Predicting Adverse Drug Events Using Pharmacological Network Models

A network-based method that uses available pharmacosafety data can predict yet-to-be-discovered adverse drug events to help reduce drug-associated morbidity and mortality. The Power of Prediction We’ve all done it: googled a combination of medical terms to describe how we feel after taking a new medication. The result is a seemingly infinite list of Web sites telling us that the nausea is normal, or that the headaches warrant another visit to the doctor. Oftentimes, important adverse effects of drugs are discovered and added to the drug label only years after a drug goes on the market. But what if scientists could know about certain adverse drug effects before they are clinically discovered? Cami and colleagues develop a mathematical approach to predicting such adverse events associated with the drugs we take, in hopes of reducing drug-related morbidity—and mortality. After its release to the market, any given drug undergoes rigorous evaluation to determine associated ADEs (adverse drug effects). This post hoc analysis is usually unable to detect rare or delayed-onset ADEs until enough clinical evidence accumulates–a process that may take years. The method devised by Cami and coauthors does not need to wait for such evidence to accumulate. Instead, it can inform drug safety practitioners early on of likely ADEs that will be detected down the line. The authors first collected a “snapshot” of 809 drugs and their 852 related adverse events that had been documented in 2005. These drug-safety associations were combined with taxonomic and biological data to construct a network that is reminiscent of a web. Cami et al. then used this drug-ADE network to train a logistic regression predictive model—basically creating a formula that would indicate the likelihood of unknown side effects of any drug in the network. The predictive capabilities of the model were prospectively validated using drug-ADE associations newly reported between 2006 and 2010. Such prospective evaluation preserves the chronological order of drug adverse event reporting, making it a realistic method for predicting future ADEs. With their network, the authors were able to predict with high specificity seven of eight drug ADEs identified by pharmacological experts as having emerged after 2005, including the relationship between the anti-diabetes drug rosiglitazone (Avandia) and heart attack. The benefit for patients? With this powerful model in place, certain unknown adverse drug effects may be discovered earlier, helping to prevent drug-related morbidity and mortality through appropriate consumer label warnings. Early and accurate identification of adverse drug events (ADEs) is critically important for public health. We have developed a novel approach for predicting ADEs, called predictive pharmacosafety networks (PPNs). PPNs integrate the network structure formed by known drug-ADE relationships with information on specific drugs and adverse events to predict likely unknown ADEs. Rather than waiting for sufficient post-market evidence to accumulate for a given ADE, this predictive approach relies on leveraging existing, contextual drug safety information, thereby having the potential to identify certain ADEs earlier. We constructed a network representation of drug-ADE associations for 809 drugs and 852 ADEs on the basis of a snapshot of a widely used drug safety database from 2005 and supplemented these data with additional pharmacological information. We trained a logistic regression model to predict unknown drug-ADE associations that were not listed in the 2005 snapshot. We evaluated the model’s performance by comparing these predictions with the new drug-ADE associations that appeared in a 2010 snapshot of the same drug safety database. The proposed model achieved an AUROC (area under the receiver operating characteristic curve) statistic of 0.87, with a sensitivity of 0.42 given a specificity of 0.95. These findings suggest that predictive network methods can be useful for predicting unknown ADEs.