Clinical Implications of Fractional Flow Reserve Measured Immediately after Percutaneous Coronary Intervention

Purpose The purpose of the present study was to �nd the independent predictors of Fractional Flow Reserve (FFR) measured immediately after percutaneous coronary intervention with drug eluting stent implantation (post-PCI FFR) and investigate if applying vessel-specic post-PCI FFR cut-off values to predict target vessel failure (TVF), a composite of cardiac death (CD), non-fatal myocardial infarction (MI) and target vessel revascularisation (TVR) or a composite of CD and MI ameliorated its predictive power.


Introduction
Fractional Flow Reserve (FFR) has become the gold standard in revascularization decisions based on extensive validation, randomized clinical studies, and registries [1][2][3][4][5][6][7][8][9][10][11].FFR measures the relative epicardial resistance to ow and can be applied not only before percutaneous coronary intervention (PCI), but also after it.Following PCI with an angiographically satisfactory result, one expects the resolution of resistance to ow and cessation of symptoms.Recent studies have however shown that in a signi cant proportion of cases, residual gradients remain even after satisfactory angiographic result [12][13][14][15] leading to low post-PCI FFR.Numerous analyses have provided evidence that FFR measured immediately after PCI carries prognostic implications and lower post-PCI FFR values are linked to poor clinical outcome [16][17][18].It was also noted that post-PCI FFR values in LAD are signi cantly lower compared to those measured in non-LAD vessels [18].Post-PCI FFR was shown to be moderately related to major adverse cardiac events (MACE), however the best cut-off to predict untoward events has been a matter of debate.
This may be related to the difference in the studied populations, the pattern of disease (focal or diffuse) but potentially also to the fact that no distinction was made to the different vessels.On the other hand, it has not been rmly established which parameters in uenced post-PCI FFR.Hence, we investigated the independent predictors of post-PCI FFR and sought to determine if applying a vessel-speci c cut-off of post-PCI FFR ameliorated its predictive power.

Patients
We included all patients who underwent FFR-guided PCI exclusively with drug eluting stent (DES) implantation and post-PCI FFR measurement at our tertiary care center with at least one-year follow-up.
We excluded only patients who had undergone heart transplantation before PCI and patients in whom the PCI was performed in bypass grafts.Patients could have chronic coronary syndrome (CCS) or acute coronary syndrome (ACS), heart failure or arrhythmias as an indication to coronary angiography.
The technique of PCI was left entirely to the discretion of the treating physician.FFR was measured using commercially available pressure wires (St.Jude Medical, St. Paul, Minnesota, now Abbott).Hyperemia was achieved mostly using intracoronary boluses of adenosine (200 µg in the left and 100 µg in the right coronary artery) or intravenous infusion of adenosine at the standard dose of 140 µg/kg/min.All patients received guideline-directed medical therapy following PCI.
Of note, all the post-PCI FFR values analyzed were the nal ones, after which no further intervention was performed.
The data that support the ndings of this study are available from the corresponding author upon reasonable request.The subjects gave informed consent.This study was performed in line with the principles of the Declaration of Helsinki.

Endpoints
The primary endpoint of the study was target vessel failure (TVF) de ned as a composite of cardiac death (CD), non-fatal myocardial infarction (MI), and target vessel revascularization (TVR).The secondary endpoint of the study was the composite of CD and MI.All deaths were considered cardiac unless a clear non-cardiac cause was present.Periprocedural myocardial infarctions related to the index procedure were not counted.Both urgent and elective TVR's were considered endpoints.Of note, no recoronary angiography was planned.In case a patient had more than one vessel included, CD was counted as related to all vessels; if a non-fatal myocardial infarction could not be clearly related to a culprit artery, all vessels of the patient in the registry were considered infarct related.Follow-up information was collected from the institutional database, telephone inquiries and from the database of other centers.In case no follow-up information was available from the above sources, the patient's unique insurance number was checked in the national death records.Endpoints were analyzed both on a vessel and on a patient level, for the latter, each patient's lowest post-PCI FFR value was used in case they had more than one vessel with post-PCI FFR measurement included.

Statistical analysis
For the multivariable modelling, linear regression was used in case of post-PCI FFR and Cox proportional hazards model was used for survival data.In each case, continuous predictors were expanded with splines rst, to check if there is any signi cant nonlinearity [19].Given the lack of it, linear models were used subsequently.In all cases, multiple imputation with predictive mean matching [20] was used to impute missing data.The Hubert-White method was used for robust covariance matrix estimation in all regression; in case of vessel-level analysis, clusters were set to patients to account for the correlation of measurements coming from the same patient.Final models are given by presenting beta coe cients (linear regression) or hazard ratios as exponentiated coe cients (Cox-regression).Results are shown with 95% con dence intervals.Variables were considered signi cant if they had p ≤ 0.05.Calculations were carried out using the R statistical software package, version 4.2.2.

Results
Between March 17, 2009, and January 19, 2021, 1024 patients had FFR-guided DES implantation at our center.Of those, 534 did not have all treated vessels assessed by post-PCI FFR measurement.After excluding those after heart transplantation and graft PCI, 434 patients and 500 arteries were included in our analysis.The owchart of patients is shown in Fig. 1.
The median age of the patient population was 65 years [IQR: 57-71], 69% were male, 49% had diabetes mellitus, of those, 27% were treated with insulin.44% of the patient population had previous PCI and 2% had coronary artery bypass surgery.The characteristics of the patients are summarized in Table 1.Of the 434 patients, 46 had 2 and 10 had 3 vessels included in our analysis.Of the 500 vessels, 333 (67%) were left anterior descending (LAD), 67 (13%) left circum ex (LCx) and 100 (20%) right coronary arteries (RCA).In 77 cases, the indication of PCI was ACS, in 62, the treated lesion was in-stent restenosis.
The distribution of pre-PCI FFR values is shown in Supplemental Fig. p < 0.001).

Predictors of post-PCI FFR
By multivariable regression analysis, LAD location (p < 0.001), male gender (p < 0.001), smaller stent diameter (p = 0.006) and lower pre-PCI FFR (p = 0.003) proved to be signi cant predictors of lower post-PCI FFR.Of note, post-PCI FFR measured in non-culprit vessels in acute coronary syndrome (ACS) and in chronic coronary syndrome (CCS) did not differ signi cantly, neither had the in-stent restenosis vs. de novo lesion category or diabetes mellitus any signi cant in uence on post-PCI FFR.This is shown in Fig. 3.
By univariate ROC analysis, a post-PCI FFR of 0.83 was found to be the best cut-off to predict TVF according to the Youden-index with a sensitivity of 45%, speci city of 86% and an area under the curve (AUC) of 0.70.
Given that the median post-PCI FFR values in the LAD were 0.07 units lower than in non-LAD vessels, we also studied these two territories separately.Two-thirds of the studied vessels were LAD, in these, the best post-PCI FFR, by univariate ROC analysis, was also 0.83 according to the Youden-index (sensitivity 60%, speci city 82%, AUC 0.75), whereas in the non-LAD vessels, accounting for one-third of the cases, the best post-PCI FFR to predict TVF was 0.91 according to the Youden-index (sensitivity 65%, speci city 61%, AUC 0.59).
The secondary endpoint of our study was the composite of CD and MI.Its independent predictors were post-PCI FFR (p < 0.001), stent length (p < 0.001), non-LAD location (p = 0.0026), and diabetes mellitus (p = 0.015), see Supplemental Fig. 3.The relationship between CD and/or MI and post-PCI FFR as a continuous variable is shown in Supplemental Fig. 4.
By univariate ROC analysis, a post-PCI FFR of 0.83 was found to be the best cut-off to predict the composite of CD and MI according to the Youden-index with a sensitivity of 54%, speci city of 86% and an AUC of 0.75.In LAD vessels, the best cut-off to predict CD and / or MI according to the Youden-index was also 0.83 with a sensitivity of 72%, speci city of 82%, and an AUC of 0.84.In non-LAD vessels, the best cut-off to predict CD and / or MI according to the Youden-index was 0.91, with a sensitivity of 70%, speci city of 61%, AUC 0.60.

Patient-level analysis
On a patient level, post-PCI FFR (p < 0.001) and stent length (p = 0.00384) were found to be independent predictors of TVF.The frequency of TVF in patients with the single lowest post-PCI FFR strata of ≤ 0.80, 0.81-0.85,0.86-0.90,0.91-0.95and > 0.95 were 47.6%, 22.2%, 11.7%, 12.5% and 2.9%, respectively.The relationship between TVF and single lowest post-PCI FFR as a continuous variable is shown in Supplemental Fig. 5.Note that this gure uses the whole follow-up for each patient, i.e., it neglects the inter-patient differences in follow-up time.
By univariate ROC analysis, a single lowest post-PCI FFR of 0.83 was found to be the best cut-off to predict TVF according to the Youden-index with a sensitivity of 51%, speci city of 85% and an AUC of 0.73.
The independent predictors of the composite of CD and MI on a patient level were post-PCI FFR (p < 0.001), stent length (p < 0.001) and diabetes mellitus (p = 0.03377).By univariate ROC analysis, a single lowest post-PCI FFR of 0.83 was found to be the best cut-off according to the Youden-index to predict CD and / or MI with a sensitivity of 62%, speci city of 85% and an AUC of 0.81 on a patient-level.

Discussion
In the present analysis, we report on our center's long-term experience of post-PCI FFR.Our salient ndings are as follows.Post-PCI FFR is in uenced by the location of the lesion (LAD vs non-LAD), gender, stent (vessel) diameter, and to a minor degree, albeit statistically signi cantly, pre-PCI FFR.Second, on a vessel-level, only post-PCI FFR, stent length and diabetes mellitus predicted TVF.We also found that different cut-off values should be applied in LAD vs non-LAD vessels.Last, importantly, post-PCI FFR, along with diabetes mellitus, and stent length independently predicted "hard" end points (CD and MI).
Numerous studies have reported on the prognostic value of post-PCI FFR to predict vessel-related MACE from the bare metal stent (BMS) [21] as well as the DES era [17,18,22].Some have reported con icting ndings on its predictive value [12,[23][24][25].A study-level meta-regression analysis by Rimac et al. [26] found an inverse relationship between post-PCI FFR and MACE, whereas a large, patient-level metaanalysis of Johnson et al. [27] showed an inverse relationship between the post-PCI FFR and the rate of untoward events.Of note, both studies included patients with BMS as well as DES implants and BMS are known to lead to higher TVR and therefore the conclusions of these analyses may be less pertinent to our present-day practice of implanting only DES.Previous reports showed con icting results as to the relationship between post-PCI FFR and "hard" end points: some found that a low post-PCI FFR was associated with higher rate of cardiac death and/or MI [16], whereas others did not [12,28].
FFR has gained widespread acceptance in the indication of PCI, and it is generally assumed that an "FFRguided" PCI -usually meaning the measurement of FFR at a distal spot in the artery of interest before PCI and performing the intervention angiography-guided -will result in the elimination of epicardial resistance to ow, myocardial ischemia, and anginal symptoms.An increasing number of reports attest to the fact that this is far from being the reality in a sizable proportion of cases, also re ected by the high percentage of patients returning for recoronary angiography because of persistent symptoms after technically successful PCI.This may be caused by microvascular angina or extracardiac causes as well as by a failure to relieve signi cant epicardial resistance to ow.
Post-PCI FFR measurement quanti es the remaining epicardial reduction of maximally achievable ow after revascularization which may be related to diffuse disease or focal, angiographically mild-looking lesions of the untreated segments, or stent-related issues (gross underexpansion, potentially malapposition, geographical miss, edge dissections, tissue protrusion, etc.).Technically, drifts can in uence the measured FFR value after PCI and a contralateral chronic total occlusion with retrograde lling from the index vessel can in uence post-PCI FFR to a variable degree [29].These different mechanisms could be elucidated by hyperemic pull-back recording at the end of the procedure which is seldom performed.Likewise, pre-PCI FFR is usually applied as a "snapshot", re ecting the total amount of epicardial resistance proximal to the sensor of the pressure wire, and in most cases no attempt is made to elucidate the relative contribution of the different segments which could potentially in uence the procedural plan.The concept of 2-dimensional FFR has been put forward: hyperemic pullback recording not only shows the severity but also the distribution of plaques [30].
Does it matter whether a post-PCI FFR of 0.83 is related to diffuse or focal disease?Emerging evidence suggests that it does.A higher pressure drop across a plaque may be related to plaque fatigue and rupture leading to ischemic events [31].However, as stated above, this is usually not clari ed, and this may be one of several reasons why the predictive value of post-PCI FFR and its cut off to predict TVF has been different among published reports.
The report by Collet et al. [32] attest to that angiography is inaccurate in assessing the physiological pattern of coronary artery disease (focal or diffuse).Applying the PPG index (Pullback Pressure Gradient), these authors showed that 36% of the cases were reclassi ed.
All the above-mentioned factors must be considered when the meaning and actual value of post-PCI FFR is contemplated.Our analysis provides one further piece of evidence: the importance of the vessel.We demonstrated that even though pre-PCI FFR values did not differ between LAD and non-LAD vessels, post-PCI FFR is signi cantly lower in the LAD, and we found that post-PCI FFR cut-off values to predict TVF as well as a composite of CD and MI were different in LAD vs. non-LAD vessels.The explanation behind this is speculative yet, and includes the fact that LAD usually perfuses a larger myocardial mass making the remaining disease functionally more relevant and emerging evidence supports that hydrostatic pressure differences also play a role [33].The latter is related to the fact that the middle and distal segments of the LAD in a supine patient run higher than the aortic root where the catheter measures the aortic pressure, therefore the pressure measured by the sensor of the PressureWire is lower [33].Before PCI, most of the hyperemic gradient stems from the disease and the hydrostatic pressure difference is negligible, whereas after successful PCI, the gradient caused by signi cant stenoses is eliminated making the hydrostatic pressure difference more relevant.
In all, post-PCI FFR can be viewed as a safety net at the end of an angiographically successful PCI indicating a suboptimal functional result which has implications not only for TVR but also for hard endpoints like CD and MI independently from other clinical or angiographic variables.Elucidating the cause of a "suboptimal" functional result by hyperemic pullback recording [34] or intravascular imaging [35] may allow for improving not only the post-intervention physiology, but also the clinical outcome.

Study limitations
Several limitations of the present analysis must be noted.First, it is a single-center cohort of a limited number of patients.Including patients who received only DES reduced the number of subjects and vessels but made our cohort more homogeneous.Some received rst-generation whereas the majority had second, and third generation DES implanted, we did not evaluate the effect of this difference.The cause of death could not be rmly established in all cases, these were considered cardiac deaths.TVR was sometimes performed without FFR measurement, only based on angiography.Events were not adjudicated by an independent Clinical Event Committee.We could not study the anginal status of our patients, hence could not establish the relationship between post-PCI FFR and quality of life.No systematic pull-back recording was performed; therefore, we could not establish the role of diffuse vs. focal nature of the remaining disease.However, in all cases, we checked for drifts and in case a drift of > 0.03 units was found, we corrected the measured value or repeated the measurement after re-equalizing.
Patients and their treating physicians were aware of the measured post-PCI FFR; however, all interventions were considered angiographically successful and only a small minority of vessels had a post-PCI FFR of ≤ 0.80.

Conclusions
In conclusion, we found that post-PCI FFR was an independent predictor of TVF as well as of the composite of CD and MI in an unselected patient population undergoing FFR-guided DES implantation.Moreover, LAD location was found to be a predictor of a lower post-PCI FFR.Our data suggest that no uniform target post-PCI FFR value exists; in order to improve the predictive power of post-PCI FFR, different cut-off values should be applied in LAD as opposed to non-LAD vessels.

Declarations
Funding: The declare that no funds, grants, or other support were received during the preparation of this manuscript.

Flowchart of patients
Legend not included with this version.
Legend not included with this version.
Legend not included with this version.