Advancing Education, Research, and Quality of Care for the Head and Neck oncology patient.
Background: Postoperative pulmonary complications (PPCs) are prevalent following major oncologic head and neck surgery with microvascular free flap reconstruction, yet current predictive models like the ARISCAT and Gupta scores fail to accurately predict PPC in this population. Additionally, head and neck cancer studies only focus on a subset of PPCs.
Objective: This study evaluated the incidence of PPC and perioperative risk factors from the preoperative period to 30-days post-surgery. A secondary goal was to develop a predictive model through machine learning and assess the effectiveness of our model in identifying PPC.
Method: In this retrospective cohort study, we analyzed 638 patients who underwent concurrent oncologic head and neck surgery with microvascular free flap reconstruction at our institution from August 2019 to May 2024. Data were collected via chart review focusing on preoperative, intraoperative, and postoperative variables. PPC was defined as the development of hypoxia/increased oxygen requirement, atelectasis, bronchospasm, pleural effusion, aspiration pneumonitis, pneumonia, lung abscess, respiratory failure, acute respiratory distress syndrome (ARDS), and cardiopulmonary collapse within 30 days of surgery.
Results: Within our cohort of 638 patients, the average age was 62.6 ±11.9 years and majority were males 434 (68%). Average BMI was 27 ±12 kg/m2. The incidence of PPC was 46% (n=292/638) with severity grade distribution as follows: Grade1: 24%, grade 2: 18%, grade 3: 19%, and grade 4: 39%. Significant PPC predictors included duration of surgery (10.1 ± 2.7 hours, p<0.005), EBL >200mL (n=379, p=0.023), advanced tumor stage (p=0.007), and immediate postoperative complications such as hematoma (p<0.001), flap dehiscence (p=0.001), partial to total flap loss (p<0.001), other flap complication (p=0.043), salivary leak (p=0.001), postoperative transfusion (p=0.036), and ICU admission (p<0.001). Multivariable analysis revealed that the odds of PPC increased with tumor stage (OR= 1.33, 95% CI: 0.16 to 0.40) and postoperative hematoma (OR= 2.16, 95% CI 0.16 to 0.40). The mortality rate at 1 month was 2%, of which 82% were patients with PPC (n=9/11, p=0.016).
The ARISCAT and Gupta scores demonstrated poor ability to discriminate between those with and without PPC ([AUC= 0.54, 95% CI: 0.45 – 0.62] and [AUC=0.47, 95% Cl: 0.38 – 0.56] respectively). Neither score was a significant predictor of PPC, with ARISCAT showing sensitivity of 79%, and specificity of 24%, whiles Gupta had sensitivity of 76% and specificity of 17%.
Conclusion: The present study confirms a high incidence of PPC (46%) after microvascular free flap reconstruction. Current models like ARISCAT and Gupta inadequately predict PPC in our cohort, underscoring the need for a tailored head and neck specific predictive model. There is an opportunity to address risk factors to decrease PPC in this population.