Novel Machine Learning Model to Predict Interval of Oral Cancer Recurrence for Surveillance Stratification.

OBJECTIVE(S): We aimed to develop a machine learning (ML) model to accurately predict the timing of oral squamous cell carcinoma (OSCC) recurrence across four 1-year intervals. METHODS: Patients with surgically treated OSCC between 2012-2018 were retrospectively identified from the Yale-New Haven Health system tumor registry. Patients with known recurrence or minimum follow-up of 24 months from surgery were included. Patients were classified into one of five levels: four 1-year intervals and one level for no recurrence (within 4 years of surgery). Three sets of data inputs (comprehensive, feature selection, nomogram) were combined with 4 ML architectures (logistic regression, decision tree (DT), support vector machine (SVM), artificial neural network classifiers) yielding 12 models in total. Models were primarily evaluated using mean absolute error (MAE), lower values indicating better prediction of 1-year interval recurrence. Secondary outcomes included accuracy, weighted precision, and weighted recall. RESULTS: 389 patients met inclusion criteria: 102 (26.2%) recurred within 48 months of surgery. Median follow-up time was 25 months (IQR: 15-37.5) for patients with recurrence and 44 months (IQR: 32-57) for patients without recurrence. MAE of 0.654% and 80.8% accuracy were achieved on a 15-variable feature selection input by 2 ML models: DT and SVM classifiers. CONCLUSIONS: To our knowledge, this is the first study to leverage multiclass ML models to predict time to OSCC recurrence. We developed a model using feature selection data input that reliably predicted recurrence within 1-year intervals. Precise modeling of recurrence timing has the potential to personalize surveillance protocols in the future to enhance early detection and reduce extraneous healthcare costs. LEVEL OF EVIDENCE: 3 Laryngoscope, 133:1652-1659, 2023.

Correlations of Radiographic and Endoscopic Observations in Subglottic Stenosis.

OBJECTIVE(S): Subglottic stenosis (SGS) represents a constellation of diverse pathologic processes that ultimately lead to narrowing of the subglottic region and can produce significant morbidity. Existing endoscopic and radiographic assessments may not be consistent in practice. METHODS: Severity of stenosis was evaluated and reported using the Cotton-Myer classification system from 33 endoscopic procedures from 32 unique subjects. Radiographic imaging within the preceding 3 month period was subsequently reviewed and narrowing was measured by a blinded radiologist. Degree of stenosis was reported as a percentage in 30 out of 33 endoscopic evaluations and subsequently compared to radiographically determined percentage of stenosis. Statistical analyzes were conducted to evaluate concordance between endoscopic and radiographic assessments. RESULTS: About 45.5% (15/33) of the evaluations were in agreement using Cotton-Myer scoring, while 27.3% (9/33) were discrepant by 1 grade and 27.3% (9/33) by 2 grades. Correlation of degree of stenosis as a percentage using Spearman (coefficient: 0.233, P-value: .214) and Pearson (coefficient: 0.138, P-value: .466) methods demonstrated very weak relationships. Radiographic scoring did not predict endoscopic classification to a significant degree using mixed effects regression. CONCLUSIONS: Radiographic and endoscopic grading of subglottic stenosis may not be reliably concordant in practice.

Thirty-day morbidity and mortality following otologic/neurotologic surgery: Analysis of the national surgical quality improvement program.

OBJECTIVES/HYPOTHESIS: To determine the rate and timing of, as well as risk factors for, postoperative morbidity and mortality following otologic and neurotologic surgery. STUDY DESIGN: Retrospective cohort study. METHODS: A total of 1,381 patients were identified in the American College of Surgeons National Surgical Quality Improvement Program for the years 2005 to 2010. Simple summary statistics, χ2 , and multivariable logistic regression were performed. RESULTS: Lateral skull base/neurotologic tumor resection (LSB) was done in 35.9%, and middle ear/mastoid procedures (MEM) were performed in 63.5%. The overall adverse event rate was 10.4%, although it was significantly higher for LSB (24.2%) and lower for MEM (2.6%). The overall mortality rate was 1.4%. Complications occurred postdischarge in 40.4% of cases. The outpatient setting (odds ratio [OR]: 0.31, 95% confidence interval [CI]: 0.15-0.65) and undergoing MEM (OR: 0.23, 95% CI: 0.12-0.47) were associated with lower risk of experiencing a complication. Impaired functional status (OR: 10.45, 95% CI: 3.65-29.89) was associated with postoperative mortality. An open wound preoperatively was associated with multiple causes of postoperative morbidity. CONCLUSIONS: Patients undergoing approaches to the skull base and neurotologic tumor resections had the higher adverse event rate. Open wounds were predictive of several postoperative complications, and poor functional status was associated with mortality. Patients with significant comorbidities should be evaluated early on in their postoperative course to prevent readmission as well as major morbidity and mortality. LEVEL OF EVIDENCE: 2b. Laryngoscope, 128:1431-1437, 2018.

Time-of-day influences on respiratory sequelae following maximal electroshock-induced seizures in mice.

Sudden unexpected death in epilepsy (SUDEP) is the leading cause of death in refractory epilepsy patients. Although specific mechanisms underlying SUDEP are not well understood, evidence suggests most SUDEP occurs due to seizure-induced respiratory arrest. SUDEP also tends to happen at night. Although this may be due to circumstances in which humans find themselves at night, such as being alone without supervision or sleeping prone, or to independent influences of sleep state, there are a number of reasons why the night (i.e., circadian influences) could be an independent risk factor for SUDEP. We explored this possibility. Adult male WT mice were instrumented for EEG, EMG, and EKG recording and subjected to maximal electroshock (MES) seizures during wakefulness, non-rapid eye movement (NREM) sleep, and rapid eye movement (REM) sleep during the nighttime/dark phase. These data were compared with data collected following seizures induced during the daytime/light phase. Seizures induced during the nighttime were similar in severity and duration to those induced during the daytime; however, seizures induced during the nighttime were associated with a lesser degree of respiratory dysregulation and postictal EEG suppression. Seizures induced during REM sleep during the nighttime were universally fatal, as is seen when seizures are induced during REM during the daytime. Taken together, these data implicate a role for time of day in influencing the physiological consequences of seizures that may contribute to seizure-induced death.NEW & NOTEWORTHY Sudden unexpected death in epilepsy (SUDEP) is the leading cause of death in patients with refractory epilepsy. SUDEP frequently occurs during the night, which has been attributed to an effect of sleep. We have shown that sleep state does indeed influence survival following a seizure. That SUDEP occurs during the night could also implicate a circadian influence. In this study we found that time of day independently affects the physiological consequences of seizures.

Comparing 30-Day Morbidity and Mortality in Pediatric and Adult Otologic Surgery.

Objective To determine differences in timing and rate of postoperative adverse events among pediatric and adult populations undergoing specific otologic procedures. Study Design Administrative database study. Setting Multi-institutional database. Subjects and Methods The National Surgical Quality Improvement Program (NSQIP) and NSQIP-Pediatric (NSQIP-P) were used to extract data from 819 adults (years 2005-2010) and 7020 children (years 2012-2014) undergoing tympanoplasty and (tympano)mastoidectomy, respectively. Simple summary statistics, χ2, and multivariable logistic regression analyses were performed. Results There were no significant differences in overall adverse event rates between adults (2.9%) and children (2.3%) ( P = .233). Adults experienced infectious complications more frequently than did children (0.4% vs 0.0%, P = .002). Postdischarge complications accounted for 83.7% of all complications. Children treated by pediatric otolaryngologists had higher readmission rates (odds ratio [OR], 2.80; 95% confidence interval [CI], 1.20-3.60; P = .002). Tympanomastoidectomy was associated with higher odds of reoperation (OR, 1.02; 95% CI, 1.01-1.03; P < .001), as was undergoing a concurrent procedure that did not include myringotomy (OR, 3.38; 95% CI, 1.47-7.79; P = .004). Conclusion Both adult and pediatric otologic surgery are safe, with patients experiencing similarly low complication rates. Most adverse events occur after discharge.

Influence of vigilance state on physiological consequences of seizures and seizure-induced death in mice.

Sudden unexpected death in epilepsy (SUDEP) is the leading cause of death in patients with refractory epilepsy. SUDEP occurs more commonly during nighttime sleep. The details of why SUDEP occurs at night are not well understood. Understanding why SUDEP occurs at night during sleep might help to better understand why SUDEP occurs at all and hasten development of preventive strategies. Here we aimed to understand circumstances causing seizures that occur during sleep to result in death. Groups of 12 adult male mice were instrumented for EEG, EMG, and EKG recording and subjected to seizure induction via maximal electroshock (MES) during wakefulness, nonrapid eye movement (NREM) sleep, and rapid eye movement (REM) sleep. Seizure inductions were performed with concomitant EEG, EMG, and EKG recording and breathing assessment via whole body plethysmography. Seizures induced via MES during sleep were associated with more profound respiratory suppression and were more likely to result in death. Despite REM sleep being a time when seizures do not typically occur spontaneously, when seizures were forced to occur during REM sleep, they were invariably fatal in this model. An examination of baseline breathing revealed that mice that died following a seizure had increased baseline respiratory rate variability compared with those that did not die. These data demonstrate that sleep, especially REM sleep, can be a dangerous time for a seizure to occur. These data also demonstrate that there may be baseline respiratory abnormalities that can predict which individuals have higher risk for seizure-induced death.

Severe epistaxis due to aberrant vasculature in a patient with STAT-1 mutation.

BACKGROUND: Signal transducer and activator 1 (STAT-1) mutations are rare and have been implicated in combined immunodeficiency, enhanced tumorigenesis, and vascular defects. METHODS: A 60-year-old woman with a novel STAT-1 mutation and resulting immunodeficiency, squamous cell carcinoma, and vascular disease presented with profuse epistaxis secondary to rupture of an aberrant artery that she developed in part because of this mutation. After unsuccessful posterior packing, embolization was initiated but subsequently aborted because of a bovine origin carotid artery and a history of multiple carotid dissections. RESULTS: After repeat posterior packing, hemostasis was achieved. No additional episodes of epistaxis occurred in the subsequent 13 months. CONCLUSION: Vascular anomalies can present challenges in epistaxis management. In patients with conditions known to cause vascular anomalies, it is critical to obtain vascular imaging before intervention.

Serotonin neurones have anti-convulsant effects and reduce seizure-induced mortality.

Sudden unexpected death in epilepsy (SUDEP) is the leading cause of death in patients with refractory epilepsy. Defects in central control of breathing are important contributors to the pathophysiology of SUDEP, and serotonin (5-HT) system dysfunction may be involved. Here we examined the effect of 5-HT neurone elimination or 5-HT reduction on seizure risk and seizure-induced mortality. Adult Lmx1b(f/f/p) mice, which lack >99% of 5-HT neurones in the CNS, and littermate controls (Lmx1b(f/f)) were subjected to acute seizure induction by maximal electroshock (MES) or pilocarpine, variably including electroencephalography, electrocardiography, plethysmography, mechanical ventilation or pharmacological therapy. Lmx1b(f/f/p) mice had a lower seizure threshold and increased seizure-induced mortality. Breathing ceased during most seizures without recovery, whereas cardiac activity persisted for up to 9 min before terminal arrest. The mortality rate of mice of both genotypes was reduced by mechanical ventilation during the seizure or 5-HT2A receptor agonist pretreatment. The selective serotonin reuptake inhibitor citalopram reduced mortality of Lmx1b(f/f) but not of Lmx1b(f/f/p) mice. In C57BL/6N mice, reduction of 5-HT synthesis with para-chlorophenylalanine increased MES-induced seizure severity but not mortality. We conclude that 5-HT neurones raise seizure threshold and decrease seizure-related mortality. Death ensued from respiratory failure, followed by terminal asystole. Given that SUDEP often occurs in association with generalised seizures, some mechanisms causing death in our model might be shared with those leading to SUDEP. This model may help determine the relationship between seizures, 5-HT system dysfunction, breathing and death, which may lead to novel ways to prevent SUDEP.