Rectal temperature is generally considered to be more accurate than axillary temperature in children, but rectal measurement is sometimes difficult to obtain in a child. To examine whether axillary temperature is equivalent to rectal temperature, Craig and colleagues reviewed the literature of studies comparing these two sites of temperature measurement in children.

Electronic searches of medicine and nursing literature were performed to identify studies in which rectal and axillary temperatures were compared. Unpublished studies were identified in a national research registry and by reviewing abstracts from conferences. Some authors of studies and manufacturers of thermometers were also contacted to provide additional information. Excluded from the meta-analysis were infants of less than 37 weeks' gestational age and children with hypothermia. Also excluded were studies in which different types of thermometers were used at the two sites and studies in which mercury thermometers were read before three minutes of contact.

A checklist was used to evaluate the quality of each study. The mean and standard deviations of the difference between rectal and axillary temperatures were determined in each study. If outcome data were not provided, the study's investigators were contacted for information on the means and standard deviations. The 40 studies that were reviewed included 5,528 children from birth to 18 years.

The analysis revealed that the mean axillary temperature was always lower than the mean rectal temperature. Large mean differences and wide limits of agreement were found between rectal and axillary temperatures. The validity of temperature readings at both sites was related to the child's age and the duration of contact with the measuring device.

The authors conclude that, in children, axillary measurement of temperature does not agree sufficiently with rectal temperature measurements to provide clinically reliable data. The wide range in the mean differences suggests that rectal temperature cannot be estimated by adding 1°C to the temperature measured at the axilla.