Evidence Review
What the trials and guidelines actually show
Evidence Review is for a question about practice rather than about one patient. It frames the question in PICO terms, sets out what current guidelines recommend and where they part company, and then works through systematic reviews and pivotal trials, giving effect sizes in absolute as well as relative terms. Each main outcome receives a certainty rating with its reasons, and the report opens with a short bottom line for practice.
Who it is for
- Consultants and registrars preparing a journal club, a protocol revision or a departmental decision
- Physicians asked by a colleague or a patient whether a newer treatment justifies a change in practice
- Residents learning critical appraisal who want a worked example of how evidence is weighed
What it does
The question made answerable
Recasts the question in PICO form and warns when it is too wide for a useful answer, so that the review deals with one population, one comparison and the outcomes patients care about.
Guidelines side by side
Sets out what each major guideline recommends, who issued it and when, how strongly it is worded, where bodies differ, and whether newer trials have been published since.
Effects in absolute terms
For each important study gives the design, who was enrolled, what was compared and the primary outcome, the relative effect and its confidence interval, and what that means in absolute terms as a risk difference and NNT or NNH.
Each study appraised
Looks at how patients were allocated and blinded, losses to follow-up, outcome switching and early termination, then at precision, agreement between studies, fit to the question, surrogate outcomes, and who paid for the work.
Certainty by outcome
Grades confidence in each key outcome on the four GRADE levels, from high to very low, and explains every step down so the reader can challenge it.
Gaps and ongoing trials
States who was excluded from the evidence and what remains unknown, and lists registered trials from ClinicalTrials.gov that are under way, so the reader knows whether the answer may soon change.
How it works
- Pose the question with its population, intervention, comparator and outcomes, and say which patient group and setting matter to you.
- Guidelines, systematic reviews, meta-analyses and randomised trials are searched first, together with registered trials on ClinicalTrials.gov.
- The engine summarises the guidelines, then works down from pooled analyses to the trials that shaped practice, turning to observational data only where randomised evidence is missing.
- Each key study is appraised and each main outcome receives a certainty rating with reasons.
- The report leads with the bottom line, followed by the evidence table, applicability and harms, gaps, and the numbered, verified references.
What to include
- The question as population, intervention, comparator and outcome
- The setting and the patient group you care about
- What you already know or have read
Example cases
- In adults over 80 with hypertension, does intensive blood pressure lowering reduce cardiovascular events compared with standard targets?
- SGLT2 inhibitors in heart failure with preserved ejection fraction: effect on hospitalisation and death?
- Is colchicine effective for secondary prevention after myocardial infarction?
What the report contains
- Bottom Line
- The Question
- What the Guidelines Say
- The Evidence
- Certainty of the Evidence
- Applicability and Harms
- Gaps and Ongoing Trials
- Key Evidence & References
Limits
- It summarises published evidence from the records it retrieves and does not re-analyse trial data. The full papers should be read before a protocol or practice is changed.
- When an exact figure cannot be confirmed from the retrieved records, the report gives the direction and rough size of the effect in words and asks the reader to confirm the number in the paper.
- Its certainty ratings follow the GRADE approach but are not a formal assessment by a guideline panel, and national guidelines and local formularies still determine practice.
Frequently asked questions
How do you calculate the number needed to treat from a trial?
Subtract the event rate in the treatment group from the rate in the control group to get the absolute risk reduction, then take its reciprocal. If events fall from 10% to 8%, the absolute reduction is 2 percentage points and the number needed to treat is 50. Always state the time frame, because a number needed to treat over five years is not comparable with one over six months.
What do the GRADE certainty levels mean?
High certainty means further research is very unlikely to change confidence in the estimate of effect; moderate means it may; low means it is likely to; very low means any estimate is very uncertain. Randomised evidence starts high and is downgraded for risk of bias, imprecision, inconsistency, indirectness or publication bias. Observational evidence starts low and can occasionally be upgraded.
Why can a relative risk reduction be misleading?
The same relative reduction produces very different absolute benefits depending on baseline risk. A 25% relative reduction prevents one event in four hundred patients whose baseline risk is 1%, but one in twenty when it is 20%. Absolute differences and numbers needed to treat show the size of the benefit for a particular patient.
Does the engine include trials that have not yet reported?
Yes. Registered trials from ClinicalTrials.gov are retrieved with the published literature, and the report lists relevant ongoing trials under its gaps section, so that a reader can see whether a pending result might change the answer.
Related articles
Clinical article
Reading a trial in ten minutes: absolute risk, number needed to treat and certainty
A ten-minute method for appraising a randomised trial: the question, the outcome, absolute risk and NNT, bias, subgroups, early stopping and certainty.
7 min read
Guide
From PubMed to the bedside: how guidelines, reviews and trials are prioritised
The evidence hierarchy and GRADE in practical terms for clinicians, and how Sana Health orders the guidelines, reviews and trials listed under each answer.
7 min read
Guide
How verified references work: why the links under an answer cannot be invented
How Sana Health builds the reference list under each answer from PubMed, ClinicalTrials.gov, DailyMed and the NCBI Bookshelf, and what verification means.
5 min read
Guide
AI clinical decision support: what it can and cannot do for a hospital doctor
A balanced look at AI clinical decision support for hospital doctors: what the evidence shows, where large language models fail, and how to use them safely.
7 min read
Related engines
A focused answer in seconds
ACS, heart failure, arrhythmia and ECGs
Clinical Scores and Calculators
The right score, calculated and interpreted