A new technique for testing how an individual will respond to a drug could be the answer to developing personalised medicines has been developed by scientists at Pfizer and Imperial College London.

The technique, reported in Nature (April 20), could also be used in clinical research to sort responders from non-responders and identify individuals that may have an adverse reaction to a medicine before they are exposed to therapeutic levels of the drug.

It’s an unfortunate fact that the majority of medicines do not actually work in many patients for whom they are prescribed, because of variations in genetic profiles. This has led to the concept of ‘personalised therapy’, in which medicines are matched to a patient to minimise side effects and optimise clinical efficacy.

But environmental factors such as nutritional status, the bacteria in a patient’s gut, age and whether the patient is suffering from an illness are also critical factors in determining the response to a drug, and simply carrying out a genomic analysis is unlikely to be sufficient to deliver a truly personalised therapy.

The new approach developed by the researchers – called pharmaco-metabonomics – is based on the biology of each individual patient by examining body fluids for metabolic signatures and using computer analyses to try to predict how that person will respond to a drug.

According to Jeremy Nicholson of Imperial, who led the research team, the first stage of the research was to examine animals’ responses to galactosamine, a compound that causes liver damage in some and has no effect in others. The scientists were able to show that it was possible to predict which animals would develop liver damage based on their metabolic profiles, determined using urine testing for many hundreds of metabolites.

They then designed a more complex experiment in which levels of natural metabolites in rats’ urine were examined using nuclear magnetic resonance spectroscopy, a powerful diagnostic tool. The animals were then exposed to a high-dose the analgesic paracetamol, which is known to cause liver damage. Mathematical modelling was used to work out how the pre-dosing metabolic profile equated to the extent of liver damage seen in the rat, and the team identified certain metabolic ‘signatures’ that predicted the severity of the toxicity.

“This new technique is potentially of huge importance to the future of healthcare and the pharmaceutical industry,” commented Nicholson. “

He suggested that the technique could be applied in clinical studies by using sub-toxic doses of drugs to build models that determine drug responses based on metabolic profiling. The resulting data could be used to either rule out patients who would have an adverse reaction to a drug candidate, select for patients who would benefit from the treatment, and determine the most effective dose to boot.

Pharmaco-metabonomics will likely sit alongside pharmacogenomics – which involves trying to determine an individual’s response to treatment simply by looking at their genetic profile – but should be a more powerful tool, said Nicholson.

It was once hoped that pharmacogenomics and the related discipline proteomics – which looks at the expression patterns of proteins in cells - would identify specific signatures that would predict drug response and toxicity, but efforts to date have not uncovered any reliable candidates.

Pharmaco-metabonomics could be a new path towards attaining that goal and “the development of more personalised healthcare for large numbers of patients,” concluded Nicholson.