Baseline predictors of post-treatment diffusion capacity after neoadjuvant chemo-immunotherapy in non-small-cell lung cancer: development and external validation of a clinical prediction model
Patrick Deniz Hurley, Guillermo Rodriguez Dávila, David Gómez de Antonio, Maria-Teresa Gomez-Hernandez, Joshil Lodhia, Marcelo Jimenez, Marco Nardini, Nuria Novoa, Peter Tcherveniakov, Elaine Teh, Alessandro BrunelliAbstract
Objectives
Routine repetition of pulmonary function tests (PFTs) following neoadjuvant chemo-immunotherapy is standard but may be redundant in selected patients. We aimed to develop and externally validate a multivariable model to predict post-treatment carbon monoxide lung diffusion capacity (DLCO), with the objective of identifying patients at low predicted risk of clinically important post-treatment impairment in whom selective omission of repeat testing could potentially be considered.
Methods
A predictive model was derived from 106 patients with locally advanced non-small cell lung cancer (NSCLC) undergoing neoadjuvant chemo-immunotherapy. Multivariable regression with backward stepwise elimination was used to test candidate predictors. The model was validated on an external cohort of 60 patients. Agreement was assessed using locally weighted scatterplot smoothing (Lowess) curves and misclassification rates relative to a clinically relevant threshold of DLCO<60%.
Results
The final model retained four independent predictors: Baseline DLCO (p < 0.001), Age (p < 0.001), Programmed death-ligand 1 (PD-L1) expression (p = 0.043), and Smoking status (p = 0.10). Applied to the external validation cohort (n = 60), the model demonstrated a median predicted DLCO of 66% (IQR 59-79) versus an observed DLCO of 73% (IQR 59-83), and Lowess curves indicated a stable relationship across the data range. The model correctly identified 100% of patients with an observed post-treatment DLCO<60% by using a predicted threshold of < 65%. All patients with a predicted DLCO>65% had an observed DLCO>60%.
Conclusions
We developed and externally validated a clinical model that reliably predicts post-neoadjuvant DLCO. The model may help identify asymptomatic patients at low likelihood of a post-treatment DLCO <60% in whom repeat PFT assessment could potentially be omitted. Prospective validation is required before this approach can be recommended in routine clinical practice.