Predicting Lung Function Following Lobectomy: A New Method to Adjust for Inherent Selection Bias

Area of Science:

• Pulmonary Medicine
• Surgical Oncology
• Biostatistics

Background:

• Overestimation of post-lobectomy lung function can lead to unnecessary surgeries.
• Previous studies on post-lobectomy lung function are biased as they only included surgical patients.
• This selection bias results in inaccurate prediction of bias and inappropriate thresholds for lobectomy eligibility.

Purpose of the Study:

• To demonstrate and adjust for selection bias in lung function prediction.
• To obtain accurate estimates of prediction bias and 95% limits of agreement.
• To establish adjusted threshold values for determining lobectomy eligibility and the need for exercise testing.

Main Methods:

• Retrospective study of patients evaluated for lobectomy.
• Utilized multiple imputations to estimate postoperative outcomes for non-surgical candidates.
• Combined data from surgical and non-surgical patients to correct for selection bias.
• Applied Bland-Altman method and bivariate normal distribution for threshold determination.

Main Results:

• Prediction bias estimates changed significantly after controlling for selection bias.
• Required predicted FEV1 > 46% and DLCO ≥53% for target postoperative FEV1 > 30% and DLCO ≥30%.
• Implementing these adjusted thresholds would alter management in 17% of cases compared to current guidelines.

Conclusions:

• Selection bias significantly affects the accuracy of lung function prediction but is correctable.
• Current threshold values for determining lobectomy eligibility require reassessment based on corrected data.