13 Research methods, statistics, critical review and evidence-based practice13.3 Experimental studies
13.3.2 N-of-1 trials
2 min read1 figure
Written by Resident Doctor Committee · Published
New in the 2027 syllabus
The College added this topic for exams from January 2027.
One person, repeated comparisons
An n-of-1 trial compares two or more treatments in a single person. The person switches between treatments over repeated periods. The order of treatments is usually randomised, and the person and assessor are blinded where this is practical.
The question it answers is personal: does one treatment help this individual more than another? A single trial cannot tell you the average effect across a population.
It suits some conditions better than others:
The condition should be stable enough to allow repeated comparisons.
Each treatment's effect should start and wear off within periods short enough to be workable.
Design and appraisal
Feature | Why it matters |
|---|---|
Randomised order | Stops one treatment lining up by chance with a trend over time |
Repeated cycles | Shows whether a difference keeps recurring, rather than reflecting one unusually good period |
Washout | Limits effects left over from the previous treatment |
Blinding | Reduces bias from expectations and from how outcomes are assessed |
Prespecified outcome | Stops anyone picking the most favourable result afterwards |
Several things can still go wrong:
Carry-over happens when the previous treatment still affects the next period, even though the person is now on something different.
Period effects happen when the condition itself changes over time, regardless of treatment.
Adherence and co-interventions should be recorded in each period. Changes in sleep, other medication or attendance can explain an apparent difference.
Irreversible treatments, and treatments with a lasting curative effect, usually rule the design out. The person cannot return to baseline, so a crossover comparison means little.
Illustrative comparison
The scores below are hypothetical. Each is the average for one period in one person, and lower scores are better.
Cycle | Treatment A | Treatment B | A minus B |
|---|---|---|---|
1 | 6 | 8 | -2 |
2 | 5 | 8 | -3 |
3 | 7 | 8 | -1 |
Work out the difference within each cycle, keeping its sign.
Mean difference = (-2 - 3 - 1) / 3 = -2 points. Descriptively, this favours A.
Compare this difference with a change you decided in advance would be meaningful, and consider its uncertainty. Three averages on their own do not show statistical significance.
Interpreting the result
Serial correlation means that observations taken close together in time tend to be related. If you treat every daily score as if it came from a separate patient, the result looks more precise than it really is.
Benefit and burden cover more than the main symptom score. Weigh up adverse effects, convenience and what the person prefers.
A planned series of n-of-1 trials, analysed in a suitable way, can support conclusions about a population. It still keeps estimates of how individuals vary.