The Field Guide · No. 10
RCTs vs. observational studies: tested or watched
In a clinical trial, also called an interventional study, researchers assign the treatment, usually by chance; in an observational study, such as a cohort or case-control study, they only record what people already do, which is why trials are better at showing cause.
Updated
A clinical trial and an observational study differ in one decision: who picks the treatment. In a clinical trial, which ClinicalTrials.gov also calls an interventional study, the researchers assign it. In an observational study they do not. They record what people were already doing, or look at data already collected, and compare how those groups fare. A randomized controlled trial, or RCT, adds one more step, assigning the treatment by chance, and that step is what lets it support cause and effect.
In an RCT, researchers randomly sort people into groups, one that gets the treatment and one that does not. Because the sorting is random, the groups end up similar in every other way, so if their outcomes differ, the treatment is the likely cause. This is why RCTs are called the gold standard: randomization is what lets a study claim cause and effect.
In an observational study, including cohort studies, researchers do not assign anything. They watch groups of people who already made their own choices, such as who eats a lot of fish or who exercises, and compare how they fare. These studies are valuable, often the only ethical or practical option, and they can uncover important patterns. But because the groups differ in countless ways beyond the one being studied, they show links, not proof of cause.
So when a study makes news, find out which kind it is. "People who did X had less Y" usually signals an observational study, a lead to follow up, not proof. "People randomly assigned to X had less Y" signals a trial that can actually support cause. The study type is the fastest way to know whether you are reading a strong finding or an interesting hint.
The two designs in official definitions
ClinicalTrials.gov, run by the National Library of Medicine, names two main types of clinical research, clinical trials and observational studies. Both may try to learn about an intervention such as a drug, behavior or device. The main difference is that clinical trial participants are assigned to get an intervention, and observational study participants are not.
The National Cancer Institute defines a randomized clinical trial as one in which participants are divided by chance into separate groups that compare different treatments or other interventions, so the groups will be similar and the treatment effects can be compared more fairly. It defines an observational study as one in which people are observed or certain outcomes are measured, with no attempt to affect the outcome.
Cohort and case-control studies
Two observational designs turn up most often. The institute describes a cohort study as one that compares an outcome in groups of people who are alike in many ways but differ by a certain characteristic, such as female nurses who smoke compared with those who do not. It describes a case-control study as one that compares people with the disease under study (cases) with a very similar group who do not have it (controls), then looks back at their medical and lifestyle histories for factors that may be associated with the disease.
RCT vs. cohort study, side by side
A cohort study is the observational design most often set against a trial, because both follow people forward and count outcomes. The difference is who decided the exposure. The table lines up the trial with the three observational designs that turn up most in health news.
| Design | Who decides the exposure | Direction in time | What it can show |
|---|---|---|---|
| Randomized controlled trial | The researchers, by chance | Forward | Cause and effect, for the treatment tested |
| Cohort study | The participants, or their circumstances | Forward, or through existing records | How often the outcome occurs in each group, and a link |
| Case-control study | The participants, or their circumstances | Backward, from the disease to past exposures | A link, reported as an odds ratio |
| Cross-sectional study | The participants, or their circumstances | One moment | How common something is, and a link |
The order of events is what separates the cohort and the trial from the other two. Both watch the exposure first and the outcome later. What the cohort cannot do is make its groups alike in everything else, and that is the job randomization does.
RCT vs. non-RCT
A non-randomized trial is not an observational study. The researchers still give the treatment, so it counts as a clinical trial, but the groups are not decided by chance. ClinicalTrials.gov asks every trial to declare its allocation: randomized means participants are assigned to intervention groups by chance, and nonrandomized means they are expressly assigned through a non-random method, such as physician choice. A trial with only one group, a single-arm trial, has no allocation to declare.
So "non-RCT" covers two different things in practice: trials without randomization, and, in looser use, observational studies. Ask which one a paper means before weighing it.
How often do trials and observational studies disagree?
Less often than the gold-standard label suggests, on average. A 2024 Cochrane review gathered 47 reviews that had compared RCTs with observational studies of the same health questions, covering 2,869 trials and 3,924 observational studies, and pooled the 34 that supplied usable data. The ratio between their effect estimates was 1.08 (95 percent confidence interval 1.01 to 1.15), which the authors describe as no difference or a very small difference, and they rated the certainty of that evidence as low. In 23 of the 34 reviews, the two kinds of study agreed on average.
An average agreement does not make any single observational study safe. The same review found small gaps for drug studies and for observational studies that did not use propensity score adjustment, and it points to differences in population, treatment, comparison and outcome as reasons two studies can disagree. One observational result can still be badly wrong, and the design gives no warning when it is.
Why assignment by chance matters
The institute gives a concrete case. If doctors could choose which patients go into which group, some might put healthier patients in the treatment group and sicker ones in the control group, even without meaning to. That could make the treatment group look better than the control group even if the treatment does nothing. Random assignment prevents that kind of bias.
Trials add a second guard. In a blinded study the patients, or the patients and their doctors, do not know which treatment is being given. ClinicalTrials.gov says this masking helps make sure the results are looked at fairly.
Not every trial uses a placebo. The institute notes that in cancer treatment trials, participants usually receive either the new treatment or the already approved standard treatment, and researchers compare the two.
What to remember
- An RCT randomly assigns people to groups, which lets it test cause and effect.
- An observational study only watches people's existing choices, so it shows links, not proof.
- "Randomly assigned" signals strong evidence; "people who did X" signals a hint to hold loosely.
From the record
Clinical studies can be divided into two broad categories: trials, in which the investigator intervenes to prevent or treat a disease, and observational studies, in which the investigator makes no intervention and patients are allocated treatment based on clinical decisions.
Asked often
Why are randomized trials called the gold standard?
Because random assignment makes the groups similar in every way except the treatment, so a difference in outcomes can be credited to the treatment itself. That is what lets an RCT support a cause-and-effect claim, which observational studies cannot do alone.
Are observational studies useless then?
Not at all. They are often the only ethical or practical way to study something, and they surface important patterns. They just show associations rather than proof, so their findings should be treated as leads to confirm, not settled cause.
Is an interventional study the same as a clinical trial?
Yes. ClinicalTrials.gov lists clinical trials, also called interventional studies, as one of the two main types of clinical research. Researchers assign participants to an intervention. In the other type, observational studies, participants are not assigned anything.
What is the difference between an RCT and a non-RCT?
In a randomized controlled trial, chance decides who gets which treatment. In a non-randomized trial the researchers still give the treatment, but people are placed in groups some other way, such as by physician choice, or there is only one group. Without randomization the groups can differ in ways that get credited to the treatment.
What is the difference between a cohort study and a case-control study?
Both are observational. A cohort study follows groups that differ by a characteristic, such as smokers and non-smokers, and compares outcomes. A case-control study starts with people who have a disease and a similar group who do not, and looks back at their histories for factors linked to the disease.
Does an RCT always need a placebo?
No. In cancer treatment trials, for example, the comparison group usually receives the already approved standard treatment, and placebos are rarely used.
Further reading
Go deeper
Paid link. If you buy a book through this link, Bookshop.org pays us a small commission and sends a share of the sale to independent bookstores.
-
Testing Treatments (opens Bookshop.org)
Imogen Evans, Hazel Thornton, Iain Chalmers and Paul Glasziou · 2011
Written for patients by clinicians and Cochrane co-founder Iain Chalmers, this book explains why fair comparisons and randomization are needed to know whether a treatment works.
Read the news better, every morning.
We send you the day's real progress in science, medicine and beyond, told plainly and sourced. Free, forever.
Free for good. One click to leave.