Voici le fruit de votre travail collectif :

The ‘google’ of side effects

A team of doctors, pharmacists and computer scientists has designed software which hunts down surfers’ comments about their use of medication.

“My doctor prescribed (me) 75mg of Effexor on Saturday and two hours later I felt really ill… Sweating, nausea, numbness in my arms and legs, an awful headache and insomnia…” This kind of comment, posted in 2007 by someone called Jane74 on the Doctissimo website, is not covered by existing drug monitoring plans.

Nevertheless, specialized websites, where people share their experiences with drugs provide a wealth of information about adverse reactions. This is what Stéphane Schück, a doctor in public health specializing in epidemiology, president and medical director of Kappa Santé, noticed anyway.

This led to the idea for a computer tool able to exploit this huge quantity of textual data, called Detec’t which will be presented from June 13 to June 16 at Centquatre in Paris, at Futur en Seine, an annual event dedicated to digital innovation.

(About) sixty products under supervision

Designed by a team of doctors, pharmacists, computer scientists, statisticians and mathematicians, Detec’t is a search engine which scans almost a dozen sites and public forums, such as Doctissimo or Atoute.org, (and) in which we enter the name of a drug. “Facebook and Twitter were not included because people don’t talk about that a lot on those sites” Stéphane Shück notes.

The tool is set up for requests pertaining to the sixty most commonly used products in France: Tahor, paracetamol, Doliprane, and psychotropic and anxiolytic drugs and sleeping pills. Detec’t processes web surfers’ posts using text-mining software components which analyze semantics: the meaning of sentences, grammar, combined with statistics programs, designed in collaboration with a research team from Inserm.

“The idea is to spot how close two words are together, and to calculate the probability that the name of a drug is associated with a side effect” Stéphane Schück explains. Kappa Santé has also developed its own algorithms which can eliminate faux amis/false friends; aspirin is often associated with headaches, but does not cause them. “We also identified a whole discussion thread about Méliane, a 3rd generation contraceptive pill, but when we checked we saw that it was actually just somebody’s first name!”

The results can be displayed in two ways: statistical distribution of the side effects of a drug, shown on/as a graph, and internet users’ comments, ranked by relevance.

Everything is made anonymous, and any proper nouns and email addresses in messages are automatically deleted. “The hard part is handling the way people talk about things related to medication: hangover has to be translated by ‘headaches’, ‘nausea’… As a result we had to build a dictionary of synonyms.”

Detec’t is still a prototype. Eventually Kappa Santé plans to make it part of the services they provide to the pharmaceutical industry, in particular to compensate for the considerable shortcomings in spontaneous reporting of adverse reactions, a legal procedure which is somewhat demanding for doctors whose duty it is.