Umbilical temperature correlation with core and skin temperatures at rest, in the heat and during physical activity

Abstract

Purpose: to determine the correlation of umbilical temperatures (T_umb) with simultaneously recorded chest wall temperature (T_chest) and rectal temperature (T_rectal) in adults during rest, heat exposure and exercise.

Methods: A total of 28 healthy men, wearing different types of clothing (athletic garb, a spandex full body heating garment, firefighter bunker gear) had average and peak umbilical, chest wall and rectal temperature measurements taken during sedentary temperature stabilisation stages, heat exposure periods and active exercise phases.

Results: Curvilinear relationships were noted between T_chest and T_umb compared with T_rectal and their association became noticeably positive and linear at approximately 35.5 °C. Polynomial regression analysis of T_rectal with linear and quadratic forms of T_chest and T_umb indicated an overall R² of 0.657 and 0.767, respectively. Bivariate analysis of a restricted data set (where T_chest and T_umb ≥35.5°), indicated that T_umb was significantly associated with T_rectal (r_average = 0.710, p <0.001; r_peak  = 0.841, p <0.001) and T_chest was also significantly associated with T_rectal, but less so (r_average = 0.570, p <0.001; r_peak = 0.699, p <0.001).

Conclusions: the umbilicus offers a non-invasive, peripheral site for measurement of temperature that more closely correlated with body core temperature than T_chest when core temperature was ≥35.5 °C.

Keywords:

• Temperature
• umbilicus
• chest
• rectal
• correlation

Introduction

Core temperature (T_core) measurement is utilised extensively in medical evaluations and research studies as it is the single best indicator of the body’s thermal status [1]. T_core (i.e. intracranial, deep thoracic, oesophageal, intraabdominal, rectal) reflects the temperature of the related anatomic region’s internal milieu. Drawbacks to the use of T_core measurements are that they are invasive, oftentimes uncomfortable, difficult to safely maintain inserted and associated with hygiene issues [2,3]. The need for a simple, non-invasive method to measure T_core is evident [2], and the use of skin temperature as an index of T_core is an attractive notion, but requires that the former be a reliable method of monitoring the latter [4]. The splanchnic abdominal organs (excluding the kidneys) of a resting human produce 33% of the body’s heat, though accounting for only 3.8% of body mass [5]. Intraabdominal T_core is generally measured with an ingested telemetric sensing capsule [6], but this method is hampered by the possibility of temperature gradients along the gastrointestinal tract, acute modifying effects of fluid and food ingestion on T_core and the uncertainty of sensor transit time [7]. Skin temperature is a function of measurement depth [4] and body orifices and depressions (e.g. external auditory canal, medial canthus of the eye, nares) are consistently warmer than flat skin surfaces because of cross-radiation of heat from their opposing skin surfaces and reduced air current effects [8,9], as well as closer proximity to internal heat sources. The umbilicus (navel) is a skin depression that is regularly identified as the warmest area of the abdominal wall on infra-red imaging studies [9–13], suggesting that it may offer a window into intraabdominal T_core. This assumption is plausible because of its anatomic features and attachments to various intraabdominal organs that may serve as thermal conduits (Table 1) [8,9,13–22]. A study of afebrile infants and children, reporting insulated umbilical temperatures (T_umb) comparable with oral and rectal readings [23], suggests that T_umb could possibly serve as a surrogate index [4] for T_core, but adult data is scarce. This study was undertaken by the National Personal Protective Technology Laboratory of the National Institute for Occupational Safety and Health (NIOSH) to determine the relationship of T_umb with other concurrent body temperature readings in adults at rest, during heat exposure and while involved in exercise. Such data could be important for researchers and those individuals who are engaged in activities where physical stress, environmental factors, or the use of encapsulating protective clothing can result in significant elevations of T_core.

Table 1. Anatomical rationale for the relationship of umbilical temperature with intraabdominal temperature.

In non-overweight persons, the umbilicus is thinner (men, 8.1 ± 4.8 mm; women, 10.6 ± 4.9 mm) [21] than the rest of the abdominal wall, as there is little fat or muscle associated with its inner surface to insulate it from core heat (fat is a poor conductor and good insulator of heat) [14,15].

The inner surface of the umbilicus is free and separated from the abdomen only by the parietal peritoneum [16] that is tightly adherent to the umbilical ring [17] and porous so that it can transmit heat radiating from the peritoneal cavity.

As a body cavity, the umbilicus allows for cross-radiation from its opposing walls, with little heat loss, so that its surface temperatures more closely approximate core temperatures [8,9,11].

The central depression of the umbilicus moderates the effects of air currents upon its skin temperature (dependent on umbilical depth) [8].

The umbilical cord remnants (hepatic ligamentum teres, medial umbilical ligaments, urachus), attached proximally to the underside of the umbilicus and distally to intraabdominal organs, are vascularised [18–20] and can transmit intraabdominal heat along a vascular temperature gradient [11].

Materials and methods

The study data was collected from subjects in three ongoing, but unrelated, NIOSH studies conducted under different conditions of clothing, ambient environments and work rates. The three studies were approved by the NIOSH Institutional Review Board and written informed consent was obtained from subjects that allowed for the inclusion of material pertaining to themselves, acknowledgement that they cannot be identified via this article, and that they have been fully anonymised. All subjects were healthy, non-smokers who were medically screened by a licenced physician prior to study entry. Each individual’s testing was completed on one laboratory visit and a physician was present during subject testing for safety purposes. Subjects (n = 28) for the current report were categorised into three groups (A, B and C) based upon the clothing worn for the studies. Group anthropometrics (±SD) and clothing are described, as follows:

Group A: 11 men (age 21.6 ± 0.9 years, height 186 ± 6.9 cm, weight 79.2 ± 6.8 kg, body mass index [BMI] 23.0 ± 2.0 kg/m²) who wore athletic shoes, shorts and tee shirts.

Group B: 7 men (age 25.6 ± 5.4 years, height 179 ± 9.1 cm, weight 80.1 ± 8.5 kg, BMI 25.0 ± 1.7 kg/m²) wearing a hooded, full body, spandex garment with internal tubing that circulates warm water for heating (Figure 1) and athletic shoes.

Figure 1. Heating garment utilised by Group B.

Group C: 10 men (age 23.4 ± 2.7 years, height 181 ± 9.2 cm, weight 79.54 ± 6.1 kg, BMI 26 ± 2.7 kg/m²) wearing shorts and a tee shirt under firefighter full bunker gear consisting of pants, jacket, hood, gloves, boots and helmet (Morning Pride Manufacturing Co., Dayton, OH), and a self-contained breathing apparatus (SCBA) with full facepiece respirator (the SCBA was not activated during the study due to its 40 min usage capacity that would not have been sufficient to carry out the study tasks).

Baseline temperature measurements (Study phase I) – Prior to measurement of rectal (T_rectal), anterior chest wall (T_chest) and T_umb baseline temperatures, stabilisation of body temperature occurred with subjects seated for 15 min in an environmental chamber at the following ambient conditions: 25 °C/50% relative humidity (RH) (Group A), 25 °C/50% RH (Group B), and 30 °C/70% RH (Group C). Group B’s heating garment was not activated during these baseline measurements. Group C subjects did not wear the SCBA tank during baseline measurements, but did wear the full facepiece respirator, without a regulator.

Heat exposure temperature measurements (Study phase II) – Subjects were seated for the same temperature measurements as obtained in the baseline studies. Group A subjects underwent passive heating seated in an environmental chamber (40 °C, 30% RH) for 30 min. Group B subjects sat in environmental chamber (40 °C, 50% RH) while wearing athletic pants and a sweatshirt over the full body heating garment that was being actively infused with 46 °C water until the subject reached a T_rectal 0.5 °C above baseline (heating phase averaged 65.8 ± 5.7 min to complete). Group C subjects sat in an environmental chamber (30 °C, 70% RH) for 15 min (SCBA tank was not worn during this phase).

Exercise temperature measurements (Study phase III) – During exercise, subjects had the same temperature measurements as obtained in the baseline studies. Group A subjects pedalled a cycle ergometer, at an initial resistance of 75 watts, to volitional fatigue (average of 26.9 ± 5.1 min to termination) in environmental chamber conditions of 40 °C and 50% RH. Group B subjects pedalled a cycle ergometer, at an initial resistance of 75 watts increasing by 25 watts every 10 min, to volitional fatigue (average of 15.4 ± 3.2 min to termination, in environmental chamber conditions of 40 °C and 50% RH (the hooded sweatshirt and pants were removed and the heating garment was not activated during this phase). Group C subjects (wearing the SCBA, but breathing through the respirator without a regulator) treadmill exercised at 40% VO₂ max for 40 min in environmental chamber conditions of 30 °C and 70% RH.

Measurement equipment – T_rectal measurements were obtained using a Precision 401 rectal thermistor probe (YSI Temperature, Dayton, OH) with a 4.7 mm tip that was inserted 13 cm into the rectum. T_chest measurements used a Precision 409 b skin thermistor (Grant Industries, Surrey, UK) with a 9.5 mm tip applied to the right anterior chest wall at the second intercostal space region and insulated with a folded sterile cotton gauze pad covered by a moisture-and-air-permeable, transparent adhesive dressing (Tegaderm^™, 3 M Company, St. Paul, MN). T_umb was measured with a Precision 401 rectal thermistor (Grant Industries), housed in a cone-shaped, pliable silicone shell that conformed to umbilical dimensions (Figure 2). The thermistor was attached with a waterproof, transparent adhesive dressing (Tegaderm) and insulated with a self-adherent neoprene patch (5.5 cm diameter, 3 mm thickness) applied over the adhesive dressing.

Figure 2. Umbilicus temperature in a silicone sleeve.

Statistical analysis – Two different forms of T_rectal, T_chest and T_umb were used in the analysis: (1) the average of the temperature taken throughout the study, and (2) the peak temperature reached throughout the course of the study. Through the varied experimental conditions and corresponding phases, a sufficient range in the variables was obtained.

The associations were visually explored through scatter plots between T_rectal, T_chest and T_umb. Visual inspection of the scatter plots revealed the appearance of a curvilinear relationship between each of the forms of the less invasive temperature measurements and T_rectal (Figure 3). The curvilinear relationships were verified through four distinct hierarchical polynomial regressions (R²) in which T_rectal was regressed on both the linear and quadratic forms of T_chest and T_umb while controlling for age, height, BMI and weight. These hierarchical regressions took the form of:

Figure 3. Average T_rectal regressed on linear and quadratic average T_umb (overall R² 0.583).

Consistent with a hierarchical regression approach, the control variables were entered in model 1, model 2 includes the addition of the linear form of the independent variable, and model 3 includes the addition of the quadratic form of the independent variable.

Hierarchical polynomial regressions were repeated for average T_chest, average T_umb, peak T_chest and peak T_umb, and the total R² (the coefficient of determination, a statistical measure of how close the data are to the fitted regression line), change in R² (ΔR²), unstandardised regression coefficients (B), and significance levels were noted for each of the corresponding steps.

Results

Subject anthropometrics and descriptive statistics for variables by study and phase are presented in Table 2. For the first set of regressions, in which the average T_rectal was modelled as the dependent variable, age was the only significant control variable (p = 0.006) (Table 2). There was a significant increase in R² when the linear form of the average T_umb was added in model 2 (ΔR² =  0.242, p <0.001). There was also a significant increase in R² when the quadratic form of average T_umb was added (ΔR² =  0.224, p <0.001). The overall R² for the final model in which the average T_umb was used as the predictor was 0.583 (Table 3; Figure 3).

Table 2. Subject characteristics and descriptive statistics by study and phase.

Characteristics	Study A	Study B	Study C
Number of subjects	11	7	10
Age (years)	21.6 ± 0.9, 20–23	25.6 ± 5.4, 21–36	23.4 ± 2.7, 21–29
Weight (kg)	79.2 ± 6.8, 68–89	80.1 ± 8.5, 62–89	79.5 ± 6.1, 67–86
Height (cm)	86 ± 6.9, 174–196	179 ± 9.1, 165–193	181 ± 9.2, 165–196
BMI (kg/m^2)	23 ± 2.0, 20–26	25 ± 1.7, 23–27	26 ± 2.7, 22–30
Variables
Study phase I
Avg. rectal Temp.	36.9 ± 0.3, 36.5–37.3	36.9 ± 0.1, 36.7–37.1	37.0 ± 0.3, 36.5–37.4
Avg. chest Temp.	32.7 ± 0.8, 31.6–34.6	33.9 ± 0.8, 32.6–34.8	34.0 ± 1.6, 31.5–36.0
Avg. umbilical Temp.	33.4 ± 0.6, 32.4–34.5	34.5 ± 1.2, 32.5–35.7	34.6 ± 0.8, 33.3–35.7
Peak rectal Temp.	37.0 ± 0.3, 36.6–37.3	36.9 ± 0.2, 36.7–37.1	37.0 ± 0.2, 36.6–37.5
Peak chest Temp.	33.1 ± 0.8, 32.2–34.9	34.4 ± 0.6, 33.1–35.0	34.3 ± 1.5, 31.8–36.2
Peak umbilical Temp.	33.7 ± 0.6, 32.6–34.8	34.6 ± 1.2, 32.9–35.8	34.8 ± 0.9, 33.3–36.1
Study phase II
Avg. rectal Temp.	36.8 ± 0.3, 36.2–37.2	37.0 ± 0.2, 36.8–37.2	37.0 ± 0.3, 36.5–37.4
Avg. chest Temp.	35.1 ± 0.4, 34.5–35.9	36.3 ± 1.1, 33.7–36.8	35.1 ± 0.9, 33.3–36.3
Avg. umbilical Temp.	35.5 ± 0.6, 34.4–36.2	36.2 ± 0.5, 35.6–37.0	35.4 ± 0.8, 33.8–36.4
Peak rectal Temp.	36.8 ± 0.3, 36.3–37.2	37.3 ± 0.2, 37.2–37.5	37.0 ± 0.2, 36.5–37.4
Peak chest Temp.	35.4 ± 0.4, 34.6–36.3	37.4 ± 0.3, 37.1–38.0	35.6 ± 0.9, 33.9–36.5
Peak umbilical Temp.	35.8 ± 0.6, 34.5–36.5	37.2 ± 0.2, 37.0–37.5	35.9 ± 0.7, 34.4–36.6
Study phase III
Avg. rectal Temp.	37.0 ± 0.3, 36.6–37.5	37.7 ± 0.2, 37.5–37.9	37.5 ± 0.3, 37.1–38.0
Avg. chest Temp.	35.9 ± 0.2, 35.5–36.3	36.7 ± 0.3, 36.1–37.2	36.5 ± 0.4, 36.0–37.0
Avg. umbilical Temp.	36.1 ± 0.5, 35.2–36.6	36.9 ± 0.3, 36.6–37.3	37.1 ± 0.3, 36.5–37.1
Peak rectal Temp.	37.2 ± 0.2, 36.9–37.6	37.9 ± 0.2, 37.7–38.2	38.4 ± 0.4, 37.5–39.0
Peak chest Temp.	36.1 ± 0.3, 35.7–36.6	36.9 ± 0.2, 36.7–37.3	37.5 ± 0.5, 36.4–38.0
Peak umbilical Temp.	36.4 ± 0.5, 35.6–36.9	36.6 ± 0.3, 36.6–37.4	38.0 ± 0.4, 37.3–38.6
Overall
Avg. rectal Temp.	36.9 ± 0.3, 36.2–37.5	37.2 ± 0.4, 36.7–37.9	37.2 ± 0.4, 36.5–38.0
Avg. chest Temp.	34.6 ± 1.5, 31.6–36.3	35.6 ± 1.5, 32.6–37.2	35.2 ± 1.5, 31.5–37.0
Avg. umbilical Temp.	35.0 ± 1.3, 32.4–36.6	35.9 ± 1.3, 32.5–37.3	35.7 ± 1.2, 33.3–37.6
Peak rectal Temp.	37.0 ± 0.3, 36.3–37.6	37.4 ± 0.4, 36.7–38.2	37.5 ± 0.7, 36.5–39.0
Peak chest Temp.	34.9 ± 1.4, 32.2–36.6	36.2 ± 1.4, 33.1–38.0	35.8 ± 1.7, 31.8–38.0
Peak umbilical Temp.	35.3 ± 1.3, 32.6–36.9	36.3 ± 1.4, 32.9–37.5	36.2 ± 1.5, 33.3–38.6

For age, height, BMI and variables the Mean ± SD and range are reported. Temperatures are in °C.

Table 3. Results for hierarchical polynomial regressions.

	ΔR^2	R^2	Adj. R^2	Β	Std Err.	t	p
Average rectal temperature as dependent variable
Model 1	0.116	0.116	0.072				0.042
Age				0.032	0.011	2.83	0.006
Height				0.037	0.038	0.96	0.339
Weight				−0.040	0.045	−0.88	0.381
BMI				0.140	0.143	0.98	0.331
Average umbilical temperature
Model 2	0.242	0.359	0.318				<0.001
Average umbilical Temp				0.138	0.025	5.43	<0.001
Model 3	0.224	0.583	0.551				<0.001
Average umbilical Temp^2				0.095	0.015	6.44	<0.001
Average chest temperature
Model 2	0.236	0.353	0.311				<0.001
Average Chest Temp				0.121	0.023	5.33	<0.001
Model 3	0.124	0.477	0.436				<0.001
Average Chest Temp^2				0.058	0.014	4.28	<0.001
Peak rectal temperature as dependent variable
Model 1	0.078	0.078	0.032				0.163
Age				0.029	0.018	1.62	0.109
Height				0.089	0.059	1.49	0.139
Weight				−0.107	0.070	−1.53	0.130
BMI				0.358	0.222	1.61	0.110
Peak umbilical
Model 2	0.425	0.503	0.471				<0.001
Peak umbilical Temp				0.249	0.030	8.17	<0.001
Model 3	0.264	0.767	0.749				<0.001
Peak umbilical Temp^2				0.120	0.013	9.34	<0.001
Peak chest
Model 2	0.402	0.480	0.447				<0.001
Peak chest Temp				0.225	0.029	7.76	<0.001
Model 3	0.177	0.657	0.631				0.014
Peak chest Temp^2				0.083	0.013	6.31	0.014

B: unstandardised regression coefficients reported; Std Err.: is the standard error of the regression coefficient; t: t-test for regression coefficient; p: significance level of the t test for the unstandardised regression coefficient.

When the linear form of average T_chest was added to the control variables, there was a significant increase in R² (ΔR² =  0.236, p <0.001). There was an additional significant increase in R² when the quadratic form of average T_chest (ΔR² =  0.124, p <0.001). The overall R² for the final model in which the average T_chest was used as the predictor was 0.477 (Table 3; Figure 4).

Figure 4. Average T_rectal regressed on linear and quadratic average T_chest (overall R² 0.477).

For the second set of regressions, in which the peak T_rectal was modelled as the dependent variable, none of the control variables significantly predicted the outcome. There was a significant increase in R² when the linear form of peak T_umb was added to the model (ΔR² =  0.425, p <0.001). There was an additional significant increase in R² when the quadratic form of peak T_umb was added in model 3 (ΔR² =  0.264, p <0.001). The overall R² for the final model in which peak T_umb was used as the predictor was 0.767 (Table 3; Figure 5).

Figure 5. Peak T_rectal regressed on linear and quadratic average T_umb (overall R² 0.767).

The linear form of peak T_chest also significantly increased the R² when added to the model (ΔR² =  0.402, p <0.001). There was an additional significant increase in R² when the quadratic form of peak T_chest was added to the model (ΔR² =  0.177, p = 0.014). The overall R² for the final model in which peak T_chest was used as the predictor was 0.657 (Table 3; Figure 6).

Figure 6. Peak T_rectal regressed on linear and quadratic peak T_chest (overall R² 0.657).

Discussion

As is indicated by the overall model R², T_umb explains more of the variance in T_rectal when both the average and peak temperatures are considered. These analyses suggest that T_umb shares a stronger association with T_rectal than the T_chest counterpart. It seems prudent, however, to further consider the application and to explore the contexts that optimise the practical utility of the use of the T_umb as a less invasive surrogate method of temperature measurement.

As can be observed from Figures 3–6, the association between each of the predictors and the forms of T_rectal noticeably becomes positive and linear at the approximate point of 35.5 °C. In order to explore the strength of the linear association at this point, bivariate correlations were computed between each of the predictors and the appropriate form of T_rectal in the restricted dataset (where T_umb and T_chest ≥35.5 °C [thereby excluding most of phase I trials]). These correlations are reported in Table 4.

Table 4. Correlations between rectal temperature with umbilical and chest temperature in restricted dataset.

	Average rectal temperature	Peak rectal temperature
Average umbilical temperature	0.710*
Average chest temperature	0.570*
Peak umbilical temperature		0.841*
Peak chest temperature		0.699*

Restricted dataset includes all observations in which both chest and umbilical temperature measurement are greater than or equal to 35.5 °C.

* , p <0.001.

Consistent with the results of the polynomial regressions, T_umb displayed higher bivariate correlations with T_rectal when compared to T_chest across the types of average and peak variables. Also consistent with the results of the polynomial regressions, the associations are higher in the peak variables when compared to the average temperatures. T_umb was significantly associated with T_rectal (r_average = 0.710, p <0.001; r_peak = 0.841, p <0.001). T_chest was also significantly associated with T_rectal, but less so (r_average = 0.570, p <0.001; r_peak = 0.699, p <0.001) (Table 4).

It has been stated that the temperature of an organ depends partially upon its depth below the skin and its anatomical relationship to the abdominal cavity [4,24]. The umbilicus serves as an appropriate example that, due to its thinness relative to the abdominal wall [21] and its attachment to the peritoneum [17], likely functions as a conduit for intraperitoneal heat. Although T_chest and T_umb are technically both skin temperature measurements, the generally higher T_umb values in this study indicate that additional heat was being dissipated across the umbilicus, the additional source being heat from the peritoneal cavity. Further, the umbilicus itself is relatively avascular and thus retains heat better than more vascular structures [25]. The amount of heat flowing to the surface from the peritoneal cavity is a function of the temperature gradient between the surface of the body and the heat-producing intraperitoneal organs, as well as the overall conductivity of the human body [26]. Principal sources of body heat in the resting individual are intraperitoneal, chiefly from the liver and, to a lesser extent, the intestines [24]. One study (75 healthy subjects), utilising inserted thermocouples, reported a morning mean liver temperature of 36.68 ± 0.03 °C and corresponding T_rectal of 36.89 ± 0.04 °C in a temperate ambient environment [27]. This data aligns with findings from another human study (11 healthy subjects) that noted a liver temperature averaging 0.44 °C below T_rectal under similar ambient conditions [28]. However, liver temperature does not equate precisely with intraperitoneal temperature in the basal state because the latter will be lower due to a peritoneal cavity temperature gradient across the abdominal wall (in normal weight individuals) and diaphragm [28,29]. Data on direct intraperitoneal temperature measurements are limited, but a study of gasless laparoscopy (thereby negating the effects of pressurised CO₂ gas on intraabdominal temperatures) reported a mean intraabdominal temperature of 36.46 ± 0.56 °C [30]. Based on available data, it appears that intraperitoneal temperature may be ∼0.2 °C less than liver temperature that is itself 0.2–0.6 °C subrectal temperature [31]. Further, although the umbilicus itself is collagenous scar tissue and relatively avascular, arterial blood flow to the umbilical skin from subdermal blood vessels and perforating vessels of the deep inferior epigastric artery [22] may attenuate some of the heat transfer to the umbilicus from the peritoneal cavity in the basal state. Also, insulated skin sites have been shown to be less effective in the thermoneutral range than in detecting the onset of heat strain [32]. An exception to this would be in persons with umbilical hernias wherein the T_umb would more directly reflect intraintestinal temperature [9]. Thus, the temperature offset between T_rectal and intraperitoneal temperature must be taken into consideration when evaluating T_umb.

T_umb values generally exceeded T_chest in this study due to the minimum air current effects and minimal radiant heat exchange on the umbilicus (dependent on umbilical depth to some degree) [9] and the umbilicus’ aforementioned intimate anatomic relationships with the peritoneal cavity and its contents (Table 1). Additionally, sensor insulation been shown to be important to prevent heat loss from the skin [33,34], even when wearing protective clothing as in study Group C [3]. However, this was not entirely the case as noted during phase II (heating) studies, when use of the heating garment resulted in T_chest that actually exceeded T_rectal (Table 1), a recognised phenomenon that occurs when the skin is warmed by an outside heat source that directly contacts the temperature sensors [34]. Of added note, during the exercise phase of study Group C, the peak and mean T_umb values were within 0.5 °C of T_rectal, a level of accuracy required between skin site and T_rectal to be considered clinically utilitarian (though this value has not been firmly established) [1,35]. This result was clearly impacted by the relatively impermeable, encapsulating firefighter ensemble worn that, combined with heat and exercise, resulted in a hot, humid microclimate conducive to the convergence of skin and core temperatures [4,36].

Fox and Solman [37] first introduced the concept of zero heat flux, wherein an insulated, heated skin surface sensor equilibrates skin temperature with deep body temperature creating a region of zero heat flow that allows for measurement of core temperature at the skin surface. An early study [38] at various ambient conditions, noted that there was good agreement between a heated sensor at the sternum with tympanic membrane and ingested sensor temperatures in warmer ambient temperatures at decreased walking speeds. Ball et al. [39] noted a mean difference of 0.1 ± 0.5 °C between a sternal zero heat flux sensor and T_rectal. Studies comparing a forehead zero heat flux sensor with T_rectal reported a mean difference of 0.9 ± 0.4 °C [40] and 0.12 ± 0.24 °C [41]. One recent study, examining the relationship between core temperature measured by ingested core temperature sensor, skin temperature measured by ceramic heat flow sensor and heat flux of subjects who treadmill walked in different ambient environments reported R² values of 0.40 for the forehead and 0.75 for the sternum [42]. A similar study by the same authors reported an R² of 0.70 for the sternum compared with T_rectal [43]. Thus, in some studies [42,43], the sternum is associated with the highest R² values that compared to T_rectal and align with the reported values for T_umb data from this study. The use of a “double sensor” (one probe adjacent to the skin and separated by a standard insulator from a second probe facing the environment) has been investigated in several studies [44–46]. Kimberger et al. [44] compared a forehead “double sensor” with oesophageal temperatures in perioperative and intensive care patients; 98% of measurements were within 0.5 °C of oesophageal values with a mean bias of 0.08 °C and limits of agreement of −0.66–0.50 °C between methods [44]. Gunga et al. [45] compared forehead “double sensor” readings with T_rectal over 36 h of bedrest and reported r = 0.704 with 0.08 ± 0.32 °C difference between methods. Another study [46], comparing T_rectal and a helmet-mounted forehead “double sensor” at 25–55% VO₂Max for 2 + h each at ambient temperatures of 10, 25 and 40 °C reported correlations of 0.49, 0.69 and 0.75, respectively. This study’s finding, that use of the umbilicus as a temperature measurement site correlates best with T_rectal when T_umb ≥35.5 °C, implies that this parameter might be most useful in monitoring situations that are associated with heat stress and hyperthermia. T_umb might be especially useful in situations where moderate-to-high physical activity is associated with elevated ambient heat and the use of restrictive clothing (e.g. firefighters, military personnel, foundry workers, etc.).

Limitations of this study include the relatively low number of subjects tested in each group and the fact that no women subjects were tested so that we cannot comment on the possible effect of gender on T_umb. We did not analyse T_umb in relation to umbilicus depth, so there exists the possibility of deeper umbilici transmitting higher temperatures [9]. Further, we did not test obese individuals whose panniculus could have served as an abdominal wall insulator [14]. Skin thickness at measurement sites (e.g. forehead skin 1.81 mm) impacts temperature measurements and, although the anterior chest wall skin thickness (1.37 mm) is less than that of the abdominal skin covering the umbilicus (1.91 mm), the underlying chest muscle layer likely would increase the functional thickness of the chest wall [47,48]. An equipment limitation observed occasionally during testing was that, due to the relatively small contact area of the T_umb sensor (4.7 mm), motion-related displacement during exercise activities, with resultant loss of contact between the umbilicus and the sensor, necessitated repeat testing. This can occur with even minor movement of the sensor [34]. In an attempt to circumvent this limitation, pilot testing using a wireless semiconductor temperature sensor with a larger surface area (I-Button, Dallas, TX) was carried out, but unsuccessful due to the sensor’s diameter (16.5 mm) exceeding the internal dimensions of the umbilicus. A similar wireless sensor, with smaller dimensions, might offer a reasonable solution to minimising T_umb sensor displacements.

Finding a skin site that correlates closely with core temperature has been the subject of much research interest for a number of years because of the varied issues associated with core temperature measurements (invasiveness, discomfort, etc.). Use of insulated skin temperature sensors, as in this study, results in closer correlations between skin and core temperatures than uninsulated sensors during exercise and heat exposure states and can allow for the development of prediction equations [3,31]. The umbilicus offers a temperature measurement site that, with the use of an appropriately-sized wireless sensor, could offer an unobtrusive site for temperature measurements that would not interfere with an individual’s ongoing activities. Further, in studies that might employ an externally applied heat source (as with this study’s heating garment), a small sensor housed in the recesses of the umbilical depression might be shielded from the effects of direct contact with a heating source. An appropriately-sized, wireless T_umb sensor could be programmed to deliver a vibratory or audible alarm to the wearer that, upon reaching a pre-programmed temperature setting, would indicate the need for heat remediation strategies to be undertaken. The umbilicus offers the possibility of a skin measurement site that correlates reasonably well with core temperature when T_rectal is ≥35.5 °C. Future studies are needed to address the overall utility of T_umb, as well as for validation of this study’s findings.

Conclusions

This study has demonstrated that, when taking average and peak temperatures into consideration, T_umb correlated well with T_rectal compared with T_chest. The umbilicus, due to its anatomical features and relationships (Table 1), offers a non-invasive, peripheral site for measurement of temperature that correlates reasonably well with core temperature when T_rectal is ≥35.5 °C. Development of an insulated, wireless skin temperature sensor, that fits snuggly within the recesses of the umbilicus, may offer a non-invasive temperature measurement capability that supplements current methods of temperature measurement.

Acknowledgements

The authors thank W. Jon Williams, PhD, Lewis Radonovich Jr., MD, and Michael Bergman, MS, of NIOSH for their manuscript reviews and comments.

Disclosure statement

The authors report no declarations of interest.

The findings and conclusions in this report are those of the author(s) and do not necessarily represent the views of the National institute for Occupational Safety and Health. Mention of a product or use of a photo does not constitute NIOSH endorsement.

Additional information

Funding

This study was conducted with internal operating funds of the U.S. national Personal Protective Technology Laboratory where all authors are employed.