To the Editor:—
We read with interest the recent article by Adnet et al. 1in which the authors compared the sniffing position with simple head extension for glottic visualization in adults who were anesthetized but not paralyzed. We find the authors’ conclusions not only lacking in statistical power, but also in rationale.
The authors showed the advantage of the sniffing position for visualizing the glottis during laryngoscopy in patients with a body mass index (BMI) greater than 30 as well as for those with head-extension limitation. They demonstrated the superiority of the sniffing position versus simple head extension in these two predictably difficult intubation scenarios. Furthermore, since the authors did not show any inferiority of the sniffing position (compared to simple head extension) in regard to glottic visualization, 2we disagree with their conclusion that “systematic application of the sniffing position offered no appreciable advantage over simple head extension for improvement of glottic visualization with use of direct laryngoscopy and a Macintosh blade.” In addition, since most patients are more comfortable with several centimeters of occipital support, which coincidentally approximates the sniffing position, and since it is certainly easier to remove a pillow placed a priori under the patient's head than to attempt to place one under the head should the indication arise, then it is logical to commence laryngoscopy with a pillow placed under the patient's occiput.
We also disagree with the method the authors chose for their sample size calculation. They calculated the appropriate sample size based upon the assumption that the sniffing position might reduce the incidence of difficult laryngoscopy to approximately the same degree as laryngeal manipulation. 1,3,4However, the authors failed to provide any rationale for such an assumption. Such an arbitrary assumption may have resulted in the calculation of too small a sample size and hence a type II error with resultant insufficient statistical power to discriminate between the experimental groups. For a given sample size, the probability of a type II error is inversely proportional to the degree of difference between the two experimental conditions. Since it was not possible to alter the degree of difference between the two experimental conditions to minimize the likelihood of a type II error, the authors should have (1) made the sample size larger or (2) reduced the population variability. Consequently, once the authors decreased the variability in their sample population by analyzing only the subset that was obese using a multivariate analysis, they demonstrated the distinct advantage of the sniffing position over simple head extension. This advantage was not apparent when they analyzed their entire sample population due to their small sample size, which led them to unjustifiably conclude that “systematic application of the sniffing position offered no appreciable advantage….”
The authors’ statistical analysis may have been further limited by the phenomenon known as convenience sampling . Convenience sampling refers to the phenomenon whereby researchers are limited to using those patients who happen to show up at their hospitals. It also refers to the nuances of the surgical schedule, the good will of the referring physicians and attending surgeons, and the willingness of patients to cooperate. At best, convenience sampling is representative of the patient population at any given institution, with absolutely no assurance that those patients are similar to patients elsewhere. Only 17 to 18% of the patients studied by Adnet et al. had a BMI greater than 30, whereas according to The National Institutes of Health, in the United States, 22.3% of adults are obese. 5Consequently, we believe convenience sampling may apply to the authors’ sample population. The small percentage of patients with a BMI greater than 30 in the study of Adnet et al. may have significantly influenced their results and, hence, their statistical analysis, resulting in their conclusion that “systematic application of the sniffing position offered no appreciable advantage over simple head extension for improvement of glottic visualization.” One would expect to demonstrate a clear superiority of the sniffing position as compared to simple head extension even with systematic application if a similar population size as Adnet's contained a greater percentage of patients with a BMI greater than 30, such as is common in the United States. Likewise, one would not be at all surprised to find no advantage of the sniffing position compared to simple head extension in a sample size population similar to that of Adnet et al. but having a smaller percentage of patients with a BMI greater than 30 even with multivariate analysis. Subsequently, when determining if a procedure should be “systematically applied,” the study sample population should resemble the patient population to which it is to be applied.
We also encountered a possible calculation error on the part of the authors. They calculated a sample size of 222 patients in each group based upon the assumption that the sniffing position might reduce the incidence of difficult laryngoscopy to the same degree as would laryngeal manipulation. However, in table 2 of the study of Adnet et al. , group A included 225 patients with a male/female ratio of 114/98 (or, 212 patients). Not included were 13 patients; however, there was no explanation for their omission from the data. Similarly, in group B, which consisted of 231 patients with a male/female ratio of 130/86 (or, 216 patients), again, 15 patients were omitted, also apparently without explanation. We wonder if this represents a simple typographical error on the part of the authors, or if this was a deliberate exclusion of certain subjects by them. Since these 28 “excluded/typographical error” patients comprise 6% of their total study patient population, their exclusion might represent a serious alteration in the results attained and, hence, their statistical analysis of the data, as well as their power of analysis since the total number for each group is less than the 222 patients per group calculated by the authors.