Coronary CTA stays robust despite missing data

April 26, 2006

Coronary CT angiography has clocked impressive numbers to detect or rule out coronary artery stenosis. But do these numbers hold up when uninterpretable data are factored into the results? The answer is yes and no, according to a study presented at the annual meeting this month of the Society of Computed Body Tomography and Magnetic Resonance.

Coronary CT angiography has clocked impressive numbers to detect or rule out coronary artery stenosis. But do these numbers hold up when uninterpretable data are factored into the results? The answer is yes and no, according to a study presented at the annual meeting this month of the Society of Computed Body Tomography and Magnetic Resonance.

On a segment-by-segment basis, coronary CTA lost sensitivity when uninterpretable segments were factored into the analysis. On a per-patient basis, however, the resulting specificity did not suffer when missing or unintrepretable data were included.

"Coronary CTA is more robust at the patient level, suggesting ultimate clinical utility despite study biases," said lead author Dr. Ruth C. Carlos, a radiologist at the University of Michigan Medical Center in Ann Arbor.

Carlos and colleagues reviewed the literature from 1991 to 2005 to evaluate the effect of missing data on sensitivity and specificity in coronary CTA studies.

They found 30 English-language peer-reviewed articles describing separate prospective observational studies. All patients underwent both CTA and conventional angiography. Four studies used 64-slice scanners, and the rest split nearly evenly between four- and 16-slice machines.

Researchers treated uninterpretable/missing data as false negatives to estimate the lower bounds of sensitivity and as false positives to estimate the lower bounds of specificity.

The published segment-level sensitivity and specificity for the various scanners were, respectively:

  • 84% and 95%, four-slice

  • 83% and 95%, 16-slice

  • 88% and 98%, 64-slice

After missing/uninterpretable segments were included, the sensitivity and specificity markedly decreased for all scanners:

  • 35% and 67%, four-slice

  • 37% and 76%, 16-slice

  • 54% and 89%, 64-slice

Patient-level data tell a different story. Sensitivity and specificity for detection of at least one-vessel stenosis were:

  • 81% and 92%, four-slice

  • 92% and 85%, 16-slice

  • 97% and 92%, 64-slice

After missing/uninterpretable cases were included, sensitivity and specificity changed minimally:

  • 81% and 92%, four-slice

  • 91% and 82%, 16-slice

  • 94% and 86%, 64-slice

At the patient analysis level, selection criteria used have resulted in a high prevalence of coronary artery disease. Coronary CTA trends toward better sensitivity than specificity, with imaging performance only minimally altered by missing/uninterpretable results bias, Carlos said.

For more information from the Diagnostic Imaging archives:

64-slice CT enters clinical practice realm

Coronary MRA struggles against success of CTA

Coronary CTA takes giant leap with 64-slice scanners

64-slice scanners build case for coronary CTA