Data analytics for real-world data integration in TKI-treated NSCLC patients using electronic health records

Abstract

Background

Real-world data (RWD) are routinely collected in clinical practice during therapeutic interventions. Data warehouses (DWHs) represent the primary source of RWD in which electronic health records (EHRs) can be rapidly analyzed via natural language processing. This study illustrates an analytic framework that systematically exploits RWD and methods to generate real-world evidence (RWE) about innovative cancer drugs. The framework has been applied to investigate real-world treatment patterns and clinical outcomes of patients with advanced non-small-cell lung cancer (aNSCLC) treated with tyrosine kinase inhibitors (TKIs).

Materials and methods

Data from a cohort of 190 epidermal growth factor receptor-positive mutation (EGFRm) patients with aNSCLC were retrospectively collected in an Italian cancer institute between 2014 and 2022. Patients were treated in first-line (1L) with osimertinib or other TKIs (non-osimertinib). A text-mining algorithm was implemented to retrieve RWD from EHRs. Survival endpoints were median time to treatment discontinuation (mTTD) and median overall survival (mOS) estimated with Kaplan–Meier curves. Time-dependent multivariate Cox analysis was carried out to overcome immortal time bias.

Results

Approximately 38% of patients received 1L osimertinib, while the remaining 62% received previous-generation TKIs. Longer mTTD [15 months; 95% confidence interval (CI) 11.9-26.4 months] was found for patients treated with 1L osimertinib compared with non-osimertinib (10 months; 95% CI 7.9-13.1 months). In multivariate analysis, osimertinib was an independent protective factor regardless of bone and brain metastases and local radiotherapy. mOS was 27 months (95% CI 21.4-39.5 months) for osimertinib versus 20.2 months (95% CI 17.6-23.1 months) for non-osimertinib.

Conclusions

Data analytics frameworks are useful tools to integrate RWE in cancer research and data-driven models are suitable to process large amounts of RWD. This study demonstrates that real-world treatment patterns and outcomes of TKIs are comparable with those found in both clinical trials and other real-world studies. RWE studies can support clinicians in investigating the best treatment strategy and decision makers to drive new health policies.