A Data Science Central Community
In a recent post on clinical trials, one person was quoted as saying:
".....(what is the) definition of a successful clinical trial?
ANSWER: “At the end of the day,” “regardless of the result, you’ve learned something.”
But is this enough? and can we do better?
Bayesian data analysis metehods are being used by some to attempt to get a better foothold on the efficacy of new drugs.........BUT using the older "gold standard" for clinical trials, the randomized double blind study, although the trial may work, it may not really give practical help for doctors in making decisions about individual pataients. One such recent trial was for the drug Avastin, and “Despite looking at hundreds of potential predictive biomarkers, we do not currently have a way to predict who is most likely to respond to Avastin and who is not,” says a spokesperson for Genentech, a division of the Swiss pharmaceutical giant Roche, which makes the drug.
Obviously, better application of "Predictive Analytics", Data Mining, and Decisioning systems are needed ..............
For more information see the NEW YORK TIMES REVIEW:
There was an article a while back mentioning a strong bias in the way people are selected to participate in these trials, not sure if it was caused by financial incentives, and not sure whether it can be fixed. It did not criticize statisticians though, who were not responsible for the bias or for selecting the participants.
I could not agree more with Dr. Miner. It is a scandal that after many millions are spent and years of research, the outcome of an RCT typically hinges on a single overall test of statistical significance. Whether or not the "main effect" is significant, we still don't know how the drug affects different types of individuals. For the past three years, I have been working on predictive analytics methods to identify factors related to efficacy or safety (see www.causalytics.com). To make personalized medicine a reality, we desperately need to develop new approaches along these lines.