UMHS LOGOUniversity of Michigan
Department of Pediatrics

Evidence-Based Pediatrics Web Site

Rectal Temperature Is Still the Gold Standard for Determining the Presence or Absence of Fever


  • Is axillary temperature effective in determining the presence or absence of fever?

Clinical Bottom Lines

  1. According to this article axillary temperature plus 1°C is equivalent to the rectal temperature. However, a Cochrane systematic review and clinical experience suggests that the axillary method is unreliable.
  2. Defining fever as 38°C, axillary temperature (+1°C) results in a LR+ of 3 and LR- 0.47. Axillary temperature without adding 1° results in LR+ of 11 and LR- 0.58 due to higher specificity, but sensitivity is much lower.
  3. In most cases we make decisions based on a patient's clinical picture and other data. Except for a few situations (e.g. neonatal period) the exact temperature value is not critical.
  4. When possible, axillary temperatures should be confirmed by rectal or oral temperatures.

Summary of Key Evidence

  1. In a prospective trial, each of 120 patients had rectal, axillary, and forehead temperatures determined. Twenty patients were in each of six arbitrarily determined categories, age < 1 month, 1-5 months, 6-11 months, 12-23 months, 2-14 years, and adults.
  2. The differences between rectal and other methods were calculated for each patient (paired difference) and the mean differences were analyzed.
  3. Rectal temperature was equivalent to axillary temperature plus 1°C for ages greater than 1 month. Less than one month, least squares linear regression analysis showed that rectal temperature was equal to axillary temperature plus 0.2°C for each week of age up to 5 weeks.
  4. Sensitivity and specificity were not calculated by the authors. They claimed that this was not a valid method for assessing diagnostic tests.

Additional Comments

  • A Cochrane literature review of multiple studies concluded that axillary temperature is an "inaccurate, inconsistent and insensitive method in infants and children aged over 1 month" and that there are not appropriate studies in febrile infants less than one month.3
  • Correlation coefficients are not a proper method for comparing methods of clinical measurement. Rather, paired differences for each patient should be employed. The authors suggest that sensitivity and specificity are not useful, and did not calculate these, instead choosing a more complex method called the limits of agreement.2


  1. Shann F, Mackenzie A. Comparison of rectal, axillary, and forehead temperatures. Arch Pediatr Adolesc Med. 1996; 150(1): 74-8.
  2. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986; 1(8476): 307-10.
  3. Duce SJ. A systematic review of the literature to determine optimal methods of temperature measurement in neonates, infants and children. The Cochrane Library. 1996: 1-124.

CAT Author: Jordan Shavit, MD

CAT Appraisers: Jon Fliegel, MD

Date appraised: October 7, 2002

Last updated June 15, 2003
Department of Pediatrics and Communicable Diseases
© 1998-2002 University of Michigan Health System