Abstract 1053P
Background
Immunotherapy has provided a significant survival benefit in up to 50% of non-small cell lung cancer (NSCLC) patients. The ongoing struggle is to accurately predict which patients will benefit. Prediction of non-response (i.e. progressive disease despite receiving immunotherapy) could enable an early decision about treatment continuation. Since serum tumor markers (STM) are known to reflect tumor mass, sequentially measured STM may provide predictive information. The aim of this study is to externally validate a previously developed STM model to predict non-response to immunotherapy soon after initiation.
Methods
In a cohort of 412 metastatic NSCLC patients treated with immunotherapy, STM were measured during the first six weeks of treatment. Non-response was defined as progressive disease on chest CT scan using RECIST criteria, clinical progressive disease, or death within six months after treatment initiation. Nine prediction methods with different combinations of STM were compared. At a specificity of 95%, the boosting model including Cyfra, CEA and NSE provided the most robust performance with a sensitivity of 74% in the training cohort and 58% in the test cohort. This model was validated in an external validation cohort of 208 metastatic NSCLC patients (Table). Table: 1053P
Training cohort (n= 412) | Validation cohort (n= 208) | |
Year of inclusion | 2014-2018 | 2019-2021 |
Mean age in years | 63 | 64 |
Treatment | ||
Mono immunotherapy | 412 (100%) | 100 (48%) |
Immuno-chemotherapy | 0 (0%) | 108 (52%) |
Lines of therapy prior to immunotherapy | ||
0 | 8 (2%) | 124 (60%) |
1 | 317 (77%) | 49 (24%) |
≥2 | 87 (21%) | 34 (16%) |
Non-response after 6 months (%) | 281 (68%) | 92 (44%) |
Results
In the external validation cohort, the Cyfra, CEA and NSE boosting model accurately predicted non-response in 46% of patients after six weeks of immunotherapy while maintaining a specificity of 92%. False positive patients could partially be explained by renal impairment or simultaneous other malignancies.
Conclusions
After six weeks of immunotherapy, our STM model is able to accurately predict non-response in nearly half of patients who will not benefit after six months. This model creates an early window of opportunity to switch to a possibly more effective second line treatment, lowering immune-related toxicity and reducing treatment costs.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.