Attn science teachers! How to demonstrate how clinical trials work?

March 11, 2013 4:41 AM

I am looking for people's experiences of demonstrating the basics of how a clinical trial works to the lay person. My audience will be pretty broad - from school children to adults - so I will need to keep it pretty sort and simple. I need something I can use when visiting schools to big events - I have heard of people using cake and chocolate to demonstrate randomisation/placebo etc - has anyone on here done that and have any pointers? TIA!
posted by ozgirlabroad to Science & Nature (9 answers total)

Not experience, just a thought. Think of an analogy that everyone can relate to - say, a trial where 10 girls wear no makeup and 10 wear makeup, and the test is how many are asked to dance, asked on a date, etc.
posted by yclipse at 4:50 AM on March 11, 2013

I guess you could use cakes made with sugar or artificial sweetener and make jokes about which kids are more hyper. Or you could *pretend* that one of the two cakes had been made with a different sweetener, but you're not sure which, or...

No personal experience though. That sounds like a hard age range to span.
posted by acm at 6:27 AM on March 11, 2013

There's the episode of Freaks and Geeks where they replace the keg with nonalcoholic beer and people act drunk anyway...
posted by ChuraChura at 7:03 AM on March 11, 2013

It might be kinda cool to get some plaque disclosing tablets and some identically-shaped/sized/flavored candies and set up a faux "trial" with them (only the experimental group will have pink teeth!).
posted by julthumbscrew at 7:10 AM on March 11, 2013 [2 favorites]

Richard Saunders demonstrates the need for proper blinding using a dowsing demonstration. It's not a clinical trial obviously but the parallels should be easy enough to make clear afterwards. How well you can get this working depends a bit on the audience too of course.
posted by edd at 8:53 AM on March 11, 2013

If I were talking about this I would go through it using an example and going through it one problem at a time, or using different real-life trials to explain each step -

1) OK, so we want to know if this treatment works - so the first thing we would want to do is just give it to people and if they get better, well, it works, right?

2) Oops, we need a control group because maybe some people just get better anyhow even if they don't do anything.

3) Oops, but what if just (swallowing the pill/rubbing in the cream/visiting the doctor) causes some people to get better? We need a placebo control.

4) Oops, but what if people know they're not taking the "real" treatment? We need to blind the condition.

5) Oops, but what if the doctors administering the treatment know they're not *giving* the real treatment - they might give the "real" treatment to people who they believe have a better chance of getting better, or they might deliberately or accidentally let the patient know which treatment they're getting.

And so on right up through concealed assignment and concealed allocation and choosing a sufficiently random and representative test group - as far as you want to go with it (or until you see eyes starting to glaze over!).
posted by mskyle at 12:03 PM on March 11, 2013 [1 favorite]

I do a quasi clinical trial sort of thing in class to demonstrate the concept of hypothesis testing.

The students are primed in previous classes to know a bit about probabilities and expectations. They know that if I flip two quarters, you'll get, on average, 25% heads-heads, 25% tails-tails, and 50% heads-tails/tails-heads. Two regular quarters acts as a "placebo."

The "drug" to be tested is a pair of quarters that I don't allow the students to handle, though I allow them to observe the outcome of the flips. Depending on my mood, I'll either have one double-headed and one double-tailed quarter (0% heads-heads, 0% tails-tails, 100% heads-tails/tails-heads) or one double-headed quarter and one regular quarter (50% heads-heads, 0% tails-tails, 50% heads-tails/tails-heads).

The "clinical trial" is to try to tell whether the "drug" is different from the "placebo" -- is there something fishy about the "drug" quarters? If so, at what point are you suspicious? Convinced? Dead certain?

This automatically gets them to think about statistical significance and other similar issues.

Perhaps you might be able to modify this with two jars full of M&Ms -- blue ones representing a cure and red representing no cure, with different ratios of red/blue for "placebo" jar vs "drug" jar, but I like the quarters because we already have some level of intuition about how the quarters are supposed to behave -- meaning that our null hypothesis is already our default expectation.
posted by cgs06 at 12:44 PM on March 11, 2013

mskyle has a great outline up there. Basically, use the scientific method as your guide.

Informed consent is a HUGE part of doing clinical research. So whatever you do, I'd have your kids/adults go through a mock informed consent process.

I think a fun thing would be to set up a mock clinical trial where you have some volunteers do something simple. I work in taste and smell research so my example is biased towards that! Smell stuff might be fun if you have kid volunteers.

Get some flavor oils. For example, Isoamyl acetate is "banana flavor." You can order this stuff online for cheap. Soak a cotton pad with each odorant and then put the pad in a plastic jar.

Have the "subjects" come up and explain the "study" to them. You're going to tell them what they will have to do in the study (i.e., smell two smells and try to identify the odor/try to tell if it is a natural or synthetic odor), any risks involved (might smell bad, might tickle the nose), etc. ask them if they agree to participate. Then you "blind" the groups. Have one group getting "natural' odors and one group getting the "synthetics." See which group more accurately can guess the odors and see which group is better to identify whether they are natural or synthetic. You can even make up a fake "fact" and test that hypothesis, i.e., 'people with blond hair are better able to identify odors than others.' It doesn't really matter what the research question is.

Then when you've gathered "data" you can have the audience help you figure out if one group performed better than the other, if blond haired people were better able to tell the odors, etc. See if people have ideas on what conclusions you can gather based on your "data." Then you can say, "well they were all synthetic flavors! Placebo effect!"

I'm just spitballin' here. If the above fails, just have people sit around cutting lots and lots of red tape :)
posted by Katine at 12:48 PM on March 11, 2013

I recently made this post you might find useful: Sense About Science
posted by Blasdelb at 4:10 PM on March 11, 2013