Machine Learning for Prediction of Immunotherapy Efficacy in Non-Small Cell Lung Cancer from Simple Clinical and Biological Data

Abstract

Background: Immune checkpoint inhibitors (ICIs) are now a therapeutic standard in advanced non-small cell lung cancer (NSCLC), but strong predictive markers for ICIs efficacy are still lacking. We evaluated machine learning models built on simple clinical and biological data to individually predict response to ICIs. Methods: Patients with metastatic NSCLC who received ICI in second line or later were included. We collected clinical and hematological data and studied the association of this data with disease control rate (DCR), progression free survival (PFS) and overall survival (OS). Multiple machine learning (ML) algorithms were assessed for their ability to predict response. Results: Overall, 298 patients were enrolled. The overall response rate and DCR were 15.3% and 53%, respectively. Median PFS and OS were 3.3 and 11.4 months, respectively. In multivariable analysis, DCR was significantly associated with performance status (PS) and hemoglobin level (OR 0.58, p < 0.0001; OR 1.8, p < 0.001). These variables were also associated with PFS and OS and ranked top in random forest-based feature importance. Neutrophil-to-lymphocyte ratio was also associated with DCR, PFS and OS. The best ML algorithm was a random forest. It could predict DCR with satisfactory efficacy based on these three variables. Ten-fold cross-validated performances were: accuracy 0.68 ± 0.04, sensitivity 0.58 ± 0.08; specificity 0.78 ± 0.06; positive predictive value 0.70 ± 0.08; negative predictive value 0.68 ± 0.06; AUC 0.74 ± 0.03. Conclusion: Combination of simple clinical and biological data could accurately predict disease control rate at the individual level.

Predicting survival and trial outcome in non-small cell lung cancer integrating tumor and blood markers kinetics with machine learning

“…Nevertheless, keeping in mind the objective to ultimately support decision making and patient stratification, a minimal (11 features), near-optimal, set of BSL variables was selected and denoted mBSL. It was defined as the first seven variables reaching the plateau (CRP, heart rate, neutrophils to lymphocytes ratio, neutrophils, lymphocytes to leukocytes ratio, liver metastases and ECOG score), complemented with four variables with established prognostic or predictive value and available in routine care: PD-L1 expression (50% cut-off) 3 , hemoglobin 30 , SLD 22 and LDH 31,32 . Applying stringent criteria to the RNAseq data (see methods), we selected 167 transcripts as candidates for final variable selection using Bolasso regression model to identify the optimal set of predictors 33 .…”

“…Baseline tumor mutational burden showed similar predictive value initially (AUC = 0.646) 11 , but led to disappointing results in a recent prospective study 46 and others found it to be more prognostic than predictive 47 . Baseline blood counts were previously reported to predict overall survival 44,[48][49][50] and treatment response (AUC = 0.74) 43 . The ROPRO score, derived from a large pan-cancer cohort and incorporating baseline clinical and biological data (27 variables) achieved a c-index of 0.69 and a 3-months AUC of 0.743 for prediction of survival in the OAK clinical trial 51 .…”

The 27-gene IO score is associated with efficacy of PD-1/L1 inhibitors independent of FGFR expression in a real-world metastatic urothelial carcinoma cohort

“…Multiple studies in bladder cancer have indicated that patients with an Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 2 have worse outcomes with ICI therapy than those with ECOG PS < 2 [34][35][36]. Prospective clinical trials with ECOG PS ≥ 2 are lacking, despite ICI therapy being an enticing option for the poor PS population.…”

Predictive Markers for Immune Checkpoint Inhibitors in Non-Small Cell Lung Cancer

“…For this, machine learning is another promising approach. Benzekry et al developed machine learning models by collecting clinical and hematological data and predicted the disease control rate of ICIs at the individual level [ 146 ]. Alternatively, instead of a simple combination of biomarkers, recent developments in high-throughput analyses such as next-generation sequencing and mass spectrometry may make it easier to identify a panel of biomarkers.…”