This study analyzed patients who underwent uniportal video-assisted thoracoscopic surgery (VATS) segmentectomy from January 2015 to September 2023. The procedures were performed through a single incision with dedicated instruments, and the inflation-deflation technique or indocyanine green was used to define the intersegmental plane. The broncovascular structures and pulmonary parenchyma were divided with motorized endo staplers. One single chest drain was inserted and connected to a digital drainage system.
Segmentectomies were classified as simple or complex based on a previously published classification. Superior segmentectomy of the lower lobes left upper division segmentectomy, lingulectomy, and basilar segmentectomies were simple. All other types of segmentectomies were considered complex.
Statistical analysis identified risk factors for prolonged air leakage (PAL). Variables assessed included patient demographics, surgical details, and clinical characteristics. The elastic net approach was used for predictive modeling, with balanced data sets for PAL and control patients.
Of 575 segmentectomies, PAL occurred in 88 patients (15.3%). The mean age was 64.8 years, and 57.9% were men. Twelve cases were converted to open thoracotomy, and three were extended to lobectomy. Patients with PAL had longer hospital stays and chest drain durations. COPD GOLD II and III were more common in the PAL group, and surgical revision was required in 13 patients. Smoking and lower FEV1 and DLCO were associated with PAL.
Lower lobe segment location and higher BMI reduced PAL risk, while additional wedge resection, liver disease, hypertension, and longer surgery time increased it. Two Predictive models showed moderate specificity and sensitivity of around 70%.
The study's findings can assist surgical planning and postoperative management to reduce PAL incidence.
Interview with Dr. med. Konstantinos Gioutsos (Bern)
What inspired you to conduct this study?
The increasing use of minimal-invasive segmentectomy in thoracic surgery and the known challenges of prolonged air leak (PAL), motivated us to investigate its risk factors. While segmentectomy offers advantages in preserving lung function, PAL remains a significant complication, leading to prolonged hospital stays and higher morbidity. Our goal was to identify modifiable factors that could help predict and reduce the incidence of PAL, ultimately improving patient outcomes and surgical efficiency.
Were there any unexpected findings?
Yes, one unexpected finding was the identification of segment II resections as a key risk factor for PAL. While upper lobe surgeries have been traditionally associated with higher PAL rates, our study highlighted that segment II resections were particularly prone to air leaks. Additionally, the relationship between diabetes and PAL was more pronounced than expected, suggesting that metabolic factors could play a significant role in post-surgical recovery.
What is the direct impact on the surgeon's work?
This study provides surgeons with a more profound understanding of the risk factors for PAL, enabling them to identify high-risk patients more accurately. For example, knowing that low BMI, segment 2 resections, increased packyears, and diabetes contribute to an increased risk of PAL can guide decision-making preoperatively. Surgeons might opt for more tailored intraoperative techniques, such as additional sealants or drainage strategies for patients with these risk factors, which could reduce the incidence of PAL and improve postoperative care.
What is your learning point from this project?
One major takeaway from this project was the importance of using multidimensional approaches – specifically, combining clinical risk factors with machine learning models to predict PAL. While statistical methods offered valuable insights, we found that incorporating machine learning could enhance the accuracy of PAL predictions. However, this approach still requires further validation and refinement. Furthermore, the differing results from various statistical analyses reflect the interactions between various clinical parameters.
Are there any subsequent projects planned?
Based on the findings, our next steps involve further refining the machine learning models to enhance their predictive power, ideally making them more applicable in clinical practice. We also evaluate conducting a prospective cohort study to validate the identified risk factors, particularly focusing on how adjusting preoperative and postoperative care for high-risk patients can reduce PAL rates.
