Performance of pre-flight assessment algorithms in interstitial lung disease validation cohorts

Dear Editor,

We read with interest the recent validation study by Bonthrone et al (2026), evaluating pre-flight algorithms for predicting hypoxic challenge test (HCT) outcomes in interstitial lung disease (ILD). ¹

We would like to highlight findings from our recently published independent external validation study of the original Barratt et al. ² pre-flight algorithm, ² in our ILD cohort at Royal Papworth Hospital. ³ In contrast to the findings reported by Bonthrone and colleagues, we observed substantially lower sensitivity of the combined PaO₂/TLCO algorithm for predicting HCT pass outcome (31.5%), despite similarly high specificity (96.7% vs 83.3%).

Our analyses suggested that the reduced sensitivity of the combined algorithm was largely attributable to the TLCO > 50% threshold, as a substantial proportion of patients passed HCT despite TLCO values below this cut-off. ³ In our cohort, our validation analyses identified resting PaO₂ > 9.30 kPa as the optimal threshold for predicting HCT pass outcome, demonstrating sensitivity of 81.1% and specificity of 72.5%, which exceeded the sensitivity observed using the combined PaO2/TLCO algorithm. We therefore proposed a simplified pre-flight assessment algorithm based solely on resting PaO2 threshold of 9.30 which had an 81% sensitivity for predicting a HCT pass.

Nevertheless, the study by Bonthrone et al. provides important additional external validation data by evaluating algorithm performance across clinically relevant pathways identifying patients likely to pass HCT, fail HCT, as well as assessing the modified BTS pre-flight pathway.

Taken together, these independent validation cohorts strengthen the evidence supporting practical pre-flight screening tools for ILD patients, particularly in settings where formal HCT is not readily available. Our findings suggested that resting PaO₂ alone may provide a simpler and more sensitive approach to pre-flight screening in ILD however, prospective validation against real-world flight outcomes remains essential before widespread implementation.