Cost-Effectiveness of Deep Learning-Enhanced CT-Derived Fractional Flow Reserve for Diagnostic Optimization in Stable Coronary Artery Disease
Author(s)
Tingting Xu, PhD1, Youli Han, phd2.
1capital medical university, beijing, China, 2capital medical university, Beijing, China.
1capital medical university, beijing, China, 2capital medical university, Beijing, China.
OBJECTIVES: Computed tomography-derived fractional flow reserve (CT-FFR), powered by advanced deep learning algorithms, is a promising diagnostic tool for coronary artery disease (CAD), offering superior accuracy and risk prediction over traditional computational fluid dynamics methods. However, its clinical and economic impacts in stable CAD patients in China remain underexplored.
METHODS: We used decision-analytic Markov models to assess the societal cost-effectiveness of deep learning-based CT-FFR compared with coronary angiography (CAG), fractional flow reserve (FFR), and quantitative flow ratio (QFR). The model tracked a cohort of patients with CT angiography (CTA)-detected stenosis (30%-90%) starting at age 50 over five 1-year cycles. Meta-regression estimated diagnostic probabilities for each strategy. Costs were derived from 12 tertiary hospitals in China. Primary outcomes included incremental cost-utility ratios (ICURs) based on quality-adjusted life-years (QALYs) and incremental cost-effectiveness ratios (ICERs) based on major adverse cardiac events (MACE) avoided. One-way deterministic and probabilistic sensitivity analyses addressed uncertainties.
RESULTS: CT-FFR with deep learning was cost-effective compared to CAG, FFR, and QFR in China. One-year ICURs were: US$-31,642.19 (95% CI: -106,453.00 to 42,968.62) vs. CAG, US$-112,215.49 (95% CI: -201,201.07 to -23,229.92) vs. FFR, and US$338,357.53 (95% CI: -219,811.42 to -896,526.00) vs. QFR. Corresponding ICERs were: US$-25,303.81 (95% CI: -42,687.70 to -7,919.93) vs. CAG, US$-48,616.70 (95% CI: -166,514.12 to 69,280.71) vs. FFR, and US$-289,614.91 (95% CI: -852,525.28 to 273,295.46) vs. QFR. Over five years, ICURs were: US$-27,864.92 (95% CI: -41,513.15 to -14,216.68) vs. CAG, US$-30,768.96 (95% CI: -55,844.35 to -5,693.58) vs. FFR, and US$3,230.49 (95% CI: -31,127.79 to 37,588.77) vs. QFR. All values were below the WHO-recommended cost-effectiveness threshold of 1-3× China’s GDP per capita. Results were robust across sensitivity analyses.
CONCLUSIONS: Deep learning-enhanced CT-FFR provides a transformative, cost-effective diagnostic approach for CAD in China, delivering significant clinical and economic value. These findings support its broader adoption to improve healthcare outcomes in resource-limited settings.
METHODS: We used decision-analytic Markov models to assess the societal cost-effectiveness of deep learning-based CT-FFR compared with coronary angiography (CAG), fractional flow reserve (FFR), and quantitative flow ratio (QFR). The model tracked a cohort of patients with CT angiography (CTA)-detected stenosis (30%-90%) starting at age 50 over five 1-year cycles. Meta-regression estimated diagnostic probabilities for each strategy. Costs were derived from 12 tertiary hospitals in China. Primary outcomes included incremental cost-utility ratios (ICURs) based on quality-adjusted life-years (QALYs) and incremental cost-effectiveness ratios (ICERs) based on major adverse cardiac events (MACE) avoided. One-way deterministic and probabilistic sensitivity analyses addressed uncertainties.
RESULTS: CT-FFR with deep learning was cost-effective compared to CAG, FFR, and QFR in China. One-year ICURs were: US$-31,642.19 (95% CI: -106,453.00 to 42,968.62) vs. CAG, US$-112,215.49 (95% CI: -201,201.07 to -23,229.92) vs. FFR, and US$338,357.53 (95% CI: -219,811.42 to -896,526.00) vs. QFR. Corresponding ICERs were: US$-25,303.81 (95% CI: -42,687.70 to -7,919.93) vs. CAG, US$-48,616.70 (95% CI: -166,514.12 to 69,280.71) vs. FFR, and US$-289,614.91 (95% CI: -852,525.28 to 273,295.46) vs. QFR. Over five years, ICURs were: US$-27,864.92 (95% CI: -41,513.15 to -14,216.68) vs. CAG, US$-30,768.96 (95% CI: -55,844.35 to -5,693.58) vs. FFR, and US$3,230.49 (95% CI: -31,127.79 to 37,588.77) vs. QFR. All values were below the WHO-recommended cost-effectiveness threshold of 1-3× China’s GDP per capita. Results were robust across sensitivity analyses.
CONCLUSIONS: Deep learning-enhanced CT-FFR provides a transformative, cost-effective diagnostic approach for CAD in China, delivering significant clinical and economic value. These findings support its broader adoption to improve healthcare outcomes in resource-limited settings.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
HTA98
Topic
Economic Evaluation, Health Technology Assessment, Medical Technologies
Topic Subcategory
Decision & Deliberative Processes
Disease
Cardiovascular Disorders (including MI, Stroke, Circulatory)