Elsevier has used big data in a large-scale study to assess the ability of animal studies to predict human safety by comparing consistency between preclinical testing and clinical trial findings.
The study was conducted in collaboration with Bayer , and involved analysis of 1,637,449 adverse events that were reported for humans, as well as the five most commonly used animals.
Data was obtained from regulatory documents of the US Food and Drug Administration (FDA) and the European Medical Agency (EMA) for 3,290 approved drugs and formulations.
Study results showed that certain animal tests are more predictive of human response compared to others. However, this depended on the species and symptoms being reported.
Elsevier said that the study could aid in deciding appropriate tests, in turn avoiding unnecessary animal testing. It expects that the findings can also potentially improve safety for patients.
Elsevier Scientific Services director Matthew Clark said: “Though generally accepted that animals predict human responses, the concordance has never been investigated on this scale before.
“Our big data study shows that through improved analysis of data, researchers can select tests based on the species that have the most predictive relationship with a human depending on the drug in question, and therefore rule out needless testing.
“This is important because it enables pharmaceutical firms to continue safely and humanely innovating, while searching for the life-changing therapies that will save many patients’ lives.”
The company used the analysis to create a dataset intended to provide an accurate means for researchers to predict human risk through parameters such as species, adverse events, and drug formulation.
Additionally, this can enable the researchers to design safe and robust clinical trials, while avoiding safety issues. The information is also expected to support evidence-based medicine.
Elsevier intends to continue advancing the analysis through collaboration with customers and their datasets, and is planning to include additional datasets on dosing to improve accuracy.