Sex Differences in Murray Law‐Based Quantitative Flow Ratio Among Patients With Intermediate Coronary Lesions

Abstract

Background

The Murray law‐based quantitative flow ratio (μQFR) is a novel technique that simulates fractional flow reserve (FFR) from a single angiographic view. However, the impact of sex differences on the diagnostic performance of μQFR has not been investigated.

Methods and Results

In this study, FFR and μQFR were assessed in 497 intermediate stenoses (30%–70% by visual estimation) from 460 patients (34.3% female). Physiological significance was defined as FFR ≤0.80 or μQFR ≤0.80. After adjusting for potential confounders, female sex was independently associated with higher FFR (P=0.048 and 0.026, respectively) and μQFR (P=0.001 for both) in both fully adjusted and stepwise backward models. μQFR provided superior diagnostic accuracy compared with angiography alone for detecting FFR ≤0.80 in both women (area under the curve, 0.93 [95% CI, 0.88–0.97] versus 0.80 [95% CI, 0.73–0.86]; P=0.001) and men (area under the curve, 0.88 [95% CI, 0.84–0.92] versus 0.73 [95% CI, 0.68–0.78]; P<0.001), with comparable performance between the sexes (P=0.175). In the multivariable analysis, sex was not a significant factor contributing to the overall disagreement between FFR and μQFR.

Conclusions

Regardless of angiographic stenosis severity, women tend to have higher FFR and μQFR values than men. Furthermore, μQFR performs similarly well in both sexes and offers improved diagnostic accuracy over angiography alone, indicating its potential as a reliable, wire‐free tool to identify functional ischemia.

Nonstandard Abbreviations and Acronyms

μQFR

Murray law‐based quantitative flow ratio

AMR

angiographic microvascular resistance

BCT

Box‐Cox transformation

CT‐FFR

coronary computed tomography angiography‐derived fractional flow reserve

FFR

fractional flow reserve

MLD

minimum lumen diameter

QCA‐%DS

quantitative coronary angiography‐derived percent diameter stenosis

QFR

quantitative flow ratio

Clinical Perspective.

What Is New?

• Women had higher continuous fractional flow reserve and Murray law‐based quantitative flow ratio (μQFR) values, even after adjusting for angiographic stenosis severity and other confounding variables.

• μQFR had comparable diagnostic performance between the sexes and significantly improved the detection of physiological significance, as defined by fractional flow reserve, over angiography alone.

• Sex was not a significant factor contributing to the overall discrepancy between fractional flow reserve and μQFR.

What Are the Clinical Implications?

• The distribution of fractional flow reserve and μQFR based on sex may suggest the potential for future optimization of these indices in sex‐specific populations.

• Results from both male and female patients show that μQFR is a dependable and wire‐free tool that exhibits higher diagnostic accuracy than angiographic diameter stenosis in detecting functional ischemia.

Fractional flow reserve (FFR) is commonly used in the catheterization laboratory to assess hemodynamic significance and aid in decision‐making for coronary revascularization. ¹ , ² Nevertheless, the use of invasive FFR has several barriers, including prolonged procedure time, increased medical costs, and hyperemia‐induced discomfort. ³ Consequently, angiography‐based quantitative flow ratio (QFR) has been developed to assess the physiological severity of coronary stenosis without the need of a pressure wire or hyperemic agents. ⁴ Although QFR demonstrated robust agreement with invasive FFR in various prospective studies, ⁵ , ⁶ , ⁷ it is constrained by the necessity for 2 angled angiographic projections to create a 3‐dimensional reconstruction of the coronary tree. Therefore, a new and improved algorithm called the Murray law‐based QFR (μQFR) has been developed, which enables automatic contour delineation and swift FFR simulation from a single angiographic view. ⁸ Its derivative index, called angiographic microvascular resistance (AMR), has been validated to have a high accuracy in the assessment of coronary microvascular dysfunction. ⁹ These emerging techniques hold the potential to improve the accessibility of physiological assessment and enhance resource use in clinical practice. ¹⁰

There are differences between women and men in the pathogenesis of atherosclerotic coronary artery disease, clinical presentation, and even prognosis. ¹¹ , ¹² , ¹³ Despite the lower prevalence of functional ischemia in women, women and men received equal benefit from an FFR‐guided revascularization strategy. ¹⁴ Women's smaller myocardial territory may contribute to their higher FFR values and a more frequent anatomic‐physiologic mismatch. ¹⁵ Nevertheless, limited data exist on the sex differences in μQFR calculation and discordance between μQFR and FFR. The objective of this study was to investigate the sex‐specific distribution of μQFR and examine its potential influence on the diagnostic performance of μQFR in identifying physiologically significant coronary lesions.

Methods

Data Availability

The corresponding authors have complete access to all data in the study and are accountable for their integrity. Anonymized data supporting the findings can be obtained from the corresponding authors upon reasonable request.

Study Population

Consecutive patients who underwent coronary angiography and simultaneous pressure wire‐based physiological evaluation for at least 1 intermediate coronary stenosis (30% to 70% stenosis by visual estimation) between July 2013 and December 2019 were considered for eligibility by screening the institutional database of our department. Subjects with significant left main coronary artery disease and myocardial infarction within 72 hours were excluded from the analysis. Additionally, interrogated vessels with in‐stent restenosis, myocardial bridging, poor angiographic image quality, and severe overlap or tortuosity were excluded based on predetermined criteria. Demographic, clinical, imaging, and physiological parameters for each subject were obtained from the electronic medical database. This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for reporting observational studies. ¹⁶

Coronary Angiography and FFR Measurement

The cardiac catheterization procedures and subsequent FFR measurement were previously described in detail. ¹⁷ A 5‐ or 6‐French catheter was used to perform standard coronary angiography via the radial artery. Angiographic projections were routinely acquired at 15 frames/s using a monoplane radiographic system (AXIOM Artis, Siemens Healthineers, Erlangen, Germany) and multiple angles were attempted to obtain the best visualization of target lesions. A preprocedural bolus of heparin (100 IU/kg) was administered, followed by intravenous infusion of an additional dose based on the activated clotting time of whole blood. Intracoronary nitroglycerin was administered as needed before angiography and physiological assessments.

The decision to measure FFR was made by the operators after angiographic coronary stenoses were visually assessed. A 0.014‐inch pressure guidewire (Certus, St. Jude Medical, St. Paul, MN) was used to measure the intracoronary pressure. The guidewire was introduced via a guiding catheter, and the pressure sensor was equalized to the aortic pressure before being positioned distal to the lesion. Coronary hyperemia was induced by administering adenosine‐5′‐triphosphate through the antecubital vein (140 μg/kg per min). The ratio of mean coronary pressure at distal segment to concurrent mean aortic pressure during steady‐state maximal hyperemia was used to calculate vessel FFR. ¹ , ¹⁸ An FFR cutoff value of ≤0.80 was considered to be indicative of hemodynamic significance. ¹

μQFR and AMR Computation

The calculation of μQFR and AMR with simultaneous quantitative coronary angiography (QCA) analysis was performed independently using commercial software (AngioPlus Core, Pulse Medical Imaging Technology Co., Ltd., Shanghai, China) by certified technicians who were blinded to clinical data and physiological results, as previously described. ⁸ In brief, semiautomated delineation of vessel contours and FFR simulation from single‐angle rather than double‐angle angiographic projections was enabled by this artificial intelligence‐based algorithm. A key frame with a sharp lumen contour at the stenotic segment was first selected for subsequent analysis from an optimal angiographic view. The vessel contours were then automatically delineated, and the reference diameter function was reconstructed by the software. Proximal and distal reference vessel diameters could be manually adjusted as needed. Finally, the Murray fractal law was used to calculate μQFR values for both main epicardial vessels and their side branches (Figure 1). The position of the pressure sensor was used to calculate μQFR by reviewing the angiographic image. Simultaneously, QCA parameters including percent diameter stenosis (QCA‐%DS), reference vessel diameter, minimum lumen diameter, and lesion length were obtained. Tandem lesions were defined as the presence of at least 2 stenoses (diameter stenosis ≥50% by visual estimation) ≥10 mm apart in the same epicardial artery, and the most severe lesion was selected for QCA analysis. The AMR ratio was calculated as distal coronary pressure to hyperemic flow velocity and was further developed into mean aortic pressure (assumed to be 86 mm Hg)×μQFR/hyperemic flow velocity. ⁹ Hemodynamic significance was indicated by a μQFR cutoff value of ≤0.80. ⁸ A μQFR‐FFR hybrid strategy was adopted, in which μQFR was used as the main diagnostic approach and wire‐based FFR measurement was performed only in a gray zone of μQFR around the cutoff value. The μQFR cutoffs were set to provide both sensitivity and specificity >90% compared with FFR.

Figure 1. Representative cases of μQFR computation in women and men.

The μQFR values of the main vessel and its side branches were calculated using the Murray bifurcation fractal law. A, The coronary angiography image reveals an intermediate lesion (red asterisk) located at the proximal segment of the left anterior descending artery in a female patient. The wire‐based FFR, μQFR, and AMR are 0.81, 0.83, and 2.29 mm Hg×s/cm, respectively. (B) Another angiographic image displays an intermediate lesion (red asterisk) located at the distal segment of the left anterior descending artery in a male patient. The wire‐based FFR, μQFR, and AMR are 0.72, 0.75, and 1.39 mm Hg×s/cm, respectively. Despite having a similar degree of coronary stenosis, the man's FFR and μQFR values are positive, whereas the woman's values are negative. AMR indicates angiographic microcirculatory resistance; FFR, fractional flow reserve; and μQFR, Murray law‐based quantitative flow ratio.

Statistical Analysis

Categorical variables were presented as counts (percentages) and compared using the chi‐square test. The normality of continuous variables was assessed using histograms and Q‐Q plots. For normally distributed variables, the mean±SD was presented and compared using Student's t test; and nonnormally distributed variables were presented as median (25th–75th percentile) and compared using the Mann–Whitney U test. The correlation between FFR and μQFR was determined using Spearman's correlation analysis, and Bland–Altman plots were used to assess agreement. A Z test based on Fisher's Z transformation was used to compare the correlation coefficients between 2 independent groups. Receiver‐operating characteristics curves were plotted to evaluate the diagnostic performance of μQFR and QCA‐%DS using wire‐based FFR as the reference standard. Delong's method ¹⁹ was used to compare the discrimination, represented by the area under the curve (AUC). A per‐vessel multiple linear mixed‐effects model with compound symmetry was used to examine the association between continuous FFR or μQFR (dependent variable) and sex (independent variable) adjusted for potential confounders (fixed effect) and the clustered nature of vessels within the same patient (random effect), using both the fully adjusted and stepwise backward models (P>0.20 for exclusion). Covariates were included at 3 levels to further evaluate their interaction, with model 1 considering sex and angiographic lesion severity, model 2 adding vessel parameters to model 1, and model 3 adding patient characteristics to model 2. Multicollinearity was assessed using variance inflation factor and tolerance. The Box‐Cox transformation (BCT) method was applied to improve the normality of the dependent variable. The optimal λ values for FFR and μQFR were selected as 3.228067 and 4.785091, respectively, based on maximum likelihood estimate. Logistic link function was used to perform a per‐vessel multivariate analysis with generalized linear mixed model to establish the effect of sex on overall FFR‐μQFR mismatch. A 2‐tailed P value of <0.05 was considered statistically significant. All statistical analyses were performed with SPSS version 25.0 (IBM Corp, Armonk, NY), Stata SE version 15.0 (Stata Corp, College Station, TX) and MedCalc statistical software version 19.0.7 (MedCalc Software bvba, Ostend, Belgium).

Ethical Approval

The study protocol was approved by the institutional review board of Zhongda Hospital, School of Medicine, Southeast University. The requirement for informed consent was waived due to our retrospective design.

Results

Baseline Patient and Lesion Characteristics

The patient selection flow chart is depicted in Figure S1. The final cohort included 497 vessels from 460 patients, of whom 158 were women (34.3%) and 302 were men (63.7%). Women were older than men (68.1±9.2 years versus 65.3±10.3 years; P=0.005) but had a lower incidence of dyslipidemia (53.3% versus 33.5%; P<0.001). In contrast, men had a higher prevalence of tobacco use (48.7% versus 2.5%; P<0.001) and a greater proportion of prior myocardial infarction (10.9% versus 4.4%; P=0.019) and stent implantation (29.8% versus 15.8%; P=0.001) (Table 1). The 2 groups were comparable in terms of body mass index, hypertension, diabetes, multivessel disease, and clinical presentation. Women had significantly lower QCA‐%DS (36.4±8.2% versus 39.4±8.1%; P<0.001) and shorter lesion length (18.3 [11.6–27.3] mm versus 21.6 [11.8–33.1] mm; P=0.044) compared with men. The distribution of vessel territory and lesion location was similar between the 2 sexes.

Table 1.

Baseline Characteristics of Included Population and Interrogated Vessels

Per‐patient analysis	Men (N=302)	Women (N=158)	P value
Age, y	65.3±10.3	68.1±9.2	0.005
Body mass index, kg/m2	24.9±3.4	24.9±3.8	0.917
Hypertension	219 (72.5)	126 (79.7)	0.089
Diabetes	92 (30.5)	42 (26.6)	0.384
Dyslipidemia	161 (53.3)	53 (33.5)	<0.001
Smoking	147 (48.7)	4 (2.5)	<0.001
Prior MI	33 (10.9)	7 (4.4)	0.019
Prior percutaneous coronary intervention	90 (29.8)	2 (15.8)	0.001
Multivessel disease	183 (60.6)	82 (51.9)	0.073
Clinical presentation			0.218
Stable angina	254 (84.1)	141 (89.2)
Unstable angina	40 (13.2)	16 (10.1)
Non–ST‐segment–elevation MI	8 (2.6)	1 (0.6)

Per‐vessel analysis	(n=329)	(n=168)
Interrogated vessel			0.600
Left anterior descending artery	220 (66.9)	115 (68.5)
Left circumflex artery	56 (17.0)	23 (13.7)
Right coronary artery	53 (16.1)	30 (17.9)
Lesion location			0.514
Proximal	167 (50.8)	78 (46.4)
Middle	141 (42.8)	81 (48.2)
Distal	21 (6.4)	9 (5.4)
Quantitative coronary angiography analysis
Diameter stenosis, %	39.4±8.1	36.4±8.2	<0.001
Reference vessel diameter, mm	3.0 (2.6, 3.5)	2.9 (2.5, 3.4)	0.477
Minimum lumen diameter, mm	1.80 (1.54, 2.16)	1.87 (1.57, 2.19)	0.188
Lesion length, mm	21.6 (11.8, 33.1)	18.3 (11.6, 27.3)	0.044
Tandem lesion	46 (14.0)	14 (8.3)	0.068
Physiological indexes
FFR	0.85 (0.79, 0.90)	0.89 (0.83, 0.92)	<0.001
FFR ≤0.80	103 (31.3)	27 (16.1)	<0.001
μQFR	0.86 (0.79, 0.91)	0.90 (0.84, 0.94)	<0.001
μQFR ≤0.80	88 (26.7)	25 (14.9)	0.003
Angiography‐derived microvascular resistance, mm Hg×s/cm	2.3±0.6	2.3±0.5	0.175
Hyperemic flow velocity, cm/s	16.4 (13.0, 20.0)	16.3 (13.7, 20.0)	0.927

Values are given as mean±SD, median (25th, 75th percentiles), or n (%). FFR indicates fractional flow reserve; MI, myocardial infarction; and μQFR, Murray law‐based quantitative flow ratio.

Sex‐Specific Distribution of FFR and μQFR

Median FFR and μQFR values per vessel were higher in women than in men, as shown in Table 1. Specifically, women had a higher median FFR (0.89 [0.83–0.92]) than men (0.85 [0.79–0.90]), as well as a higher median μQFR (0.90 [0.84–0.94]) than men (0.86 [0.79–0.91]), with a statistically significant difference between the 2 groups (P<0.001 for both). Men had a higher proportion of FFR ≤0.80 (31.3%) and μQFR ≤0.80 (26.7%) compared with women (16.1% and 14.9%, respectively). These differences were statistically significant (P<0.001 for FFR, and P=0.003 for μQFR). Subgroup analysis revealed that women had a higher FFR and μQFR values in the 30% to 39% QCA‐%DS category (Figure S2). When categorized into quartiles by lesion length, women had significantly higher FFR values for women in the first and second quartiles, and μQFR values were significantly higher in the first quartile (Figure S2).

Tables 2 and 3 show a positive association between female sex and continuous FFR (P=0.048 and 0.026, respectively) and continuous μQFR (P=0.001 for both) after BCT in both fully adjusted and stepwise backward models. To assess the interaction between sex and other confounders, we developed 3 separate linear mixed‐effect models (Table S1). In model 1, which included sex and angiographic lesion severity as covariates, sex was an independent predictor of BCT‐FFR (P=0.002) and BCT‐μQFR (P=0.013). The addition of vessel parameters to the analysis (model 2) maintained the statistical significance of sex as a predictor of BCT‐FFR (P<0.001) and BCT‐μQFR (P<0.001). In model 3, where patient characteristics were added to model 2, sex had a significant impact on BCT‐QFR (P=0.001), whereas its effect on BCT‐FFR was close to the threshold (P=0.048).

Table 2.

Multivariate Analysis for Continuous FFR After Box‐Cox Transformation

	Fully adjusted model		Stepwise backward model
Variable	Estimate±SE	P value	Estimate±SE	P value
Intercept	−0.088±0.025	<0.001	−0.033±0.013	0.012
Female sex	0.008±0.004	0.048	0.008±0.004	0.026
Quantitative coronary angiography‐derived percent diameter stenosis/10%	−0.018±0.003	<0.001	−0.023±0.002	<0.001
Lesion length/10 mm	−0.005±0.001	<0.001	−0.004±0.001	0.003
Minimum lumen diameter, mm	0.012±0.004	0.001	…	…
Multivessel disease	−0.015±0.004	<0.001	−0.016±0.003	<0.001
Left anterior descending artery location	−0.049±0.004	<0.001	−0.051±0.004	<0.001
Proximal segment	−0.002±0.003	0.527	…	…
Angiography‐derived microvascular resistance, mm Hg×s/cm	0.025±0.003	<0.001	0.023±0.003	<0.001
Tandem lesion	0.0002±0.0048	0.972	…	…
Age/10 y	0.002±0.002	0.285	…	…
Body mass index/10 kg per m2	−0.005±0.005	0.301	…	…
Hypertension	0.006±0.004	0.111	…	…
Diabetes	0.006±0.003	0.063	0.005±0.003	0.177
Dyslipidemia	0.002±0.003	0.616	…	…
Smoking	−0.006±0.004	0.150	−0.006±0.004	0.131
Prior myocardial infarction	−0.009±0.006	0.145	−0.014±0.006	0.014
Prior percutaneous coronary intervention	−0.005±0.004	0.275	…	…
Stable angina	0.005±0.005	0.266	…	…

Per‐vessel multiple linear mixed‐effects model with compound symmetry was performed to show the association between continuous FFR after Box‐Cox transformation (dependent variable) and sex (independent variable) adjusted on confounders (fixed effect) and clustered nature of vessels within the same patient (random effect). Hyperemic flow velocity and reference vessel diameter were not included due to multicollinearity. FFR indicates fractional flow reserve.

Table 3.

Multivariate Analysis for Continuous μQFR After Box‐Cox Transformation

	Fully adjusted model		Stepwise backward model
Variable	Estimate±SE	P value	Estimate±SE	P value
Intercept	−0.055±0.012	<0.001	−0.029±0.009	0.001
Female sex	0.007±0.002	0.001	0.007±0.002	0.001
Quantitative coronary angiography‐derived percent diameter stenosis/10%	−0.023±0.001	<0.001	−0.026±0.001	<0.001
Lesion length/10 mm	−0.007±0.001	<0.001	−0.006±0.001	<0.001
Minimum lumen diameter, mm	0.006±0.002	0.001	…	…
Multivessel disease	−0.008±0.002	<0.001	−0.009±0.002	<0.001
Left anterior descending artery location	−0.029±0.002	<0.001	−0.030±0.002	<0.001
Proximal segment	0.002±0.002	0.238	0.003±0.002	0.041
Angiography‐derived microvascular resistance, mm Hg×s/cm	0.032±0.002	<0.001	0.031±0.002	<0.001
Tandem lesion	−0.007±0.002	0.006	−0.007±0.002	0.004
Age/10 y	0.0004±0.0008	0.613	…	…
Body mass index/10 kg per m2	−0.005±0.002	0.040	−0.004±0.002	0.071
Hypertension	0.003±0.002	0.135	0.002±0.002	0.186
Diabetes	0.001±0.002	0.625	…	…
Dyslipidemia	0.003±0.002	0.063	0.003±0.002	0.068
Smoking	0.001±0.002	0.675	…	…
Prior myocardial infarction	0.005±0.003	0.119	…	…
Prior percutaneous coronary intervention	−0.001±0.002	0.495	…	…
Stable angina	0.002±0.002	0.277	…	…

Per‐vessel multiple linear mixed‐effects model with compound symmetry was performed to show the association between continuous μQFR after Box‐Cox transformation (dependent variable) and sex (independent variable) adjusted on confounders (fixed effect) and clustered nature of vessels within the same patient (random effect). Hyperemic flow velocity and reference vessel diameter were not included due to multicollinearity. μQFR indicates Murray law‐based quantitative flow ratio.

Correlation and Agreement Between FFR and μQFR

Spearman's correlation coefficients were strong between FFR and μQFR in both women (rho=0.790; P<0.001) and men (rho=0.771; P<0.001), with no significant difference in correlation coefficients between the sexes (P=0.611). Bland–Altman plots showed good agreement between μQFR and FFR in both women (mean difference: 0.008±0.041) and men (mean difference: 0.006±0.058) (Figure 2).

Figure 2. Scatterplots and Bland–Altman analysis for the association between FFR and μQFR.

A and C, Strong correlation was observed between FFR and μQFR in both sexes, as evidence by Spearman's correlation analysis (women: rho=0.790, men: rho=0.771). B and D, Good agreement between FFR and μQFR was demonstrated in Bland–Altman analysis for women (mean difference: 0.008±0.041) and men (mean difference: 0.006±0.058). FFR indicates fractional flow reserve; and μQFR, Murray law‐based quantitative flow ratio.

Diagnostic Performance of μQFR and Angiography

The receiver operating characteristic curves demonstrated a strong discriminatory ability in both women (AUC, 0.93 [95% CI, 0.88–0.97]) and men (AUC, 0.88 [95% CI, 0.84–0.92]), with no significant difference between the 2 groups (P=0.175). In detecting FFR ≤0.80, μQFR showed a superior improvement over QCA‐%DS in both women (AUC, 0.93 [95% CI, 0.88–0.97] versus 0.80 [95% CI, 0.73–0.86]; P=0.001) and men (AUC, 0.88 [95% CI, 0.84–0.92] versus 0.73 [95% CI, 0.68–0.78]; P<0.001), as illustrated in Figure 3. Table 4 displays the diagnostic parameters of μQFR and QCA‐%DS for detecting functional ischemia.

Figure 3. Diagnostic performance of μQFR and angiography for detection of functional ischemia according to sex.

The per‐vessel diagnostic performance of μQFR and QCA‐%DS was expressed as the area under the curve for detecting functional ischemia (FFR ≤0.80). In both women (A) (AUC: 0.93 vs 0.80; P=0.001) and men (B) (AUC: 0.88 vs 0.73, P<0.001), μQFR exhibited better performance than QCA‐%DS. No significant difference was found in AUCs between the 2 sexes (P=0.175). AUC indicates area under the curve; FFR, fractional flow reserve; QCA‐%DS, quantitative coronary angiography‐derived percent diameter stenosis; and μQFR, Murray law‐based quantitative flow ratio.

Table 4.

Per‐Vessel Diagnostic Performance of μQFR and Angiography Stratified by Sex

Parameters	Men		Women
	μQFR	QCA‐%DS	μQFR	QCA‐%DS
Accuracy	83.3 (78.8–87.2)	70.8 (65.6–75.7)	92.9 (87.9–96.3)	84.5 (78.2–89.6)
Sensitivity	66.0 (56.0–75.1)	21.4 (13.9–30.5)	74.1 (53.7–88.9)	22.2 (8.6–42.3)
Specificity	91.2 (86.7–94.5)	93.4 (89.3–96.2)	96.5 (91.9–98.8)	96.5 (91.9–98.8)
Positive predictive value	77.3 (68.6–84.1)	59.5 (44.3–73.0)	80.0 (62.2–90.7)	54.5 (28.3–78.5)
Negative predictive value	85.5 (81.8–88.5)	72.3 (70.1–74.3)	95.1 (91.1–97.4)	86.6 (84.1–88.8)
LR+	7.5 (4.8–11.6)	3.2 (1.7–5.9)	20.9 (8.6–50.8)	6.3 (2.1–19.1)
LR−	0.4 (0.3–0.5)	0.8 (0.8–0.9)	0.3 (0.1–0.5)	0.8 (0.7–1.0)

Values are given as n or % with 95% CIs. Positive test results are defined as μQFR ≤0.80 and QCA‐%DS ≥50%. LR+ indicates positive likelihood ratio; LR−, negative likelihood ratio; QCA‐%DS, quantitative coronary angiography‐derived percent diameter stenosis; and μQFR, Murray law‐based quantitative flow ratio.

Discordance Between FFR and μQFR

When vessels were categorized by μQFR, there was no significant difference in the prevalence of functional ischemia between women and men across all 4 quartiles, and FFR showed a higher value in women than in men in the third quartile (0.89–0.92) (Figure S3). On the other hand, when vessels were categorized by FFR, men had a comparable frequency of μQFR ≤0.80 as women across all 4 quartiles of FFR, but the third quartile (0.87–0.91) showed higher μQFR values in women compared with men (Figure S4). Accuracy per‐vessel varied across different μQFR ranges, with low accuracy in the μQFR range of 0.76 to 0.80 for women and 0.81 to 0.85 for men, when using μQFR ≤0.80 as the cutoff value (Figure S5). Using a prespecified assumption of sensitivity and specificity >90%, the gray zones of the μQFR were 0.84 to 0.86 in women and 0.81 to 0.85 in men (Figure S6). A μQFR‐FFR hybrid strategy would require further wire‐based FFR measurements in only 10.1% (women) and 20.4% (men) of vessels. According to the AUC analysis, the μQFR gray zones with sensitivity and specificity ≤95% were 0.82 to 0.86 in women and 0.75 to 0.88 in men, resulting in further invasive coronary pressure measurements in 15.5% and 45.3% of cases, respectively.

Using FFR ≤0.80 as the reference standard, men showed higher rates of overall discordance than women for μQFR (16.7% versus 7.1%; P=0.003) and QCA‐%DS (29.2% versus 15.5%; P=0.001) (Figure 4). To investigate whether sex is an influence factor of FFR‐μQFR discordance, we conducted univariate and multivariate analyses to adjust for confounding variables. Table S2 displays patient and lesion characteristics according to the presence of FFR‐μQFR mismatch. There were notable differences in the distribution of sex, dyslipidemia, multivessel disease, QCA‐%DS, minimum lumen diameter, lesion length, and AMR between the 2 groups. After adjustment for other confounders, multivariate analysis showed that female sex was not responsible for a lower frequency of FFR‐μQFR mismatch (Table 5).

Figure 4. Frequency of mismatch in μQFR and angiographic diameter stenosis versus FFR.

There were significant differences in the classifications of FFR versus μQFR (A) and QCA‐%DS (B) between women and men (overall P<0.001 and=0.002, respectively). The rates of total mismatch were higher in men than women for both μQFR (16.7% vs 7.1%; P=0.003) and QCA‐%DS (29.2% vs 15.5%; P=0.001). FFR indicates fractional flow reserve; QCA‐%DS, quantitative coronary angiography‐derived percent diameter stenosis; and μQFR, Murray law‐based quantitative flow ratio.

Table 5.

Generalized Linear Mixed Model Examining the Association Between Sex and Overall FFR‐μQFR Mismatch

Variables	Univariate			Multivariate
	OR	95% CI	P value	OR	95% CI	P value
Female sex	0.363	0.188–0.699	0.002	0.736	0.359–1.511	0.403
Hypertension	0.678	0.388–1.185	0.172	0.592	0.335–1.048	0.072
Diabetes	1.421	0.836–2.413	0.194	1.271	0.756–2.137	0.366
Dyslipidemia	1.981	1.177–3.334	0.010	1.731	1.032–2.905	0.038
Smoking	2.165	1.299–3.608	0.003	1.637	0.941–2.850	0.081
Prior percutaneous coronary intervention	1.873	1.084–3.235	0.024	1.848	1.069–3.193	0.028
Multivessel disease	1.969	1.110–3.495	0.021	1.230	0.668–2.262	0.506
Quantitative coronary angiography‐derived percent diameter stenosis/10%	2.115	1.523–2.935	<0.001	1.572	1.030–2.399	0.036
Minimum lumen diameter, mm	0.460	0.262–0.806	0.007	0.758	0.392–1.465	0.409
Lesion length/10 mm	1.433	1.190–1.725	<0.001	1.320	1.081–1.613	0.007
Angiography‐derived microvascular resistance/mm Hg×s/cm	0.566	0.345–0.930	0.025	0.861	0.506–1.464	0.579

Per‐vessel univariate and multivariate analysis for overall FFR‐μQFR mismatch (dependent variable) was performed using a generalized linear mixed model with logistic link function to adjust confounders (fixed effect) and clustered nature of vessels within the same patient (random effect). FFR indicates fractional flow reserve; OR, odds ratio; and μQFR, Murray law‐based quantitative flow ratio.

Discussion

In this study, we examined the sex‐related distribution of an artificial‐intelligence‐based μQFR algorithm and its impact on diagnostic performance for identifying functional ischemia in patients with intermediate coronary stenosis. The main results are as follows: (1) women tend to have higher FFR and μQFR independently of angiographic lesion severity; (2) μQFR is superior to angiography alone in detecting the physiological significance, regardless of sex; (3) μQFR has a similar discriminatory power between the sexes when FFR ≤0.80 is used as the reference standard; and (4) sex is not an independent predictor of overall FFR‐μQFR discordance after adjusting for other confounders. To the best of our knowledge, this is the first study to investigate the impact of sex differences on the diagnostic accuracy of μQFR compared with FFR. Our findings support the reliability of μQFR and its superiority to conventional angiography in both women and men.

Sex Differences in FFR

In line with previous studies, ¹⁴ , ²⁰ , ²¹ , ²² our study observed that women tend to have higher FFR values than men. The difference could be attributed to several factors, including older age, less severe diameter stenosis, lower subtended myocardial mass, or increased microvascular resistance. ¹⁵ , ²³ In contrast, microvascular resistance did not vary between the sexes in our study population, which is consistent with other studies of microcirculatory function. ²⁴ , ²⁵ Despite being associated with several factors, continuous FFR remained significantly affected by sex in the multivariate analysis. A possible explanation for this phenomenon could be the variations in plaque characteristics. A recent post hoc analysis by Kim et al ²¹ indicated that women had a lower frequency of lipid‐rich plaque, low attenuation plaque, and positive remodeling on coronary computed tomography angiography images. Their analysis demonstrated that sex was not an independent predictor of continuous FFR and functional ischemia after adjusting for stenosis severity, high‐risk plaque characteristics, and left ventricular mass. Our previous study also supported this observation, reporting that high‐risk plaque characteristics such as thin‐cap fibroatheroma, lipid‐rich plaque, and thrombus, identified by optical coherence tomography, were negatively associated with 3‐dimensional QFR. ²⁶ After controlling for patient characteristics, our study found a marginal association between sex and FFR values. These findings imply that other factors or confounders may be responsible for the observed difference in FFR values between men and women.

Sex Differences in μQFR

In the context of another computational index, coronary computed tomography angiography‐derived FFR, there is also a similar sex‐related pattern. Specifically, women have a higher coronary volume to myocardial mass ratio and coronary computed tomography angiography‐FFR value, which are both associated with a lower probability of revascularization. ²⁷ , ²⁸ Furthermore, coronary computed tomography angiography‐FFR can accurately detect lesion‐specific ischemia in both sexes and performs better than noninvasive computed tomography angiography alone. ²⁹ , ³⁰ However, information regarding sex differences in the angiography‐derived μQFR algorithm is limited. Hou et al ³¹ found a lower rate of lesions with contrast‐derived QFR ≤0.80 and complete revascularization in the female group in a study of 353 patients with ST‐elevation myocardial infarction. Our study is novel in demonstrating that, although sex is associated with μQFR, there is no significant difference in the diagnostic performance of μQFR between women and men when FFR ≤0.80 is used as the reference standard. Additionally, μQFR is more accurate than QCA‐%DS in identifying flow‐limiting lesions regardless of sex. The frequency of FFR‐μQFR disagreement is significantly associated with angiographic lesion severity but not with sex. These findings support the reliability of μQFR in both sexes, although further studies are required to explain the mechanism behind the sex‐specific distribution of μQFR. ³²

Independent Predictors of FFR‐μQFR Discordance

Our study revealed that 13.5% of the interrogated vessels exhibited misclassification between FFR and μQFR, which is consistent with existing incidence estimates. ³³ , ³⁴ According to a meta‐analysis of individual patient data, the discordance rate between FFR and μQFR was 13% in prospectively enrolled patients, ranging from 7.3% to 17%. ³³ Previous studies have suggested that the discordance may be due to patient characteristics, anatomic lesion severity, microvascular function, and even aortic valve area. ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ Our results similarly indicated that dyslipidemia, prior revascularization, QCA‐%DS, and lesion length were independent predictors of discordance between FFR and μQFR, whereas sex was not a significant contributor. Interestingly, although AMR was a significant indicator of sex differences, it was not an independent predictor of discordance between the 2 physiological measures, even though it differed significantly between the 2 groups according to the presence of FFR‐μQFR discordance. This phenomenon may be attributed to the overlapping effect of included covariates. Furthermore, compared with direct pressure sensors, AMR relies more on the anatomy on the 2‐dimensional image and may not fully reflect subtle variations in individual microvascular function. We hypothesize that underlying mechanisms, such as plaque vulnerability, plaque eccentricity, and diffuse narrowing, affect wire‐based FFR measurement, explaining the positive association between lesion severity and FFR‐μQFR misclassification. ³⁸ , ³⁹

Clinical Implications

Although women have higher FFR and μQFR values than men with similar degrees stenosis, the same cutoff value of ≤0.80 has been universally used in both sexes. The 2‐year data from the FAME (Fractional Flow Reserve Versus Angiography for Multivessel Evaluation) substudy revealed that the rates of major adverse cardiac events were comparable between the sexes (20.3% versus 20.2%; P=0.923) regardless of the therapy used, even after baseline comorbidities were adjusted. ¹⁴ In another subgroup analysis of the 3V‐FFR‐FRIENDS (Influence of Total Atherosclerotic Burden Assessed by 3‐Vessel Fractional Flow Reserve (FFR) on the Clinical Outcomes of the Patients With Multi‐Vessel Disease) study, Kim et al ⁴⁰ concluded that sex did not affect the prognostic effect of the total anatomic and physiologic disease burden at the 2‐year follow‐up. Nevertheless, it is too soon to draw a definitive conclusion due to the limited adverse events in a stable population. A retrospective single‐center study of 1331 patients reported inconsistent findings that women had more significant 1‐year vascular adverse outcomes, especially in deferred lesions. ²² Redefining the optimal threshold of FFR or μQFR for women might aid in mitigating future adverse events. This analysis also demonstrated sex differences in the minimum accuracy distribution and gray‐zone ranges (Figures S5 and S6). Nevertheless, before clinical application, a sex‐specific approach to revascularization decision‐making necessitates further discussion. Additional subgroup analyses of FAVOR III China (NCT03656848) and FAVOR III Europe‐Japan (NCT03729739) are necessary to assess sex‐specific outcomes of QFR‐guided versus angiography‐guided and QFR‐guided versus FFR‐guided strategies. ⁴¹ , ⁴²

Study Limitations

Several limitations should be considered when interpreting the results. First, the decision to perform FFR measurements was based on the comprehensive assessment of the operators, which could introduce selection bias due to the single‐center, retrospective design. Second, almost three quarters of the lesions were physiologically nonsignificant according to FFR criteria. This might result in insufficient power to correct the interaction between variables. Nevertheless, our findings reflect the real‐world scenario that visual assessment might overestimate the severity of angiographic diameter stenosis when compared with QCA analysis. ⁴³ Despite a relatively lower frequency of ischemic lesions, the proportion of FFR‐μQFR discordance was similar to those previously reported. ⁴ , ³⁴ Therefore, additional studies with a larger sample size and a more extensive range of lesion severity are required to confirm our results. Third, the lack of follow‐up data limited our ability to determine the prognostic value of μQFR using a sex‐specific approach. We also did not examine the effect of sex on lesion‐specific μQFR and μQFR difference across the lesion, which could be novel indicators for diagnosing focal or diffuse disease patterns and guiding stent implantation. ⁴⁴ , ⁴⁵ Prospective, multicenter randomized controlled trials are required to investigate the influence of μQFR on long‐term outcomes. Furthermore, other clinical information was not available for this analysis, such as plaque composition, vessel volume, and left ventricular mass, which have been linked to the interaction between sex and physiological indices.

Conclusions

In conclusion, women tend to exhibit higher FFR and μQFR values when adjusted for lesion severity and other confounders. Nevertheless, μQFR has comparable diagnostic performance in both female and male patients, providing additional diagnostic value in identifying functionally significant lesions beyond angiography alone. This finding further underscores the reliability of this adenosine‐free computational approach, especially in a sex‐specific context.

Sources of Funding

This work was supported by the Jiangsu Provincial Key Research and Development Program (BE2022852), Jiangsu Provincial Medical Key Discipline (ZDXK202207), and the National Natural Science Foundation of China (82070295).

Disclosures

Dr Shengxian Tu received research grants from Pulse Medical Imaging Technology. The remaining authors have no disclosures to report.

Supporting information

Tables S1–S2

Figures S1–S6