How to Read a Nutrition Study Without Getting Fooled

Why Nutrition Science Seems So Contradictory

One week eggs are killing you, the next week they’re a superfood. Coffee causes cancer on Monday and prevents Alzheimer’s on Friday. This chaos isn’t because scientists are incompetent — it’s because nutrition is genuinely difficult to study, and media incentives favor sensational headlines over accuracy.

This guide will arm you with the tools to evaluate nutrition studies yourself, so you can stop being whiplashed by headlines.

Study Design: The Most Important Factor

Observational Studies vs. Randomized Controlled Trials

Observational studies watch people who already eat a certain way and track outcomes. They’re useful for generating hypotheses but cannot prove causation. When you read “people who eat X have lower rates of Y,” that’s observational data.

The problem: people who eat more vegetables also tend to exercise, smoke less, and have higher incomes. Untangling which factor caused the health benefit is often impossible.

Randomized controlled trials (RCTs) assign people to groups randomly and control variables. These can demonstrate causation. However, nutrition RCTs are expensive, often short-term, and compliance is difficult (people don’t stick to assigned diets).

When evaluating claims like whether intermittent fasting is safe, look for RCT evidence rather than relying solely on observational associations.

Correlation vs. Causation

This is the single most abused concept in nutrition media. Ice cream sales and drowning deaths are correlated — because both increase in summer. That doesn’t mean ice cream causes drowning.

Similarly, many nutrition correlations have hidden confounders. The famous “red wine is healthy” narrative likely reflected the fact that moderate wine drinkers in studies were wealthier, more educated, and had better overall health behaviors than non-drinkers.

How to spot it: Look for language like “associated with” or “linked to” — these are observational claims. Only interventional studies can use “causes” or “prevents” with confidence.

Relative Risk vs. Absolute Risk

This is how headlines manufacture fear. Consider this real example:

• “Processed meat increases colorectal cancer risk by 18%!” (relative risk)
• Absolute translation: your lifetime risk goes from about 5% to about 5.9%

Both statements are technically accurate, but they feel very different. Always look for absolute numbers. A 50% relative risk increase from a rare baseline might be trivial, while a 10% relative risk increase from a common baseline might be significant.

Sample Size and Statistical Power

A study with 12 participants cannot reliably detect small effects. General rules of thumb:

• Under 30 participants: very preliminary, treat as hypothesis-generating
• 30-100 participants: small, may detect large effects
• 100-500: moderate, can detect medium effects
• 500+: well-powered for most nutrition questions
• Meta-analyses combining thousands: most reliable

When you see a dramatic supplement claim based on a study of 15 people, appropriate skepticism is warranted.

Funding Bias

Industry-funded studies are not automatically wrong, but they are 4-8x more likely to reach conclusions favorable to the sponsor. This has been demonstrated across pharmaceutical, food, and supplement research.

How to check: Look at the “Conflicts of Interest” or “Funding” section at the bottom of papers. If a sugar study was funded by Coca-Cola, apply extra scrutiny.

Similarly, when assessing whether caffeine has long-term health risks, note that much coffee research has been funded by the coffee industry.

P-Hacking and Multiple Comparisons

If you measure 20 different outcomes, one of them will likely show a “statistically significant” result by pure chance (at the p < 0.05 threshold). This is called the multiple comparisons problem.

Red flags:

• A study measured many outcomes but only reports one significant finding
• The reported outcome wasn’t declared in advance (no pre-registration)
• The “significant” result has a p-value hovering just below 0.05

Pre-registered studies (where outcomes are declared before data collection) are much more trustworthy.

Publication Bias

Studies with exciting positive results get published. Null results (“we found nothing”) often sit in file drawers. This creates a distorted literature where effects look larger than reality.

How to account for it: Systematic reviews that search for unpublished data, funnel plots in meta-analyses, and pre-registered studies all help address this bias.

Duration Matters

A 2-week diet study tells you almost nothing about long-term health. Weight loss studies under 6 months rarely capture regain. Supplement studies under 8 weeks may miss both benefits that take time to develop and harms that accumulate slowly.

For chronic disease claims, look for studies lasting months to years, or long-term cohort data.

Your Practical Toolkit

Next time you encounter a nutrition headline, ask:

1. What type of study? (Observational → can’t prove causation)
2. How many people? (Under 50 → very preliminary)
3. How long? (Under 4 weeks → limited conclusions)
4. Who funded it? (Industry → extra scrutiny)
5. Relative or absolute risk? (Always convert to absolute)
6. Has it been replicated? (Single study → wait for more)

These six questions will protect you from 90% of misleading nutrition claims. Science literacy isn’t about having a PhD — it’s about knowing which questions to ask.