Evidence and research methods
How to Read Antidepressant Effect Sizes Without Overclaiming
How mean differences, standardized effects, response odds, placebo change, duration, and reporting bias shape antidepressant evidence.
Health & Evidence Library
Evidence and research methods
Tools for reading studies, interpreting tests and effect estimates, recognizing uncertainty, and asking whether evidence applies.
207 guides · Page 3 of 4
Evidence and research methods
How to Read an N-of-1 Trial: A Randomized Experiment for One Person
How repeated treatment periods, randomization, blinding, washout, outcomes, and analysis turn one person's treatment comparison into evidence.
Evidence and research methods
How to Read a Stepped-Wedge Trial When Treatment and Time Move Together
How randomized rollout, secular trends, clustering, transition periods, and time-adjusted analysis determine a stepped-wedge result.
Evidence and research methods
How to Read a RoB 2 Risk-of-Bias Assessment
How Cochrane RoB 2 assesses bias in a specific randomized-trial result across randomization, deviations, missing data, measurement, and selective reporting.
Evidence and research methods
How to Read a Responder Analysis
Responder analyses turn continuous scores into yes or no. Learn what the threshold clarifies, what information it loses, and what to check.
Evidence and research methods
How to Read a Methods Section Like a Reviewer
A practical methods-section audit for checking whether a study's design, participants, outcomes, and analysis support its claims.
Evidence and research methods
How to Read a Crossover Trial: Periods, Carryover, and Paired Evidence
How crossover trials use within-person comparisons, when washout works, and how period effects, dropout, and analysis shape validity.
Evidence and research methods
How to Read a Cluster Randomized Trial Without Overcounting the Evidence
How group randomization changes sample size, analysis, recruitment bias, estimands, and interpretation in a cluster trial.
Evidence and research methods
How to Interpret an E-Value Without Overclaiming
What an E-value quantifies, how to benchmark it, and why it cannot rule out bias or turn an observational association into causation.
Evidence and research methods
How to Audit Difference-in-Differences and Interrupted Time Series Studies
A practical guide to the counterfactual assumptions, graphs, time trends, and modern checks behind two common quasi-experiments.
Evidence and research methods
How Surrogate Endpoints Are Validated and Used
Why biomarker correlation is insufficient, how trial-level validation works, and how FDA distinguishes validated and reasonably likely surrogates.
Evidence and research methods
How SGLT2 Inhibitors Became Heart Failure Therapy
EMPA-REG, DAPA-HF, EMPEROR, and DELIVER moved SGLT2 inhibitors from glucose lowering to heart failure care across ejection fraction.
Evidence and research methods
How Overdiagnosis Is Measured, and Why the Estimates Vary
Why overdiagnosis is a population estimate, how trials, cohorts, pathology studies, and models measure it, and why denominators matter.
Evidence and research methods
How Corrections, Expressions of Concern, and Retractions Change a Paper
A guide to the notices journals use to repair the scholarly record, what each status means, and how readers can verify the current version.
Evidence and research methods
How Audit Trails Make Clinical-Trial Data Traceable
ICH E6(R3), ALCOA principles, metadata, and secure audit trails explain how a clinical result can be traced back to its source.
Evidence and research methods
How a Patient-Reported Outcome Measure Earns Trust: Reading It Through COSMIN
A guide to content validity, structure, reliability, error, construct validity, responsiveness, interpretability, and feasibility in PROM research.
Evidence and research methods
How a Clinical Prediction Model Earns Trust
How TRIPOD+AI and PROBAST+AI frame model reporting, validation, calibration, clinical utility, fairness, and transportability.
Evidence and research methods
GRADE Imprecision: Confidence Intervals, Thresholds, and Information Size
How GRADE judges whether uncertainty crosses a decision threshold, when optimal information size matters, and why significance is not enough.
Evidence and research methods
FAIR Research Data: More Than Posting a Download Link
How identifiers, metadata, standard protocols, vocabularies, licenses, and provenance make research data usable by people and machines.
Evidence and research methods
Every Extra Trial Test Spends Some of the False-Positive Budget
Multiple endpoints, groups, time points, subgroups, and interim analyses raise false-positive risk unless a prespecified testing strategy controls it.
Evidence and research methods
Directed Acyclic Graphs: A Reader's Guide to Causal Adjustment
How causal diagrams distinguish confounders, mediators, colliders, and selection variables before statistical adjustment begins.
Evidence and research methods
Diagnostic Error and Patient Safety: Building a Safer Diagnostic Process
How diagnostic error is defined, why uncertainty is not the same as negligence, and how teams and patients can close safety gaps.
Evidence and research methods
Compare the Published Trial With the Outcome Plan Written Before Results
Outcome switching lets a result chosen after analysis appear prespecified. Registry histories, protocols, and analysis plans reveal what changed and when.
Evidence and research methods
Cohort or Case-Control? How Observational Designs Answer Different Questions
How sampling distinguishes cohort from case-control studies, what each estimates, and where confounding, selection, and measurement distort results.
Evidence and research methods
Build Model-Informed Drug Development Around the Decision
Use ICH M15 to connect a drug-development question, context of use, model risk, evaluation criteria, reporting, and a bounded decision claim.
Evidence and research methods
Bayesian and Frequentist Trial Results Answer Different Questions
How to interpret p values, confidence intervals, priors, posterior probabilities, and credible intervals without mixing the frameworks.
Evidence and research methods
Allocation Concealment and Blinding Guard Different Parts of a Trial
Allocation concealment protects random assignment before it occurs; blinding limits biased behavior and assessment afterward.
Evidence and research methods
A Subgroup Result Needs an Interaction Test, Not Two Separate P Values
ICEMAN helps readers judge whether an apparent treatment difference across patient groups is credible rather than a chance finding.
Evidence and research methods
A Living Systematic Review Is a Maintained Evidence System
Living reviews use continual surveillance, explicit update triggers, and version control to keep high-priority evidence syntheses current.
Evidence and research methods
A Clinically Important Difference Is Not Just a Significant Difference
The minimal clinically important difference connects a score change to patient importance, but its value depends on context, anchors, and uncertainty.
Evidence and research methods
Why the Same Test Result Means Different Things in Different Patients
How pretest probability, likelihood ratios, and the Fagan nomogram show why one lab result can mean certainty in one patient and little in another
Evidence and research methods
Why the Headline and the Study So Often Disagree
Exaggeration in health headlines usually starts in the university press release, not the newsroom. Here is how to trace any claim back to the study.
Evidence and research methods
Why Placebo Response Is So Large in Depression Trials
Why placebo response runs high in antidepressant trials, and how a strong placebo arm compresses the measured drug-placebo gap.
Evidence and research methods
Why Guidelines Stopped Recommending Aspirin for Most Healthy Adults
How newer randomized trials reshaped aspirin advice for healthy adults, weighing bleeding against benefit and why established disease is different.
Evidence and research methods
Why Antidepressant Withdrawal Numbers Disagree So Much
Withdrawal-incidence estimates swing from about 15% to over 40%. Here is how study design, not biology, drives most of that gap.
Evidence and research methods
Why '1 in 100' Beats '1 Percent': Communicating Risk Clearly
Natural frequencies like '1 in 100' and icon arrays are read more accurately than percentages, because they keep the denominator in plain sight.
Evidence and research methods
Who Gets to See the Raw Trial Data? IPD Sharing Explained
What a trial's data-sharing statement really promises, who can obtain individual participant data, and why reanalysis can confirm or overturn a result.
Evidence and research methods
What a Meta-Analysis Can and Cannot Tell Us in Diabetes Research
How systematic reviews and meta-analyses synthesize evidence, why they rank high in the evidence hierarchy, and where their limits lie.
Evidence and research methods
Two Ways to Count a Trial: Intention-to-Treat and Per-Protocol
Intention-to-treat preserves randomization and estimates a real-world effect. Per-protocol answers a narrower question and can let selection bias back in.
Evidence and research methods
Relative Risk, Absolute Risk, and Number Needed to Treat: How to Read a Statin Trial
How to turn a statin trial's headline relative risk reduction into absolute benefit and number needed to treat, so one result stops sounding two ways.
Evidence and research methods
Reading Tumor Response on a Scan: What RECIST Can and Cannot Tell You
How RECIST 1.1 turns serial scans into response categories, why immunotherapy defies those rules, and how iRECIST handles pseudoprogression.
Evidence and research methods
Predictive Versus Prognostic Cancer Biomarkers Explained
How predictive and prognostic cancer biomarkers differ, why the interaction test matters, and what PD-L1 and TMB actually tell you.
Evidence and research methods
Prebunking: The Science of Inoculating People Against Health Misinformation
What recent trials show about prebunking and inoculation theory for resisting health misinformation, and where the effects fade.
Evidence and research methods
P-Hacking: The Hidden Choices That Manufacture Significance
How undisclosed analysis choices manufacture statistical significance, and why disclosure and preregistration are the fix.
Evidence and research methods
Number Needed to Treat: How Many People to Help One
Number needed to treat counts how many people must take a treatment for one of them to benefit, turning a trial result into a figure you can picture.
Evidence and research methods
Number Needed to Harm, and How to Read It Against Number Needed to Treat
Number needed to harm counts how many people are treated before one is harmed. Read it beside number needed to treat to weigh both sides.
Evidence and research methods
Intention to Treat: Why a Fair Trial Counts Everyone It Enrolled
What intention-to-treat analysis means, why it keeps randomized trial results honest, and how dropping participants can distort what a study appears to show.
Evidence and research methods
Insulin Sensitivity and Insulin Response: Why the Two Numbers Only Make Sense Together
Insulin sensitivity and insulin response are two different measurements, and diabetes risk lives in how they relate, not in either value alone.
Evidence and research methods
How to Tell Whether a Systematic Review and Meta-Analysis Is Trustworthy
A practical order for reading a systematic review: the question, the search, risk of bias, the forest plot, heterogeneity, and the studies that went missing.
Evidence and research methods
How to Spot Spin in a Study Abstract or Press Release
Spin is the gap between what a study found and how it is sold. Read the methods, the named outcome, and the absolute numbers first.
Evidence and research methods
How to Spot a Predatory Journal
Check a journal against DOAJ, COPE membership, and the Think. Check. Submit. list rather than trusting its site or a flattering invitation email.
Evidence and research methods
How to Read a PRISMA Flow Diagram in a Systematic Review
A practical walkthrough of the PRISMA flow diagram, the numbers that make a systematic review auditable, and the patterns that should make you pause.
Evidence and research methods
How to Read a Kaplan-Meier Curve: What the Steps, the Spread, and the Censoring Marks Mean
A Kaplan-Meier curve shows the share still event-free over time. Here is how to read its steps, gaps, and censoring marks honestly.
Evidence and research methods
How to Read a Health Headline Without Getting Fooled
A four-part checklist for decoding a health headline: the study design, who was studied, absolute versus relative effect, and who paid.
Evidence and research methods
How to Read a Forest Plot Without the Jargon
A plain-language guide to reading the forest plot in a meta-analysis: the lines, the boxes, the diamond, and what they really show.
Evidence and research methods
How to Read a Conflict of Interest Disclosure
How to read an ICMJE conflict of interest disclosure, why sponsored studies tilt toward favorable results, and why naming a conflict does not cancel it.
Evidence and research methods
How to Read a Cochrane Plain-Language Summary
A part-by-part guide to Cochrane plain-language summaries: where to find the question, the certainty rating, the search date, and the honest bottom line.
Evidence and research methods
How to Read a Cardiology Guideline: Class of Recommendation and Level of Evidence
How to read ACC/AHA cardiology guidelines: what Class of Recommendation and Level of Evidence mean, and why the two labels are rated separately.
Evidence and research methods
How to Read a Cancer Survival Statistic
Five-year, relative, and net survival explained, plus why lead-time and length-time bias make survival a weak test of whether screening saves lives.
Evidence and research methods
How the USPSTF Built Its First Adult Anxiety Screening Recommendation
How the USPSTF reached its 2023 Grade B for anxiety screening, why it stopped short for adults 65 and older, and how to read a split verdict.