The trial with one participant

There's a type of trial where the subject and the entire population are the same person. Some consider it the strongest evidence there is — for you specifically.

An ordinary trial splits a hundred or a thousand patients between two groups and compares the groups afterwards. An N-of-1 trial does something different: it takes one patient and distributes the treatments over time instead of over people. Same method — randomisation, ideally blinding — but compressed into one person.

That sounds like a weakness. One patient surely can't say anything about a population. But that isn't what the trial is attempting either. The question isn't "does this work in general". The question is "does this work for you" — and that's a different question, which needs a different design.

Why an average of a thousand people can be the wrong answer

A review of the literature on weather and migraine has already put its finger on one of the problems: some patients are sensitive to low air pressure, others to high. Pool them into one large cohort, and the two effects can statistically cancel each other out, even though both are real for the patients they apply to. The average says "no connection". That's true for the average and false for both of the groups it's made of.

An N-of-1 trial sidesteps the problem by not asking about the average at all. It only asks about you.

Add many of them together, and you still get an average

The interesting part is that you don't have to choose between the individual and the group. A planned series of N-of-1 trials across several comparable patients can be pooled statistically, typically using methods such as individual patient data meta-analysis or Bayesian hierarchical models. The result is two figures at once: a population average, and each individual person's deviation from it.

That's a different kind of average from the one that hides the two opposing groups. Here nobody disappears into the figure — you can always go back and see exactly how far this particular person sat from the middle.

What Aneroid isn't

It's tempting to call Aneroid an N-of-1 trial. It isn't one, and it's worth saying clearly why. An N-of-1 trial randomises a treatment over time — period A, period B, in random order, often blinded, so neither patient nor clinician knows which period is which. Aneroid gives no treatment. There's no period A and B to distribute. There's only the pressure, which does what it does, and your answers about how you're feeling.

What the app borrows isn't the trial design. It's the thinking: that the question "does this work for me" isn't answered by reading an average out of a paper, but by measuring yourself, over time, with something to compare against. The three daily questions with no pressure drop are the part of that thinking the app can actually build — a control period, not a randomised treatment.