Abstract 85P
Background
Immune checkpoint inhibitors (ICIs) have revamped the clinical outcomes of patients (pts) affected by advanced non-oncogene addicted non-small cell lung cancer (NSCLC). The aim of our study is to develop an ensemble clustering algorithm (Clust) integrated with a logistic regression model (Pro) to prognostically stratify NSCLC pts treated with ICIs (NSCLC-Pro ClustAI).
Methods
Data extraction involved, retrospectively, all consecutive pts with advanced NSCLC treated with ICIs at Department of Medical Oncology, Ancona, Italy. Baseline features included clinicopathological variables and comorbidities commonly available in daily clinical practice. An unsupervised clustering analysis was used to identify groups within the dataset that hold prognostic significance. Clust was built by stacking both the K-means algorithm and the Gaussian Mixture Model. Subsequently, Pro was used to predict clusters’ label. The metrics used to evaluate the average performance of Pro on test sets were ACCuracy (ACC) and Area Under the Curve (AUC). The model was developed in Python on-cloud using the Google Colab service.
Results
A total of 89 NSCLC pts with complete data available were included in the final analysis. The two generated clusters were: cluster 1 (n=54) and cluster 2 (n=35) (Table). Considering clinical outcomes, median progression free survival was 9.96 months for cluster 1 and 3.78 months for cluster 2 (p = 0.007), and median overall survival was 14.25 months for cluster 1 and 9.14 months for cluster 2 (p=0.05). The average ACC and AUC across all splits achieved by Pro model for pts classification into clusters were 0.978 and 0.981, respectively. From the feature ranking, it emerged that the one with higher importance was CONUT score.
Table: 85P
Dataset stratification
| Cluster 1 (n=54) % | Cluster 2 (n=35) % | p | |
| Age* | 67.0 | 73.0 | <0.001 |
| NLR* | 4.9 | 6.1 | 0.28 |
| BMI* | 24.4 | 25.4 | 0.03 |
| Sex | |||
| Male | 68.5 | 82.9 | 0.13 |
| Female | 31.5 | 17.1 | |
| ICI | |||
| Monotherapy | 51.9 | 100.0 | <0.001 |
| Combination | 48.1 | 0.0 | |
| Therapy line | |||
| 1st | 83.3 | 91.4 | 0.61 |
| ≥2 | 16.7 | 8.6 | |
| Metastatic sites | |||
| 0-1 | 50.6 | 54.3 | 0.67 |
| ≥2 | 49.4 | 45.7 | |
| ECOG PS | |||
| ≥2 | 3.7 | 2.9 | 0.83 |
| 0-1 | 96.3 | 97.1 | |
| Smoking | |||
| No | 20.4 | 5.7 | 0.06 |
| Yes | 79.6 | 94.3 | |
| Cardiovascular disease | |||
| No | 83.3 | 71.4 | 0.18 |
| Yes | 16.7 | 28.6 | |
| Diabetes mellitus | |||
| No | 87.0 | 57.1 | 0.001 |
| Yes | 13.0 | 42.9 | |
| Hypertension | |||
| No | 61.1 | 14.3 | <0.001 |
| Yes | 38.9 | 85.7 | |
| Statin use | |||
| No | 83.3 | 28.6 | <0.001 |
| Yes | 16.7 | 71.4 | |
| CONUT score | |||
| Low | 88.9 | 22.9 | <0.001 |
| High | 11.1 | 77.1 | |
| PD-L1 | |||
| 0% | 24.1 | 2.9 | |
| 1-49% | 40.7 | 2.9 | <0.001 |
| ≥50% | 35.2 | 94.1 |
∗Values expressed as median
Conclusions
By using easy to obtain and reproducible clinicopathological features, we demonstrated that NSCLC-Pro ClustAI is a promising and accurate prognostic model for stratifying NSCLC pts receiving ICIs.
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
L. Cantini: Financial Interests, Institutional, Ownership Interest: Fortrea Inc.; Financial Interests, Institutional, Stocks/Shares: Fortrea Inc. R. Berardi: Financial Interests, Institutional, Advisory Role: AstraZeneca, Boehringer Ingelheim, Novartis, Merck, Otsuka, Eli Lilly, Roche. All other authors have declared no conflicts of interest.