J
THORAC CARDIOVASC SURG
84:585-594, 1982
Neuropsychological dysfunction following elective cardiac operation I. Early assessment TII'o neuropsychological tests were administered to 227 men and women, ages 25 to 69 years, before and after coronary bypass and cardiac valve operations to provide current information regarding the incidence of postoperative decrements in neuropsychological dysfunction and the factors associated with them. Biographical, psychological, and medical-surgical data were studied together with changes in scores on the Trail Making Test and the Visual Reproduction (VR) Test of the Wechsler Memory Scale (WMS). Postoperative decrements greater than one standard deviation were observed in each of the four scores derived from these testings for II % to 17% of the patients. Yet 70% of all patients remained within one standard deviation of original performance on all four scores. Among the preoperative correlates of significantly reduced test performance were age greater than 60 years, end-diastolic pressure greater than 30 mm Hg, moderate to severely enlarged heart size on preoperative x-ray film, and use of propranolol or chlordiazepoxide hydrochloride. Significant perioperative correlates included measure of duration of operation (such as total time of operation greater than 7 hours, time on the pump greater than 2 hours, and aortic cross-clamp time greater than 2 hours), total estimate of blood loss greater than 2,000 ml, hypotension, difficult intubation, and insertion of an intra-aortic balloon. Postoperative factors significantly associated with declines in test scores included electrolyte (Na", K+, 0-) abnormalities, longer stay in the intensive care unit, bizarre behavior or disorientation, and depression score. These findings suggest that those patients with more precarious heart function, a more protracted operation, and/or increased metabolic disturbances are especially prone to neuropsychological dysfunction following cardiac operations.
Judith A. Savageau, M.P.H.,* Babette-Ann Stanton, Ph.D.,* C. David Jenkins, Ph.D.,* and Michael D. Klein, M.D.,** Boston, Mass.
T
he surgical treatment of acquired and congenital cardiac lesions has developed vastly over the past two decades. Numerous studies have appeared in the literature reporting observations of postoperative behavioral disturbance and neurologic dysfunction among patients undergoing cardiac operations.v " Kimball" has noted that frequency of these postoperative complications ap-
From Boston University School of Medicine, Boston, Mass. Supported by Research Grant HL20637 from the National Heart, Lung and Blood Institute. Received for publication Aug. 24, 1981. Accepted for publication Jan. 13, 1982. Address for reprints: Judith A. Savageau, M.P.H., Boston University Medical Center, School of Medicine, 720 Harrison Ave., Suite 800, Boston, Mass. 02118. • Department of Behavioral Epidemiology, Division of Psychiatry. •• Department of Cardiology. 0022-5223/82/100585+ 10$01.00/0
© 1982 The C. V. Mosby Co.
pears to be greater than is observed following other types of operations. Despite improvements which have occurred over time in both the nature and technique of the operation and the method of perfusion, neuropsychological dysfunction continues to be observed in some patients in the postoperative period. The reported incidence of behavioral disturbance and neurologic dysfunction following cardiac operations with extracorporeal circulation varies from 20% to 80%, depending on the definition of endpoints, type of circulatory support, patient selection criteria, and years in which the operation was performed.t"" Many attempts have been made to correlate the incidence of neurologic dysfunction with various preoperative, perioperative, and postoperative factors. Possible etiologic factors include advanced age" I L 12 and preoperative presence of neurologic or cerebrovascular disease. I, 8, 12 Profound hypothermia" and biochemi585
The Journal of
586 Savageau et al.
cal changes during the operation':' may also contribute in some cases. Microemboli entering cerebral capillaries during the operation, including emboli from both the heart and the perfusion apparatus, have been hypothesized to be a major cause of neurologic dysfunction.?: 15-17 Extended duration of extracorporeal circulation or hypotension leading to inadequate cerebral perfusion has also been suggested as a possible contributor to cerebral hypoxia.P- 8. 10. 12. 18-20 Postoperative psychological abnormalities either may be related to neurologic damage" or may represent a response to the stress of facing a life-threatening operation and exposure to the abnormal environment of a surgical intensive care unit. 22 Neuropsychological testing has been utilized to study short-term and long-term effects of cardiac procedures on neurologic functioning.'?' 16 Such tests seek to distinguish dysfunctions of emotional origin (functional) from structural or physiological changes in the central nervous system (organicj.P- 24 Some of the factors thought to be related to neurologic dysfunction could possibly be eliminated or their effects minimized, but this requires a knowledge of their relative importance. The nature of the behavioral and neurologic changes, and the causes thereof, have not been clearly defined. In an attempt to understand better the extent of central nervous system dysfunction following cardiac operations and the factors associated with it, we examined patient characteristics and medical-surgical circumstances before, during, and after the operation.
Procedure Data were obtained from 227 patients from four hospital centers: University Hospital and Beth Israel Hospital, Boston, Massachusetts; University of Massachusetts Medical Center, Worcester, Massachusetts; and The Hospital of the Albert Einstein College of Medicine, New York, New York. All patients were participants in a prospective study of the recovery process following elective cardiac operations requiring extracorporeal circulation (1979 to 1980). Criteria for selection into the larger prospective study included the following: (1) age-from 25 to 69 years; (2) fluency in English with the ability to give informed consent; (3) absence of any major noncardiac impairment that might interfere with the postoperative recovery process, including diagnoses of psychiatric or neurologic disorders; (4) availability for a preoperative interview to be conducted in the hospital 1 to 3 days prior to the surgical procedure; (5) stable symptomatic heart disease not necessitating intensive care unit surveillance immedi-
Thoracic and Cardiovascular Surgery
ately prior to the operation; and (6) present admission to the hospital being the first and only cardiac operation for the patient. Of all patients eligible to participate, based on those criteria, 84% accepted the invitation to enter our study. The 227 subjects studied for the purposes of this paper consisted of 184 (81%) men and 43 (19%) women. Ages of the participants ranged from 25 to 69 years, with a mean age of 54.5. Seventy-six percent (172) of the sample underwent coronary artery bypass grafting, 13% (29) underwent valve replacement (either aortic or mitral), and the remaining 11% (26) had multiple procedures (either both valves or some combination of valve replacement and bypass grafting). Information on these participants was gathered by means of a standardized interview and self-administered questionnaire 1 to 3 days prior to operation, plus an interview conducted approximately 9 days postoperatively. The interview and self-report questionnaire covered medical history, smoking and dietary habits, sleep problems, life change stress, physical functioning, work and exercise habits, a variety of items on family and social life (attitudes and satisfactions), and expectations regarding postoperative activities. All these data were combined with information from the patient's hospital records, including items of medical history, results of preoperative cardiac catheterization, listings of types and dosages of prescribed medications, and data on the operation itself and any perioperative or postoperative complications. All patients were alert, oriented, and medically stable at the time of preoperative examination, and none of the patients had documented neurologic or psychiatric impairment preoperatively. All tests were conducted by examiners trained in the administration of these tests, and the preoperative and postoperative interviews were almost always conducted by the same examiner. The following tests were administered: (1) Trail Making Test, Parts A and B, from the Halstead-Reitan Battery25 and (2) Visual Reproduction (VR) from the Wechsler Memory Scale (WMS), Form I (postoperative) and Form II (preoperativej.t" 27 The Trail Making Test Part A (TM-A) is composed of 25 circles containing the numbers 1 through 25 on a white sheet of paper. The examinee is instructed to connect the circles in numerical order as quickly as possible. Part B (TM-B) is composed of 25 circles containing the numbers 1 through 13 and the letters A through L. Here, the examinee is instructed to connect the circles in ascending order, as quickly as possible, while alternating between numbers and letters (e.g.,
Volume 84
Neuropsychological dysfunction, I
Number 4 October, 1982
IA, 2B, etc.). Errors are brought to the attention of the examinee, who is instructed to correct them. The score standardly used is the elasped time in seconds. In addition, we recorded the total errors for each part of the test. In the VR task from the WMS, the examinee is shown each of three cards for a period of 10 seconds. After each card is taken away, the examinee is required to reproduce the geometric design from memory. The drawings are scored according to the directions in the WMS Manual. The VR test was given preoperatively but not postoperatively for the first few months of the study. Therefore, the sample size of the preoperative versus the postoperative differences will be smaller for this one test. We opted to use the two forms of the WMS, Form II preoperatively and Form I postoperatively, on the presumption that they were equivalent and in order to avoid confounding the test-retest difference with a practice effect. However, several months into the study it became clinically apparent that Form I of the WMS VR test was more difficult than Form II of the same test. To test this empirically we gave both Form I and Form II of the VR test consecutively at the 9 day postoperative examination to all patients in the most active of our four hospitals for several months (N = 78). The mean score for Form II was higher by 3.2 points. The median difference was 3 points for the 78 pairs of scores. Although a study planned in advance to test and adjust for differences in these two forms may have had a more sophisticated design and may have worked out a more complex algorithm for adjustment, given our circumstances we decided that, when making comparisons between the forms, the most equitable adjustment would be to add 3 points to the Form I score (postoperative) to compensate for its greater difficulty than Form II. These will be referred to as adjusted postoperative scores and adjusted differences. The difference between the preoperative and postoperative scores were the variables used in the statistical analysis of the data. It has been stated that intraindividual changes in scores from preoperative test to postoperative test seem to be the best measure of postoperative mental function.i" Patients were later classified as showing deterioration in function on a particular test if the postoperative test scores were more than one standard deviation lower than their preoperative score." All factors thought to be related to neurologic dysfunction were initially analyzed separately through correlation coefficients and one-way analyses of variance. However, in view of the probable interrelationship
5 87
of many of these factors, the data were reconsidered through multiple regression analyses. Results
Neuropsychological functioning was significantly reduced postoperatively compared to preoperatively for three of the four test scores studied. * Table I displays univariate statistics for the outcome scores plus onesample t statistics to test if a significant difference existed between the preoperative and postoperative test scores. Preoperative and postoperative test scores were significantly different in all tests except for YR. Only for the TM-B times did the average decline seem to be of clinical importance. The variability of performance across individuals was considerable, however, and the correlates of those differences will be examined in detail later. The results of a two-way analysis of variance testing the effects of sex and surgical procedure on the test score differences indicated that neither sex nor surgical procedure was significantly related to the outcomes, and no interaction effect existed between the two variables (i.e., there was no effect present due to the joint influence of the two factors, above and beyond the effects of each variable separately). Thus all subsequent analyses studied the subject population as a whole. The finding of no significant differences in neuropsychological changes by category of heart operation is an important finding which permits the data herein reported to be considered representative of both valve and bypass patients in clinical populations comparable to ours. The numerous factors associated with neurologic dysfunction in our sample can be divided into three categories: preoperative, perioperative (occurring in the operating and recovery rooms), and postoperative (from intensive care unit stay to time of discharge). Tables II to IV present the correlation and one-way analysis of variance results relating these various factors to the outcome variables. In instances where data were collected as continuous variables (e.g., duration of anesthesia), the text will present the results of these variables as they correlate with the outcomes, as well as one-way analysis of variance results after the variables have been categorized (e.g., I to 3 hours, 3 to 5 hours, 5 to 7 hours, 7 to 9 hours, etc.). Of the preoperative factors investigated, the follow*InitialIy, we also tabulated errors for TM-A and means for this are shown in Table I, but the participants in this study made so few errors on this subtest (only 3% made more than one error) that further analysis of these error scores was not warranted.
The Journal of
5 8 8 Savageau et al.
Thoracic and Cardiovascular Surgery
Table I. Average scores on neuropsychological tests for patients before and after cardiac operation
Visual Reproductiont (No. of correct parts) Range Mean
SO Trail Making Test, Part A-times (sec) Range Mean
SO Trail Making Test, Part A - errors Range Mean
SO Trail Making Test, Part B -times (sec) Range Mean
SO
Trail Making Test, Part B - errors Range Mean
SO
Preoperative test score
Postoperative test score
Test score differences *
One-sample t test
0-14 11.218 2.583
4-17 11.165 3.376
-6 - +8 -0.053 3.250
-0.187 (p = 0.437)
15-180 42.659 22.812
17-177 45.739 24.800
-130 - +71 -3.080 22.318
-2.07 (p = 0.039)
0-5 0.173 0.534
0-4 0.235 0.536
-3 - +4 -0.062 0.709
-1.31 (p=0.190)
32-360 99.860 41.673
38-480 119.680 75.273
-420 - +90 -19.820 64.256
-4.60 (p < 0.001)
0-8 0.905 1.353
0-16 1.405 2.253
-II - +7 -0.500 2.203
-3.38 (p = 0.001)
'For all test scales, negative difference scores are indicative of decreased test performance in the postoperative period.
t Adjusted for test difficulty differences between preoperative and postoperative scores.
ing were significantly related to at least one of the test score differences: age, admitting systolic blood pressure, preadmission use of propranolol (lnderal), preoperative use of chlordiazepoxide hydrochloride (Librium) (prescribed after admission to hospital), heart size (based on the radiologists' interpretations of the transverse cardiac diameter on the preoperative x-ray film and recorded in the hospital chart as "normal," "slight, " "moderate," or "severe" enlargement), and end-diastolic pressure (measured during cardiac catheterization) . Postoperative test score decrements were noted with increasing age, especially for those patients over 60 years. This association was significant for all tests except VR, though a trend toward significance was present for this test. Elevated systolic blood pressure on admission was also related to poorer scores postoperatively on TM-B time and errors and VR; for all test scores, poorest performance was observed among those patients with systolic blood pressure greater than 140 mm Hg. Higher end-diastolic pressure was significantly correlated only with poorer adjusted VR test performance, although for all four tests, poorer postoperative performance occurred in those patients with an enddiastolic pressure greater than 30 mm Hg. Those patients judged from the x-ray film to have
any degree of heart enlargement performed less well postoperatively on TM-B time scores, and similar trends (p > 0.05) were observed for TM-B error and TM-A time scores. Patients taking daily doses of propranolol greater than 160 mg preoperatively also performed better postoperatively on the adjusted VR and TM-B error scores, with a similar trend evident on TM-B time scores. In comparison, those patients who were prescribed chlordiazepoxide hydrochloride preoperatively slowed down much more postoperatively on the TM-B time scores, even though there were no differences by drug dosage for preoperative test scores. Of the various perioperative factors examined, those that were significantly associated with the outcomes included the following: insertion of an intra-aortic balloon, difficult intubation, hypotension, administration of propranolol during the operation, total estimated blood loss, duration of anesthesia, aortic cross-clamp time, extracorporeal circulation time (i.e., time on the heart-lung pump), and total duration of the operation. Those 12 patients who required an intra-aortic balloon to be inserted during the operation performed less well postoperatively on all three of the Trail Making test variables in the analysis, with a similar trend present for the VR test. Difficult intubation (in only seven pa-
Volume 84 Number 4 October, 1982
Neuropsychological dysfunction, I
5 89
Table II. Preoperative correlates of decline in postoperative test performance Test score differences Predictors
No.*
Trail Making, Part A -times
Correlation coefficients t Age Admitting systolic blood pressure End-diastolic blood pressure Mean changes for subgroupss Age
30-39 40-49 50-59 60-69 Heart size on x-ray film Normal Slight enlargement Moderate/ severe enlargement Preadmission use of propranolol No use 1-160 mg 161-320 mg >320 mg Preoperative use of chlordiazepoxide hydrochloride No use Use
-0.1026 -0.0719 -0.0875
I
Trail Making, Part B -times
I
Trail Making, Part B -s-errors
I
Visual Reproduction
-0.1739§ -0.2181§ -0.0212
-0. I826§ -0.1622§ -0.0397
-0.1568§ -0.2309§ -0. I474§
-0.40 0.97 -0.08 -0.88
10 98 93 66
-11.90 1.91 -1.41 -8.5711
-5.50 -11.02 -11.11 -42.7~1
0.00 -0.09 -0.24 -1.3311
156 37 27
-1.11 -4.54 -9.15
-9.78 -35.73 -38.2311
-0.33 -0.35 -1.39
0.14 -1.10 0.06
64 90 52 21
-1.52 -1.88 -2.79 -13.67
-30.95 -22.48 -7.98 -3.95
-1.11 -0.55 0.10
o. J(~I
-0.84 -0.21 0.03 I. 9411
162 65
-3.55 -1.91
-12.95 -37.16/1
-0.55 -0.38
0.16 -0.66
'Sample sizes will not always total 227 because of occasionally missing data. tNegative coefficients signify that higher values of the predictor are associated with poorer postoperative performance. +Negative differences signify decrement in postoperative performance. §Product-moment correlation significant at p < 0.05. [One-way analysis of variance significant at p < 0.05.
tients) and administration of propranolol during the operation (in 68 patients) were significantly related to poorer postoperative performance on the VR test, whereas intraoperative hypotension was associated with decreased performance on TM-B time scores (the same trend for the latter variable was also observed with TM-B error scores.) In addition, a total estimated blood loss of more than 2,000 ml was associated with decreased test performance on TM-B time and error scores as well as adjusted VR scores. These same test scores were significantly related to aortic cross-clamp time in that times greater than 120 minutes were associated with poorer postoperative test scores (a similar trend also being present for TM-A). Longer time on the heartlung pump was significantly correlated with TM-B error scores and the VR; however, on all four tests, decrements in test performance postoperatively were evident in persons on the pump for more than 2 hours. Longer durations of anesthesia were significantly associated with decrements in both TM-A and TM-B time scores and VR (with a similar trend for TM-B error
scores) postoperatively; decreased test scores were especially apparent in those having anesthesia for more than 7 hours. This same negative correlation (Le., increased time with decreased test performance) was observed between duration of operation and TM-B time scores. When this variable was analyzed categorically, a trend toward significance was observed on all four tests, with poorer performance observed for those with more than 7 hours of operation. Postoperative factors observed to be associated with neuropsychological dysfunction (at the 9 day postoperative test) included a hematocrit value below 30% during the first 12 hours after the operation, an elevated creatinine level after the first 24 hours, * electrolyte abnormalities after the first 12 hours, * bizarre behavior and disorientation after the first 5 days (as noted in the hospital chart), depression (scored from the 9 day follow-up interview using the depression scale from the *Laboratory values outside of the normal prevailing ranges at each of the four hospital laboratories were used as indicators of "abnormal values. "
The Journal of Thoracic and Cardiovascular
590 Savageau et al.
Surgery
Table Ill. Perioperative correlates of decline in postoperative test performance Test score differences Predictors Correlation coefficientst Total duration of operation Aortic cross-clamp time Duration of extracorporeal circulation Duration of anesthesia Total estimated blood loss Mean changes for subgroups's Insertion of intra-aortic balloon: No Yes Difficult intubation: No Yes Hypotension: No Yes Administration of propranolol: No Yes
No.*
Trail Making, Part A -times -0.1046 -0.0930 -0.1244 -0. 1968§ -0.0538
I
Trail Making, Part B -times -0.1615§ -0. 1426§ -0.0995 -0. 1947§ -0.1267§
I
Trail Making, Part B -errors
I
Visual Reproduction
-0.1205 -0.2147§ -0.1374§ -0.1011 -0.1594§
-0.0467 -0. 1503§ -0.1541§ -0.234O§ -0.1304§
211 12
-2.14 -21.2711
-17.70
-0.43
-70·~1
-2.3~1
0.02 -1.88
216 7
-2.89 -4.17
-20.12 -2.14
-0.50 -0.71
0.06 -4.6711
187 35
-3.13 -2.54
-16.09 -40.2911
-0.40 -1.09
-0.05 -0.10
146 68
-2.52 -4.88
-16.73 -25.28
-0.44 -0.61
0.37 -1.2211
*Sample sizes will not always total 227 because of occasional missing data. tNegative coefficients signify that higher values of the predictor are associated with poorer postoperative performance. :j:Negative differences signify decrement in postoperative performance. §Product-moment correlation significant at p < 0.05.: [One-way analysis of variance significant at p < 0.05.
Profile of Mood States);" and length of stay in the intensive care unit. Patients who spent 4 or more postoperative days in the intensive care unit showed greater slowing on TM-B. Those who felt more depressed at the time of the postoperative neuropsychological testing were not slowed in Trail Making performance, but they did decrease in the memory and spatial skills required for the VR test. Elevated creatinine levels after the first 24 hours, suggestive of renal difficulties, were related to poorer performance on the TM-A time, TM-B time, and TM-B error scores. Serum electrolyte abnormalities after the first 12 hours postoperatively were associated with significantly slower motor responses to TM-A and marginally decreased VR (p < 0.10). Patients with hematocrit levels less than 30% during the first 12 hours most frequently performed less well on the VR test, though this association was of borderline statistical significance. Bizzarre behavior during the first 5 days (abstracted from medical and nursing entries in the hospital chart) was significantly related to poorer TM-B error scores, and medical or nursing notation of "disorientation" was associated with strikingly poorer values on all three Trail Making variables. (Only two of
the 21 persons noted to have either disorientation or bizarre behavior displayed both aberrations.) When comparing preoperative and postoperative test scores in terms of our ad hoc criterion for significant decrement in test performance, we observed that 12% of this patient group showed significant decrement on the TM-A time; i.e., the postoperative test scores were more than one standard deviation (22 seconds) lower than the preoperative test scores. A total of 12% of our sample displayed significant decrement postoperatively on the TM-B time test; i.e., postoperative test scores were more than 64 seconds longer than the preoperative scores. For the error scores of this same test (TM-B), 11% of the patients performed significantly less well postoperatively (as indicated by having two more errors on the 9 day test). Seventeen percent of those patients with preoperative and postoperative VR scores had significantly decreased (more than 3 points lower) postoperative test scores after adjustment for the difference in difficulty of the two test forms. Of the 227 patients studied, 160 (70%) did not show significant postoperative deterioration on any of the four test scores. Forty-three patients (19%) performed significantly less well postoperatively on only one test, whereas an additional 15 patients (7%) had sig-
Volume 84
Neuropsychological dysfunction, I
Number 4
59 1
October, 1982
Table IV. Postoperative correlates of decline in postoperative test performance Test score differences Predictors
No. *
Trail Making, Part A -limes
I
Trail Making, Part B -limes
Correlation coefficients t
Length of stay in intensive care unit Postoperative depression Mean changes for subgroupss: Hematocrit 30% (during first 12 hours) No Yes Elevated creatinine level (during first 24 hours) No Yes Electrolyte (Na,+ K,+ abnormalities (after first 12 hours) No Yes Bizarre behavior (during first 5 days) No Yes Disorientation (during first 5 days) No Yes
en
-0.0430 -0.1132
135 92
I
-0.2596§ -0.Q708
Trail Making Part B -errors
I
Visual Reproduction
-0.1234 -0.0470
-0.0858 -0.2521§
-2.75 -3.57
-20.34 -19.06
-0.54 -0.44
0.30 -0.72/1
-1.86
-15.99
-0.39
-49.74~
-1.41~
-0.16 0.63
-15.31 -24.58
-0.42 -0.58
0.48 -0.4711
-18.54 -57.80
-0.45 -2.800
0.05 -2.00
-15.32 -82.07#
-0.41 -1.800
0.06 -1.27
195 28
-11.64~
115 112
-6.7~
0.49
220 5
-2.48 -17.40
210 16
-13.81~
-2.25
*Negative coefficients signify that higher values of the predictor are associated with poorer postoperative performance. tNegative differences signify decrement in postoperative performance. :j:Sample sizes will not always total 227 because of occasional missing data. §Product-moment correlation significant at p < 0.05. [One-way analysis of variance significant at p < 0.10. 1I0ne-way analysis of variance significant at p < 0.05. #One-way analysis of variance significant at p < 0.001.
nificantly lower postoperative scores on two of the tests. Eight more patients (3%) showed significant deterioration on three of the tests, and one patient displayed significant decrement on all four tests. Multiple regression analysis
At first glance, one may be a bit overwhelmed at the variety and number of factors that appear to be related to postoperative neurologic dysfunction as measured by the two neuropsychological tests utilized. Because ofthe possible interrelationships between many of these independent variables, multiple regression analysis was employed to determine which variables might be the more important predictors of poorer test performances in the postoperative period when taken into consideration with all other factors related to test performance at a univariate level. This statistical procedure was performed for each of the four outcome test scores: TM-A times, TM-B times and errors, and YR. Of the seven variables that were univariately related to TM-A time scores, multiple regression analysis revealed that duration of anesthesia, perioperative inser-
tion of an intra-aortic balloon, and electrolyte abnormalities more than 12 hours after operation provided the best subset of predictors of postoperative reduction in TM-A performance. Each of the three variables is adjusted for the effects of the other two in the regression analysis. For the TM-B time scores, six of the 12 univariately related variables made significant independent contributions to predicting poorer outcomes postoperatively by the multiple regression. These included admitting systolic blood pressure, heart size on preoperative x-ray film, preoperative use of chlordiazepoxide hydrochloride, duration of operation, number of days spent in the intensive care unit, and disorientation after the first 5 postoperative days. When taken together, those variables related to the TM-B error scores included age, admitting systolic blood pressure, preadmission use of propranolol, intraoperative insertion of an intra-aortic balloon, aortic cross-clamp time, and bizarre behavior after the first 5 postoperative days. For YR, the results of the multiple regression analysis indicated that, when combined, seven of the 13
592 Savageau et al.
significant variables provided the best equation for decreased test performance postoperatively, adjusting for the effects of the other variables. The seven variables were age, admitting systolic blood pressure, preoperative use of chlordiazepoxide hydrochloride, aortic cross-clamp time, administration of propranolol during the operation, difficult intubation, and postoperative depression. Discussion We first attended to possible sources of bias in these findings, such as selective drop out and variation in time of postoperative testing (since patients were never retested until they had left the intensive care unit). The 227 patients whose data are here presented represent 91% of the cohort of eligible patients entering the study in this time period. The remaining 23 patients included nine (4% of total) who died or were too sick to retest. These nine patients were older than the 227 here analyzed (60.0 versus 54.5 years), had markedly higher operative blood loss (3,000 ml versus 1,974 ml), much longer duration of anesthesia (9 hours, 54 minutes versus 6 hours, 44 minutes), and longer median stay in the intensive care unit (6 days versus 3 days). However, the other 14 patients not followed (including three refusals and II others with data awaiting editing or missed by the field investigators) were virtually identical to the sample of 227 on the aforementioned variables (all of which related both to severity of illness and to neuropsychological dysfunction in our sample of 227). We therefore estimate that if all nine patients in the death and illness group would have shown neuropsychological dysfunction had they been well enough to be retested, our estimates of dysfunction at the 9 day postoperative examination are probably within 4% for the kind of surgical population which is defined by our inclusion criteria. We observed that 11% to 17% of patients had a loss of function one standard deviation or greater on each of these specific test scores, and 30% of patients had such a loss on anyone of the four scores. (Each of these figures might be augmented by as much as 4%, if one presumes the severely ill and nonsurviving dropouts also incurred such damage.) These figures are in substantial agreement with data previously reported. 2. 3. 8 Although every effort was made to test all patients exactly 9 days after operation, this was not always possible due to scheduling problems on some weekends, temporary indisposition of patients, and in a few cases delayed recovery keeping the patient in the intensive care unit. We tested the relation of the number of days after operation that the postoperative examina-
The Journal of Thoracic and Cardiovascular Surgery
tion was administered to test performance and found no significant correlation for all scores except one (time on TM-A, r = 0.26). Hence, in general, those deviations from schedule which did occur did not create a source of bias confounding the findings. It should be noted that the correlation of TM-B time and TM-B error scores was only r = 0.43 and that somewhat different sets of medical variables are correlated with deterioration in performance on the two indices. Therefore, it is valuable to study both time and errors, because they probably reflect somewhat different neuropsychological processes. An important substantive finding was that older patients had greater decrements in test performance on all tests except the TM-A. Advanced age has been observed to be associated with lower performance on both parts of the Trail Making Test in other study populations."? The present finding suggests that the declines which come with aging in renal, metabolic, cerebrovascular, circulatory, sensory, and musculoskeletal functioning may also make older patients (as a group) more at risk neuropsychologically when they are subjected to a major heart operation and its attendant procedures. The finding that preoperative use of propranolol was associated with less of a decline in postoperative test scores has a plausible explanation. Among the adverse reactions to propranolol are short-term memory loss, visual disturbances, slightly clouded sensorium, and slowing of performance. Although some of our patients were receiving high dosages of propranolol (greater than 320 mg per day) preoperatively, almost none of them were receiving such high doses of this drug at the time of the postoperative testing. One-way analyses of variance of preoperative test scores showed differences among groups for the VR test (significant at p = 0.02) and TM-B time (marginal at p = 0.07). For TM-B errors, persons taking more than 320 mg of propranolol made an average of 1.4 errors, whereas the no use group made only 0.7 errors (analysis of variance not significant). These preoperative differences account in large measure for significance of the differences observed for the change scores. The preoperative use of chlordiazepoxide hydrochloride, on the other hand, although not associated with any preoperative differences in scores, was related to a decrease in test performance postoperatively when patients were, in nearly all instances, no longer receiving chlordiazepoxide hydrochloride. The reason for this is not apparent and warrants investigation. Several significant correlates of performance decrement seem to reflect preoperative weakness of the
Volume 84 Number 4
Neuropsychological dysfunction, I
593
October. 1982
heart. These include elevated end-diastolic pressure and heart enlargement on the x-ray film, both determined preoperatively, as well as use of the intra-aortic balloon during the operation, and to some extent, by inference, hypotension during the operation. Other perioperative factors significantly associated with reduced test performance seem to be related primarily to the severity of trauma associated with the operation. The four time variables-time on the pump, aortic cross-clamp time, duration of anesthesia, and total length of the operation-are not only indicators of the duration of the surgical stress, but are also lengthened by the occurrence of traumatic complications. Other variables reflecting the extent of surgical stress are extent of blood loss, difficult intubation, and in some respects, hypotension. For two of these factors, insertion of the intra-aortic balloon and difficult intubation, the rarity of occurrence created statistical problems in testing for significance of differences. Furthermore, use ofthe intra-aortic balloon (but not difficult intubation, as would be expected) was associated with longer surgical times, and observed neuropsychological effects may be secondary thereto. Administration of propranolol during the operation was associated with decrement in VR scores. This may also be consistent with extent of surgical stress in that this treatment is often given to correct supraventricular tachycardia, which often raises the risk of reduced blood perfusion to the brain. The postoperative neuropsychiatric observations of bizarre behavior and disorientation during the first 5 postoperative days were predictive of poorer test performance at 9 days. Since patients were not tested postoperatively while behaving bizarrely or disoriented, we are left with the possible hypotheses either that the Trail Making Test was picking up subclinical cognitive abnormalities or that we were observing in affected patients a rapidly remitting brain syndrome characterized by overt clinical disturbance in the first few days and only subtle effects on visual motor performance (but not memory) by the 9 day examination. The small numbers involved make these hypotheses tentative, pending replication. The Profile of Mood-State (POMS) depression scores, obtained at the same time as the neuropsychological tests, denoted only six persons at or above the means of about 20 (raw score) published in the POMS Manual for the standardizing groups of adult psychiatric outpatients." Nevertheless, even at far lower levels of depression (raw score of 6 or greater, the top quartile for our group), we found a significant average decrement in VR scores. Contrary to expectation, these levels of
depression were not associated with significantly slower times on the Trail Making Test. The implications of these findings of subtle psychoneurologic dysfunction in over 20% of heart surgery patients will become clearer when these same functions are retested 6 months after the operation. It can also be determined at that time if these short-term dysfunctions are predictive of longer term impairments in the broader recovery process. This work is currently underway in the recovery study. We are pleased to acknowledge the cooperation and assistance of Dr. Dwight E. Harken, HarvardMedical School; Dr. Robert Berger and Dr. John McCormick, University Hospital, Boston, Mass.; Dr. Ronald Weintraub and Dr. Robert L. Thurer, BethIsrael Hospital, Boston, Mass; Dr. T. J. Van der Salm, Dr. O. N. Okike, Dr. Bruce S. Cutler, and Dr. Ira Ockene, University of Massachusetts Medical Center, Worcester, Mass.; and Dr. Robert W. M. Frater, Albert Einstein Medical Center, New York, N. Y. REFERENCES Silverstein A, Krieger HP: Neurologic complications of cardiac surgery. Arch Neurol 3:601-605, 1960 2 HellerSS, Frank KA, MaimJR, BowmanFO, Harris PO, Charltan MH, Kornfeld OS: Psychiatric complications of open-heartsurgery. N Engl J Med 283:1015-1020, 1970 3 Tufo HM, Ostfeld AM, Shekelle R: Central nervous system dysfunction following open-heart surgery. JAMA 212:1333-1340,1970
4 Kilpatrick DG, Miller WC, Allain AN, Huggins MB, Lee WH: The use of psychological test data to predict openheart surgery outcome. A prospective study. Psychosom Med 37:62-73, 1975 5 Branthwaite MA: Neurological damage related to openheart surgery. Thorax 27: 748-753, 1972 6 Kimball CP: A predictive study of adjustment to cardiac surgery. J THoRAc CARDIOVASC SURG 58:891-896, 1969 7 Branthwaite MA: Prevention of neurological damage during open-heart surgery. Thorax 30:258-261, 1975 8 Javid H, Tufo HM, Najafi H, Dye WS, Hunter JA, Julian OC: Neurological abnormalities following open-heart surgery. J THORAC CARDIOVASC SURG 58:502-509, 1969 9 Gilman S: Cerebral disorders after open-heart operations. N Engl J Med 272:489-498, 1965 10 Brierly JB: Brain damage complicating open-heart surgery. A neuropathological study of 46 patients. Proc R Soc Med 60:858-860, 1967 11 Barash PG: Cardiopulmonary bypass and postoperative neurologic dysfunction. Am Heart J 99:675-677, 1980 12 Branthwaite MA: Detection of neurological damage during open-heart surgery. Thorax 28:464-473, 1973 13 Drew CE: Profound hypothermia in cardiac surgery. Br Med Bull 17:37-42, 1961 14 Michenfelder JD, Theye RA: The effects of profound hypocapnia and dilutional anemia on canine cerebral me-
5 9 4 Savageau et al,
The Journal of Thoracic and Cardiovascular Surgery
15
16
17
18
19
20 21
22
tabolism and blood flow. Anesthesiology 31:449-457, 1969 Aguilar MJ, Gerbode F, Hill JD: Neuropathologic complications of cardiac surgery. J THORAC CARDIOVASC SURG 61:676-685, 1971 Aberg T: Effect of open-heart surgery on intellectual function. Scand J Thorac Cardiovasc Surg 15: 1-63, 1974 (Suppl) Kessler J, Patterson RH: The production of microemboli by various blood oxygenators. Ann Thorac Surg 7:221228, 1970 Stockard JJ, Bickford RG, Schauble JF: Pressure dependent cerebral ischemia during cardiopulmonary bypass. Neurology 23:521-529, 1973 Ehrenhaft JL, Claman MA, Layton JM, Zimmerman GR: Cerebral complications of open-heart surgery. Further observations. J THoRAc CARDIOVASC SURG 42:514-526, 1961 Bass RM: Cerebral damage during open-heart surgery. Nature 222:30-33, 1969 Blachly PH, Kloster FE: Relation of cardiac output to postcardiotomy delerium. J THoRAc CARDIOVASC SURG 52:422-427, 1966 Kornfeld DS, Zimberg S, Maim JR: Psychiatric complications of open-heart surgery. N Engl J Med 273:287292, 1965
23 Frank KA, Heller SS, Kornfeld DS, Maim JR: Long-term effects of open-hean surgery on intellectual functioning. J THoRAc CARDIOVASC SURG 64:811-815,1972 24 Landis B, Baxter J, Patterson RH, Tauber CE: Bender Gestalt evaluation of brain dysfunction following openheart surgery. J Pers Assess 38:556-562, 1974 25 Reitan RM: Trail Making Test. Manual for Administration, Scoring and Interpretation, Indianapolis, 1958, Department of Neurology, Section of Neuropsychology, Indiana University Medical Center 26 Wechsler D: A standardized memory scale for clinical use. J Psychol 19:87-95, 1945. 27 Wechsler D, Stone CP: Instruction Manual for the Wechsler Memory Scale. New York, 1945, The Psychological Corporation 28 Gruvstad M, Hane M: Changes in mental functions after open-heart surgery operations. Methodological aspects. Scand J Thorac Cardiovasc Surg 10:215-220, 1975 29 McNair DM, Lorr M, Droppleman LF: Profile of Mood-States: Manual, San Diego, 1971, Educational and Industrial Testing Service 30 Davies ADM: The influence of age on trail making test performance. J Clin Psychol 24:96-98, 1968