Clinical and angiographic predictors of stroke and death from carotid endarterectomy: systematic reviewBMJ 1997; 315 doi: http://dx.doi.org/10.1136/bmj.315.7122.1571 (Published 13 December 1997) Cite this as: BMJ 1997;315:1571
- Correspondence to: Dr P M Rothwell Department of Clinical Neurology, Radcliffe Infirmary, Woodstock Road, Oxford OX2 6HE peter.
- Accepted 9 July 1997
Objective: To identify risk factors for operative stroke and death from carotid endarterectomy.
Design: Systematic review of all studies published since 1980 which related risk of stroke and death to various preoperative clinical and angiographic characteristics, including unpublished data on 1729 patients from the European carotid surgery trial.
Main outcome measure: Operative risk of stroke and death.
Results: Thirty six published studies fulfilled our criteria. The effect of 14 potential risk factors was examined. The odds of stroke and death were decreased in patients with ocular ischaemia alone (amaurosis fugax or retinal artery occlusion) compared with those with cerebral transient ischaemic attack or stroke (seven studies; odds ratio 0.49; 95% confidence interval 0.37 to 0.66; P<0.00001). The odds were increased in women (seven studies; 1.44; 1.14 to 1.83; P<0.005), subjects aged ≥75 years (10 studies; 1.36; 1.09 to 1.71; P<0.01), and with systolic blood pressure >180 mm Hg (four studies; 1.82; 1.37 to 2.41; P<0.0001), peripheral vascular disease (one study; 2.19; 1.40 to 3.60; P<0.0005), occlusion of the contralateral internal carotid artery (14 studies; 1.91; 1.35 to 2.69; P<0.0001), stenosis of the ipsilateral internal carotid siphon (five studies; 1.56; 1.03 to 2.36; P=0.02), and stenosis of the ipsilateral external carotid artery (one study; 1.61; 1.05 to 2.47; P=0.03). Operative risk was not significantly related to presentation with cerebral transient ischaemic attack versus stroke, diabetes, angina, recent myocardial infarction, current cigarette smoking, or plaque surface irregularity at angiography. Multiple regression analysis of data from the European carotid surgery trial identified cerebral versus ocular events at presentation, female sex, systolic hypertension, and peripheral vascular disease as independent risk factors.
Conclusions: The risk of stroke and death from carotid endarterectomy is related to several clinical and angiographic characteristics. These observations may help clinicians to estimate operative risks for individual patients and will also facilitate more meaningful comparison of the operative risks of different surgeons or at different institutions by allowing some adjustment for differences in case mix.
Although carotid endarterectomy has been shown to reduce the risk of stroke in selected patients, there is about a 5% operative risk of stroke and death
The cost effectiveness of the operation is limited by the associated morbidity and mortality
The risk of stroke and death from carotid endarterectomy is higher in women than in men
Other clinical characteristics associated with an increased risk include cerebral as opposed to ocular transient ischaemic attack, age over 75 years, systolic hypertension, and peripheral vascular disease
Angiographic characteristics associated with an increased operative risk include occlusion of the contralateral internal carotid artery and stenosis of the ipsilateral carotid siphon or external carotid artery
The absolute benefit derived from carotid endarterectomy is limited by the morbidity and mortality caused by the operation itself. The balance of risk and benefit is particularly fine in patients in whom the risk of stroke while they are receiving medical treatment is very low to begin with, such as those with asymptomatic stenosis.1 2 3 To maximise the cost effectiveness of endarterectomy we need to identify subgroups of patients who are at high risk of stroke without surgery but who have a relatively low operative risk of stroke and death.4 Much work has gone into prediction of the risk of major stroke on medical treatment alone,5 6 but there has been less research into the factors which predict operative risk. Although there have been at least 126 reports of the morbidity and mortality of carotid endarterectomy published over the past 15 years,7 many of which have made some attempt to identify possible risk factors, there has been no systematic review of risk factors for operative stroke and death. Such a review is required both to target endarterectomy more effectively and to help to correct for case mix in clinical research and audit. We performed a systematic review of all studies published since 1980 which related the risk of stroke and death from carotid endarterectomy to preoperative clinical and angiographic characteristics. The review also includes unpublished data from the European carotid surgery trial.8 9 In addition, we performed a multiple regression analysis of the data from the European carotid surgery trial to identify which of the potential risk factors identified in the systematic review were independent predictors of operative outcome.
The methods of the systematic review have been published previously.7 Briefly, studies were identified by a single observer from CD ROM (Cambridge Medline, 1980-6) by using the search strategies carotid endarterectomy and carotid surgery. The Cochrane Collaboration's stroke database was also searched,10 but there was no systematic hand searching of journals. The reference lists of all papers identified electronically were searched.
Articles were included in this review if they fulfilled the following criteria: (a) prospective or retrospective (case note review) study; (b) reported numbers of strokes and deaths occurring within 30 days of carotid endarterectomy (or similar time period); (c) endarterectomy performed for symptomatic stenosis, asymptomatic stenosis, or a combination of the two, but not explicitly for acute stroke; (d) operative risks defined per operation rather than per patient in studies in which some patients underwent bilateral endarterectomy; (e) operative risk stratified according to one or more clinical or angiographic characteristics assessed before surgery; and (f) no evidence of a systematic policy for patients with different characteristics to be operated on by different surgeons or at different institutions. Unpublished data on the 1729 patients who underwent carotid endarterectomy as part of the European carotid surgery trial8 9 were included. The North American symptomatic carotid endarterectomy trial continues to randomise patients and no data on surgical morbidity were available.11
The risk of stroke and death was defined per operation. The overall relative odds of stroke and death for one preoperative characteristic versus another were calculated by using the Mantel-Haenszel method.12 Differences between studies in the relation between each clinical characteristic and operative risk were assessed by using the χ2 test for heterogeneity. We used data on the 1729 patients in the European carotid surgery trial to perform multiple regression analysis to identify which of the potential risk factors identified in the review were independent predictors of operative stroke and death. A forward conditional stepwise logistic regression analysis (SPSS for windows, version 6.1) was used. Variables were entered at a significance level of P=0.05 and removed at a level of P=0.1.
A total of 126 studies reporting the risk of stroke and death associated with carotid endarterectomy were identified from our literature search. Six studies contained no useful data and 84 did not stratify operative risk of stroke and death by baseline characteristics. In addition to the unpublished data from the European trial a total of 36 studies fulfilled our inclusion criteria.3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 One study47 was excluded because the data reported overlapped with those of a subsequent study.23 There were 122 major strokes or deaths within 30 days of endarterectomy in the 1729 patients in the European carotid surgery trial.
The clinical and angiographic characteristics for which the relation with operative risk was reported are summarised in Table 1. Five clinical characteristics were associated with a significantly increased operative risk of stroke and death: surgery for cerebral symptoms (stroke or transient ischaemic attack) compared with monocular ischaemia (amaurosis fugax or retinal artery occlusion) (fig 1); female versus male sex (fig 2); age above or below 75 years (fig 3); and systolic blood pressure >180 mm Hg and peripheral vascular disease (fig 4). There was no significant heterogeneity between studies in the odds of stroke or death for any of these risk factors (Table 1).
Three angiographic characteristics were associated with an increased risk of operative stroke or death: contralateral internal carotid artery occlusion (fig 5); stenosis of the intracranial portion of the ipsilateral internal carotid artery and of the ipsilateral external carotid artery (fig 6). None of these analyses showed significant heterogeneity between studies (Table 1). There was no significant relation between operative risk and the presence of diabetes, angina, recent myocardial infarction, plaque surface irregularity, or a current history of smoking (Table 1). There was no overall difference between the operative risk associated with cerebral transient ischaemic attack versus stroke (fig 7), although there was significant heterogeneity (χ2 14.0; df=6; P=0.04; Table 1). Meta-analysis of the six published studies suggested that stroke was associated with an increased risk (odds ratio 1.46; 95% confidence interval 1.04 to 2.04), whereas the data from the European carotid surgery trial suggested a significantly lower risk (0.62; 0.42 to 0.91).
Of the 14 risk factors in Table 1 which were entered into a multiple logistic regression analysis of the data from the European carotid surgery trial, four were found to be significant independent predictors of operative stroke or death (Table 2).
This is the first systematic review of the clinical and angiographic factors associated with an increased risk of stroke and death from carotid endarterectomy. The analysis was restricted to defining the overall relative odds of stroke and death according to the presence or absence of each potential risk factor in those studies in which the information was available. In view of the clear variation between different studies in the reported risks7 we did not determine absolute risks. Restriction of the analysis to relative odds allowed us to include, when appropriate, data from reports of surgery for asymptomatic stenosis as well as symptomatic stenosis. Even though the reported absolute risks of surgery for asymptomatic stenosis are consistently lower than in patients with symptomatic stenosis,48 there is no evidence that the relative effect of clinical characteristics such as age or sex on the operative risk of stroke and death will differ. This is supported by the lack of significant heterogeneity between studies for any of the risk factors for which data from studies of symptomatic patients were combined with data from studies of asymptomatic patients.
Clinical presentation and operative risk
We have shown that the risk of stroke and death from carotid endarterectomy depends on the type of presenting symptoms. The odds of stroke and death in patients with only ocular symptoms (amaurosis fugax or retinal artery occlusion) are less than half those in patients with cerebral transient ischaemic attacks. Although not strictly comparable, the 2.5% (95% confidence interval 1.8 to 3.5) risk of stroke and death in the eight reports of endarterectomy for ocular symptoms alone is similar to the 3.4% (2.4 to 4.3) overall risk derived from 25 studies of endarterectomy for asymptomatic stenosis.48 It is interesting to note that the risk of stroke on medical treatment is also lower in patients with amaurosis fugax than in patients with cerebral transient ischaemic attack.4 5 49
Contrary to expert guidelines on carotid endarterectomy50 51 we found no overall difference in the operative risk of stroke and death in patients operated on for stroke compared with those operated on for transient cerebral ischaemia, although there was significant heterogeneity between studies. We suspect that the increased risk in patients operated on for stroke in the published studies may well have been due to the inclusion in some of these studies of patients with stroke in evolution or very recently completed stroke. In the European carotid surgery trial, in which patients operated on for stroke had a significantly lower operative risk than patients operated on for transient cerebral ischaemia, surgeons were advised not to operate until at least one month after the occurrence of a stroke. The suggestion that surgery for transient ischaemic attacks has a lower operative risk than surgery for stroke may also be due to a failure to distinguish between ocular and cerebral transient ischaemic attacks. The real dichotomy is not between transient ischaemic attack and stroke but between ocular and cerebral ischaemia. This should be reflected in future guidelines. The expected operative risks for endarterectomy for asymptomatic stenosis and ocular ischaemia are in the region of 2-4%, whereas the risk of endarterectomy for cerebral ischaemia is about double this.
Sex and operative risk
Operative morbidity and mortality in coronary artery surgery is higher in women than in men.52 We therefore examined the effect of sex on the risk of stroke and death from carotid endarterectomy. In the seven studies which reported data, the overall odds of stroke and death were increased by 44% in women. A significantly increased operative risk in women was mentioned in two further studies, but insufficient data were given to allow inclusion in the review.53 54 The increased risk reported in women might be due to publication bias—authors reporting data only if there is an interesting trend. No studies, however, reported a significantly increased operative risk in men. The higher operative risk observed in women compared with men might be accounted for by differences in other factors, such as age, presenting symptoms, or coexisting illness. Multiple regression analysis of the data from the European carotid surgery trial, with correction for the other variables listed in Table 1, however, failed to reduce the association between female sex and operative risk (Table 2). It is unclear why women should be at greater operative risk than men. Possibly the fact that the carotid arteries in women are, on average, about 40% smaller than in men makes the operation technically more difficult.
Other clinical and angiographic characteristics and operative risk
The risk of stroke and death due to endarterectomy was increased in patients aged 75 and over, those with systolic hypertension, and those with peripheral vascular disease. The association of peripheral vascular disease with increased operative risk was reported in another study,55 but no data were given. With regard to hypertension, it is unclear whether treatment before surgery would be beneficial. The increased risk of surgery in patients aged 75 years and over is not large, and age should not be regarded as a contraindication to surgery. There was no evidence that angina, recent myocardial infarction, diabetes, or smoking are associated with an increased risk of stroke or death due to endarterectomy, although more data are required before any useful conclusions can be drawn. They may, of course, be associated with other postoperative morbidity, such as myocardial infarction or chest infection.
Given that patients with irregular stenoses are at increased risk of ipsilateral ischaemic stroke on medical treatment,56 57 the observation that angiographic irregularity of the plaque surface is not clearly associated with an increased risk of stroke with surgical treatment is important. Patients with irregular or ulcerated plaques are probably, therefore, particularly likely to benefit from endarterectomy, although more data on the relation between plaque surface morphology and operative risk are required.
There was a trend towards an increased operative risk with occlusion of the contralateral internal carotid artery in 12 of the 14 studies (see fig 5). Because of the relatively low prevalence of contralateral occlusion in the studies reviewed (407/8626; 4.7%), however, this reached significance in only two. This is also likely to account for the lack of a significant association between contralateral occlusion and operative risk in the multiple regression analysis of the data from the European carotid surgery trial.
Limitations of this study
The studies included in the systematic review were of varying methodological quality. Some were retrospective and only a minority of the remainder had independent assessment of outcome by a neurologist. Although this may have led to an underestimation of the absolute operative risk in some studies,58 we have assumed that this will not bias the relative odds of stroke and death due to surgery with respect to particular risk factors. We have also assumed that the “effect” of the risk factor on operative risk is relatively independent of other treatments—such as anticoagulation, patching, or shunting—the use of which may have varied between studies. Our previous experience of similar analyses suggests that these assumptions are reasonable. In a systematic comparison of the operative risk of carotid endarterectomy for asymptomatic compared with symptomatic stenosis we identified 25 studies in which the operations had been carried out by the same surgeon or in the same institution.48 Even though there were major differences between the studies in terms of methodology and other treatments given, a trend towards a reduced operative risk associated with asymptomatic stenosis was found in 24 (96%) of the studies, and there was virtually no quantitive heterogeneity of effect between studies (χ2 13.6; df=24; P=0.96). In the present study, one of the meta-analyses showed borderline significant heterogeneity (see Table 1). Given that this study includes 12 meta-analyses, however, one significantly heterogeneous result would reasonably be expected by chance. The remainder of the analyses showed relatively little heterogeneity, suggesting that our assumptions were reasonable.
As discussed above with reference to the increased operative risk reported in women, publication bias is a further potential problem with the study. It is quite possible that some of the studies we included looked at the interaction of several risk factors with operative risk but published only those which were “interesting” or significant. This should, however, introduce bias in both directions—that is, analyses are probably equally likely to be published whether or not a particular risk factor indicates an abnormally high or an abnormally low operative risk. Funnel plots of the analyses reported in our paper do not show any obvious skewing suggestive of publication bias.
The associations between clinical and angiographic characteristics and the risk of stroke and death from endarterectomy in the systematic review cannot, of course, be corrected for the possible confounding effect of other variables. The risk factors identified in the multiple regression analysis of the European data, however, are independent to the extent that the association with operative risk has been corrected for the other potential risk factors in Table 1. This does not, of course, mean that the associations are necessarily causal. For example, the association between peripheral vascular disease and an increased operative risk is probably due to some confounding factor which has not been measured in this study. In terms of identifying patients at high risk of operative stroke and death, however, it is association rather than causation that is important.
To improve the cost effectiveness of carotid endarterectomy patients should be selected on the basis of their likely risk of stroke without surgery as well as their risk of stroke and death from endarterectomy. Although we have identified risk factors for increased surgical risk, it does not follow that the presence of these risk factors will necessarily reduce the possible absolute benefit of endarterectomy. Certain characteristics may be associated with an increased risk of stroke on medical treatment as well as surgical treatment and will not therefore be associated with reduced absolute benefit from surgery. Other characteristics, such as female sex, seem to indicate an increased surgical risk but are not related to the risk of stroke on medical treatment alone and may therefore be useful in identifying patients with less to gain from endarterectomy.
We have defined several clinical and angiographic characteristics that are associated with an increased risk of stroke and death from carotid endarterectomy. These observations may help clinicians to estimate the possible operative risks for individual patients and will also facilitate more meaningful comparison of the operative risks of different surgeons or different institutions by allowing some adjustment for differences in case mix. The risk factors derived from the systematic review and those derived from analysis of the data from the European carotid surgery trial, however, need to be validated on a large independent dataset before they can be used routinely in clinical practice.
We acknowledge the European Carotid Surgery Trial Collaborative Group for the collection of the data from the European carotid surgery trial used in this study.
Funding: PMR and JS were funded by the Medical Research Council.
Conflict of interest: None.