METHODS

The patients and methods have been described in detail elsewhere.^2,3 Briefly, the Oxford research ethics committee approved the trial. Patients gave written consent to participation. Inclusion criteria were patients undergoing non-emergency first time CABG with no previous cerebrovascular accident and an ejection fraction > 30%.

Lexipafant (British Biotech, Oxford, UK) is a highly potent and specific antagonist of PAF and in vitro binds to the PAF receptor with an affinity seven times that of PAF itself. It has an intravenous half life of about one hour and is well tolerated without serious adverse reactions.

Randomisation by minimisation was by a double blind process so that the treatment groups would be balanced with respect to age, sex, number of diseased vessels, previous percutaneous coronary intervention, surgeon, aspirin, and left ventricular function.

After loading doses of 0.4 mg (low dose) or 4 mg (high dose), lexipafant was infused at rates of 0.4 mg/hour and 4 mg/hour over 24 hours, achieving plasma concentrations of roughly 10 ng/ml and 100 ng/ml, respectively.

Anaesthesia was standardised with morphine, scopolamine, fentanyl, pancuronium, etomidate, nitrous oxide, halothane, and propofol. Benzodiazepines were not used.

CABG was performed through a median sternotomy using non-pulsatile CPB and a temperature of 34°C. Topical cooling, ventricular venting, and arterial filters were not used. A membrane oxygenator and alpha stat control of acid base management were used and the mean arterial pressure maintained at 50–60 mm Hg with pharmacological manipulation if necessary. Aortic clamping and induced fibrillation or cold crystalloid cardioplegia (15 patients) were used for distal anastomoses and a side biting clamp for proximal anastomoses.

Power calculations based on data from the mid-1980s suggested that three groups of 50 patients would be sufficient to obtain a 20% reduction in neuropsychological impairment with 80% power at the 1% level of significance at three months.

Cognitive assessment

Patients underwent a standard battery²⁴ of 10 cognitive tests performed by one examiner (SMB) before surgery and at five days and three months after CABG. Only four tests (auditory verbal learning test (AVLT), adult memory and information processing battery (AMIBP), trail-making test B (TMTB), and verbal fluency test (VFT)) were normally distributed and these were used to calculate a composite cognitive index (see below).

In the absence of agreed consensus as to the optimal method for defining impairment^1,2 three methods of analysis were used:

• Group mean: performance in each test was analysed with repeated measures analysis, using a within (time) and a between (lexipafant group) subject factor, with three levels.

• 20% definition: test deterioration was a > 20% fall in the preoperative score. The number of tests with > 20% deterioration was summed from zero to four.

• Composite cognitive index: the composite cognitive index was an estimate of overall performance, calculated from the mean of each patient’s four test z scores (raw scores converted from mean = 0, SD = 1) at each time point.

Statistical analysis

S-PLUS2000 (MathSoft Inc, Seattle, Washington, USA) was used for statistical analysis. Differences between the groups were analysed using analysis of variance for continuous, Kruskal-Wallis for ordinal, and χ² test for nominal data. Within subject changes (from preoperation to discharge and to three month scores) were analysed using paired t tests and between groups differences using Kruskal-Wallis. To account for multiple comparisons a significance was determined as p < 0.005 (that is, 0.05/8 = 0.006, where 8 represents four tests at two time points).

To examine specifically the effects of drug dose, linear mixed effects models²⁵ were fitted to each of the four psychometric tests using fixed covariates of lexipafant dose, test time, age, and duration of CPB. A quadratic effect was applied to test time to account for the early deterioration and subsequent improvement.

RESULTS

Overall completeness of data

The study was completed by 140 of the 150 patients enrolled. Of the 10 patients who did not complete the study, six died (three within one week of surgery and three during follow up) and the remaining four patients failed to attend at three months.

Demographics

Table 1 shows that the groups were similar regarding preoperative and intraoperative factors.

Cognitive results

Of the 150 trial patients 135 underwent all three assessments (although not necessarily completing the full battery). Reasons for patients not being assessed at both postoperative time points were death (six), stroke (two), refusal (four), medical instability (two), and irregular hospital discharge (one).

Group mean analysis

All tests were impaired at discharge (table 2). By three months all had improved except for AVLT. The most important predictor of the change in scores was the time of testing.

20% definition

The three treatment groups were marginally different at discharge (Kruskal-Wallis one way analysis of variance χ² 6.20, p = 0.05), although not at the three month assessment (Kruskal-Wallis one way analysis of variance χ² 0.11, p = 0.9) (tables 3 and 4).

Cognitive composite index

A repeated measures analysis with one within group factor (time of assessment) and one between group factor (lexipafant group), each with three levels, was used (table 5). The time factor again was significantly different (F = 69.8, p = 0.000). The lexipafant group factor was not significantly different (F = 1.45, p = 0.238). The interaction between the two factors was also insignificant (F = 1.12, p = 0.349). This indicates that the pattern of change for the three treatment groups was not significantly different.

Predictors of changes in cognitive function

There were no differences in baseline scores for any test between the three groups (table 6). The linear mixed effects model showed significant effects for timing of the test and age in predicting worse cognitive performance, but no significant effect of lexipafant or duration of CPB. Importantly, there was no detectable adverse reaction attributable to lexipafant.

DISCUSSION

Cognitive dysfunction after CABG is common and the rationale for the use of a PAF antagonist to reduce it is compelling. CABG using CPB induces high concentrations of PAF,^21,22 which causes cerebral dysfunction in the clinical setting^17–20 and which is ameliorated by PAF antagonists in experimental brain dysfunction.^13–16

A major and unique strength of this trial was the use of three commonly used definitions of cognitive impairment. Our trial confirmed that age and timing of the test were the strongest predictors of cognitive impairment.¹ The failure to identify duration of CPB as a predictor of decline is probably due to the relatively narrow spectrum of our CPB times.

It is unlikely that the dose of lexipafant was insufficient to achieve treatment effects. As a plasma concentration of lexipafant of 2 ng/ml blocks exogenous effects of PAF, a dose of 100 mg/24 hour, providing plasma concentrations between 50 and 150 ng/ml, should block exogenous PAF release in direct cell to cell interactions.

In retrospect, however, our trial may have been underpowered to detect a protective effect of lexipafant, as the incidence of cognitive impairment was less than had been expected. At three months only one test (delayed recall) had not returned to its baseline score. This is consistent with the findings of others⁵ and may also explain, in part, our failure to detect a difference in cognitive impairment in patients undergoing CABG with or without CPB.² If the overall improvement in surgical outcome is the result of ongoing refinements to the conduct of CPB, then the negative findings of our study (which used studies from over a decade ago to calculate power) provide a valuable service to future studies. On the basis of the extent of cognitive impairment we detected postoperatively in this trial, 500 patients would have been needed to show efficacy of lexipafant.

While several drugs have been mooted as potential neuroprotective agents against cognitive dysfunction, there is only one other current large trial of a putative neuroprotective drug in CABG.⁵ In a randomised trial of 171 patients Arrowsmith and colleagues reported that remacemide, an N-methyl-d-aspartate receptor antagonist, did not significantly reduce cognitive dysfunction but did result in greater preservation of learning.⁵

Even accepting that the incidence of cognitive impairment is less than that observed a decade ago¹ we do not advocate a complacent attitude. As cognitive dysfunction is an age related phenomenon it will remain an important consideration as a progressively elderly population with more comorbidity undergoes CABG.¹ This is likely to be of even greater relevance in open operations where the risk of cerebral dysfunction is greater.²⁶

Conclusions

Despite persuasive experimental evidence for the neuroprotective effects of PAF antagonists, we failed to detect such an effect in a large clinical trial. It is possible that our trial was, in retrospect, underpowered given the apparent reduction in overall cognitive impairment resulting from CABG that has occurred over the past decade. As older patients with greater comorbidity are referred for CABG, however, the clinical need for innovative cerebral protective strategies remain compelling.