OBJECTIVES : Following their platinum-combined induction bevacizumab and pemetrexed monotherapies are the suggested substances for maintenance therapy in patients with non-small cell lung carcinoma (NSCLC). Several studies state that the use of bevacizumab in combination may improve the overall survival (OS) and progression-free survival (PFS) compared to pemetrexed. Our aim was to perform a survival analysis based on real-world data to calculate and compare the OS and PFS in both treatment groups.

METHODS : Electronic medical records (EMRs) of patients older than 18 years with advanced NSCLC were acquired from our local EMR database at a university hospital. Patients were grouped into two retrospective cohorts based on their first-line treatment. A data mart exported into Excel tables was analyzed using built-in functions and the Kaplan-Meier method. Dates of therapy induction and death were collected from structured fields, while data regarding disease progression (RECIST) were collected from unstructured cancer care team reports.

RESULTS : A total of 117 patients were included in group A (first-line bevacizumab), and 147 patients were included in group B (first-line pemetrexed). In the entire cohort, median age was 60 years and 56,8% of the patients were male. In terms of age, sex, presence of brain metastasis and Charlson Comorbidity Index there was no significant difference between the two groups. The median PFS was 240 days in group A and 163 in group B (P=0.0057), while the overall survival was 525 and 388 (p= 0.0016) days respectively.

CONCLUSIONS : Collecting information that describes disease progression from unstructured fields of text using a semi-automated method is a challenging task. Our study proves that OS and PFS can be calculated using EMRs as a source, however this method has its limitations and needs to be tested on a larger patient population. Natural language processing methods may improve the quality of results.