German researchers are building a database to predict the progression of MS
October 15, 2003
By Hannah Cleaver
The integration of computer and statistical sciences with biomedical research is breaking new ground in Munich, Germany, where researchers are working to create a “virtual” patient based on biometric data, which they say could one day replace real patients administered placebo in some clinical trials.
Scientists at the Sylvia Lawry Centre for Multiple Sclerosis Research have compiled data on disease progression in nearly 20,000 patients, with the aim of being able to predict how a patient with multiple sclerosis (MS) would fare with no treatment and to use that as the comparison with the patient receiving the treatment under examination.
Using a “virtual” patient could resolve some ethical problems with placebo treatments and reduce the cost of clinical trials, said Albrecht Neiss, a researcher with an interest in genetic epidemiology who heads the center.
Neiss told The Scientist his group was working with pharmaceutical companies, which have contributed case study information to the database he is compiling.
Developing it is an incremental process, he explained. “The first stage is to use a much smaller control group than normal and fill the numbers out with information from the database. Then we aim to move onto the second stage, which would not use any new placebo patients, just using case studies from the database as a control group. The third stage would be to be able to create virtual patients which would match those in the study getting the treatment under trial.”
The characteristics of patients assigned to treatment—such as gender, age, and onset of disease—would be compiled, and a composite match would be created from the cases in the database.
This could create a “twin” patient whose development of the disease was predicted using biological statistics and whose progression could be compared with the effects of treatment. Neiss said his team was hoping to reach that third stage in about 18 months.
He said that the new databank did not yet include genotypic information, but that the project had been in contact with a genetics project in Oxford, England. “In the future there may be, if we continue to collect data for the database. It may be common to use genetic information in a clinical trial, and we could add this data to what we have.”
Using the statistical method to model MS was quite difficult, as its progression is complicated, he said. “The most difficult technical problem is that the patient cases are very heterogeneous, making it difficult to draw common ideas about how the patient's condition will progress.”
Although the project already has between 17,000 and 20,000 case studies in its database, this number is seriously reduced when it comes to finding the raw data needed to create a virtual patient with specific attributes. “For example, you might need a virtual patient where the time of onset of disease is known. Many of the cases in the database may not have that particular piece of information.”
But if the system can deal with MS, it could handle diseases with different profiles. “If we can do it with MS, we could do it for other neurological diseases such as Alzheimer's and Parkinson's. And it could also work with some cancers where there is a large amount of data available on the disease's progression.”
Neiss said the responses from many other scientists has been positive, although he admitted getting the approval of regulatory bodies would probably take longer.
“We made contact very early in the process with the Food and Drug Administration in the USA, and they are very interested in what we are working on,” he said.
Francis P. Crawley, secretary general of the European Forum for Good Clinical Practice, told The Scientist that this kind of computer modelling will become increasingly important in the development of new drugs, for scientific and economic reasons. But as things stand, computer models will not be enough for health authorities to grant marketing approval for drugs, he said.
"The question for science is to what extent can we model human biology on a computer," he said. "My guess is that we can do that to a large extent…but there will be limitations in practice and a priori limitations."
The use of the modelling system for other diseases may depend heavily on the characteristics of those conditions. Richard Sylvester, assistant director of the European Organisation for Research and Treatment of Cancer, said cancer trials only very rarely use placebo.
“Most cancer study patients in the control group receive an active treatment,
generally the treatment that is considered to be the standard treatment
for the type of cancer under study,” he said. The World Medical Association's
Declaration of Helsinki sets out ethical principles for when placebo can
acceptably be used in clinical trials.
Copyright © 2003, The Scientist Inc.