EDUCATION / UNDERSTANDING EVIDENCE
Make sense of the evidence.
Learn the terms you will meet in the source library, then practice reading three real examples. Start with the question a study asked, the comparison it made, and the uncertainty in its result.
Keep three questions nearby
- Who was studied?Read the population, sample, and follow-up.
- Compared with what?Check what each group received.
- What does the result establish?Look at the outcome, uncertainty, and limits.
PUT THE TERMS TO WORK
Three examples from the collection
Each example uses the same facts as its source page. Open the original paper for the full methods and results.
Manoogian et al. · 2024
Read the count, comparison, and units together
- Participants
- 122 randomized; 108 completedExplain randomized trial
- Comparison
- A personalized 8–10-hour eating window plus usual care versus usual care and nutrition counseling alone.Explain comparison group
- Reported finding
- HbA1c improved by an additional 0.10 percentage points on average (95% confidence interval: −0.19 to −0.003). The effect was small.Explain percentage points
What would you check before turning this result into a broad health claim?
Check who entered the trial, which participants the analysis includes, what both groups received, and the estimate with its interval. Keep the measured outcome distinct from future health outcomes.
Dote-Montero et al. · 2025
A headline and a primary outcome can emphasize different things
- Population
- Adults with overweight or obesity in Spain
- Reported finding
- The main outcome, MRI-measured visceral fat, did not clearly differ between groups. Weight decreased more in the eating-window groups as a secondary outcome.Explain primary and secondary outcomes
Which outcome should you identify before interpreting the weight-loss headline?
Identify the trial’s main outcome first, then read additional outcomes in that context. Keep each finding attached to the outcome actually measured.
Chen et al. · 2026
Check the evidence date behind a new publication
- Published
- January 21, 2026
- Evidence search through
- January 3, 2023Explain evidence-search cutoff
- Included evidence
- 41 randomized trials; 2,287 participantsExplain systematic review
Does the publication year tell you which later trials were included?
Use the documented search cutoff to understand the review’s coverage. A newer publication date does not automatically mean newer studies were searched for or incorporated.
A REFERENCE WHILE YOU READ
Research glossary
Definitions, a reading tip, and a source from the collection where the term is useful.
Randomized trial
Researchers use chance to assign participants to intervention groups. This helps make the comparison fairer by reducing systematic differences in who receives each intervention.
When reading a studyRandom assignment describes the design. Still check who was studied, how the trial was conducted, and which participants were included in the reported results.
Definition sources: National Cancer Institute dictionary
See it in context: TIMET: a modest glucose change with usual care
Comparison group
The group against which an intervention is assessed. It may receive usual care, another intervention, a placebo, or no intervention.
When reading a studyRead what both groups actually received. A trial comparing two supported eating plans asks a different question from a trial comparing a plan with unchanged habits.
Definition sources: National Cancer Institute dictionary
See it in context: Does an eating window add to calorie restriction?
Crossover study
Participants receive different interventions in successive periods, with the sequence assigned by the study. In a randomized crossover trial, chance determines that sequence.
When reading a studyThe same people contribute to more than one condition. Do not add the condition counts as though they were different participants; check the order and interval between conditions.
Definition sources: National Cancer Institute dictionary
See it in context: Early eating in a small, controlled experiment
Observational study
Researchers observe exposures, characteristics, or outcomes without assigning the exposure or intervention being studied.
When reading a studyPeople with different eating patterns can differ in other ways. An observed association alone does not establish that the eating pattern caused the outcome.
Definition sources: National Cancer Institute dictionary: Observational study; Cochrane: Handbook, chapter 25: Assessing risk of bias in a non-randomized study
See it in context: Explore public dietary data—and its limits
Primary and secondary outcomes
A primary outcome is a study’s designated main result, selected before the study begins. Secondary outcomes are additional measured results.
When reading a studyIdentify the main question before interpreting a headline. A favorable secondary finding does not change what happened to the primary outcome.
Definition sources: National Cancer Institute dictionary: Primary endpoint; National Cancer Institute dictionary: Secondary endpoint
See it in context: Early, late, or self-selected: the primary outcome matters
Systematic review
A review that uses explicit, planned methods to find and assess studies addressing a defined question, with stated rules for which studies qualify.
When reading a studyCheck the question, search dates, included studies, and limitations. A review’s title or size does not by itself establish the certainty of its conclusions.
Definition sources: Cochrane
See it in context: Cochrane: little clear advantage over dietary advice
Meta-analysis and network meta-analysis
Meta-analysis statistically combines results from separate studies. Network meta-analysis compares several interventions using both direct comparisons and indirect evidence through shared comparators.
When reading a studyCombining studies does not remove their limitations. Indirect comparisons depend on whether the different study groups and settings are sufficiently comparable.
Definition sources: Cochrane: Handbook, chapter 10: Analysing data and undertaking meta-analyses; Cochrane: Handbook, chapter 11: Undertaking network meta-analyses
See it in context: Earlier windows look promising; shorter is not always better
Confidence interval
A range expressing statistical uncertainty around an estimate under the analysis assumptions. Read it alongside the estimated difference and its units.
When reading a studyFor an absolute difference, zero is the no-difference value; for a ratio, it is one. An interval spanning that value does not prove equality. Precision is also different from freedom from bias.
Definition sources: Cochrane
See it in context: Glucose evidence in prediabetes and type 2 diabetes
HbA1c (A1C)
A blood test reflecting average blood glucose over roughly the previous three months. It measures the proportion of hemoglobin with attached glucose and is reported as a percentage.
When reading a studyRead the study’s population and the size of the change. Interpreting an individual result requires clinical context; the timer does not measure HbA1c.
Definition sources: National Institute of Diabetes and Digestive and Kidney Diseases
See it in context: TIMET: a modest glucose change with usual care
Percentage points
The absolute difference between two values expressed as percentages. Relative percentage change uses the starting value as its denominator.
When reading a studyIllustrative arithmetic: 6.0% to 5.9% is a fall of 0.1 percentage points. These are teaching values, not participant results from a study.
Definition sources: European Union data visualisation guide
See it in context: TIMET: a modest glucose change with usual care
Biomarker and health outcome
A biomarker is a measurement indicating a biological process or response, such as a laboratory value. It differs from directly measuring symptoms, daily functioning, or survival.
When reading a studyUsing a biomarker as a substitute for a clinical outcome requires validation for that context. A short-term marker change should not automatically be described as disease prevention or longer life.
Definition sources: U.S. Food and Drug Administration: About Biomarkers and Qualification; FDA–NIH BEST Resource: Validated Surrogate Endpoint
See it in context: Early eating in a small, controlled experiment
Evidence-search cutoff
The date through which a review searched for the evidence covered by its analysis. It can be earlier than the date the review was published.
When reading a studyLook for whether later searches were fully incorporated. A publication date alone does not tell you how recently the evidence was gathered.
Definition sources: Cochrane
See it in context: Earlier windows look promising; shorter is not always better
CHECK THE DEFINITIONS
Original methods and education sources
These sources explain research methods and measurements. They are separate from the fasting studies in the collection.
- National Cancer Institute dictionaryRandomized clinical trial
- National Cancer Institute dictionaryControlled clinical trial
- National Cancer Institute dictionaryCrossover study
- National Cancer Institute dictionaryObservational study
- National Cancer Institute dictionaryPrimary endpoint
- National Cancer Institute dictionarySecondary endpoint
- CochraneHandbook, chapter 1: Starting a review
- CochraneHandbook, chapter 10: Analysing data and undertaking meta-analyses
- CochraneHandbook, chapter 11: Undertaking network meta-analyses
- CochraneHandbook, chapter 15: Interpreting results and drawing conclusions
- CochraneHandbook, chapter 25: Assessing risk of bias in a non-randomized study
- National Institute of Diabetes and Digestive and Kidney DiseasesThe A1C Test & Diabetes
- European Union data visualisation guidePercentages versus percentage points
- U.S. Food and Drug AdministrationAbout Biomarkers and Qualification
- FDA–NIH BEST ResourceValidated Surrogate Endpoint
- CochraneReporting search dates
These explanations help with reading research. They do not interpret your medical results or recommend a treatment. For the collection’s scope and source checks, read our editorial method.