Is it possible to decrease the incidence of peri‐operative stroke associated with acute peri‐operative beta‐blocker administration?

Acute peri-operative beta-blockade significantly decreases 30-day non-fatal myocardial infarction (from about 5% to 3%, with a relative risk reduction of 35% in randomised controlled trials) [1], and initiation of peri-operative beta-blocker protocols are associated with improved postoperative survival following non-cardiac surgery (OR 0.52 (95% CI 0.33–0.83) for 30-day mortality, and OR 0.64 (95% CI 0.51–0.79) for 1-year mortality) [2]. However, these benefits are at the expense of an increased incidence of non-fatal stroke (from 0.31% to 0.67%, a relative risk increase of 116% in randomised controlled trials) [1]. Although the importance of stroke associated with beta-blockade only assumed prominence following the PeriOperative ISchemic Evaluation (POISE) trial [3], it has been a consistent signal in all the prospective randomised trials of peri-operative beta-blockers that have reported stroke as an outcome [1]. To realise the full cardioprotective utility of acute peri-operative beta-blockade, the risk of stroke needs to be decreased. A number of factors are associated with peri-operative stroke: a pre-operative history of stroke or transient ischaemic accident; prior use of clopidogrel or ticlopidine; significant peri-operative bleeding; atrial fibrillation; clinically significant hypotension; renal disease; valvular heart disease; and urgent surgery following hip fracture [3-5]. Each of these may involve a potential pathway through which peri-operative stroke is precipitated when beta-blockers are administered. While simply avoiding beta-blockers in patients with these risk factors may decrease the risk of peri-operative stroke, this may exclude a significant number of patients likely to benefit from the cardioprotection conferred by beta-blockade. In addition, excluding patients with risk factors will not necessarily guarantee safety from stroke in patients receiving beta-blockers. This is because hypotension is independently associated with peri-operative stroke [3], and the magnitude of hypotension is greater following acute peri-operative beta-blockade (OR 1.62 (95% CI 1.44–1.82)) with a number-needed-to-harm of 17 [1]. When beta-blockers are given as premedication or at induction of anaesthesia, blood pressure is generally better maintained than when therapy is continued [6-9]. Perhaps the magnitude of hypotension associated with beta-blockade could be reduced through more careful monitoring of blood pressure and aggressive management of any hypotension. The effect of such an intervention can be estimated by making assumptions about (i) the reduction in excess strokes during and after the intervention, and (ii) the period over which the intervention is applied. For example, if an intervention were 100% successful, then the predicted stroke rate would be expected to fall from 0.67% (as reported with beta-blockers) to the background stroke rate as reported in the placebo or control arms of various trials (i.e. ∼0.31%, or 3 per 1000 patients) [1]. However, peri-operative strokes are not evenly distributed over the postoperative period. About one third of strokes within 30 days of surgery occur 24–48 h following surgery, with the remaining two-thirds occurring up to 30 days postoperatively (Table 1) [10-16]. If preventing beta-blocker associated hypotension for the first 24–48 h postoperatively were to reduce the stroke rate during this intervention period to the placebo rate (0.31%), and further results in a halving of the excess stroke rate up to 30 days postoperatively (i.e. from 0.67% to 0.49%, which is halfway to 0.31% [1]), then it is possible to estimate that 1 in every 420 subjects will no longer experience a stroke. This is the best-case scenario; the true impact of preventing beta-blocker associated hypotension is likely to be far less favourable, as it is possible that only 15% (95% CI 5–35%) of the additional stroke risk associated with beta-blockers is modifiable with management of hypotension [3]. If, in fact, only 15% of the excess strokes are prevented during the first 24–48 h, and if there is no effect of the intervention on the stroke rate over the remaining 30-day postoperative period, any hypotension-prevention protocol confined to just the first 24–48 h would make almost no difference to the predicted peri-operative stroke rate (Table 2). These simple considerations lead to the important conclusion that targeting hypotension alone in the peri-operative period is unlikely to realise significant clinical benefit associated with postoperative stroke. A successful intervention plan should therefore include multiple potential therapies (of which one could be a hypotensive intervention plan), and should be continued over a period of time exceeding 48 h because, as noted above, more than half of the strokes occur more than 48 h after surgery, up to the 30th postoperative day (Table 1). Specific interventions that may be useful include better management of pre-operative antiplatelet agents, prevention and treatment of atrial fibrillation, and the use of statins (which are known to decrease stroke following non-cardiac surgery in those at risk) [17]. To maximise the success of any interventions, the time periods associated with the highest proportion of peri-operative strokes should be targeted. Our current data are very limited and we need more accurate information on exactly when strokes occur following non-cardiac surgery (Table 1). Secondly, we need accurate data on both the control and treatment incidences of peri-operative stroke. Certain patient groups or those undergoing specific surgical procedures may be at increased risk of stroke compared with others, since studies report higher stroke rates following total hip arthroplasty [5], hemicolectomy and pulmonary resection [4] than are reported in other operations [1]. Possibly, the best data to address these needs will come from the Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Study [18]. Whichever approach is taken to decrease peri-operative stroke, it is imperative that we measure the success of the intervention. Considering the very extreme case of an intervention that prevents all excess strokes over the 30-day period (i.e. reducing the stroke rate from 0.67% to 0.31% [1]), in a controlled clinical trial, a test for a reduction in the stroke rate requires 5240 individuals per group to achieve a statistical power of 80% at a 5% significance level (Matlab v7.5.0.342, R2007b, The MathWorks Inc., 2007, Natick, MA, USA) [19]. In reality, the intervention would be applied over a much shorter time period postoperatively and the reduction in stroke rates is expected to be smaller, requiring even larger samples sizes (Fig. 1). These prohibitive sample sizes suggest that it might be preferable to measure the success of any intervention on a surrogate outcome known to be associated with stroke, such as hypotension. For example, to achieve the same power in a test for a reduction in hypotension, only 560 subjects per group are required for an expected reduction in incidence of hypotension from 17% to 11.6% [1], and 2310 subjects if the reduction in incidence was from 17% to ∼14.5% (i.e. halving the excess hypotension associated with beta-blocker). If the proposed intervention had a significant effect on a surrogate outcome such as blood pressure, then this would justify a larger trial assessing its effect on the primary outcome of stroke. Required sample size (in thousands) for a controlled clinical trial to test for a reduction in the incidence of stroke, to achieve a power of 80% at a significance level of 5%. The x-axis shows the predicted stroke event rate in the control group (R2), while the y-axis shows the absolute reduction in the stroke event rate in the intervention group (R1 is the event rate in the treatment/intervention group). The contours indicate the required sample size (in thousands). Successfully demonstrating a reduction in peri-operative stroke associated with beta-blockade is a challenge, but the recent publication by Wallace et al. [2] suggests that there are institutions with the infrastructure necessary to tackle this problem. The San Francisco Veterans Administration Medical Center has initiated a peri-operative beta-blocker protocol that has been adopted by others. Collaboration and reporting of cardiovascular and stroke outcomes by institutions such as these, following common adaptations to the protocol to decrease peri-operative beta-blocker associated stroke, may assist in quantifying and documenting the success of these interventions. Research into improving cardiovascular outcomes after surgery is well justified, and beta-blockers have a positive impact. Yet these drugs have adverse impact on equally important cerebrovascular outcomes. Cardiovascular and cerebrovascular outcomes have hitherto been considered as separate entities in previous trials. It is now necessary to consider the two together and design trials that seek the optimum balance between the two. No external funding and no competing interests declared.