Neurocognitive, Quality of Life, and Behavioral Outcomes For Patients With Covert Stroke After Cardiac Surgery. Exploratory Analysis Of Data From A Prospectively Randomized Trial

Abstract

Background:

Asymptomatic brain ischemic injury detected with diffusion weighted magnetic resonance imaging (DWI) is reported in over one-half of patients after cardiac surgery. There are conflicting findings on whether DWI detected covert stroke is associated with neurocognitive dysfunction after surgery and it is unclear whether such ischemic injury affects quality of life or behavioral outcomes. The purpose of this study was to perform exploratory analysis on whether covert stroke after cardiac surgery is associated with delayed neurocognitive recovery 1 month after surgery, impaired quality of life, anxiety, or depression.

Methods:

Analysis of data collected in a prospectively randomized study in patients undergoing cardiac surgery testing whether basing mean arterial pressure (MAP) targets during cardiopulmonary bypass to be above the lower limit of cerebral autoregulation versus usual practices reduces the frequency of adverse neurological outcomes. A neuropsychological testing battery was administered before surgery and then 1 month later. Patients underwent brain MRI between postoperative days 3 to 5. The primary outcomes was DWI detected ischemic lesion; the primary endpoint was change from baseline in domain specific neurocognitive Z-scores 1-month after surgery. Secondary outcomes included a composite indicator of delayed neurocognitive recovery, quality of life measures, state and trait anxiety, and Beck Depression Inventory scores.

Results:

Of the 164 patients with postoperative MRI data, clinical stroke occurred in 10 patients. Of the remaining 154 patients, 85 (55.2%) had a covert stroke. There were no statistically significant differences for patients with or without covert stroke in the change from baseline in Z-scores in any of the cognitive domains tested adjusted for sex, baseline cognitive score, and randomization treatment arm. The frequency of delayed neurocognitive recovery (no covert stroke, 15.1%; covert stroke, 17.6%, p=0.392), self-reported quality of life measurements, anxiety rating, or depression scores were not different between those with or without DWI ischemic injury.

Conclusions:

Over one-half of patients undergoing cardiac surgery demonstrated covert stroke. In this exploratory analysis, covert stroke was not found to be significantly associated with neurocognitive dysfunction 1 month after surgery, evidence of impaired quality of life, anxiety, or depression, albeit a type II error can not be excluded.

Introduction

While clinically overt stroke affects 1% to 5% of patients after cardiac surgery (depending on the patient risk profiles) covert stroke is more prevalent occurring in up to 61% of patients.^1–3 The latter form of brain injury is detected with brain magnetic resonance diffusion weighted imaging (DWI). This imaging modality detects restriction in water diffusion due to ischemia-induced cytotoxic edema resulting in a high-intensity signal within minutes of injury and lasting for several weeks.^4,5 Brain DWI detected ischemic injury can be an incidental finding including during imaging as part of a research protocol where patients have no neurological deficits. The clinical significance of DWI lesions in the absence of clinically correlated neurological abnormalities is unclear.

Covert brain infarction is associated with cognitive impairment and dementia in adults enrolled in longitudinal population-based studies.^6,7 Early and delayed neurocognitive dysfunction after cardiac surgery has been the focus of multiple observational and interventional investigations yet its mechanism remains largely unknown.^1,3 Understanding whether there is a relationship between objective evidence of ischemic brain injury after cardiac surgery and neurocognitive dysfunction could enhance the understanding of the clinical significance of these incidental findings. Several studies have investigated for this relationship but the findings are inconsistent and inconclusive.^2,8 Importantly, prior investigations have not evaluated the impact of covert stroke on quality of life or behavioral end-points.

We recently completed a randomized clinical trial assessing the potential benefits of cerebral autoregulation monitoring for mean arterial pressure (MAP) management in patients undergoing cardiac surgery.^9,10 Using this novel database, we sought to perform exploratory analysis on whether covert stroke after cardiac surgery is associated with delayed neurocognitive recovery 1 month after surgery. We further sought to evaluate whether covert stroke is associated with impaired quality of life, anxiety, or depression 1 month after surgery.

Methods

This was a secondary analysis of data collected from 460 patients enrolled in a randomized study where MAP during cardiopulmonary bypass (CPB) was targeted to be above the lower limit of cerebral autoregulation compared with usual care.^9,10 Analysis of domain-specific cognitive performance and a composite neurocognitive end-point from before surgery baseline to 1 month after surgery and DWI results were a pre-specified end-point of that study. Patients undergoing coronary artery bypass graft (CABG) surgery, with or without cardiac valve surgery, or ascending aortic surgery that required CPB were enrolled in the parent study. Inclusion criteria were age ≥ 55 years with risk factors for neurologic complications including history of stroke, and/or carotid bruit, and/or hypertension, and/or diabetes. Exclusion criteria included: 1) contraindication to MRI; 2) estimated glomerular filtration rate of ≤60 mL/min; 3) emergency surgery; 4) hepatic dysfunction; 5) inability to attend outpatient visits; and 6) visual impairment or inability to speak and read English. The Johns Hopkins Medical Institutes, Baltimore, MD and Northwestern University Feinberg School of Medicine, Chicago, IL Institutional Review Boards approved the protocol and written informed consent was obtained from all patients. Enrollment occurred from October 2, 2009 to May 10, 2016 at The Johns Hopkins Hospital, Baltimore, MD and from October 26, 2016 to January 10, 2019 at Northwestern Memorial Hospital, Chicago, IL.

Perioperative care

Patients received midazolam, propofol, fentanyl, isoflurane, and a skeletal muscle relaxant for anaesthesia induction and maintenance. Heparin was given to achieve an activated clotting time > 450 s. The patients received non-pulsatile CPB with a membrane oxygenator and flow of 2.0 to 2.4 L/min/m² with α-stat pH management. Arterial pH and PaCO₂ were maintained within a normal range by continuous in-line monitoring calibrated with arterial blood gases. The patients received routine institutional postoperative care. Cerebral blood flow autoregulation was monitored during surgery with transcranial Doppler (Compumedics DWL, El Paso, TX) with direct arterial pressure monitoring using ICM+ software (Cambridge Enterprises, University of Cambridge, UK) as we have described.^9–12 Patients were randomized after anesthesia induction using a computer generated list to either have their MAP targets during CPB to be above the lower limit of autoregulation or based on usual institutional practice. Randomization was performed by statisticians at The Johns Hopkins Bloomberg School of Public Health using an internet accessed process. The patient, surgeons, and outcome assessors were blinded to the assignment. Perfusionist, anesthesiologist, and surgeons were not masked since they managed blood pressure during CPB. Blood pressure was raised with phenylephrine (100 mcg) given into the CPB circuit and by reducing isoflurane concentrations to 0.7% after first ensuring target CPB flow. Intraoperative care after CPB and postoperatively in the ICU was not specified in the study protocol.

Neuropsychological Testing

Neuropsychological testing was performed 1 to 2 days before surgery and then 1 month after surgery assessing cognitive domains known to be affected by cardiac surgery.^9,10,13 The battery consisted of the Rey Auditory Verbal Learning Test; Rey Complex Figure Test; Controlled Oral Word Association Test; Symbol Digit Modalities Test; Trail Making A and B; and the Grooved Pegboard Test. The test results were grouped into the following cognitive domains: attention (Rey Auditory Verbal Learning Test I correct); memory (Rey Auditory Verbal Learning Test V correct, Rey Auditory Verbal Learning Test IX correct); visuoconstruction (Rey Complex Figure Test copy trial score); verbal fluency (Controlled Oral Word Association Test letters F, A, S); processing speed (Symbol Digit Modalities Test correct, Trail Making Test A); executive function (Trail Making Test B); and fine motor speed (Grooved Pegboard, dominant and non-dominant hand). A Z-score was calculated from the average of the individual test Z-scores, which were calculated using the mean and standard deviation of baseline tests from all participants with available data.^9,10,14 These were then re-normalized using the mean and standard deviation of the tests at baseline. Timed tests were multiplied by ‘−1’ so that higher scores represented better performance. A global neurocognitive outcome was generated as a sum of the Z-scores on each individual cognitive domain. A difference in score between pre- and post-testing > 0.3 was considered a marker of cognitive decline.^9,10 The patients were assessed at the same testing periods with the SF-36 short form of the RAND Medical Outcomes Study health-related quality of life questionnaire.¹⁵ This questionnaire is a multi-item scale that measures 8 health-related concepts: physical function, social function, physical role, emotional role, mental health, energy, pain, and general health perceptions. The Beck Depression Inventory was administered to assess depressive symptoms at each testing session.^16,17 This 21-item questionnaire assesses self-reported symptoms including such items as self-dislike, suicidal ideation, insomnia, and sadness. The score for each item ranges from 0–3. The 21 items are summarized on a scale that ranges from 0 to 63. Scores of 14–19, 20–28, 29–63 are considered to represent mild, moderate, and severe depression, respectively. A cut-off of >13 was used as a dichotomous definition of depression. Anxiety was assessed using the Beck State-Trait Anxiety Inventory that assesses self-reported ratings of anxiety on a four-point Likert scale.¹⁸ This inventory rates two types of anxiety: state anxiety or anxiety related to an event; and trait anxiety, or anxiety as a personal characteristic.

Neurologic Evaluation

Patients were evaluated at screening and on postoperative day 5 with the National Institutes of Health Stroke Scale.¹⁹ A neurologist was consulted when a neurological abnormality was detected after surgery. Physicians and nurses evaluated the patients daily for any acute mental status changes including abnormalities in arousal, memory, or cognition. A diagnosis of clinical delirium was entered into the medical record using International Classification of Diseases-10^th Revision (Codes F05 or G93.40) criteria.

Magnetic Resonance Imaging

Brain MRI was performed as part of the study protocol between postoperative days 3 and 5 or sooner when neurological deficits were noted by the clinical staff. The imaging sequences included axial DWI from which apparent diffusion coefficient maps were calculated. These data were acquired on a 3T MRI instrument. DWI images were obtained using a multi-slice, isotropic sequence, with bmax = 1000 sec/mm² and TR/TE = 10,000/120 msec.²⁰ The images were analyzed independently by a neurologist and a neuroradiologist masked to patient randomization. Based on the location, size, and pattern of distribution, the acute ischemic infarct were classified into one of four categories: small subcortical infarction (lesion <1 cm in diameter located in the basal ganglia, thalamus, or brainstem); territorial infarction (lesion located in a vascular distribution, usually resulting from large vessel occlusion); watershed infarction (lesion located in the border zone between vascular territories); multiple emboli (multiple lesions in one or several different vascular territories), or mixed watershed and embolic pattern. The number of acute DWI lesions were counted, and the total infarct volume traced and measured volumetrically by a technician using Image J software.²¹

Primary Exposure

The primary exposure of interest was the presence of DWI ischemic lesions on postoperative MRI.

Study Outcomes

Change in Z-scores on neuropsychological tests from baseline to 1 month after surgery; quality of life measurements; anxiety and depression assessments; the composite neurocognitive dysfunction end-point 1 month after surgery; and clinical outcomes of interest.

Statistical Analysis

This analysis used available data from the parent randomized trial. In that study a sample size of 490 patients was calculated to provide 80% power for detecting a 45% reduction in the composite neurological outcome with the intervention. That sample size, however, was based on the assumption that 122 patients in the control group and 122 patients in the autoregulation group would have postoperative MRI data. The study enrolled 232 patients in the control and 228 in the autoregulation group but MRI data was obtained only in 79 and 85 patients in these groups, respectively. The Data Safety and Monitoring Committee for the parent trial reviewed the data on January 4, 2019 and recommended that the trial be stopped based on futility of the intervention on the primary outcome.

Descriptive statistics summarized all baseline demographic, clinical, and treatment-specific variables by the presence or absence of DWI lesions. Patients with clinical evidence of stroke were omitted from this analysis of covert stroke. Primary analyses utilized multivariable linear regression models to assess the association between having a DWI lesion and the change in domain-specific neurocognitive Z-score from baseline to 1 month after surgery, adjusting for age, sex, randomized treatment arm, and baseline neurocognitive Z-score. Similar models also included interaction terms between sex and DWI lesion presence to assess whether the association between DWI lesion and neurocognitive change was different by sex. Sensitivity analyses considered models using multiple imputations for the primary endpoints, cognitive domain z-scores. Specifically, data were imputed using fully conditional specification methods (FCS) with 20 imputation sets incorporating baseline characteristics summarized in Table 1 (e.g., age, sex, race, surgical procedure, and risk factors), all outcomes of interest at baseline and one-month follow-up (individual cognitive tests, SF-36, Beck State Trait Anxiety measures, and Beck Depression Inventory scores) as well as other clinically relevant variables (e.g., treatment arm and Mini-Mental State Examination). Due to small counts, a Fisher’s exact test was used to compare proportions with cognitive decline between groups.

Table 1.

Patient data for the cohort with brain MRI data-stratified by presence or absence of diffusion weighted imaging evidence of acute ischemic brain injury. Ten patients with a clinically overt stroke were omitted from the current analysis. The data represent number of subjects for each category with percentage in parenthesis unless otherwise noted.

	Cohort with MRI Data N=154	No Diffusion Weighted Imaging Lesions N=69	Positive Diffusion Weighted Imaging Lesions N=85
Age (years)*	70 [65, 76]	69 [65, 75]	70 [64, 76]
Female sex	41 (26.8)	20 (29.0%)	21 (25.0%)
Race  White  African American  Hispanic  Asian  Multiple  Other	122 (79.7) 16 (10.5) 2 (1.3) 3 (2.0) 0 9 ( 5.9)	56 (81.2%) 6 (8.7%) 1 (1.4%) 2 (2.8%) 0 4 (5.8%)	66 (78.6%) 10 (11.9%) 1 (1.2%) 2 (2.4%) 0 5 (6.0%)
Hypertension	137 (89.0)	61 (88.4%)	76 (89.4%)
Diabetes	70 (45.5)	32 (46.4%)	38 (44.7%)
Prior MI	48 (31.2)	22 (31.9%)	26 (30.6%)
Atrial fibrillation	30 (19.5)	17 (24.6%)	13 (15.3%)
Prior stroke	14 (9.4)	6 (8.8%)	8 (9.9%)
COPD	21 (14.1)	10 (14.9%)	11 (13.4%)
Current tobacco abuse	19 (12.6)	8 (11.8%)	11 (13.3%)
Highest level of education (years)*	14 [12, 16]	14 [12, 16]	14 [12, 16]
Type of Surgery  CABG  CABG/AVR  CABG/MVR  CABG/AVR/MVR  AVR  MVR  AVR/MVR  Aortic root replacement  CABG/aortic root replacement  TVR	79 (51.3) 19 (12.3) 5 (3.2) 31 (20.1) 13 (8.4) 2 (1.3) 5 (3.2) 0 0 0	39 (56.5%) 5 (7.2%) 2 (2.9%) 11 (15.9%) 7 (10.1%) 2 (2.9%) 3 (4.3%) 0 0 0	40 (47.1%) 14 (16.5%) 3 (3.5%) 20 (23.5%) 6 (7.1%) 0 2 (2.4%) 0 0 0
Duration of CPB (min)*	100 [76, 138]	99.0 [74.0, 132.0]	106.00 [78.00, 140.50]
Duration of aortic cross-clamping (min)*	69.5 [54.0, 94.0]	66.0 [55.0, 90.0]	75.00 [52.00, 96.00]
MAP at the LLA (mean±SD, mmHg)	66.4 (11.4)	65.8 (12.1)	66.9 (10.9)
Product of the magnitude and duration that MAP was < LLA* (mmHgxhr)	4.57 [1.76, 13.55]	4.83 [1.9, 14.5]	4.57 [1.6, 13.1]
Average MAP during CPB (mean±SD, mmHg)	74.1 (7.9)	73.6 (7.8)	74.5 (8.1)

MI=myocardial infarction; COPD=chronic obstructive pulmonary disease; LLA=mean arterial pressure (MAP) at the lower limit of cerebral blood flow autoregulation.

^*median[interquartile range]

Similar methods were employed for additional endpoints including the SF-36 self-reported quality of life measures, Beck State Trait Anxiety measures, and Beck Depression Inventory scores. Of note, Role Limiting Physical Health and Role Limiting Emotional Health scales were dichotomized to represent “No Problems” vs “Any Problems” due to violation of distributional assumptions. Multivariable logistic regression models, with similar adjustment, were estimated for these binary endpoints. A similar model was fit for the binary indicator of Beck Depression Inventory Score > 13.

Generalized linear regression, using Poisson distribution and log link with robust standard errors, assessed associations between sex and presence of postoperative DWI lesion, adjusting for randomized treatment arm. Within the subgroup of patients having a DWI lesion, we compared lesion-specific outcomes between men and women, adjusting for randomized treatment arm using similar models. Continuous outcomes (e.g., lesion volume) were log transformed and considered in multivariable linear regression models with similar adjustment. Risk ratios were reported for binary outcomes and linear regression coefficients were reported for continuous outcomes, each with corresponding 95% confidence intervals. Additional analyses explored associations between postoperative DWI lesion with clinical end-points of interest (e.g., clinical delirium) using generalized linear models for binary endpoints and linear regression models for continuous endpoints, as described previously.

Results

A patient flow diagram is shown in Figure 1. This diagram depicts patients in the parent study where blood pressure targets during CPB were either managed based on autoregulation metrics versus usual care. In the present secondary analysis of those data, postoperative MRIs were available from 164 patients. Of these patients, 10 patients had a clinical stroke and were excluded from analysis. Of the remaining 154 patients with MRI data, 85 (55.2%) patients had a covert stroke. Patient demographic and medical information for the entire cohort and based on presence or absence of covert stroke are listed in Table 1. Similar data for patients with and without postoperative MRI data are listed in Supplemental Table 1. Patients with postoperative MRI data were similar to patients without MRIs with the exception of a higher frequency of prior myocardial infarction. The median [interquartile range] number of DWI lesions was 3.0 [1.0, 6.0] and volume 0.99 [0.44, 4.00]. The majority of the acute ischemic lesions (67.9%) had embolic characteristics, 23.5% demonstrated a hypoperfusion “watershed” pattern, and 16.0% a mixed embolic-watershed pattern.

Figure 1.

Patient flow diagram

Both baseline and one-month neuropsychological testing results for at least one domain were available from 127 (83%) patients with covert stroke as shown in the patient flow diagram (Figure 1). Thirty-three of these patients were missing data from one or more cognitive domains. Box and whisker plots of the neuropsychological test results for patients with complete data at baseline and one month after surgery for each domain based on the presence or absence of covert stroke is shown in Figure 2. The data represent a change from preoperative baseline with a negative number indicating decline in Z-score. The multivariable linear regression results evaluating the relationship between covert stroke and domain-specific neurocognitive results are listed in Table 2. These data are presented as a complete data analysis and multiple imputations for missing data. There were no statistically significant differences in the change from baseline in the domain specific neurocognitive results 1 month after surgery between patients with and without covert stroke, adjusting for age, sex, baseline cognitive domain Z-score, and randomized treatment group. These results remained consistent in sensitivity analysis using multiple imputations. When considered as a composite outcome, there was no statistically significant difference in the frequency of delayed neurocognitive recovery 1 month after surgery for patients with and without covert stroke (no covert stroke, 15.1%; covert stroke, 17.6%, Fisher’s exact p=0.392)

Figure 2.

Cognitive outcomes at 1 month

Table 2.

Multivariable linear regression model for association between covert stroke and change from baseline for each neurocognitive end-point 1 month after surgery. The data are presented based on complete case analysis and with multiple imputations for missing data. Models are adjusted for age, sex, treatment arm, and baseline cognitive test results.

Domain	Number with complete data	Complete Case Analysis Estimate (Standard Error)	P-Value	Multiple Imputations Estimate (Standard Error)	P-Value
Attention	122	−0.003 (0.170)	0.985	−0.142 (0.216)	0.511
Memory	122	0.042 (0.115)	0.719	−0.020 (0.132)	0.878
Visuoconstruction	119	−0.117 (0.135)	0.387	−0.004 (0.148)	0.980
Verbal Fluency	118	−0.124 (0.112)	0.273	−0.139 (0.120)	0.250
Processing Speed	120	0.004 (0.108)	0.972	0.054 (0.107)	0.618
Executive Function	107	−0.098 (0.128)	0.444	−0.010 (0.121)	0.937
Fine Motor Speed	108	−0.053 (0.093)	0.572	−0.030 (0.109)	0.787

Quality of life assessment results from 1 month after surgery adjusted for randomized treatment group of the parent trial and baseline scores are listed in Table 3. There was no statistically significant relationship between covert stroke and any quality of life measure. Patients with covert stroke had lower scores on state anxiety scores (transient reaction to adverse situation at a specific moment) than those without brain injury. There were no statistical differences between these groups in trait anxiety scores (an individual’s tendency to express anxiety). While there were no differences in the average Beck Depression Inventory scores (higher score represent more depressive symptoms) for patients with or without covert stroke, more patients in the latter group were classified as depressed than those with covert stroke.

Table 3.

Rand Corporation SF-36 self-reported quality of life measures, Beck State Trait Anxiety measures, and Beck Depression Inventory scores 1 month after surgery for patients with and without a brain MRI diffusion weighted imaging evidence of new ischemic injury without a clinical neuropsychological deficit (covert stroke). Beck Depression Inventory scores are dichotomized to those with a score > 13 indicating depression. The data are listed as mean (±standard deviation)^* or as the number and percentage of patients with a score on a specific domain.

Domain	No DWI Lesion N=69	DWI Lesion Present N=85	Estimate from Linear or Logistic* Regression Model (Standard Error)	P-Value1.
Physical Function	42.4±27.0	46.6±26.5	2.115 (4.913)	0.668
Role Limit Physical Health No Problems Any Problem	1 (1.9%) 51 (98.1%)	1 (1.5%) 66 (98.5%)	1.473* (4.562)	0.798
Role Limit Emotional Health No Problem Any Problem	29 (55.8%) 23 (44.2%)	36 (53.7) 31 (46.3%)	0.788* (1.572)	0.599
Energy Fatigue	44.9±23.0	47.3±21.7	2.279 (3.927)	0.563
Emotional Well-Being	80.5±18.2	81.8±15.3	0.597 (2.666)	0.823
Social Function	58.6±23.8	58.2±29.6	−1.403 (5.243)	0.790
Pain	55.8±28.0	59.4±23.5	−0.249 (4.961)	0.960
General Health	65.3±20.9	67.2±18.8	2.037 (3.106)	0.514
State Trait Anxiety (State)	32.1±10.79	30.5±10.2	−3.521 (1.648)	0.035
State Trait Anxiety T (Trait)	31.1±10.2	30.6±11.0	−1.777 (1.629)	0.278
Beck Depression Inventory Score	6.7±5.7	6.4±5.7	−0.906 (0.723)	0.213
Patients with Beck Depression Inventory Score > 13	10 (18.5%)	7 (10.0%)	0.206* (2.185)	0.043

¹P-value from multivariable linear or logistic regression model adjusting for randomized treatment arm in the parent study and baseline QOL measure

^*Estimate from logistic regression model is on the log odds scale

There was no difference in the frequency of clinically detected delirium (n=9 [6.0%] vs n=7 [10.1%], p=0.925) or the frequency of postoperative atrial fibrillation (n=31 [36.5%] vs n=27 [29.1%], p=0.738) between patients with or without covert stroke, respectively. The duration of hospitalization in the ICU (median [interquartile range] hours, 44.0 [26.0, 66.0] vs 40.0 [25.0, 52.5], p=0.613), or on the postoperative ward (median [interquartile range], days, 7.0 [6.0, 9.2] vs 7.0 [6.0, 9.0] p=0.744) was not different between patients with and without covert stroke, respectively. There were no deaths within 30 days of surgery for patients with or without covert stroke.

Discussion

In study we found evidence of covert stroke after cardiac surgery in 55.2% of 154 patients. The majority of the ischemic lesions had embolic characteristics with a quarter demonstrating a hypoperfusion pattern. In exploratory analysis there was no significant relationship between brain covert stroke and neurocognitive end-points 1 month after surgery when considered on a complete data basis or after multiple imputations. The presence of covert stroke was not associated with alterations in quality of life, severity of anxiety, or depression.

There have been several reports describing DWI ischemic lesions in patients after cardiac surgery. A systematic review of 13 studies involving 446 patients reported DWI lesions in 29% of 127 patients (range 16% to 61% of patients).⁸ The size of these studies ranged from 13 to 86 patients with only two studies reporting on 50 or more patients. The DWI lesions were 1 to 10 mm in size, most often multiple, located in all vascular territories, and typically not associated with clinical or overt stroke. Neuropsychological end-points were reported in 12 of the studies with inconsistent findings. There was much heterogeneity in the extent of the neuropsychological testing batteries, and timing of testing with the majority reporting cognitive outcomes two to seven days after surgery. More recently, secondary analysis of MRI data from a randomized study of empiric MAP management strategies during CPB has been reported.² In that study, DWI lesions were found in 52.8% and 55.7% of the 169 patients in the two different MAP management groups. These investigators found a relationship between delayed neurocognitive recovery at hospital discharge and DWI ischemic lesions. Taken together existing investigations have reported conflicting findings on whether covert stroke after cardiac surgery is associated with delayed neurocognitive recovery. It remains possible, though, that in our study the ischemic lesions occurred in brain areas not assessed with the employed neuropsychological battery. Perhaps, a more extensive neuropsychological battery would identify more specific neurocognitive deficits associated with new DWI ischemic lesions. Our study extends prior studies showing a lack of an association between covert stroke and self-reported physical health, emotional health, energy/fatigue, social function, pain, or perceptions of general health. Symptoms of depression in patients undergoing cardiac surgery are associated with risk for postoperative complications, reduced functional recovery, impaired quality of life, and mortality.^22–24 In our study, the presence of covert stroke did not impact Beck Depression Inventory scores or increase the frequency of depression or anxiety.

The presence of neurocognitive complications of cardiac surgery have been appreciated for nearly seven decades.²⁵ It is widely believed that there are shared mechanisms between stroke and neurocognitive dysfunction; the varied clinical manifestations suggested to be dependent on the extent and location of injury.^1,3 For multiple reasons including that it is a more frequent complication than stroke, neurocognitive end-points have been the primary outcome for multiple neuroprotective clinical trials.^1,3 Our data and that of other studies demonstrate that objective evidence of brain injury can occur in the absence of clinically apparent neurocognitive dysfunction and vice versa. These findings suggest that clinical trials of neuroprotective strategies should evaluate multiple neurological end-points in addition to neurocognitive function including, perhaps, neuroimaging and/or brain specific biomarkers.^26–28

The patients in this study were enrolled in a clinical trial where patients were randomized to have MAP targets during CPB set to be above the lower limit of cerebral autoregulation versus usual care. In our analysis we adjusted for randomized treatment group. As a secondary analysis of the primary data, though, our results should be viewed with caution as we may be underpowered for these analyses. A larger sample size possibly would have identified a link between DWI ischemic lesions and neurocognitive, quality of life, or behavioral deficits. Limitations to our study include missing data. Reasons for missing MRI data for patients whom otherwise met study enrollment criteria include patient refusal after surgery, unavailability of a scanner before patient discharge, retention of epicardial pacemaker leads, unexpected retained facial metal detected with x-ray before the MRI, and patient intolerance of the MRI. The number of patients with MRI data, though, is similar to that in other trials of MAP management during CPB.² Patients with missing MRI data had similar characteristics as those with this data (Supplemental Table 1). We can not exclude a selection bias though in obtaining postoperative MRI data. Missing postoperative cognitive data was mostly due to patients refusing further testing. Cognitive testing is challenging to patients, particularly those with cognitive impairment. Indeed, most patients were able to complete at least some of the tests but refused others. Our complete-case analyses may introduce bias, due the fact that missing data may not be completely at random. We additionally performed a sensitivity analysis using multiple imputations to allow for inclusion of participants with incomplete data under a missing at random assumption. Enrollment in the study over a 9-year period could have influenced our results that failed to account for improvement in perioperative care over longer period of time.

In conclusion, we found that over 50% of patients undergoing cardiac surgery demonstrated postoperative DWI ischemic injury. In this exploratory analysis, covert stroke was not found to be significantly associated with neurocognitive dysfunction 1 month after surgery, evidence of impaired quality of life, anxiety, or depression, albeit a type II error can not be excluded.

Key Points.

Question:

Is there a relationship between covert stroke detected with sensitive brain MRI modalities after cardiac surgery and neurocognitive dysfunction, altered quality of life, and adverse behavioral outcomes?

Findings:

Covert stroke was detected in over one-half of patients after cardiac surgery but it was not associated with worse neurocognitive outcomes, altered quality of life, or evidence of higher rates of anxiety or depression than those without covert stroke.

Meaning:

The presence of acute MRI evidence of ischemic brain injury can occur without significantly impacting neurocognitive, quality of life, or behavioral outcomes 1 month after surgery.

Glossary of Terms

CABG

Coronary artery bypass graft

CPB

Cardiopulmonary bypass

DWI

Diffusion weighted imaging

MAP

Mean arterial pressure

MRI

Magnetic resonance imaging