It's time to review the domain Indirectness
As before GRADE starts at High and rates down for indirectness in the evidence.
Direct evidence is considered to come from studies of populations (P), interventions (I), comparisons (C), and outcomes (O) that match the clinical question you are seeking to answer with evidence. The answered PICO must match the asked PICO.
Populations must match
If you are interested in children, the studies must be done in children. If you are interested in outpatients, studies must be done in ambulatory, not hospital settings.
Interventions must match
If you are interested in a pharmaceutical treatment, for example, all the characteristics of treating with that therapy must match your question, things like the dosing, route of delivery, etc.
Comparisons must match
If you are interested in how a new intervention fares against the current best intervention, studies using placebo don't meet your criterial for directness.
Seems pretty straightforward, right?
Outcomes must match
Here's where it can get a little trickier for panelists. Especially clinical panelists who are used to accepting surrogate outcomes as clinical indicators. These are things like blood pressure, lipid levels, HbA1C. These surrogate outcomes may be of interest as clinical intermediaries to outcomes important to patients like strokes and heart attacks. But they are not themselves the outcomes important to patients.
For GRADing evidence, in addition to requiring that outcomes must match your pre-defined PICO outcomes, direct outcomes must be patient centered, important clinical outcomes. They are things that matter to patients like having a stoke or heart attack or spending time in a hospital or long-term care or premature death. The intermediate, surrogate, or proxy outcomes are rated down for indirectness.
There's another slightly tricky concept for indirectness we must consider.
Remember I mentioned studies that compare the intervention of interest to a placebo when you are interested in comparison to the current best intervention? When your question is about the comparative effectiveness of an intervention compared to another intervention and you only find studies comparing each intervention to placebo, you would have to make indirect comparisons. Since you can't make head to head comparisons of intervention A to intervention B, you would have to compare intervention A to intervention C (in this case a placebo) and intervention B to intervention C. When the relative effects of the two interventions of interest have to be compared through a third comparator, the evidence would be rated down for indirectness.
So GRADE rates down for 4 types of indirectness:
- The population is not the one specified. For example, attempting to infer the effect on children when studies are conducted in adults.
- The intervention is not the one specified. This may be if the way it is delivered or the setting in which it is delivered differs from the clinical question
- The outcome is not the one specified and may be an intermediate or surrogate outcome.
- The comparison is not what you specified and may not be able to be evaluated head to head with your intervention.
TheEvidenceDoc November 17, 2017
And as always, for more detail on this topic or to work ahead, check out the GRADE HANDBOOK