (11%) control dogs reportedly vomited or regurgitated after undergoing anesthesia, and dogs that vomited or regurgitated were significantly (P = 0.017) more likely to develop postoperative pneumonia than were dogs that did not.

Seven of the 17 (41%) affected dogs and 2 of 56 (4%) control dogs had tetraparesis prior to anesthesia. These proportions were significantly (P = 0.001) different, and dogs with tetraparesis prior to anesthesia were 17.85 times as likely (95% CI, 2.7 to 188) to develop postoperative pneumonia.

Nine of the 17 affected dogs had cervical lesions, 6 had thoracolumbar lesions, and 2 had lumbosacral lesions. By contrast, only 5 of the 56 control dogs had cervical lesions, 45 had thoracolumbar lesions, and 6 had lumbosacral lesions. Dogs with cervical lesions were significantly (P < 0.001) more likely to develop postoperative pneumonia than were dogs with lesions in other locations.

Dogs with postoperative pneumonia were significantly (P = 0.002) more likely to have undergone anesthesia more than once (7/17) than were control dogs (4/56). Duration of the first anesthetic episode was significantly (P = 0.031) longer for affected dogs (mean ± SD, 277 ± 87 minutes) than for control dogs (228 ± 61 minutes). There was no significant (P = 0.705) difference in duration of the second anesthetic episode between groups.

Information on drugs administered during the first anesthetic episode was available for 11 affected and 52 control dogs (Table 2). We did not detect any significant differences between groups with regard to drugs administered during the first anesthetic episode. Information on drugs administered during the second anesthetic episode was available for all 7 affected dogs and all 4 control dogs. Dogs that developed postoperative pneumonia following the second anesthetic episode were significantly (P = 0.015) more likely to have received hydromorphone than were control dogs. Dogs that received fentanyl as an analgesic after the second anesthetic episode were significantly (P = 0.024) less likely to develop postoperative pneumonia than were dogs that did not receive fentanyl. We did not detect other significant differences between groups with regard to drugs administered during the second anesthetic episode.

None of the dogs in either group had postanesthetic seizures. One control dog developed pneumothorax and underwent respiratory arrest after being extubated following the first anesthetic episode. None of the dogs had any bleeding complications during the first anesthetic episode; 1 control dog had bleeding from a small sinus during the second anesthetic episode. None of the dogs required any blood products.

Discussion

Results of the present study suggested that at our institution, the incidence of postoperative pneumonia in dogs anesthetized for diagnosis or treatment of IVDD had increased in recent years, with 17 of 369 (4.6%) dogs anesthetized between 2002 and 2006 but only 2 of 338 (0.6%) dogs anesthetized between 1992 and 1996 developing postoperative pneumonia. Risk factors for the development of postoperative pneumonia included undergoing > 1 anesthetic procedure, a longer duration of the first anesthetic episode, undergoing MRI, having a cervical lesion, having tetraparesis prior to undergoing anesthesia, and vomiting or regurgitating after undergoing anesthesia.

In the present study, we examined dogs anesthetized during 2 periods to determine whether there were differences that might explain the higher incidence of postoperative pneumonia in recent years. There were no significant differences in age, weight, sex, or location of IVDD between the 2 groups. However, the diagnostic and surgical procedures that were performed, number of anesthetic events, and duration of hospitalization differed between the 2 study periods.

Previously, myelography was the most frequently used diagnostic imaging method in dogs suspected to have IVDD, but in recent years, MRI and CT have become more popular, and in human medicine, MRI is currently the diagnostic procedure of choice when evaluating patients with cervical myelopathies^22,23 because it provides adjunctive information beyond that which can be obtained with myelography.^22,24 During the years covered by the present study, the veterinary hospital at the University of Pennsylvania did not have an on-site MRI facility and dogs were referred to an offsite facility for MRI and allowed to recover from anesthesia before being transferred back for treatment. Our results suggested that dogs that underwent MRI were significantly more likely to develop pneumonia than were dogs in which other diagnostic imaging modalities were used, and MRI appeared to be an independent risk factor for the development of postoperative pneumonia even when we controlled for the number of anesthetic events. It is possible that the restricted access to patients during MRI makes it difficult to monitor them closely. Additionally, many of these patients were transported back to the teaching hospital immediately after recovery from anesthesia, often while they were still groggy, and it is possible that inadequate nursing and monitoring during transport facilitated the occurrence of gastrointestinal tract or respiratory events that increased the risk that the dogs would develop aspiration pneumonia. Despite the higher risk of pneumonia that may be associated with the use of MRI, this modality still remains the diagnostic method of choice for dogs with cervical spinal cord disease.

Dogs examined during the later period in the present study were anesthetized significantly more often than were dogs examined during the earlier period. In part, this was a result of a preference for MRI during the later period and the need to transport dogs to an off-site facility for this procedure, followed by a second anesthetic episode when the dogs were returned to the teaching hospital for treatment. Hypoxemia commonly occurs during general anesthesia, largely as a result of atelectasis and overperfusion of nonventilated areas of lung, which leads to impaired gas exchange.²⁵ Atelectasis may persist for several days after anesthesia²⁶ and may be associated with the development of pneumonia.²⁷ Therefore, it may contribute to the risk of postoperative pneumonia, especially in dogs anesthetized more than once. Dogs examined during the later period in the present study were hospitalized for significantly longer than dogs in the earlier study period, which may also have reflected the more frequent use of multiple anesthetic episodes during the later period, and the longer hospitalization times may have increased the risk of nosocomial pneumonia.

In the present study, duration of the first anesthetic episode was significantly longer for dogs that developed postoperative pneumonia than for unaffected control dogs. Similar results were found in a recent study⁷ of pulmonary complications following laparotomy in dogs. Alveolar macrophages play an integral role in protecting the lung from infection. Studies^28,29 in people suggested that volatile anesthetics can suppress the function of alveolar macrophages, and duration of anesthesia and surgery in people has been found to be significantly correlated with death of alveolar macrophages.³⁰ It is possible, therefore, that prolonged anesthesia in dogs results in the loss of alveolar macrophages, which may increase the risk of pneumonia. In addition, a longer anesthesia time may reflect the presence of more severe disk disease or the level of expertise of the surgeon, which are also possible risk factors for complications.

Because of the small number of dogs in the present study, it was difficult to identify associations between particular drugs given during the perioperative period and the development of postoperative pneumonia. It is possible that administration of certain opioids during the perioperative period may increase sedation, leading to respiratory compromise, or increase the risk of gastroesophageal reflux, increasing the risk of aspiration pneumonia. We found that dogs that received hydromorphone during a second anesthetic episode were significantly more likely to develop postoperative pneumonia, whereas dogs that received fentanyl after a second anesthetic episode were significantly less likely to develop pneumonia. Pure P opioid receptor agonists are commonly used for analgesia, but they promote respiratory depression that can result in hypoventilation. In a previous study,³¹ for instance, human patients who received hydromorphone postoperatively had a significantly greater risk of respiratory depression or desaturation. Furthermore, opioids have been shown to reduce lower esophageal sphincter pressure in humans and dogs,^32–34 and this may increase the risk of gastroesophageal reflux and aspiration. It may be that dogs that received hydromorphone during the second anesthetic episode in the present study had more central and respiratory depression resulting in hypoventilation or had more episodes of gastroesophageal reflux, compared with dogs that received fentanyl. However, this conclusion must be interpreted cautiously given the small numbers of animals.

In the present study, we found that dogs with cervical lesions had a significantly greater risk of developing postoperative pneumonia than did dogs with thoracolumbar or lumbosacral lesions. The cause is likely multifactorial. Dogs with cervical lesions were more likely to undergo anesthesia more than once, which was itself identified as an independent risk factor for the development of postoperative pneumonia. In addition, cervical spinal cord dysfunction may be intrinsically associated with a greater risk of pneumonia. Patients with cervical spinal injury may have impaired innervation of the respiratory muscles, resulting in decreased tidal volume, atelectasis, impaired cough, or difficulty clearing secretions. Bronchial hyperresponsiveness may be caused by loss of sympathetic innervation to the airway with unopposed parasympathetic innervation.³⁵ In a recent study,⁵ 5% of dogs undergoing cervical spinal cord surgery required positive-pressure ventilation to treat perioperative hypoventilation. In our study, only 1 dog developed severe enough hypoventilation to require positive-pressure ventilation, and this dog developed ventilator-associated pneumonia but did not fulfill our criteria for development of postoperative pneumonia because signs did not develop until > 48 hours after anesthesia. Even in dogs with cervical lesions without respiratory nerve dysfunction, tetraparesis and immobility represent additional concerns, especially for larger dogs, because of the risk of persistent atelectasis. We found that dogs with tetraparesis prior to anesthesia were significantly more likely to develop postoperative pneumonia in the present study, and many previous studies^36–38 have shown that human patients with tetraplegia have a substantial risk of pneumonia.

Gastrointestinal tract dysfunction may also contribute to the risk of postoperative pneumonia in dogs with cervical spinal cord disease. Dysphagia is common in human patients who undergo anterior cervical spinal cord surgery, with reported prevalence ranging from 5% to 45%.³⁹ Human patients with cervical spinal cord injury can have abnormal function of the upper esophageal sphincter, which has been associated with aspiration and pulmonary complications.⁴⁰ Gastroesophageal reflux has been reported in 17% of dogs undergoing anesthesia, with or without surgery,^41,42 and may reflect reduced lower esophageal sphincter tone.^16,41,42 If regurgitation occurs, gastroesophageal reflux may increase the risk of aspiration pneumonia. It is possible that lower or upper esophageal sphincter dysfunction, or both, was present in the dogs in our study with cervical lesions, predisposing them to aspiration pneumonia.

Regardless of the location of IVDD, we found that dogs with postanesthetic vomiting or regurgitation were significantly more likely to develop pneumonia than were dogs that did not vomit or regurgitate after anesthesia. Other studies^8,19 have documented similar results. The postanesthesia period is a critical time when patients are not fully conscious and upper airway reflexes are impaired. In addition, patients are likely to receive analgesics during this period, which may increase sedation and further increase airway vulnerability.

Hemilaminectomy was performed significantly more frequently than dorsal laminectomy during the later period in the present study, although this did not appear to be associated with development of postoperative pneumonia. This difference likely reflected the trend toward the use of hemilaminectomy in recent years because it provides better exposure and does not cause as severe instability.^43–45 Despite the differences identified between study periods, there was no significant difference in survival rates.

The present study had several limitations inherent to retrospective studies. Dogs included in the study were examined during 2 study periods at a veterinary teaching hospital and managed by multiple surgeons and anesthetists with various degrees of experience. As a result, many variables pertaining to diagnostic and therapeutic management varied. Furthermore, not every dog underwent preanesthetic thoracic radiography or echocardiography to rule out preexisting pulmonary parenchymal or cardiac disease. The inadvertent inclusion of some dogs with preexisting cardiopulmonary disease may have resulted in overestimation of the incidence of postoperative pneumonia. In addition, it is possible that some episodes of perianesthetic vomiting or regurgitation were not witnessed or recorded in the medical record, causing erroneous assessment of the association between vomiting and regurgitation and development of postoperative pneumonia. Because of the use of an off-site facility for MRI, it was impossible to determine what role MRI alone, versus the entire referral process, had on development of postoperative pneumonia. Finally, it was difficult to interpret information about perioperative drug administration given the small numbers of patients.