What Makes a Good Study?

Not all research is equal. Well-designed trials follow strict rules so their conclusions can be trusted; poorly designed ones can produce results that look impressive but mean little. Here is how to tell them apart.

1. A control group

A treatment must be compared against a control. The placebo effect for conditions like pain and depression can be surprisingly large, so a placebo-controlled trial gives some participants an inert substance — and the treatment has to beat it by a meaningful margin. If 58% improve on the herb but the same share improve on placebo, the herb probably is not doing much. A study may instead compare against an established treatment such as fluoxetine. Controls also let researchers measure side effects accurately. Be skeptical of uncontrolled "preliminary" trials.

2. Randomization

Participants should be assigned to the treatment or control group at random. Randomization helps separate the effect of the treatment from differences between the people in each group.

3. Double-blind design

Neither the patients nor the doctors should know who is getting the real treatment until the study ends. This prevents expectations from biasing the results. Single-blind studies, where doctors know, are more vulnerable to bias.

4. Sample size

Larger studies — ideally hundreds of patients across multiple centers — are more representative and statistically stronger. A small study (say, 30 people) can still be valid if it is properly controlled, randomized, double-blind and well analyzed.

5. Natural-history control

Some trials add a third, untreated group to observe how the condition progresses on its own, providing extra context for the results.

The bottom line

The most trustworthy studies are controlled, randomized and double-blind. Large, multi-site trials with hundreds or thousands of participants offer the strongest evidence of all.