Singular science
March 13, 2020 7:47 PM   Subscribe

“N of 1” studies aim to answer medical questions one person at a time. "As with most research studies, N of 1 studies gain their power through data points. But instead of taking a few measurements from many people, researchers can conduct many measurements from one person over time."

By Chris Woolston for Knowable Magazine (01.09.2020)
By studying individuals instead of groups, researchers can learn more about which quirks really matter for the big picture. If a large number of N of 1 studies showed, for instance, that people with certain genetic markers responded well to a particular diet, that diet could be recommended to others with that marker. “Finding the differences that matter versus the ones that we can ignore — that’s where all of the action is,” Schork says.

... The average blood sugar reading after eating a piece of bread, for example, was 44 milligrams of sugar per deciliter, but some people had less than 15 milligrams while others scored around 80. Such wide variation underscores the need to study nutrition on an individual basis
posted by spamandkimchi (7 comments total) 11 users marked this as a favorite
 
If a large number of N of 1 studies showed, for instance, that people with certain genetic markers responded well to a particular diet, that diet could be recommended to others with that marker.

Isn’t this just an N of a large number study?
posted by vorpal bunny at 10:47 PM on March 13, 2020 [3 favorites]


As someone who is both trans and taking psych meds, I’ve often referred to my life as an N-of-1 study. It feels surprisingly validating to see self-experimentation being taken seriously.
posted by WCWedin at 10:55 PM on March 13, 2020 [4 favorites]


I've tried to do this with my doctor while getting my blood pressure medication right. I'm not taking anything too complicated, the usual diuretics and ACE inhibitors. But between my intolerance for Lisinopril and the challenge of getting the dosage right there's been some experimentation.

What bugs me is there's no methodical experimentation. Certainly not guided from my doctor. He's open to the idea in a "try different dosages and tell me what works". But measuring blood pressure accurately is difficult, since it varies so much by time of day. And I don't know how best to do the experiment design. Do I try a different set of dosages every day? Every week? What combination of things should I try and in what order?

It seems like it'd be simple enough to design a, say, eight week program for a patient to try out two or three different medications in two or three different doses. But no one offered me that and I'm left trying to work it out on my own. Meanwhile, my doctor gets excited when I walk in one day and my BP is down 10 points from last time. And I try to explain "but doctor that's because I had a relaxing walk over here instead of a stressful drive" and he just brushes that away. In the end we just added medications until we hit the magic point but I'm left wondering if I should try taking away some of the first medications since they were not effective alone.
posted by Nelson at 7:15 AM on March 14, 2020 [4 favorites]


So uh...case studies have existed for a while now?
posted by Young Kullervo at 8:28 AM on March 14, 2020 [1 favorite]


Yeah, this feels a little dated, only because N of 1 studies have been used extensively in lots of fields for a long time and are considered good evidence, though I would not consider them the highest level of evidence, which is typically reserved for things like meta analyses.

Isn’t this just an N of a large number study?

Yeah, exactly, but I think the author is making the argument that N of 1 studies can be powerful because they reveal how some intervention affects a single person over time, instead of at the group level between groups, as is how typically RCTs in the past have been constructed. Increasingly, however, our statistical models include both individual changes as well as individual differences in baselines as well as group level treatment effects (these are called mixed effects models, and they are increasingly widely used).
posted by Lutoslawski at 9:44 AM on March 14, 2020 [2 favorites]


Michael Apted's Up series is an N of 14 study.
posted by Harvey Kilobit at 11:38 AM on March 14, 2020 [1 favorite]


Essentially a within subjects design with a very low N. Perception research in psychology has been using extremely low N studies with huge numbers of within subject trials for ages. I'm not sure how they have fared in psychology's replication crisis / statistical awakening.
posted by srboisvert at 1:12 PM on March 14, 2020


« Older Now that's using your noodle   |   If kin work paid just minimum wage Newer »


This thread has been archived and is closed to new comments