Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Which anaesthetic technique?

General anaesthesia involves using gases and intravenous drugs to send us to sleep and keep us asleep. Injection of local anaesthetic drugs in or around the spine constitutes neuraxial blockade and may confer additional benefits. Is there any difference between them in terms of harmful outcomes? A meta-analysis suggests there may be, but it is also an object lesson in caution and sensitivity analysis. It is also important because it helps us think about rare but serious adverse events ( Bandolier 85 ).


The review was exemplary in the way it searched for papers, found additional information, contacted authors, and extracted data. The aim was to find all trials where patients were randomised to neuraxial blockade or not. Patients receiving neuraxial blockade could also have general anaesthesia. Considerable effort went into extracting all useful data, but here we concentrate on mortality within 30 days of randomisation.


There were 141 trials with 9,559 patients. There were 247 deaths within 30 days, recorded in 35 trials. There were nine trials with at least 10 deaths per trial, and these are shown in the Figure as a L'Abbé plot. For only three of these smaller trials was there a large effect of neuraxial blockade, and in these three there was an extraordinarily high death rate with control of over 15%. For six other trials in which the death rate with control was below 15%, the death rates with neuraxial blockade and control were about the same.

Figure: L'Abbé plot of nine trials with at least 10 deaths

This could be interpreted as some form of heterogeneity, and the L'Abbé plot was first suggested as an aid to detecting whether or not all trials in a meta-analysis were giving the same sort of result. Clearly here they were not.

So some sensitivity analysis would seem in order. Much sensitivity analysis according to methodological issues had been done in the original paper, but not one according to event rates. The Table shows the results obtained for all 141 trials, and for those with more or fewer than 10 deaths. Clearly the latter show the largest treatment effect.

Table: Mortality with neuraxial blockade and control in a meta-analysis and sensitivity analysis

      Deaths/Total (%)    
      Neuraxial blockade    
Condition Trials Patients (% total) Present Absent Relative risk (95% CI) NNT (95% CI)
All trials 141 9559 (100) 103/4871 (2.1) 144/4688 (3.1) 0.7 (0.5 to 0.9) 98 (60 to 265)
Trials with fewer than 10 deaths 132 7067 (74) 32/3537 (0.9) 44/3530 (1.2) 0.7 (0.5 to 1.2) n/c
Trials with more than 10 deaths 9 2492 (26) 71/1334 (5.3) 100/1158 (8.6) 0.6 (0.5 to 0.8) 30 (19 to 77)
More than 10 deaths, more than 100 patients (death rate in control < 10%) 4 1889 (20) 49/1054 (4.6) 48/835 (5.7) 0.9 (0.6 to 1.4) n/c
More than 10 deaths, fewer than 100 patients (death rate in control > 10%) 5 603 (6) 22/280 (7.9) 52/323 (16.1) 0.5 (0.3 to 0.8) 12 (7.5 to 32)
All trials with a death rate with control of less than 10% 136 8956 (94) 81/4591 (1.8) 92/4365 (2.1) 0.8 (0.6 to 1.1) n/c
n/c = NNT not calculated because no significant difference on relative risk

In those trials with more than 10 deaths, four with more than 100 patients per group have a death rate of below 10% and show no statistically significant effect of neuraxial blockade. Five trials had fewer than 100 patients per group and a death rate with control of over 10%, and show a very large effect.

If we exclude all trials with death rates of over 10% from the combined analysis, then the effect of neuraxial blockade is very small, perhaps reducing the death rate from 2.1% to 1.8%, a reduction that is not statistically significant, but in 136 trials with 94% of the patients.


When interpreting results of clinical trials or meta-analyses to our patients, we are told to ask whether the patients in the trial or analysis are like our patients, in terms of demography or severity of disease. Here we find that an apparent overall beneficial effect of neuraxial blockade derives from a few small trials with exceptionally high death rates, three of which were published in the early 1980s. When we exclude these, no statistical effect remains.

Few surgical procedures have death rates above 10%, and it is pertinent to ask whether these results apply to our patients. The authors of the meta-analysis are understandably cautious over any claims, and the aim here is not to argue against the use of neuraxial blockade in anaesthesia. Rather it is to use this as an example of justified caution of where an overall conclusion can be influenced by a few small and unrepresentative trials.


  1. A Rodgers et al. Reduction in postoperative mortality and morbidity with epidural or spinal anaesthesia: results from overview of randomised trials. BMJ 2000 321: 1-12.
previous story in this issue