Explain why people improve without any treatment working.
A hundred people with back pain take a new remedy. Six weeks later, seventy say they feel better. What has been shown?
Yes.Without a comparison group you cannot separate the remedy from the passage of time. Many conditions improve on their own, and that improvement is what the remedy gets credited with.
Not quite.That would need to know how many would have improved with nothing. The trial as described cannot say.
Not quite.The 'why' is a later question. This design cannot establish the 'whether'.
Almost everything gets compared against getting better on its own. A study without a control group is measuring time and calling it treatment.
Which of these can make people improve, with no active treatment at all?
The condition running its natural course.
Joining at the worst point, then drifting back to normal.
Expecting to feel better.
Being watched and asked about it weekly.
The drug's chemistry.
Yes.Four routes to improvement before the chemistry is even considered. A trial's entire job is to hold all four constant so the fifth can be seen.
People usually join a study when their symptoms are at their worst. Slide forward in time and watch what happens with no treatment at all.
Week 0Week 3Week 6
Symptom score8.5/10
Their usual level5/10
Week 0They joined today, because today was bad. That is what made them seek help.
Symptom score6.5/10
Their usual level5/10
Week 3Drifting back toward their normal. No treatment involved.
Symptom score5.1/10
Their usual level5/10
Week 6Back at their usual level — a 40% improvement, from nothing at all.
This effect appears in almost every study of a fluctuating condition. Why?
Yes.It is called regression to the mean, and it is the most reliable source of false improvement in medicine. Anything that recruits people at their worst gets a free-looking recovery.
Not quite.That happens too, and it is a separate effect. This one occurs even if nothing whatever changes.
Not quite.The pattern appears with objective measurements as well.
Move the control to see what changes.
Someone says a treatment worked for them personally. What is the reasonable response?
Yes.Both halves matter. Dismissing the improvement is wrong — it happened. Attributing it to the treatment is the step the evidence cannot support from one case.
Not quite.The feeling is real data. It is the cause that is unidentified.
Not quite.Single cases have launched a great deal of good research. They cannot settle a question of cause.
In a trial, 70 of 100 treated people improve, and 60 of 100 untreated people improve. How many percentage points of improvement can the treatment account for?
points
Yes.70 − 60. The untreated group is what turns an impressive-sounding 70% into a real, and much smaller, finding — and that arithmetic is why the control group exists.
Put these in order of how much they can establish about cause.
Tap them in order — first to last.
One case→Many cases→Compared with similar people→Randomly assigned
Each step buys the removal of a different kind of doubt.Yes.Each step removes a class of alternative explanation. The last one removes the whole family of hidden differences at once, which is the subject of the next lesson.
Lesson complete
Most people improve anyway, so a study with no comparison group measures time.