Original ResearchClinical decision support algorithm based on machine learning to assess the clinical response to anti–programmed death-1 therapy in patients with non–small-cell lung cancer
Graphical abstract
Introduction
Immune checkpoint inhibitors (ICIs), including programmed cell death protein 1/programmed death-ligand 1 (PD-1/PD-L1) inhibitors, have resulted in prolonged survival and were approved as first- and second-line therapies in patients with recurrent/metastatic non–small-cell lung cancer (NSCLC) [1,2]. PD-L1 protein expression has been approved as a biomarker for investigating the efficacy of PD-1/PD-L1 inhibitors. PD-L1 expression is enriched in anti–PD-1/PD-L1 inhibitor therapy responders [3,4]. Nevertheless, fewer than 30% of patients with NSCLC respond to anti–PD-1 inhibitors [5]. Moreover, a substantial proportion of PD-L1–positive patients show no response to therapy, whereas a subset of PD-L1–negative patients do show. Therefore, PD-L1 expression alone may not comprehensively reflect the complexity of the tumour microenvironment and response to PD-1 inhibitors.
With increasing use of anti–PD-1 therapy, various clinical characteristics related to treatment response in patients with NSCLC have been identified, such as neutrophil/lymphocyte ratio before immunotherapy, smoking history, performance status, sex, the presence of metastases and driver mutations and pathology [[6], [7], [8], [9], [10], [11], [12], [13], [14]]. These factors are routinely examined in clinical practice before anti–PD-1 therapy and are collected in electronic medical records (EMRs). However, none of the clinical factors can accurately predict the response to anti–PD-1 therapy; thus, a model integrating these factors is needed.
Recently, machine learning (ML)–based methods have been developed to predict disease progression and treatment response in various diseases [15,16]. ML tools can identify key features from complex data sets associated with a specific purpose and interest. The ML techniques, multilayer neural network (MNN), ensemble learning, support vector machine (SVM) and penalised regression, have been widely used in recent studies for developing predictive models to facilitate effective and accurate decision-making [8,17,18]. However, ML has not yet been extensively studied or used in ICI-treated patients with cancer.
This study explored and validated various predictive algorithms using an ML approach based on the clinicopathological factors of anti‒PD-1 therapy‒treated patients with NSCLC from prospectively collected data obtained from EMRs. We aimed to establish a clinical decision support system for prescribing anti‒PD-1 therapy in patients with NSCLC using clinicopathological information routinely collected during clinical practice.
Section snippets
Patients
Patients with histologically confirmed stage IV NSCLC who were treated with anti‒PD-1 therapy (nivolumab 2 mg/kg every 2 weeks or pembrolizumab 200 mg fixed dose every 3 weeks) at Yonsei Cancer Center (Seoul, Korea) between March 2014 and April 2020 were included (n = 192). Patients were divided into a discovery cohort—patients consecutively enrolled from March 2014 to January 2018—to explore the optimal algorithm and the validation set—patients enrolled after January 2018—to confirm the
Patient characteristics
In total, 192 patients with advanced NSCLC treated with anti‒PD-1 therapy were included for analysis. The discovery (n = 142) and validation cohorts (n = 50) showed no significant differences in patient characteristics, including age, performance status, driver mutation status, line of therapy and site of baseline metastasis (Table 1).
Treatment outcomes in the discovery set and role of PD-L1 as a predictive biomarker
In the discovery set, 56 responders and 86 non-responders were classified based on definitions. The clinical response of total patients is shown in Fig. 1A. The
Discussion
Immunohistochemistry(IHC)-based PD-L1 expression is the only approved predictive marker for anti‒PD-1 therapy, but its accuracy is not sufficient to discriminate the responders from non-responders. The associations of various clinical factors or blood test values in routine practice with treatment outcomes after anti‒PD-1 therapy are increasingly being reported. However, the practical clinical application of these findings is limited because these clinical factors and blood test results have
Authors’ contributions
Beung-Chul Ahn: Conceptualisation, Methodology, Resources, Investigation, Formal analysis, Roles/Writing - original draft, Writing - review & editing Jea-Woo So: Conceptualisation, Methodology, Resources, Investigation, Formal analysis, Roles/Writing - original draft, Writing - review & editing Byoung Chul Cho: Conceptualisation, Methodology, Funding acquisition, Project administration, Resources, Roles/Writing - original draft, Writing - review & editing Hye Ryun Kim: Conceptualisation,
Availability of data and material
All data generated or analysed during this study are included in this published article and its supplementary information files.
Conflict of interest statement
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Funding
This research was supported by the Bio & Medical Technology Development Program of the National Research Foundation funded by the Ministry of Science and ICT [grant numbers NRF-2019M3A9B6065231, 2017M3A9E8029717, 2017M3A9E9072669]. This study was supported by a Dongin Sports research grant of Yonsei University College of Medicine (6-2019-0128).
References (29)
Pretreatment neutrophil-to-lymphocyte ratio as a marker of outcomes in nivolumab-treated patients with advanced non-small-cell lung cancer
Lung Canc
(2017)Neutrophil-to-Lymphocyte ratio (NLR) and Platelet-to-Lymphocyte ratio (PLR) as prognostic markers in patients with non-small cell lung cancer (NSCLC) treated with nivolumab
Lung Canc
(2017)Cancer immunotherapy efficacy and patients' sex: a systematic review and meta-analysis
Lancet Oncol
(2018)Liver metastasis is associated with poor progression-free survival in patients with non–small cell lung cancer treated with nivolumab
J Thorac Oncol
(2017)Italian nivolumab expanded access Program in nonsquamous non-small cell lung cancer patients: results in never-smokers and EGFR-mutant patients
J Thorac Oncol
(2018)Machine learning-based predictors for immune checkpoint inhibitor therapy of non-small-cell lung cancer
Ann Oncol
(2019)Machine learning applications in cancer prognosis and prediction
Comput Struct Biotechnol J
(2015)The blockade of immune checkpoints in cancer immunotherapy
Nat Rev Canc
(2012)Predictive correlates of response to the anti-PD-L1 antibody MPDL3280A in cancer patients
Nature
(2014)- et al.
Checkpoint inhibitors for malignant melanoma: a systematic review and meta-analysis
Clin Cosmet Invest Dermatol
(2017)
Recent development in clinical applications of PD-1 and PD-L1 antibodies for cancer immunotherapy
J Hematol Oncol
Pembrolizumab for the treatment of non-small-cell lung cancer
N Engl J Med
Impact of smoking on efficacy of PD-1/PD-L1 inhibitors in non-small cell lung cancer patients: a meta-analysis
Onco Targets Ther
Prognostic clinical factors in patients affected by non-small-cell lung cancer receiving Nivolumab
Expet Opin Biol Ther
Cited by (17)
AI/ML advances in non-small cell lung cancer biomarker discovery
2023, Frontiers in Oncology
- 1
These authors contributed equally.