Incidence, morbidity, and mortality of pulmonary complications in free flap reconstruction: limitations of predictive models

Authors: Abdul-Rahman N-H, Harris M, Bottegal M, Sridharan S, Spector M, Snyderman C.

Affiliation: University of Pittsburgh School of Medicine; Department of Otolaryngology.

Journal: Otolaryngology–Head and Neck Surgery, Sept. 2025.

PMID: 40407209

Key takeaways

• Clinically significant postoperative pulmonary complications (PPCs) occurred in 27% of head and neck microvascular free flap (MVFF) patients.

• Independent predictors of PPCs: advanced tumor stage (OR 1.29), longer surgery duration (OR 1.08 per hour), and postoperative hematoma (OR 2.98).

• PPCs markedly worsened survival: HR 3.94 for mortality; 1-year survival 75.6% with PPCs vs 89.4% without.

• This study built and internally validated a head and neck free flap–specific risk calculator (nomogram; AUC 0.65), outperforming generic tools

  • ARISCAT (general preop lung‑risk score): AUC 0.51

  • Gupta (NSQIP pneumonia/respiratory failure): AUC 0.45

Background

Postoperative pulmonary complications are frequent after major oncologic head and neck surgery but definitions and prediction tools have been inconsistent for this population.

Objective

Quantify PPC incidence, identify risk factors and survival impact after MVFF, evaluate ARISCAT and Gupta models, and propose an MVFF-specific nomogram.

Methods

• Design/setting: Retrospective review at a tertiary academic center. Level of evidence: III.

• Cohort: 638 MVFF reconstructions (August 2019–May 2024).

• Exclusions: Prior head and neck radiation therapy (RT) patients were excluded.

• Exposure variables: Pre-, intra-, and postoperative factors (e.g., tumor stage, operative time, estimated blood loss, hematoma, transfusion).

• Outcome: PPCs within 30 days; graded with Modified Clavien-Dindo scale (grades 2–5 = clinically significant).

  • Grade 1 (mild): Hypoxia or atelectasis only; no treatment required.

  • Grade 2 (moderate): Bronchospasm; pleural effusion not needing invasive intervention; or atelectasis needing intervention/with effusion or hypoxia.

  • Grade 3 (severe): Pneumonia, lung abscess, or significant atelectasis needing invasive intervention and/or with respiratory failure or pneumonia.

  • Grade 4 (life-threatening): ARDS, respiratory failure, or cardiopulmonary collapse.

  • Grade 5: Death from a PPC.

• Statistics: Univariate and multivariable logistic regression for PPCs; survival analysis (Cox) for mortality; model performance by AUC; creation of a nomogram with Youden-index cutoff.

Results

• Incidence: Grades 2–5 PPCs occurred in 27% of patients.

• Univariate signals: Longer surgery (mean 10.06 ± 2.67 h; P = .006), estimated blood loss ≥200 mL (P = .006), stage III/IV (P = .013), hematoma (P < .001), postoperative transfusion (P = .037).

• Multivariable predictors of grades 2–5 PPCs:

  • Tumor stage: OR 1.29 (95% CI 1.06–1.57; P = .012)

  • Surgery duration: OR 1.08 (95% CI 1.01–1.17; P = .031)

  • Hematoma: OR 2.98 (95% CI 1.50–5.94; P = .002)

• Tumor location: No association with PPC risk on univariate testing (e.g., oral cavity vs laryngeal/hypopharyngeal; cutaneous vs aerodigestive).

• Resource use and adverse outcomes: ICU transfer 17.7% with PPCs vs 3.6% without (P < .001); length of stay 12.4 days overall (range 2–87).

• Mortality: In-hospital 1.4% (n=9)—all with grade 5 PPCs. One-year mortality 13.48% overall; deaths at 1 year 24.1% with PPCs vs 9.6% without (P < .001). PPC independently predicted death (HR 3.94, 95% CI 1.69–9.22; P = .002).

• Modeling: MVFF-specific model AUC 0.65; ARISCAT ~0.51; Gupta ~0.45. Nomogram high-risk cutoff ≥27% predicted probability.

Conclusion

Clinically significant PPCs are common after MVFF for head and neck cancer and are strongly associated with ICU use, longer hospitalization, and higher short- and long-term mortality; general PPC prediction scores underperform, supporting a tailored, head and neck–specific risk model.

Strengths and limitations

• Strengths

  • Clinically meaningful PPC definition (grades 2–5) tied to treatment and outcomes.

  • Large, contemporary MVFF cohort with linkage to ICU use, length of stay, and 1‑year mortality.

  • Head‑and‑neck–specific risk calculator demonstrates superiority over generic tools (ARISCAT, Gupta).

  • Predictors are actionable (tumor stage, operative duration, postoperative hematoma).

• Limitations

  • Single‑center, retrospective design limits external validity.

  • Prior head and neck RT patients were excluded, narrowing generalizability to a high‑risk population.

  • Model performance is modest (AUC ~0.65) with likely high specificity and lower sensitivity; thresholding may miss at‑risk patients.

  • Mixed timing of predictors: inclusion of postoperative hematoma limits preoperative triage utility.

  • Coarse covariate granularity: airway strategy (planned tracheostomy vs early extubation), bony vs soft‑tissue reconstruction, extent of neck dissection, and ERAS compliance not fully explored.

  • Outcome adjudication details (e.g., pneumonia criteria, inter‑rater reliability) not described in depth.

Clinical relevance

• Make pulmonary prevention a core reconstructive objective. Standardize airway plans (planned tracheostomy vs early extubation), pulmonary hygiene, early mobilization, and incentive spirometry alongside flap checks.

• Shorten and stage operations thoughtfully. Use two‑team approaches, efficient sequencing, and avoid unnecessary delays to reduce operative time.

• Prioritize meticulous hemostasis and early detection of bleeding. Given hematoma’s association with PPCs and mortality, reinforce OR and PACU protocols for recognition and intervention.

• Use risk tools judiciously. The study’s nomogram can guide level of care (e.g., step‑down vs ICU) and counseling, but do not rely on ARISCAT/Gupta; combine the nomogram with clinical judgment, especially for advanced stage tumors and very long cases.

• Triage resources. Consider closer postoperative monitoring and dedicated respiratory therapy for patients with advanced tumor stage, prolonged operative duration, or intraoperative events.