Letters

Reduction of postoperative mortality and morbidity

Little information was given on inclusion criteria

Research into modern anaesthesia techniques and perioperative medicine is needed

Authors' reply

Little information was given on inclusion criteria

EDITORRodgers et al report a meta-analysis of 141 trials comparing general anaesthesia with neuraxial blocks.¹ They conclude that their data should result in more widespread use of spinal or epidural anaesthesia. The challenge for clinicians is deciding which of their patients (if any) these results apply to, but Rodgers et al provided little information about the inclusion criteria for the trials examined. The applicability of a meta-analysis is difficult to assess when heterogeneous patient groups are combined. Also, although a spinal or epidural anaesthetic might be reasonably standard, there are many general anaesthetic agents and these may not be comparable.

Figure 1 of the meta-analysis shows that four trials contributed 31 deaths to the overall mortality difference of 41.^2-5 In those four studies the mortality from general anaesthesia ranged from 8% to 27%, compared with 3.1% for all trials combined.

We wonder whether any information was collated on antithrombosis prophylaxis. In 2001 most patients at risk of venous thrombosis having a general anaesthetic will receive prophylaxis including anticoagulation drugs at low doses. Of the four trials referred to above, three are over 15 years old, and at least one specifically excluded patients receiving low dose anticoagulation.³

Figure 2 of the meta-analysis indicates an apparent benefit of neuraxial block in orthopaedic patients.¹ Mortality after vascular, urological, and general surgery showed no significant difference. We therefore question the conclusion of Rodgers et al that their result is applicable to all surgical patients. In the vascular group (the only non-orthopaedic group that approached significance) there was a difference of eight deaths. That difference would be reduced to just one death by eliminating a single trial in which the mortality after general anaesthesia was 18%.⁵ We would be interested in the views of Rodgers et al on the relevance of that trial to institutions where mortality may be much lower.

It could be a mistake to apply the conclusions of a meta-analysis to any particular patient if that patient's characteristics are different to those in the initial trials or if aspects of their management differ. From the data presented, we could not come to the same conclusions as the authors.

Timothy J McCulloch, staff specialist anaesthetist. 
John A Loadsman, staff specialist anaesthetist. 
Department of Anaesthetics, Royal Prince Alfred Hospital, Camperdown, New South Wales 2050, Australia

Research into modern anaesthesia techniques and perioperative medicine is needed

EDITORRodgers et al in their meta-analysis to resolve one of the more contentious issues in anaesthesia over recent yearsnamely, whether there is there any advantage of regional over general anaesthesia.¹ From the results they conclude that regional techniques (spinal or epidural anaesthesia) reduce postoperative mortality and decrease the incidence of other serious complications such as pneumonia and pulmonary embolism (although the effects of regional anaesthesia on myocardial infarction and renal failure were inconclusive).

Rodgers et al suggest that these data support the more widespread use of regional anaesthetic techniques. The data have been meticulously gathered and researched from a large number of trials but we would like to point out a few areas of concern.

• Only 56 trials (40%) looked at outcome. Thirteen of these (23%) followed patients up for a period lasting more than 30 days, while 19 (34%) followed patients up for less than seven days
• One hundred and sixteen (82%) of the 141 papers that met the inclusion criteria were published before 1990. This means that data from these studies are now at least 10-12 years old
• Anaesthetic techniques, equipment, and drugs have changed quite dramatically in recent years, and, therefore, studies predating these advances may have lost some of their relevance
• Use of some of the older volatile agents has declined, and newer agents with fewer cardiovascular side effects are now in widespread use
• Stopping cardiac drug treatment preoperatively is no longer recommended
• Heparin prophylaxis for venous thrombosis and pulmonary embolism is now much more common,² particularly since the introduction of low molecular weight heparins
• Ten trials looked at outcome in vascular surgical patients (a group known to be at increased risk of perioperative complications), eight of which were published after 1990, and there were more myocardial infarctions and cardiac deaths in the patients who had received a regional anaesthetic.

We broadly agree with the conclusion of Rodgers et al. Morbidity and mortality seem to be reduced in patients receiving regional anaesthesia with a trend towards reduced cardiac and renal complications. However, the paucity of recent data in which hard outcome measures are assessed indicates the need for more research directed towards understanding the impact of modern anaesthesia techniques and perioperative medicine on patient outcomes.

Helen Higham, research fellow. 
helen.higham{at}nda.ox.ac.uk

Pitabas Mishra, specialist registrar in anaesthetics. 
P Foëx, professor of anaesthetics. 
Nuffield Department of Anaesthetics, John Radcliffe Hospital, Headington, Oxford OX3 9DU

Authors' reply

EDITORMcCulloch and Loadsman raise issues about applicability of meta-analyses of heterogeneous patient populations, implying that generalisation should be restricted to patients closely similar to those in the included trials. Because of the large number of trials, only limited data on eligibility criteria could be published, even in the web site version. The key issue, however, aiding generalisibility is not "representativeness," but consistency between different trials, especially if observed across heterogeneous patient groups.¹ The proportional effects of neuraxial blockade were broadly consistent, justifying the pooling process. Absolute risk reductions therefore increased with increasing event rates. For example, neuraxial blockade reduced deep vein thrombosis by 44% (SE 11) in trials that employed screening and 46% (22) in other trials, and so absolute differences were greater in screening versus other trials (12% v 0.5%, respectively). Therefore, trials with most events contributed most of the net difference in events. However, treatment effects were not clearly restricted to such trialsfor example, in trials that observed more than or less than 10 deaths, mortality reductions were 33% (13) and 27% (20) respectively.

Most patient groups have lower mortality than those in the meta-analysis. But there were clear reductions, overall and in several different surgical groups, for other important outcomes such as thromboembolism, pneumonia, and bleeding. For these outcomes, it seems appropriate to require very good direct evidence of lack of benefit before safely concluding that neuraxial blockade is not effective in some particular group.³ We did not observe such evidence.

Applying trial results to individual patients should ideally entail combining a typical proportional reduction from a meta-analysis with a patient's estimated absolute risk.² An updated meta-analysis with data from individual participants could improve estimates of proportional reductions and more reliably identify any subgroup effects. A key challenge for the clinician remains estimating a patient's absolute risk, however, since this is likely to vary substantially more than any subgroup differences in proportional effects of neuraxial blockade.

Higham et al point out correctly that most trials did not report clinical outcomes. But we aimed to collect data from all trials, irrespective of their original aims. Length of follow up inevitably varies between trials, and we separately analysed events within 30 days (usually within 20) and deaths after 30 days.

Both letters point out that thromboprophylactic practices have changed in recent decades, but we excluded trials in which they were systematically different between the randomised groups. Concomitant anaesthestic practice has also changed, but the overall reduction in mortality was 28% (13) for trials published before 1990 and 35% (21) for trials published in the 1990s.

Anthony Rodgers, codirector. 
Natalie Walker, research fellow. 
Clinical Trials Research Unit, Department of Medicine

Stephan Schug, professor. 
Division of Anaethesiology University of Auckland, Private Bag 92019, Auckland 1, New Zealand

Andrew McKee, consultant anaesthetist. 
Department of Anaesthetics, Green Lane Hospital, Epsom, Auckland 1003

Andre van Zundert, consultant anaesthetist. 
Department of Anesthesiology, Intensive Care and Pain Therapy, Catharina Hospital, NL-5623 EJ Eindhoven, Netherlands