The correlation between CT and duplex evaluation of autogenous vein bypass grafts and their relationship to failure

Abstract

Objective

Duplex ultrasound(DUS) for vein bypass graft(VBG) surveillance is confounded by technical and physiologic factors that that reduce the sensitivity for detecting impending graft failure. In contrast, three-dimensional computed tomographic angiography(CTA) offers high fidelity anatomic characterization of VBGs but its utility in detecting at risk grafts is unknown. The current study sought to analyze the correlation between DUS and CTA for detection of vein graft stenosis and evaluate the relationship of the observed abnormalities to VBG failure.

Methods

Consecutive lower extremity VBG patients underwent surveillance with concurrent DUS and CTA at 1-week, 1-month, 6-month and 12-months postoperatively. A standardized algorithm was used for CT reconstruction and extraction of the lumen geometries at 1 mm intervals. At each time interval, CT derived cross-sectional areas were co-registered and correlated to DUS peak systolic velocities(PSV) within six pre-designated anatomic zones and then analyzed for outcome association. Vein graft failure was defined as pathological change within a given anatomic zone resulting in thrombosis, amputation or re-intervention within the six month period following the observed time point.

Results

Fifty-four patients were recruited and 10(18%) experienced failure within 18 months of implantation. The expected inverse relationship between CSA and PSV was only weakly correlated(Spearman rank coefficient= −0.19). Moderate elevations in PSV ratio(PSV_r 2–3.5) were frequently transient, with 14 of 18 grafts(78%) demonstrating ratio reduction on subsequent imaging. A PSV_r ≥ 3.5 was associated with a 67% failure rate. CT stenosis <50% was highly correlated with success(zero failures); however, high-grade(>80%) CT stenosis was more likely to succeed than fail(25%). Eighteen patients had significant discordance between CT and DUS. While 14 of these patients had CT stenosis > 70% with a PSV_r < 3.5, only 2 subsequently failed. Conversely, 3 of 4 subjects with CT stenosis < 70% but PSV_r > 3.5 experienced graft failure. Focused analysis of these cases using computational fluid dynamic modeling demonstrated that vein side branches, local tortuosity, regional diameter variations, and venovenostomies to be the drivers of these discrepancies.

Conclusion

This analysis demonstrated that a PSV_r ≥ 3.5 is strongly correlated with VBG failure while the natural history of moderately elevated PSV_r(2–3.5) is largely clinically benign. Although minimum stenosis on CT scan was highly predictive of success, high grade CT stenosis was infrequently associated with failure. The interaction of anatomic features with the local flow dynamics were identified as the primary confounder for a direct correlation between CT and DUS.

Introduction

The consequences of lower extremity autogenous vein bypass graft(VBG) thrombosis are potentially devastating, with approximately 15–40% of affected individuals subsequently undergoing major amputation^{1, 2}. Given these poor outcomes, graft surveillance strategies are frequently employed with pre-emptive remediation to prevent graft loss. Numerous singleinstitution studies have evaluated and supported the role of duplex ultrasound surveillance(DUS)^3–6, however the merits of surveillance have been challenged by several randomized clinical trials that demonstrated a lack of a clear benefit in bypass patency or limb salvage rates^7–9. As such, current recommendations for postoperative VBG monitoring include routine physical exam, pulse palpation and measurement of ankle-brachial indices, without routine ultrasound evaluation¹⁰. Among the difficulties is that DUS may fail to detect impending graft thrombosis in up to half of all occlusion events². This apparent lack of sensitivity has led to calls for improved risk stratification and surveillance techniques.

At its core, DUS is a physiologic study that examines the local variations in graft hemodynamics to identify focal areas of lumen narrowing. Physiologic perturbations in the hemodynamic environment of the vein graft, coupled with the inherent technical variability associated with study acquisition, may negatively influence the accuracy of DUS. Recent advances in the speed and accuracy of anatomic, cross-sectional imaging technologies may avoid some of these limitations implicit in physiologic-based DUS imaging. In particular, the high resolution and substantial reproducibility of computed tomographic angiography(CTA) seems well equipped for this task. The role of CTA for procedure planning and vessel monitoring has been highlighted in the cardiac literature, where CTA has been shown to be useful in evaluation of coronary vessel area and correlated to intravascular ultrasound¹¹. Additionally, the submillimeter spatial resolution and the ability to perform computational fluid dynamic modeling from high quality CT imaging can provide unique morphologic and functional characterization of vascular anatomy that is lacking in DUS and conventional arteriography¹². To date, no studies have examined the role of CTA in postoperative VBG monitoring and compared this imaging modality to DUS.

Focused on these issues, the purpose of the current study was to analyze the correlation between DUS and CTA for detection of vein graft stenosis and evaluate the relationship of the observed abnormalities to VBG failure.

Methods

Study design

Over the last decade, the vascular surgery group at the University of Florida has had an established interest in understanding the intersection of systemic inflammation and local hemodynamic forces as the fundamental drivers of vein graft biology. Within the construct of our prospective, translational studies¹³, serial contrast-enhanced CT and DUS scans were performed, offering the ideal data set to examine the relative value of these imaging studies in clinical care. The current study focuses on the subset of 54 patients who underwent infrainguinal autogenous vein graft bypass between 2007 and 2012 at the Malcolm Randall Veterans Affairs Medical Center in Gainesville, Florida.

Patients were followed post-operatively to monitor graft patency with concurrent DUS and CTA imaging performed at 1 week, 1 month, 6 months and 12 months post-operation. Eleven (20%) patients failed to reach the 12-month imaging follow-up time point (death, N =3; amputation, N =3; loss to follow-up, N = 5). The study protocol was approved by the Institutional Review Board and written informed consent was obtained from each participant.

Duplex ultrasound surveillance

To facilitate communication about VBG anatomy, we partition the lower extremity into 6 zones of equal length corresponding to the proximal, mid-, distal thigh and the proximal, mid-, distal calf. Note that most grafts naturally do not span all six anatomic zones. Ultrasound scanning was typically performed using a 7–15 MHz transducer(5-MHz;tunneled grafts) along the entire VBG, including the proximal and distal anastomosis and corresponding inflow/outflow arteries. Color flow pulsed Doppler spectral waveform analysis was performed and the highest peak systolic velocity(PSV) measurement obtained with a ≤ 60° insonation angle within the pre-determined anatomic zones was recorded.

CTA image acquisition and vein graft reconstruction

CT scans of the VBGs were performed with intravenous contrast at 1-mm slice intervals. Acquired scans were analyzed using our previously published reconstruction algorithms¹⁴. Briefly, Amira(Visualization Sciences Group, Burlington, VA) was used to visualize the vein graft using an isosurface module below a threshold intensity value which is usually 100. The VolumeEdit module was used to remove all remaining non-graft structures. These altered images were then segmented using the Vascular Modeling Toolkit(www.vmtk.org) with a level set algorithm. The graft images were then further processed to remove surface irregularities using Geomagic Studio(3D Systems, Rock Hill, SC). Next, the centerline of the graft was created using VMTK, and the areas of cross-sections orthogonal to this centerline were calculated at 1-mm intervals using a customized Matlab script(MathWorks, Inc., Natick, MA). A rolling average technique that incorporated the seven nearest neighbors was employed in order to approximate the limited spatial resolution of DUS and to minimize the effect of extremely large or small cross-sectional area(CSA) variation on the dataset. The minimum CSAs within each zone, which theoretically would correspond to the maximum DUS-derived PSV, were selected to represent the CSA for that zone(Figure 1).

Figure 1.

Technique for vein graft reconstruction. A. CT arteriography with 1-mm slice thickness of the vein bypass graft was obtained at 1-week, 1-month, 6-month and 12-month postoperative intervals. The lower extremity is divided into six anatomic zones: three in the upper leg corresponding to the proximal, middle and distal thigh, and three in the lower leg corresponding to the proximal, middle and distal calf. A single, representative slice from each of the four zones which this graft spans is shown with the graft circled in yellow. B. The isosurface module in Amira software was used to visualize the lower extremity tissues and then remove all non-graft tissue by properly setting an intensity threshold. The vein graft is blue and the native arteries are red. C. Graft reconstruction was completed the with Vascular Modeling ToolKit(VMTK) based on the level set algorithm. The graft centerline was created and crosssections orthogonal to it were calculated at 1-mm intervals. D. Cross-sectional areas of the vein bypass graft are calculated at each millimeter along the graft at the four post-operative time points.

Data set derivation

The reconstructed, CT-derived graft geometries across all time points were co-registered and correlated with the corresponding anatomic DUS zones. Thus, our data set consisted of both a DUS PSV measurement and a corresponding CT CSA measurement for each zone per graft, at each of the four time points(l week, 1 month, 6 months, 12 months), for a total of 656 graft*zone*time points. To parallel the clinical approaches currently in use, we transformed our 656 data pairs of CT CSA and DUS PSV measurements into CT percent stenosis and PSV ratios.¹⁵ CT percent stenosis was calculated for each zone by subtracting from 1 the ratio of the minimum cross-sectional area and the mean cross-sectional area within the zone:

$$\mathit{\text{CT}}\%\mathit{\text{stenosis}}=1-\frac{{\mathit{\text{Zone CSA}}}_{\mathit{\text{minimum}}}}{{\mathit{\text{Zone CSA}}}_{\mathit{\text{mean}}}}$$ (1)

We chose this definition in order to maximize the chance that the calculated CT stenosis ratio would reflect the narrowest point within the zone, which presumably would be the same point from which the highest PSV in that zone would be recorded. DUS PSV ratios were measured using standardized interpretation criteria¹⁶. Except in the case of the most proximal zone in a graft, a zone’s velocity ratio was defined by the following equation:

$${\mathit{\text{PSV}}}_{(\mathit{\text{ratio}})}=\frac{{\mathit{\text{PSV}}}_{\mathit{\text{zone}}}}{{\mathit{\text{PSV}}}_{\mathit{\text{proximal zone}}}}$$ (2)

For the most proximal zone in each graft, the PSV_r was defined as the velocity in that zone divided by the velocity in the zone immediately distal that region. Any ratio less than 1 was considered to be equal to 1.

End-points, definitions, and statistical analyses

The primary end-point was development of CT and/or DUS evidence of stenosis. Stenosis was defined as CT stenosis ≥ 50% and/or PSV_r ≥ 2.0. The secondary end-point of the analysis was VBG failure. VBG failure was defined as any CT and/or Duplex ultrasound detected stenosis identified within a given anatomic zone of the bypass between 1 and 18 months postoperatively resulting in re-intervention, thrombosis, or amputation above the ankle within the 6 months of the initially detected stenosis. Any re-intervention within 1 -month of bypass implantation was considered a technical failure and excluded from this definition. Additionally, any evidence of CT or DUS stenosis that was identified within 1 month postoperatively that had no evidence of progression but subsequently underwent remediation was not classified as a failure. Finally, limb amputation events that occurred for non-graft failure events(e.g. severe tissue loss progression, infection, etc.) were not designated as failures.

Outliers within the analysis were defined by a CT stenosis ≥ 70% and/or PSV ratio ≥ 3.5. These values were chosen due to the known changes in pressure and flow associated with critical stenosis of a vessel¹⁷ and most common DUS criteria for imminent graft occlusion that correlates to a >70% diameter-reducing stenosis¹⁸. A prospective uniform protocol for confirmatory angiography and reintervention was present throughout the duration of the study. Patients underwent confirmatory arteriography(with or without re-intervention) if they developed a PSV ratio ≥ 3.5 and/or significant CT stenosis(> 70%) within a specific anatomic zone of the VBG. The decision regarding the timing and type of remediation was left to the operating surgeon’s discretion. Means comparison was performed in SPSS, version 22(IBM, Chicago, IL) using either Fisher’s exact test or Mann-Whitney rank-sum test. A value of P < .05 was considered statistically significant.

Results

Study population

A total of 54 patients underwent infra-inguinal autogenous VBG with a mean age of 63.2±8.3 years and a majority were male(98%;N=53). Indications included disabling claudication(6%;N=3), rest pain(44%;N=24) and tissue loss(50%;N=27). Additional details regarding the operative characteristics for the study population are highlighted in Table I.

Table I.

Patient Demographics and Operative Characteristics

Patient Demographics, No. (%)	N = 54
Age, mean ± SD	63.2 ± 8.3
Male	53 (98%)
Race
Caucasian/Non-hispanic	42 (78%)
African American/Black	10 (18%)
Other	2 (4%)
Operative Characteristics
Indication
Claudication	3 (6%)
Rest pain	24 (44%)
Tissue loss	27 (50%)
Conduit
Single leg vein	35 (65%)
Single arm vein	3 (5%)
Spliced vein	16 (30%)
Valve status
Lysed	33 (61%)
Not lysed	12 (22%)
Composite	9 (17%)
Vein Configuration
Non-reversed	24 (44%)
Reversed	17 (32%)
Combinationa	8 (15%)
In-situ	5 (9%)
Inflow
Common femoral	40 (74%)
Superficial femoral	9 (17%)
Popliteal	5 (9%)
Outflow
Popliteal	11 (20%)
Tibial	39 (72%)
Dorsalis pedis	4 (8)

SD, standard deviation

^aCombination = one segment reversed, other segment is non-reversed

DUS and CT correlation analysis

The relationship between CT CSA and DUS PSV was initially explored by plotting each of the 656 paired values in Figure II. Only a weak association between the measurements was found(Spearman rank correlation coefficient = −0.19). Most notable was that those segments with either a small CT CSA(<30mm²) or a low DUS PSV(< 200cm/s) demonstrated substantial variability in the corresponding velocity or area measurements(Figure 2A).

Figure 2.

Correlating imaging modalities. A. Duplex ultrasound peak systolic velocity and CT cross-sectional area demonstrates a weak inverse correlation; Spearman rank correlation coefficient = − 0.19. B. Correlation between duplex ultrasound peak systolic velocity ratio and CT percent stenosis. As predicted, a majority of the low peak systolic velocity ratios were associated with a minimal CT percent stenosis. However, several discrepancies between CT and DUS are noted for moderate to several stenoses.

Following conversion of CT CSA and DUS PSV into percent stenosis and PSV ratio, respectively, an improvement in the general correlation was observed(e.g. low PSV_r associated with a low CT % stenosis), although significant variability was present for the outlier data points. Figure 2B highlights this CT percent stenosis -DUS PSV_r discordance among the majority of outlier points, irrespective of outcome. For example, only 2 of the 18 points(11%) with a CT stenosis > 70% had a DUS PSV ratio ≥ 3.5. In fact, 11 of the 18 points(61%) with a significant stenosis on CT had a DUS velocity ratio of < 2. Similarly, only two of six points(33%) with a DUS PSV ratio ≥ 3.5 had a CT stenosis > 70%.

DUS PSV ratio, CTA stenosis and outcome

Each of the 656 graft*zone*time DUS: CTA data pairs were adjudicated to be either a clinical success or failure, and 10 failure events were identified(Supplemental Tables I and II). The relationship between PSV_r, CT % stenosis and graft failure is displayed in Table II. The distribution of both PSV ratios and CT percent stenosis varied significantly as a function of outcome(P<.001), however only PSV_r stratification identified a subset of grafts which was likely to fail. A velocity ratio ≥ 3.5 was associated with a 67% failure rate. Notably, 18 grafts with available subsequent imaging were identified to have a moderate PSV_r elevation(2–3.5) and 14(78%) had a successful clinical outcome with evidence of PSV_r stabilization or reduction with time(Figure 3). Two of the three grafts in this group that failed underwent endovascular(N=1) and open bypass revision(N=1) with documented reduction in PSV_r following remediation. The third patient suffered VBG occlusion that resulted in limb amputation. In contrast, a CT stenosis < 50% was highly correlated with success(zero failure), however even greater degrees of CT-demonstrated stenosis(>80%) were more likely to succeed than fail(25%).

Table II.

Relationship between duplex ultrasound peak systolic velocity ratio, CT scan cross sectional area diameter reduction and graft failure^a

	Failed	Patent	P-valueb
DUS PSV Ratio
< 2.0	3 (0.5%)	622 (99.5%)
2.0 – 3.5	3 (12%)	22 (88%)
> 3.5	4 (67%)	2 (33%)	<.001
CT Stenosis
0 – 49%	0 (0%)	578 (100%)
50 – 59%	2 (5%)	35 (95%)
60 – 69%	5 (22%)	18 (78%)
70 – 79%	2 (14%)	12 (86%)
> 80%	1 (25%)	3 (75%)	<.001

^aFailure = pathological change within a given anatomic zone resulting in thrombosis, amputation or re-intervention within the six month period following the observed time point; 656 graft*zone*time point represented

^bFisher’s exact test

Figure 3.

Fate of graft segments with moderately elevated U/S velocity ratios. T₀ = time point when the DUS derived peak systolic velocity ratio is 2.0 – 3.5. T₁ = subsequent time point. At six months, the majority of these grafts remained patent (88%). Of these grafts that remained patent, 73% had a velocity ratio reduction at the subsequent time point.

Time dependent analysis of DUS and CT measurements

For each patient*zone*time point, we calculated the percent change of both the DUS PSV and CT CSA from the time point directly preceding it. The points that failed had an interval increase in PSV(median 17%) during the time period prior to failure, while those that succeeded had an interval decrease(median 7%) in PSV, although the difference failed to meet significance(P=.25). In contrast, the difference in the interval change in CT CSA was significantly different between the outcome groups. While VBGs that remained patent had essentially no change in CSA(median increase< 1%), those that failed had a median decrease in CSA of 54% compared to the prior study(P< .001) (Figure 4).

Figure 4.

Comparing measurement change over time between imaging modalities. Peak systolic velocity percent change from the prior study was not significantly different between those grafts that failed and those that remained patent (P = .249). However, the change in CT cross-sectional area did vary significantly as a function of graft outcome (P < .001).

DUS-CT discordance and 3D reconstruction analysis

To understand causation of the imaging modality discordance, reconstructed CT images of each graft containing an outlier DUS-CT data pair mismatch were examined. In all cases, the CT-reconstructed image revealed an anatomic irregularity of the graft that was most likely responsible for creating the discrepancy which is tabulated in Table III. For the subset of patients with a CT stenosis >70% but a corresponding PSV_r < 3.5, there existed either a ligated side branch(N=5), venovenostomy(N=3), normal regional morphologic variation(N=4) or an imaging reconstruction artifact(N=1) in the zone of interest; any of these anatomic elements could create a high calculated CT stenosis in the absence of any true intra-luminal pathology. Conversely, points having no critical stenosis on CT scan(< 70%) but having a PSV_r ≥ 3.5, were uniformly noted to have spatially complex configurations, and presumed non-uniformities in the local flow patterns, as the underlying cause of the discrepancy. Three of the four grafts had tortuosity which impacted the accuracy of the CT reconstruction and/or affected the measured DUS velocities. Interestingly, three of these four grafts failed, suggesting clinically significant pathology was present within these grafts.

Table III.

Three dimensional CT reconstruction analysis to investigate significant CT and DUS imaging discordance

Abnormal CTa, normal DUSb
Patient	CT stenosis (%)	DUS (PSV ratio)	Reason for discordance	Outcome
1	77%	2.3	side branch	Patent
2	86%	1.0	side branch	Patent
3	76%	1.7	side branch	Patent
4	74%	1.5	side branch	Patent
5	73%	2.5	side branch	Failed
6	72%	2.0	venovenostomy	Patent
7	70%	1.0	venovenostomy	Patent
8	71%	1.0	venovenostomy	Patent
9	80%	1.0	regional variation	Patent
10	76%	1.0	regional variation	Patent
11	73%	1.2	regional variation	Patent
12	80%	2.9	regional variation	Patent
13	76%	1.0	reconstruction artifact	Patent
14	72%	3.3	tortuosity	Failed

Normal CTc, abnormal DUSd
Patient	CT stenosis (%)	DUS (PSV ratio)	Reason for discordance	Outcome
15	14%	3.9	tortuosity	Patent
16	58%	6.7	tortuosity	Failed
17	63%	6.4	tortuosity & varicose vein	Failed
18	62%	3.9	side branch	Failed

^aabnormal CT = >70% cross sectional area reduction

^bnormal DUS = PSV ratio ≤ 3.5

PSV, peak systolic velocity

^cnormal CT = < 70% cross sectional area reduction

^dabnormal DUS = PSV ratio ≥ 3.5

Discussion

Multiple studies have investigated the role of postoperative DUS surveillance of VBGs^{3–9, 19–21} however, our study is the first to prospectively collect and compare serial concurrent DUS-CT data while relating it to graft outcome. We discover that CT CSA and DUS PSV are weakly correlated. While conversion to CT percent stenosis and DUS PSV ratios improves this correlation, considerable discordance is still observed, particularly in those regions of critical (high-grade) stenosis. Our analysis corroborates previous findings that a PSV_r ≥ 3.5 is strongly associated with VBG failure while the natural history of moderately elevated PSV ratios(2.0–3.5) is relatively benign. In contrast, although a minimal stenosis on CT(<50%) was highly correlated with successful outcome, high-grade CT stenosis(>70%) was infrequently associated with VBG failure. The discrepancy between DUS and CT in detecting critical VBG stenosis is most frequently related to variation in vein graft morphology.

Approximately one-third of lower extremity vein bypass grafts will either fail or require revision within the first post-operative year²² and the optimal approach to graft surveillance for identifying and repairing lesions which may lead to graft occlusion is not clear. Vein graft failure occurs at three distinct temporal phases including early (0–30 days; technical factors, intrinsic defects of the conduit, hypercoaguability), mid-term (neointimal hyperplasia) and late (progression of native atherosclerotic disease) time-points. Further confounding these patterns of failure are observations from our laboratory. We have previously examined the complex flow patterns in the tertiary structure of VBGs and demonstrated marked variations in blood velocity that appear independent of a clearly defined anatomic stenosis.²³ Such local variations may contribute to the lack of sensitivity of DUS in detecting a significant proportion of early graft occlusion events. These findings in conjunction with multiple RCTs^7–9 disputing the role of DUS monitoring and a recent analysis by Oresanya et. al.², highlight the need for better risk stratification and surveillance protocols after VBG surgery.

CT scanning is increasingly utilized by vascular surgeons and its application continues to expand as three dimensional imaging techniques become more commonly integrated into the management of aortic^{24, 25}, carotid²⁶, and lower extremity²⁷ disease. For example, some groups now perform pre-operative vein mapping, once obtained exclusively by ultrasound, using CT scans²⁷. In addition, finite element analysis, which uses CT data to model fluid dynamics and vessel wall forces, has been shown to have value in AAA rupture risk prediction.^{24, 28, 29} While cost as well as repeated contrast and radiation exposure of a post-operative routine CT surveillance program would almost certainly be prohibitive, given that the role of CT imaging in the care of the vascular patient is expanding, an initial analysis of its usefulness in monitoring VBGs is timely.

Cross-sectional imaging is a purely anatomic study, which uses graft geometry to identify regions of luminal narrowing. In contrast, duplex ultrasonography is fundamentally a physiologic study that examines the intersection of local geometry and graft flow to identify areas of hemodynamic perturbation. Whether a direct measurement of luminal narrowing or an assessment of regional alterations in blood flow velocities is a better predictor of impending vein graft failure remains an unanswered question, and a central focus of this investigation.

The weak association between CT and DUS may reflect technical limitations in performing these studies as much as it does true discordance between graft luminal area and blood velocity, since the data produced by both modalities are subject to multiple potential confounders. DUS PSV measurements are affected by multiple non-pathological elements including physiologic variables(e.g. graft inflow velocity), anatomic features(e.g. graft morphology, tandem stenosis) and methodological factors(e.g. duplex operator technique). CT derived measurements have fewer potential confounders, but the necessary transformation of these high-resolution anatomical data into fewer, useable data elements reduces the granularity of the CT data so that, when compiled with DUS data, potential correlative patterns between the two modalities may be obscured.

For example, non-pathological graft morphology irregularities such as venovenostomy, ligated side branches, valve leaflets and normal variation in CSA may create the appearance of stenosis in the absence of actual intra-luminal narrowing. Also, the calculated CT stenosis ratio may not accurately reflect the degree of stenosis at the exact point within the zone at which the maximum PSV measurement was obtained. Not surprisingly, examination of outlier data elements demonstrated that all grafts had spatially complex configurations that could reduce the reliability of CT and/or DUS measurements(Table III).

Closer inspection of two bypass grafts using 3D CT reconstruction and computational fluid dynamic modeling provides further insight into the cause of discordance between the two imaging modalities(Figure 5 and Supplemental Videos 1 and 2). At one-year post operation, Patient 1 was found to have a significantly elevated PSV_r = 6.4 at the most proximal segment of his common femoral to anterior tibial artery bypass graft, while the corresponding CT showed only a moderate stenosis of 63% in that region. Concern for a failing graft prompted an angiogram one week later which identified an occluded graft, but since the patient had no remaining autogenous conduit or evidence of limb threat, no intervention was performed. The elevated PSV_r appears to have accurately identified true pathology within the graft, whereas a much greater CT stenosis would have been expected given the subsequent angiographic finding corroborating evidence of a critical stenosis leading to graft failure. Examination of the reconstructed graft image demonstrates that there is significant tortuosity and varicosity in the proximal VBG, irregularities which may have reduced the accuracy of the reconstructed CT images.

Figure 5.

Discordance between imaging modalities. A. CT-reconstructed grafts with regions of interest circled in red. B. Graft cross-sectional areas with regions of interest circled in red. C. Enlargement of regions of interest. Graft 1 has curvature without significant variation in graft cross-sectional area. Graft 2 has an outpoutching from a ligated venous branch which creates a perceived significant stenosis. D. Particle velocity mapping throughout the cardiac cycle. Graft 1 shows increased particle velocity near the graft wall at the inlet during peak systole. Just distal there is an area of flow recirculation where the velocities are quite low. Hence, the velocity ratio is elevated in this region. Graft 2 shows complex flow with vortices, but at peak systole, both inlet and outlet have high velocities. Hence, the velocity ratio ~ 1. Notably, there are no areas of re-circulation.

Patient 2 had a superficial femoral artery to posterior tibial artery cephalic VBG which at 1-month post-operation was found on CT scan to have significant stenosis(~86%) in the mid-graft region, while the DUS showed completely normal flow characteristics (PSV_r=1.0). Review of the reconstructed CT images reveals an architectural abnormality in the VBG resulting from ligation of a large venous side branch, from the segment of arm vein harvested from the antecubital fossa, which was now located at the center of the graft. This focal dilation served to increase the mean CSA of this anatomic zone, resulting in a large calculated CT stenosis ratio in the absence of actual pathology. Particle velocity analysis shows that at peak systole, there is little velocity difference between the inlet and outlet regions, and correspondingly a normal PSV_r, despite the clear variations in lumen area(Figure 5).

An interesting observation is the natural history of VBG with moderate stenosis(PSV_r 2.0–3.5) identified on DUS. Although VBGs with a moderate DUS stenosis are at an increased risk of failure compared to those with ratios < 2.0(relative risk=24), they still demonstrate only a 12% failure rate(Table 2), and the velocity ratios tend to improve with time(Figure 3). Notably, previous retrospective analyses have demonstrated that intermediate stenoses(200 cm/sec < PSV < 300 cm/sec; 2 < PSV_r < 4) are dynamic, with 20% to 30% of these lesions demonstrating significant regression within 2 years^{5, 30}. These observations are congruent with our recent work demonstrating that vein grafts have considerable plasticity and, in response to hemodynamic and biologic influences, undergo oscillatory patterns of adaptation with eventual regression of early wall thickening so that the majority of grafts do not progress to critical stenosis.³¹

A corollary to the intermediate DUS stenoses study was the time dependent analysis of DUS and CT which demonstrated that PSV_r variation appears to be independent of outcome. This likely stems from the fact that ultrasound assesses graft physiology which is affected by many non-pathologic factors which vary over time, thus prohibiting the emergence of clear temporal patterns. On the other hand, failed grafts experienced a 54% decrease in luminal area on CT compared to the preceding study, whereas those that remained patent had essentially no change(Figure 4). Clearly, grafts that fail demonstrate a trend of significantly greater luminal narrowing over time than those that remain patent. CT, unlike DUS, is likely able to detect temporal patterns because it offers a high-resolution representation of graft anatomy which is inherently subjected to fewer confounding influences.

The current role of DUS in post-VBG monitoring remains controversial as a result of RCTs^7–9 and consensus guidelines¹⁰ refuting its utility being diametrically opposed to a large array of cohort studies supporting routine surveillance^{3–5, 18, 19, 21}. Complicating identification and management of vein graft stenosis is the dynamic remodeling that occurs after implantation, during which appearance and subsequent regression of focal regions of luminal narrowing are frequently observed^{5, 32}. The emergence of new imaging techniques may provide alternative and/or complementary methods for VBG surveillance that leads to more reliable risk stratification and subsequent prediction of outcome.

Conclusion

This analysis demonstrated that PSV_r ≥ 3.5 is strongly correlated with VBG failure while the natural history of moderately elevated PSV_r(2–3.5) is largely clinically benign. Although minimum stenosis on CT scan was highly predictive of success, high grade CT stenosis was only weakly associated with failure. The interaction of anatomic features with the local flow dynamics were identified as the primary confounder for a direct correlation between CT and DUS. Based on the results of this analysis, DUS appears to be useful in determining outcomes of VBGs. Future study is needed to better understand the relationship between DUS and CT to establish the optimal strategy of VBG monitoring for detecting pathology that can be remediated prior to graft failure.

Supplementary Material

2. Supplemental Video 1.

This graft zone has only a moderate CT stenosis of 63% but a significantly elevated velocity ratio of 6.4. The tortuous course of the vein graft at this point causes a significant increase in blood flow velocity without a correspond decrease in luminal area. Computational fluid dynamics shows increased particle velocity near the graft wall at the inlet, but just beyond there is an area of flow recirculation where the velocities are quite low. Hence, the velocity ratio is increased.

3. Supplemental Video 2.

This graft zone has a significant CT stenosis of 86% in the absence of any increase in duplex ultrasound measured peak systolic velocity ratio. The vein caliber in the proximal portion of the zone is reduced compared to that of the more distal segment. This is clearly a result of a ligated side branch vessel, as this portion of vein would sit within the antecubital fossa, rather than any intra-luminal pathological stenosis. However, the large difference between the minimum and the mean cross-sectional areas within this zone creates an elevated CT stenosis by our definition. Computational fluid dynamics shows complex flow with vortices, but at peak systole, both inlet and outlet have high velocities. Hence, the velocity ratio is 1. Notably, there are no areas of re-circulation.