Real-world utilization of pembrolizumab dosing schedules in Belgium: a multi-center retrospective study using a common data model and natural language processing

Aim: Pembrolizumab's six-week extended dosing (ED), approved as an alternative to the standard three-week schedule (SD), offers flexibility in treatment plans. Materials & methods: In this retrospective study across four Belgian hospitals, we described real-world pembrolizumab utilization using a pipeline that processed unstructured data with natural language processing (NLP) and mapped both structured and unstructured data to an OMOP-CDM database. Physicians and data analysts collaboratively developed the project-specific data dictionary and performed validation. Results: A total of 303 patients were included, 56% with non-small cell lung cancer (NSCLC). Most patients (72%) received monotherapy, with 89% starting on SD. For NSCLC first-line monotherapy, the median time on treatment (rwTOT) was 106 days in the ‘SD only’ cohort and 666 days in the ‘SD to ED’ cohort, while the median time to next treatment (rwTTNT) was 280 days for ‘SD only’ (median rwTTNT not reached for ‘SD to ED’). Dosing cohort distributions showed that 74% of patients in the ‘SD to ED’ group had NSCLC, compared with 77% with urothelial cancer in the ‘ED only’ group. The mean (StdDev) gap between administrations was 23 (7) days for SD and 43 (5) days for ED. Most patients receiving ED did so after starting on SD, and all ED initiators stayed on ED. Dosing patterns remained stable pre- and during COVID-19. Conclusion: This study describes real-world pembrolizumab dosing patterns across four Belgian hospitals and illustrates how integrating NLP and structured data within an OMOP-CDM framework can support such analyses. Patients switching from SD to ED showed longer rwTOT and rwTTNT than those on SD alone, though potential immortal time bias requires caution in interpretation.