Big data’s big bias: bringing noise and conflicts to US drug regulation

Big data can be used cautiously to examine real world outcomes and to improve surveillance of drug safety. For example, it has been used to identify overuse of some interventions and can show drug and device complications in real world settings rather than idealized controlled trials.12 However, big data are a noisy mess, and analyses by entities with profit motives may identify spurious associations that support fast track approvals and indication creep (broadening the indications for drugs and devices).

The Reagan-Udall Foundation curates real world evidence or “big data” derived from routinely collected health data from insurance claims, electronic health records, voluntary registries, and social media. The US drug and device regulator, the Food and Drug Administration, says that such data can speed up research, “saving time and money” for “therapeutic development, outcomes research [and] safety surveillance.”3

In January, Robert Califf, then FDA commissioner, announced the launch of Innovation in Medical Evidence Development and Surveillance (IMEDS), a foundation project that he said would collect and analyze big data to identify “important safety issues.”4

However, critics of the move say that big data are poor for identifying adverse events and the system may expose patients to overtreatment and associated harms. Financial conflicts of interest, they worry, could influence the way big data are used, including exploitation of the weaknesses inherent in observational data to win FDA approval for new uses of drugs and devices and to exonerate drugs of previously detected harms. There is evidence and precedent to support …