01State which proposition is uncertain

The phrase evidence is limited could mean few studies, missing methods, disagreement between results or no direct measurement of the claimed outcome. These are different problems and can lead to different conclusions.

Cochrane’s GRADE guidance assesses certainty for a body of evidence about a particular outcome. It considers risk of bias, inconsistency, indirectness, imprecision and publication bias, with documented reasons for the assessment.Cochrane Handbook — Chapter 14: Summary of findings and certainty of evidence (opens in a new tab)

That structure is a reminder to locate the uncertainty, not a shortcut for assigning a score. This guide does not perform a GRADE assessment of any peptide or designate a body of evidence as high or low certainty.

For an original hypothetical study, the observed result is an increase in a reporter signal. The uncertain proposition might be whether the signal reflects a functional change, rather than whether the instrument recorded a higher number.

02Replace vague hedges with specific explanations

Each revised sentence still needs the relevant source and context. The wording does not become accurate merely because it sounds restrained. For example, no eligible study was found requires an actual search with a stated scope.

When the result is imprecise, describe the range of effects compatible with the reported interval where that is possible. An estimate can be uncertain about magnitude even when all values in its interval point in the same direction.

When the problem is indirectness, state the mismatch. Evidence about a cell readout may be direct for that readout while remaining indirect for a human functional claim.

03Keep formal ratings and ordinary prose distinct

If you quote a review’s formal certainty rating, preserve the outcome, comparison and reason attached to it. Do not transfer a rating from one outcome to every claim about the compound.

A paper being peer reviewed, frequently cited or recently published does not itself supply a certainty assessment. Nor does a narrow confidence interval account for every potential source of bias.

In the hypothetical reporter study, an honest summary can say that functional relevance remains untested in the reported experiment. It does not need to invent a percentage probability that the broader claim is true.

04Make positive and negative statements equally precise

A study that did not detect a difference does not necessarily establish the absence of every meaningful effect. Conversely, uncertainty should not be used to imply that a benefit is likely simply because it has not been ruled out.

Separate the finding from the next research question. The finding might concern one signal in one model; the next question might be whether it corresponds to a functional effect. Presenting that question as future work keeps it from becoming an implied result.

Avoid repeating a generic more research is needed sentence after every paragraph. Explain what information would resolve the particular uncertainty, such as a direct outcome measurement, clearer material identity or a more informative comparison.

Finally, check the title and takeaway against the body. A carefully qualified discussion cannot repair an overconfident headline that most readers encounter first.