Validation of an ingestible temperature data logging and telemetry system during exercise in the heat.

Aim: Intestinal temperature telemetry systems are promising monitoring and research tools in athletes. However, the additional equipment that must be carried to continuously record temperature data limits their use to training. The purpose of this study was to assess the validity and reliability of a new gastrointestinal temperature data logging and telemetry system (e-Celsius™) during water bath experimentation and exercise trials. Materials and
Methods: Temperature readings of 23 pairs of e-Celsius (TeC) and VitalSense (TVS) ingestible capsules were compared to rectal thermistor responses (Trec) at 35, 38.5 and 42°C in a water bath. Devices were also assessed in vivo during steady-state cycling (n = 11) and intermittent running (n = 11) in hot conditions.
Results: The water bath experiment showed TVS and TeC under-reported Trec (P<0.001). This underestimation of Trec also occurred during both cycling (mean bias vs TVS: 0.21°C, ICC: 0.84, 95% CI: 0.66-0.91; mean bias vs. TeC: 0.44°C, ICC: 0.68, 95% CI: 0.07-0.86, P<0.05) and running trials (mean bias vs. TVS: 0.15°C, ICC: 0.92, 95% CI: 0.83-0.96; mean bias vs. TeC: 0.25, ICC: 0.86, 95% CI: 0.61-0.94, P<0.05). However, calibrating the devices attenuated this difference during cycling and eliminated it during running. During recovery following cycling exercise, TeC and TVS were significantly lower than Trec despite calibration (P<0.01).
Conclusion: These results indicate that both TeC and TVS under-report Trec during steady-state and intermittent exercise in the heat, with TeC predicting Trec with the least accuracy of the telemetry devices. It is therefore recommended to calibrate these devices at multiple temperatures prior to use.

Introduction

Gastrointestinal temperature, measured via an ingestible telemetry device, has been shown to be a valid index of core temperature, with responses displaying similar profiles as measures made via the rectum or esophagus. Intestinal temperature assessment is also a popular technique in operational and occupational settings. Such wireless technology (e.g. VitalSense, CorTemp) provides a comfortable and practicable means of monitoring thermal strain. Although an expensive method relative to minimally invasive and reusable alternatives (e.g., axilla, oral or temporal temperatures), intestinal temperature provides greater validity as an index of core temperature. However, in athletic settings the additional equipment that must be carried to receive data from the ingested capsule might impact upon performance. Moreover, it may be prohibited to carry additional equipment during sanctioned competition, thus limiting the use of such devices to monitoring training.

Recent technological advances have led to ingestible sized capsules that are capable of both live temperature telemetry and data storage. Negating the need for additional cumbersome equipment, the e-Celsius™ system (BodyCap, Caen, France) presents as a promisingly useful tool for monitoring core temperature within training and competition. Previous studies show the capsules meet standard electrical thermometry performance requirements and have yielded similar results when compared to existing implantable telemetry systems in rats. Despite this, intestinal temperature profiles typically display slower responses to rapid body temperature changes in humans when compared to other measurement sites (i.e. pulmonary artery or esophagus) used in clinical settings. Therefore direct comparisons of gastrointestinal telemetry systems to a clinical criterion standard are often used to assess their measurement accuracy during exercise in the heat.

To date, the validity of the e-Celsius™ system as a device for monitoring steady state and intermittent exercising temperature responses in the heat compared to existing invasive and non-invasive methods remains unknown. Therefore, the aim of this study was to first use a water bath experiment to determine the validity and reliability of the e-Celsius™ system compared to another validated commercially available ingestible telemetry system (VitalSense, Philips Respironics, Bend, Oregon, USA), and a criterion standard used in laboratory research; rectal temperature. A secondary aim was to compare the performance of the device during both steady-state cycling and intermittent running in hot-humid conditions. It was hypothesized that the e-Celsius™ system would perform similarly to the existing system and criterion standard, and would therefore be a valid and reliable tool for use in laboratory and field settings.

Materials and methods

Participants

A total of 16 participants took part in the study, which was conducted at the Athlete Health and Performance Research Center laboratories at Aspetar, Qatar. Experiments consisted of low-intensity steady state cycling (n = 11) and/or intermittent soccer-simulating treadmill running (n = 11) in laboratory conditions. Six participants completed both trials. Cycling participants' average age, height, body mass and maximal oxygen uptake (VO2max) were 33 ± 6 y, 179 ± 6 cm, 78.3 ± 5.3 kg, 54.6 ± 5.4 ml/kg/min, whereas running participants were 33 ± 6 y, 179 ± 6 cm, 78.3 ± 10.2 kg and 56.2 ± 5.4 ml/kg/min. All participants were recreationally trained cyclists and team sport players and provided written informed consent prior to participation. The study was approved by the institute's ethics review committee (Anti-Doping Lab Qatar) and conforms to the declaration of Helsinki.

Ingestible telemetric temperature systems

The e-Celsius™ capsule is 17 mm long and 8.2 mm in diameter. Temperature data is sampled and transmitted every 30 s, and is displayed in real time on an external receiver (e-Viewer, BodyCap, Caen, France) when within ∼1 m of the ingested capsule. When not in contact with the receiver, data is stored within the capsule and may be downloaded to the receiver at the end of a data acquisition period for further analysis. The VitalSense capsule is 23 mm in length and 8.7 mm in diameter. Once activated, data can be immediately transmitted by radio telemetry a minimum of every 15 s to a wireless ambulatory chest strap that contains a variety of sensors to monitor physiological status (EquiVital Life Monitor, Hidalgo Ltd. Cambridge, UK). The data collected can be displayed and stored on a laptop via Bluetooth®.

Water bath experiment

Prior to each experimental trial one e-Celsius™ and one VitalSense temperature capsule were activated and assigned to a dedicated data receiver. Receivers were synchronised on activation to ensure temperature measurements of the eCelsius™ (TeC) and VitalSense (TVS) systems occurred simultaneously and data could be later averaged over the course of a 1 min period. All capsules (23 pairs) and rectal temperature (Trec) thermistors underwent a 3-point calibration in a water bath (WNB 14, Memmert GmBH, Swabach, Germany) at 35, 38.5 and 42°C. Temperature of the bath was set according to an uninsulated digital thermistor (MAC flexible probe, Ellab, Hillerød, Denmark) with a temperature variation of ±0.1°C across 15–45°C, which was used as a reference standard. Capsules were lowered into a thin meshed metal container so that they did not touch the interior surfaces of the water bath. The reference temperature and rectal thermistors were secured to the edge of the bath by tape so that they were fully immersed by a minimum of 5 cm and placed in close proximity of the capsules. Care was taken to ensure capsules did not come into contact with each other while water was constantly circulated in the bath. Following a 15 min period of stable reference temperature readings at 35°C, a 20 min measurement period began where temperature readings were sampled every minute from each device (i.e., reference temperature, TeC, TVS and Trec). At the end of this period the process was repeated at 38.5 and 42°C, respectively. Following the water bath experiment each thermistor temperature could then be manually corrected using raw values in the equation:where; the Observed Value is the raw (i.e. uncorrected) data transmitted by the respective device at a given time and the Intercept and Slope are values obtained from regression analysis of the device vs. the reference temperature over the 3-point calibration.

Pre-experimental procedures

In an initial visit, participants completed an exercise test to determine VO2max on an electronically braked cycle ergometer (Schoberer Rad Meßtechnik; SRM, Jülich, Germany) and/or motorised treadmill (Cosmed T170 DE Med, hp-cosmos, Nussdrof-Traunstein, Germany) in ∼20°C and ∼50% relative humidity (RH) conditions. Participants completing both experimental trials undertook both VO2max tests on separate days. They were required to abstain from vigorous physical activity and alcohol consumption for 24 h and caffeine intake for 12 h prior to attending the laboratory.

The cycling test consisted of pedalling at a self-selected cadence while resistance was applied in the order of 5 W every 10 s until volitional exhaustion, or cadence fell below 60 rpm, despite strong verbal encouragement. The treadmill test consisted of running at a speed of 12.5 km/h at a 0% gradient for a 2 min period, after which the gradient was increased by 2% every 2 min until the point of volitional fatigue. During each test heart rate was measured continuously using a telemetry belt strapped to the chest (T31, Polar Electro, Kempele, Finland) and pulmonary gas exchange was measured via on-line analysis of breath-by-breath gases (MasterScreen CPX, Carefusion GmBH, Germany). VO2max was defined as the mean VO2 over the final min of the incremental cycling or running test, while maximal aerobic power output in the cycling test was determined as the highest mean power output over 30 s.

Experimental procedures

Participants reported to the laboratory on the day before each experimental trial. They were provided with one e-Celsius™ and one VitalSense telemetry capsule and were instructed to ingest both ∼8 hours prior to experimentation with water, followed by a standardised light meal to aid transit into the gastrointestinal tract. All trials were completed on separate days with a minimum of 24 h between each visit. In cases where participants completed both the cycling and running trial, separate pairs of capsules were ingested to standardise time between ingestion and experimentation.

On arrival to the laboratory for the experimental trial, nude body mass was measured before participants self-inserted a rectal thermistor (MRB rectal probe, Ellab, Hillerød, Denmark) 15 cm beyond the anal sphincter for measurement of Trec. For the cycling trials, participants wore cycling shorts, socks and cycling shoes, and during the running trials wore shorts, socks and running shoes. During each trial they wore a heart rate monitor and an EquiVital Life Monitor ambulatory belt that housed the TVS receiver. Participants then sat quietly in the laboratory (22.3 ± 2.6°C and 50 ± 9% RH) while heart rate and Trec were measured each min over a 5 min period. Participants then moved into an environmental chamber to complete the cycling or running protocol. Average ambient conditions during the cycling and running trials were 35.1 ± 0.3°C and 52.3 ± 3.5% RH, and 32 ± 0.8°C and 41 ± 7.8% RH, respectively. Cycling trials consisted of cycling on the SRM ergometer at 30% of maximal aerobic power output for an initial 10 min, followed by a further 50 min at 45% of maximal aerobic power output. The intermittent soccer-specific running task consisted of an initial 10 min self-paced warm up on the motorised treadmill, followed by a 5 min rest/stretching period before the task began. The running task was designed to simulate the typical activity profile of 45 min of soccer. The duration, speed and pattern of the movements of the running task are displayed as a 15 min block in Figure 1, which was repeated 3 times consecutively. Heart rate and Trec were recorded at 5 min intervals, while ambient conditions were recorded every 10 min throughout the trials. Participants were permitted to drink water ad libitum during the trial. All water was placed into the environmental chamber ∼90 min before trial commencement to allow it to equilibrate to the conditions, thereby minimising any potential confounding effects on telemetry capsule temperature.

Figure 1.

Intermittent treadmill speeds over 15 min of the intermittent running protocol. Each 15 min block was repeated 3 times consecutively. Adapted from Clarke et al. and Drust et al.

At the end of each trial, participants immediately exited the environmental chamber and lay supine in the laboratory (24.2 ± 2.3°C and 45 ± 8% RH) for a period of 20 min. During this time data from the trial in the TeC capsule were downloaded and post-exercise TVS and Trec were monitored. Laboratory and environmental chamber conditions were monitored throughout the trials using a wet bulb globe temperature monitor (Kestrel 4400 Heat Stress Meter, loftopia, LLC, Birmingham, MI, USA).

Statistical analyses

Data was coded and analyzed using computer software (SPSS Version 20, Chicago, Il, USA). Mean difference (bias) and 95% confidence intervals between each thermistor and the reference temperature in the water bath experiment were computed using a 2-tailed paired t-test. Mean difference (bias) and 95% confidence intervals between TeC, TVS and Trec in the water bath and laboratory experiments were computed using a 2-tailed paired t-test considering Trec as the criterion standard for body core temperature. The differences between each telemetry capsule and Trec recording at each of the 3 temperature points during the water bath experiment were compared using One-way Analysis of Variance (ANOVA) to establish whether the differences vary at different temperature points. Absolute changes in temperature during exercise and recovery were converted to rates (i.e., Δ°C/min) and compared to assess the sensitivity of measurement sites and devices. In order to assess the inter-measure agreement, Intra Class Coefficients (ICC) for agreement and 95% confidence intervals (CI) were reported. Limits of Agreement (LOA) at 95% were computed as the product of the standard deviation of the mean difference between the thermistors and a reference value of 1.96. Standard error of the measurement (SEM)/typical error was estimated from the square root of the mean square residual from the ANOVA output. Data are reported as mean ± SD unless otherwise stated. The level of significance was set at P = 0.05.

Results

Both TeC and TVS reported significantly different temperatures compared to the reference temperature across the three calibration temperatures (bias: 0.34°C, 95% CI: 0.31 – 0.37°C; bias: 0.18°C, 95% CI: 0.18 – 0.19°C; respectively, P < 0.001). Although a significant difference was noted between Trec and reference temperature at 35°C (bias: 0.03°C, 95% CI: 0.02 – 0.04°C, P < 0.01), no difference was observed at 38.5 and 42°C. Both TeC and TVS were significantly lower than Trec across all temperature ranges (P < 0.001; Table 1). TeC also consistently under-reported both Trec and TVS (P < 0.001). One-way ANOVA indicated that the difference between the reference temperature and both telemetry capsules at 35°C was significantly different to that at 42°C with a mean difference of −0.03°C (95% CI: −0.05 - −0.01°C; P<0.01), while no other differences were observed. The SEM values were increasing with higher calibration temperature points. Relative to Trec, overall TVS showed better performance (95% LOA ±0.30 and SEM: 0.11) compared to TeC (95% LOA ±0.34 and SEM: 0.12).

Table 1.

Reliability statistics for Tre, TVS and TeC thermistors undergoing a three-point calibration in a temperature controlled water bath (n = 23).

		Mean Bias (°C) & 95% Confidence Interval	Intraclass Correlation Coefficient & 95% Confidence Interval	Limits of Agreement (95%)	Standard Error of Measurement (°C)
Trec – TVS	35.0	0.13 (0.12 - 0.15)*	0.04 (–0.03 - +0.11)	±0.26	0.09
	38.5	0.17 (0.15 - 0.18)*	–0.05(–0.13 - +0.25)	±0.28	0.10
	42.0	0.20 (0.19 - 0.22)*	0.03 (–0.03 - +0.94)	±0.34	0.12
Overall		0.17 (0.16 - 0.18)*	0.99 (0.98 - 0.99)	±0.30	0.11
Trec – TeC	35.0	0.19 (0.18 - 0.21)*	0.02 (–0.03 - +0.08)	±0.30	0.11
	38.5	0.24 (0.22 - 0.25)*	0.05 (–0.04 - +0.12)	±0.29	0.10
	42.0	0.26 (0.24 - 0.28)*	0.04 (–0.03 - +0.04)	±0.40	0.14
Overall		0.23 (0.22 - 0.24)*	0.99 (0.89 - 0.99)	±0.34	0.12
TVS – TeC	35.0	0.06 (0.05 - 0.07)*	0.54 (0.31 - 0.68)	±0.20	0. 07
	38.5	0.07 (0.06 - 0.08)*	0.35 (0.12 - 0.52)	±0.21	0.07
	42.0	0.06 (0.05 - 0.07)*	0.54 (0.37 - 0.66)	±0.23	0.08
Overall		0.06 (0.06 - 0.07)*	0.99 (0.99 - 0.99)	±0.21	0.08

*Significant difference, P<0.01

Cycling trial

Prior to cycling exercise, raw/uncorrected temperature values from the capsules reported significantly lower than raw Trec values (P < 0.05; Fig. 2A, Table 2). There were no differences in raw Trec and raw TVS during cycling except at 50 and 55 min (0.22°C, 95% CI: 0.02–0.43 and 0.27°C, 0.06–0.49°C, respectively; P < 0.05). Raw TVS recordings were also significantly lower than raw Trec during the recovery period following exercise (P < 0.001). Raw TeC recordings were significantly lower than raw Trec at baseline and from 5 min of exercise until the end of the recovery period following cycling (P < 0.05). Raw TeC was also significantly lower than raw TVS at various time points throughout baseline, cycling and recovery (P < 0.05; Fig. 2A).

Figure 2.

Raw (A) and corrected (B-D) temperature responses to 10 min cycling at 30% of maximal aerobic power output followed by 50 min at 45% of maximal aerobic power output. *Significant difference between Trec and TVS, P < 0.05; #Significant difference between Trec and TeC, P < 0.05; †Significant difference between TVS and TeC, P < 0.05.

Table 2.

Reliability statistics for TVS and TeC against Trec, and for TeC against TVS during cycling in the heat and post-exercise recovery in cool conditions (n = 11). Data presented is raw uncorrected values.

		Mean Bias (°C) & 95% Confidence Interval	Intraclass Correlation Coefficient & 95% Confidence Interval	Limits of Agreement (95%)	Standard Error of Measurement (°C)
Trec – TVS	Pre	0.13 (0.03 - 0.23)*	0.86 (0.45 - 0.97)	±0.29	0.11
	Cycle	0.11 (0.06 - 0.16)*	0.90 (0.84 - 0.93)	±0.59	0.21
	Post	0.55 (0.47 - 0.62)*	0.45 (–0.08 - +0.79)	±0.50	0.18
Overall		0.21 (0.16 - 0.25)*	0.84 (0.66 - 0.91)	±0.66	0.24
Trec – TeC	Pre	0.31 (0.22 - 0.34)*	0.58 (–0.10 - 0.88)	±0.44	0.16
	Cycle	0.34 (0.28 - 0.40*	0.76 (0.27 - 0.90)	±0.73	0.26
	Post	0.78 (0.68 - 0.89)*	0.27 (–0.08 - +0.63)	±0.69	0.25
Overall		0.44 (0.38 - 0.49)*	0.68 (0.07 - 0.86)	±0.79	0.29
TVS – TeC	Pre	0.18 (0.04 - 0.32)*	0.72 (0.11 - 0.92)	±0.41	0.15
	Cycle	0.24 (0.19 - 0.28)*	0.84 (0.50 - 0.93)	±0.58	0.25
	Post	0.25 (0.14 - 0.36)*	0.62 (0.22 - 0.81)	±0.68	0.21
Overall		0.24 (0.19 - 0.28)*	0.82 (0.50 - 0.92)	±0.59	0.21

*Significant difference, P<0.05

Corrected temperature values altered the relationship between Trec, TVS and TeC during exercise (Fig. 2B and 2C). During the recovery period, corrected TVS and TeC were significantly lower than Trec (P < 0.01; Fig. 2B and 2C). Corrected TeC and TVS did not differ throughout the trial except at 5 and 60 min during cycling exercise (P < 0.05; Fig. 2D). The average increase in the corrected values for Trec (0.033 ± 0.006°C/min) during cycling was significantly greater than that of TeC (0.026 ± 0.008°C/min; P < 0.05), but similar to TVS (0.028 ± 0.007°C/min, P = 1.22). During the 20 min recovery period after exercise, the temperature change per minute was similar across all 3 measurements (∼0.049°C/min, P > 0.05).

Throughout the cycling trial, telemetry devices were significantly biased, under-reporting raw Trec values (P < 0.01; Fig. 3A and 3B). Although bias was significantly away from 0 (P < 0.05), raw TVS showed excellent reliability during cycling with an ICC: 0.90 (95% CI: 0.84–0.93) and SEM: 0.21°C (Table 2). On the other hand, raw TeC provided significantly greater bias (P < 0.05). Moreover, although the ICC was as high as 0.76, it provided wider CI (0.27 to 0.90). Although systematic bias was present for both raw TVS and Raw TeC, this was not proportional (i.e. bias did not increase proportionally with magnitude of the temperature measured; Fig. 3A and 3B) and appears to have been introduced mostly during the recovery period. A significant difference was also noted between telemetry devices (P < 0.01; Table 2, Fig. 3C). During exercise, the SEM was 0.24 and 0.29°C for raw TVS and raw TeC relative to raw Trec, respectively (P<0.05). While the SEM decreased during post-exercise recovery, the under prediction of raw Trec by telemetry increased (Table 2). Raw TVS capsules reported significantly higher values than raw TeC throughout the cycling trial, with a mean bias of 0.24 (95% CI: 0.19–0.28°C).

Figure 3.

Bland-Altman plots of raw TVS and TeC vs. Trec (A and B) and TVS vs. TeC (C) prior to exercise, during 60 min of cycling and 20 min of post-exercise recovery.

Running trial

One participant had passed the VitalSense capsule prior to commencing the trial. Therefore, data are presented for 10 participants for all TVS running analyses. During the running trial, raw Trec values were significantly higher than raw TVS from 10–35 min (P<0.05), whereas they were higher than raw TeC from 5 to 15 min of exercise and from 30 min to end of the recovery period, except at 5 min post (P<0.05; Fig 4A). Raw TeC were significantly lower than raw TVS until 10 min into the running task and again at the end of exercise (P < 0.05).

Figure 4.

Raw (A) and corrected (B-D) temperature responses to a 10 min self-paced warm up and 45 min of intermittent treadmill running, followed by 20 min of recovery in cool conditions. *Significant difference between Trec and TVS, P < 0.05; #Significant difference between Trec and TeC, P<0.05; †Significant difference between TVS and TeC, P < 0.05.

Corrected temperature values for Trec, TVS and TeC were not different during exercise and recovery (Fig. 4B and 4C), whereas corrected values resulted in a higher TVS compared to TeC at baseline until the beginning of the trial (P < 0.05; Fig. 4D). When the increases in corrected temperature were compared, the increase in Trec (0.031 ± 0.009°C/min) was significantly greater than that of TVS (0.027 ± 0.009°C/min; P < 0.01) and TeC (0.026 ± 0.009°C/min; P < 0.01). Following exercise, TVS decreased at a faster rate than Trec (−0.054 ± 0.021 vs −0.054 ± 0.015°C/min; P < 0.05), while TeC declined at a similar rate to Trec (P = 0.12).

Overall, raw TVS and raw TeC were lower than raw Trec during intermittent running and post-exercise recovery (P < 0.01; Table 3, Fig. 4A). Although bias was significantly away from 0 (P < 0.05), the ICC for agreement was excellent for raw TVS during the run (ICC: 0.90 (95% CI: 0.70 to 0.96)) and raw TeC (ICC: 0.83 (95% CI: 0.51–0.92)). The data indicated a higher raw Trec bias compared to raw TVS and raw TeC, respectively (P < 0.01; Table 3, Fig. 5A and 5B). Although systematic bias was present, as the line of equality lies away from the mean bias, this was not proportional (i.e., bias did not increase proportionally with magnitude of the temperature measured; Fig. 5A and 5B). As with the cycling trial, the bias appears to have been introduced mostly during the recovery period. Furthermore, a mean bias was observed for raw TVS compared to raw TeC throughout (Fig. 5C). During the intermittent running task, a significant mean bias for raw Trec was observed between the telemetry devices (P < 0.01; Table 3).

Table 3.

Reliability statistics for TVS and TeC against Trec, and for TeC against TVS during during intermittent running in the heat and post-exercise recovery in cool conditions (n = 10). Data presented is raw uncorrected values.

		Mean Bias (°C) & 95% Confidence Interval	Intraclass Correlation Coefficient & 95% Confidence Interval	Limits of Agreement (95%)	Standard Error of Measurement (°C)
Trec – TVS	Pre	0.03 (–0.16 - +0.23)	0.73 (0.22 - 0.93)	±0.53	0.19
	Run	0.18 (0.13 - 0.23)*	0.90 (0.70 - 0.96)	±0.48	0.17
	Post	0.14 (0.05 - 0.23)*	0.91 (0.81 - 0.96)	±0.46	0.20
Overall		0.15 (0.11 - 0.19)*	0.92 (0.83 - 0.96)	±0.51	0.18
Trec – TeC	Pre	0.17 (–0.06 - +0.39)	0.61 (0.10 - 0.87)	±0.66	0.24
	Run	0.26 (0.20 - 0.33)*	0.83 (0.51 - 0.92)	±0.66	0.24
	Post	0.29 (0.19 - 0.40)*	0.84 (0.47 - 0.94)	±0.70	0.25
Overall		0.25 (0.20 - 0.30)*	0.86 (0.61 - 0.94)	±0.67	0.24
TVS – TeC	Pre	0.15 (0.03 - 0.27)*	0.87 (0.34 - 0.97)	±0.32	0.12
	Run	0.11 (0.07 - 0.15)*	0.95 (0.89 - 0.97)	±0.42	0.15
	Post	0.14 (0.03 - 0.25)*	0.89 (0.78 - 0.94)	±0.67	0.24
Overall		0.12 (0.08 - 0.16)*	0.94 (0.88 - 0.96)	±0.48	0.17

*Significant difference, P<0.05

Figure 5.

Bland-Altman plots of raw TVS and TeC vs. Trec (A and B) and TVS vs. TeC (C) prior to exercise, during 45 min of intermittent running and 20 min of post-exercise recovery.

Discussion

This study sought to determine the validity and reliability of the e-Celsius™ system from BodyCap as a method for monitoring core temperature during exercise in hot/humid environments. This is the first study to examine the performance of this core temperature data logging and monitoring device in relation to an existing telemetry system (i.e. VitalSense) and medical grade thermistors in both water bath and in vivo human exercise experiments. The main findings from this study are that i) both TeC and TVS reported lower temperatures across the three calibration temperatures relative to the reference value and Trec, with TeC also under-reporting TVS, ii) the e-Celsius™ capsules consistently reported lower temperatures than both the Trec and TVS during steady-state cycling, as well as at the onset and termination of intermittent running, iii) and during rapid and large changes in body temperature with exercise in the heat, both TeC and TVS significantly under-reported Trec throughout cycling and intermittent running. However, calibrating the devices significantly attenuated this difference during steady state cycling and eliminated it during the intermittent running task. Hence, it is strongly recommended to calibrate intestinal telemetry systems prior to use.

For evaluating the validity and reliability of devices such as intestinal telemetry systems, an informed decision is required when considering whether the device meets agreement standards. The device must provide low bias, high ICC with the reference temperature, low SEM and narrow 95% LOA. Moreover, the validity statistics should be consistent across different exercise conditions, like steady-state cycling and intermittent running in hot conditions. The differences between devices during the water bath experiment were not equal across the 3 temperatures. This suggests that the relationship between our reference thermistor and the capsules was not linear. A potential limitation of the present study is that we did not manipulate bath temperature over time and record the responses of devices at discrete intervals. This would allow a non-linear correction to be applied to the raw data. Notwithstanding, the largest difference observed in the relationship between the telemetry devices and reference temperature was 0.03°C (both TVS and TeC). Given this minimal difference and the sensitivity of the rectal thermistor in reporting to the nearest 0.1°C, a calibration of this nature would likely not have made a detectable difference to the corrected data presented.

Chapon et al. previously calibrated a prototype of the e-Celsius™ system and observed 88% of capsules used reported temperatures within ±0.2°C of their criterion standard. In the present investigation, the 23 capsules used in the water bath experiment reported similar lower (0.23°C) and upper (0.25°C) temperatures when calculated using a 95% limits of agreement. Adjusting these limits to a similar level to that of Chapon et al., the lower and upper levels of agreement remain unchanged across the 3 points (35, 38 and 42°C). Despite the differences in sensitivity of the reference probe used in this study (0.1°C) and that of Chapon et al. (0.01°C), the mean bias and range of TeC values from our water bath experiment are slightly larger (0.24 ± 0.14°C). Conversely, the average difference between our reference thermistor and TVS was 0.18 ± 0.14°C, while the difference between the 2 capsules in the water bath experiment is similar to that previously reported. Therefore, it is recommended that intestinal temperature telemetry capsules undergo calibration prior to use.

Although previous studies report varying relationships, intestinal temperature tends to be higher than that of the rectum during passive rest in temperate environments. In the present study, both TeC and TVS were lower than Trec prior to cycling, whereas similar temperatures were observed before the running task (Fig. 2A and 4A). When corrected, both telemetry systems reported a temperature similar to that of Trec at rest, which is in agreement with numerous other observations. Data from the water bath experiments also indicated a consistent under-reporting of Trec by both TVS and TeC (Table 2). Therefore the differences at rest in the present study compared to previous work are likely due to the inherent difference between both temperature telemetry systems to the rectal thermistor used.

During exercise, a large amount of evidence suggests acceptable agreement between intestinal and rectal temperatures at levels of moderate hyperthermia up to temperatures of ∼39.5°C. A review by Byrne and Lim highlighted that the bias between rectal and intestinal temperature varies in magnitude and direction. However, the authors concluded there is an acceptable agreement of <0.4°C between temperatures. During the running trial, there were no differences in corrected temperatures across all 3 thermistors (Fig. 4B and 4C). However, this is in contrast to the cycling trial, which was conducted under conditions of greater heat stress. At rest and during the first few minutes of cycling, corrected TVS and TeC were similar to Trec. However, after 50 min of exercise Trec was significantly higher than corrected TeC (Fig. 2C). While the average differences between TeC and Trec at this point was within the acceptable limit (∼0.25°C), the lower and upper 95% confidence intervals for these time points ranged from 0.03–0.51°C, respectively. This suggests a proportional systematic bias and a poor level of agreement when core temperature approaches ∼39°C during exercise in the heat (Fig. 3A and 3B). After the cessation of exercise and transfer into cooler conditions, the mean bias increased further as core temperature declined. A more rapid decrease in intestinal temperature compared to Trec following exercise is a typical response, yet these changes tend to be slower than temperature measured via the esophagus. Therefore this should be considered when using telemetry devices where it may be necessary to lower an individual's core temperature rapidly.

Kolka et al. observed intestinal temperature responded more rapidly than that of rectal temperature during moderate and intense exercise in warm conditions (29.5°C). In a more recent study by Teunissen et al., changes in rectal and intestinal temperature were similar (0.75 and 0.8°C, respectively) after 10 min of submaximal exercise followed by 8 min of maximal self-paced cycling in similar conditions to that used by Kolka et al. Easton et al. also observed no differences in rectal and intestinal temperature during 40 min of steady state cycling followed by a 16 km time trial. This is in contrast to the raw temperatures noted during cycling trial in the present investigation (Fig. 2A). However when corrected, the relationship between Trec, TVS and TeC during cycling became much stronger, with variations in temperature potentially stemming from the capsules being in different locations. Indeed, the differences in corrected temperatures might indicate the capsules were located differently along the gastrointestinal tract (Figs. 2D and 4D). While it is not possible to know the exact location and movement of the capsules, exercise has been shown to increase peristaltic velocity, which may alter their position along the duodenum and therefore alter the temperature variability of the telemetry devices. Notwithstanding, Domitrovich et al. studied the variation between telemetry capsules consumed 24 h and 40 min prior to 45 min of exercise and showed no differences in temperature between the 2 capsules. This suggests that starting location of the capsule (i.e., upper or lower gastrointestinal tract) may not necessarily alter the temperature responses during steady-state exercise.

The discrepancy between corrected rectal and intestinal temperatures during the recovery period in cool conditions following cycling in the heat may relate to the distinct redistribution of blood flow during each of these phases. For instance, increases in core and skin temperature during exercise-heat stress result in a constriction of central vascular beds, such as in the splanchnic circulation, resulting in a repartitioning of blood flow to the periphery to dissipate heat. This reduction in splanchnic flow occurs in proportion to the level of heat stress and exercise intensity. Given the large amount of energy required to alter tissue temperature across the intestinal tract, the discrepancy in corrected temperature during the cycling recovery phase may be due to a large return of cooled blood from peripheral vascular beds, resulting in altered temperature responses across the intestine and rectum. Accordingly, when rapidly moving from a hot/humid to cool environments, as may be necessary following the onset of heat illness symptoms, gastrointestinal temperature may significantly under predict that of the rectum.

Summary

This study investigated the validity and reliability of the e-Celsius™ temperature data logging and monitoring system from BodyCap compared to an existing temperature telemetry system and a medical precision thermistor. Results of the water bath experiment indicate a large measurement error and variability in both the e-Celsius™ and VitalSense capsules, with e-Celsius™ capsules consistently reporting slightly lower values than VitalSense. This trend was also observed during the cycling and running exercise trials. However, following a correction of the values, both telemetry devices recorded similar temperatures to that of Trec during cycling and intermittent running. Both devices reported lower temperature values during the recovery phase of the cycling trial, which likely relates to rapid changes in whole-body temperature and blood flow redistribution. It is therefore strongly recommended to calibrate ingestible telemetry devices prior to use, and to consider the possible under prediction of rectal temperature before using intestinal telemetry as a substitute for more obtrusive means of monitoring core temperature.