Free
Correspondence  |   December 2005
Graphical Display of Data Could Reveal Errors in Statistical Tests
Author Notes
  • HAMPERSANDNUMBERSIGNx000F4;pital Ambroise ParAMPERSANDNUMBERSIGNx000E9;, Boulogne Billancourt, France.
Article Information
Correspondence
Correspondence   |   December 2005
Graphical Display of Data Could Reveal Errors in Statistical Tests
Anesthesiology 12 2005, Vol.103, 1314. doi:0000542-200512000-00031
Anesthesiology 12 2005, Vol.103, 1314. doi:0000542-200512000-00031
To the Editor:AMPERSANDNUMBERSIGNx02014;
I read with interest the article about intraoperative remifentanil infusion by Lee et al.  I have major concerns regarding the nonsignificance of chi-square tests for the behavioral pain score during the first 15 min in the recovery room. The bar representation in their figure 1 is appropriate for expressing these results and clearly shows a different comportment of patients in the two groups. After redoing the analysis by extrapolated number of patients according to bar height, it seems that, as expected, significant activity for chi-square tests is very high at all times studied (). Calculations were made with JMP 5.1 (SAS Institute, Cary, NC). The comment in the text says that 60% of patients in the remifentanil group versus  40% in the nitrous oxide group have a behavioral pain score of 0. That does not match the figure in the article. Furthermore, the percentage of total patients exceeds 100% at T5 for the remifentanil group.
Table 1. Numbers of Patients Classified by Behavioral Pain Scores 
Image not available
Table 1. Numbers of Patients Classified by Behavioral Pain Scores 
×
If these results are not type errors, we could have question about the morphine titration. In our study, the morphine titration was based essentially on behavioral scale during the first 15 min after extubation, because of the difficulty to have correct pain assessment just based only on the visual analog pain scale score at this moment. Another difference with our study is the use of fentanyl at induction and morphine at skin incision. The time to first dose of morphine does not appear in the results. The authors' conclusion could be right, but the discrepancies in the presented data may alter this finding. Opioid tolerance is not always clinically significant because of patient variability, surgery duration, opioid dosage, or concomitant medication. Graphics can help us to show clinical evidence, and statistical tests are used to confirm and valid ideas revealed by data. The high publication pressure should not deserve statistical review. Analysis and criticism are the guaranties of medical research.
HAMPERSANDNUMBERSIGNx000F4;pital Ambroise ParAMPERSANDNUMBERSIGNx000E9;, Boulogne Billancourt, France.
References
Lee L, Irwin M, Lui SK: Intraoperative remifentanil infusion does not increase postoperative opioid consumption compared with 70% nitrous oxide. Anesthesiology 2005; 102:398AMPERSANDNUMBERSIGNx02013;402Lee, L Irwin, M Lui, SK
Guignard B, Bossard AE, Coste C, Sessler DI, Lebrault C, Alfonsi P, Chauvin M: Opioid tolerance: Intraoperative remifentanil increases postoperative pain and morphine requirement. Anesthesiology 2000; 93:409AMPERSANDNUMBERSIGNx02013;17Guignard, B Bossard, AE Coste, C Sessler, DI Lebrault, C Alfonsi, P Chauvin, M
Schriger D, Cooper R: Achieving graphical excellence: Suggestions and methods for creating high-quality visual displays of experimental data. Ann Emerg Med 2001; 37:75AMPERSANDNUMBERSIGNx02013;87Schriger, D Cooper, R
Hawkins BA: More haste, less science (letter). Nature 1999; 400:498Hawkins, BA
Table 1. Numbers of Patients Classified by Behavioral Pain Scores 
Image not available
Table 1. Numbers of Patients Classified by Behavioral Pain Scores 
×