Background: Few small sample size studies have reported lymphocyte count was prognostic for survival in small-cell lung cancer (SCLC). This study aimed to validate this finding, to build prediction model for overall survival (OS) and to study whether novel models that combine lymphocyte-related variables can predict OS more accurately than a conventional model using clinical factors alone in a large cohort of limited-stage SCLC patients. Methods: This study enrolled 544 limited-stage SCLC patients receiving definitive chemo-radiation with pre-radiotherapy lymphocyte-related variables including absolute lymphocyte count (ALC), platelet-to-lymphocyte ratio (P/L ratio), neutrophil-to-lymphocyte ratio (N/L ratio), and lymphocyte-to-monocyte ratio (L/M ratio). The primary endpoint was OS. These patients were randomly divided into a training dataset (n=274) and a validation dataset (n=270). Multivariate survival models were built in the training dataset, and the performance of these models were further tested in the validation dataset using the concordance index (C-index). Results: The median follow-up time was 36 months for all patients. In the training dataset, univariate analysis showed that ALC (P=0.020) and P/L ratio (P=0.023) were significantly correlated with OS, while L/M ratio (P=0.091) and N/L ratio (P=0.436) were not. Multivariate modeling demonstrated the significance of ALC (P=0.063) and P/L ratio (P=0.003), and the improvement for OS prediction in combined models with the addition of ALC (C-index =0.693) or P/L ratio (C-index =0.688) over the conventional model (C-index =0.679). The validation dataset analysis confirmed a modest improvement of C-index with the addition of ALC or P/L ratio. All these models showed reasonable discriminations and calibrations. Conclusions: This study validated the significant value of pre-radiotherapy ALC and P/L ratio on OS in limited-stage SCLC. The combined model with ALC or P/L ratio showed additional OS prediction values than the conventional model with clinical factors alone.
All Science Journal Classification (ASJC) codes