Long-term cognitive and neurovascular changes after carotid endarterectomy

Background: Carotid endarterectomy (CEA) has been associated with both cognitive decline and improvement, but the underlying neurovascular mechanisms are unclear. The aim of this study was to investigate the relationship between neurovascular indices and cognitive changes after CEA. Methods: We studied 55 patients with severe (  70%) symptomatic or asymptomatic carotid stenosis before and six months after CEA. A wide array of neuropsychological tests was arranged in eight cognitive domains and cognitive functions specific to hemisphere ipsilateral to operation. Differences in cognitive performance between patients and 38 matching healthy controls were studied with linear mixed models. Neurovascular functioning and microembolic signals were assessed with transcranial Doppler ultrasound of the middle cerebral artery. Associations between neurovascular indices and cognitive change were assessed with linear regression analyses. Results: On group level, the CEA patients improved more than controls in working memory, whereas no cognitive deterioration was detected. Also on individual level, improvement was most frequently observed in working memory. Worse preoperative cerebrovascular reactivity was related with improvement in cognitive functions of the ipsilateral hemisphere. Low preoperative pulsatility index was associated with improvement in executive functioning and ipsilateral cognitive functions. Poorer preoperative blood flow velocity associated with improvement in complex attention. Microembolic signals were rare. Conclusion: The present findings suggest that CEA may have beneficial long-term effects on cognition. These effects may specifically involve patients with impaired preoperative circulatory adaptive mechanisms.

Carotid stenosis has been associated with a degree of cognitive deterioration even in asymptomatic patients (1,2).The commonly suggested underlying mechanisms are hemodynamic impairment caused by the stenotic lesion as well as continuous silent embolization from the atherosclerotic plaque, with more evidence supporting the former (1,3).The protective mechanisms against hypoperfusion caused by carotid stenosis are both anatomical and physiological: redistribution of blood flow through collateralization, which is dependent on individually variable anatomy, and physiological compensation within the vasculature through autoregulation.Autoregulation refers to changes in vascular resistance, mainly by arterial contraction or dilatation, as a response to varying blood pressure (4).When carotid stenosis starts to diminish blood flow, cerebral arteries adapt to the lowered blood supply and pressure by reducing vascular resistance.
Transcranial Doppler ultrasound of cerebral arteries (TCD) ( 5) is an accessible noninvasive bedside technique which can provide versatile information on effects of carotid stenosis on brain vasculature.The middle cerebral artery (MCA) mean blood flow velocity (MFV) corresponds roughly to changes in absolute cerebral blood flow (6,7) which the progression of carotid stenosis will ultimately decrease.The pulsatility index (PI) is a measure of vascular resistance distal to examined artery, calculated as the difference between peak systolic velocity (PSV) and minimal diastolic velocity (EDV) divided by MFV (PI = (PSV − EDV)/MFV) (8,9).Both high and low PI measures may be considered as indicators for disorders.On the one hand, increasing arterial stiffness results in increment of vascular resistance and PI values (10).On the other hand, the brain may try to reduce vascular resistance to compensate for inadequate circulation e.g., in the case of carotid stenosis.Thus, low PI values may correspond to hypoperfusion of the brain.Cerebrovascular reactivity (CVR), i.e., the vasomotor response to changes in gaseous blood levels, is another way to assess hemodynamic reserve of the brain.It is usually measured with either the breath-holding index (BHI; reactivity to breath-holding for at least 30 seconds) (11,12), carbon dioxide (CO 2 ) reactivity (reactivity to inhaled CO 2 ) (13), common carotid artery compression (14) or reactivity to acetazolamide administration (Diamox test) (15,16).If cerebral arteries are maximally dilated due to constant decrease in cerebral blood flow, cerebrovascular reactivity will become exhausted (11).TCD is also used to record microembolization from the atherosclerotic carotid plaque to cerebral circulation, which has been shown to associate with stroke risk in carotid stenosis patients (17,18).
The standard treatment of symptomatic carotid stenosis is carotid endarterectomy (CEA) which is well established for stroke prevention (19)(20)(21).However, the reversibility of cognitive decline and prospect of cognitive change are less clear, as reviewed by several authors (22)(23)(24)(25)(26)(27)(28)(29).Hence, carotid endarterectomy is not recommended for the prevention of cognitive impairment (29).There is some evidence from earlier studies that TCD indices of arterial flow, autoregulation and cerebrovascular reactivity may at least partly explain why some patients show cognitive benefit from CEA and others do not (30)(31)(32)(33), but in all studies these associations have not been found (34).We are aware of only one study examining the associations between preoperative microembolization and postoperative cognitive change after CEA, with no significant associations found (35), and no studies investigating postoperative change in microembolization in the context of cognitive change after CEA.In sum, cerebrovascular measures of cerebral blood flow, autoregulation and cerebrovascular reactivity seem potential candidates to explain and predict postoperative cognitive changes after CEA, but more systematic understanding of how cerebrovascular functioning contributes to domain-wise cognitive performance in a steady phase after CEA is still lacking.
We studied neurovascular functioning together with a comprehensive cognitive assessment before and six months after CEA in patients with a severe carotid stenosis to evaluate their significance in domain-specific cognitive change.The primary outcome measure was domain-wise cognitive change between pre-and postoperative assessments.A set of neurovascular parameters served as secondary outcome measures.Based on our earlier findings (36), and considering the vulnerability of especially anterior brain regions in case of carotid hypoperfusion, we hypothesized that cognitive domains most dependent on frontal networks would be the first to show association with hemodynamic indices.

2
Materials and Methods

Setting
This study was performed as a prospective substudy of the Helsinki Carotid Endarterectomy Study 2 (37) -Brain and Eye Substudy (HeCES-BEST).The data was collected in Helsinki University Hospital, Helsinki, Finland between March 2015 and December 2018.The study was approved by the local ethical committee and performed according to the ethical standards of the Declaration of Helsinki.

Participants
The number of participants in each assessment is shown in Figure 1.We studied 58 consecutive, independent Finnish-speaking elective carotid endarterectomy patients exhibiting severe (70%) carotid stenosis in computed tomography angiography (CTA) according to North American Symptomatic Carotid Endarterectomy Trial (NASCET) criteria (38).Both asymptomatic and symptomatic (amaurosis fugax, reduced visual acuity, or transient ischemic attack; TIA) patients were included.We excluded patients with evidence for cerebral stroke (focal symptoms or signs of infarct on computed tomography scan; CT) within six months to avoid the possibility that observed cognitive change after CEA would be related to rehabilitation from stroke rather than CEA per se.Other exclusion criteria were major psychiatric diseases requiring continuous treatment, substance abuse, contraindication to magnetic resonance imaging (MRI) or Mini-Mental State Examination (MMSE) score <25.Of the original study population, three subjects missed the six-month assessment and were excluded from the final study population: one patient was excluded because of postoperative cerebral stroke, one failed to respond to contact and one declined.Thus, the final study population comprised 55 patients.All participants gave written informed consent.
The patients were examined prior to CEA (range 1-79 days, median 2 days; the operation of one subject was postponed because of hematological illness) and on average six months (range 131-273 days, median 181 days) after CEA with neuropsychological assessment, TCD ultrasound of cerebral arteries and a three-tesla MRI of the brain.
The control group consisted of 38 healthy nonmedicated subjects with no history or signs of potential cardiovascular, neurological, inflammatory, or metabolic disease, major psychiatric morbidity or substance abuse.The controls, matched with the patients for age, gender, education level and occupation level, performed neuropsychological assessments at same intervals as the patients (range 159-222 days between assessments, median 180 days).In control subjects, MRI and TCD assessments were performed only at baseline.Carotid ultrasonography was used to rule out carotid stenosis in controls.

Surgical procedure
CEA was performed with standardized routine by experienced vascular surgeons, mostly by one of the authors (PV).Local anesthesia was preferred.Shunting was performed if the patient developed symptoms in local anesthesia or there was a significant drop in the TCD velocity in the general anesthesia patients.NIRS and TCD were used as cerebral monitoring on top of local anesthesia.CEA routinely lasted 60-90 minutes and internal carotid artery closure times varied between 15 and 30 minutes.Patch closure was the preferred technique.Target blood pressure value was above systolic 170 mmHg during closure and below 150 mmHg upon opening.As a completion control, we used transit time flow measurement and selective ultrasound control.There were no perioperative or 90day complications among the included patients.Further details of the CEA procedure have been described earlier (37).

Baseline characteristics
We scored education into three levels according to the Finnish educational system: basic level (compulsory education; 6-9 years), middle level (vocational training, matriculation examination and/or bachelor's degree; 8-15 years), or higher level (university level master's degree or higher; minimum 16 years).Occupation was scored based on the International Standard Classification of Occupations (ISCO-08) occupational skill levels on a four-code scale, one referring to simple physical or manual routine tasks and four to tasks that require complex problem solving and decision making based on extensive theoretical knowledge in a specialized field (39).Cardiovascular risk factors such as diagnosis of diabetes, hypertension or hypercholesterolemia, statin use (more than three months regular use prior to baseline), smoking (smoker/quit smoking within one year or non-smoker/quit smoking more than one year before), alcohol consumption (Alcohol Use Disorders Identification Test AUDIT) (40), low physical activity (less than recommendations) and body mass index (BMI) were recorded.Depressive symptoms and anxiety were assessed with the Hospital Anxiety and Depression scale (HADS) (41).

Neuropsychological assessment
The comprehensive neuropsychological test battery, shown in

2.6
Transcranial ultrasound TCD of cerebral arteries was assessed at baseline in a subpopulation of 46 patients and 36 controls upon availability of personnel and TCD facilities.Of this subpopulation, 39 patients were assessed also 6 months postoperatively.In control subjects, TCD assessments were performed only at baseline.
TCD examinations were performed according to a standardized routine in the same place in the afternoon hours.The participants were asked to refrain from smoking and coffee in the preceding hours.After the instructions they were allowed to rest for five minutes in supine position before blood pressure and TCD measurements.The cerebral arteries were insonated through the transtemporal window using SonaraTek device (SONARA™ Viasys Healthcare, USA) with a hand-held 2MHz PW transducer by an experienced ultrasonographer using a standardized protocol.The highest flow parameters and PIs in the first segment of each cerebral artery were recorded at the end of the resting period during normal spontaneous inhalation of room air.The BHI test was performed as described earlier (11).Three 30-second breath-holding episodes with 3-minute breaks were recorded, and the mean value of two technically best measurements was used.In 22 patients preoperatively and 21 postoperatively as well as 3 controls the BHI measurements were not successful either due to their inability to perform the 30-second apnea reliably or the relative insufficiency of the transtemporal window, or registrations were not available due to a storage error which prevented their retrieval and blinded re-analysis.
We used the following TCD variables to assess pre-and postoperative hemodynamics of ipsilateral MCA: ipsilateral BHI, ratio of ipsilateral and contralateral MFV and ipsilateral PI <0.8 (30).In addition, contralateral PI in patients with no 70% stenoses in contralateral carotid artery was used as an index of general arterial stiffness and vascular resistance.Postoperative change in hemodynamics was assessed with change scores in ipsilateral BHI and ratio of ipsilateral and contralateral MFV.For the controls, we used the mean of right and left MFV and PI values.
A subpopulation of 30 patients was assessed preoperatively and 26 patients postoperatively (18 patients both pre-and postoperatively) for microembolic signals (MES) upon availability of personnel and TCD facilities, with an ambulatory registration of ipsilateral MCA flow with a portable device (Atys TCD-X, Atys Medical, Soucieu-en-Jarrest, France).We excluded 3 measurements due to analyzable recording times below 30 minutes, resulting in 28 preoperative and 25 postoperative measurements.Median analyzable recording time was 1 h 57 min (range 45 min -3 hours 8 minutes).

Statistical analyses
Statistical analyses were performed with IBM SPSS 25 and 28.We standardized the neuropsychological test results relative to the controls' baseline performance and inverted timed zscores so that negative z-scores always indicated inferior cognitive performance.Next, we averaged the standardized results within each cognitive domain.All analyses of cognitive performance were carried out within each cognitive domain.P-values below 0.05 were counted significant.
We used Mann-Whitney U tests, independent samples t-tests, χ² tests, Fisher's exact tests, Wilcoxon signed-rank tests and McNemar tests for univariate analyses.Differences between the patient and control groups as well as interactions between group and measurement (baseline and six months) in cognitive performance were assessed with linear mixed models adjusting for age, sex and education class.Beforehand, reflection and square root transformation were applied to processing speed and reflection and logarithmic transformation to complex attention.We chose compound symmetry as repeated effect covariance structure based on Bayesian information criteria.Significant interactions were assessed post hoc between groups within both measurements with Bonferroni corrected pairwise comparisons based on estimated marginal means.
To allow for normal variation and practice effects, we used the reliable change index (RCI) (54)(55)(56) to assess postoperative cognitive performance within the patient group.Postoperative cognitive decline (POCD) and improvement (POCI) were defined based on RCI with a cut-off criterion of ±2.Associations between cardiovascular risk factors, TCD variables and domain-wise RCI were evaluated with separate linear regression models controlling for symptomatic stenosis, age, sex and education class.Beforehand, square root transformation was applied to executive functioning and logarithmic transformation to complex attention.We calculated local effect sizes of statistically significant models with Cohen's F 2 based on the R squares of the models (57,58).Both linear mixed models and regression models were first performed including all patients and then excluding the patients who were bilaterally operated during the follow-up time.

Baseline characteristics
Baseline characteristics of the patient and control groups are shown in Table 2. Age, gender, occupational level or education class of the patient population did not differ from the controls (Table 2).During the research process, 28 controls were diagnosed with hypercholesterolemia.Of the patients, 24 (44%) were asymptomatic and 31 (56%) symptomatic.Symptoms included amaurosis fugax in 19 (35%), hemispheric TIA in 10 (18%) and reduced visual acuity due to neovascular glaucoma or branch retinal arterial occlusion in 2 (4%) patients.For practical and logistical reasons, the most urgent patients could not be included in the study.The median delay from index symptom to surgery in symptomatic patients was 12,5 days.There were no significant differences in age, sex, education class, occupation level, baseline cognitive performance or postoperative cognitive change in any cognitive domain between asymptomatic and symptomatic patients (all p-values >0.05, Mann-Whitney U tests, χ² tests and Fisher's exact tests).
The operation was performed on the left side for 33 (60%) and on the right side for 22 (40%) patients.Of the patients, 8 had  70% contralateral stenosis and 4 underwent contralateral CEA during the 6 months follow-up time.The bilaterally operated patients did not differ significantly from the other patients in age, sex, education class, occupation level, any cognitive domain at baseline or in domain-wise postoperative cognitive change (all p-values >0.05, Mann-Whitney U tests and Fisher's exact tests).
The three patients who dropped out of postoperative assessments were inferior to the other patients in baseline motor dexterity (U=16.00,p=0.01,Mann-Whitney U test) but did not differ from the other patients in age, sex, education class, occupation level or any other cognitive domain at baseline (all p-values >0.05, Mann-Whitney U tests and Fisher's exact tests).

Pre-and postoperative cognitive performance between patients and controls
Cognitive performance of patients and controls is shown in Table 3. Results of linear mixed models assessing interactions in cognitive performance between group (patients and controls) and measurement (baseline and six months) are reported as follows.There was an interaction between group and measurement in working memory (F(1,91)=4.31,p=0.04).Post hoc analysis showed that the patient group performed inferior to control group at baseline (t(108.77)= 0.38, p=0.007) but no longer at six months (t(108.77)= 0.19, p=0.17).When patients who underwent bilateral CEA during follow-up time were removed from analysis, there was a trend in interaction (F(1,87) = 3.12, p=0.08).
There were no other interactions between group and measurement, but there was a significant main effect of group showing inferior performance in patients compared with controls in all remaining cognitive domains:

Neurovascular functioning
Neurovascular functioning of patients and controls is shown in Table 5.To sum up, patients exhibited significantly poorer ipsilateral BHI than controls (U=166.0,p<0.001,Mann-Whitney U test) and there were more patients than controls with low (<0.8)PI at baseline (p=0.04,Fisher's exact test).In addition, the patients with no severe contralateral carotid stenosis had a significantly higher PI value in the contralateral side at baseline than controls (U=361.  1 Difference within patients between baseline and six months. 2Difference between patients' baseline and controls. 3In controls, mean of right and left sides is used. 4Contralateral PI is only analyzed within patients without severe (70%) contralateral stenosis.*p<0.05,**p < 0.01, ***p<0.001.BHI, breath holding index; MFV, mean blood flow velocity; PI, pulsatility index; N/A, not applicable.MES's were detected in only 3 (11%) patients preoperatively and 3 (19%) patients postoperatively.Each patient exhibiting MES's was symptomatic, resulting in 16% symptomatic patients with preoperative MES's and 30% symptomatic patients with postoperative MES's.Due to the small number of patients with MES's, further analyses were not conducted.

Predictors of postoperative cognitive change
There were statistically significant predictors of cognitive change in linear regression analyses in the domains of working memory, complex attention and executive functioning as well as ipsilateral cognitive functions (Table 6).Hence, other cognitive domains were excluded from further inspection.
Looking at cardiovascular risk factors, hypertension was inversely associated with postoperative change in complex attention (=-0.35,p=0.04) and directly associated with postoperative change in working memory (=1.01,p=0.04).The effect sizes in these associations were small (F²<0.15).The observed trend between hypercholesterolemia and change in executive functioning is subject to unreliability due to a very small proportion of patients without hypercholesterolemia (four patients).
Statin use related with postoperative change in executive functioning (=0.30,p=0.001), with a moderate effect size (F²>0.15).After excluding bilaterally operated subjects from analyses, the results remained essentially unchanged, but the association between hypertension and working memory was no longer significant (=0.90, p=0.06).
Regarding neurovascular parameters, BHI at baseline was inversely related with postoperative change in ipsilateral cognitive functions (=-3.20,p=0.01) and there was a near-significant inverse trend between BHI at baseline and postoperative change in working memory (=-2.67,p=0.06), both with large effect sizes (F²>0.35).There was an inverse association between preoperative ratio of ipsiand contralateral MFV and complex attention (=-0.66,p=0.046) and a trend between postoperative change in ratio of ipsi-and contralateral MFV and executive functioning (=0.46,p=0.08), with moderate effect sizes (F²>0.15).Furthermore, there were significant associations between low preoperative ipsilateral PI and greater improvement in ipsilateral cognitive functions (=1.70,p=0.002) and executive functioning (=0.39,p=0.008) with moderate effect sizes (F²>0.15).When bilaterally operated subjects were excluded from analyses, the results remained mostly unchanged, but the associations between BHI and ipsilateral cognitive functions (=-3.04,p=0.08) and preoperative ratio of ipsi-and contralateral MFV and complex attention (=-0.56,p=0.06) no longer reached statistical significance.These results are displayed in Figure 2.

Discussion
This study was set out to explore associations between neurovascular mechanisms and cognition in non-stroke carotid patients undergoing CEA.Our results indicate cognitive improvement particularly in working memory.The findings also corroborate the association between indices reflecting arterial flow parameters, autoregulation and cerebrovascular reactivity and postoperative cognitive improvement (30)(31)(32)(33), with an emphasis on the functions of anterior cerebrum and ipsilateral hemisphere.
CEA patients as a group improved significantly more than controls in the domain of working memory, whereas no group-level deterioration was observed.Working memory was most frequently improved also on individual level, although the frequencies of domain-wise postoperative cognitive improvement were in general rather low, presumably owing to the adopted strict definition of postoperative cognitive decline and improvement.Notably, after excluding the four patients who were operated on bilaterally during the follow-up, there was only a trend of working memory improvement.Taken together, the present results suggest that working memory may be the most sensitive cognitive domain to postoperative changes after CEA, but on group level, this improvement is modest and more pronounced after bilateral operation.
The anterior cerebral cortex, perfused by carotid arteries via the anterior and middle cerebral arteries, subserves both working memory (59) and other complex cognitive functions such as executive functioning (60).Why, then, is working memory in the present study more sensitive to postoperative changes six months after CEA than other cognitive domains?Methodological reasons may in part explain our findings.We used established neuropsychological tests to assess each cognitive domain, but also modern, probably more sensitive computerized tests were used in the assessment of working memory, complex attention and processing speed.Furthermore, although generally considered as tests of working memory, the computerized N-back tests used in this study also test other cognitive abilities, such as attentional and executive control (61,62).Indeed, in our previous work of a historical cohort using only traditional paper-and-pencil assessment methods, postoperative cognitive improvement in working memory was rare, whereas executive functioning was the most improved domain in a follow-up time of one week and three months after CEA (36).In this light and considering the importance of anterior cerebrum in executive functioning (60), it is somewhat surprising that improvement in executive functioning domain was rare in this study.It is probable that inconsistent results are at least partly related to differences in study populations (proportion of symptomatic patients, index symptoms) due to study protocols and changes in surgery indications.Differences in follow-up times can hardly explain deviant findings because it is plausible that with increasing cerebral restructuring after restored carotid circulation, cognition would rather improve than deteriorate in later surveillance after CEA (22).While the six-month interval for follow-up adopted in the present study allows for both passing of perioperative stress and time for cerebral restructuring, it is possible that cognitive improvement could be even more pronounced at longer surveillance.In sum, the present study implicates that at six months after CEA, cognitive improvement is mainly detected in sensitive tests and demanding performance requiring working memory as well as attentional and executive control.
In the present study, only few cardiovascular risk factors showed predictive value on postoperative cognitive functioning after CEA.There were contradictory findings on the associations between hypertension and postoperative cognitive change.Continuous (regularly more than three months prior to baseline) statin use associated with postoperative change in executive functioning with a moderate effect size.This finding corroborates earlier studies showing an association between statin use and lower incidence of postoperative cognitive decline after CEA (63,64) and presumably reflects vulnerability to postoperative decline in patients with sub-optimally treated cardiovascular disease.
Looking at neurovascular functioning, the patients had poorer cerebrovascular reactivity, measured with breath-holding index, than controls at baseline.This result corroborates earlier findings on the association between carotid stenosis and cerebrovascular reactivity (65,66).Furthermore, there was an association between lower preoperative cerebrovascular reactivity and cognitive improvement six months after CEA, notably with large effect sizes.More specifically, impaired cerebrovascular reactivity associated with improved ipsilateral cognitive functions, corroborating previous findings (32), although no longer significantly after excluding bilaterally operated patients.In addition, there was a near-significant trend between impaired cerebrovascular reactivity and improved working memory.There were no significant group-level postoperative changes in patients' cerebrovascular reactivity, contrasting earlier findings (66-69), nor were there significant associations between postoperative change in cerebrovascular reactivity and cognitive functioning, also reported earlier (31).This discrepancy probably relates to Type II error due to restricted sample size in cerebrovascular reactivity measurement.However, even in a limited cohort, our results suggest a rather robust association between preoperative cerebrovascular reactivity and cognitive change after CEA, particularly in bilaterally operated patients.
In contrast to earlier findings (70), there were no significant group-level changes in preoperative mean blood flow velocity (MFV) between patients and controls, nor in ipsilateral MFV after revascularization within patients.However, looking at individual contrasts in MFV between ipsi-and contralateral hemispheres, we observed an association between lower preoperative ipsilateral MFV and improving complex attention, but it no longer reached significance after excluding bilaterally operated patients.There was also a trend between postoperative improvement in ipsilateral MFV and executive functioning.These results are in accordance with earlier findings at the acute phase, one day after CEA (30) and further suggest that a simple measure of cerebral blood flow may have some predictive value on postoperative cognitive functioning still at six months after CEA.
Reduced vascular resistance, measured with ipsilateral pulsatility index (PI), was common among CEA patients and related with cognitive improvement.Ipsilateral PI was low more often among patients than controls at baseline and increased postoperatively, in line with previous findings (70)(71)(72).Low preoperative PI associated with postoperative improvement in ipsilateral cognitive functions and executive functioning with a moderate effect size.Earlier, an association between low preoperative PI and cognitive improvement has been found at the acute phase after CEA (30).Our results further implicate that the association between reduced preoperative ipsilateral vascular resistance and postoperative cognitive improvement is still prevalent at a stable phase after CEA.
The patients with no contralateral carotid stenosis had a significantly higher contralateral PI value than controls at baseline, indicating increased vascular resistance related to general vascular stiffness (10).The finding is not surprising given the differences in atherosclerotic disease and cardiovascular risk factors between patients and healthy controls.Contralateral PI did not associate with postoperative cognitive change, suggesting that the patients' ability to gain cognitive benefit from CEA is not heavily affected by general arterial stiffness and vascular resistance per se; rather, it is mediated by hemodynamic changes resulting from the stenosed carotid artery.
Microemboli are often proposed as a potential source of cognitive decline in carotid stenosis (73) but associations between microembolic signals and cognitive functioning in asymptomatic carotid patients (74) or preoperative microemboli and cognitive change after CEA (35) have not been found.The paucity of microembolic signals in this study suggests that the hemodynamic changes outweigh their contribution to cognitive change.A larger study would be needed to assess the interplay and synergy between flow parameters and microemboli.
The main limitation of the present study is the restricted sample size that may risk Type II error, i.e., missed associations between potential predicting variables and postoperative cognitive performance.Especially the number of patients undergoing TCD measurements, in particular BHI assessments, was limited.There is also a risk of Type I error due to the exploratory nature of the study with a wealth of uncorrected comparison.Nevertheless, our findings on fairly consistent associations between neurovascular parameters and cognitive performance in a relatively small dataset, with mostly moderate to large effect sizes, implicate their weight in the prediction of postoperative cognitive changes after CEA, overshadowing the predictive power of traditional cardiovascular risk factors.These preliminary results serve guidance to future larger-scale studies towards incorporating vascular physiological parameters among the predictors of domain-wise cognitive change in a stable phase after CEA.
There are also methodological limitations in the neurosonological assessment.End-tidal CO 2 concentrations were not monitored during the breath-holding test, which may add to the variability of the results.Another arguable weakness of the present study is that, in contrast to most of the earlier studies studying hemodynamic changes and cognition in CEA patients (30,33,34), both symptomatic and asymptomatic patients were included in this study.In order to avoid confusing postoperative cognitive change and spontaneous rehabilitation from stroke, patients with radiological or clinical signs of acute/subacute ischemic stroke were excluded.In addition, for practical and logistical reasons, the more urgent patients were naturally excluded.Hence, the patients included in this study (56% symptomatic patients with no recent strokes, median delay from index symptom to surgery 12,5 days) do not represent the usual severity and distribution of CEA indications in our hospital (81% symptomatic patients, delay from symptom to surgery 12 days) (75).However, it must be noted that cerebrovascular changes are more probable in symptomatic patients who may not have proper collateral circulation to compensate for reduced blood supply from the carotid arteries.Our sample size did not allow for separate analyses within symptomatic and asymptomatic patient groups, but we included symptom status, albeit not significantly associated with either preoperative cognitive level or postoperative cognitive change, in statistical models looking for predictive elements of cognitive change.
Notwithstanding these limitations, the present study has considered many of the pitfalls of earlier studies investigating cognitive changes after CEA, pointed out in recent reviews (3,22,76).In contrast to many earlier studies using only screening tests or a limited selection of neuropsychological tests, we used an extensive neuropsychological testing battery organized in multiple well-covered cognitive domains, including tests sensitive for functions of both hemispheres.Other strengths include neuropsychological assessments conducted by a specialized neuropsychologist, comparison with a control group and statistical methods used to account for learning effects and the definition of postoperative cognitive decline and improvement, and a long follow-up time.Moreover, this study presents a comprehensive evaluation of neurovascular parameters in relation to cognitive outcome after CEA.

Conclusions
In conclusion, our study suggests that group-wise cognitive changes at a stable phase, six months after CEA are subtle, with modest improvement of working memory particularly in bilaterally operated patients.Moreover, our findings indicate that neurovascular mechanisms, in particular reduced preoperative cerebrovascular reactivity and vascular resistance, may explain individual differences in postoperative cognitive changes particularly in cognitive functions operated in the ipsilateral hemisphere and anterior cerebrum.Although carotid revascularization cannot in general be recommended based on cognitive expectations, it is possible that patients who benefit cognitively from carotid surgery could be identified by focusing on preoperative neurovascular functioning.Future studies investigating cognitive functioning after carotid revascularization should focus on cognitive functions dependent on frontal networks as well as hemisphere-specific cognitive functions and their relations to preoperative cerebrovascular reactivity and vascular resistance.

Declaration of interest
Declarations of interest: none.

Funding
This work was supported by the Wilhelm and Else Stockmann Foundation, the Instrumentarium Science Foundation, the Dorothea Olivia, Karl Walter and Jarl Walter Perklén's memory Foundation, the Sigrid Jusélius foundation and Helsinki University Hospital governmental subsidy for clinical research.The first author was supported by the Päivikki and Sakari Sohlberg Foundation, the Finnish Cultural Foundation, the Finnish Concordia Fund, the Alfred Kordelin foundation, the Finnish Foundation for Cardiovascular Research and Helsinki University Hospital governmental subsidy for clinical research.

Figure 1 .
Figure 1.Patients and controls in each assesement.TCD, Transcranial Doppler ultrasound of cerebral arteries; MES's, microembolic signals.*Assessed upon availability of personnel and TCD facilities.

Figure 2 .
Figure 2. Unadjusted associations between neurovascular parameters and domain-wise, untransformed postoperative cognitive change.P-values are from linear regression analyses adjusted for age, sex, education class and symptomatic carotid stenosis.BHI, breath holding index; MFV, mean blood flow velocity; PI, pulsatility index.

Table 1
-sequencing of the Flexible Attention Test (FAT) (computerized modification of Trail making B; time to complete) (43,44,50) Number-symbol-sequencing of the FAT (computerized test resembling Trail making B/set shifting; time to complete) (43,44,50) Number-month-sequencing of the FAT (computerized test resembling Trail making B/set shifting; time to complete) (43,44,50) Number-month-sequencing backwards of the FAT (computerized test resembling Trail making B/set shifting; time to complete) (43,44,50) Verbal tests included in ipsilateral cognitive functions for patients with left-sided CEA, 2 visual tests included in ipsilateral cognitive functions for patients with right-sided CEA.
LearningWord list of the WMS-III (12 words, sum of three trials) (45)1Modified Rey visual design learning test (12 drawings, sum of three trials) (51)2Delayed memory Delayed recall of WMS-III Word list (45)1Recognition of the WMS-III Word list (45) 1 Delayed recall of the Rey visual design learning test (51) 2 Recognition of the Modified Rey visual design learning test (51) 2 Reasoning Similarities of the Wechsler Adult Intelligence Scale IV (WAIS-IV) (52) 1 Block Design of the WAIS-IV (52) 2 Processing speed Stroop Color subtest (time to complete) (46) Number sequencing of the FAT (computerized modification of Trail making A; time to complete) (43,44,50)

Table 3 .
Cognitive performance of patients and controls.

Table 5 .
Neurovascular functioning in patients and controls.Data are presented as median (interquartile range) or N (%).p-values are from Wilcoxon signed-rank tests and McNemar tests in within-group analyses, and Mann-Whitney U tests and Fisher's exact tests in between-group analyses.