Traditionally, newly developed drugs have been tested on animals so that any toxic effects can be caught before the drug is used by humans. However, putting animals through these types of experiments just for our own sake is ethically problematic. Furthermore, extrapolating the meaning of test results from non-human animals to humans limits the accuracy of those conclusions.

One alternative to animal testing is the use of human-based computer models. If such models are proven to be effective, they can replace testing on animals and potentially cost research less time and money, and cost far fewer animal lives. This study sought to evaluate how well human-based action potential (AP) models predict drug-induced arrhythmias (heart rate irregularities) when compared to traditional experimental methods that involve animals. 

The O’Hara-Rudy dynamic model, which was created using data from over 140 real human hearts, served as a baseline to create 1,213 AP models that served as the controls for this study. Each model was created with its own unique ionic properties, thus allowing the models to simulate the range of differences in heart function across real human populations.

A total of 62 drugs with known cardiac effects (including some specifically meant to treat arrythmia) were considered. The effect of a range of concentration levels of each drug was simulated on the control models using Virtual Assay software developed at Oxford University. These computer simulations were tested specifically on their ability to classify drugs according to the drug’s risk (or lack thereof) for causing TdP, a serious and potentially fatal arrythmia. For comparison, the results from the computer models were compared to results from traditional experiments that tested the same drugs on animal models (e.g. rabbit hearts).

The simulations attempted to identify which drugs had a risk of causing TdP in two main ways: first, by looking for increases in the model’s action potential duration (APD) and second, by looking for repolarization abnormalities (RA). The APD method’s accuracy suffered from a relatively low specificity, meaning the method was prone to categorizing even safe drugs as risky. By contrast, the RA method was a lot better at recognizing safe drugs, and it was overall much more accurate in its classifications. This was illustrated by results which were based on simulating each drug’s effects in all of the AP models in order to make a binary classification of the drug as either risky or safe. Also, it’s worth noting that these results are based off of simulations of the 49 drugs with the more clear-cut known classifications: 24 were standardly known to be “high risk” for causing TdP, and 25 were known to not pose a risk of causing TdP.  

When using the APD-based method, the simulations classified 80% of these 49 drugs correctly. Of the drugs that were commonly known to not increase the risk of TdP, the ADP method correctly categorized 64% as safe and miscategorized 36% as risky. By contrast, the RA method had a much higher specificity: of the drugs that were commonly known to not increase the risk of TdP, the RA method correctly categorized 92% as safe and miscategorized 8% as risky. Combined with its 100% sensitivity rate (all high-risk drugs were correctly categorized as risky), the RA method successfully categorized 96% of the 49 drugs.

These results indicate that when using the RA method to identify a drug’s risk of causing TdP, computer simulations display “sensitivity, specificity and accuracy higher or comparable to the ones obtained through animal studies.” For example, using the RA method, computer simulations classified 89% of the full set of 62 drugs correctly. By contrast, a previous study using the rabbit heart model to measure a drug’s risk of causing TdP correctly classified only 75% of the 64 drugs that were tested. 

Animal studies often try to identify TdP risk by looking for prolonged intervals, which indicate that the heart is taking too long to “rest” between beats. However, previous studies have found that classifications of TdP risk based on prolonged intervals in animal models only matches up with the clinically-known TdP risk status of a drug about 49% of the time. In fact, the FDA has approved 89% of the drugs that have been found to cause prolonged intervals. Using computer models also has another advantage: because an entire population of models with differing baseline physiologies were used, researchers could study the effects of a drug on specific heart types. This could allow researchers to identify subpopulations that are at higher risk for developing drug-induced TdP based on the ionic characteristics of their hearts. 

In conclusion, the study found that computer simulations are accurate enough to reduce the need for animal testing when it comes to identifying a given drug’s risk of causing TdP. Though the authors stop short of advocating complete and immediate elimination of animal models for drug-induced TdP risk, they believe that at least partial replacement of animal models by virtual models can improve drug testing quality and reduce costs. Animal advocates can use this study to educate others about how alternatives to animal testing work, and what benefits the alternatives bring. Additionally, students can advocate for the use of computer models (rather than animal models) in their science classes. Finally, those pursuing cardiology research can conduct additional studies into the use of computer simulations for heart problems.