Patient Summit USA 2012

Oct 29, 2012 - Oct 30, 2012, Philadelphia, USA

The right mix of payer buy-in, patient support, digital engagement and partnerships

Clinical Trials: How Much Benefit Do Computer Models Add?

Following last week's announcement that new computer models have successfully predicted negative side effects in hundreds of current drugs, Rochelle Sampy asks whether computer modelling is any substitute for traditional patient trials...



Can researchers develop drugs that are safer for patients by using a computer system rather than a trained medical reviewer to determine adverse side-effects? Is this method of predictive testing more likely to benefit the safety of patients or increase the profit of pharmaceutical companies rather than doing both?

Once a drug has reached the market, patients are less likely to use it if it has a high number of side-effects. In order to increase drug adoption, many efforts are made to relate severe side-effects to specific genetic biomarkers before the clinical trials, so that patients can benefit from a prescribed drug while avoiding adverse drug reactions.

It is stated that the method of in silico prediction of potential side-effects to develop safer, more effective drugs for patients can replace the traditional experimental detection process of in vitro safety profiling as this is considered challenging in terms of cost and efficacy.

A week ago, Novartis Institutes for Biomedical Research (NIBR) and SeaChange Pharmaceuticals created a set of computer models that predict possible adverse side-effects of existing drugs through using the UCSF School of Pharmacy’s ‘similarity ensemble approach’ which uses similarities between the shape of drugs and other compounds to determine negative drug reactions. The computer model identified 1,241 possible side-effects for 656 drugs of which 151 were new drug side effects that were not on Novartis’ drug database.

Michael Keiser, co-founder of SeaChange, stated that this method will not replace existing safety approaches like animal or lab testing but will simply prioritize what lab tests should be done or which research compounds will require fewer tests than others.

However, is it responsible to promote the need for fewer tests for certain research compounds due to results from an inert object?

The human body is unique as are individual patient preferences and characteristics. The computer programme will only produce results depending on the information it was provided with and will not account for any individuality. There is a strong probability that the clinical human testing that follows this computer data will result in useful “big data” that no sophisticated computer system could ever produce.

In 2008, researchers at Duke University proposed more testing at the clinical trials stage as this would be a cost-effective way of reducing adverse side effects for the patient population after approval. This is in line with the Simon Davies' (Teenage Cancer Trust) perspective that clinical trials should be opened up to greater numbers through an opt-out scheme as reported last week.

Even if side effects were known before clinical trials through a computer method, a larger number of patients could be used to test these findings so that patients could directly inform other patients of their experiences, regardless of what companies publish about the trials.

However, through any combination of in silico and in vitro testing, will pharmaceutical companies and patients be faced with more adverse side effects than they can handle? Will this reduce the number of patients taking part in clinical trials, invariably leading pharmaceutical companies to reach even less of their testing targets?

Overall, predicting side-effects of a drug has always been difficult whether through human or animal testing or computer methods. One thing for certain is that patients do need to be better informed of possible side-effects before taking a drug. Even with pre and post-clinical data, expert medical reviewers should still individually test each patient before they are prescribed a new drug. However, no amount of testing can predict the exact way that each individual patient will react to a particular drug and that is the hardest part.



Patient Summit USA 2012

Oct 29, 2012 - Oct 30, 2012, Philadelphia, USA

The right mix of payer buy-in, patient support, digital engagement and partnerships