Let's parse this latest article and the study it refers to:
The article says that people who received abiraterone (Zytiga) plus prednisone lived a 15.8 months vs. those who got a placebo plus prednisone lived 4.6 months less, or 11.2 months. These are median figures. The difference was statistically significant at the p<.0001 level.
So what does does this mean? And how can we reconcile the competing arguments being made in the articles.
First, note the extension of life in the trial is only 4.2 months, which is a good thing but hardly a great thing in broader scheme of things.
Second, note that the increase can be "statistically significant" even though the actual number of months is quite short. That may seem to defy common sense but it is a result of the way statistical significance is calculated. If you had a large enough group of people participating in the trial even a week-long increase in survival time can be "statistically significant".
Third, the study found that abiraterone was effective in lowering psa in only 29.5% of the people in the trial. Meaning 70%+ got no benefit at all. If the study reported the media survival time within the 29.5% who benefitted, it would like be much longer than the media survival in the abiraterone group as whole, 70% of whom got no benefit at all. The overall median is very misleading.
Thus, two doctors or scientists, looking at exactly the same report and data can legitimately say opposite things: One can say "there is no meaningful difference in life expectancy, and the treatment is essentially palliative" and the other can say "there is a significant difference in life expectancy" and both can be correct. Cup half empty? Cup half full?
But two other things need to be considered as well.
First, the trial was only 20 months long. It could be that, in those patients for whom the drug is effective, that the real difference in life expectancy is much more than 4.2 months. This would only become clear through a longer trial.
Second, reporting just the differences in median life expectancy can be misleading. (google the definition of "median.") For example, it could be that the drug does nothing at all for half the people, and completely cures the other half. A short trial, looking only at medians, would completely miss the fact.
Regarding 142's point (142, you probably know this already) often when trials are stopped before completion it isn't because the drug isn't working or has intolerable side effects, it's because the drug is working so well that it is unethical to keep half the patients on the placebo. The trial is ended early so that all the patients who were randomly assigned to the placebo group can be offered the treatment which is now known to work. That may be the case here, and that would be a good argument to raise with any insurance company that is balking.
Regarding Sonpavde's statement that "optimum sequence of therapy" needs to be further studied but that one drug at a time is probably best... this seems to be little more than a hunch at this point. One of the most curable cancers right now is leukemia, and the reason some varieties are curable is that 3-4 different treatments are given simultaneously. The result is that the cancer is not able to mutate out of the way of the treatment since it is is being attacked from multiple directions at once. That kind of logic is why there is adjuvant treatment protocols (ADT, Radiation, Surgury.) It's likely that is more treatment options that are less toxic come on the market, more combinations will be tried and some of these combinations will work better than running the drugs one at a time and giving the cancer a chance to recover and become more dangerous between each treatment. But it's a tough experiment to run ethically, since if you give the treatments simultaneously early on, then you have that many less "bullets" to use later on if the first attempt fails. It's very tough for a patient to say "use all the bullets NOW!"
Post Edited (proscapt) : 10/8/2012 10:47:14 AM (GMT-6)