| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
Exergen Corporation, Watertown, MA
To the Editor:
I read with concern the recent paper on "a new temporal-artery thermometer" by Suleman et al (1), which concluded that the tested Philips device was "insufficiently precise for clinical use" for children. The authors make no comment that their earlier published abstract of the same study concluded it was "sufficiently accurate and precise for clinical use" for children (2).
A simple analysis shows flaws with the data which led to the contradictory conclusions (Table 1):
|
Since I am the inventor of temporal artery (TA) thermometry (3), and co-inventor (4) of a method which solves the perspiration problem described by Suleman et al, but not employed in their study (1), I have concern over this paper implying the TA method is the problem. The data clearly suggest the Philips-supplied unknown faulty thermometer and incomplete instructions as the problem.
References
University of Louisville, Louisville, KY
In Response:
A cardinal rule of clinical studies is to avoid arbitrarily stopping before the a priori designated number of patients has been enrolled, because stopping early increases the chances of statistical errors, especially when only a small number of patients have been studied. This study is an excellent example: had we stopped enrollment after only 6 of 16 children, we would have incorrectly concluded that the SensorTouch temporal artery thermometer was sufficiently accurate for perioperative use.
Pompei’s assumption that the population or thermometer was changed during the study is unjustified. The entire study was conducted in a single postanesthesia care unit, using identical enrollment criteria. As one might expect, demographic and morphometric characteristics and initial and maximum postoperative temperatures were similar in the two populations (Table 1). A single investigator made each of the measurements. Measurement technique was based on a thorough reading of the instruction manual supplied by Philips, Inc.
|
When evaluating the validity of a thermometer, it is important to retain information associated with individual measurements, otherwise averaging makes devices appear better than they actually are. We thus report the values for each measurement in each individual. Similarly, the inaccuracies we report are based on statistical evaluation of individual measurements. However, the proper statistical approach is quite different when the question is whether the initial pediatric patients differed from the subsequent ones. Pompei’s "simple analysis" comparing the two groups is incorrect because it falsely assumes that the 246 observations are independent. Six patients contributed 96 paired measurements that were reported in our abstract, whereas 10 subsequent patients contributed an additional 150 data pairs to the 246 that were reported in our paper. The appropriate way to compare these two groups is to first average the inaccuracy in each patient, and then compare these values in the first group (n=6) with those in the second (n=10). Done this way, the two groups do not differ significantly (P=0.1).
Performance of the SensorTouch system was considerably worse in adults than in the pediatric patients. This is consistent with thicker subcutaneous tissue in adults. It is thus unsurprising that the amount of inaccuracy correlated with patient age in the pediatric group (r=0.62, P=0.01). As it happened, there was only a single child over the age of five years among the first six patients whereas there were three, including the two oldest, in the subsequent ten patients. Worse accuracy in these older patients explains the slight, non-significant difference between the groups that Pompei notes.
We consider temperature measurement systems and sites suitable for perioperative monitoring when nearly all the measurements are within 0.5°C of core temperature (i.e., a 1°C span around core temperature). This requirement is somewhat arbitrary, but one that we’ve used consistently. In each case, though, we report the actual results; this give clinicians the option to accept wider limits in clinical situations where speed, cost, or ease-of-use are more important than accuracy. There might well be situations, such as home use, in which the performance of the temporal artery thermometer is considered adequate.
Pompei’s focus on slight differences between our initial and subsequent pediatric patients does not alter the more important conclusion that the accuracy and precision of temporal artery thermography was remarkably poor in the adults. Specifically, 89% of the temporal artery temperatures differed by more than 0.5°C, and the sensitivity for detecting fever was zero.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
