When Tysabri is used, "one in 1,000 patients will get a horrendous and usually fatal disease," said Dr. Larry Steinman of Stanford University, who co-discovered the drug. "The potential benefit and rare harm makes for a difficult choice that doctors and patients will have to enter into."
I love how he KNOWS that these people WILL get PML. Gee, if that was the case, why weren't there more cases on PML?
scoobyjude wrote:But there's a small chance Tysabri could cause a horrible side effect. In roughly one in 1,000 patients, the drug unleashes a brain virus that can kill patients or leave them blind, paralyzed and unable to speak.
So where did they get 1/1000? I'm guessing that they said there were about 3000 patients on the original trial and 3 got PML. But what you cant say from that is that the probability of anyone else taking TYSABRI getting PML is 1/1000. Doncha just love stats??
I do It was the principal component of my BSc (many years ago now)HarryZ wrote:Well, if you believe in math, stats and the laws of probability, ...
Sorry Harry you cant say that. The best you can say is that 1/1000 is an estimate of the chance of contracting PML because of Tysabri use. And that estimate must be surrounded with a whole bunch of stuff (confidence intervals etc etc) before you can make any inference.HarryZ wrote:... you can say that there is a 1/1000 chance of someone else coming down with PML.
But you're spot on thereHarryZ wrote:But I think it's fair to say that we just don't know all the answers about its connection to PML and it's going to take some time before we do.
Sorry Harry you cant say that. The best you can say is that 1/1000 is an estimate of the chance of contracting PML because of Tysabri use. And that estimate must be surrounded with a whole bunch of stuff (confidence intervals etc etc) before you can make any inference.
better2gether wrote:In the Tysabri mono therapy trial (Affirm trial), there was no single patient contracted PML for two years.
So using the 3/3000 = 1/1000 chance of PML argument we get 0/600 = no chance of PML. Er, maybe but not a valid inference to make from this set of data.
Users browsing this forum: No registered users