Gender-based outcomes and the etiology of sex-related mechanisms in patients treated with CEA are a topic of considerable debate in the recent vascular literature [
34,
35,
38,
39,
64,
71,
77,
88]. Till now, there are no generally accepted and obliging guidelines regarding the preoperative selection of CEA patients subjected to the gender [
59,
89,
90].
RCT data
Subgroup analysis of the large CEA-randomised trials have suggested that the benefit from CEA would be less in women compared to men for both, symptomatic and asymptomatic carotid stenosis. In particular, in the European Carotid Surgery Trial (ECST) and the Asymptomatic Carotid Surgery Trial (ACST) women appeared to have higher risks of perioperative complications, and in the moderate (50 % to 69 %) stenosis arm of the North American Symptomatic Carotid Endarterectomy Trial (NASCET) women did not appear to benefit from surgery at all. In this subset of female patients CEA was associated with a marginal or reduced long-term benefit combined with at least 2-fold increase in the perioperative complication rate as compared to men. It has been hypothesized, that the lower degree of benefit in women was caused partly by a (slightly) higher operative risk in combination with a lower natural history risk of stroke in women as compared to men [
5‐
7]. However, these assumptions are based on post-hoc subgroup analyses on results of these RCTs and have therefore some inherent limitations. Women comprised only a third of patients in these trials (between 28 and 34 %) and this relatively small sample size has left them underpowered to detect important differences between women and men. In addition to that, selection criteria for these randomized controlled trials may partly explain the discrepant findings and thus limiting their generalizability to the typical endarterectomy population. As a result, these gender-related results are difficult to interpret due to significant design limitations [
5] and due to a lack of adjustment for other important prognostic factors. In the recently published International Carotid Stenting Study (ICSS), women had a higher 120-day event rate for CEA compared to men (7.6 % vs 4.2 %) [
91]. On the contrary, in the Carotid Revascularization Endarterectomy vs Stenting Trial (CREST), women undergoing CEA had a lower periprocedural event rate compared to men (3.8 % vs 4.9 %) [
87].
The present unselected meta-analysis of RCTs revealed that female patients undergoing CEA had a higher combined risk of death and stroke after the intervention than did male patients (Peto OR, 1,162; 95 % CI, 1.067-1.266; P = .001). We analyzed the robustness of the observed outcomes by performing sensitivity analyses. Low heterogeneity among the studies existed (I2 = 15 %), and the likelihood of publication bias was low (P = .69). In the TSA, the cumulative z curve crossed the traditional boundary but not the TSMB, suggesting a lack of firm evidence for an RRI of 20 % in the female group compared with the male group in RCTs for the endpoint under investigation. As a result, the higher perioperative risk of stroke and death after CEA in women observed in large RCTs is likely to be a statistical artifact due to small female patient numbers. In addition to that, because of low absolute difference between the gender related outcomes (Peto OR, 1,162; 95 % CI, 1.067-1.266; P = .001), the clinical significance of this finding is unclear. As well, we found a significant difference in overall perioperative stroke rates for CEA when men were compared with women (Peto OR, 1.204; 95 % CI, 1.073-1.351; P = .002). Sensitivity analysis revealed that there might be a statistically significant difference between the two genders regarding the end-point 30-day stroke rate by using the fixed-effects model, as well (Peto OR, 1.398; 95 % CI, 1.106-1.765; P = .005), but again the clinical significance of this small difference may be narrow. In this sensitivity analysis, even moderate heterogeneity among the studies existed (I2 = 45.8 %) but the likelihood of publication bias was low (P = .73).
It has also been suggested that the RCT results may be confounded by referral bias [
86]. Women are less likely than men to be selected for both, cardiac and peripheral vascular surgery [
92] and recent studies have demonstrated that women are less likely to receive CEA or angioplasty than men in the same situation [
86]. Therefore, it might be possible that those women who actually undergo surgery are at higher risk of complications than those who do not. As a result, it is unclear whether these RCT results can be generalised to the non-trial setting [
93,
94]. It is therefore essential to analyze whether the RCT results of CEA are also seen in routine clinical practice.
Non-RCT data
The case series reporting combined 30-day stroke and mortality rates subjected to the sex as their primary aim found CEA as safe and beneficial in women as in men when applying the random-effects model (Peto OR, 1.202; 95 % CI, 0.925-1.561; P = .168). However, when applying the fixed-effects model for the same subset of data, there was a statistically significant difference between the two genders regarding the combined end-point 30-day mortality and stroke rate (Peto OR, 1.299; 95 % CI, 1.089-1.548; P = .004). Moderate heterogeneity among the studies existed (I2 = 32.9 %), and the likelihood of publication bias was low (P = .23). However, in TSA the cumulative z curve crossed the traditional boundary but not the TSMB, suggesting a lack of firm evidence for an RRI of 20 % in the female group compared with the male group in case series for the endpoint under investigation and gender differences as their primary aim. Thus, although those studies have shown no difference in perioperative stroke and mortality between men and women in the random effects model, criticism may be pointed to the size and power of those studies, raising the possibility of a type II error as demonstrated by TSA. The same was true when analyzing the endpoint 30-day stroke in case series with sex-related differences as a primary aim, with no difference in the association between sex and the end-point when applying the random-effects model (Peto OR, 1.322; 95 % CI, 0.922-1.895; P = .129), but a statistically significant difference between the two genders regarding the end-point 30-day stroke rate when applying the fixed-effects model (Peto OR, 1.235; 95 % CI, 1.024-1.490; P = .027). Moderate heterogeneity among the studies existed (I2 = 63.4 %), and the likelihood of publication bias was low (P = .44).
The case series reporting combined 30-day stroke and mortality rates subjected to the gender as a secondary aim suggest that CEA is associated with significantly increased risk for periprocedural death and stroke in women when compared with men when applying the random-effects model and the fixed-effects model, respectively (Peto OR, 1.390; 95 % CI, 1.148-1.684; P = .001, and Peto OR, 1.400; 95 % CI, 1.180-1.662; P < .000, respectively). Moderate heterogeneity among the studies existed (I2 = 13.4 %), and the likelihood of publication bias was low (P = .82). However, the cumulative z curve crossed the traditional boundary but not the TSMB, suggesting a lack of firm evidence for an RRI of 20 % in the female group compared with the male group in case series for the endpoint under investigation and gender differences as a secondary aim. Thus, although those studies have shown a difference in perioperative stroke and mortality between men and women in the random-effects and the fixed effects model, criticism may be pointed to the size and power of those studies, raising the possibility of a type II error as demonstrated by TSA. The same was true when analyzing the endpoint 30-day stroke in case series with gender differences as a secondary aim, with a significant difference in the association between sex and the end-point when applying the random-effects model (Peto OR, 1.403; 95 % CI, 1.052-1.871; P = .021) and the fixed-effects model (Peto OR, 1.403; 95 % CI, 1.052-1.871; P = .021), as well. No heterogeneity among the studies existed (I2 = 0 %), and the likelihood of publication bias was low (P = .71).
Databases
We found no difference in overall perioperative stroke rates and combined death and stroke rates for CEA when men were compared with women in database analyses. In extensive sensitivity analyses, we demonstrated the robustness of all observed outcomes under investigation. Low heterogeneity among the studies existed, and the likelihood of publication bias was low for all comparisons. Although registries and state-wide databases lack the granular details of patient demographics, comorbidities and procedure preferences, they represent large diverse populations without the institutional selection bias. Therefore, presumably, these databases reflect routine real-world medical practice as compared to databases from randomised controlled trials that usually include tertiary care and university centres only with carefully defined patient selection criteria and practitioner credentialing. These findings probably suggest that medical, daily, population wide practice is rather different from that in large centres.
Meta-regression analysis investigated potential effects of publication date of each study, age, hypertension, diabetes mellitus, coronary artery disease, peripheral artery disease, dyslipidemia, smoking status, and symptomatic or asymptomatic carotid disease on perioperative adverse events associated with CEA subjected to the gender. An interesting finding of the meta-regression analysis is that only the category “year of publication” and “symptom status” were significant confounders for the log odds ratio for stroke and mortality in male and female CEA patients. This essentially means that in older studies, the difference in stroke or mortality is large, becoming less as the years pass, and this reduction in difference is statistically significant. These results are in concordance with the study by Rockman et al. [
64]. When stratified by the presence of preoperative symptoms, asymptomatic male and female patients undergoing carotid intervention had a nearly identical rate of postoperative stroke and in-hospital death. However, symptomatic women undergoing carotid artery interventions had a significantly higher rate of postoperative stroke than symptomatic men (3.8 % vs 2.3 %,
P = .03). The clinical significance of this finding is unclear. It is possible, of course, that symptomatic female patients would be at a higher risk for future stroke if no intervention were performed and still benefit from intervention vs medical management.
Own results
A former meta-analysis of the existing literature performed in 2005 by Bond et al. [
95] found that women undergoing CEA did have a higher rate of operative stroke and death than men (odds ratio, 1.31; 95 % confidence interval, 1.17-1.47,
P < .001). In the present meta-analysis, the effect of sex on the operative risk of CEA in case series was consistent with those observed in the RCTs. Given the potential concerns about the generalisability of observations made in trials to routine clinical practice, it was important to determine whether the increased operative risk of stroke and death in women observed in the trials of CEA were likely to be seen in routine clinical practice. In contrast to the results presented by Bond et al. [
95], we have shown that the effects of sex on the operative risk of CEA in published series from routine clinical practice are not consistent with those observed in the RCTs and even differ between cases series with gender considerations as primary aim and those with gender aspects as a secondary aim and database analyses. Whereas the unselected overall meta-analysis, and the meta-analysis of case series with gender aspects as a secondary outcome showed a significantly increased risk for 30-day stroke and death in women compared to men, meta-analysis of databases and case series with sex-related outcomes as a primary aim demonstrated no increase in operative risk of stroke and death in women compared to men. As a result, the findings of reports in which the gender association was the primary subject of study were highly consistent with those in large databases in which the gender observations was one of many associations reported. Our unselected analysis found significant differences in overall stroke and mortality outcomes between women and men after CEA. In addition, there were also a differences found in stroke and mortality among asymptomatic and symptomatic patients from both sexes. These results do not support the generalisability of the analyses of the overall effects of CEA from the unselected study data to routine clinical practice.
Over the decades 1980 and 2015 optimal medical treatment has been changed tremendously. There is moderate strength of evidence among three quality-A randomized controlled trials (RCTs) (the Veterans Affairs Cooperative Study [VA], the Asymptomatic Carotid Atherosclerosis Study (ACAS), and the Asymptomatic Carotid Atherosclerosis Trial [ACST]) that carotid endarterectomy (CEA) and best medical therapy (BMT) can reduce the risk of ipsilateral stroke as compared with best medical therapy alone, which was demonstrated by all three trials. However, the results from these trials are not applicable to contemporary clinical practice, as they do not compare CEA with contemporary best medical therapy and under conditions of real-world adherence and persistence, respectively. Surgeons in contemporary clinical trials with up-to-date best medical treatment have also achieved CEA periprocedural death and stroke rates lower than those in pivotal trials. For example, in the Carotid Revascularization Endarterectomy vs. Stenting Trial (CREST), the death/stroke rates for symptomatic patients was 3.2 % and for asymptomatic patients was 1.4 %. To date, there is no RCT that has analysed the impact of contemporary best medical treaetment of carotid artery stenosis subjected to the sex as a primary aim of the study.
In the present study, among the 24 studies [
2,
10,
11,
56,
59,
60,
62‐
67,
69,
72‐
79,
85‐
87] published after the year 2004 with a contemporary best medical treatment and reporting combined 30-day stroke and mortality rates there was no difference in the association between sex and the combined end-point when applying the random-effects model as well as when using the fixed-effects model. In addition to that, TSA showed that confidence can be put into these results. Although, the studies published after the year 2004 represent a more contemporary management of patients with carotid artery stenosis, the interpretation of these results regarding the impact of best medical treatment alone on these sex-stratified outcomes should be done with caution.
Possible reasons for gender differences
The reasons for the postulated perioperative risk difference remain speculative and the overall evidence for outcome differences by sex-specific characteristics is limited in the literature. Potential explanations for higher surgical risks in women include the older age of onset of cerebrovascular disease in women [
96]. Other explanations for the sex disparities in benefit from CEA may be attributed to further reasons listed in Table
2 [
97‐
105].
Table 2
Possible reasons for gender differences for carotid endarterectomy
Epidemiology | | |
| Older age of onset of cerebrovascular disease in women | |
Plaque characteristics | | |
| Higher rates of carotid artery stenosis, | |
Lower and more stable plaque burden for the same degree of carotid artery stenosis in females compared with males |
Anatomy | | |
| Female carotid arteries are higher- velocity vessels with increased outflow/inflow ratio, | |
Women tend to have atherosclerotic plaque relatively localized and mainly distributed in the common carotid as opposed to the proximal internal carotid artery, usually seen in men |
| Smaller ICA size in women that might in turn lead to a higher incidence of early (immediate thrombosis, postoperative microembolization) and late (recurrent stenosis, ipsilateral stroke) postoperative complications | |
Surgery | | |
| Higher surgical risk in women | |
Pathophysiology | | |
| Lower cerebrovascular reserves in women, as cerebrovascular reactivity after hypercapnia was found to be more impaired in postmenopausal women compared with men of the same age | |
Limitations
Although we believe that our results are likely to be valid, our study does have some potential shortcomings. First, the studies included in the review were of varying methodological quality. Some were retrospective and only a minority of the remainder had independent assessment of outcome by a neurologist. However, although the absolute operative risk will therefore have been underestimated in some studies, this should not have biased the within-study relative odds of stroke and death due to surgery by sex. However, the reliability of the meta-analyses of the within-study comparisons is supported by the consistent results, with very little statistical heterogeneity between studies, in the present review. The same argument applies to the fact that the use of ancillary treatments, such as the use of patching, shunting or local anaesthetic, will also have varied between studies. Secondly, publication bias is a potential problem with analyses of published data. It is possible that some of the studies looked at the interaction of several risk factors with operative risk, but only published those that were ‘interesting’ or statistically significant. However, funnel plots did not show any obvious skewing suggestive of publication bias (data not shown) and the lack of heterogeneity between studies indicates that selective reporting of extreme results (either associations with low operative risk or high operative risk) was uncommon. Adequate power is difficult to achieve in institutional or even multicenter studies to make meaningful comparisons of rare events, but our analysis has an advantage in that a large patient cohort was used to calculate pooled outcome estimates, such as mortality and stroke. As in other meta-analyses, given the lack of data in each trial, we did not adjust our analyses for medications used during and following the procedure. Although detailed sensitivity analyses on many variables were undertaken, given heterogeneity in the study protocols, clinically relevant differences could have been missed and are best assessed in a meta-analysis of individual patient data. The subgroup analyses might suffer from multiple testing. As a result, the results of the sensitivity analyses are best described as secondary and hypothesis generating only. In addition to that, our report is limited by the heterogeneous groups of patients entering the meta- analysis models. No adjustments for differences in clinical characteristics of the study populations, such as presenting symptom status and atherosclerotic comorbidity, could be made. Further more, each of these studies reported the results of operations performed by multiple surgeons and, in the latter case, in multiple institutions. As a result, there was no standardization of the surgical approach with respect to the method of anesthesia, the use and method of cerebral monitoring, the use of an indwelling shunt, closure of the arteriotomy with a patch, and other factors. These variables were also not controlled in the ACAS [
5] and could potentially confound an analysis of the impact of gender on surgical outcome.