On-site Machine Learning-Based FFRCT Feasible, Informative: TARGET

This approach could cut down on unneeded angiograms, while catching obstructive CAD that might otherwise be missed.

On-site Machine Learning-Based FFRCT Feasible, Informative: TARGET

NEW ORLEANS, LA—A strategy that involves on-site machine learning to facilitate computed tomography-derived fractional flow reserve (FFRCT) in patients with new-onset stable coronary artery disease and intermediate stenoses on coronary CT angiography (CCTA) appears to help clinicians identify who needs to undergo further evaluation, new data from the randomized TARGET trial show.

Compared with standard care, this approach to FFRCT reduced the proportion of patients within 90 days who underwent coronary angiography in the absence of obstructive CAD or failed to undergo intervention in the presence of obstructive CAD. Moreover, the FFRCT strategy increased the overall rate of revascularization and lowered costs.

Overall, the TARGET findings indicate “an on-site machine-learning based CT-FFR strategy is feasible, safe, and effective,” said Yundai Chen, MD (Chinese PLA General Hospital, Beijing), who presented the data virtually during a featured clinical research session today at the American College of Cardiology/World Cardiology Congress (ACC/WCC) 2023 meeting. The results were simultaneously published online in Circulation.

Several studies have demonstrated FFRCT’s ability to noninvasively pinpoint functional myocardial ischemia, the researchers note. Among them, the PLATFORM trial demonstrated how FFRCT provides information that can guide testing in stable CAD, and at the same time saves money. Others, like FORECAST and an analysis of PROMISE trial data, dampened hopes that FFRCT could lower costs.

All of these studies relied on computational fluid dynamics to determine FFR values, with images transferred for off-site analysis, whereas TARGET involves on-site FFRCT measurement done by machine learning (DEEPVESSEL FFR; Keya Medical Technology). The on-site testing modality that received CE Mark in 2018 was approved by China’s National Medical Products Administration in 2020 and received 510(k) clearance from the US Food and Drug Administration in 2022.

TARGETing Patients for Angiography

Conducted at six Chinese centers, TARGET randomized 1,216 patients with stable CAD (72.9% with angina) and stenoses of 30% to 90% on CCTA to either the on-site FFRCT care pathway or to standard care.

  • With FFRCT, if values were ≤ 0.80 in one or more major coronary arteries, the patient was referred for invasive angiography; for values > 0.80, the next step was optimal medical therapy.
  • With standard care, local physicians chose whether to perform additional stress testing, typically exercise ECG; patients with positive stress tests underwent invasive angiography or revascularization if indicated.

Median time from enrollment to the post-CCTA strategy was 4 days with FFRCT and 9 days with standard care, a nonsignificant difference. Ultimately, 69.2% of the FFRCT group and 79.4% of the standard group underwent invasive angiography.

The likelihood that such angiography would turn up nonobstructive CAD, or that obstructive CAD would be found but interventions not performed, within 90 days (the primary endpoint) was significantly lower with the FFRCT strategy than with standard care (28.3% vs 46.2%; P < 0.001).

Use of the FFRCT approach increased the proportion of patients who underwent revascularization by 1 year (49.7% vs 42.8% with standard care; P = 0.02), with a similar likelihood of 1-year MACE between the two groups (HR 0.88; 95% CI 0.59-1.30). Both groups saw improvements in quality of life and symptoms, as well.

Additionally, costs trended 9% lower with FFRCT versus standard care (P= 0.07), though the TARGET investigators point out that testing costs and reimbursement models in China differ from those in the US and Europe.

The Novelty of On-site FFRCT

Pamela Douglas, MD (Duke Clinical Research Institute, Durham, NC), past president of the ACC, told TCTMD that what stands out most about the TARGET trial is the novelty of its on-site FFRCT analysis. Speaking with TCTMD, she noted that as a researcher in this area, she’s led trials funded by HeartFlow (PLATFORM and PRECISE) testing its proprietary technology wherein FFRCT is calculated off-site.

An on-site approach does have the potential to be cheaper and to more rapidly return results, said Douglas.

But she pointed out several reasons why it’s difficult to compare TARGET against other trials.

All the big-name studies that have come before in the CT space—PROMISE, SCOT-HEART, PRECISE, and FORECAST—were “de novo investigations of chest pain, so they were people at their first presentation,” she said.

But here in TARGET, everyone underwent up-front CCTA. “What comes out at the back end is a much higher rate of disease than one would expect and much higher rate of cath than one would expect in a primary-strategy chest pain trial,” said Douglas, which would dilute the impact of FFRCT. Moreover, most of the stress testing in the standard-care arm was exercise ECG (87.5%), whereas in the United States imaging stress tests would predominate.

In clinical practice, if CCTA is used as a first-line test, Douglas said the question then becomes: “What should you do next if you have an intermediate lesion?” For her, she said, “it’s kind of a bit of a no-brainer, in that the FFRCT is just a software analysis,” albeit very sophisticated, whereas ordering a stress test requires the patient to come back in at a later date and is not without risk. 

The on-site machine learning aspect of TARGET is important, the researchers note. “The advantages of using a deep-learning algorithm is that it provides the possibility for on-site deployment, avoiding the need for transferring sensitive medical data, shortening the calculation time, and increasing clinician participation.” While it’s also possible to calculate FFR using computational fluid dynamics on-site, this strategy is complex and requires substantial resources, they explain.

The convenience of machine learning will facilitate FFRCT’s application in broader scenarios, the investigators write, adding, “An on-site CT-FFR strategy is pragmatic and may be preferred to meet the demands of clinical practice under a variety of clinical settings.”

  • Yang J, Shan D, Wang X, et al. On-site computed tomography-derived fractional flow reserve to guide the management of patients with stable coronary artery disease: the TARGET randomized trial. Circulation. 2023;Epub ahead of print.

  • The TARGET trial was sponsored by grants from the National Key R&D Program of China and Beijing NOVA Program.
  • Chen reports consultant fees/honoraria from Cordis and Medtronic.
  • Douglas reports research grants to her institution from HeartFlow and funding from the National Institutes of Health for the PROMISE trial.