Predicting reoperation and readmission for head and neck free flap patients using machine learning.

BACKGROUND: To develop machine learning (ML) models predicting unplanned readmission and reoperation among patients undergoing free flap reconstruction for head and neck (HN) surgery. METHODS: Data were extracted from the 2012-2019 NSQIP database. eXtreme Gradient Boosting (XGBoost) was used to develop ML models predicting 30-day readmission and reoperation based on demographic and perioperative factors. Models were validated using 2019 data and evaluated. RESULTS: Four-hundred and sixty-six (10.7%) of 4333 included patients were readmitted within 30 days of initial surgery. The ML model demonstrated 82% accuracy, 63% sensitivity, 85% specificity, and AUC of 0.78. Nine-hundred and four (18.3%) of 4931 patients underwent reoperation within 30 days of index surgery. The ML model demonstrated 62% accuracy, 51% sensitivity, 64% specificity, and AUC of 0.58. CONCLUSION: XGBoost was used to predict 30-day readmission and reoperation for HN free flap patients. Findings may be used to assist clinicians and patients in shared decision-making and improve data collection in future database iterations.

Preoperative Circulating Tumor HPV DNA and Oropharyngeal Squamous Cell Disease.

Importance: The utility of preoperative circulating tumor tissue-modified viral human papillomavirus DNA (TTMV-HPV DNA) levels in predicting human papillomavirus (HPV)-associated oropharyngeal squamous cell carcinoma (HPV+ OPSCC) disease burden is unknown. Objective: To determine if preoperative circulating tumor HPV DNA (ctHPVDNA) is associated with disease burden in patients with HPV+ OPSCC who have undergone transoral robotic surgery (TORS). Design, Setting, and Participants: This cross-sectional study comprised patients with HPV+ OPSCC who underwent primary TORS between September 2021 and April 2023 at one tertiary academic institution. Patients with treatment-naive HPV+ OPSCC (p16-positive) and preoperative ctHPVDNA levels were included, and those who underwent neck mass excision before ctHPVDNA collection were excluded. Main Outcomes and Measures: The main outcome was the association of increasing preoperative ctHPVDNA levels with tumor size and lymph node involvement in surgical pathology. The secondary outcome was the association between preoperative ctHPVDNA levels and adverse pathology, which included lymphovascular invasion, perineural invasion, or extranodal extension. Results: A total of 70 patients were included in the study (65 men [93%]; mean [SD] age, 61 [8] years). Baseline ctHPVDNA levels ranged from 0 fragments/milliliter of plasma (frag/mL) to 49 452 frag/mL (median [IQR], 272 [30-811] frag/mL). Overall, 58 patients (83%) had positive results for ctHPVDNA, 1 (1.4%) had indeterminate results, and 11 (15.6%) had negative results. The sensitivity of detectable ctHPVDNA for identifying patients with pathology-confirmed HPV+ OPSCC was 84%. Twenty-seven patients (39%) had pathologic tumor (pT) staging of pT0 or pT1, 34 (49%) had pT2 staging, and 9 patients (13%) had pT3 or pT4 staging. No clinically meaningful difference between detectable and undetectable preoperative ctHPVDNA cohorts was found for tumor size or adverse pathology. Although the median preoperative ctHPVDNA appeared to be higher in pT2 through pT4 stages and pN1 or pN2 stages, effect sizes were small (pT stage: η2, 0.002 [95% CI, -1.188 to 0.827]; pN stage: η2, 0.043 [95% CI, -0.188 to 2.600]). Median preoperative log(TTMV-HPV DNA) was higher in active smokers (8.79 [95% CI, 3.55-5.76]), compared with never smokers (5.92 [95% CI, -0.97 to 1.81]) and former smokers (4.99 [95% CI, 0.92-6.23]). Regression analysis did not show an association between tumor dimension or metastatic lymph node deposit size and preoperative log(TTMV-HPV DNA). After univariate analysis, no association was found between higher log(TTMV-HPV DNA) levels and adverse pathology. Conclusions and Relevance: In this cross-sectional study, preoperative ctHPVDNA levels were not associated with disease burden in patients with HPV+ OPSCC who underwent TORS.

Transoral robotic surgery with free flap reconstruction: Functional outcomes of 241 patients at a single institution.

BACKGROUND: Transoral robotic surgery (TORS) for oropharyngeal malignancy optimizes oncologic outcomes while preserving functionality. This study identifies patterns of functional recovery after TORS with free flap reconstruction (FFR). METHODS: Retrospective cohort study at a tertiary care center of patients with primary oropharyngeal tumors treated with TORS with FFR between 2010 and 2022. Patients were categorized into: adjuvant chemoradiation or radiation, or no adjuvant therapy (NAT). Functional outcomes were measured by functional oral intake scale (FOIS). RESULTS: 241 patients were included. FOIS declined at first postoperative appointment (median = 7.0 to 2.0, IQR = [7.0, 7.0], [2.0, 4.0]), and progressively improved to 6.0 (5.0, 6.0) after 1 year, with NAT having the highest FOIS (7.0, p < 0.05). Predictors of poor long-term FOIS included RT and hypoglossal nerve (CN XII) involvement (p < 0.05). CONCLUSIONS: TORS with FFR leads to good long-term function with minimal intake restrictions. Radiation therapy and CN XII involvement increase risk of worse functional outcomes.