Building and Validating a Lymphocyte Nadir Based Model to Predict Survival in Patients with Limited Stage-Small Cell Lung Cancer

Abstract

Background: Increasing evidence indicates host immunity participate in cancer progression and metastases. It has been reported that the lymphocyte nadir is an independent prognostic marker for survival in various cancers. We have previously reported a survival significance of baseline lymphocyte in limited stage small cell lung cancer (LS-SCLC). Here we hypothesize that treatment induced lymphopenia, like the lymphocyte nadir during the course of treatment, in combination of clinical factors can predict survival better than conventional models in patients with LS-SCLC. Method: This is a retrospective study of 616 patients from a single Institution. Consecutive patients with LS-SCLC treated with thoracic radiation (with or without concurrent chemotherapy) from 2013 to 2017 were included. Additional eligibility included availability of complete-blood-count data from baseline and at least two time points during the course of radiotherapy. These 616 patients were randomly divided into a training dataset (n=308) and a validation dataset (n=308). The primary endpoint was overall OS. Univariate proportional hazard (PH) cox model was used to assess potential clinicopathological predictors on OS. The multivariable Cox PH model was constructed by the forward selection. According to the final Cox model built using significant variables from training dataset, we calculated the risk score for every patient and validate the predictive valuable of the risk score on OS in the validation set. Result: Under univariate analysis, younger age (HR 1.02 per 1 yr, 95%CI 1.006-1.043, p=0.008), female gender (HR 1.40, 95%CI 0.95-2.07, p=0.09), earlier stage (stage I-II vs stage III, HR 2.2, 95%CI 1.11-4.33, p=0.02), concurrent chemotherapy (concurrent vs. not, HR 0.61, 95%CI 0.42-0.88, p=0.01) and a higher lymphocyte nadir (HR 0.48, 95% CI 0.20-1.14, per 103 lymphocytes/μL, p=0.097) was significantly associated with increased OS in the training dataset. Using lymphocyte nadir in combination of significant clinical factors from univariate analysis, we developed a multivariable Cox PH model (lymphocyte nadir: HR 0.39, 95% CI 0.16-0.99, per 103 lymphocytes/μL, p=0.048) with concordance (C)-index of 0.63. In the validation dataset, the multivariable model revealed that lymphocyte nadir had a borderline significance on OS (HR 0.45, 95% CI 0.19-1.06, per 103 lymphocytes/μL, p=0.067) with a comparable c-index of 0.60. Moreover, the risk score calculated using the coefficients from the final Cox model built using the training dataset remained to be a significant predictor for OS (HR 2.04, 95% CI 1.36-3.07, per 1 risk score increase, p<0.0001) in validation dataset. Conclusion: This may be the first study validated a survival predictive model based on lymphocyte nadir in a large sample of patients with LS-SCLC. Should it should be validated in an external dataset, this model might provide some prediction for each patient and provide an opportunity to individualize treatment based on the individual’s survival probability.