Introduction
Second-generation drug eluting stents (DES) have effectively inhibited neointimal hyperplasia and hence substantially reduced the incidence of in-stent restenosis (ISR) [1, 2]. Nonetheless, recurrent ISR still occurs in late or very late phases after DES implantation in up to 7–10% of cases [3, 4], and interventional treatment of DES-ISR remains challenging. Several studies have demonstrated the efficacy of drug-coated balloons (DCB) for the treatment of ISR [5–9]. DCBs transfer an antiproliferative drug, in most cases paclitaxel, onto the vessel wall during the short time of balloon inflation, resulting in efficient inhibition of smooth muscle cell proliferation and neointimal hyperplasia [10, 11], thus circumventing the need to implant additional metallic layers in the vessel, whilst achieving a comparable net therapeutic performance to DES [9, 12, 13].
Quantitative flow ratio (QFR) is an innovative implement of computational physiology based on three-dimensional (3D) reconstruction of anatomy and hemodynamic simulation, which has shown excellent correlation and agreement with invasive wire-based fractional flow reserve (FFR) [14–16]. Previous studies have demonstrated that QFR can be used to evaluate patients with ISR [17, 18], and post-procedural QFR had the ability to predict future clinical vessel-oriented composite endpoints (VOCE) [19]. On this basis, several studies have recently explored the ability of conventional 3D-based QFR to predict the clinical outcome of ISR treated with DCB [20, 21]. Our current study appraised the prognostic value of the new-generation QFR, based on Murray bifurcation fractal law (μQFR), aided by artificial intelligence [22], to predict the incidence of vessel-oriented composite endpoint after DCB angioplasty, using data from a previous DCB-ISR trial.
Methods
Study design
This was a post-hoc analysis of a prospective, multicenter, randomized controlled clinical trial comparing the efficacy of two different kinds of DCBs, Shenqi (Shenqi Medical, Shanghai, China) or Sequent Please (B. Braun Melsungen AG, Melsungen, Germany), for the treatment of first-occurrence DES-ISR between December 2016 and January 2018 [11]. Patients were excluded from the current analysis if the thrombolysis in myocardial infarction (TIMI) grade flow was < 3 at baseline or after DCB angioplasty, the angiography recordings were deemed of insufficient quality for μQFR analysis, or nonstandard angiographic DICOM imaging encoding failed to analyze µQFR (Fig. 1).
The study complied with the principles of good clinical practice and with the Declaration of Helsinki for investigation in human beings. The study protocol was approved by the Institutional Review Board and Ethics Committee at each participating center, and all patients provided written informed consent before receiving DCB treatment.
Angiographic follow-up and endpoint definition
Patients were routinely scheduled an angiographic follow-up of 9 ± 1 months, even though some angiograms were performed earlier or later if clinically indicated. The endpoint of the study was VOCE at 1-year follow-up, defined as a composite of cardiac death, target vessel-related myocardial infarction, and ischemia-driven target vessel revascularization (TVR). A residual lesion was defined as diameter stenosis (DS) ≥ 50% in vessels ≥ 1.5 mm by visual assessment, not located within the in-segment (a ISR lesion complete treated segment + 5 mm adjacent margins) treatment by DCB angioplasty.
Quantitative coronary angiography (QCA)
Quantitative measurements of coronary angiograms were analyzed offline using QAngio XA 7.3 (Medis Medical Imaging System BV, Leiden, the Netherlands) by two well-trained observers, blind to patients’ information, at the central angiographic core laboratory, according to standard methodology [11]. Appropriate angiographic projections were selected to avoid excessive vessel foreshortening and overlap. Reference and minimal lumen diameters, percentage of DS, and lesion length were measured before and immediately after the procedure, and at follow-up. Restenosis patterns were assessed by Mehran classification [23].
μQFR computation
Two experienced and qualified analysts performed μQFR analysis offline, using Angioplus Galley software (Pulse Medical Imaging Technology, Shanghai, China). A single angiographic projection displaying the target vessel from the ostium to the distal segment, with the corresponding side branches and encompassing the whole target lesion, was selected as meeting the requirements of μQFR analysis. The frame with optimal definition of the target lesion was chosen as the key frame for the measurements. Lumen contour and coronary flow velocity were automatically delineated and computed, respectively, aided by artificial intelligence. In cases of inaccurate lumen delineation, minor manual editing was allowed by adding additional points markers along the lumen contour.The reference diameter was calculated along the target vessel according to the Murray fractal law, resulting in reference step-down at bifurcations. Then the physiological indexes of the main vessel and side branch were subsequently derived. In the presence of an eccentric lesion, 3D QCA analysis was performed after selecting a second projection of the target vessel, > 25° apart from the main projection. 3D μQFR was computed from the reconstruction used for 3D-QCA. A paradigmatic example is shown in Figure 2.
Statistical analysis
Continuous variables were expressed as mean ± standard deviation or median (interquartile range) according to the data distribution determined by the Kolmogorov-Smirnov test, and they were compared using Student’s t-test or the Mann–Whitney U test, as appropriate. Categorical variables were described as counts (percentage) and compared using Pearson’s c2 test or Fisher’s exact test, as appropriate. Receiver operator characteristic (ROC) curve analysis was performed to determine the optimal post-procedural μQFR cut-off value to predict VOCE, as determined by the Youden index. Kaplan–Meier survival analysis was performed, comparing the groups defined by the μQFR cut-off with the log-rank test. Multivariate Cox regression analysis was performed to search for independent predictors of VOCE. Proportional-hazards assumption was tested on the basis of Schoenfeld residuals. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated. All statistical analyses were performed using SPSS version 22.0.0 (IBM Corporation, Armonk, New York, USA) and MedCalc version 14.12 (MedCalc Software, Ostend, Belgium). A two-sided p-value < 0.05 was considered as statistically significant.
Results
The study comprised 216 lesions in 216 patients. Forty-seven patients were excluded from the current study due to TIMI grade flow < 3 at baseline or after DCB angioplasty (8 patients), or due to insufficient quality (12 patients) or nonstandard angiographic DICOM imaging encoding (27 patients) for μQFR analysis, thus resulting in 169 patients being successfully analyzed: 20 patients with VOCE and 149 patients without VOCE (Fig. 1). One-year clinical follow-up was completed in all eligible patients with a median follow-up period of 353 days (340–371 days).
Baseline characteristics
Baseline clinical and procedural characteristics of 169 patients and lesions finally enrolled in the study are presented in Tables 1 and 2. VOCEs occurred in 20 patients at 1-year follow-up. There were no significant differences in age, gender, body mass index, hypertension, hyperlipidemia, current smoking, medical history, clinical presentation, left ventricular ejection fraction, and number of diseased arteries between patients with and without VOCE. Diabetes mellitus and family history of coronary heart disease were significantly more prevalent among patients with VOCE than among event-free patients (Table 1).
The distributions of target vessel, ostial lesion, bifurcation lesion, Mehran restenosis pattern, and type of DCB applied did not differ between groups. Baseline QCA parameters showed smaller reference vessel diameter (2.37 ± 0.40 vs. 2.60 ± 0.42, p = 0.02) and minimum lumen diameter (0.74 ± 0.35 vs. 0.95 ± 0.39, p = 0.02) in patients with VOCE than in event-free patients, while the mean lesion length and diameter stenosis at baseline and immediately after the procedure were similar in both groups. Procedural variables were similar between patients with or without VOCE. Patients with VOCE showed lower μQFR values than event-free patients, both at baseline (0.50 [0.37–0.75] vs. 0.77 [0.63–0.85], p = 0.001) and post-procedure (0.88 [0.79–0.94] vs. 0.96 [0.91–0.98], p < 0.001), whilst μQFR improvement, defined as difference values between post-procedure and baseline, was larger in the VOCE group (0.30 [0.18–0.41] vs. 0.18 [0.10–0.34], p = 0.01). The proportion of patients with post-procedural μQFR ≤ 0.80 was larger in the VOCE group than in the event-free group (6 [30.00%] vs. 6 [4.03%], p < 0.001) (Table 2).
Variables |
VOCE (n = 20) |
Non-VOCE (n = 149) |
P |
Age [years] |
59.90 ± 11.57 |
62.88 ± 9.58 |
0.20 |
Female gender |
6 (30.00%) |
35 (23.49%) |
0.58 |
BMI [kg/m2] |
25.12 ± 3.15 |
25.60 ± 3.31 |
0.54 |
CAD risk factors: |
|||
Diabetes mellitus |
13 (65.00%) |
56 (37.58%) |
0.03 |
Hypertension |
15 (75.00%) |
109 (73.15%) |
0.86 |
Hyperlipidemia |
9 (45.00%) |
49 (32.88%) |
0.28 |
Current smoker |
6 (30.00%) |
33 (22.15%) |
0.59 |
Medical history: |
|||
Previous MI |
9 (45.00%) |
64 (42.95%) |
0.86 |
Previous PCI |
20 (100.00%) |
149 (100.00%) |
> 0.999 |
Family history |
6 (30.00%) |
14 (9.39%) |
0.03 |
Clinical presentation: |
0.25 |
||
Silent ischemia |
0 (0.00%) |
15 (10.07%) |
|
Stable angina |
1 (5.00%) |
20 (13.42%) |
|
Unstable angina |
19 (95.00%) |
114 (76.51%) |
|
LVEF [%] |
61.40 ± 7.67 |
60.56 ± 7.23 |
0.63 |
No. of diseased arteries: |
0.86 |
||
1 |
6 (30.00%) |
55 (36.91%) |
|
2 |
10 (50.00%) |
68 (45.64%) |
|
3 |
4 (20.00%) |
26 (17.45%) |
Variables |
VOCE (n = 20) |
Non-VOCE (n = 149) |
P |
Target vessel: |
0.81 |
||
RCA |
6 (30.00%) |
57 (38.26%) |
|
LAD |
11 (55.00%) |
70 (46.98%) |
|
LCX |
3 (15.00%) |
22 (14.76%) |
|
Ostial lesion |
2 (10.00%) |
4 (2.68%) |
0.15 |
Bifurcation lesion |
5 (25.00%) |
47 (31.54%) |
0.55 |
Restenosis pattern: |
0.41 |
||
Mehran I |
4 (20.00%) |
50 (33.56%) |
|
Mehran II |
13 (65.00%) |
73 (48.99%) |
|
Mehran III |
3 (15.00%) |
26 (17.45%) |
|
Shenqi DCB |
11 (55.00%) |
78 (52.35%) |
0.82 |
QCA parameters: |
|||
Lesion length [mm] |
15.67 ± 5.61 |
14.99 ± 8.39 |
0.73 |
RVD [mm] |
2.37 ± 0.40 |
2.60 ± 0.42 |
0.02 |
MLD [mm] |
0.74 ± 0.35 |
0.95 ± 0.39 |
0.02 |
DS [%] |
67.95 ± 12.10 |
63.20 ± 12.53 |
0.11 |
DS after DCB [%] |
17.58 ± 15.17 |
21.11 ± 10.10 |
0.32 |
Procedural data: |
|||
Predilation |
20 (100.00%) |
149 (100.00%) |
> 0.999 |
Cutting balloon |
5 (25.00%) |
38 (25.50%) |
0.96 |
DCB diameter [mm] |
2.87 (2.62–3.50) |
3.00 (2.75–3.50) |
0.48 |
DCB length [mm] |
20.00 (20.00–25.50) |
20.00 (17.00–26.00) |
0.91 |
DCB pressure [atm] |
9 (8–9) |
9 (8–10) |
0.60 |
DCB Inflation time [s] |
60 (60–60) |
60 (60–60) |
> 0.999 |
No. of DCB used > 1 |
1 (5.00%) |
3 (2.01%) |
0.40 |
μQFR measurements: |
|||
Baseline μQFR |
0.50 (0.37–0.75) |
0.77 (0.63–0.85) |
0.001 |
Post-procedural μQFR |
0.88 (0.79–0.94) |
0.96 (0.91–0.98) |
< 0.001 |
Post-procedural μQFR ≤ 0.80 |
6 (30.00%) |
6 (4.03%) |
< 0.001 |
μQFR improvement |
0.30 (0.18–0.41) |
0.18 (0.10–0.34) |
0.01 |
Residual lesion after DCB treatment |
8 (40.00%) |
21 (14.09%) |
0.01 |
Diameter difference DCB — RVD |
0.61 (0.43–0.82) |
0.43 (0.29–0.59) |
0.004 |
Length difference DCB — lesion |
6.36 (2.62–9.32) |
7.26 (2.65–11.88) |
0.59 |
Patients with residual lesion after DCB-ISR treatment were more likely to develop VOCE than patients without residual lesion (8 [40.00%] vs. 21 [14.09%], p = 0.01). Moreover, patients with greater differences between DCB diameter and reference vessel diameter (RVD) seemed more prone to develop VOCE (0.61 [0.43–0.82] vs. 0.43 [0.29–0.59], p = 0.004). The length mismatch between DCB and lesion was similar between groups (Table 2).
Definition of potential post-procedural cut-off value
Receiver-operating characteristic curve analysis identified post-procedural μQFR ≤ 0.89 as the optimal cut-off value to predict the occurrence of VOCE, with sensitivity 55% and specificity 74% (AUC: 0.74; 95% CI: 0.67–0.80; p < 0.001). Nevertheless, there was no significant predictive value of post-procedural percent diameter stenosis (%DS) for VOCE (AUC: 0.61; 95% CI: 0.53–0.68; p = 0.18). The ROC curve for μQFR improvement also showed moderate predictive value for VOCE, with the best cut-off value > 0.20, and with sensitivity of 75% and specificity of 58% (AUC: 0.67; 95% CI: 0.60–0.74; p = 0.001) (Fig. 3).
p = 0.18); B. Post-procedural μQFR (AUC 0.74, 95% CI 0.67–0.80; p < 0.001); C. Post-procedural μQFR improvement (AUC 0.67, 95% CI 0.60–0.74, p = 0.001); AUC — area under curve; CI — confidence interval.
Clinical outcomes
Clinical outcomes stratified by post-procedural μQFR are shown in Table 3. Patients who achieved μQFR > 0.89 after DCB treatment had significantly fewer VOCEs than those with μQFR ≤ 0.89 (6.77% vs. 30.55%, p < 0.001), mainly attributed to a higher incidence rate of TVR (6.77% vs. 27.78%, p = 0.001). Two patients presented with more than one event; both developed target vessel myocardial infarction followed by TVR. One patient with post-procedural μQFR ≤ 0.89 died of cardiac death, while the other, with post-procedural μQFR > 0.89, did not. Kaplan-Meier curves also confirmed that post-procedural μQFR ≤ 0.89 had a remarkably higher incidence rate of VOCE (Fig. 4).
Variables |
μQFR ≤ 0.89 (n = 36) |
μQFR > 0.89 (n = 133) |
P |
Cardiac death |
1 (2.78%) |
0 (0.00%) |
0.21 |
Target vessel MI |
2 (5.56%) |
0 (0.00%) |
0.04 |
Target vessel revascularization |
10 (27.78%) |
9 (6.77%) |
0.001 |
Target lesion revascularization |
7 (19.44%) |
8 (6.01%) |
0.02 |
VOCE |
11 (30.55%) |
9 (6.77%) |
< 0.001 |
After correcting for potential confounders (diabetes mellitus, family history, ostial lesion, lesion length, residual lesion, differences in diameter DCB-RVD), post-procedural μQFR ≤ 0.89 remained associated with a 6-fold increase in the risk of VOCE (adjusted HR: 5.94; 95% CI: 2.33–15.09; p < 0.001; Table 4, Model a1). μQFR improvement > 0.20 was also associated with suboptimal clinical result at 1-year follow-up (HR: 3.75; 95% CI: 1.31––10.68, p = 0.01; Table 4, Model b1). Considered as a continuous variable, post-procedural μQFR was associated with a lower incidence of VOCE in the multivariate analysis (HR: 0.34; 95% CI: 0.23–0.51; p < 0.001), whereas μQFR improvement (each 0.10 increase) was associated with a higher incidence of VOCE (HR: 1.31; 95% CI: 1.04–1.66; p = 0.02; Table 4, Model a2 and Model b2).
HR (95% CI) |
P |
|
Model a1 |
||
Post-procedural μQFR |
5.94 (2.33–15.09) |
< 0.001 |
Diabetes mellitus |
2.64 (1.03–6.78) |
0.04 |
Difference of DCB diameter and RVD (per 0.10-mm increase) |
1.34 (1.10–1.62) |
0.003 |
Model a2 |
||
Post-procedural μQFR (per 0.10-mm increase) |
0.34 (0.23–0.51) |
< 0.001 |
Diabetes mellitus |
1.61 (0.55–4.66) |
0.16 |
Difference of DCB diameter and RVD (per 0.10-mm increase) |
1.25 (1.03–1.50) |
0.02 |
Model b1 |
||
μQFR improvement |
3.75 (1.31–10.68) |
0.01 |
Diabetes mellitus |
3.15 (1.23–8.05) |
0.02 |
Residual lesion after DCB treatment |
3.12 (1.21–8.03) |
0.02 |
Difference of DCB diameter and RVD (per 0.10-mm increase) |
1.33 (1.11–1.59) |
0.002 |
Model b2 |
||
μQFR improvement (per 0.10-mm increase) |
1.31 (1.04–1.66) |
0.02 |
Diabetes mellitus |
3.10 (1.21–7.92) |
0.02 |
Residual lesion after DCB treatment |
3.75 (1.47–9.56) |
0.01 |
Difference of DCB diameter and RVD (per 0.10-mm increase) |
1.32 (1.11–1.57) |
0.001 |
Discussion
To the best of our knowledge, this post-hoc study of a previous DCB-ISR trial investigated the prognostic value of Murray law-based QFR, empowered by artificial intelligence, after DCB- -ISR treatment for the first time. A low μQFR after DCB-ISR angioplasty was an independent predictor of clinical adverse events at 1-year follow-up. Post-procedural μQFR showed an optimal predictive value for the occurrence of VOCE, and a moderate predictive value was also observed for μQFR improvement.
Coronary interventions have classically relied on the assessment of anatomic stenosis observed in angiography, even though the stenosis severity scarcely correlates with the physiological significance as a flow-limiting lesion [24, 25] and has modest predictive value for future clinical events [26]. Decision-making based on physiology has consistently proven its superiority over purely angiographic guidance in most clinical scenarios of stable coronary heart disease, thus being endorsed in international guidelines for clinical practice as the highest standard of care [27, 28]. The QFR is a novel tool to derive physiology parameters in the coronary arteries, based on 3D angiographic reconstruction and computerized hemodynamic simulation. This emerging method can efficiently identify functionally significant lesions, whilst overcoming the drawbacks of traditional wire-based invasive physiology [14, 15]. Previous studies have demonstrated that post-procedural QFR is significantly associated with clinical outcomes after percutaneous coronary intervention (PCI) [19, 29]. In the present study, we applied the most advanced and refined version of QFR, called Murray-law based QFR (µQFR), which is characterized by calculating the reference vessel diameter according to fractal geometry. This adds extra accuracy to the estimation, especially for challenging bifurcation lesions, thus achieving excellent agreement with fractional flow reserve [22].
Our study proved an inverse relationship between post-procedural μQFR and adverse clinical events after treatment of DES-ISR with DCB, an association that could not however be verified for post-procedural %DS. This finding is in line with a large corpus of evidence consistently proving the superiority of morphofunctional or computational methods over purely morphologic approaches to assess severity and prognosis [14, 16, 30]. Angiographic residual %DS has been regularly used for the assessment of procedural success in routine clinical practice [31], even though its prognostic value for future clinical events might be disputable [32]. Among morphofunctional parameters, post-procedural μQFR showed the best prognostic value for VOCE at 1-year follow-up, superior to other parameters like μQFR improvement. These findings clearly point out the paramount importance of functional residual lesion after treatment, in this case assessed by means of post-procedural μQFR, and hence indirectly suggest the need to optimize the functional (rather than the angiographic) result after PCI, in line with multiple studies.
Besides functional residual stenosis, i.e., post-procedural μQFR, and the subsequent importance of PCI optimization, an adequate DCB sizing seems also to play a relevant role for future events [33]. Obviously, an insufficient DCB balloon sizing might result in incomplete surface contact, and ultimately to inappropriate drug transfer onto the vessel wall. Nonetheless, the current study showed that the larger the oversizing of DCB diameter in regard to RVD, the higher the risk of developing VOCE at 1-year follow-up. An oversized DCB might create edge dissections of inflicting additional insult to the vessel, which might interfere with the optimal healing process, thus creating the substrate for future events. Intracoronary imaging is currently the best ancillary tool available for both accurate sizing and targeted PCI optimization, so it might be instrumental to optimize these two variables identified by our study as predictors for clinical events after treatment of DES-ISR with a DCB, namely, post-procedural μQFR and DCB diameter mismatch. Evidence about the clinical correlates of a refined interplay between physiology and intracoronary imaging to optimize PCI results is becoming increasingly strong [34].
As in many other publications, diabetes mellitus was also an independent risk factor for adverse events in the current study, probably due to a multifactorial etiology. Diabetes has been associated with more intense plaque progression and has been shown to elicit an exaggerated neointimal hyperplasia reaction [35, 36]. Like other clinical scenarios, the percutaneous treatment of ISR in diabetes patients is particularly challenging and deserves careful attention.
Limitations of the study
There are several limitations in the current study. First, this study is a post-hoc analysis of a previous DCB-ISR trial, which was not initially designed to investigate the prognostic value of μQFR on ISR. Therefore, some cases did not meet the acquisition requirements for μQFR computation. Nonetheless, this attrition of the initial sample size occurred at random and was therefore unlikely to result in selection bias. All cases were analyzed offline, which inevitably affects the precise analysis. Second, the cut-off value of post-procedural μQFR to predict vessel-oriented composite events was not exclusive, varied widely due to multiple factors, including the observed population, incidence of clinical events, lesion and procedural characteristics, etc., and should be validated in a future large randomized controlled trial. Lastly, intravascular imaging was not mandatory in the previous trial; conversely, intravascular morphology information would be beneficial to understand the underlying mechanics of low μQFR values and adverse events.
Conclusions
Post-procedural μQFR after treatment of ISR with DCB was inversely associated with the occurrence of subsequent adverse clinical events and may be considered as a promising predictor.
Acknowledgments
The authors would like to thank the dedicated efforts from the clinical research collaborators who participated in the previous DCB-ISR trial.