|Home | About | Journals | Submit | Contact Us | Français|
We studied surgeons’ practice patterns in the use of completion imaging (duplex or arteriography), and their association with 30-day stroke/death and 1-year restenosis after carotid endarterectomy (CEA).
Using a retrospective analysis of 6115 CEAs, we categorized surgeons based on use of completion imaging as rarely (<5% of CEAs), selective (5% to 90%), or routine (>90%). Crude and risk-adjusted 30-day stroke/death and 1-year restenosis rates were examined across surgeon practice patterns. Finally, we audited 90 operative reports of patients who underwent re-exploration and characterized findings and interventions. We analyzed the effect of re-exploration on outcomes.
Practice patterns in completion imaging varied: 51% of surgeons performed completion imaging rarely, 22% selectively, and 27% routinely. Crude 30-day stroke/death rates were highest among surgeons who routinely used completion imaging (rarely: 1.7%; selectively: 1.2%, routinely: 2.4%; P = .05). However, after adjusting for patient characteristics predictive of stroke/death, the effect of surgeon practice pattern was not statistically significant (odds ratio [OR] for routine-use surgeons, 1.42; 95% CI, 0.93-2.17; P = .10; selective-use surgeons, 0.75; 95% CI, 0.40-1.41; P = .366). Stenosis >70% at 1 year showed a trend toward lowest rates for surgeons who performed completion imaging (rarely: 2.8%, selectively: 1.1%, and routinely: 1.1%; P = .09). This effect became statistically significant for selective-use surgeons after adjustment (hazard risk [HR] for selective-use surgeons, 0.52; 95% CI, 0.29-0.92; P = .02). Overall, 178 patients (2.9%) underwent operative re-exploration. Routine-use surgeons were most likely to perform re-exploration (7.6% routine, 0.8% selective, 0.9% rare; P < .001). An audit of 90 re-explored patients demonstrated technical problems, the most common being flap, debris, and plaque. Rates of stroke/death were higher among patients who underwent re-exploration (3.9% vs 1.7%; P = .03); however, this affect was attenuated after adjustment (OR, 2.1; 95% CI, 0.9-5.0; P = .08).
The use of completion imaging during CEA varies widely across our region. There is little evidence that surgeons who use completion imaging have lower rates of 30-day stroke/death, although selective use of completion imaging is associated with a small but a significant reduction in stenosis 1 year after surgery. We also demonstrate an association between re-exploration and higher risk of 30-day stroke/death, although this effect was attenuated after adjustment for patient-level predictors of stroke/death. Future work is needed to direct the selective use of completion imaging to prevent stroke, rather than cause unnecessary re-exploration.
Many surgeons have used intraoperative completion imaging to ensure a technically sound operation during carotid endarterectomy (CEA).1-10 Proponents argue that completion imaging allows surgeons to identify causes of and avert early complications, as well as to ensure good long-term outcomes by avoiding restenosis.1-8,11,12 Completion imaging with arteriography3,4,9,10 and duplex ultra-sound1-13 have been studied extensively and have excellent sensitivity (97% to 100%) and specificity (100%)5,14 for the intraoperative detection of technical errors during CEA.
Despite evidence that completion imaging can effectively detect technical problems during CEA, the true utility of completion imaging remains a topic of debate.9,15 It is unclear if completion imaging is routinely necessary, which abnormalities require re-exploration, and what effect completion imaging has on outcomes such as 30-day stroke/death or restenosis.4,7-9,15 Studies attempting to answer these questions have largely consisted of small, single-center series,1-8,11,12 or were not adequately powered to detect differences in outcomes.2-7,11,12
Moreover, although proof that completion imaging is helpful has been difficult to establish, some have suggested that it may be harmful.9,15 For example, defects of uncertain significance detected on completion imaging might lead to complications caused by unnecessary surgical re-exploration. A recent literature review suggested that approximately 10% of patients who undergo completion imaging are re-explored at the time of surgery,1-8,10-13,16-25 potentially unnecessarily increasing the risk of stroke or technical complications.1-8,11-13
The purpose of this study was to describe the use of completion imaging in our region, and its effect on outcomes after CEA. Using different surgeon practice patterns in the use of completion imaging as our exposure variable, we studied the associations between completion imaging and 30-day stroke/death, as well as restenosis ≤1 year after surgery.
For this report, we used data collected by the Vascular Study Group of New England (VSGNE), a regional cooperative quality improvement initiative developed to improve vascular health care. Further details on the registry used by VSGNE have been published previously26 and are available at http://www.vsgne.org.
Data were examined from 6379 CEAs performed in 5638 patients who underwent CEA by 73 participating surgeons across 11 study hospitals, between January 2003 and January 2010. Of these, we excluded 118 CEAs done concomitantly during coronary bypass grafting and 146 redo CEAs. This left 6115 primary, isolated CEAs in our cohort for analysis.
Cases involving the use of completion imaging (defined as duplex ultrasound or contrast arteriography) were identified in the VSGNE data set. Completion imaging, interpretation, and decision making regarding potential carotid artery re-exploration was performed at the discretion of the operating surgeon and was not controlled by any research protocol.
Based on the use of completion imaging by individual surgeons in our data set (Fig 1), as well as by a literature review,9 we categorized surgeon practice patterns as rare (<5% of a surgeon's CEAs), selective (5% to 90%) or routine (≥90%) use of completion imaging. We then evaluated patient outcomes after CEA across each of the three groups.
Our study had two main outcome measures: 30-day postoperative stroke/death, and >70% stenosis of the ipsilateral carotid artery at 1-year follow-up. Major stroke was defined as disability causing non-independent living status; minor stroke was defined as nondisabling.
For our second main outcome measure, >70% restenosis at 1-year, we studied only patients who had undergone duplex evaluation of their CEA at 1 year. Because not all of the patients who underwent CEA procedures in 2009 have completed follow-up, these procedures were excluded from the restenosis analyses (1116 of 6115; 18%). Of the remaining 4999 patients, 858 (17%) were lost to follow-up, and an additional 350 (7%) were excluded from analyses because they did not undergo duplex imaging at their follow-up. This left 3791 patients (76%) for analyses of restenosis at 1 year.
Stenosis on follow-up duplex scan was categorized as <50%, 50% to 70%, >70%, 70% to 80%, >80%, or occlusion. We defined a clinically significant stenosis as >70%.
First, we examined univariate associations between individual patient characteristics and the use of completion imaging with χ2 and Fisher exact tests. To gain insight into the patient-level factors associated with undergoing a completion imaging study, all variables with values of P < .20 were used to develop a multivariate logistic regression model to predict which patients were most likely to undergo completion imaging.
Next, we performed univariate comparisons with our main outcome measures across each of the three categories of surgeon practice pattern in completion imaging. The 30-day stroke/death rate was also analyzed independent of practice pattern and reported in a distinct patient-level analysis. This variable was analyzed in a categoric fashion.
Life-table analysis was used to calculate the incidence of stenosis at 1-year follow-up. Kaplan-Meier survival curves and log-rank tests were used to compare 1-year stenosis rates across the three surgeon practice patterns.
To risk adjust rates of 30-day stroke/death, we used a previously published logistic regression model that predicted the patient-level risk of 30-day stroke/death in the VSGNE as described in our previous work.27 Covariates in this model were used to determine the effect of surgeon practice pattern on 30-day stroke/death rate, adjusted for the patient characteristics known to be associated with stroke or death. A similar Cox proportional hazards model was constructed for stenosis at 1 year, using our previously published model for risk factors associated with restenosis.28 We purposefully included surgeon practice pattern, independent of its significance, to allow us to discern its effect on our main outcome measures. We also examined surgeon volume and adjusted for variations in volume in our multivariate analysis.
To gain insight into what aspects of completion imaging prompt re-exploration, as well as what actions are taken during re-exploration, we sampled 90 operative reports from 178 patients who underwent re-exploration from two institutions where these data were available (operative records from the remaining institutions were not available for analysis).
Findings on completion imaging were categorized in five ways: (1) flap/debris/plaque, (2) thrombosis, (3) dissection, (4) increased/decreased velocity, or (5) other. We then categorized the intervention undertaken for each case as (1) tacking/removal of debris, (2) angiography with or without endovascular intervention, or (3) other.
We then examined the effect of re-exploration in the multivariable models described above on 30-day stroke or death.
Overall, we studied 6115 CEAs, performed in 5638 patients by 73 surgeons. Completion assessment of patency was performed in 5554 CEAs (91%); however, in 3520 (58%) this consisted only of Doppler insonation, and not completion imaging, defined as duplex scanning or arteriography. Completion imaging was used in 2033 CEAs (33%) comprising DUS imaging in 1919 (94%), arteriography in 94 (5%), and both arteriography and DUS imaging in 20 (1%). Patient, surgeon, and hospital characteristics were compared between CEAs in which completion imaging was and was not performed (Table I, A; Appendix Table I, online only).
Several patient-level and operative factors were associated with the use of completion imaging (Table I, B; Appendix Table II, online only). This model had reasonable discriminative ability (area under the receiver operating curve = 0.72). Unfortunately, intraoperative factors that may further delineate the decision to use completion imaging are not contained in our data set and therefore were not available for analysis.
Of the 73 surgeons in our study, 37 (51%) rarely, 16 (22%) selectively, and 20 (27%) routinely used completion imaging (Fig 1). Although 54% of CEAs were performed by the 37 surgeons who rarely used completion imaging, 29% of CEAs were performed by the 20 surgeons who routinely used completion imaging. The 16 surgeons who selectively used completion imaging performed 17% of CEAs in our study.
Surgeon volume during the study period was nearly identical in the groups of surgeons classified as rare (mean, 90 cases per surgeon; range, 5-585) and routine (mean, 88; range 5-378) and was lower in the selective group (mean, 40; range, 5-174). However, within each category of surgeon practice pattern, there was wide variation in individual surgeon case volume. Because our main outcome measures occurred relatively infrequently, this limited our ability to analyze the effect of volume on 30-day stroke/death at the individual surgeon level.
Crude 30-day stroke/death rates were significantly lower among surgeons who selectively used completion imaging and were higher in surgeons who routinely used completion imaging (1.7% rarely, 1.2% selectively, 2.4% routinely, P = .05; Table II). When we risk-adjusted 30-day stroke/death rates for patient characteristics known to independently predict stroke/death, including age (≥70 years), contralateral internal carotid artery occlusion, use of antiplatelet agent, presence of congestive heart failure, emergency procedure, preoperative ipsilateral cortical symptoms,27 we found that the differences across surgeon practice patterns were not statistically significant, but still demonstrated the same direction and magnitude of effect for selective-use surgeons (odds ratio [OR], 0.75; 95% confidence interval [CI], 0.40-1.41; P = .366) and for routine-use surgeons (OR, 1.42; 95% CI, .93-2.17; P = .106; Table III).
In our second main outcome measure, restenosis >70% at 1 year, crude differences across surgeon practice patterns were not statistically significant (2.8% rarely, 1.1% selective, 1.1% routine, P = .09). However, Kaplan-Meier survival curves demonstrated a trend toward a slightly lower risk of restenosis among surgeons who performed selective or routine completion imaging (Fig 2). Of all patients diagnosed with restenosis >70% at 1 year, only four underwent reintervention (repeat CEA or carotid stent) due to symptoms.
When we risk-adjusted 1-year restenosis rates for patient characteristics known to independently predict restenosis, including type of closure (primary vs patched), contralateral carotid artery stenosis, and dialysis,28 selective use of completion imaging was associated with a significantly lower risk of stenosis at 1 year, with a hazard ratio (HR) for restenosis in selective-use compared to rare-use of 0.52 (95% CI, 0.29-.092, P = .024; Table IV). However, routine use of completion imaging had little effect on the risk of restenosis, with an HR for restenosis in routine-use compared to rare-use of 0.92 (95% CI, 0.61-1.38, P = .676). As in our prior work, we confirmed that there are significant associations between primary closure, contralateral carotid artery stenosis, and restenosis.28
Overall, intraoperative carotid artery re-exploration occurred in 178 CEAs, comprising 2.9% of all CEAs and 8.8% of CEAs with completion imaging. The rate of re-exploration was significantly higher among routine-use surgeons compared with selective or rare-use surgeons (7.6% routine, 0.8% selective, 0.9% rare, P < .001 between routine/selective, P = .001 between routine/rare). Of the 31 patients who underwent re-exploration without completion imaging, 25 had Doppler insonation of the endarterectomy before re-exploration.
Of these 178 CEAs, we audited 90 available operative reports from cases where re-exploration was prompted by findings on completion imaging. The most common finding on completion imaging was flap/debris/plaque (77 of 90, 86%; Fig 3). Examples of these abnormalities are shown in Fig 4.
The most common finding upon re-exploration was also flap/debris/plaque (67 of 88, 76%). In 10 of 88 patients (11%), however, the abnormal completion imaging study represented a false positive, in that no identifiable technical defects were detected on re-exploration. Two of these 10 patients underwent intervention (additional tacking sutures) even though no discrete abnormality was found (Fig 3).
Overall, 88 arteries were surgically re-explored, and 2 additional patients underwent arteriography and stent placement without surgical re-exploration. In 79 of 90 re-explorations (88%) the intervention involved removing or tacking excess debris/plaque. In 8 cases (9%), no intervention was performed.
Among the 178 patients who underwent re-exploration, 6 strokes (3 minor, 3 major) and 1 death occurred ≤30 days. These events were evenly distributed across the categories of completion imaging findings. Crude 30-day stroke/death rates were 3.9% in the 178 cases with arterial re-exploration compared with 1.7% in those patients who were not re-explored (P = .028). However, these differences were attenuated when we risk-adjusted for patient characteristics predictive of 30-day stroke or death (adjusted OR, 2.1; 95% CI, 0.9-5.0; P = .076).
To investigate our presumption that completion imaging is used most commonly in CEAs where stroke risk is higher, we compared the rates of 30-day stroke/death between patients who underwent completion imaging with those who did not. Crude 30-day stroke/death rates were higher in patients where completion imaging is performed (2.6% vs 1.3%; P < .001). Further, after adjusting for patient characteristics associated with 30-day stroke/death, the odds of stroke/death remained significantly higher in patients undergoing completion imaging (OR, 1.9; 95% CI, 1.2-2.7; P = .002), suggesting that differences in outcome were not due to differences in patient characteristics.
Surgeon practice pattern in the use of completion imaging varies widely across New England, and the use of completion imaging is not associated with a significantly lower risk of 30-day stroke or death. Rather, routine use of completion imaging was associated with a trend toward a higher risk-adjusted 30-day stroke/death, and selective use was associated with a trend toward lower risk-adjusted 30-day stroke/death rates. Further, surgeons who selectively use completion imaging have a significant reduction in restenosis ≤1 year after surgery, although the absolute effect size was small. Lastly, it remains uncertain whether re-exploration based on completion imaging is associated with higher rates of perioperative stroke/death, although our results suggest that the risk of adverse outcomes are higher in those patients who undergo re-exploration.
Several prior investigators have studied the relationship between completion imaging and stroke/death after CEA (Table V). In a 1997 study in a study of >9000 CEAs, Rockman et al9 reported no significant difference in 30-day stroke/ death rates between patients undergoing completion imaging and those who did not (3.6% vs 3.3%, respectively). Studies by Dykes et al2 (1997) and Lipski et al5 (1995) further support these findings. Moreover, Rockman et al9 reported no difference in 30-day stroke/death rates be tween surgeons whose use of completion imaging was classified as routine (≥90% of CEAs) and those who never performed completion imaging (2.9% vs 3.6%). These data reflect our results, which demonstrated no statistically significant difference in 30-day stroke/death across surgeon practice pattern in completion imaging. Instead, our findings suggest a trend toward increased risk of stroke or death among surgeons who routinely use completion imaging, even after adjusting for patient factors predictive of stroke or death.
Our study specifically examined surgeons who selectively use completion imaging. We hoped that this category might add insight into the effect of surgeon “judgment,” and indeed, crude and adjusted rates of stroke/death were lowest in this group. One might reason that routine use of completion imaging may be overly sensitive and might detect more defects than necessary, resulting in unnecessary re-exploration and increased stroke risk. Our data offer suggestive support for this theory: crude and adjusted stroke risks were highest among surgeons who routinely performed completion imaging, and these surgeons had significantly higher rates of re-exploration.
Several prior studies have examined the relationship between completion imaging and stenosis after CEA (Table V).1-13,15-25 Studies by Dykes et al2 (1997) and Lipski et al5 (1995), showed less residual stenosis in the immediate postoperative period in cases using intraoperative completion imaging, and Kinney et al18 (1993) reported significantly lower rates of recurrent stenosis up to 48 months after surgery in those cases where completion imaging was performed. In our study, rates of >70% carotid artery stenosis at 1 year after CEA were significantly lower among surgeons who selectively use completion imaging. However, because we did not collect data on immediate postoperative DUS results, our 1-year stenosis rate may reflect recurrent or residual stenosis, or both. Additionally, our follow-up extends only for 1 year, limiting application of our findings in defining long-term restenosis.
Although completion imaging studies can identify correctible technical defects,1,2,16,29 no standardized recommendations presently exist to identify which lesions justify carotid artery re-exploration and which are best left alone. Several studies have demonstrated that not all lesions are associated with an increased risk of perioperative complications and, thus, do not require immediate operative revision.3,4,30 However, most would agree that the correct thresholds to inform the decision for intraoperative re-exploration, except for in cases with extreme findings, have yet to be established.
Unfortunately, re-exploration itself might increase the risk of stroke, possibly due to increased ischemic time or other manipulation during re-exploration. For example, one study by Zannetti et al10 showed that patients who underwent surgical revision as a result of completion imaging findings actually experienced strikingly worse perioperative neurologic outcomes. In our region, crude 30-day stroke/death rates were 3.9%, significantly higher in patients who underwent re-exploration than in patients who did not (1.7%), although this difference was attenuated with risk adjustment. Although this suggests that re-exploration itself may be associated with a higher risk of stroke or death, it is unknown what might have happened to these patients if they had not been re-explored. It is reasonable to infer that surgeons who re-explore patients do so because they believe the risk of stroke/death is higher if the technical abnormality is not corrected.
Many will question why we simply did not compare the risk of stroke/death between patients who underwent completion imaging and those who did not. We reasoned that cases involving completion imaging, especially among selective users, likely represent technically challenging operations, wherein the surgeon chose to interrogate the repair using completion imaging. Although our risk adjustment can account for some proportion of these excess patient-level risks, there remains a significant possibility that confounding by indication may be present.31 Therefore, we focused on the exposure variable in our study, surgeon practice pattern.
However, little is known about what conditions prompt surgeons to perform completion imaging during CEA. Further, it remains unclear what completion imaging findings should prompt re-exploration, and which findings can be managed conservatively. Our study showed that surgeons who routinely use completion imaging are significantly more likely to perform re-exploration, and this suggests that routine-use surgeons are either finding more significant defects than selective-use or rare-use surgeons or have a lower threshold for intervention. Our future work will seek to identify those settings wherein completion imaging will be most likely to identify only those lesions that necessitate operative revision, without being “too sensitive” and prompting unnecessary re-exploration.10
Our study has several limitations. First, data in the VSGNE reflect observational data, not evidence from a randomized trial, and certain elements and details surrounding patient events, such as the timing and extent of stroke or operative indications for re-exploration, are limited.
Second, practice patterns across the VSGNE are not standardized. Thus, decisions such as when to perform completion imaging and when to re-explore the carotid artery likely vary across centers as well as among individual surgeons, introducing potential selection bias.
Third, variation in surgeon practice pattern over time is a possibility. However, when we examined the individual surgeons in the selective use of completion imaging category, we found that only 6% moved across categories, indicating that surgeons who fall in the 5% to 90% range are truly selecting when to perform completion imaging and when it is not necessary based on various personal thresholds and indications.
Fourth, although we recognize that the descriptive data pertaining to re-exploration are limited by the fact that only half of re-explored patients were included in our chart review, all quantitative analyses included every re-explored patient in our cohort.
Fifth, DUS criteria for restenosis, either at the time of completion imaging or at follow-up, varied across institutions. However, these criteria did not change over time, limiting any bias introduced to within-center variability, with no changes across centers or over time.
Although completion imaging studies identify correctable technical lesions after CEA, there is little evidence to suggest that surgeons who routinely or selectively use completion imaging have lower rates of 30-day stroke or death. However, our study demonstrates that surgeons who selectively use completion imaging have a trend toward lower rates of 30-day stroke or death, and a small, but a statistically significant reduction in stenosis at 1 year after CEA. Lastly, it remains uncertain whether or not re-exploration based on completion imaging is associated with higher rates of perioperative stroke or death, although our study suggests that the risks of adverse outcomes are higher in those patients who undergo re-exploration.
We would like to thank the following people for their generous contributions to this study: Yuanyuan Zhao, MA (Dartmouth Hitchcock Medical Center), Robert A. Cambria, MD (Eastern Maine Medical Center), Andres Schan zer, MD (University of Massachusetts Medical Center), and Jens Eldrup-Jorgensen, MD (Maine Medical Center).
Competition of interest: none.
Presented at the Thirty-seventh Annual Meeting of the New England Society for Vascular Surgery, September 24-26, 2010, Rockport, Me.
The editors and reviewers of this article have no relevant financial relationships to disclose per the JVS policy that requires reviewers to decline review of any manuscript for which they may have a competition of interest.
Conception and design: JW, PG, DS, BN, DW, JC
Analysis and interpretation: JW, PG, JC
Data collection: JW, PG, DB, JV, DS, BN, DW, JC
Writing the article: JW, PG
Critical revision of the article: JW, PG, DB, JV, DS, BN, DW, JC
Final approval of the article: JW, PG, DB, JV, DS, BN, DW, JC
Statistical analysis: JW, PG
Obtained funding: Not applicable
Overall responsibility: JW