Secondary Logo

Journal Logo

Correspondence

Epidural anaesthesia for pain relief in labour

Zuokumor, P.; Columb, M.

Author Information
European Journal of Anaesthesiology: August 2003 - Volume 20 - Issue 8 - p 674-675

EDITOR:

We read with interest the paper by Aşik and colleagues who evaluated the effects of epidural bupivacaine 0.2% compared with ropivacaine 0.2% combined with fentanyl for analgesia during labour [1]. The authors enrolled 60 patients with 28 and 25 patients in bupivacaine-fentanyl (BF) and ropivacaine-fentanyl (RF) groups, respectively, completing the study. In their results, they concluded that there was a significantly less incidence (7 versus 15) of instrumental delivery in the RF group (P < 0.05). Using Fisher's exact test, a two-sided P = 0.094 should be returned. Surprisingly, the authors declared that a Fisher's exact test was employed. This is obviously misleading even if the one-sided P is considered (= 0.053). Estimation of the 95% confidence interval (CI) for the difference in proportions of −0.01 to 0.52, although including zero, suggests an effect. This is further supported by the use of a two-sided mid-P = 0.068, which is close to significance [2]. A more appropriate approach is to combine the adverse outcomes (Table 1) and consider the frequencies of normal deliveries. This returns a Fisher's exact two-sided P = 0.17 with a 95% CI difference for the proportions of normal deliveries of −0.06 to 0.48. This issue has already been addressed [3]. The most appropriate analysis is to analyse the two groups by the three outcomes using the Fisher-Freeman-Halton exact test, which returns a two-sided P = 0.18. An intention to treat analysis, which would have included three more Caesarean deliveries, would have militated further against finding significance. Therefore, it appears that the authors have applied undue weight to their conclusions about obstetric outcomes by attempting to cherry pick results.

Table 1
Table 1:
Data abstracted from Aşik (2002), Tables 2 and 3[1].

In Tables 1 and 2 of their paper, the authors also stated that there were no significant differences between the groups. This may not be entirely correct. Crude reanalysis of the summary statistics in Table 2 suggests that the duration of pain relief after the loading dose of local anaesthetic was significantly longer in the BF group (P = 0.006, 95% CI difference 4.3-23.9 min). The total volume infused to delivery was significantly larger for the BF group (P = 0.024, 95% CI difference 1.7-23.7 mL). It then follows that the total doses of both local anaesthetic and fentanyl were significantly larger in the BF group (P = 0.024, 95% CI difference 3.4-47.4 mg or μg, respectively). The frequency of boluses was also significantly higher in the BF group (P = 0.0002, 95% CI difference two to six boluses). Note too that the subjects in the BF group were more advanced in labour (wider cervical dilatation). As this is a pretest variable, then randomization implies that any difference must be due to chance and a significance test is inappropriate. However, it is possible to estimate a sampling probability of 0.003 for this. The clinical relevance of this finding is, however, not very clear, as the durations of labour were similar. However, some discussion is warranted given the finding by Capogna and colleagues of the association of bupivacaine requirements with cervical dilatation [4]. The authors were correct in stating that there were a significant number of patients with motor block in the BF group. However, this is hardly surprising, as the BF group had, as mentioned above, a significantly higher total dose of local anaesthetic and fentanyl, in addition to the potential potency issues correctly considered by the authors.

In conclusion, it would appear that there are more questions than answers. Certainly, more attention is warranted to the statistical analyses before any conclusions can be drawn. If the authors have used one-sided P, then this should be highlighted in the study design and justified with the direction of the expected outcome explained. By convention, reported P are two-sided and the use of one-sided values should not only be highlighted, but also justified as they are rarely ever appropriate [2,3,5]. Failure to do so may lead the reader to question the potential for bias in analyses and conclusions.

P. Zuokumor

M. Columb

Department of Anaesthesia and Intensive Care Medicine; South Manchester University Hospital Trust; Manchester, UK

References

1. Aşik I, Göktug I, Gülay N, Alkiş N, Uysalel A. Comparison of bupivacaine 0.2% and ropivacaine 0.2% combined with fentanyl for epidural analgesia during labour. Eur J Anaesthesiol 2002; 19: 263-270.
2. Columb MO. One-sided P values, repeated measures ANOVA; two sides of the same story! Anesth Analg 1997; 84: 701-702.
3. Columb MO, Polley LS. Potencies and probabilities: one-sided P values suggest a one-sided story. Anesth Analg 2001; 92: 278-279.
4. Capogna G, Celleno D, Lyons G, Columb M, Fusco P. Minimum local analgesic concentration of bupivacaine increases with progression of labour. Br J Anaesth 1998; 80: 11-13.
5. Altman DG. Practical Statistics for Medical Research. London, UK: Chapman & Hall, 1991: 170-171.
© 2003 European Academy of Anaesthesiology