Association of clinical covariates and survival with the DELFI score

We performed univariate analyses comparing the distribution of the DELFI score to baseline clinical and laboratory covariates age, smoking, and serum inflammatory markers using a Wilcoxon rank sum test. In addition, we evaluated the relationship of the DELFI score and cancer risk with and without baseline covariates age, smoking, and sex using logistic regression.

To assess whether the DELFI score was associated with prognosis, we categorized high-risk lung cancer patients according to whether they were more likely to have cancer than not (DELFI score > 0.5). To assess whether this categorization was associated with survival among lung cancer patients in a univariate analysis, we used a log rank test to compare survival curves (Supplementary Table ⁶). While we did not optimize the 0.5 cutoff, we verified that inference from the log rank test was robust to other possible choices that would achieve higher or lower specificities, including a range of DELFI score cutoffs of 0.3–0.95. In addition to the univariate analysis, we evaluated whether the DELFI score was independently associated with cancer-specific survival in a multivariable Cox proportional hazards model that included age, histological subtypes of primary lung cancer, clinical staging, as well as treatment modality.