‘Risk prediction models can be used for targeted screening for asymptomatic carotid stenosis,’ researchers find

247

Most prediction models can “reliably” identify individuals at high risk of significant asymptomatic carotid stenosis (ACS), a new study shows.

Michael H. F. Poorthuis, MD, of the Clinical Trial Service Unit and Epidemiological Studies Unit in the Nuffield Department of Population Health at the University of Oxford, U.K., et al made the discovery in a systematic review and data validation that drew on more than half a million individuals who attended commercial vascular screening clinics in both the U.K. and U.S.

Furthermore, the best‐performing prediction models identified over one third of all cases by targeted screening of individuals in the highest decile of risk only, the researchers found.

The findings were published in the Journal of the American Heart Association.

In introducing their rationale, the authors noted that ischemic stroke is the first presentation of cardiovascular disease in about 25% of cases, with 15–20% of ischemic stroke cases are associated with extracranial carotid artery stenosis. The prevalence of moderate (≥50%) and severe (≥70%) ACS in the general population has been estimated to be 2% and 0.5%, respectively, they pointed out.

“Because of this low overall prevalence, population‐level screening for ACS with duplex ultrasound is not recommended in current guidelines,” Poorthuis et al wrote. “However, targeted screening of high‐risk individuals might be worthwhile, and risk stratification tools or prediction models have been developed to provide individualized risk estimation for ACS. Before recommending targeted screening, risk prediction tools should be assessed for discrimination, calibration and likely ability to detect false‐positive and false‐negative cases in an independent external population.”

The systematic review included studies that addressed development and/or validation of diagnostic prediction models to detect ACS of 50% or greater and assessed prediction models in both general and high‐risk populations but not in diseased populations at higher risk of ACS, the authors added.

Some 975 studies were reviewed, with sic prediction models identified—three each for moderate and severe ACS.

Validation utilized data from 596,469 individuals. Discrimination and calibration were assessed.

“In the validation cohort, 11,178 (1.87%) participants had ≥50% ACS and 2,033 (0.34%) had ≥70% ACS,” the investigators continued. “The best model included age, sex, smoking, hypertension, hypercholesterolemia, diabetes mellitus, vascular and cerebrovascular disease, measured blood pressure, and blood lipids. The area under the receiver operating characteristic curve for this model was 0.75 (95% CI, 0.74–0.75) for ≥50% ACS and 0.78 (95% CI, 0.77–0.79) for ≥70% ACS. The prevalence of ≥50% ACS in the highest decile of risk was 6.51%, and 1.42% for ≥70% ACS. Targeted screening of the 10% highest risk identified 35% of cases with ≥50% ACS and 42% of cases with ≥70% ACS.”

Elaborating on their findings, the authors concluded “the present study showed that most prediction models had modest discrimination but could reliably identify a cohort of cases at high risk of ACS.”

Further, they wrote: “The prevalence of ACS in the decile(s) at highest predicted risk of ACS was considerably higher than the overall prevalence, thereby substantially reducing the number of individuals needed to screen to detect ACS. Further research should determine the optimum thresholds required for a targeted screening by considering the number needed to screen, the diagnostic yield, the absolute reduction of stroke risk by prophylactic treatment, and cost‐effectiveness of different approaches.”

SOURCE: doi.org/10.1161/JAHA.119.014766

LEAVE A REPLY

Please enter your comment!
Please enter your name here