Development of a Resting Energy Expenditure Estimation in Patients Undergoing Targeted Temperature Management with a Surface Gel Pad Temperature Modulating Device

Abstract

Targeted temperature management (TTM) directly impacts energy expenditure via temperature modulation and shivering associated with temperature modulating devices. We hypothesized that resting energy expenditure (REE) can be accurately estimated utilizing data obtained from a surface gel pad temperature modulating device (TMD) and demographic factors. Baseline demographic data, along with concurrent temperature, sedation, and shivering data, and indirect calorimetry (IDC) were collected from patients undergoing TTM. The data from the IDC and temperature modulation device (TMD) were synchronized and averaged over 60-second intervals to provide simultaneous comparisons. Heat transfer (calories) was calculated from the TMD by an equation that assessed water temperature from the TMD to the patient, water temperature returning to the TMD, water flow rates, and device mode. A linear regression model was used to determine factors associated with REE as measured by IDC. A difference in the mean between REE and estimated REE was used to assess accuracy. There were 48 assessments conducted in 40 subjects [mean (standard deviation)] age: 58 (14) years, 60% female, body surface area (BSA): 2.0 ± 0.3 who underwent simultaneous assessments. Target temperature was 36–37°C in 75%, with a median Bedside Shivering Assessment Score of 0 (range 0–2). Factors associated with REE on multivariable linear regression included older age (p < 0.001), male sex (p = 0.004), higher BSA (p < 0.001), higher patient temperature (p < 0.001), and lower heat transfer (p = 0.003). Adjusted prediction coefficients from this model were then tested against REE by a Bland–Altman analysis. The difference between difference in resting energy estimation (REEdiff) and measured REE by IDC was 6.2 calories/min (REEdiff: 95% confidence interval: −14.1 calories, 26.5 calories, p = 0.5). We believe that the heat transfer data from the TMD coupled with clinical characteristics of patients can be utilized to calculate the REE for every minute of TTM. These data can be utilized to mitigate the consequences of shivering and malnutrition during TTM.

Introduction

Current temperature modulation devices (TMDs) provide an array of information indicating the workload necessary to achieve a specified target temperature. These data reflect the metabolic impact of shivering and/or ongoing fever. Shivering has important physiologic consequences, including increased resting energy expenditure (REE) (Badjatia et al., 2008; Hata et al., 2008), which can lead to malnutrition of patients who are critically ill.

Successful early identification of shivering facilitates targeted management and avoidance of oversedation (Choi et al., 2011). Ineffective shivering management may negate the benefits associated with cooling and increases systemic oxygen (O2) consumption (Badjatia et al., 2007; Oddo et al., 2010). Shivering in its initial state may be subtle enough that the casual observer may not even be aware of its occurrence. Moreover, shivering is episodic and can be easily missed if only assessed periodically. To best address the onset of shivering in its earliest state (i.e., microshivering), and capture short episodes, monitoring energy expenditure continuously would be ideal.

The purpose of this study was to develop a model to accurately assess energy expenditure in patients undergoing targeted temperature modulation with the Arctic Sun^® 5000 (AS5000) Temperature Management System (Bard Medical, Inc.) We hypothesized that basic clinical demographic information, coupled with heat transfer data from the AS5000, could be utilized to approximate the simultaneously collected REE assessed by indirect calorimetry (IDC).

Methods

Patient selection and data collection

This was an analysis of REE data prospectively collected as part of routine care for all patients undergoing targeted temperature management (TTM) with the AS5000 in a 22-bed Neurocritical Care Unit. Concurrent clinical physiologic, IDC, and TMD data were time synchronized and averaged over 60-second intervals, allowing for direct comparison of the TMD heat transfer and REE. Data related to temperature goal, admission diagnosis, as well as age, sex, race, height, and weight, were collected from review of the electronic medical record.

This study was conducted institutional review board (IRB) approval from the University of Maryland.

Heat transfer calculation

The AS5000 was the TMD utilized for this study. This surface gel pad TMD provides bedside clinicians with data regarding core body temperature (Tc) as well as water supply temperature (Ts), and water flow rate (Q). In addition, the device collects the temperatures of the water supplied to the surface gel pads (Ts) and the water returning (Tr) to the machine after returning from the surface gel pads to the device. The delta in Ts and Tr, as well as Q, provides the basis for calculating heat transfer.

Patient temperature data as well as the Ts (C⁰) and Tr (C⁰), and water flow rate Q (L/min) were recorded at 1-minute intervals and used to calculate heat energy transfer. The uncorrected heat transfer rate was computed as the heat difference between the amount of energy in the water leaving the TMD and the water returning to the TMD. Heat energy transfer was calculated using the following equations: $q u = 60 \times σ \times Q \times C p \times (T r - T s)$

where qu = uncorrected heat transfer rate (Kcal/h), σ = density of water (1 kg/L), Cp = heat capacity of water (1 kcal/[kg·^oC]), and 60 = conversion factor (per minute to per hour). As defined in the above equation, heat being removed from the patient has a positive value.

The heat transfer rate then was corrected for ambient heat losses. The ambient heat losses are energy lost in the water supply lines and energy lost from the surface gel pads. To measure the ambient heat loss from the fluid supply line, the AS5000 has calibrated in-line thermistor probes attached to the supply and return water ports of the AS5000. An ambient thermistor probe is also connected to the TMD. The AS5000 was operated in manual mode with a water target of 4°C and 42°C. The ambient heat loss from the fluid delivery line was computed as follows: $H l \times A l = 60 \times σ \times Q \times C p \times (T r - T s) ∕ (T r - T a)$

where Hl × Al = heat transfer coefficient normalized for surface area (Kcal/[h·°C]) and Ta = ambient temperature (°C). The value of the fluid delivery line heat loss was as follows: $h l \times A l = 3.7 K c a l ∕ h ∕^{0} C$

Energy loss from the surface gel pads has an ambient heat transfer coefficient of the following: $H p \times A p = q p ∕ (T s - T a) = 6.1 K c a l ∕ h ∕^{0} C$

where qp = pad heat loss (Kcal/h).

The corrected heat transfer rate was computed as follows:

The total energy removed from a patient was computed as the integral of the corrected heat transfer rate over time. In equation form: $C u m u l a t i v e E n e r g y T r a n s f e r = \sum q c, i \times Δ t i$

where qc,i = corrected heat transfer rate (Kcal/min) and Δ ti = time interval (1 minute).

Indirect calorimetry

IDC studies were performed for 30 minutes in each patient as part of routine assessment of nutritional status in the ICU using a Vmax Spectra (Sensormedics) to measure inspired and expired concentrations of O2 and carbon dioxide (CO2). In mechanically ventilated patients, the circuit system was connected to the O2 delivery and exhaust systems of the ventilator. In nonmechanically ventilated patients, the circuit was connected to an air-tight canopy that covered the patients' head and neck and delivered a measured constant flow of air (21% O2). Both methods allowed for continuous measurements of O2 and CO2 concentration in the inspired and expired air, allowing calculation of O2 consumption (Vo2, mL/min) and CO2 production (Vco2, mL/min).

Regular device calibration was performed according to the manufacturer's guidelines to ensure accuracy of the O2 and CO2 sensory equipment. Each IDC assessment started with establishing a steady state of a 30-minute interval during which average minute O2 consumption (Vo2) and CO2 production (Vco2) changed by <5% and <10%, respectively. Continuous measurements were averaged every 60 seconds throughout the entire IDC session. IDC studies were not performed in patients who required Fio2 ≥ 60% or who were known to be having seizures.

Statistical methods

The goal of this study was to develop a method to approximate REE as measured by IDC using data obtained from the AS5000. We utilized a Bland–Altman analysis (Bland and Altman, 1986) to compare heat transfer from the AS5000 with REE from calorimetry. The difference was then modeled with relevant cofactors of Bedside Shivering Assessment Score (BSAS), age, gender, weight, and height to develop a predictive factor for AS5000 derived assessment of REE. We believed that 40 subjects were required to adequately perform these analyses. Data from the first 20 patients were utilized to develop a predictive model that was tested on data from a subsequent 20 patients.

Results

There were 48 measurements in 40 subjects during the study period. Table 1 demonstrates the baseline characteristics of the subjects. Many of the subjects were maintained at 37°C during each IDC with a median BSAS of 0 (0–1) during each IDC measurement.

Table 1.

Baseline Characteristics of Patients

Characteristics	n = 40
Age, years	57 (14)
Female, n (%)	23 (57.5)
Diagnosis, n (%)
Cardiac arrest	8 (20)
Ischemic stroke	6 (15)
ICH	13 (32.5)
SAH	5 (12.5)
Status epilepticus	5 (12.5)
BSA	2.0 (3)
BMI	22.8 (14.4)
TTM goal
33–35°C	10 (25)
36–37°C	30 (75)

BMI, body mass index; BSA, body surface area; ICH, intracerebral hemorrhage; SAH, subarachnoid hemorrhage; TTM, targeted temperature management.

IDC data

There were 1444 minutes of IDC data recorded in all study subjects during the study period. On univariate analysis, REE (calories/min) correlated with age (r = −0.1, p < 0.001), body surface area (BSA) (r = 0.34, p < 0.001), heat transfer (r = 0.20, p < 0.001), and patient temperature (r = 0.35, p < 0.001). REE was noted to be higher in male subjects (2054 ± 599 calories/min vs. 1907 ± 513 calories/min, p < 0.001). In a multivariable linear regression model, male sex, age, BSA, heat transfer, and patient temperature were all independently associated with REE (Table 2).

Table 2.

Factors Associated with Resting Energy Expenditure

Factor	Unstandardized coefficient		Standardized coefficient beta	t-score	p
Factor	Beta	SE	Standardized coefficient beta	t-score	p
Constant	−5318.816	342.543		−15.53	<0.001
Age	−4.145	0.896	−0.098	−4.625	<0.001
Male	−77.402	25.989	−0.065	−2.978	0.003
BSA	804.234	35.918	0.476	22.391	<0.001
Core temperature	162.755	9.085	0.407	17.914	<0.001
Heat transfer	0.179	0.063	0.064	2.85	0.004

F score—173.999, p < 0.001. R² = 0.377.

SE, standard error.

AS 5000 heat transfer data

There were 1444 minutes of data recorded from the AS 5000, corresponding to each IDC session. Heat transfer data correlated with age (r = 0.15, p < 0.001), core body temperature (r = 0.466, p < 0.001), and REE (r = 0.18, p < 0.001). As expected, the correlation with To (r = −0.8, p < 0.001) and Tr (r = −0.723, p < 0.001) is strongly correlated with heat transfer.

Resting energy expenditure estimation

A Bland–Altman analysis was performed comparing the estimated REE (REEest) calculated using the adjusted predictive values from the linear regression model with REE measurements with the IDC. Additional comparisons of REE were made using commonly used BMR estimation models: American College of Chest Physicians (ACCP)(Cerra et al., 1997); Mifflin–St. Jeor (Mifflin et al., 1990); and Harris–Benedict (Bendavid et al., 2020) equations. As shown in Table 3 and Figure 1, the only equation found to be in agreement with the minute by minute REE measurement was the REEest. All other equations were noted to have a significant difference in the mean REE measurements.

FIG. 1.

Measures of agreement between REE and REEest. Bland–Altman graphs of agreement between IDC measured REE (mean REE) and (a) temperature modulating device calculated REE (REEest), (b) Harris–Benedict equation, (c) Mifflin–St. Jeor equation, (d) ACCP equation. Solid line represents mean difference. Hashed lines represent 95% CI around mean difference. ACCP, American College of Chest Physicians; CI, confidence interval; IDC, indirect calorimetry; REE, resting energy expenditure; REEest, estimated REE.

Table 3.

Bland–Altman Analysis of Resting Energy Expenditure

Measurements	REE equations
Measurements	REEest	ACCP	Harris–Benedict	Mifflin–St. Jeor
REE range (kcal)	1040–2853	1123–3458	1241–3047	1055–2735
Difference mean ± SE	6.2 (10.3)	−250.5 (16.4)	−54.6 (13.9)	111.8 (87.9)
Significance	0.55	0.001	0.001	0.012

ACCP, American College of Chest Physicians; REE, resting energy expenditure; REEest, estimated REE.

Discussion

In this prospective cohort study, we found that age, sex, BSA, core body temperature, and heat energy transfer from the AS 5000 were all independently associated with IDC measured REE. Furthermore, the developed REEest equation based on the prediction coefficients from our linear regression model reliably assessed IDC energy expenditure measurements. This REEest model was more accurate than traditionally utilized equations for estimating REE in this population.

Understanding a patient's REE in the critical care setting is important to accurately provide nutritional support and avoid the sequelae of malnutrition (Moonen et al., 2021). Measurement of REE by IDC is considered the gold standard but is cumbersome and often not performed given numerous estimating equations have been shown to accurately predict REE (Wichansawakun et al., 2015). However, none of the existing equations considers the impact that TTM may have on REE.

TTM fundamentally impacts REE given the dynamic, continuous heat transfer from TMDs that is necessary to achieve a targeted core body temperature. Heat transfer from a TMD is influenced by the patient's age, sex, BSA, the delta between core temperature, and targeted temperature goal. Clinically, the BSAS measurement (Badjatia et al., 2008) and water temperature display have been considered surrogates of the amount of heat transfer during TTM (Madden et al., 2017). The BSAS is often only recorded hourly, and therefore limited in its ability to measure more continuous changes. Water temperature can be influenced by factors unrelated to heat transfer (e.g., amount of BSA coverage by surface gel pads, number of surface gel pads being utilized).

Heat transfer, by itself, does not give a true accounting of REE, as it does not account for additional influential factors, such as age, sex, and BSA. The REEest equation we propose in this study overcomes many of these limitations by combining heat transfer data with clinical characteristics such as age, sex, BSA, and core body temperature to provide a continuous measure of REE.

There are significant implications of having continuous REE measurement in patients undergoing TTM. First, it provides clinicians with an accurate assessment of the effect TTM may be having on the patient's metabolic state. This has been the traditional role of shiver assessment tools such as the BSAS or bispectral index (BIS) monitor (May et al., 2018). However, the BSAS is intermittently measured and only provides a snapshot of the impact of shivering. The BIS monitor has only been shown to be effective in patients receiving neuromuscular blockade (NMB) (May et al., 2018), and it is not established whether this monitor would be effective in discerning shivering from patient movement. None of the subjects in this study received NMBs.

Having a continuous shivering detection monitor will allow for more accurate and timely intervention of shivering and may result in overall lower use of sedatives. A targeted approach toward shivering that emphasizes early intervention based upon mild symptoms is a good method to avoid moderate to deep sedation in TTM patients (Choi et al., 2011). A more accurate measurement of REE will also help avoid over/underfeeding, which may occur in patients undergoing TTM. This may especially be a significant concern in patients undergoing hypothermia, where gastric resorption may influence nutritional delivery (Dobak and Rincon, 2016). Patients requiring prolonged normothermia treatment may also benefit from more accurate nutritional delivery and mitigate the sequelae of malnutrition.

There are limitations to our analysis worth consideration. The primary limitations are our relatively small sample size for patients and depths of TTM. We recognize that this limits the conclusiveness of our findings, but maintains that the approach outlined in our study can be easily replicated and should be done across multiple institutions and various depths and durations of TTM. In addition, we did not analyze sedative usage, which clearly influences the REE. However, this was intentional, given our primary goal was to identify a method by which to reliably estimate energy expenditure, regardless of sedation, given that nonintubated, nonsedated patients also undergo TTM for prolonged periods.

In conclusion we believe that REE can be accurately measured continuously in patients undergoing TTM with the AS5000 by heat transfer measurements coupled with core body temperature, age, sex, and BSA. Future multicenter studies should be performed to validate our preliminary findings across the depth and duration of TTM goals.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

Funding Information

The conduct of the study was funded by Becton-Dickinson.

References

Badjatia

, Kowalski

, Schmidt

, et al. Predictors and clinical implications of shivering during therapeutic normothermia. Neurocrit Care, 2007; 6:186–191.

Badjatia

, Strongilis

, Gordon

, et al. Metabolic impact of shivering during therapeutic temperature modulation: the Bedside Shivering Assessment Scale. Stroke, 2008; 39:3242–3247.

Bendavid

, Lobo

, Barazzoni

, et al. The centenary of the Harris-Benedict equations: how to assess energy requirements best? Recommendations from the ESPEN expert group. Clin Nutr, 2020; 40:690–701.

Bland

, Altman

. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 1986; 1:307–310.

Cerra

, Benitez

, Blackburn

, et al. Applied nutrition in ICU patients. A consensus statement of the American College of Chest Physicians. Chest, 1997; 111:769–778.

Choi

, Ko

, Presciutti

, et al. Prevention of shivering during therapeutic temperature modulation: the Columbia anti-shivering protocol. Neurocrit Care, 2011; 14:389–394.

Dobak

, Rincon

. Enteral feeding during induced hypothermia after intracerebral hemorrhage. Neurology, 2016; 86.

Hata

, Shelsky

, Hindman

, et al. A prospective, observational clinical trial of fever reduction to reduce systemic oxygen consumption in the setting of acute brain injury. Neurocrit Care, 2008; 9:37–44.

Madden

, Hill

, May

, et al. The implementation of targeted temperature management: an evidence-based guideline from the Neurocritical Care Society. Neurocrit Care, 2017; 27:468–487.

10.

May

, Riker

, Gagnon

, et al. Continuous surface EMG power reflects the metabolic cost of shivering during targeted temperature management after cardiac arrest. Resuscitation, 2018; 131:8–13.

11.

Mifflin

, St Jeor

, Hill

, et al. A new predictive equation for resting energy expenditure in healthy individuals. Am J Clin Nutr, 1990; 51:241–247.

12.

Moonen

HPFX

, Beckers

KJH

, van Zanten

ARH

. Energy expenditure and indirect calorimetry in critical illness and convalescence: current evidence and practical considerations. J Intensive Care, 2021; 9:8.

13.

Oddo

, Frangos

, Maloney-Wilensky

, et al. Effect of shivering on brain tissue oxygenation during induced normothermia in patients with severe brain injury. Neurocrit Care, 2010; 12:10–16.

14.

Wichansawakun

, Meddings

, Alberda

, et al. Energy requirements and the use of predictive equations versus indirect calorimetry in critically ill patients. Appl Physiol Nutr Metab, 2015; 40:207–210.