Robert J Heine1
, MD, PhD, FRCPE; Stephen Knowles, MD; Byron J. Hoogwerf, MD; Haoda Fu, PhD; David H. Manner, PhD
1Eli Lilly & Company, Indianapolis, IN, USA
Several glucose-lowering agents with established long-term safety are available. Rightfully, the safety hurdle for diabetes drugs is high. A critical aspect to patients, and thus in drug development, is ensuring that long-term benefits exceed potential adverse events (AEs).
Assessment of safety starts in the preclinical phase and continues throughout the lifecycle of the drug. Drug safety assessment in diabetes is complex due to the overlap between potential AEs and health risks associated with diabetes (e.g. CVD, biliary tract diseases, pancreatitis, cancer, fractures). It is fair to assume that clinically meaningful risks, e.g. a 30% increased risk of CVD, can be excluded by the time of drug approval. The reasonable exclusion of this risk requires randomized clinical trials with approximately 7,000-10,000 patients. To assess the risk of rare events, the “Rule of 3” can be applied with sample sizes >30: if a certain event did not occur in a sample with n subjects, the interval from 0 to 3/n is the 95% CI for the rate of occurrences in the population.1 E.g., if 0 events are observed (n=1000 over 1 year), then the 95% CI for the event rate is [0, 0.3%], where 0.3%=3/1000/year.
The exclusion of lower incidence events requires different approaches. Post-marketing drug surveillance currently relies, among others, on AE report systems that are susceptible to important biases. These biases need to be understood and signals detected by these systems require validation in appropriately designed cohort studies.
With the development of precision medicine, we will be better able to identify patients who benefit the most with the lowest risk of developing AEs. Additionally, with the fast adoption of electronic health records we may expect the development of more sensitive and robust post-marketing surveillance systems that aid in establishing the balance between benefits and risks. This should be the ultimate objective.
1. Eypasch E; Lefering R; Kum CK; Troidl H. Probability of adverse events that have not yet occurred: A statistical reminder. BMJ 1995;311(7005):619–620.