Observational Studies vs Randomized Trials
An observational study shows that two things tend to go together; a randomized trial assigns people to a treatment by chance, so it can test whether one thing causes the other.
On this page
- The difference in one table
- Why observational studies can mislead
- Why trials can miss things too
- Example 1: multivitamins, where the two disagreed
- Example 2: fiber, where they point the same way
- How to weigh them
- Related
The difference in one table
| Observational (cohort) study | Randomized controlled trial (RCT) | |
|---|---|---|
| Who decides what people eat or take | The people themselves | Chance (randomization) |
| Typical size | Tens of thousands | Dozens to a few thousand |
| Typical length | Years to decades | Weeks to a few years |
| How intake is measured | Questionnaires, sometimes one urine sample | Assigned, sometimes provided |
| What it can show | Association ("linked to") | Cause, for the outcome measured |
| Main weakness | Other differences between groups | Short, narrow, hard to stick to |
Why observational studies can mislead
People who choose a habit often differ in other ways. Someone who takes a vitamin every day may also see a doctor more, smoke less or already be ill and trying to help themselves. These hidden differences are called confounders.
Researchers adjust for the confounders they know about. They cannot adjust for the ones they did not measure.
Measurement is the other weak point. O'Donnell et al., 2014 (PURE) followed 101,945 people in 17 countries, but estimated sodium intake from a single morning urine sample. See the PURE sodium study explained and sodium on keto.
Why trials can miss things too
Randomization balances the groups, so a difference at the end is likely due to the treatment. But trials have their own limits:
- Narrow population. A trial in older men tells you about older men.
- Short length. Diseases like cancer and heart disease take years to show up.
- One dose, one product. A trial tests one pill or one diet, not the whole category.
- Adherence. People assigned to a diet do not always follow it.
Example 1: multivitamins, where the two disagreed
| Study | Type | Who | Finding |
|---|---|---|---|
| Mursu et al., 2011 | Observational (Iowa Women's Health Study) | 38,772 older women | Multivitamin use associated with slightly higher total mortality (HR 1.06) |
| Gaziano et al., 2012 | RCT (PHS II) | 14,641 male physicians 50+, median 11.2 years | Daily multivitamin gave a modest reduction in total cancer (HR 0.92) |
| Sesso et al., 2012 | RCT (PHS II) | Same trial | No reduction in major cardiovascular events (HR 1.01) |
| Khan et al., 2019 | Umbrella review of 277 trials | Many trials pooled | Most supplements, multivitamins included, had no significant effect on mortality or cardiovascular outcomes |
How to read this: the cohort found slightly higher risk among women who took multivitamins. The trial in men found a small cancer benefit and no heart benefit. They studied different people, measured different outcomes and used different designs, so they are not direct contradictions.
The cohort cannot say whether the vitamins caused anything; the women who took them may have differed in other ways. The trial cannot say what happens in women. Taken together with the umbrella review, the overall picture is no clear effect on death or heart disease. More in vitamins on keto and do multivitamins work?.
Example 2: fiber, where they point the same way
| Study | Type | Finding |
|---|---|---|
| Reynolds et al., 2019 | Comparison of highest vs lowest fiber intake | 15–30% lower all-cause and cardiovascular mortality, heart disease, stroke, type 2 diabetes and colorectal cancer; largest reductions at 25–29 g/day |
| Aune et al., 2011 | Prospective studies | Each 10 g/day of total fiber associated with about 10% lower colorectal cancer risk |
| Threapleton et al., 2013 | Prospective cohorts | Each 7 g/day of total fiber associated with 9% lower cardiovascular and heart disease risk |
| Brown et al., 1999 | 67 controlled trials | 2–10 g/day of soluble fiber gave small, significant falls in total and LDL cholesterol |
| Zhao et al., 2018 | Randomized trial in type 2 diabetes | A high-fiber diet promoted certain gut bacteria; HbA1c improvement tracked their increase |
How to read this: the large cohort studies link more fiber with lower disease risk. The trials cannot run long enough to count heart attacks or cancers, but they do show fiber changing cholesterol and blood sugar markers in the expected direction. When cohorts and trials agree like this, confidence is higher than either gives alone.
The percentages here are relative. "15–30% lower" compares the highest and lowest fiber groups, not your own odds. See relative vs absolute risk, fiber on keto and soluble vs insoluble fiber.
How to weigh them
- One cohort study alone: a reason to look closer, not a reason to change your diet.
- One trial alone: strong evidence for that population, that dose and that time span.
- Cohorts and trials agree: the strongest case available for most diet questions.
- They disagree: check whether they studied the same people, the same dose and the same outcome before deciding one is wrong.
KetoDick on this: he warns against confirmation bias, saying a keto forum will answer every question with keto, and suggests some forum weight loss may come from tracking, meal prep and paying attention rather than from keto itself (video 1m1hPgaRIg4). That is a confounder in plain terms: something other than the diet that changed at the same time. He also says the research stands apart from the personalities who promote it (video 1JbK1_ClDmY).
If a study leads you toward a supplement and you take blood pressure medication or have kidney disease, check with your doctor first, especially for potassium or magnesium.
Related
Guide: Chapter 12: Reading nutrition news and studies · How to read an abstract · Diet headlines · Red flags in diet advice · Fiber on keto · Vitamins on keto · Sodium
Blog: Do multivitamins work? · The multivitamin memory study · The PURE sodium study explained · Soluble vs insoluble fiber · Relative vs absolute risk
Videos: Keto is dangerous and warnings from friends · Men's Health hack job on keto
Last verified: 2026-10-31