Physicians can use incidental findings on chest CTs to identify patients at risk of cardiovascular disease. Researchers reported online May 27 in Radiology that a validated, imaging-based risk predictor had good discriminative ability and calibration for stratifying at-risk patients.
Pushpa M. Jairam, MD, PhD, of University Medical Center Utrecht in Utrecht, the Netherlands, and colleagues, developed and validated a predictive model based on data from the PROVIDI (Prognostic Value of unrequested Information in Diagnostic Imaging) study. PROVIDI was a retrospective, case cohort study that enrolled 23,443 patients from hospitals in the Netherlands who underwent chest CT scans between 2002 and 2005.
For the ancillary study, Jairam and colleagues selected two different cohorts for derivation and validation analyses. The derivation group included 10,410 patients from seven hospitals, with a subgroup featuring a 10 percent random sample for scoring for survival analyses. The validation group had 1,653 patients from one hospital with a subgroup of 211 randomly selected patients.
Radiologists had performed chest CTs using multidetector CT scanners on all patients. They graded scans for calcification in the coronary arteries, the thoracic aorta and cardiac valves. They also determined cardiac and thoracic diameters and ascending and descending thoracic aorta diameters.
They used the Dutch National Registry of Hospital Discharge Diagnoses and the National Death Registry to assess cardiovascular events, which were defined as all coronary events, angina pectoris, cerebrovascular events, peripheral artery disease and heart failure. In the external validation part of the study, they evaluated whether the prediction tool could stratify patients as low, intermediate or high risk (less than 10 percent, 10 percent to less than 20 percent and more than 20 percent, respectively).
At a follow-up of 3.7 years, they found 1,148 mean annualized cardiovascular events for 29.3 events per 1,000 person years. They determined that patient age and sex, CT indication, left anterior descending coronary artery calcifications, mitral valve calcifications, descending aortic calcifications and cardiac diameter offered the best fitting predictive value.
The C index was 0.71 in the validation cohort, which they pointed out was comparable to the Framingham score. The model also provided good stratification ability. In the final model they found similar rates by risk: 10 percent vs. the observed 10-year risk of 9 percent for low risk; 37 percent vs. 12 percent for intermediate risk; and 52 percent vs. 50 percent for high risk.
“[T]he use of markers of subclinical target organ damage for CVD [cardiovascular disease] risk prediction, like cardiovascular calcifications incorporated in the imaging based model, provide a novel strategy and adequate estimation for CVD risk, irrespective of the conventional risk factor status,” they wrote.
Jairam et al observed that use of CT chest scans is growing, making implementation of the prediction model as a screen feasible. They also noted that the event rate of 29.3 events per 1,000 person years was more than double what is reported for the general population, which showed that there is a proportion of high-risk patients who might benefit from identification and possible intervention.
On the downside, it was not clear that early drug treatment would lead to improved outcomes or savings in healthcare costs. The risk predictor was limited to a white, Dutch patient population, they cautioned, and needed to be validated in other patient populations.