Patients
This retrospective study was conducted as a preliminary study after approved by the Research Ethics Committee of Hiroshima City Asa Citizens Hospital and Hiroshima University (E-466-3, E-1554-2). Written, informed consent was obtained from all participants. In this study, 20 consecutive patients who underwent EC-IC bypass for severe steno-occlusive disease of the intracranial internal carotid artery (ICA) or middle cerebral artery (MCA) at Asa Citizens Hospital from September 2017 to April 2020 were reviewed. The inclusion criteria for this study were the following: 1) patients with symptomatic intracranial steno-occlusive disease who had rCBF of the ipsilateral MCA less than 32 mL/100 g/min (corresponding to 80% of the normal value), and a CVR less than 10% on a quantitative SPECT study with acetazolamide (ACZ) challenge [11]; and 2) patients with available preoperative and postoperative (3–6 months or later post surgery) CVR studies available. Exclusion criteria were as follows: 1) allergy to contrast media; 2) renal dysfunction (estimated glomerular filtration rate < 30 ml/min/1.73 m2); or 3) medical illness, physical disability, or speech impediment precluding the VC task. Twenty patients were selected on the basis of the inclusion criteria (5 women; mean age at the time of bypass 66.9 years). The etiology included atherosclerosis (n = 18) and Moyamoya disease (n = 2) (Supplemental Table 1).
Assessment of cognitive functions
To assess attention deficit, four kinds of VCs (Kana, Triangle, Symbol, Number) included in the Clinical Assessment for Attention (CAT), which is a standardized test for attention deficit, were used as previously described [19]. Participants used a pencil to cross out a target stimulus dispersed within rows of randomly placed interfering stimuli displayed on a sheet. These tasks were scored as speed (completion time) and accuracy. Accuracy was based on the ratio of the number of correct answers to the total number of items (% correct answers) or the number of accurate answers compared to the number of total responses (both correct and incorrect responses) (% accurate answer).
It is known that the scores of the VC depend on age. To correct by age, age-matched values were calculated as (VC score)/(age specific mean VC value); the lower the age matched % correct answer and % accurate answer, the greater the attentional disturbance. Moreover, the higher the age-matched completion time, the greater the attentional disturbance.
The changes in the age-matched time and accuracy scores between before and after EC-IC bypass of each of the 4 VCs were scored as “1” for “improved (time was faster, and correct rate was increased)” and as “0” for “not improved (time was equal or slower, and the correct rate was equal or decreased)”. Then, the speed score was calculated as the sum of the speed values of the four VCs (score range from 0 to 4). The accuracy score was calculated as the sum of the values of the % correct answers and the % accurate answers of the VCs (score range from 0 to 8).
Measurement and analysis of cerebral blood flow
Scans for rCBF were performed just before and 10 minutes after injection of 1.0 g of acetazolamide. Regional cerebrovascular reactivity (rCVR) was calculated as follows: rCVR(%) = [(acetazolamide challenge rCBF–resting rCBF)/resting rCBF] × 100. A three-dimensional stereotactic surface profile program (3D-SSP, Nihon Medi-Physics, Tokyo, Japan) was used to spatially normalize the local distribution. The change of CVR (post-operative CVR – pre-operative CVR) was calculated, and the anatomical classification was evaluated using the stereotactic extraction estimation method: (SEE method LEVEL3) [18].
Operative procedure
Under general anesthesia, with continued antiplatelet medication perioperatively, a skin incision was made just over the superficial temporal artery (STA) frontal branch or parietal branch. Under microscopy, meticulous STA dissection was conducted. Then, the skin incision was extended toward the forehead, and a skin flap was reflected. The frontal branch of the STA was dissected. After craniotomy, an STA-MCA single or double anastomosis was performed between each STA branch and the recipient M4 (cortical MCA branch). Successful bypass was confirmed by microvascular Doppler evaluation.
Statistical analysis
Normality assumptions for each value were analyzed using the Shapiro-Wilk test, and mean (95% confidential interval: CI) was calculated for normal distribution and the median (interquartile range) were calculated for the non-normal distribution. To compare differences between two groups, Fisher’s exact test was used for categorical variables, and the Mann-Whitney U-test was used for quantitative variables. The level of significance was set at p < 0.05.
To test the correlations between the time or the accuracy score and the CVR change in the 31 brain areas on each side (62 in total), bivariate analysis (Spearman’s rank correlation coefficient, ρ) was performed. A cross-correlation matrix was constructed after calculation of Spearman’s rank correlation coefficient (ρ). Then, the hierarchical clustering was built from the bottom-up by joining the closest clusters at each step according to the defined distance and linkage functions [20]. The result was represented by a dendrogram.
Stepwise multiple linear regression analysis based on the Akaike information criterion (AIC) was used to estimate the independent effects of predictor variables on the Speed or Accuracy score (forward–backward selection method). These predictor variables were as follows: age, sex, laterality of the operation, and the CVR changes in the 31 brain areas on each side (62 in total). To assess multicollinearity, the variance inflation factor (VIF) was calculated.
A least absolute shrinkage and selection operator (LASSO) regression model was used to choose the optimized subset of brain regions related to CVR change to predict the Accuracy or the Speed score. First, 10-fold cross-validation was used to obtain the best hyperparameters, while all factors were degenerate. Thus, lambda was set up in ways that 6 values were residual through trial and error.
To compare the performance of the LASSO model, three metrics were calculated between the actual Speed or Accuracy score and the predicted Speed or Accuracy score obtained from all folds: the coefficient of determination (R2), the squared correlation coefficient (r2), and root-mean-square error (RMSE). The values were considered significant at p < 0.05. All data were analyzed using R (version 3.6.2).