I'm reading some of the educational material for a new therapy that can help treat PPMS (Primary Progressive Multiple Sclerosis).

I'm having difficulty understanding their use of statistics.

Let's first talk in general terms:

If person A has 100 oranges, and person B has 75 oranges, then person B has 25% fewer oranges than person A.

We compute this as:

(100-75)/100 = 0.25 = 25%

Now let's get to the educational material. It states:

Overall, the percent of people with 12-week CDP was 32.9% of Ocrevus-treated people vs. 39.3% of placebo-treated people. Another way of looking at this data is to say that people treated with Ocrevus were 24% less likely to have disability progression than those treated with placebo.

I don't get it. Using the same formula as above, (39.3-32.9)/39.3 = .163 = 16%

Yet the educational material uses a figure of 24%, not 16%.

How do they obtain that figure, and is it a common (fair, straightforward) or uncommon (misleading) way to represent that data?

(Please note that I am in no way suggesting anything is fair or misleading. I am only stating that I don't understand and am confused by what I have read.)


1 Answer 1


24 % is the decrease in odds ratio, where, for each treatment option, $$ odds = \frac{n\ cases}{n\ controls} = \frac{n\ cases}{n\ total-n\ cases} $$

Thus, odds of having CDP in the treated group are $32.9/(100-32.9) = 0.490$; odds in the placebo group: $39.3/(100-39.3) = 0.647$. Ratio of these odds is $0.490/0.647=0.757$, which is 24.3 % smaller than 1 (odds ratio under no effect).

Odds ratio probably is the most common way of presenting effects in epidemiology studies with case-control outcome, but I agree with you that the writers' presentation is unclear - "less likely" could refer to any other related measures, like odds difference, risk ratio, risk difference etc. As to which measure is more appropriate - entire chapters are written about that, just look up any textbook of epidemiology.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.