Diagnostic Performance of Noninvasive Fractional Flow Reserve Derived From Coronary Computed Tomography Angiography in Suspected Coronary Artery Disease: The NXT Trial (Analysis of Coronary Blood Flow Using CT Angiography: Next Steps)

OBJECTIVES: The potential for FFRCT to noninvasively identify ischemia in patients with suspected CAD has not been sufficiently investigated. The goal of this study was to determine the diagnostic performance of noninvasive fractional flow reserve (FFR) derived from standard acquired coronary computed tomography angiography (CTA) datasets (FFRCT) for the diagnosis of myocardial ischemia in patients with suspected stable coronary artery disease (CAD).

METHODS: This prospective multicenter trial included 254 patients scheduled to undergo clinically indicated ICA for suspected CAD. Coronary CTA was performed before ICA. Evaluation of stenosis (>50% lumen reduction) in coronary CTA was performed by local investigators and in ICA by an independent core laboratory. FFRCT was calculated and interpreted in a blinded fashion by an independent core laboratory. Results were compared with invasively measured FFR, with ischemia defined as FFRCT or FFR ≤0.80.

RESULTS: The area under the receiver-operating characteristic curve for FFRCT was 0.90 (95% confidence interval [CI]: 0.87 to 0.94) versus 0.81 (95% CI: 0.76 to 0.87) for coronary CTA (p = 0.0008). Per-patient sensitivity and specificity (95% CI) to identify myocardial ischemia were 86% (95% CI: 77% to 92%) and 79% (95% CI: 72% to 84%) for FFRCT versus 94% (86 to 97) and 34% (95% CI: 27% to 41%) for coronary CTA, and 64% (95% CI: 53% to 74%) and 83% (95% CI: 77% to 88%) for ICA, respectively. In patients (n = 235) with intermediate stenosis (95% CI: 30% to 70%), the diagnostic accuracy of FFRCT remained high.

CONCLUSIONS: FFRCT provides high diagnostic accuracy and discrimination for the diagnosis of hemodynamically significant CAD with invasive FFR as the reference standard. When compared with anatomic testing by using coronary CTA, FFRCT led to a marked increase in specificity. 

PMID: 24486266

Posted in Computed Tomography, Invasive Imaging, Radiography and tagged , , , , , .


  1. A number of patients (n=47 or 13%) were excluded by the CT-FFR core laboratory, secondary to limited image quality, which appears to include calcification. In addition 56 patient were excluded after initial acceptance, including 23 for “other reasons.”

    In an “intention-to-diagnose” apporach, should these patients be included in the analysis of sensitivity and specificity?

  2. Thanks for the interest in the NXT trial Paul. When assessing the diagnostic accuracy of a diagnostic test the “intentention-to-diagnose” approach is preferable for transparent reporting of non-evalauble results. The NXT trial proportion of vessels with a non-evauable coronary segment (proximal to the segment in which FFRct was calculated) was 4.5% (22/484). In order to obtain the most conservative estimate of the “FFRct versus corCT” diagnostic performance, non-evaluable segments were assumed to be non-stenotic in the corCT accuracy analyses. As holds true for corCT and other invasive testing modalities, FFRct analysis cannot be performed in all patients. Based on previous experiences (DISCOVER-FLOW and DeFACTO) a minimum of image quality as assessed by an image quality score (based on the summed effects of motion, noise, contrast, and blooming on vessel lumen interpretability) was acquired (and pre-study defined) for patients to enter the study. Accordingly, 44 patients did not enter the study. Another 56 (exept for one “unspecified” case) patients did not enter the study because of other pre-study defined causes for “exclusion”. The causes of study screen-failure are described in great detail in Figure 2 in the JACC ms. Of note, 22 patients were not included in the study because of (study pre-defined) issues related to the quality of the reference standard (FFR) per se. Screen failures (based on pre-study defined criteria) are not to be considered “non-evaluable”, and thus should not be included in the NXT secondary end-point analyses assessing the FFRct diagnostic accuracy.
    With regard to coronary calcification per se, FFRct as shown in both the NXT trial (subanalyses presented in the JACC paper) and previous studies by Min J and Leipsic J
    seems rather diagnostic robust with high accuracy measures even in the event of an Agatston score >400. My own experience is that calcification in the event of good image quality is very rarely a problem for FFRct calculation or accuracy (when compared with the outcomes in cath lab), however, the combination of calcification and other artifacts (especially motion) -as for corCT interpretation- may compromize FFRct accuracy.

  3. Thank you for these explanations, Bjarne!
    I hope that some of my colleagues from the editorial team will comment as well.

  4. Thanks for all the added information!
    Really interesting to know that coronary calcification may not be the most limiting factor in calculating FFRct.

Leave a Reply

Your email address will not be published. Required fields are marked *