Safety signaling is a term that is used specifically in the pharmacovigilance field. There hasn't been a unanimous definition of the term and hence different organizations provide differing views. Two global bodies such as the World Health Organization and the Council for International Organizations of Medical Sciences provide definitions which broadly define a safety signal as reported information from multiple events, observations, experiments, etc that suggests a possible causal relationship between an adverse event and a drug and that this relationship was previously either unknown or incompletely documented. Signal detection is the act of discovering such signals from the event the data is collected during the trials.
Prior to the widespread use of computers, case reports used to be filed manually on paper. After the introduction and widespread use of digital resources in the clinical research industry, electronic records and databases have become more popular due to their flexibility, accessibility, and capacity to hold large amounts of trial related information in one place. The storage of all event-related reports, individual case reports, and other documentation in a single place help in detecting signals since more than one report is required for a signal to be generated.
Individual case safety report is the major form of documentation that is used to detect, prioritize, and evaluate safety signals. Here, each patient has an ICSR about the list of events that have occurred during the trial to him/her. Currently, these ICSRs are in an electronic format and hence can be linked up with electronic databases maintained for the clinical trial. This helps in a smooth process for signal detection.
But ICSRs alone are not enough in keeping track of signals since there are a lot of patients involved in a trial. Hence, data mining algorithms come into the picture here. They are used to analyze the list of ICSRs. Once there is a certain number of ICSRs, data mining algorithms are deployed in order to detect safety signals. The same algorithms also can be deployed on patient registries in order to detect safety signal anomalies. The interconnection between the various documents through programming algorithms is significantly beneficial compared to manually comparing paper documents by people.
The criteria for deploying a data mining algorithm also will depend on the occurrence of the adverse event. If the event is found to be rare, it is likely the threshold for deployment will be lower since there will only be a few cases reporting the event. The data mining algorithms used for safety signal detection in documents employ either classical statistical techniques or Bayesian techniques.
Spontaneous adverse event reports are another means of detecting safety signals. Here, the report is entered into the trial database by the pharmaceutical company involved in the trial after enforcement by the regulatory agency. The Food and Drugs Administration (FDA) maintains a database called Adverse Event Reporting System (AERS). All spontaneously reported adverse events are entered into the AERS database because the FDA aims to develop it into a vehicle that can enable the agency to perform signal detection activities.