This Article Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Columb, M. O. Articles by Campbell, D. C. Search for Related Content PubMed PubMed Citation Articles by Columb, M. O. Articles by Campbell, D. C.

---

LETTERS

Potencies and Probabilities: One-Sided P Values Suggest a One-Sided Story!

Department of Intensive Care Medicine South Manchester University Hospital Trust Manchester, United Kingdom *Department of Obstetric Anesthesiology University of Michigan Medical Center Ann Arbor, MI

We read with interest the Discussion by Campbell et al. (1) in their paper regarding the comparison of bupivacaine and ropivacaine for analgesia in labor. Some corrections are required to basic mathematics that have misrepresented results from previous studies (2,3). The authors stated that the dose required for ropivacaine was 40% larger than that for bupivacaine in the minimum local analgesic concentration studies. A potency ratio (%weight/volume) for bupivacaine:ropivacaine of 0.6 was quoted in those studies. Basic mathematics then imply that the dose of bupivacaine is 40% less than that of ropivacaine and that the dose ropivacaine is 1.67 times that of bupivacaine, or 67% more!

The authors are correct in stating that the full dose-response relationships have not as yet been defined and that there is the possibility that the curves may approximate for bupivacaine and ropivacaine at the upper extremity. This is of pharmacodynamic interest, as the differing slopes would then imply differing "receptor" interactions and mechanisms of action within the homologous pipecoloxylidine series. With up-down sequential allocation, testing is concentrated at 50% effective concentration values (± 1 SD). This coincidentally ranges from the 16th to 84th centiles or the straight part of the cumulative dose-response curve. So the slope is likely to be better estimated than by using predetermined doses where potential errors at the extremes of the distribution exert more weight in regression methods, thus affecting slope estimates (4,5). The heterogeneity in minimum local analgesic concentration estimates that the authors describe has been described previously (6,7) and serves to remind us how difficult it can be to compare studies, which the authors continue to do later in the Discussion! This variability is often missed in fixed recipe study designs that are often at the upper end of the dose-response curve (8).

More disturbing, however, is the misuse of one-sided P values, which unfortunately I (MOC) have had to highlight previously in the Journal (9). The P values (Fisher’s exact) quoted in the paper can only be explained by the use of one-sided tests. The use of one-sided P values in this setting implies that any beneficial effect attributable to bupivacaine would have to be a result of chance! This can imply a potential bias in favor of a particular outcome. One-sided P values may also suggest an exaggeration of the effect size of the measured outcome. Such biases are suggested in the selected use of the contingency table detailing obstetric outcomes ( Table 1) and forceps delivery in particular. Essentially combining the adverse outcomes (as shown simply by analyzing the frequencies of normal deliveries) makes even the one-sided P value not significant! By convention, reported P values are two-sided and the use of one-sided values should not only be highlighted but also justified, as they are rarely appropriate (9,10). In conclusion, the uses of one-sided P values and significant tests must always be described and justified with the direction of expected outcome explained. Failure to do so may lead the reader to question the potential for bias in analyses and conclusions.

1. Campbell DC, Zwack RM, Crone LAL, Yip RW. Ambulatory labor epidural analgesia: bupivacaine versus ropivacaine. Anesth Analg 2000; 90: 1384–9.[Abstract/Free Full Text]
2. Capogna G, Celleno D, Fusco P, et al. Relative potencies of bupivacaine and ropivacaine for labour analgesia. Br J Anaesth 1999; 82: 371–3.[Abstract/Free Full Text]
3. Polley LS, Columb MO, Naughton NN, et al. Relative analgesic potencies of ropivacaine and bupivacaine for epidural analgesia in labor: implications for therapeutic indexes. Anesthesiology 1999; 90: 944–50.[Web of Science][Medline]
4. Wetherill GB, Levitt H. Sequential estimation of points on a psychometric function. Br J Mathematical Statistical Psychology 1965; 18: 1–10.[Web of Science][Medline]
5. Armitage P, Berry G. Statistical methods in medical research. 3rd ed. Oxford: Blackwell Scientific Publications, 1994: 169–71.
6. Columb MO, Lyons G, Polley LS, et al. Minimum local analgesic concentration of bupivacaine in labor: a metanalysis [abstract]. Proceedings of SOAP 1997:125.
7. Columb MO. Dose, concentration and volume: A triangulated circular argument! Int J Obstet Anesth 1998;7:64–5.
8. Reynolds F. Does the left hand know what the right hand is doing? An appraisal of single enantiomer local anaesthetics. Int J Obstet Anesth 1997; 6: 257–69.[Medline]
9. Columb MO. One-sided P values, repeated measures ANOVA; two sides of the same story! Anesth Analg 1997;84:701–2.
10. Altman DG. Practical statistics for medical research. London: Chapman & Hall, 1991: 170–1.

Response

Associate Professor College of Medicine University of Saskatchewan; Director of Obstetric Anesthesiology and Research Department of Anesthesiology Royal University Hospital Saskatoon, Saskatchewan, Canada

In Reply—

I thank Drs. Polley and Columb for their interest in our recent publication in Anesthesia & Analgesia (1) and appreciate the opportunity to respond to their concerns. First of all, Dr. Columb has co-authored several recent publications that describe the determination of the concentration of a local anesthetic necessary to produce epidural analgesia (visual analog scale for pain [VASP] <10/100 by 30 min) in 50% of laboring parturients. This body of work has raised awareness within the Obstetric Anesthesia community of the possibility that local anesthetics such as bupivacaine and ropivacaine may not be equipotent. The 50% effective concentration (EC₅₀) ratio (%w/v) for bupivacaine:ropivacaine of 0.6 was indeed independently determined by two recent studies coauthored by Dr. Columb (2,3). Unfortunately, this research model focuses entirely on only one point on the concentration (dose) response curve (i.e., EC₅₀). Of importance, anesthesiologists providing epidural labor analgesia do not routinely (if ever) utilize such ineffective concentrations in clinical practice. Secondly, the upper end of the concentration (dose) response curves (i.e., EC₉₀, EC₉₅) for these local anesthetics have yet to be determined as acknowledged by Drs. Columb and Polley. Consequently, EC₉₀ or EC₉₅ (%w/v) ratio comparisons are unknown, and therefore any conjecture related to the clinically relevant potencies of the two local anesthetics is premature and potentially misleading! Drs. Columb and Polley’s suggestion that the "up-down sequential allocation" method of determining the EC₅₀ on the "straight part of the cumulative dose-response curve" allows the EC₅₀ to be "better estimated" may be true, but certainly not clinically relevant. To overcome "potential errors at the extremes of distribution" (i.e., EC₉₀ determinations), I suggest appropriate studies be undertaken with enrollment of large numbers of patients rather than the 30 per group used to "estimate" EC₅₀s (2,3). "Basic mathematics" aside, proclamations that bupivacaine and ropivacaine are not equipotent based on EC₅₀ studies have yet to be validated and are extremely premature.

Our investigation compared concentrations of bupivacaine and ropivacaine (0.08%) combined with 2 µg/mL fentanyl that had not been previously described. Both medication combinations proved effective labor analgesia with mean times to VASP 0/100 of approximately 12 min in both groups. Only one study to date (coauthored by Dr. Columb) reported the EC₅₀ of bupivacaine, combined with 2 µg/mL fentanyl, to be 0.048% (4). We acknowledged this result in our discussion to illustrate the inability of the EC₅₀ studies to be extrapolated to the clinical relevant portions of the concentration (dose) response curves for each local anesthetic. As Drs. Columb and Polley suggest "the dose of bupivacaine (required) is 40% less than that of ropivacaine" if the goal is to provide "effective" (VASP <10/100 x 30 minutes) in only 50% of laboring women (2,3). Our study results suggest that the initiation of epidural analgesia with 0.08% ropivacaine plus 2 µg/mL fentanyl approximates at least the EC₉₅ (if not higher) for ropivacaine. Therefore, if the concentration (dose) response curves for the two local anesthetics are truly parallel, then 40% less bupivacaine, or only 0.048% should represent the EC₉₅ for bupivacaine. However, Dr. Columb and co-workers reported that this concentration produced effective epidural labor analgesia in only 50%, not 95%, of laboring women! Consequently, our data does not support such a large concentration (%w/v) ratio of 0.6 for bupivacaine:ropivacaine at concentrations that produce effective labor analgesia. In fact, recent investigations (5,6), including our own (1), have reported similar total requirements of bupivacaine and ropivacaine over the entire duration of labor, rather than only at the time of initiation, using patient-controlled epidural anesthesia. These reports provide preliminary evidence that the EC₉₅s may be similar, in combination with fentanyl, at clinically relevant concentrations. I agree with Drs. Columb and Polley that this is certainly of interest and requires further investigation.

I am surprised that Drs. Columb and Polley devoted so much effort criticizing the analysis of a secondary outcome of the study that we clearly stated requires further investigation. I am intrigued by the criticism of the one-sided Fisher’s exact test pertaining to the incidence of forceps deliveries between the groups. Of note, we found that "all" women in the ropivacaine group were able to ambulate compared with significantly fewer women in the bupivacaine group. Do Drs. Columb and Polley actually believe that any local anesthetic (bupivacaine) that causes significantly more motor blockade would actually have less chance of producing a forceps delivery than one that does not? Our use of the one-sided P value was based on our expected outcome that bupivacaine, which produced significantly more motor blockade, could only cause more forceps deliveries, not fewer.

In summary, I thank Drs. Columb and Polley for their interest in our recent article. I am hopeful that, in our continuing attempts to improve epidural labor analgesia, the issue of relative potencies of these and newer local anesthetics (with or without fentanyl) will be addressed in future investigations using clinically relevant concentrations.

References

1. Campbell DC, Zwack RM, Crone LAL, Yip RW. Ambulatory labor epidural analgesia: bupivacaine versus ropivacaine. Anesth Analg 2000; 90: 1384–9.
2. Capogna G, Celleno D, Fusco P, et al. Relative potencies of bupivacaine and ropivacaine for analgesia in labour. Br J Anaesth 1999; 82: 371–3.
3. Polley LS, Columb MO, Naughton NN, et al. Relative analgesic potencies of ropivacaine and bupivacaine for epidural analgesia in labor: implications for therapeutic indexes. Anesthesiology 1999; 90: 944–50.
4. Columb MO, Lyons G. Determination of the minimum local analgesic concentrations of epidural bupivacaine and lidocaine in labor. Anesth Analg 1995; 81: 883–7.
5. Meister GC, D’Angelo R, Owen M, et al. A comparison of epidural analgesia with 0.125% ropivacaine with fentanyl versus 0.125% bupivacaine with fentanyl during labor. Anesth Analg 2000;90:632–7.
6. Owen MD, D’Angelo R, Gerancher JC, et al. 0.125% ropivacaine is similar to 0.125% bupivacaine for labor analgesia using patient-controlled epidural infusion. Anesth Analg 1998;86:527–31.

This Article Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Columb, M. O. Articles by Campbell, D. C. Search for Related Content PubMed PubMed Citation Articles by Columb, M. O. Articles by Campbell, D. C.