How to Read a Cardiology Guideline: Class of Recommendation and Level of Evidence

Every recommendation in a modern American College of Cardiology and American Heart Association (ACC/AHA) guideline carries two separate labels. The Class of Recommendation (COR) tells you how strongly the writing committee wants you to act, from Class I ("should") down through Class III ("do not"). The Level of Evidence (LOE) tells you how confident anyone can be in the data behind that advice, from LOE A (multiple high-quality randomized trials) down to LOE C-EO (expert opinion). The two are rated independently, and that is the point most readers miss: a strong recommendation is not the same as strong evidence, and reading the labels together lets you tell a randomized-trial mandate from a committee's best guess.

What the Class of Recommendation actually says

COR is a statement about the size and certainty of net benefit, phrased as an instruction. The 2015 ACC/AHA framework, published in Circulation in 2016 as "Further Evolution of the ACC/AHA Clinical Practice Guideline Recommendation Classification System," fixed standard language for each class so the wording itself signals strength.

Class I is strong: the benefit greatly outweighs risk, and the guideline uses verbs like is recommended, is indicated, or should be performed.

Class IIa is moderate: benefit still outweighs risk, but less decisively. The signature phrase is is reasonable or can be useful.

Class IIb is weak: benefit is only marginally greater than or equal to risk. The language softens to may be reasonable or may be considered.

Class III comes in two flavors that the framework deliberately separates. Class III: No Benefit means the intervention does not help (is not recommended, is not indicated). Class III: Harm means it is expected to injure (should not be performed, is potentially harmful, associated with excess morbidity or mortality). Collapsing those two into a single "don't" loses real information: one says "pointless," the other says "dangerous."

The verbs are not decoration. When a committee chooses may be reasonable over is recommended, that gradient is the recommendation.

What the Level of Evidence actually says

LOE describes the quality, quantity, and consistency of the data, independent of how strong the recommendation is. The current system uses five tiers.

LOE A rests on high-quality evidence from more than one randomized controlled trial, or meta-analyses of such trials. LOE B-R is moderate-quality evidence from one or more randomized trials (the R means randomized). LOE B-NR is moderate-quality evidence from well-designed nonrandomized, observational, or registry studies (NR for nonrandomized). LOE C-LD means limited data: small studies, registries with methodological limits, or physiologic reasoning. LOE C-EO is expert opinion, the consensus of the writing group when direct evidence is thin.

The split of the old "Level B" and "Level C" into randomized versus nonrandomized components was the main 2015 change. It lets a reader see, at a glance, whether "moderate evidence" came from a randomized design or from observation, a distinction that changes how much a single study should move your confidence.

Why COR and LOE are assigned independently

Here is the part that trips people up. A high Level of Evidence does not force a high Class of Recommendation, and a low Level of Evidence does not forbid a strong one.

You can have a Class I recommendation supported only by LOE C-EO. A classic example is a practice so obviously beneficial, or so ethically impossible to randomize, that no trial exists or ever will, yet withholding it would be indefensible. The committee's confidence in the action is high even though the evidence tier is low. Conversely, a large randomized trial (LOE A) can produce only a Class IIb recommendation if the measured benefit is small, uncertain in a given subgroup, or offset by cost or harm. Strong data, weak mandate.

That independence is not a flaw in the system; it is the point. COR integrates evidence plus judgment about benefit, risk, feasibility, and values. LOE reports only the data's reach. Reading them as one number throws away the distinction between "we are sure this helps a lot" and "we are sure the trial was well done."

How much of cardiology rests on strong evidence

This matters because the strongest evidence is scarcer than most readers assume. In a 2019 analysis in JAMA, Fanaroff and colleagues reviewed 26 current ACC/AHA guidelines and found that only 8.5 percent of recommendations were supported by LOE A, while 50 percent were LOE B and 41.5 percent were LOE C. Even among the strongest Class I recommendations, only about 14 percent carried LOE A, meaning the large majority of "should do" statements were not backed by multiple randomized trials. The proportion of LOE A recommendations had not meaningfully risen compared with a decade earlier.

None of that means guidelines are unreliable. It means the labels are doing exactly their job: telling you which recommendations are anchored in randomized data and which represent careful consensus filling an evidence gap. A Class I, LOE A recommendation and a Class I, LOE C-EO recommendation ask the same action of you, but they invite very different levels of scrutiny when a new trial lands or an individual patient does not fit the mold.

Reading a recommendation in practice

Put the two labels together. Start with the COR to learn what the committee wants done and how forcefully. Then read the LOE to learn how solid the ground is. A Class IIb, LOE C-LD line is a tentative suggestion built on thin data, reasonable to weigh against patient preference and to revisit as evidence matures. A Class I, LOE A line is about as close to settled as cardiology gets. When guidelines are updated, watch for a recommendation's LOE climbing as trials report, or its COR shifting as the benefit-risk picture changes; the movement of these labels over successive editions is often the real story.

This is an educational overview of how evidence is graded, not medical advice; decisions about any specific patient belong to that patient and their treating clinicians.