Design: retrospective diagnostic accuracy study with multivariable logistic regression, calibration and decision curve analysis
Population: patients with carotid atherosclerosis imaged with both CT angiography and high-resolution vessel wall MRI, January 2021 to December 2024
Primary outcome: discrimination of symptomatic from asymptomatic carotid plaque
Effect: integrated model area under the curve 0.876 (95% CI 0.828–0.924) overall and 0.919 (0.867–0.971) in the moderate-stenosis subgroup, with greatest net benefit on decision curve analysis

The carotid with 50 to 69% stenosis is the one that defeats a stenosis-based report: the number does not decide anything, and the patient's symptoms usually have. Plaque-RADS was designed to add plaque morphology to that judgement. A retrospective study asked whether its ancillary features add anything on top, in patients who had both computed tomography angiography and high-resolution vessel wall magnetic resonance imaging between 2021 and 2024.

On multivariable logistic regression, degree of stenosis, Plaque-RADS category, plaque burden, remodelling index, enhancement ratio and mean perivascular fat density were each independently associated with symptomatic plaque. An integrated model combining all of them discriminated symptomatic from asymptomatic plaque with an area under the curve of 0.876 (95% confidence interval 0.828 to 0.924) across the whole cohort. In the moderate-stenosis subgroup — the group where the question is actually difficult — it reached 0.919 (0.867 to 0.971). Calibration was good and decision curve analysis showed the ancillary-feature models gave the greatest net benefit.

The reportable consequence is concrete. Perivascular fat density, remodelling index, enhancement ratio and plaque burden are measurable on studies departments already acquire, and reporting them turns the moderate-stenosis carotid from a number into a risk statement the stroke team can use. The caution is that this is retrospective, single-cohort and internally validated, so the discrimination figures are almost certainly optimistic — report the features, and let the clinical team weigh them, rather than issuing the model's probability as though it were validated.

Report plaque burden, remodelling index, enhancement ratio and perivascular fat density alongside the Plaque-RADS category.
Do this particularly in the 50–69% stenosis group, where stenosis alone decides nothing.
Do not issue a computed risk probability from this model — it has not been externally validated.
Standardise how perivascular fat density is measured locally before reporting it, or the numbers will not be comparable between reporters.
Say explicitly in the report which features are present and which were not assessable on the sequences available.

Why it matters

It gives the reporting radiologist something to say about the carotid where the stenosis measurement is least informative and the referral decision is hardest.

The statistics, in plain English

An area under the curve of 0.919 in the moderate-stenosis subgroup is high, but it was derived and tested in the same retrospective cohort, so it is an optimistic estimate rather than a measured performance — models built this way typically lose discrimination on external data. The subgroup result is also based on fewer patients than the overall figure, which is why its interval (0.867–0.971) is wider despite the higher point estimate. Decision curve analysis showing net benefit is reassuring about the direction but inherits the same optimism. The independence of the six predictors comes from a single multivariable model, not from replication.