Complications Associated with Endoscopic Removal of Percutaneous Endoscopic Gastrostomy (PEG) Tubes in Children

INTRODUCTION Percutaneous endoscopic gastrostomy (PEG) tube placement for enteral feeding is a standard procedure in selected children with severe chronic feeding problems or dysphagia (1,2). Complications of PEG placement include perforation of other abdominal organs during placement, peritonitis, stoma site infection, necrotizing fasciitis, tube migration and duodenal dislodgement, gastric ulceration or perforation, and mechanical failure (3,4). Complications of removal of the initial PEG tube are rarely reported. We report two children who experienced significant complications associated with endoscopic removal of the initial PEG. We discuss the surgical and anesthetic implications of these cases and make recommendations on how to avoid these complications. CASE REPORTS Case 1 A 4.5-year-old girl with dysphagia and cerebral palsy underwent elective PEG replacement. The initial PEG (20 Fr MIC-Key, Medical Innovations Corporation, Draper, UT) was placed at age 2 years because of malnutrition, dysphagia, and persistent regurgitation. Ulcerative esophagitis resistant to proton pump inhibitor therapy complicated her gastroesophageal (GE) reflux. A Nissen fundoplication was performed at age 4 years. The PEG change was scheduled electively 6 months after fundoplication. Her weight at removal was 15 kg, well below the third percentile for age. Standard endotracheal intubation and inhaled anesthesia was used for the procedure. An Olympus GIF XP 160 gastroscope (outer diameter 5.6 mm; Olympus Optical Company, Tokyo, Japan) was introduced. The PEG internal retention dome (IRD) was snared with a standard endoscopic polypectomy snare, and the gastrostomy tube cut at the skin surface. Removal of the IRD was attempted by way of the esophagus. Upward passage of the snared IRD beyond the GE junction was difficult because of the fundoplication. After multiple attempts, the IRD was maneuvered into the esophagus. It then became wedged in the upper esophagus, and simultaneously there was a decrease in airway compliance and oxygen desaturation. Oxygenation improved with change to hand ventilation with high pressure and 100% inspired O2. Attempts to push the IRD back into the stomach with the gastroscope were unsuccessful. The gastroscope was removed, leaving the snare attached to the IRD. A laryngoscope was inserted to attempt removal under direct vision. Applying traction to the snare moved the IRD into the upper esophagus, lodging it at the level of the upper esophageal sphincter. At this stage, the manual pressure required to ventilate decreased. During further retrieval attempts, the snare wire broke, and the IRD became lodged in a transverse plane in the cricopharyngeus muscle. Inspection of the upper esophagus with a laryngoscope revealed that the IRD was impacted with the cut end of the feeding tube facing rostrally and the cricopharyngeus muscle tightly surrounding the dome. Multiple attempts to remove the device with Magill forceps were unsuccessful despite deepening the anesthetic with propofol and muscle relaxant and using a Miller rather than a Macintosh laryngoscope blade. We passed a McGill's forceps beyond the cricopharyngeus in an unsuccessful attempt to mechanically compress the retention dome. The IRD was finally removed with force, using a long-toothed artery forceps. Subsequent rigid esophagoscopy demonstrated superficial esophageal abrasions with no evidence of esophageal perforation. After esophagoscopy, the patient was extubated but developed postextubation stridor requiring nebulized epinephrine and supplemental oxygen to maintain adequate O2 saturations. The patient was observed overnight in the intensive care unit and was discharged home 24 hours later without evidence of upper airway compromise. Seven days later, she returned to the emergency department with increased oral secretions and general malaise. Lateral neck radiograph revealed posterior pharyngeal swelling with no evidence of submucosal fluid collection or abscess. She was admitted for observation. After 48 hours of conservative management, she made a complete recovery and was discharged. Case 2 A 2.5-year-old girl with cystic fibrosis (weight 8.9 kg) presented for elective replacement of a 20 Fr Medical Innovations Corporation PEG tube with a Bard button gastrostomy (18 Fr, 1.7 cm; Bard International, Billerica, Massachusetts) feeding device. The initial PEG had been inserted 3 months prior without complications. Standard endotracheal intubation after mask induction of anesthesia was performed. An Olympus GIF XP160 gastroscope (outer diameter 5.6 mm; Olympus Optical Company, Tokyo, Japan) was used. The PEG IRD was snared and the gastrostomy tube severed at skin level. The device was successfully withdrawn through the GE junction on the first attempt. During passage through the upper esophagus, some resistance was felt, but excessive traction force was not required. A low-profile button device was inserted percutaneously with an introducer and the intragastric position endoscopically confirmed. The patient's immediate postoperative recovery was complicated by four vomiting episodes (3 containing specks of altered blood). Overnight, she became febrile, distressed, irritable, refused medical examination, and held her neck stiffly. Fevers of 38 to 39.5°C persisted for 1 week. Initial investigations revealed neutrophilia (total white cell count of 16.1 × 109/L with 11.6 × 109/L neutrophils); thrombocytosis (550-650 × 109/L); persistently elevated C-reactive protein of 78 mg/L (normal < 8 mg/L); and erythrocyte sedimentation rate of 55 mm/hr (normal 0-6 mm/hr). Three blood cultures yielded no bacterial pathogens. Lumbar puncture revealed normal cerebrospinal fluid. Ebstein-Barr virus and cytomegalovirus serology were negative. She was empirically treated with intravenous flucloxacillin and cefotaxime. An abdominal ultrasound showed no intra-abdominal fluid collections, and a radiologic stomagram confirmed proper positioning of the button gastrostomy. A nuclear medicine bone scan and echocardiogram revealed no infective source. As fever and neck stiffness persisted, her antibiotic treatment was changed from flucloxacillin to vancomycin to cover possible methicillin-resistant staphylococcal infection. A lateral neck radiograph followed by throat examination under anesthesia were performed, both of which were normal. A computerized tomographic (CT) scan of her neck revealed an abscess of the retropharyngeal space with possible mediastinal extension. Because a throat swab had grown profuse multiresistant Enterobacter spp., her intravenous antibiotic regimen was changed to clindamycin and gentamicin. Her fever and general condition improved gradually over the following 4 days. She was discharged home for 1 week of oral clindamycin, followed by 2 weeks oral amoxicillin/clavulanic acid. She developed no recurrence of the retropharyngeal abscess. DISCUSSION PEG tubes are commonly used in children with feeding difficulties and complex medical problems for nutritional rehabilitation. Approximately 80% of these children are regarded as "high risk" (ASA grade 3 and 4) for general anesthesia (5). This fact may explain the relatively high morbidity and mortality associatedwith initial PEG placement in children (3,4). In consideration of the worldwide frequency with which PEGs are inserted, replaced, and removed, complications appear to occur rarely. In adults, the recommendation for PEG removal is a nonendoscopic technique involving skin level division and spontaneous passage of the IRD (the "cut-and-push" technique). This is associated with low incidence of complications (2.7%) in a frail, medically unfit patient population (6). A retrospective audit of 397 adults requiring PEG removal also recorded a low complication rate, the main being persistent gastrocutaneous fistula (7). The "cut-and-push" technique is not recommended in children. A rare complication is the upward migration of the IRD, producing esophageal obstruction (8,9). In one case series, 5 of 11 children where spontaneous passage of the IRD was aimed for after the cut and push manuever developed complications: one death from esophageal perforation, three esophageal impactions requiring endoscopic retrieval, and one gastrocutaneous fistula containing the retained bumper two years after PEG the procedure (10). Significantly, in these patients, endoscopic retrieval did not prevent subsequent esophageal stricture development (9). A general recommendation resulting from these reports is to retrieve the PEG's IRD endoscopically in children. Both patients reported here suffered potentially life-threatening problems related to endoscopic PEG removal. The first child had significant upper airway obstruction at the cricopharyngeus, which caused tracheal compression distal to the endotracheal tube and impaired ventilation. In this patient, considerable force was required to remove the retained IRD with resultant pharyngeal injury noted on subsequent rigid esophagoscopy. This complication has been previously reported in children (8,11). Obstruction in our patient occurred despite the 2.4 cm diameter of the PEG's flexible disc, which is smaller than coins children frequently swallow without complication. The risk of impaction in the esophagus is likely related to the combination of the patient's small stature and the relative large size of the IRD with the residual 1 to 2 cm stalk protruding from its bumper. The stalk's added dimension impeded its removal, especially through the upper esophageal sphincter. The second patient developed a retropharyngeal collection despite an apparently uncomplicated endoscopic retrieval of the IRD. We hypothesize that removing the stalk through the cervical esophagus caused superficial trauma and an entry for bacterial colonization and invasion. In the setting of poor nutrition and cystic fibrosis, this may have caused the retropharyngeal infection. It is important to note that most cases of traumatic perforation resulting from diagnostic gastroscopy occur in the weak muscled upper esophageal sphincter area (12). This, coupled with the temporal relationship between endoscopy and retropharyngeal infection development, adds weight to our proposal. Unfortunately, the diagnosis of retropharyngeal infection can be delayed because of atypical symptoms, especially in children. We advocate that clinicians should consider this diagnosis because the potential for life-threatening infection and upper airway obstruction is high. Chest and lateral neck radiographs are essential and may detect pneumomediastinum, left-sided pneumothorax and pleural effusion, or widening of the retropharyngeal space. CT scans are useful in assessing the extent of retropharyngeal/mediastinal soft tissue infiltration. CT findings of acute mediastinitis secondary to perforation include esophageal thickening, extraluminal gas or contrast medium, pleural effusion, and single or multiple abscesses (13). The complications described in these two cases likely resulted from mismatch of the IRD with its large stalk and the narrow cervical esophagus in a small child. The IRD can often only be removed with difficulty, particularly if the gastrostomy plastic has aged, thus losing elasticity. The risk of this complication could be reduced by using a smaller PEG initially in young children, or alternatively, a low-profile button (14). A softer, more pliable IRD is not available and may lack durability. An alternative is the Bard Fastrac pull PEG device (20 Fr) that has an inflatable internal balloon. The balloon is deflated by severing the external gastrostomy. The device is then removed through the gastrocutaneous tract by application of traction without endoscopy. The product information states that the force required is 50% lower than required for removal of the Bard silicone PEG. In retrospect, the Fastrac may have been a preferred initial device for case 1. Our concluding recommendation is that after a technically difficult endoscopic removal of an IRD, children need to be carefully monitored for signs of esophageal perforation or retropharyngeal infection.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.