Randomized controlled trial

Suppose you want to find out whether a new intervention actually changes an outcome. The simplest approach is to compare people who tried the intervention with people who did not. However, people who volunteer to take part may already pay more attention to their health. In that case, even if the results differ, it is difficult to tell whether the difference is due to the intervention or to differences in the participants' original characteristics.

A randomized controlled trial addresses this problem through the way it assigns participants to groups. Participants are divided randomly, using something like a computerized draw or a coin toss. One group (the intervention group) receives the method being tested, while the other (the control group) receives no treatment or continues its usual care. After a set period, the results of the two groups are compared.

Suppose you want to test whether a new fertilizer helps plants grow better using twenty potted plants. If you select only the ten plants that look particularly healthy to receive the fertilizer, the comparison is not fair. You would be giving fertilizer to plants that were likely to grow well anyway.

What if you draw lots for each plant, giving fertilizer to ten and none to the other ten? Conditions such as sunny placement, soil condition, and initial size are distributed similarly between the two groups. The greatest advantage of random assignment is that even conditions the researchers have not thought of are distributed evenly.

• Random assignment. Factors that could affect the outcome, such as age, health status, and lifestyle habits, are distributed evenly between the two groups. The same applies to factors that have not been measured.
• A control group. A comparison group makes it possible to separate out improvements that occur naturally over time and changes arising from the expectation of receiving treatment. These changes relate to Regression to the mean and the Placebo effect.
• Blinding. Keeping participants or the people measuring outcomes unaware of which group participants belong to can reduce the influence of expectations or preconceptions on the results. See Blinding for details.
• A predefined plan. Deciding which outcomes to measure and how to measure them before a study begins makes it harder to select and report only favorable findings after seeing the results.

The British Medical Research Council's (MRC) 1948 trial of streptomycin for tuberculosis is often cited as a milestone that brought random assignment in clinical trials to widespread attention. The method later came to be used not only for drugs but also for studies comparing exercise programs, education, and lifestyle interventions. Today, guidelines such as CONSORT set out how studies should be reported, so papers often describe their procedures using a common framework.

• What was compared? First, look at what treatment each of the intervention and control groups received. The meaning of the results differs depending on whether the control group received 'no treatment' or 'continued usual care.'
• How many people took part? With few participants, chance differences can appear large. Looking at the Confidence interval alongside the results can help you judge how precise the estimate is.
• How many people dropped out? If many people do not complete the study, the balance between the two groups can be disrupted, making it difficult to compare them on the same equal footing as at the start.
• Is the difference small or large? A statistically significant difference is not the same as a difference large enough to notice in everyday life. Look at both the Effect size and the Minimal clinically important difference.
• Is there more than one study? When several studies address the same question, their findings are often brought together in a Systematic review or Meta-analysis. Looking for these resources can give you a more reliable grasp of the overall findings.

Not every question can be investigated this way. Observational studies are used to examine exposures such as smoking that are difficult to assign randomly, very rare outcomes, or changes that occur over decades. Researchers also continually grapple with the difficulty of fully applying blinding to interventions such as hands-on treatment or exercise, where it is hard to conceal who is delivering the intervention. That is why it is important to understand the strengths and weaknesses of each study design and choose a method suited to the question.