Two axes, not one ranking
A large estimated benefit can have low certainty. A small estimated benefit can have high certainty. Conflating size with certainty makes both findings harder to understand.
Imagine a fictional review of a support program. Several studies measure short-term satisfaction well, but few follow participants long enough to assess daily function. The review can reasonably have different certainty ratings for those outcomes. Giving the whole program one label would erase that distinction.
What a rating should explain
Cochrane's GRADE guidance considers risk of bias, inconsistency, imprecision, indirectness, and publication bias when judging certainty by outcome. A useful review explains which concerns changed the rating, rather than supplying a label alone.
For example, uncertainty may arise because the studies are small, because their results differ, or because they studied a population unlike the one in question. Those limitations suggest different research needs.
A decision still needs a context
Suppose two fictional options have similar uncertain benefits, but one is inexpensive and easy to stop while the other carries substantial burdens. The same evidence rating does not make the choices equivalent. Consequences, alternatives, and preferences still matter.
This is a decision example, not a recommendation for a treatment. A certainty rating cannot choose on someone's behalf, and a recommendation may incorporate considerations beyond that rating.
Reading the sentence after the label
“Low certainty because follow-up was short” tells you something different from “low certainty because the comparison groups differed at baseline.” Looking for that explanation turns a vague impression of weak evidence into a specific account of what remains unresolved. It also makes later updates easier to interpret: did the new research address the limitation that mattered?
