Institutional members access full text with Ovid®

Agreement Between an Ingestible Telemetric Sensor System and a Mercury Thermometer Before and After Linear Regression Correction

Challis, Graeme G BKin (Hons); Kolb, Jon C PhD

Clinical Journal of Sport Medicine: January 2010 - Volume 20 - Issue 1 - pp 53-57
doi: 10.1097/JSM.0b013e3181c96837
Original Research

Objectives: To evaluate the use of the CorTemp ingestible sensor system to monitor athletes in the prevention of thermal injuries, 3 objectives were established: (1) to determine the agreement between the system and a mercury thermometer and quantify the effect of exceeding the recommended yearly manufacturer calibration, (2) to establish the effect of individual sensor correction on agreement, and (3) to determine the quantity of data required for effective correction.

Design: Validation study.

Participants: Ninety-four ingestible sensors.

Interventions: (1) Five comparisons were made between each sensor and a mercury thermometer across the range of 33 to 41°C. This was performed immediately, and 14 to 18 months, after factory calibration. (2) Linear regression equations were created and used to correct sensor readings at ∼37°C; the corrected value was compared with the mercury thermometer. (3) Linear regression equations were created for each sensor using 2 to 5 data points.

Main Outcome Measures: Systematic bias and random error (95%).

Results: (1) Systematic bias ± random error (95%) was 0.73% ± 0.23% (≈0.27 ± 0.09°C) and 0.54% ± 0.28% (≈0.22 ± 0.11°C) immediately and 14 to 18 months post factory calibration, respectively. (2) Regression correction improved agreement through reductions in systematic bias. (3) Individualized equations using a minimum of 3 comparisons were required to reduce agreement to ±0.10°C; the use of 5 comparisons minimized the number of readings exceeding ±0.10°C.

Conclusions: (1) The CorTemp system inflates temperature measurements compared with a mercury thermometer. (2) Individual sensor calibration is warranted. (3) Correction equations should use a minimum of 3, preferably 5, comparisons. After regression correction, the system displays satisfactory accuracy for preventative monitoring.

From the Human Performance Lab, Faculty of Kinesiology, University of Calgary, Calgary, Alberta, Canada.

Submitted for publication July 14, 2009; accepted October 4, 2009.

Supported by a grant from the Olympic Oval High Performance Fund (J.C.K.). G.G.C. was supported by Natural Sciences and Engineering Research Council and University of Calgary-Faculty of Graduate Studies Graduate Funding.

The authors state that they have no financial interest in the products mentioned within this article.

The authors have no conflict of interest.

Reprints: Jon C. Kolb, PhD, Human Performance Lab, Faculty of Kinesiology, University of Calgary, 2500 University Dr NW, Calgary, AB T2N 1N4 Canada (e-mail:

© 2010 Lippincott Williams & Wilkins, Inc.