How do we know a medicine actually works?

Suppose you have a cold. You drink a cup of hot lemon and honey. Four days later the cold is gone.

Did the lemon and honey cure your cold?

Obviously not — colds go away by themselves. But notice how completely the evidence fits anyway. You did a thing, and then you got better. If you did not already know that colds resolve on their own, that story would be fairly convincing.

This is the central problem in all of medicine. People get better on their own. Diseases come and go in cycles. Symptoms are worst on the day you finally decide to try something, which means they are very likely to be milder tomorrow no matter what you do.

So "I took it and I improved" is almost worthless as evidence. Not because people lie. Because getting better is what bodies usually do.

The fix: compare against something

The solution is almost embarrassingly simple, and it took a very long time to become standard.

Split people into two groups. Give one group the medicine. Give the other group something that looks identical but contains nothing — a placebo. Then compare.

Now you are not asking "did people get better?" You are asking the only question that matters: did more people get better with the drug than without it?

The second group is called the control group. It is the single most important idea in clinical research, and it is the thing to look for first whenever anyone claims a treatment works.

Two more details that matter

Randomise. Decide who goes in which group by chance, not by choice. If a doctor picks, they will tend — honestly, without meaning to — to put sicker patients in the treatment group, because they want to help them. That alone can ruin the result.

Blind it. Ideally nobody knows who got what, not the patients and not the doctors measuring the outcome. People who know they got the real drug report feeling better. People measuring results they are hoping for see improvement. This is not dishonesty; it is how human perception works, and the only reliable fix is to not know.

Put those together and you have a randomised controlled trial. It is the best tool medicine has, and it is a deeply unglamorous idea: compare against something, choose by chance, and don't tell anyone.

What to look at in a result

When you see a trial result, ask for the numbers in both arms.

"Improved by 60 percent" means nothing on its own. If 50 out of 100 improved on the drug and 45 out of 100 improved on placebo, that is a very different medicine from one where 50 improved and 10 did not.

The gap between the two groups is the effect of the drug. Everything else was going to happen anyway.

Try this

1. Go to ClinicalTrials.gov: https://clinicaltrials.gov
2. Search for a condition you know something about — eczema, asthma, migraine.
3. Open a completed trial with results posted, and find the Outcome Measures.
4. Write down the number for the treatment group and the number for the placebo group. Then subtract.

That subtraction is the honest answer to "does it work?" Everything else is presentation.

One question to take away

If a trial has no control group at all, what exactly has it proved?