So here is a fast guide to statistics for marketing people.

Most statistics that we deal with have to do with clinical trials. So let’s start there. A clinical trial enrolls a group of patients who make up the patient population. This number is usually expressed as n and appears in the study as n=202. I have no idea what n stands for.

You would think a big patient population is a good thing, and often it is, but a study only has to be large enough to make certain statistical assessments. In other words, before the study commences some math geeks sit down and say, “we can say this or that would be statistically significant if we could test this on 47 patients.” When that happens, the study is “powered” to reveal certain things. Sometimes studies do not enroll enough patients or do not get enough of the expected responses to make statistical claims. At that point they are called “underpowered” and everybody is depressed.

Studies try to find things that are statistically significant. This means that mathematically, it is believed that the results could not have been caused by chance.

For example, let’s say a study has an endpoint of all-cause mortality or death from any cause. And let’s say during the course of a study of 10 people, one guy gets hit by a bus. Now that counts as death so you could say that 10% of the people in this clinical trial died from being hit by a bus. That happens to be a random event; it’s dumb luck. It would not be statistically significant. Significance of the statistical variety is achieved only when you can more or less mathematically determine that chance could not possibly have played a role.

In general, statistical significance is expressed by the p-value (I don’t know what p stands for, either). Significance occurs when p is 5% or less, often expressed something like p=0.04 or p<0.001). Think of it this way. If p=0.04, then there is a96% chance that chance was not involved.

The whole world seems to have agreed that p has to be 5% or less for statistical significance to occur. Sometimes you’ll see weasel studies where a study “trended toward significance” or “verged on significance” (like p=0.051) but if it is more than 0.05 than it ain’t significant.

That’s statistical significance and that is a big deal. However, studies also have clinical significance and that is a much squishier term. Clinical significance means that it matters to clinicians. Some studies are statistically significant but not clinically significant.

If a study is close to significant but not quite significant, you will often see a result marked as NS (that I do know: not significant).

Clinical trials will state results but you can’t get too hung up on looking at the actual numbers; you want to see significance. For instance, a study might find that a particular surgery reduced the rate of death from kidney failure by 23%. Now if it says 23% NS, you are dealing with something that does not really matter because even though 23% sounds good, the results were not significant. But let’s say that study found the surgery reduced death from kidney failure by 15% and that was reported 15% (p=0.001). Now you have something. Even though 15% sounds like less of a big deal than 23%, it’s all in the p-value.

Sometimes you will find that the population is described in “patient characteristics at baseline” in terms of age, race, drugs taken, physical characteristics, medical diagnoses, and so on. In this case, you’ll run into “median” and “mean” values, and they are not the same. Now to a mathematician, both of these are types of average, but to a regular marketing guy, “mean” is “average” and median is something else.

Let me explain. Let’s say you have 5 patients in your study and their ages are 19, 25, 40, 50, and 52.

The mean age (what we mere mortals often call the “average”) is 37.2. You get that by adding the numbers up (19+25+4-+50+52=186 and then dividing by the number of numbers 186 divided by 5 is 37.2).

The median age is the age in the middle when you line up all the ages in order. The median age in this study is 40. (Of the five numbers, 19, 25, 40, 50, and 52, the number 40 appears in third position).

Many clinical trials will state what sort of statistical tests they used in terms of gathering data. Consult your local biostatistician for details.

Share on Facebook

 Share now!