ABSTRACT

Prior to regulatory approval, while a drug is in development, randomized clinical trials represent the primary sources of safety information. Such experiments are generally regarded as the highest level of evidence, leading to an unbiased estimate of the average treatment effect (Atkins et al. 2004). Unfortunately, most trials suffer from insufcient sample size and lack of applicability to reliably estimate the risk of other potential safety concerns for the target population (Berlin et al. 2008; Waller and Evans 2003). Even if one leverages meta-analytic tools, rare side effects, long-term outcomes (both positive and negative), and effects in patients with comorbidities may still be unknown when a product is approved because of the relatively small size and short duration of clinical trials. For products intended to treat chronic, nonlife-threatening conditions that occur in large populations, the International Conference on Harmonisation (Azoulay et al. 2012) recommends a baseline safety database that involves at least 1500 patients on average with at least

CONTENTS

9.1 Background ................................................................................................. 141 9.2 Opportunity for a Risk Identication System ....................................... 143 9.3 Review of Observational Databases ........................................................ 146 9.4 Review of Observational Database Methods ......................................... 147 9.5 Methodological Progress toward a Risk Identication System .......... 151 9.6 Findings from the Observational Medical Outcomes Partnership .... 153

9.6.1 Performance .................................................................................... 156 9.6.2 Database Heterogeneity ................................................................ 160 9.6.3 Parameter Sensitivity .................................................................... 164 9.6.4 Need for Calibration ...................................................................... 164

9.7 Summary ..................................................................................................... 164 References ............................................................................................................. 165

a  6-month exposure time to reliably (i.e., 95% of the time) identify events happening at the 1% level (U.S. Food and Drug Admin 1999). In other words, events that occur less frequently than 1 in 100 patients are not expected to be detected under this recommendation.