Fatal Hemothorax Following Management of an Esophageal Foreign Body

Case Report

A 10.8-year-old, 3.8-kg, spayed female toy poodle was referred to the University of Missouri–Veterinary Medical Teaching Hospital (UMC-VMTH) with a 1-week history of gagging and vomiting. Signs had begun soon after the dog had been given a compressed vegetable chew treat.^a The referring veterinarian had completed a serum biochemical profile and thoracic radiographs 5 days previously. In retrospect, radiographs showed an opacity within the esophagus cranial to the diaphragm, but at the time they were interpreted by the referring veterinarian as a pulmonary opacity believed to be compatible with pneumonia or pulmonary edema. The dog was treated with amoxicillin/clavulanic acid, furosemide, and intravenous (IV) fluids. When the dog failed to improve, it was referred to the UMC-VMTH.

On presentation, physical examination was unremarkable. Thoracic radiographs delineated a soft-tissue opacity within the caudal thoracic portion of the esophagus, which was dilated in the area of opacity [Figure 1]. Several small mineral opacities were located in the area of the esophageal dilatation as well. Although the pulmonary parenchyma in the area of the opacity could not be evaluated, the remainder of the lung fields were unremarkable. Mature neutrophilic leukocytosis (segmented neutrophils, 20.5 × 10³/μL; reference range, 3.0 to 11.5 × 10³/μL) and monocytosis (3.0 × 10³/μL; reference range, 0.15 to 1.35 × 10³/μL) were identified on complete blood count (CBC), while the only abnormal findings on serum biochemical analysis were a slight increase in anion gap (18 mmol/L; reference range, 7 to 15 mmol/L) and increased alkaline phosphatase (272 U/L; reference range, 22 to 116 U/L). Urinalysis was unremarkable with a specific gravity of 1.041.

Upper gastrointestinal (GI) endoscopy was performed on the day of presentation. With the dog in right recumbency and under general anesthesia, endoscopic evaluation was unremarkable to the level of the distal esophagus. The distal esophagus caudal to the heart contained a green, irregular, mucus-covered foreign body identified as a part of the previously ingested, compressed chew treat. Attempts to dilate the esophagus and advance the foreign body into the stomach through the cardia were unsuccessful. Attempts to retract the foreign body with a through-the-channel endoscopic biopsy instrument, a wire-pronged foreign body retriever, and a wire-loop foreign body retriever were unsuccessful and resulted in small pieces of the foreign body breaking apart. A rigid forceps designed for laparoscopic use was passed alongside the endoscope. While visualizing the foreign body endoscopically, the forceps was used to hold and apply gentle traction to the chew treat. The foreign body was advanced orally, resulting in its successful removal. Visual inspection of the esophageal mucosa after removal of the foreign body revealed severely reddened, ulcerated mucosa, with an irregular gray area on the dorsolateral surface of the esophagus [Figure 2]. The gastric mucosa appeared normal. The dog was placed in left recumbency, the ventrolateral abdomen was shaved and prepared routinely, and a 20-French (FR), 25-mm, mushroom silicone percutaneous endoscopic gastrostomy (PEG) tube^b was placed to allow enteral nutritional support while avoiding esophageal irritation from foodstuffs.

Initial recovery from the procedure was uneventful, and treatment was initiated with IV fluids (i.e., lactated Ringer’s solution), ampicillin (84 mg/kg body weight, IV q 8 hours), and cimetidine (20 mg/kg body weight, IV q 8 hours). Dexamethasone (0.5 mg/kg body weight, intramuscularly [IM] q 24 hours) was also administered in an attempt to minimize risk of esophageal stricture formation. Twenty-four hours after placement of the PEG tube, feedings with a recuperative diet^c (15 mL q 6 hours) were initiated through the tube. Sucralfate slurry (500 mg, per os [PO] q 8 hours) was initiated on the third day, and dexamethasone was discontinued. Over the next several days, the IV fluids were discontinued and the tube feedings were increased in volume to meet the dog’s calculated caloric requirements (230 kcal per day).

The dog was discharged from the UMC-VMTH 8 days after removal of the foreign body. At discharge, physical examination was unremarkable, but a CBC identified a mildly regenerative anemia (packed cell volume [PCV], 24%; reference range, 36% to 55%; reticulocytes, 5%), tentatively attributed to GI ulceration with blood loss, and persistent mild neutrophilia (segmented neutrophils, 22.9 × 10³/μL; reference range, 3.0 to 11.5 × 10³/μL). Biochemical analysis was not repeated at that time. Medications included sucralfate (500 mg slurry PO q 8 hours, 1 hour prior to cimetidine) and cimetidine (20 mg via gastrostomy tube q 8 hours). The owner was advised to continue tube feedings but to reintroduce oral feedings with a small volume of a moist-consistency diet beginning 14 days after removal of the foreign body. The 14-day delay in oral alimentation was designed to minimize trauma to the damaged esophageal mucosa in an effort to allow time for healing. The owner was also instructed to return to her regular veterinarian 5 days after discharge for the dog’s recheck examination and follow-up CBC; results indicated that the neutrophilia had resolved, and the normocytic/normochromic anemia was improved at that visit.

The dog was represented to the UMC-VMTH 17 days after discharge for removal of the feeding tube. The owner had discontinued use of the feeding tube altogether only 1 week after discharge and had reintroduced solid foods earlier than she had been advised. As the dog was eating well with no regurgitation or vomiting, the owner requested that the tube be removed. Physical examination was unremarkable, and the only abnormality identified on CBC and serum biochemical analysis was a mild normocytic/nor-mochromic anemia (PCV, 33%).

The dog was anesthetized to facilitate endoscopic examination of the esophagus and removal of the PEG tube. After premedication with butorphanol, propofol was used to induce anesthesia and allow placement of an endotracheal tube; anesthesia was maintained with isoflurane. Endoscopic examination revealed a normal cranial esophagus. Approximately 2- to 3-cm proximal to the lower esophageal sphincter, a short (approximately 0.5 cm in length) section of narrowed lumen was observed in the area previously obstructed by the foreign body [Figure 3]. The strictured area was easily dilated with a 16-FR dilating balloon^d filled with air to 20 to 25 psi, with no visible mucosal bleeding. The mucosal surface of the stomach showed several small gastric erosions and nonbleeding ulcers. The intraluminal “mushroom” tip portion of the PEG tube was identified and grasped using rat-toothed grasping forceps passed through the biopsy channel of the endoscope. The external portion of the PEG tube was severed flush with the body wall in preparation for removing the mushroom portion of the tube via traction through the esophagus and mouth. Because the mushroom tip was dropped from the instrument’s grasp several times during attempted removal, there was an approximately 10- to 15-minute delay between the time the esophagus had been balloon-dilated and the time the mushroom tip was eventually withdrawn. Mild resistance was encountered as the mushroom tip entered the distal esophagus while steady traction was applied to the endoscope, but the mushroom tip passed the area of prior stricture with a minimal exertion of force.

The dog became severely bradycardic and hypotensive immediately after the mushroom portion of the tube was removed through the mouth. Atropine was administered, and IV 0.9% saline was bolused at 90 mL/kg body weight per hour. Spontaneous respiration ceased, and mechanical ventilation was begun at 12 breaths per minute. Assuming the most likely cause of hemodynamic instability was an esophageal rupture with resulting pneumothorax, thoracocentesis was performed on the left side but produced only about 4 mL of air. The endoscope was reinserted into the esophagus, but only a minute amount of blood was visualized and no obvious esophageal perforation was seen [Figure 4]. The dog was moved to a sternal recumbent position, and thoracocentesis was repeated on the right side of the chest. This produced large volumes of a sanguineous fluid; the fluid’s packed cell volume and total protein, as well as the circumstances of its rapid appearance, strongly suggested that it was blood. This fluid was returned to the dog via auto-transfusion as it was removed from the thorax. In addition, 90 mL of whole blood was obtained from a blood donor for immediate transfusion. The dog’s owner could not be reached to grant permission for thoracotomy, so conservative therapy was continued. The heart rate increased to 110 to 140 beats per minute. Blood pressures improved and were stable at a mean of near 60 mm Hg measured oscillometrically while the dog was in sternal recumbency, but they decreased dramatically when the dog was laterally recumbent. The rate of fluid administration was decreased to 10 mL/kg per hour within 30 minutes and after administration of the whole blood. Inhalant anesthesia was initially discontinued but was reinstituted when the animal began struggling during the clinician’s attempts to investigate the cause of hemothorax.

Thoracic radiographs obtained approximately 20 minutes after removal of the PEG tube were compatible with a pronounced right-sided pleural effusion and left-sided pneumothorax [Figure 5]. No air was observed in the mediastinum, as would have been expected with an esophageal rupture. A dynamic esophagram was performed after instilling 20 mL of a 1:3 ratio of contrast agent^e to sterile saline. Good filling of the esophagus was noted, with no leakage of contrast into the mediastinum. Continuous right-sided pleural evacuation, which had originally produced large volumes of sanguineous effusion, now produced very little fluid. Repeated thoracic radiographs taken approximately 1 hour after removal of the PEG tube revealed much less pleural fluid on the right, but a large volume of air was still present in the left pleural space. Blood pressures and heart rate remained stable. Although the source of the intrathoracic bleeding remained unknown, the rate of blood loss seemed to have decreased substantially. Because active bleeding seemed to have slowed considerably, chest tubes were not inserted. The dog was taken to the intensive care unit, and left-sided thoracocentesis was performed to reduce the pneumothorax. The dog remained intubated and continued to receive oxygen as well as isoflourane as needed.

After removing approximately 50 mL of air from the left-sided pleural space, the dog again became hemodynamically unstable with worsening bradycardia and hypotension. The dog was administered an immediate bolus of another 60 mL of whole blood, and the rate of saline administration was again increased to 90 mL/kg body weight per hour. The dog’s right hemithorax was wiped with antiseptic solution, and an emergency thoracotomy incision was performed at the sixth intercostal space. A large volume of blood flowed from the open chest, and a large, solid blood clot was removed. The source of the bleeding was not immediately apparent. In an effort to increase visualization in the face of continuous hemorrhage, the descending aorta was cross-clamped with a Satinsky vascular clamp as close to the heart as possible. Evaluation of the thoracic cavity revealed bilateral avulsion from the descending aorta of the vertebral arteries of the eighth and ninth thoracic vertebrae. Two of the avulsed vessels were rapidly ligated with 3-0 Vicryl. The Satinsky clamp was replaced in a tangential fashion for partial occlusion on the remaining avulsed vessels, allowing the aortic clamp to be removed after 4 minutes. Ligatures were placed over the other two avulsed vessel ends using 3-0 Vicryl. During the course of the exploratory thoracotomy, another 250 mL of whole blood was administered. Although hemostasis was eventually achieved, a very large volume of blood had been lost. Despite autotransfusion of the fluid produced by thoracocentesis and transfusion of a total of approximately 400 mL of fresh whole blood, the dog’s peripheral blood packed cell volume was only 4%. Resuscitation efforts were stopped when it became apparent that there was no chance for functional recovery, and the dog was pronounced dead approximately 1.5 hours after removing the feeding tube.

Necropsy was permitted, and numerous pathological conditions were noted. Multiple, <0.1-cm diameter, arterial branches from the thoracic aorta were avulsed, and many of these branches had been ligated. Hemorrhage was noted in the overlying parietal intercostal pleura adjacent to thoracic vertebrae. White, thread-like strands were attached to the caudal one-third of the esophageal serosa, and there was moderate, multifocal, serosal hemorrhage. A circumferential, tan to brown ulceration from 1 to 4 cm in length with smooth, white edges was noted on the mucosal surface of the caudal one-third of the esophagus. In the center of the ulceration was a linear, 1.6-cm partial-thickness laceration that extended to the serosa and a 0.3 ×0.1-cm serosal perforation.

Histopathologically, several areas of hemorrhage were seen in the adipose tissue surrounding the aorta. The aortic tunica intima was multifocally thickened by a small amount of eosinophilic fibrillar material, and the associated internal elastic membrane was disrupted. At the level of an avulsed and ligated arterial branch, the aortic tunica media was regionally necrotic with moderate vacuolization and exhibited a decreased number of smooth muscle cells with pyknotic nuclei [Figure 6]. On a Verhoeff’s-van Gieson’s elastic-stained section, medial elastic fibers in this necrotic area were fragmented and separated. There was mild, multi-focal, adventitial mineralization. Valve leaflets in adjacent veins were thickened and hyalinized.

In the area of the esophageal perforation, the mucosa was regionally ulcerated, and mucosal epithelium at the edge of the ulceration was necrotic. The submucosa contained a moderate amount of granulation tissue that extended into the muscularis externa. Myofiber degeneration and regeneration were noted in the muscularis externa adjacent to the granulation tissue. The granulation tissue regionally contained a small amount of blood, a low number of neutrophils, and few macrophages and lymphocytes. A diverticulum extended from the edge of the perforation into the submucosal granulation tissue, and it partially undermined the mucosa. Distal to the perforation, there was moderate esophageal-serosal edema as well as mild, multifocal hemorrhage in the serosa and lamina propria. Histopathological changes in the surrounding tissues suggested that the esophageal laceration had occurred at the time of the original tissue injury rather than during the endoscopic procedures completed 1 to 2 hours prior to the dog’s death, but the tiny perforation within the ulcerated area of esophagus may have occurred acutely. Several histopathological changes common to elderly dogs were identified in other organs and were unrelated to the dog’s death.

Discussion

The ultimate cause of death in this dog was exanguination from avulsion of the dorsal intercostal arteries from the aorta. Vascular disease within the aorta almost certainly contributed to the propensity of the arteries to avulse. Nutrient supply to the outer layer of the thoracic aorta depends on blood flow from the vasa vasorum, contained within the periaortic connective tissues. Impairment of vasa vasorum flow produces medial degeneration and necrosis, as well as decreased aortic elasticity.¹² Medial degeneration of the aorta observed in this dog was similar to histopathological descriptions of changes observed in dogs experimentally deprived of vasa vasorum blood flow.²³ Medial degeneration of the aorta and subsequent diminution of aortic distensability and elasticity have been described in association with many conditions, including Marfan’s syndrome, diastolic hypertension, dissecting aortic aneurysm, syphilitic aortitis, aging, and following experimental disruption of vasa vasorum blood flow.¹²⁴⁵ In the single case report of naturally occurring medial aortic necrosis in a dog, no underlying cause was identified.⁴ In the authors’ experience, this histopathological finding is rare even in elderly dogs. Although experimental disruption of intercostal vasculature results in aortic medial necrosis in dogs,³ avulsion of the intercostal vessels in this dog almost certainly occurred acutely. The 1.5-hour time between vessel avulsion and the dog’s death would not have been sufficient to result in the observed histopathological changes of aortic medial necrosis. The authors speculate that the original injury to the esophagus produced a full-thickness laceration that went undetected clinically. The subsequent local inflammatory reaction not only resulted in adhesion formation between the esophagus and the aorta, but likely also disrupted blood flow through the vasa vasorum, resulting in aortic medial necrosis.

Adhesions between the aorta and esophagus likely were formed weeks previously, following the initial damage to the esophagus from the foreign body. The authors theorize that the slight cranial traction applied to the esophagus as the bulb of the feeding tube was pulled past the point of stricture was sufficient to pull the aorta cranially along with the esophagus. Because the dorsal intercostal arteries pass directly into the musculature of the thoracic wall, they offer little free mobility. Apparently the intercostal arteries remained fixed in place while the aorta moved slightly with the esophagus, resulting in the described avulsions. It is also possible that dilatation of the esophageal stricture contributed to avulsion of the intercostal arteries, but this seems to be a less likely scenario. Balloon dilatation uses only radial forces generated from inside the esophageal lumen. These forces would not be expected to result in significant cranial or caudal movement of the intercostal arteries, even if the arteries were trapped within a network of fibrous adhesions. Additionally, the 10- to 15-minute time lapse between esophageal dilatation and the acute onset of hemodynamic instability (which began within seconds of removing the mushroom tip of the feeding tube) would argue against vascular rupture occurring at the time of balloon dilatation.

This case demonstrated a very rare but ultimately fatal complication associated with the removal of a PEG feeding tube. The use of PEG feeding tubes has become quite common and is considered a relatively safe procedure in most cases. Although complication rates ranging from 44% to 58% are reported, most are of minor consequence.^6–8 Known complications include laceration of tissues during tube placement, malpositioning or migration of the tube, peristomal cellulitis or abcessation, peritonitis, subcutaneous emphysema, pneumoperitoneum or pneumoretroperitoneum, nausea or vomiting associated with tube feeding, aspiration of vomited foodstuffs, food leakage from the tube site, obstruction of the tube, premature removal of the tube, and difficulties associated with the owner’s use of the feeding tube.^6–15 Although poorly documented in the literature, obstruction of the GI tract with the bulb portion of the feeding tube is another potential anticipated complication when the tube is severed from the bulb either purposefully or accidentally. Most of the described complications occur either at the time of tube placement or during the course of its use. The complication described in this case report occurred not during use of the feeding tube, but rather when the tube was removed.

Had the authors been aware of the formed adhesions, alternate methods of PEG tube removal could have been utilized. It has been standard practice in the authors’ hospital to remove the bulb portion of the PEG tube via endoscopy, but alternative removal techniques are available. The external portion of the feeding tube may simply be cut off while leaving the bulb of the tube in the stomach.⁹ Although the bulb may pass through the GI tract of a larger dog, the small size of this dog precluded this option. The bulb portion of flexible PEG feeding tubes may be pulled through the gastrostomy site externally, often by inserting a stylet through the tube in an attempt to deform and elongate the bulb during its removal.⁹ The silicone tubes in use at the authors’ hospital at the time this dog was seen have the advantage of long-term stability, but the material does not have the flexibility required for removal through the gastrostomy site. In retrospect, surgical laparotomy for removal of the tube would have been another viable option, particularly in light of the known esophageal narrowing. The authors had anticipated that the gastrostomy bulb would pass through the dilated portion of the esophagus without incident. In fact, very little tension was appreciated when the bulb was pulled through the narrowed section of esophagus. Unfortunately, the authors had no suspicion that adhesions were present between the esophagus and the vasculature and that vascular integrity was compromised. Even if such lesions were suspected, they would not have been identified by additional routine imaging studies.