This Article Abstract Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Grieco, G. Articles by Jonas, S. Search for Related Content PubMed PubMed Citation Articles by Grieco, G. Articles by Jonas, S.

(Stroke. 1996;27:858-874.)
© 1996 American Heart Association, Inc.

Articles

Evaluating Neuroprotective Agents for Clinical Anti-Ischemic Benefit Using Neurological and Neuropsychological Changes After Cardiac Surgery Under Cardiopulmonary Bypass

Methodological Strategies and Results of a Double-Blind, Placebo-Controlled Trial of GM₁ Ganglioside

Giacinto Grieco, MD; Marilyn d'Hollosy, PsyD; Alfred T. Culliford, MD Saran Jonas, MD

From the Departments of Neurology and Surgery (A.T.C.), New York University Medical Center (NY).

	Abstract

Top Abstract Introduction Subjects and Methods Results Discussion Statistical Appendix References

 
Background and Purpose Many neuroprotective agents (NPAs) are effective in acute experimental cerebral ischemia in animals. None have proven effective in human stroke trials. Even short treatment delays cause substantial efficacy loss. Cardiac surgery under cardiopulmonary bypass (CS-CPB) causes cerebral ischemia with cognitive impairment at a predeterminable time point and should permit efficient screening of NPAs for stroke benefit. We sought to develop sensitive methods to assess dysfunction from CS-CPB in a double-blind trial of the NPA GM₁ ganglioside.

Methods Eighteen GM₁ and 11 Control patients received GM₁ 300 mg or placebo, two doses intravenously, before nonemergency CS-CPB. Independent examiners administered structured neurological examinations and neuropsychological test batteries at Baseline and 1 day (Acute Postop; neurological only), 1 week (Early F/U), and 6 months (Long-term F/U) postoperatively; using defined procedures they employed ordinal Clinical Change Scores (CCSs) to quantify neurological cerebral, neurological noncerebral, and neuropsychological performance changes. Several methods to analyze CCSs and neuropsychological test score changes were evaluated.

Results The most sensitive indicators were the mean Acute Postop Neurologist's CCS-Cerebral (P<10^-5) and the mean Early F/U Neuropsychologist's CCS (P<.01), with statistically nonsignificant differences favoring GM₁. No significant mean changes in Neurologist's CCS-Noncerebral or any Long-term F/U CCSs occurred. CCS distributions and neuropsychological test score mean changes showed similar temporal patterns, with less sensitivity to change. When, as usual in prior CS-CPB studies, impairment was defined by neuropsychological test score declines (increases ignored), results were spurious.

Conclusions The strokelike cerebral dysfunction (maximal acutely, with eventual recovery) that occurs after CS-CPB is useful to screen NPAs for clinical efficacy. CCSs based on detailed neurological examination and neuropsychological testing are sensitive measures; refinement of this approach should enhance the efficiency of the CS-CPB model. Further testing of GM₁ is warranted.

Key Words: cardiopulmonary bypass • cerebrovascular disorders • clinical trials • cognition • gangliosides • neuroprotection

	Introduction

Top Abstract Introduction Subjects and Methods Results Discussion Statistical Appendix References

 
Numerous compounds, termed NPAs, have been shown effective in reducing histopathologic and functional sequelae of experimental global and/or focal ischemia.^{1 2} In acute stroke trials, however, NPAs have not convincingly demonstrated neuroprotection in humans.^{1 3} This discrepancy between animal and human study outcomes has been attributed in large part to the at least several-hour delay between stroke onset and initiation of treatment in human trials.³ In animal studies, the benefits of NPAs are maximal when they are first given before or within minutes after an insult.¹ As administration is delayed by up to several hours, benefits progressively diminish, then disappear. The preferred experimental models are those in which measurable but not devastating sequelae are produced in nearly all animals. We infer that the ideal human settings in which to evaluate an NPA are those in which cerebral ischemic injury could be expected to occur at a predictable time, with resulting limited deficits measurable in a large percentage of subjects. In this situation, as with animal studies, prophylactic or very early treatment could be instituted, and potential therapeutic benefits efficiently assessed.¹ CS-CPB provides such a setting. During CS-CPB, patients are subject to multiple cerebral emboli during a limited, predeterminable period of time^{4 5 6 7 8 9} and frequently develop ischemia-related measurable CNS impairment.^{8 9 10 11 12 13} If protection against the effects of cerebral ischemia is indeed possible in humans, it should be more readily demonstrable in CS-CPB than in stroke. If so, CS-CPB could be employed to in effect screen NPAs for potential benefit in human stroke, with the use of much smaller populations than needed for stroke trials. If an NPA shows no benefit in CS-CPB, then an acute stroke study is probably unwarranted. If a benefit is shown, then the NPA may also be efficacious in stroke, and an acute stroke study is probably justified. We present the results of a pilot trial of GM₁ ganglioside in CS-CPB and consider the implications for clinical studies of NPAs.

The objectives of this pilot study were (1) to investigate the acute prophylactic neuroprotective benefit of GM₁ ganglioside in reducing CNS dysfunction from CS-CPB, including estimating the magnitudes of any treatment effects and estimating population sizes for potential confirmatory studies (statistically significant treatment effects were not expected in this small study), and (2) to develop sensitive methods for assessing CNS dysfunction resulting from CS-CPB that would be appropriate for evaluating potential treatments.

	Subjects and Methods

Top Abstract Introduction Subjects and Methods Results Discussion Statistical Appendix References

 
Patient Selection Criteria
Candidates were adult men and women who were to undergo nonemergency CABG surgery for symptomatic coronary artery disease and/or nonemergency heart valve replacement for valvular heart disease. They were required to have the following: normal cognition; normal motor function (minor, stable peripheral deficit that would not affect performance on the Grooved Pegboard Test was allowable); competency in English; no clinically significant depression or anxiety; no recent psychoactive substance use disorder; no history of major psychiatric illness; no recent experimental drug treatment; and no prior ganglioside treatment. Patients with general circulatory insufficiency were excluded. Signed, witnessed, informed consent was obtained from all study participants. Institutional review board approval was granted. The study was performed in accordance with institutional guidelines.

Study Design and Schedule of Procedures
This was a double-blind, placebo-controlled, parallel-group pilot study of 30 patients, with unbalanced randomization to either of two treatment groups, GM₁ ganglioside or placebo, in a ratio of 2:1, respectively. The unbalanced randomization was used to increase the likelihood of detecting GM₁ effects in a limited-size study.

All patients underwent Baseline neurological examination and neuropsychological testing, generally 1 day preoperatively, before they received any test medication. Two doses of intravenous study medication (GM₁ 300 mg or placebo) were administered: one the evening before and one the day of surgery, with exceptions described below. An Acute Postop neurological examination was scheduled for the day after surgery. Early and Long-term F/U neurological examinations and neuropsychological testing were performed approximately 1 week and 6 months postoperatively, respectively. To preclude interobserver bias, neurological examinations and neuropsychological testing were performed independently by a neurologist and neuropsychologist. The examiners also independently arrived at CCSs, as described below.

Study Medication
Gangliosides are complex sialic acid–containing glycosphingolipids that are normal components of vertebrate cell plasma membranes and are particularly abundant in the nervous system. They consist of a hydrophilic negatively charged oligosaccharide moiety (on the basis of which they are classified) linked to a hydrophobic ceramide moiety (an N-acylated long-chain base of variable length). The molecular species classified as GM₁ contain a specified tetrasaccharide moiety with a single sialic acid residue.

In in vitro and in vivo studies of experimental CNS ischemia^{14 15} or trauma,¹⁶ GM₁ ganglioside has been shown to exert acute protective effects when administered before or within hours after insult, and longer-term enhancement of neural regeneration, when administered for days or weeks.^{17 18} The early protective effect may be mediated by antagonism of excitotoxin-mediated stimulation of -aminobutyric acid receptors without interfering with physiological activation of such receptors.^{19 20} Trials in acute human stroke have yielded suggestive evidence of efficacy with 100 mg IM or IV daily dosing for from 2 to 4 weeks (initial dose of 100 or 200 mg), with better results in post hoc defined "early-treated" subgroups.^{21 22 23 24}

Neurological Examination
The neurological examination consisted of six major categories of assessment: (1) alertness and cognition ("bedside" mental status testing, including level of consciousness, orientation, aural comprehension, verbal expression, immediate and delayed recall, calculations, and recall of recent US presidents); (2) cranial nerves (including confrontation visual field testing; visual pursuit and saccades; pupillary size, shape, and reactivity; fundoscopy; hearing; mandibular, facial, palatal, oropharyngeal, and lingual motor function; and sensation in the trigeminal distribution); (3) motor system (strength, tone, and bulk of spinal nerve–innervated musculature); (4) reflexes (muscle stretch and plantar); (5) gait and coordination; and (6) sensation (pinprick, light touch, proprioception, and vibration). The results of neurological examinations were reported on standardized forms; an "other" category was provided for reporting any abnormalities (eg, mood disturbance, tremor) that did not fit into the aforementioned categories. To avoid interobserver differences, all Baseline, Acute Postop, Early F/U, and Long-term F/U neurological examinations were performed by the same attending neurologist (G.G.). Eleven of the 30 patients received a second preoperative neurological examination after the Baseline one, since surgery was not performed the day after the Baseline examination, as required by the protocol; no changes were found. Three patients had their last preoperative examination 2 days before surgery, and one had it 4 days before surgery.

Each patient's post-Baseline neurological examinations were performed with a copy of the Baseline (and no other) examination form in hand so as to facilitate detection of changes in functioning. At Acute Postop neurological examinations, it was common for some aspect(s) of functioning not to be fully evaluable: typically, gait could not be assessed because the patient was still confined to bed; often motor effort was diminished because exertion resulted in incisional pain; fundoscopy often was impractical; and the gag reflex was not tested in patients with a known exaggerated response to avoid inducing needless sternal pain. These limitations did not adversely affect the process of determining the neurologist's CCSs for cerebral functioning.

Neurologist's CCSs
The examining neurologist compared the findings from each postsurgical neurological examination with those from the Baseline examination to identify and score all observed differences. If some aspect of functioning assessed at Baseline could not be assessed at a later examination, it was presumed not to have changed; nearly all such occurrences were at the Acute Postop exam. Changes in peripheral sensation resulting from surgical incisions were ignored. All other changes were assessed in the following manner: (1) a determination was made as to whether the change represented an alteration in cerebral (ie, cerebral hemispheres or brain stem) or noncerebral (eg, peripheral course of cranial nerves, spinal cord, spinal nerve roots) neurological functioning or whether it represented a psychiatric change; (2) if an etiology for the observed change could be determined (eg, recent narcotic administration resulting in altered cognition or level of consciousness), it was documented; (3) the severity and direction of the change were taken in combination and rated as 3=marked improvement, 2=moderate improvement, 1=mild improvement, 0.5=minimal improvement, 0=no change, -0.5=minimal worsening, -1=mild worsening, -2=moderate worsening, or -3=marked worsening.

For each postsurgical neurological examination (Acute Postop, Early F/U, and Long-term F/U), one composite change score was assigned, using the same scale described above, for each of the two domains of neurological function: cerebral and noncerebral. If no changes had been detected in a domain, the composite score assigned was zero; otherwise a score was assigned on the basis of a clinical assessment of overall change in that domain.

Neuropsychological Test Battery
A battery of standardized tests^{25 26 27 28 29 30 31 32 33 34 35 36} yielding 22 scores was used (Table 1) to probe orientation, mental control, memory, language, visual-perceptual-motor skills, intellect, and fine motor speed in a manageable amount of time (<1.5 hours) in an acute-care environment. Alternate forms were used on different visits for those tests that provided them.^{25 29 32 35} Test-administration–related variance was limited by having all administrations (except one of the baseline sessions) performed by the same examiner (M. d'H.). A score of 11 (maximum of 14) on the General Information subtest of the Randt Memory Test and an age-adjusted scaled score at or above the fifth percentile for the Information subtest of the Wechsler Adult Intelligence Scale (revised) were required at baseline for enrollment.

View this table: [in this window] [in a new window]   Table 1. Neuropsychological Test Battery

Neuropsychologist's CCSs
The neuropsychologist assigned a CCS for each Early F/U and Long-term F/U test battery administered. The CCSs were determined subsequent to calculation and compilation of the neuropsychological test scores and were based on the neuropsychologist's clinical assessment of overall qualitative and quantitative differences from baseline performance. The possible scores were as follows: 3.5=marked improvement, 2.5=moderate improvement, 1.5=mild improvement, 0.5=minimal improvement, 0=no change, -0.5=minimal worsening, -1.5=mild worsening, -2.5=moderate worsening, and -3.5=marked worsening.

Surgical and Safety Data
A standard reporting form was devised for anesthetic and study personnel to record surgical data, including surgery-related times of interest (for 12 events, starting with anesthetic induction, first incision, and pump start, and continuing through discharge from the intensive care unit), blood loss, and intraoperative complications. The examining neurologist and the study nurse reviewed the operative records, nursing notes, progress notes, and laboratory test results to verify the entries and to discover potential adverse events.

Statistical Methods
Means and SDs were calculated (by treatment group [GM₁ or Control] and for the overall population) for demographic variables; surgery-related time periods; time intervals between assessments; neuropsychological test scores at Baseline, Early F/U, and Long-term F/U; and the Neurologist's and Neuropsychologist's CCSs. Upper and lower 95% confidence limits were calculated for the means of the CCSs. For neuropsychological test results, only raw scores were used in analyses, since only intraindividual changes were of interest; adjustments for level of education and other factors, as allowed for scoring some of the tests,^{25 32} were not made. Scores for several neuropsychological tests showed highly skewed distributions and were therefore transformed appropriately before statistical significance testing: the two Grooved Pegboard and the two Trail Making Test scores (all timed tasks) were transformed by taking 100/raw score; the Symbol Digit Modalities score was transformed by taking log₁₀ (raw score+1). Between-group demographic comparisons were made by paired t tests. CCSs (by treatment group and overall) at each of the three postoperative assessments were analyzed by Student's t tests, with between-group comparisons made by paired t tests. As reported in "Results," there were few instances in which changes in cerebral functioning on neurological examination were due to identifiable causes. Therefore, no additional analyses of Neurologist's CCSs-Cerebral were performed to take into account the effects of such causes. Neuropsychological test scores were analyzed both univariately and multivariately for between-group and for postoperative differences versus Baseline. Comparisons were made by paired t tests, ANOVA, and ANCOVA. Individual patient demographic data were considered as possible covariates to reduce error variance in the analyses even though no bias between groups was found. Principal component factor analysis with rotation was used to examine whether groups of correlated scores could be combined into meaningful summary scores.

The SDs of the 22 neuropsychological test scores at Baseline for the overall population were used to classify each Early and Long-term F/U test score as follows: "Up," or improved by, 2 SDs from the Baseline score; "Up," or improved by, 1 SD from the Baseline score; changed by <1 SD from the Baseline score; "Down," or worsened by, 1 SD from the Baseline score; and "Down," or worsened by, 2 SDs from the Baseline score. Analyses were made of neuropsychological test score improvements and worsenings of 1 SD and 2 SDs (results in the "Statistical Appendix"). For Early and Long-term F/Us, "Net" scores were calculated for each patient equal to the number of scores Up 1 SD minus the number of scores Down 1 SD (see "Discussion" and "Statistical Appendix" for further considerations). Means and SDs (by group and overall) were calculated for the per-patient numbers of scores Up 1 SD, the numbers Down 1 SD, and the Net scores at the two follow-up times. Upper and lower 95% confidence limits were calculated for the means of the Net scores. Net scores (by treatment group and overall) at Early and Long-term F/U were analyzed by Student's t tests, with between-group comparisons made by paired t tests. Since few patients had scores that changed by 2 SDs at any follow-up visit, statistical analyses similar to those done for changes 1 SD would not have been appropriate and were not done. The distributions of test score changes (Up 1 SD, changed <1 SD, or Down 1 SD) by treatment group and for the overall population at Early and Long-term F/U were compared by ² tests.

All tests of statistical significance were two-tailed.

Dosing Errors and Deviations
Two patients unintentionally did not receive any study medication. One patient complained of right forearm and thigh itching immediately on initiation of infusion of the first dose; the infusion was stopped with <15% of the dose administered, and no further doses were given. Two patients received reduced second doses (one 67%, the other 64%, of the full dose) in error. As these deviations from the planned dosing schedule occurred, it was decided (with treatment codes unbroken) to follow each patient for all scheduled observations, to analyze the two patients who received reduced second doses with whatever group(s) they had been assigned to (both were in the GM₁ group), and to combine the data from the two untreated patients and the patient with the aborted first dose (GM₁ group) with the placebo patients' data to form the Control group.

	Results

Top Abstract Introduction Subjects and Methods Results Discussion Statistical Appendix References

 
Anesthesia and Perioperative Events
Twenty-nine of the 30 operations were performed by the same attending cardiac surgeon (A.T.C.). The same basic anesthetic and CPB protocol was used for all patients: membrane oxygenator, arterial filter, pump primed with crystalloid, alpha-stat protocol, thiopental induction, and balanced anesthesia with oxygen, enflurane, sufentanil, and midazolam, with metocurine and/or pancuronium for muscle relaxation. Pressors, inotropic agents, and antiarrhythmics were used as needed. Intraoperative blood losses were not significant. There were no instances of significant hypotension, hypertension, or systemic hypoxemia. Aside from the surgical procedures themselves and the use of extracorporeal bypass, there were no recognizable preoperative or intraoperative events that could be deemed to have possibly resulted in CNS sequelae. One postoperative event did: one patient (placebo) developed an air leak around his chest tube on postoperative day 3, as well as sternal wound dehiscence and pulmonary insufficiency, resulting in anoxic encephalopathy, on postoperative day 5. He was thereupon discontinued from the study (thus, he had an Acute Postop but no Early or Long-term F/U evaluation). Except for the surgical incisions (resulting in interruption of small-diameter peripheral nerve branches), there were no recognizable events at any time that could be deemed to have resulted in peripheral nervous system deficits (eg, pressure neuropathies).

Patient Population, Demographics, and Surgical Characteristics
Of the 19 patients randomized to GM₁, 18 were analyzed in the GM₁ group. The Control group consisted of the 9 placebo patients, the 2 untreated patients, and the 1 GM₁ patient with the aborted first dose (see Table 2 for enrollment figures). Acute Post-op neurological examinations were performed on all 30 patients and Early F/U assessments (neurological examinations and neuropsychological tests) on 29. Four patients (1 GM₁, 1 placebo, 2 untreated) of the 29 were subsequently lost to follow-up. Twenty-five patients had Long-term F/U assessments, on average 9 months postoperatively: one (GM₁ group) had only neuropsychological testing (no neurological examination).

View this table: [in this window] [in a new window]   Table 2. Patient Enrollment and Survival

Patients undergoing CABG surgery without valve replacement were overwhelmingly male and constituted the majority of both treatment groups (Table 2). The two groups were well balanced with respect to demographic variables, surgical times, and time intervals between each of the assessments and surgery (Table 3), with no statistically significant differences between them. The mean Baseline scores for the two groups were comparable for each of the 22 neuropsychological tests and were also comparable to published norms (data not shown).

View this table: [in this window] [in a new window]   Table 3. Demographics, Surgical Times, and Assessment Intervals

Neurological Examinations: Qualitative Observations
At Acute Postop neurological examinations, deficits in cerebral functioning included lethargy, small pupils (either of these even without narcotic administration), disorientation, difficulty concentrating, impaired immediate or delayed recall, impaired remote memory, dyscalculia, dysnomia, and dysarthria. No patients had developed strokelike deficits or hypoxic encephalopathy. CNS depressants may have resulted in impaired cerebral functioning in 7 patients (3 GM₁ and 4 Control) of the 22 with negative Neurologist's CCSs-Cerebral. At Early F/U, all patients were fully alert, none were receiving narcotics, and none had medication-induced impairment of neurological functioning. At Long-term F/U, one patient (GM₁ group) had impairment of concentration and immediate recall secondary to a severe acute reactive depression linked to a stressor unrelated to his medical condition, and was given a "moderately worse" Neurologist's CCS-Cerebral; all others had no more than mild improvement or worsening.

Five patients (4 GM₁ and 1 Control) had negative Noncerebral CCSs at the Acute Postop evaluation: 4 had hyporeflexia or areflexia, and 1 had nonspecific tremulousness. At Long-term F/U, changes in noncerebral performance were common (see below): a number of patients demonstrated generalized improvement in motor strength, in apparent association with increased physical activity; elderly patients often demonstrated mild worsening in hearing, proprioception, or vibration sensation.

Neurologist's CCSs
For the Acute Postop examination, there was statistically highly significant worsening in Neurologist's CCSs-Cerebral for both treatment groups and for the study population as a whole (Fig 1 and Table 4). The magnitude of mean Cerebral CCS worsening for the GM₁ group (-0.694) was less than for the Control group (-1.042), but the treatment difference (0.348) was not statistically significant (see "Sample Size Estimates" and "Discussion" for further considerations). At Early F/U, while there was still statistically highly significant worsening of mean Neurologist's CCSs-Cerebral (-0.278 and P=.008 for the GM₁ group and -0.259 and P=.007 for both groups combined; the Control group value of -0.227 did not attain statistical significance), the change was a fraction of that at Acute Postop, and the treatment difference (0.051, again not significant) was much less. By Long-term F/U, differences from Baseline in cerebral performance had essentially vanished: no mean changes were statistically significant. The return of cerebral functional performance to Baseline levels over time can also be seen by looking at the distributions of Neurologist's CCSs (negative or worse, versus zero or unchanged, versus positive or better) in Fig 2 and Table 4. From the Acute Postop to the Long-term F/U evaluations, the proportion of all (GM₁+Control) patients with negative Neurologist's CCSs-Cerebral decreased from nearly 75% to 12.5%, while the proportion of patients with zero scores increased from less than 25% to 75%.

View larger version (28K): [in this window] [in a new window]   Figure 1. Neurologist's CCSs-Cerebral.

View this table: [in this window] [in a new window]   Table 4. Clinical Change Scores

View larger version (24K): [in this window] [in a new window]   Figure 2. Distribution of Neurologist's CCSs-Cerebral for All patients (GM1+Control).

In contrast to the Cerebral CCSs, the Noncerebral CCSs (Table 4) demonstrated minimal changes in mean scores (all magnitudes <0.2, with only one—for all patients at Acute Postop—attaining statistical significance), with no clear trends over time. On the other hand, the proportion of patients with zero scores (Table 4) declined from 83% at Acute Postop to 37.5% at Long-term F/U, while the proportions of patients in the worse and the better categories increased, for the reasons previously given.

Neuropsychologist's CCSs
The mean Early F/U Neuropsychologist's CCSs demonstrated (Fig 3 and Table 4) statistically significant worsening for the two groups separately (-0.722, P=.032 for GM₁; -1.045, P=.028 for Control) and in combination (-0.845, P=.002). As with the Acute Postop Neurologist's CCSs-Cerebral, the difference in favor of GM₁ was not statistically significant, and approximately 75% of all patients had negative scores (Fig 4 and Table 4). (Although 7 patients had positive Early F/U Neuropsychologist's CCSs, 6 had the lowest possible positive score.) By Long-term F/U, as with the Neurologist's CCSs-Cerebral, differences from Baseline in mean scores had essentially vanished; no changes were statistically significant. The distribution of CCSs showed that the proportion of all patients unchanged from Baseline increased from 0% at Early F/U to 40% at Long-term F/U.

View larger version (24K): [in this window] [in a new window]   Figure 3. Neuropsychologist's CCSs.

View larger version (23K): [in this window] [in a new window]   Figure 4. Distribution of Neuropsychologist's CCSs for All patients (GM1+Control).

Neuropsychological Test Score Changes
As seen in Table 5, at Early F/U nearly all tests (19 of 22 for the combined groups) had worse mean scores than at Baseline, while at Long-term F/U nearly equal numbers of tests had mean improvement as had worsening (10 and 11, respectively, for the combined groups). The differences were statistically significant for only a few tests. At Early F/U, all statistically significant differences were in the direction of worsening. These results indicate that performance was, on the whole, worse at Early F/U than at Baseline, while it was not changed overall at Long-term F/U.

View this table: [in this window] [in a new window]   Table 5. Group Mean Changes From Baseline in Neuropsychological Test Scores

Principal component factor analysis yielded a partition of the neuropsychological tests into four sets of correlated scores. Summary values suggested by the analysis did not demonstrate significant changes (results not shown). The results of additional analyses of neuropsychological test score changes are presented in the "Statistical Appendix." These also demonstrated worsened performance at Early but not Long-term F/U, without significant treatment differences, but with less sensitivity than Neuropsychologist's or Neurologist's CCSs-Cerebral.

Sample Size Estimates
One of the objectives of this trial was to estimate sample sizes for potential future studies to confirm possible treatment benefits. The statistically nonsignificant treatment differences seen in favor of GM₁ for the Acute Postop Neurologist's CCSs-Cerebral and for the Early Postop Neuropsychologist's CCSs correspond to 5.80% and 4.61%, respectively, of the ranges of possible scores. Sample size estimates for these two end points (for =0.05 and ß=0.2) yielded 150 and 530 patients per treatment arm, respectively (Table 6). In contrast, the smaller treatment differences of 0.85% and 0.81% of the ranges of possible scores for the Early Postop Neurologist's CCSs-Cerebral and for the Net scores, this time in favor of the control group, resulted in larger sample size estimates of 2798 and 1913 for the same and ß errors.

View this table: [in this window] [in a new window]   Table 6. Sample Size Estimates

Subgroup Analyses
We have presented the results of analyses of data for all available patients ("survivors") at each assessment for the GM₁, Control, and combined (GM₁+Control) groups. The results, not presented, of analyses of data for the following subgroups at each assessment showed inconsequential differences from those presented: placebo survivors; nonplacebo survivors in the Control group; Control patients with Early F/U data; GM₁+Control patients with Early F/U data; GM₁ patients with Long-term F/U data; Control patients with Long-term F/U data; and GM₁+Control patients with Long-term F/U data.

	Discussion

Top Abstract Introduction Subjects and Methods Results Discussion Statistical Appendix References

 
NPA Trials in Stroke and CS-CPB
NPAs are of profound interest as potential prophylaxis and treatment for stroke. In stroke treatment trials to date, NPAs have not clearly demonstrated clinical benefit. We know of no studies of the prophylactic use of NPAs in populations at risk for stroke. Stroke studies require thousands of patients per treatment arm for prophylaxis,³⁷ several hundred for treatment,³⁸ and major commitments of time and resources. Prophylaxis studies require long periods of observation, typically 1 or more years, for infrequent events. In treatment studies, many potential subjects are excluded because of failure to be evaluated and started on study medication before expiration of the treatment "window" (typically 6 hours from stroke onset).²⁴ Small-scale pilot treatment studies may provide important safety information and disclose logistical problems, but they have not been useful for providing evidence of efficacy. Clinical investigation of potential stroke treatments is hampered by the fact that studies in relatively small stroke populations do not provide adequate power to identify agents and best doses for subsequent full-scale trials.

These problems may be effectively addressed by finding a clinical surrogate for stroke that would permit evaluation of efficacy with far fewer patients observed for less time than needed in a stroke study. CS-CPB appears to be an excellent candidate. CS-CPB has been associated with CNS morbidity, including stroke, anoxic encephalopathy, and impaired cognitive performance. Stroke and overt encephalopathy, being very infrequent, need not concern us here. On the other hand, impairment in cognitive functioning, recognized since the earliest days of CS-CPB, is a frequent occurrence, even with modern pump and filtration devices. Both microembolism and hypoperfusion have been implicated in the pathogenesis of such deficits. Pathological and imaging studies strongly support an ischemic etiology for the cognitive changes observed. Because cognitive change is so common and since the time of insult can be known in advance, CS-CPB should be useful for evaluating NPAs as prophylactic or acute (well within narrow therapeutic time windows) anti-ischemic therapy, using a fraction of the patients, and taking a fraction of the time, that would be needed for stroke studies. CS-CPB trials could be used to select drugs and dosing regimens for large-scale stroke studies, in addition to identifying treatments for cognitive impairment associated with CS. For studies of NPAs in CS-CPB to be carried out efficiently, neurological and neuropsychological assessments of cognitive change need to be both valid and sensitive. Noncognitive neurological deficits also need to be assessed.

If, on a cellular level, the mechanisms of neuronal injury in stroke and in postcardiac-surgery cognitive impairment are similar, we should observe clinical similarities. In stroke, particularly if not extensive and not complicated by cerebral edema or hemorrhage, deficits are maximal acutely. Recovery, if it occurs, tends to be most rapid early on and reaches a plateau after several months. Confirming the appropriateness of CS-CPB as a stroke model, our results show that this pattern, with a foreshortened time scale, also applies to cognition after CS-CPB.

Assessing Cognitive Change After CS-CPB by Neuropsychological Testing
Early reports of cognitive disruption after CS-CPB were based on clinical observations (cognitive disruption may have been more overt with early CPB equipment and techniques).³⁹ Subsequently, formal cognitive testing was introduced and became routine. Commonly, a battery of standardized neuropsychological tests is administered preoperatively and at one or more postoperative time points. Neuropsychological testing is usually performed with a descriptive goal in mind: to characterize a subject's level of performance in one or more areas of cognition, usually with respect to norms (eg, for age and level of education). When carried out on individuals suspected of having an acquired impairment, such testing is usually performed to identify overt deficits after they have been incurred. Without a premorbid basis for comparison, the deficits are identified by test scores being worse than expected by, for example, 1 SD. In such situations, the individual can only be characterized by comparison to a reference population. When, however assessing cognition both before and after a potentially acute disruptive event such as cardiac surgery, the goal should be to assess intraindividual change, particularly subtle change, in performance. This calls for a different approach to interpreting test results and, if one is studying a population of such patients, for some means of quantifying change in function over an entire test battery. There is as yet no established method for such quantification. In almost all reported studies in CS-CPB, in keeping with the traditional use of population-based norms, only neuropsychological test score decreases are taken into consideration, with subjects classified simply as "impaired" or not at each of the postoperative assessments. Typically, a given threshold of decline (usually of 1 SD, determined from the baseline scores of the subject cohort) in any single test score is defined as meaningful, and a subject is defined as "impaired" if an arbitrary minimum number (for example, 2 of 10) of test scores show threshold or greater declines.^{8 9 10 11 12 13 40 41 42 43} We term this method an MNTD analysis.

We were concerned that MNTD analysis is inappropriate to characterize neuropsychological change after CS-CPB because of the following: (1) It is a population-norm–based rather than an individual-change–based approach. (2) If one looks only for declines, one can find only declines. (3) Having a dichotomous (impaired or not impaired) as opposed to a graded or continuous outcome variable reduces sensitivity to change and to detecting the effects of treatments, different anesthetic protocols, and other factors. (4) Taking score declines into consideration only if they exceed a threshold results in anomalous outcomes. For example, a patient with declines, in nearly all test scores, that almost reach threshold would be classified as not impaired, while another patient with the minimum required number of threshold declines, but otherwise unchanged or increased scores, would be classified as impaired. (5) Not taking score increases into consideration ignores the "learning" phenomenon observed with certain tests. Failure to improve on such tests may represent a deficit. (6) After administering a set of neuropsychological tests free of learning effects to a population of normal subjects at two time points, random variation would result in equal numbers of score increases and decreases, normally distributed. If tests with learning effects were included, there would be an excess of increases. Given enough tests, individual subjects would likely have both declines and increases, even exceeding a 1 SD threshold, and could be inappropriately classified as impaired.

The last point is best demonstrated mathematically. Equation 3 from the "Statistical Appendix" tells us that if a set of N tests free of learning effects is administered on two occasions, the probability of at least k results decreasing by, for example, 1 SD each is the same as the probability of k results increasing by 1 SD each. Using Equation 3, we calculated the probabilities of finding from zero to 3 test score declines (the same probabilities apply for increases) for three different thresholds of change, for batteries of 10 and 22 tests (Table 7). For a 1 SD threshold, for example, the probabilities of a patient having 1 or 2 declines of 10 scores are 82.2% and 48.7%, respectively. With a 22-test battery, the corresponding probabilities rise to 97.8% and 88.5%, respectively. Since the test scores are not completely independent and since intraindividual variation should be less than interindividual variation, the percentages of declines (and of increases) actually observed are expected to be less than the predicted ones. We show in the "Statistical Appendix" that our results fulfill this expectation.

View this table: [in this window] [in a new window]   Table 7. Probabilities That Random Variation Will Result in Minimum Numbers of Declines (Increases) in Test Scores (for Batteries of 10 and 22 Tests)

To rigorously separate out the effects of anesthesia and repeated cognitive testing, it would be necessary to conduct a CS-CPB study with two control groups comparable in terms of age, education, and possibly cardiovascular disease: one undergoing surgery with general anesthesia of a duration similar to that of the CS-CPB group but without CPB or major vessel manipulation, and one not undergoing surgery. All subjects would have to undergo the same assessments at similar time intervals, ideally by the same examiners who would be blind to an individual's group. On the other hand, to adequately assess the results of therapeutic intervention on neuropsychological test performance after CS-CPB, those control groups would not be necessary for detecting outcome differences. Whether assessing the effects of CS-CPB itself or of a therapeutic intervention, meaningful and sufficiently sensitive methods that take into consideration the magnitudes and directions of changes in all administered tests are needed. MNTD analysis does not do so and results in artifactual labeling of subjects as impaired. Whatever analytic method is used must take into account the reported studies, which show that the changes (improvements and worsenings) in performance (on individual tests or entire batteries) associated with CS-CPB, and consequently the maximum benefit possible from any intervention, are of a magnitude similar to that of the measurement errors associated with the neuropsychological testing process itself. CS-CPB–associated cognitive change should not be thought of as an impairing "event" that either occurs or does not occur in any one individual but rather as a continuum subsuming both improvements and worsenings; where a particular subject falls on this continuum is subject to measurement error. Estimates of the degree of CS-CPB–associated dysfunction must be based on population statistics, not inferred rates of impairment. The Neuropsychologist's CCS appears to be an appropriate and sensitive global statistic. We anticipate more focused test selection and further refinements in assessing change to enhance the sensitivity of this approach in future studies.

Neurological Examination and the Assessment of Change in Function
Preoperative neurological examinations were performed with the intention of assuring that patients did not have any excludable conditions and to serve as bases of comparison for postoperative examinations. The postoperative examinations were conducted primarily to identify patients who may have suffered stroke, transient ischemic attack, cerebral hypoxia, or other CNS injury, since such could result in cognitive impairment detectable by neuropsychological testing; there were no such cases. The postoperative examinations were also conducted to determine whether any mechanical or peripheral neurological impairments had arisen, since these could affect performance on some tests; again, there were no such instances. Since we did not expect to observe clinically significant changes at Early F/U and expected that medication effects would obscure the results of Acute Postop examinations, mental status testing was intentionally not extensive. We had not anticipated the usefulness of neurological examinations in this setting. Bedside mental status testing together with a detailed neurological examination, performed in consistent fashion, turned out to be a sensitive means of assessing change in cerebral functioning in CS patients. We found that changes in neuropsychological test scores can be associated with clinically detectable alterations in neurological and/or cognitive functioning. It is likely that with increased structure to and greater depth of mental status testing, sensitivity to cognitive change and to potential treatment effects would be substantially enhanced.

Stroke Versus Cognitive Change After CS-CPB
In most cases of stroke, deficits are maximal acutely, recovery (if any) is greatest in the first weeks or months postictus, and there unequivocally is persistent residual impairment. Cases of progressing deficit or of total recovery are infrequent. By 3 to 6 months after the episode, most patients have reached a plateau where further change, barring a new stroke or other catastrophe, is gradual. Treatment trials thus typically extend to 3 to 6 months postictus, with the degree of recovery at the last assessment being a principal outcome measure. Any benefit from substantially prolonging the trial period (eg, to 1 or 2 years) would be more than offset by higher attrition and confounding events such as recurrent stroke, cardiac morbidity, and mortality from all causes. With CS-CPB, cognitive impairment (as opposed to the infrequent overt stroke), while also maximal acutely, has a foreshortened time scale of recovery, with the greatest recovery occurring in days to weeks and complete or near complete recovery the rule in several months. Thus, in a treatment trial in CS-CPB, the final efficacy assessment needs to be performed rather soon after surgery. We propose that it be approximately 1 week later (or just before discharge, if that be earlier) for the following reasons: (1) The acute physiological alterations resulting from surgery, anesthesia, and related pharmacological interventions (including narcotic administration) have resolved. (2) Since the patient is still in the hospital, the risk of being lost to follow-up is essentially nil. (3) Postponing the final efficacy assessment will diminish the likelihood of measuring treatment effects because of continuing recovery; greater variability in time interval from surgery; and patient dropouts.

The use of a 6-month or 1-year postoperative assessment would make it impossible to detect a treatment effect since patients will have essentially fully recovered by then.

Implications for GM₁
When this trial was designed, human safety data were limited to doses of no more than 300 mg. Animal studies, however, suggested an acute neuroprotective effect might require a 1000-mg single dose in humans. We took advantage of the long half-life of GM₁ (>24 hours) to give two 300-mg doses, typically within 8 to 12 hours of each other (one the evening before and one shortly before surgery) so as to achieve intraoperative tissue concentrations approximately equivalent to those expected after a 500-mg single dose. By the use of pretreatment and higher doses than in prior GM₁ stroke studies, it was hoped that a potential therapeutic benefit could more readily be detected. The observed treatment differences suggest a possible therapeutic effect of GM₁ in CS-CPB and, by inference, that higher acute dosing than heretofore used in stroke trials (100 or 200 mg initial dose, then 100 mg daily^{21 22 23 24} ) would be appropriate. (Since this study was completed, the safety and tolerance of a loading dose of 1000 mg, followed by daily dosing of 200 mg, have been shown in a pilot study in patients with Parkinson's disease.⁴⁴ ) With improvement in study methodology, and the option to use higher doses (more likely to demonstrate acute protective benefits) than in this trial, future clinical investigation of GM₁ in CS-CPB may result in more apparent clinical effects and/or reduction of sample size requirements.

Conclusions
(1) After nonemergency CS-CPB, in patients with normal cognition and no symptoms or signs of cerebrovascular compromise preoperatively, there was impairment of cerebral (including cognitive) functioning. The impairment was most pronounced and statistically highly significant at the Acute Postop assessment (by neurological examination) approximately 1 day after surgery. Impairment at this time point could only partly be accounted for by administration of CNS depressants. By all measures (Neurologist's CCSs, Neuropsychologist's CCSs, Net per-patient number of score changes of 1 SD, and population-based distributions of score changes), there was significant cerebral dysfunction/cognitive impairment at Early F/U but no significant change from Baseline performance at Long-term F/U (6 to 21 months after surgery).

(2) In terms of relative sensitivity to change from Baseline and to possible treatment differences, Acute Postop Neurological Cerebral CCSs appeared to be most sensitive, followed closely by the Early F/U Neuropsychological CCSs. The Early F/U Net score and Neurological Cerebral CCSs appeared to be substantially less sensitive.

(3) MNTD analyses of neuropsychological test changes yield rates of apparent impairment that vary substantially with the parameters (threshold for decline and minimum number of declines required) arbitrarily chosen. MNTD analyses are likely to result in misleading estimates of frequency of impairment (typically, overestimates) and may even lead to the conclusion that there has been impairment when performance has actually improved. Such would reduce if not eliminate the possibility of detecting treatment effects or of comparing rates and severities of cognitive deficits in patients subjected to different anesthetic and bypass techniques.

(4) At Long-term F/U, there were no demonstrable alterations, other than age-related declines and nonspecific motor improvements, in noncerebral neurological performance.

(5) Cognitive change after CS-CPB, even in patients cognitively normal preoperatively and who are without known cerebrovascular compromise, appears to be a useful clinical model to efficiently evaluate NPAs for potential efficacy in stroke. Refinement of statistical methods, selection of more appropriate neuropsychological tests, and use of more rigorous bedside mental status testing should substantially enhance the usefulness and sensitivity of CS-CPB in evaluating NPAs.

(6) Treatment with GM₁ ganglioside (two 300-mg doses IV, preoperatively) was associated with a non–statistically significant benefit compared with Control in Neurologist's mean Cerebral CCSs at the Acute Postop assessment and in Neuropsychologist's mean CCSs at Early F/U. Sample size estimation indicates that 150 patients per treatment arm would be needed (with =0.05 and ß=0.20) to demonstrate a statistically significant effect on the Acute Postop Cerebral CCS. Improved test methods and administration of higher doses could reduce sample size requirements. Further study of GM₁ ganglioside is needed to better elucidate its potential therapeutic benefits for this indication.

	Selected Abbreviations and Acronyms

 

Acute Postop = 1 day after surgery CABG = coronary artery bypass graft CCS(s) = Clinical Change Score(s) CNS = central nervous system CS-CPB = cardiac surgery under cardiopulmonary bypass Early F/U = follow-up approximately 1 week after surgery Long-term F/U = follow-up approximately 6 to 21 months after surgery MNTD = minimum number of threshold declines MNTI = minimum number of threshold increases NPA(s) = neuroprotective agent(s)

	Acknowledgments

 
Study medication and financial support were provided by Fidia Pharmaceutical Corporation, Washington, DC. We are particularly indebted to Katherine M. Marschall, MD, and Susan Roggen Tanke, RN, of the Department of Anesthesiology at NYU Medical Center for their invaluable assistance in capturing data relating to the perioperative period. We are indebted to Linda Chin, RN, for her tireless review of hospital charts and completion of reporting forms, and to Howard Thaler, PhD, of the Department of Biostatistics of the Memorial-Sloan Kettering Hospital, New York, NY, for analyses of demographic and surgical data, mean changes in test scores, and "Up," "Down," and "Net" scores.

	Footnotes

 
Reprint requests to Giacinto Grieco, MD, Department of Neurology, New York University Medical Center, NB 7W 11, 550 First Ave, New York, NY 10016.

¹ This assumption is for reasons of simplification. In reality, the results are not completely independent, and adjustment factors for interdependency would have to be introduced in the expression.

	Statistical Appendix

Top Abstract Introduction Subjects and Methods Results Discussion Statistical Appendix References

 

View larger version (35K): [in this window] [in a new window]   Figure 5. Patients with neuropsychological test score changes of 1 SD at Early F/U (All patients: GM1+Control).

View larger version (33K): [in this window] [in a new window]   Figure 6. Patients with neuropsychological test score changes of 1 SD at Long-term F/U (All patients: GM1+Control).

View larger version (24K): [in this window] [in a new window]   Figure 7. Neuropsychological test score changes from Baseline of 1 SD. Shown are number of scores worsened (down) by 1 SD, number improved (up) by 1 SD, and Net scores, per patient, for all patients (GM1+control).

View larger version (22K): [in this window] [in a new window]   Figure 8. Distribution of neuropsychological test score changes from Baseline. Shown are percentage of scores worsened (down) by 1 SD versus changed <1 SD vs improved (up) by 1 SD for all patients (GM1+Control).

View this table: [in this window] [in a new window]   Table 8. Distributions of Changes from Baseline in Neuropsychological Test Scores at Early Postoperative Follow-up

View this table: [in this window] [in a new window]   Table 9. Distributions of Changes from Baseline in Neuropsychological Test Scores at Long-term Follow-up

View this table: [in this window] [in a new window]   Table 10. Neuropsychological Test Score Changes of 1 SD

View this table: [in this window] [in a new window]   Table 11. Neuropsychological Test Score Distributions

Determination of the Probability of One or More Test Results Being Abnormal Out of a Set of N Test Results
Let N be a positive integer 1.

Let n and k be nonnegative integers N.

Let {T} be a set of N different tests {t₁, t₂, . . . t_i, . . . t_n} in a battery.

Let p_iinf be the probability that the result of test t_i is subnormal (ie, performance on test t_i is, per primum, below a minimally acceptable level or has, on retesting, worsened by a threshold amount or greater compared with baseline).

Let p_isup be the probability that the result of test t_i is above normal (ie, performance on test t_i is, per primum, above normal or has, on retesting, improved by a threshold amount or greater compared with baseline).

Let p_{inot inf} be the probability that the result of test t_i is not subnormal.

Let P_inf t_i|{T} be the probability that for all tests in {T}, only the result of t_i is subnormal.

Let P_inf n/N|{T} be the probability that exactly n of N test results in {T} are subnormal.

Let P_inf n/N|{T} be the probability that at least n of N test results in {T} are subnormal.

Let P_inf n/N|{T} be the probability that at most n of N test results in {T} are subnormal.

Similarly define P_sup t_i|{T}, etc, and P_{not inf} t_i|{T}, etc.

Then:

P_inf k/N|{T}=1-P_inf(k-1)/N|{T}

(1)

Now,

P_inf /N|{T}=P_{not inf}N/N|{T}

Assuming the results of the N tests to be independent of one another,¹

If p_1inf=p_2inf=. . .=p_kinf=. . .=p_Ninf=p_inf, then it can be shown by induction that

(2)

Substituting (2) in (1) yields:

Furthermore, if the distributions of the test scores or the test score changes are sufficiently normal or can be normalized by appropriate transformation(s) [essentially, this holds if the tests do not have substantial floor or ceiling effects(s)],

p_supp_inf (where p_1sup=p_2sup=. . .=p_Nsup=p_sup),

P_supk/N|{T}P_infk/N|{T}, and

(3)

If we define P_sup,inf k, j/N|{T} as the probability of exactly k test results being above normal, and of exactly j being subnormal, then it also follows that

P_sup,infk, j/N|{T}P_sup(k+j)/N|{T}=P_infk+j)/N|{T}, and, trivially,

P_sup,infk,k/N|{T}P_inf2k/N|{T}.

Additional Analyses of Neuropsychological Test Score Changes
Because of the problems inherent in MNTD analyses, we turned to three other methods to assess change in neuropsychological performance, using the following: (1) Net scores, (2) the distributions of threshold-or-greater changes, and (3) Neuropsychologist's CCSs. The Net score approach provides a graded, pseudocontinuous outcome measure for individuals. The distribution of test score changes results in a quasi-continuous measure for populations. Both approaches take increases in scores into account but still suffer from conceptual deficiencies. Primary among these are the assumptions that change in performance on one test is adequately described as a trichotomous outcome (worse, no change, or better), that any change of at least threshold in one test is equivalent to an at-least-threshold change in any other test, and that overall cognitive change is an additive function of changes in individual neuropsychological tests. For the primary analysis of neuropsychological test score changes, we therefore chose Neuropsychologist's CCSs. We here present the results of MNTD, Net score, and score distribution analyses of our data. Although our position is that impairment rate is not an appropriate outcome measure, we shall attempt to estimate the rate of impairment in our study population since it has been adduced in nearly all prior CS-CPB studies.

Tables 8 and 9 present the numbers and percentages of patients showing different minimum numbers ("Index No.") of 1 and 2 SD decreases, increases, and both increases and decreases from Baseline in neuropsychological test scores at, respectively, Early and Long-term F/U. The figures from the "Scores Decreased by. . ." columns give the results of all possible MNTD analyses for both 1 and 2 SD thresholds. Correspondingly, the figures from the "Scores Increased by. . ." columns give the results for what one could term MNTI analyses. The percentages for 1, 2, and 3 changes of 1 SD for all patients are graphically displayed in Figs 5 and 6. We see, for example, that for all patients at Early F/U, MNTD results for a threshold of 1 SD and for 1, 2, and 3 declines yield apparent impairment rates of 93%, 86%, and 59%, respectively. However, the corresponding percentages for score increases yield apparent improvement rates of 86%, 48%, and 21%, while the percentages for simultaneous increases and decreases (79%, 41%, and 10%) are not much lower than these latter. (As expected, the apparent rates of impairment and improvement, by MNTD and MNTI analyses, are less than the predicted rates in Table 7.) We conclude from these data and from the results of CCS analyses that MNTD analysis does not yield a meaningful estimate of the true frequency of impairment. One way to estimate the proportion of patients suffering neuropsychological decline would be to calculate the differences between corresponding "scores decreased by. . . " and "scores increased by. . . " rates. For index numbers of 1 through 4, the percentages for all study subjects are 6.9, 37.9, 37.9, and 31.0, respectively. Interestingly, from the Early F/U cerebral CCSs (Table 4), one would estimate that approximately one third of patients, net (percentages "worse" minus percentages "better"), irrespective of treatment group, suffered cognitive decline and/or other CNS impairment detectable by bedside evaluation, while the estimate from the Early F/U Neuropsychologist's CCSs would be approximately one half of patients. On the other hand, when we turn to the Long-term F/U results, we see that the percentages for 1, 2, and 3 score increases (84%, 60%, and 40%) of 1 SD now equal or exceed the corresponding percentages for declines (80%, 48%, and 40%), implying no overall impairment. This shift to a small excess of increases is consistent with expectations (see item 5 under "Assessing Cognitive Change After CS-CPB by Neuropsychological Testing" in "Discussion") for retesting a population that has not undergone overall change in cognitive functioning, as the Neurologist's CCSs-Cerebral (percentages "worse" equal percentages "better") and the Neuropsychologist's CCSs (percentages "better" equal to or slightly greater than percentages "worse") in Table 4 suggest. In two large studies using the same MNTD criteria (minimum of two declines of 10 scores, with a 1 SD threshold), impairment rates of 77%¹³ and 73%⁴⁵ at approximately 1 week postoperative were reported. From Table 7 we would estimate that somewhat <48% of patients would have been inappropriately classified impaired as an artifact of the MNTD method. Thus, it is more likely that the "true" rate of impairment in those studies was somewhat >30%, in keeping with the estimates from our study.

We now turn to the results of "Down," "Up," and "Net" score analyses, summarized in Table 10 (a "Down" score having worsened by 1 SD from Baseline, and an "Up" score having improved by 1 SD; the "Net" score equals the number of "Up" scores minus the number of "Down" scores a patient had). The results for all patie