(B45) Feasibility of Assessing Comparative Long-Term Safety for Gene Therapy Products by Creating an External Control Arm using Real-World Data Sources

Publication Number: 655

Background: Gene therapy products are often designed to achieve therapeutic effect through permanent or long-acting changes in the human body. Post-authorization, the FDA and EMA therefore typically require long-term follow-up (i.e., up to 15 years) of patients who receive these products in clinical trials to collect data on delayed adverse events following administration. However, as background rates of events of interest might be unknown, long-term safety derived from an external control group could provide insights into the excess risk associated with gene therapy products.

Objectives: We aimed to investigate the feasibility of creating an external control arm for comparative safety purposes using two real-world data sources for the first gene therapy products approved by the FDA in 2017 (Kymriah®and Yescarta®).

Methods: Patients with relapsed or refractory (R/R) diffuse large B-cell lymphoma (DLBCL) treated with non-gene therapies meeting the pivotal gene therapy trial criteria were identified using ICD10, ICD9, NDC ,CPT, and HCPCS codes between September 1, 2017 and February 8, 2023, in a US claims database and an electronic medical records (EMR) database for healthcare organizations worldwide. Baseline characteristic variables were assessed in terms of availability and completeness. Subsequently, baseline characteristic distributions and length of continuous follow-up (until end of study period or gene therapy initiation) were compared for each product to those reported in the literature for similar patients .

Results: Adult patients with R/R DLBCL (at least 2 prior lines of therapy) in the US, were identified in the claims (n=26,245 ) and EMR (n= 26,061 ) databases.
Baseline age and sex were nearly 100% complete in both databases. Race and ethnicity were 84% and 77% complete respectively in the EMR database and available upon request in the claims database. Charlson comorbidity scores, prior lines of DLBCL therapy, and prior stem cell transplant information are available in both databases, with data completeness checks being more complex and ongoing. Median continuous follow-up was 5 years (range 0-16 years) in the EMR database and 3 years (range 0-6 years) in the claims database.

Conclusions: Both databases satisfy the criterion of the minimum sample size of the patient population (1500 patients) as requested in post-marketing requirement. Checking for completeness of data for certain baseline characteristics (i.e., Charlson comorbidity scores, prior lines of DLBCL therapy, prior stem cell transplants, ECOG status, histology, disease stage, and metastases) is complex and in some instances requires linkage to other data sources via tokenization.