Notes on Evidence-Based Medicine and Data Transparency

What Epidemiology Is

Epidemiology is the science of how we know in the real world whether something is good for you or bad for you.
It helps explain why sensational headlines (e.g., about coffee, Wi‑Fi, or olive oil) can be inconsistent or misleading.
The aim is to unpick the evidence behind claims; this is socially useful and a powerful explanatory tool.

EBM = critical appraisal of evidence to determine what really helps or harms people.
It relies on data, not authority or prestige.
Academic critique (e.g., Q&A in journals) is normal and valuable.

Authority alone is unreliable in science; reasons matter more than titles.
Example: Gillian McKeith (TV diet guru) with an unaccredited PhD and dubious credentials; illustrates why credentials aren’t enough.

Red wine and breast cancer: lab findings (enzyme behavior with grape compounds) do not translate to personal risk reduction; alcohol increases breast cancer risk.
Observational claims (olive oil + vegetables reducing wrinkles) reflect associations, not causation; confounding factors matter.
What we need: well-conducted human studies, ideally trials, to establish causation.

Observational studies: snapshot associations; prone to confounding (socioeconomic status, lifestyle).
Trials (randomized, controlled) are the gold standard for establishing causation.

A well-designed trial compares a new treatment to the best available alternative, not just to nothing.
The fish oil trial (3{,}000 children) lacked a proper control group and was vulnerable to placebo effects and natural maturation.
Placebo effect: beliefs and expectations can change outcomes; two sugar pills can outperform one; even a saltwater injection can outperform a dummy pill.
Trials must be properly controlled to attribute effects to the treatment.

Media and marketers can distort evidence; misreport, cherry-pick, or oversimplify.
Industry trials use similar tricks with slightly more sophistication, but still aim to shape outcomes.
Key idea: control groups, appropriate comparators, and realistic dosing are essential for fair trials.

Industry-funded trials are more likely to report positive results, but negative data are often missing.
The missing data problem skews the apparent effectiveness of treatments.
A top-priority problem: having access to all trial data is essential for accurate judgments.

Funnel plots: graphs that show trial size vs. effect; asymmetry suggests publication bias or missing small negative trials.
Data withholding can be inferred from statistical patterns or stories of withheld trials.
Reboxetine example: 76% of trials were withheld from clinicians and patients.

About half of antidepressant trial data have been withheld historically; similar issues exist with other medications.
The Nordic Cochrane Group and European agencies have struggled to obtain complete data for systematic reviews.
Tamiflu data access illustrates large-scale implications for policy and public health.
Ethical problem: decisions cannot be made without all information.

Sunlight (data transparency) is the best disinfectant for science.
Problems persist due to a culture of tediousness around data sharing.
Actionable takeaway: promote full data availability and rigorous, independent analysis to improve real-world decisions.

Epidemiology helps verify real-world health claims beyond sensational headlines.
Prioritize evidence over authority; seek randomized, placebo-controlled trials with appropriate comparators.
Be wary of observational studies and confounding factors.
Recognize and account for placebo effects in trial design.
Be vigilant about publication bias and data withholding; demand complete data access.
Use transparency and independent reviews (e.g., Cochrane-style analyses) to inform decisions.