Reading The Evidence

Start at the source

Inspect the source record behind a displayed fact. Check its date, study identity, and context. A citation establishes where a statement came from; it does not make every possible interpretation of that statement valid.

Atlas surfaces comparator evidence. It does not recommend treatment, diagnose disease, declare benefit-risk, or establish that one intervention is more effective than another.

Preserve comparison context

Before comparing values, check the population, arm or regimen, endpoint, denominator, timepoint, and data cutoff where those fields apply. Similar labels can refer to materially different measurements.

Protocol declarations describe planned work. Observed results describe reported findings. Keep them distinct when both are present.

Inspect an evidence packet

An evidence packet contains the prepared evidence associated with selected records. Review its source links, coverage, missing fields, conflicts, and any quarantined or unavailable material.

A partial packet may contain useful accepted evidence. Its gaps remain part of the result and should travel with any summary you make.

Interpret missingness correctly

• Missing: a requested field was not available in the displayed evidence.
• Conflicting: sources or records disagree in a way that remains unresolved.
• Quarantined: material was excluded from accepted evidence because it did not pass the applicable checks.
• Not searched: the product did not complete that source search for this question.

None of these states means that a biological effect, safety issue, program, or result does not exist.

Read model-assisted research

Use the research answer as a starting point for inspecting its cited evidence. Check that each material statement is supported by the displayed sources and applies to the same comparison context.

The current research action synthesizes selected eligible evidence. It does not independently browse for more sources or establish exhaustive coverage.