Cognitive decline after carotid endarterectomy: systematic review and meta-analysis

BACKGROUND Postoperative cognitive decline (pCD) occurs frequently (6 to 30%) after carotid endarterectomy (CEA), although there are no exact estimates and risk factors are still unclear. OBJECTIVE The objective of this study was to determine pCD incidence and risk factors in CEA patients. DESIGN We performed a systematic review and meta-analysis of both randomised and nonrandomised trials following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. DATA SOURCES We searched Cochrane, PubMed/Medline and Embase databases from the date of database inception to 1 December 2018. ELIGIBILITY CRITERIA We selected longitudinal studies including CEA patients with both pre-operative and postoperative cognitive assessments. Primary outcome was pCD incidence, differentiating delayed neurocognitive recovery (dNCR) and postoperative neurocognitive disorder (pNCD). dNCR and pNCD incidences were expressed as proportions of cases on total CEA sample and pooled as weighted estimates from proportions. Postoperative delirium was excluded from the study design. Secondary outcomes were patient-related (i.e. age, sex, diabetes, hypertension, contralateral stenosis, pre-operative symptoms, dyslipidaemia and statin use) and procedure-related (i.e. hyperperfusion, cross-clamping duration and shunting placement) risk factors for pCD. We estimated odds ratios (ORs) and mean differences through a random effects model by using STATA 13.1 and RevMan 5.3. RESULTS Our search identified 5311 publications and 60 studies met inclusion criteria reporting a total of 4823 CEA patients. dNCR and pNCD incidence were 20.5% [95% confidence interval (CI), 17.1 to 24.0] and 14.1% (95% CI, 9.5 to 18.6), respectively. pCD risk was higher in patients experiencing hyperperfusion during surgery (OR, 35.68; 95% CI, 16.64 to 76.51; P < 0.00001; I2 = 0%), whereas dNCR risk was lower in patients taking statins before surgery (OR, 0.56; 95% CI, 0.41 to 0.77; P = 0.0004; I2 = 19%). Sensitivity analysis revealed that longer cross-clamping duration was a predictor for dNCR (mean difference, 5.25 min; 95% CI, 0.87 to 9.63; P = 0.02; I2 = 49%). CONCLUSION We found high incidences of dNCR (20.5%) and pNCD (14.1%) after CEA. Hyperperfusion seems to be a risk factor for pCD, whereas the use of statins is associated with a lower risk of dNCR. An increased cross-clamping duration could be a risk factor for dNCR