Analysis of Study Results

The probability curves (that describe the probability of illness or recovery as time progresses) are called Kaplan-Meier curves, because of the specific statistical analysis that Drs. Kaplan and Meier developed. If you would like to know more, there are several good tutorials on the web: one such tutorial is issued by Cancerguide.org.

We consider one more example, relevant to medication testing, and to statins in particular. The curves present the percentage of people who develop coronary heart disease (CHD) as a function of time after the start of the trial. (1) presents the ideal curve that would be obtained if FactStatin were 100% effective after a certain period. (1) levels off after a period of a year - this means that the medication is totally effective (having taken it for a year), and we have eradicated heart disease once and for all. (2) is the result for the group medicated by FactStatin, and (3) is the result for the placebo group. Unfortunately, we have not yet invented the ultimate cure. Real statins, compared to a placebo, exhibit behavior like that in (2) and (3). After 5 years (in this fictitious example), 8% of the participants who took the placebo developed CHD, whereas 6% of the people who took FactStatin developed CHD. This difference is denoted by the vertical arrow (V). Similar to the pharmaceutical industry claims, CHD is reduced by 25% when statins are taken.

But that is not the whole story. There are other aspects that should be considered: we have given medication to 100 people, in order to prevent two CHD events. Or have we prevented them?

1. Some of the people will experience side effects; a few of them - serious.
2. We really haven't prevented the additional 2% of events - we've just delayed them!

The side effects are a separate issue that we'll discuss in the context of the actual studies. The delay, denoted by the horizontal arrow (H) in the drawing, shows that if the study were to continue for approximately another year and a half, the dashed continuation of the FactStatin curve (2) would have a height equal to the height the placebo curve (1) had after 5 years - 8%. So really, CHD hasn't decreased - 2 out of 100 people have on average 18 months more till they get the disease. This is the same time-extension effect we described in the previous page.