(P-452) Identification of Appropriate Safety Metrics in Long-term Studies

Background: Choosing the appropriate metrics to estimate the risk of safety outcomes is critical to accurately characterize the safety profile of a drug. However, few papers provide guidance on the best metrics when assessing long-term safety of drugs.

Objectives: The goal of the study is to identify an appropriate metric to estimate the risk of safety outcomes in long-term real-world studies.

Methods: We compared the crude incidence, exposure-adjusted incident rate (EAIR), the complement of Kaplan-Meier (KM), and the Nelson-Aalen (NA) estimator of cumulative hazard. We performed simulations to understand the characteristics (bias, mean squared error, 95% coverage probability) of each metric, using a combination of various event background rates (rare versus moderate) and dropout rates (no dropout, 20%, 40%, 60%, 80% dropouts). Also, we studied the effects of non-constant hazards on the performance of the metrics.

Results: The crude incidence was substantially more biased as the dropout rate increased. The biasedness increased as the background event rate and hazard rate increased. The EAIR demonstrated low bias and the coverage probability was >95% when the hazard rate was relatively constant over time. The background event rate nor dropout rate affected the performance. However, EAIR did not perform well when the hazard rate was not constant. The performance worsened particularly for AEs with long latency period. The complementary of KM performed better than the crude incidence and EAIR. However, it underestimated the risk compared to the cumulative hazard of Nelson-Aalen estimator.

Conclusions: Cumulative hazards of Nelson-Aalen estimator maintained the lowest bias with nominal coverage probability regardless of hazard rates, dropout rates, and background rates.