To the Editor:—
Klock et al. 1report that equivalent minimum alveolar concentration (MAC) fractions of sevoflurane and desflurane yielded different responses to endotracheal tube manipulation. We identify three problems in their report.
First, the statistical analysis is questionable. The primary comparison is the incidence of a moderate or severe response to inflation of an endotracheal tube cuff (2 of 32 subjects with sevoflurane vs. 8 of 32 subjects with desflurane). The authors claimed statistical significance using the chi-square test. The P value for the chi-square test with the Yates continuity correction is 0.08; only the uncorrected chi-square test yielded a statistically significant value (P = 0.04). The convention in statistics is that for a 2 × 2 table (i.e. , two groups and two possible outcomes, as in the current study), the Yates correction should be applied. According to Zar, 2“when one calculates the chi-square statistic, the theoretical chi-square distribution is being approximated. This approximation is a very acceptable one, except when v = 1 [i.e. , one degree of freedom, as in a 2 × 2 table] (in which case the Yates correction for continuity usually should be employed).” The P value obtained from the Fisher exact test (which does not rely on an approximation to the theoretical chi-square distribution) is 0.08. Should Klock et al. choose to report a nonstandard statistical approach, they have an obligation to their readers to reveal and justify this decision.
Second, Klock et al. state, “The slope of the regression curve between severity of coughing and heart rate increase was significantly greater for desflurane than for sevoflurane (coefficient of correlation ± SE of the coefficient was 4.4 ± 2.0 for desflurane and 2.3 ± 1.04 for sevoflurane, P < 0.01).” If by “coefficient of correlation” the authors mean “correlation coefficient,” their statement does not make sense: correlation coefficients are bounded between −1 and 1. More likely, the authors used incorrect terminology, intending “coefficient of correlation” to mean “slope.” If so, their analysis is flawed: this difference is not statistically significant.
Finally, MAC fractions imposed in the two experimental groups were probably not equivalent, evidence for another bias in experimental design. Klock et al. calculated anesthetic doses using a MAC value of 6% for desflurane; this value is in the range typically reported. Their MAC value for sevoflurane, 2.05%, the largest of the seven values reported for that age range 3(others range from 1.58 to 1.95%), 4–9is 11% above the average of all the values. The opportunity for bias can be understood by considering the relation between anesthetic dose and airway responsiveness. A published report 10indicates that the concentration–effect relation for coughing is steep. Based on the data of Neelakanta and Miller, 10we determined that the Hill coefficient exceeds 18. Applying this Hill coefficient to the data of Klock et al. indicates that the ED50s for cough suppression for the two anesthetics differ by only 8% (fig. 1). Correcting for the 11% bias in MAC values negates this potential difference in potency.
We conclude that the three experimental flaws compromise the conclusions of Klock et al. regarding differences in the effects of sevoflurane and desflurane on the response to airway stimulation.
The authors thank Edmond I Eger, M.D. (Department of Anesthesia, University of California, San Francisco, California), for his insights.