Before launching a new drug to the market, it is tested on thousands of people, but adverse reactions (side effects) may not be detected until many more patients have used the drug. Once the drug is on the market, clinicians are responsible for recognizing and reporting suspected side effects, which are collected in so-called spontaneous reporting systems. However, a number of recent, highly publicized drug safety issues showed that adverse effects of drugs may be detected too late, when millions of patients have already been exposed.
With the aim of improving this system, ALERT will analyse data from electronic health care records (EHRs) of over 30 million patients from The Netherlands, Denmark, the United Kingdom, Spain and Italy, using a variety of computational techniques. Those techniques, including text mining and epidemiological computing, will help to retrieve information from the data and detect "signals", such as combinations of drugs and suspected ADRs that require further investigation.
The focus of the ALERT project will be on side effects in children, as relatively little is known about them and children are particularly vulnerable. Moreover, the interdisciplinary research team will attempt to find a way to discriminate between signals that do indeed indicate an ADR and spurious signals, which might even result in withdrawal of a useful drug from the market. Also from a commercial and regulatory perspective the cost of a false-positive signal is significant.
To discriminate between true signals and spurious signals, in ALERT a possible biological explanation is sought for each signal. This process of signal substantiation requires that the signal be placed in the context of our current understanding of possible biological mechanisms. ALERT will use to the maximum the currently available databases that contain information about these biological mechanisms and augment that understanding with in silico models and simulations of the behaviour of drug and biological systems. ALERT will also rely on experimental screening to test the causal hypothesis generated during the substantiation of signals.
The project partners emphasize that this kind of analysis is a continuous process: "As more patient data become available and medical, biological and molecular knowledge expands, previous conclusions will need to be revisited. In order to deal with this constant process of revision, ALERT will focus on automated procedures as much as possible."
In the United Kingdom, academics will use a research database of about 10 million patients from the not-for-profit QResearch, a partnership between the University of Nottingham and Egton Medical Information Systems. Other institutions involved in the project that will analyse their own respective databases include:
- The Arhus University Hospital in Denmark;
- Erasmus University Medical Center in The Netherlands; and
- The University of Santiago de Compostela in Spain.