The thesis thatfound nothing, andwhy it passed

Last June a student defended a study comparing two commonly taught techniques for the same presentation. She had recruited forty-one patients from our clinic, randomised them properly, and used an outcome measure she had not chosen after seeing the data.

The result was nothing. No difference between the groups that survived contact with a confidence interval.

Why that is a good thesis

The panel gave it one of the highest marks of the year, and three people asked afterwards whether that was a mistake. It was not.

A thesis is assessed on the quality of the question, the method, and the honesty of the interpretation. It is not assessed on whether the profession comes out looking good. A study that is well built and finds nothing tells you more than a study that is badly built and finds something.

The alternative is worse. If we only reward positive results, we teach students to go looking for them — to change the outcome measure late, to drop the inconvenient participants, to describe a null result as a trend. Those habits do not stay in the thesis. They follow people into practice.

What she actually concluded

Her discussion is the part worth reading. She did not conclude that the techniques do not work. She concluded that her study could not distinguish between them at the sample size available, and she calculated what sample size would have been needed. That number was 260.

That is a useful finding for the next student who takes the question on. It is also an honest description of the limit of what one person can do in eighteen months alongside a full clinical load.

The uncomfortable part

We are a school that trains people to do a manual therapy. It would be convenient if every study we ran confirmed that the therapy works. Several of them do not, and we publish those internally alongside the rest.

Students notice quickly whether a school means what it says about evidence. The fastest way to lose them is to teach method in year four and then flinch in year five.