Women with severe symptomatic carotid stenosis were found to have a significantly lower prevalence of plaque hemorrhage than men, independent of other major vascular risk factors. Women with ≥70% stenosis were at lower risk of recurrent ischemic events than men, but this difference was found to be at least partially mediated by the lower prevalence of MRI PH.
We found an estimated twofold lower prevalence of MRI PH in female patients with symptomatic carotid stenosis of ≥50% with an estimated fourfold reduction when adjusting for other known vascular risk factors. Our findings are generally well in line with a large histological study 
and several MRI studies 
but conflict with other reports 
. This may be explained by true differences between studied populations or technical differences in determining MRI PH. Hence we conducted a meta-analysis which confirmed an approximately twofold higher prevalence of PH in men. Moreover, we did not observe significant heterogeneity between studies suggesting no substantial effect from different techniques. Some negative studies may have been underpowered. Interestingly, MRI PH studies that did not show significant sex difference reported generally much lower PH prevalence compared to the histological studies 
. Indeed, a likely type II error may also explain the lack in our study to demonstrate male predominance of PH in the moderate stenosis subgroup. This subgroup contained fewer PH+ patients and only revealed a similar trend. The nature of this sex difference remains unclear, but may relate to more efficient repair as more fibrous or fibroatheromatous plaques are found in women in histology studies 
. Also carotid plaques appear to heal faster in women than in men which may explain why the benefit from CEA decreases rapidly in women after 2 weeks 
We found a lower incidence of any recurrent ischemic symptom in women with 15% of compared to 30% of men experiencing a recurrent ipsilateral AmF, TIA or stroke during clinical follow-up. This is well in line with findings in a larger group of patients with similar symptomatic carotid disease, where the five year cumulative risk of stroke was significantly lower in women 
. Importantly, when we stratified the groups according to MRI PH status, incidence rates of recurrent events became similar for PH+ men and women, and for PH- men and women suggesting that the lower PH prevalence in women may mediate their observed lower recurrent event rate compared to men.
To test a possible mediation effect of PH, we undertook repeated multivariate Cox regression survival analysis that was limited to the subgroup of patients with severe carotid disease for whom we could establish lower prevalence of PH in women compared to men. We found female sex predicts a longer event-free survival time compared to men controlling for time from initial symptom to MRI and age. This effect was abolished when adding MRI PH, itself a significant predictor of shorter event-free survival in the model. Thus, prevalence of carotid plaque hemorrhage can be inferred to mediate at least partially the sex differences in the risk of recurrent ischemic events.
Sex differences in the prevalence of PH may not be the only mediating factor of lower recurrence risk in women. In fact, at the time of presentation, women were older, had higher degrees of stenosis and were less likely to smoke than men. Amongst these factors, only lower rates of smoking could plausibly exert a mediation effect on lower risk of recurrent events. Smoking is an established vascular risk factor and there is good epidemiological evidence for lower prevalence of smoking in older women 
. Indeed, adding smoking in the multivariate regression model abolished the sex effect on event-free survival, but smoking status itself failed to reach significance (p
0.08) in predicting recurrent events. This does not allow ruling out a partial mediation effect from smoking; nevertheless, if contributing this would be a weak affect that is also clinically irrelevant as encouragement to discontinue smoking will be offered to all patients.
There are several limitations to this study; the study was a pooled retrospective analysis across several studies. Nevertheless, individual studies were prospective, consecutive and with high recruitment rates. In addition, standardised protocols were deployed and we are not aware of conceivable potential for bias between sexes. The small number of observed events limited the power of the study and the negative result for smoking has to be cautiously interpreted. The follow-up period was variable and relatively short as delay to CEA was only determined by clinical considerations and logistical constraints of clinical services. These factors are, however, unlikely to have introduced a bias and in fact we did not observe significant sex differences in time from symptom and follow-up period.
We used a single MRI contrast as marker of PH for this study and found 64% prevalence of PH. This is higher than previously reported from multiple MRI sequences, but in very good agreement with large histological series 
confirming the previously reported high accuracy of this MRI technique