In all studies listed there was a reduction in the number of breast cancers in the treated group by about 50%. This means a relative risk reduction (RRR) for the treated group of 50% compared to the control group. The problem with this RRR number is that in order for us to make sense of the magnitude of the effect on the population, in this case the benefit of the medicines, we need to know the level of risk in the control group or how many bad outcomes happen without the treatment. Treatments with large risk reductions have a small effect in conditions that are rare (very few bad outcomes in the control group). On the other hand even a modest risk reduction can have a large effect if the bad outcome in the control group is large.
Two examples will help, each with a 50% RRR. In the first the bad outcome happens 80% of the time in the control group but only 40% in the treated group. In the second the bad outcome happens 8% of the time but only 4% in the treated group. In each case a 50% relative risk reduction but obviously the outcome in the first example is of greater magnitude and in the first case you treat 100 to help 40 people, but in the second you treat 100 to help only 4.
There is another way to help us understand the magnitude of effect when we consider treatment and very large studies with numbers not so easily managed, particularly in brains like mine. Absolute risk reduction (ARR) is just the difference between to the two outcomes, but this then can be converted to the more useful Number Needed to Treat (NNT). This number, which is the multiplicative inverse or reciprocal of ARR is another way to understand the magnitude of effect.
For the first example the ARR 0.8-0.4=0.4 and 1/0.4=2.5. 2.5 is the Number Needed to Treat (NNT) to prevent one bad outcome (the same as treating 100 to help 40).
In the second example ARR 0.08-0.04=0.04 and NNT then 1/0.04=25. This is just another way to say treat 100 to help 4. These measures are helpful when the numbers are more cumbersome. The ideal NNT would be 1, where everybody treated saw a benefit.
In the NSABP P-1 study that showed a breast cancer risk reduction for those treated with tamoxifen, the control group of 6599 women had 175 invasive breast cancers and the tamoxifen group of 6576 had 89 invasive breast cancers.
The rate of cancer in the control group is 0.027 and in the tamoxifen group 0.014 so the RRR is 48%, but the ARR is 1.3 giving a NNT of 77. This means 77 women need to take tamoxifen for one to benefit.
That is part of the reason we offer treatment only to those at high risk (more likely to have breast cancer). We also need to weigh adverse effects in all cases, but particularly when we prescribe for an asymptomatic person to prevent a bad outcome.
The NNT for raloxifene in the CORE trial at 8 years is 67, but raloxifene has another benefit, reducing the risk of vertebral fractures. In the MAP.3 trial the NNT for exemestane is 94.
Again, this information is to prompt thoughtful discussion and consideration and is not intended as medical advice for any one person. These prescription drugs have saved many lives and will continue to save lives when used according to guidelines.
For those who want even more about statistics and medicine consider the article in Bloomberg Businessweek http://www.businessweek.com/magazine/content/08_04/b4068052092994_page_2.htm
Together we can prevent 75,000 breast cancer cases each year.