Predictive Energy Equations Inaccurately Estimate Metabolic Demands of Older Adult Trauma Patients

Abstract

Introduction:

Suboptimal nutrition promotes unfavorable outcomes in trauma patients, particularly among those aged 60 and over. While many institutions employ predictive energy equations to determine patients’ energy requirements, mounting evidence shows these equations inaccurately estimate caloric needs. In this pilot randomized controlled trial, we sought to quantify the discrepancy between predictive equations and indirect calorimetry (IC)—the gold standard for determining energy requirements—in the older adult trauma population.

Methods:

This is a nested cohort study within a pilot randomized control trial in which 32 older adult trauma patients were randomized 3:1 to receive IC-guided nutrition delivery versus standard of care. IC requirements of patients in the intervention arm were compared to Mifflin St. Jeor (MSJ), Harris-Benedict (HB), and the American Society for Parenteral and Enteral Nutrition-Society of Critical Care Medicine (ASPEN-SCCM) predictive energy equations.

Results:

Twenty patients underwent IC to assess measured resting energy expenditure (mREE), yielding a mean (standard deviation) mREE of 23.1 ± 4.8 kcal/kg/d. MSJ and HB gave mean predictive resting energy expenditures of 17.5 ± 2.0 and 18.5 ± 2.0 kcal/kg/d in these patients, demonstrating that IC-derived values were 32.1% and 25.0% higher, respectively. When patients were stratified by body mass index (BMI), MSJ, and HB more severely underestimated caloric requirements in individuals with BMI <30 versus BMI 30–50. While the mean mREE fell within the mean predictive resting energy expenditure range prescribed by ASPEN-SCCM equations (21.4 ± 4.1 to 26.2 ± 4.3 kcal/kg/d), individuals’ IC-derived values fell within their personal range in 8 of 20 cases.

Conclusions:

The MSJ and HB predictive energy equations consistently and significantly underpredict metabolic demands of older adult trauma patients compared to IC and perform worse in lower BMI individuals. ASPEN-SCCM equations frequently overpredict or underpredict resting energy expenditure. While these findings should be confirmed in a larger randomized control trial, this study suggests that institutions should prioritize IC to accurately identify the metabolic demands of older trauma patients.

Introduction

Predictive resting energy expenditure (pREE) equations have been utilized since the early 20th century to estimate human basal metabolic rate (BMR).¹ While the Harris-Benedict equation was the first, several iterations of this and other equations have become well-known over the last hundred years. However, more recently, the inaccuracies of these predictive equations have also been identified by those who study and care for malnourished patients. The early pREE equations use anthropomorphic data points, including gender, height, weight, and age, to estimate caloric requirements. Only after the creation of the Penn State equation in the early 21st century were other important contributors to BMR, including temperature and minute ventilation, included.²

Although some data suggest improvement in the accuracy of more updated pREE equations from less than 20% to greater than 60%, there remain deficiencies in accurately identifying patient energy expenditure, particularly in the critically ill.³ Calorimetry is the most accurate way to determine the caloric needs of patients. While direct calorimetry measures the caloric expenditure of an individual through the production of heat by resting metabolism, indirect calorimetry (IC) extrapolates upon this concept by measuring the concentrations of oxygen and carbon dioxide through respiration to compute measured resting energy expenditure (mREE) and respiratory quotient.³ IC is currently considered the gold standard in determining caloric needs in critically ill patients at the bedside, and its use has been strongly recommended by recent European Society for Clinical Nutrition and Metabolism and American Society for Parenteral and Enteral Nutrition (ASPEN) guidelines.^4–9 Nevertheless, its adoption across America is minimal.

The BMRs of traumatically injured and critically ill patients are variable and differ from those who are uninjured.³ This presents a unique challenge regarding providing adequate nutrition in both the inpatient and postdischarge settings. Suboptimal nutrition can complicate hospitalization and adversely affect functional status and overall recovery. Most institutions routinely utilize predictive energy expenditure equations to estimate the nutritional needs of patients, yet several studies over the last few decades have shown that this leads to inaccurate estimation and delivery of nutrition.^10–12 Specifically, as it relates to the older adult population—defined by the World Health Organization as adults over 60 y of age—the metabolic needs of the injured patient are complex. It is known that aging leads to a gradual decline in resting energy expenditure (REE) and BMR, with whole-body resting metabolic rate declining by 1%−2% per decade after 20 y of age, primarily attributed to decreases in metabolically active fat-free mass.^13–16 However, debate persists as to whether this decline is solely attributable to changes in body composition, as adjustments for fat-free mass fail to fully explain the observed differences in resting metabolic rate between young and older adults. Nevertheless, to our knowledge, the performance of predictive equations in accurately estimating energy expenditure has yet to be assessed in the traumatically injured older adult population.¹⁷

In our study, we aimed to bridge this existing knowledge gap by hypothesizing that traditional pREE equations significantly underestimate the resting energy needs of older adult trauma patients relative to IC. To test this hypothesis, we conducted IC measurements to ascertain mREE in this patient cohort and compared these values to the results of traditional pREE equations. Our objective was to assess the accuracy of standard pREE equations against IC in predicting precise nutritional requirements, aiming to enhance the clinical management of nutritional support in these patients.

Materials and Methods

Study population and design

This is a nested cohort study within The StructurEd Nutrition Delivery Home pilot randomized controlled trial registered as “Personalized Nutrition Delivery to Improve Resilience in Older Adult Trauma Patients” on ClinicalTrials.gov (NCT05544162), which was designed to evaluate a precision nutrition pathway against the standard of care in nutrition delivery for older adults with trauma. As part of this pathway, the intervention utilized IC to establish personalized caloric needs, starting at 72-h postadmission and repeated at approximately 5-d intervals. Once deemed appropriate for a full liquid diet, patients received tailored nutrition interventions, including oral nutrition supplements.

Eligible participants were adult trauma patients aged over 60, anticipated to survive their hospital stay, and able to tolerate oral nutrition. Exclusion criteria included patients expected to have life-sustaining treatment withdrawn, those with traumatic brain injury, prisoners, or those unable to provide consent. The study received ethical approval from the Duke Health Institutional Review Board in Durham, North Carolina (Pro00110867). Following informed consent, patients were allocated to the intervention or control group within 2 d of admission, using a 3:1 computer-generated randomization scheme to maintain balance between groups.

Measurements and study variables

Demographic and anthropometric data were collected at the time of intensive care unit (ICU) admission, while injury assessment data were abstracted from the electronic medical record upon hospital discharge. mREE data were obtained for patients in the intervention group using the Q-NRG Metabolic Monitor (COSMED Rome, Italy). Before all measurements, the calorimeter’s internal blower, turbine flowmeter, and gas analyzers were calibrated using test gases (ambient air and calibration gas 5% CO₂, 16% O₂, balanced with N₂).

Patients were temporarily excluded from IC assessment under the following conditions: FiO2 >70%, hemodynamic instability, positive end-expiratory pressure > 16 mmHg, inability to tolerate IC (i.e., nausea, claustrophobia), or per ICU attending clinical judgment. IC data were selected from at least 5-min intervals that met steady-state conditions, defined by a variance of oxygen consumption and carbon dioxide production < 10% as per published validation data for the Q-NRG device. Measurements not meeting these criteria were excluded from the final analysis.

Calculations of pREE were performed on all patients at 72 h after admission, including patients in the control group to help establish an expected range for pREE measurements in this population. Three commonly utilized pREE equations were chosen for comparative analysis to IC measurements: Mifflin St. Jeor (MSJ), Harris-Benedict (HB), and the ASPEN/Society of Critical Care Medicine (ASPEN-SCCM) Clinical Guidelines for the Provision and Assessment of Nutrition Support Therapy in the Adult Critically Ill Patient (Table 1). Anthropometric and clinical data used for pREE equations were correlated with admission height and weight recorded in the medical record. All pREE calculations were performed according to patient sex and age.

Table 1 –

Predictive energy equations.

MSJ

Men: pREE = 10 × weight + 6.25 × height − 5 × age + 5

Women: pREE = 10 × weight + 6.25 × height − 5 × age − 161

HB

Men: pREE = 66.47 + 13.75 × weight + 5.0 × height − 6.75 × age

Women: pREE = 665.09 + 9.56 × weight + 1.84 × height − 4.67 × age

ASPEN/SCCM

ASPEN-SCCM Lower-End of Range

BMI <30 kg/m2: pREE = 25 × weight

BMI 30 – 50 kg/m2: pREE = 16.9 × weight

ASPEN-SCCM Upper-End of Range

BMI <30 kg/m2: pREE = 30 × weight

BMI 30 – 50 kg/m2: pREE = 21.5 × weight

All units of measure are as follows: height: cm; weight: kg; age: y.

Statistical analysis

Descriptive statistics were used to report demographic and injury assessment data. Final REE data in kcal/kg/day were calculated by dividing pREE (as determined by MSJ, HB, or ASPEN-SCCM) or mREE(asdeterminedby IC) bykg of weight at 72 hpostadmission. For the intervention group, mREE as a percentage of pREE was obtained by dividing mREE by pREE values for all three predictive energy equations. These data were further stratified by separating patients in the intervention group into BMI categories specified by the ASPEN-SCCM predictive energy equations: BMI <30 kg/m² and BMI 30–50 kg/m². Continuous data were evaluated for distribution normality using Kolmogorov–Smirnov tests and by checking normality visually with Q-Q plots. Approximately normally distributed continuous data were compared using paired t-tests and reported as mean ± standard deviation. Non-normally distributed continuous data were compared using Mann–Whitney U tests and reported as median ± interquartile range. In all instances, a two-sided P value of <0.05 determined significance. Three patients in the intervention group who could not tolerate IC due to claustrophobia or low oxygenation were excluded from the analysis.

Results

Between February and November 2023, our pilot study successfully enrolled 32 patients. During this period, we encountered one unexpected mortality but no withdrawals of consent. Demographic specifics of the entire study population along with injury severity score and hospital length of stay are detailed in Table 2, and are further stratified by ASPEN-SCCM BMI category (BMI <30 kg/m² and BMI 30–50 kg/m²) in those receiving IC in Table 3. Demographic variables, injury severity scores, and hospital lengths of stay were balanced between both intervention and control groups and between BMI categories.

Table 2 –

Demographics, injury severity score, and hospital length of stay of intervention and control groups.

Variables	Intervention (n = 23)	Control (n = 9)	P Value
Age (years)
Median ± IQR	69 ± 10	73 ± 6	0.19
Range	61–94	68–90
Sex (%)			>0.10
Female	12 (52)	7 (78)
Male	11 (48)	2 (22)
BMI (kg/m2)
Mean ± SD	29.48 ± 6.62	29.09 ± 8.42	0.89
Range	16.52–43.89	19.78–46.84
Race (%)			>0.10
Black	5 (22)	1 (11)
White	18 (78)	7 (78)
Not reported	0 (0)	1 (11)
Ethnicity (%)			>0.10
Hispanic/Latinx	1 (4)	0 (0)
Non-hispanic/Latinx	21 (91)	8 (89)
Not reported	1 (4)	1 (11)
Injury severity score
Median ± IQR	10 ± 4	9 ± 5	0.23
Range	2–22	4–17
Hospital length of stay			0.87
Median ± IQR	6 ± 5	6 ± 3
Range	2–24	3–20

IQR = interquartile range; SD = standard deviation.

Table 3 –

Demographics of patients in the intervention group stratified by ASPEN-SCCM BMI category.

Variables	BMI <30 kg/m2 (n = 12)	BMI 30–50 kg/m2 (n = 11)	P Value
Age (y)
Median ± IQR	71 ± 13	66 ± 8	0.29
Range	61–94	61–85
Gender (%)			>0.50
Female	6 (50)	6 (55)
Male	6 (50)	5 (45)
BMI (kg/m2)
Median ± IQR	24.21 ± 2.73	33.41 ± 2.24	<0.0001
Range	16.52–28.42	30.41–43.89
Race (%)			>0.50
Black	2 (17)	3 (27)
White	10 (83)	8 (73)
Ethnicity (%)			>0.10
Hispanic/Latinx	1 (8)	0 (0)
Non-hispanic/Latinx	11 (92)	10 (91)
Not reported	0 (0)	1 (9)
Injury severity score
Mean ± SD	10.9 ± 5.2	14.2 ± 4.6	0.13
Range	2–22	9–22
Hospital length of stay			0.95
Median ± IQR	6.5 ± 5	5 ± 7
Range	2–12	3–24

IQR = interquartile range; SD = standard deviation.

Demographically, the control and intervention groups were majority female with an average age over 70 y old and statistically similar along age, sex, and BMI. Racial and ethnic representation in the intervention group was diverse. The study population included a wide range of injury severity scores and hospital lengths of stay balanced between both groups.

The nature of the injuries among participants was predominantly blunt trauma, with only one individual in the intervention group experiencing a penetrating injury. The blunt injuries were categorized by cause, including falls, vehicle-related accidents, and crush incidents, and were further classified by anatomic location: chest, abdomen, neck, back, upper limbs, pelvic region, and lower limbs. The majority of these injuries involved skeletal damage, primarily fractures of the manubrium and ribs (both unilateral and bilateral), with some cases presenting with pneumothorax or hemothorax. Additionally, there were instances of spinal, scapular, upper and lower limb, and pelvic fractures. Notably, two patients in the intervention group sustained blunt trauma to solid abdominal organs. Table 4 provides an extensive overview of the injuries sustained by the study participants.

Table 4 –

Summative assessment of injuries sustained by study participants.

Injury (percent)	Intervention (n = 23)	Control (n = 9)
Type
Blunt	22 (95.7)	9 (100.0)
Penetrating	1 (4.3)	0 (0.0)
Mechanism
Fall	11 (47.8)	4 (44.4)
Vehicular accident	12 (52.2)	4 (44.4)
Crush	0 (0.0)	1 (11.1)
Injury location
Chest	21 (91.3)	9 (100.0)
Sternal/Manubrial fracture	3 (13.0)	6 (66.7)
Unilateral rib fractures	13 (56.5)	6 (66.7)
Bilateral rib fractures	7 (30.4)	2 (22.2)
Number of rib fractures
One	2 (8.7)	0 (0.0)
Two	2 (8.7)	2 (22.2)
Three	2 (8.7)	0 (0.0)
Four	6 (26.1)	2 (22.2)
Five	1 (4.3)	1 (11.1)
Six or more	7 (30.4)	3 (33.3)
Hemothorax/Pneumothorax	6 (26.1)	4 (44.4)
Abdomen	2 (8.7)	0 (0.0)
Neck/Back	10 (43.5)	3 (33.3)
Scapular fracture	5 (21.7)	0 (0.0)
Spinal fracture	5 (21.7)	3 (33.3)
Location of spinal fractures
Cervical	1 (4.3)	0 (0.0)
Thoracic	0 (0.0)	2 (22.2)
Cervical & thoracic	1 (4.3)	0 (0.0)
Lumbar	3 (13.0)	1 (11.1)
Upper extremity	2 (8.7)	0 (0.0)
Pelvis	1 (4.3)	0 (0.0)
Lower extremity	2 (8.7)	1 (11.1)

The results of pREE calculations for the 20 intervention group patients who tolerated IC using the MSJ, HB, and ASPEN-SCCM equations are shown in Table 5 alongside the mean mREE determined by IC for this group. The MSJ and HB predictive equations consistently underestimated metabolic requirements, with the difference between mREE and pREE reaching statistical significance. mREE values represented 132.1 ± 25.4% and 125.0 ± 22.2% of the pREE determined by MSJ and HB, respectively. When patients were stratified by BMI group, the mean mREE and pREE gap, as determined by MSJ and HB, widened in the BMI <30 group (Table 5). Intolerance to IC in the intervention arm either by face mask or canopy hood was secondary to claustrophobia or inadequate oxygenation.

Table 5 –

pREE and IC-derived mREE in older adult trauma patients.

Study arm	REE (kcal/kg/d)					P value*
	MSJ	HB	ASPEN-SCCM LER	ASPEN/SCCM UER	mREE	MSJ	HB	ASPEN-SCCM LER	ASPEN-SCCM UER
Control (n = 9)	17.1 ± 2.1	18.4 ± 2.1	22.3 ± 4.0	27.2 ±4.3	-	-	-	-	-
Intervention (n = 20)	17.5 ± 2.0	18.5 ± 2.0	21.4 ± 4.1	26.2 ± 4.3	23.1 ± 4.8	<0.0001	<0.0001	0.06	<0.01
BMI <30 (n = 11)	18.6 ± 2.0	19.6 ± 1.9	25.0	30.0	25.7 ± 4.4	<0.001	<0.001	0.61	<0.01
30 < BMI <50 (n = 9)	16.2 ± 0.9	17.1 ± 1.2	16.9	21.5	20.0 ± 3.2	<0.01	<0.05	<0.05	0.20

Study arm	mREE % of pREE
	MSJ	HB	ASPEN-SCCM LER	ASPEN-SCCM UER
Intervention (n = 20)	132.1 ± 25.4	125.0 ± 22.2	109.8 ± 19.5	89.0 ± 14.9
BMI <30 (n = 11)	139.1 ± 27.3	131.5 ± 22.2	102.8 ± 17.6	85.6 ± 14.6
30 < BMI <50 (n = 9)	123.6 ± 21.2	117.1 ± 20.6	118.4 ± 19.1	93.1 ± 15.0

LER = lower end of range; UER = upper end of range.

Data are means ± standard deviations.

Kilograms reflect patients’ admission weight. Three patients in the intervention group who could not tolerate IC due to claustrophobia or low oxygenation were excluded from the analysis.

^*P values were calculated relative to mREE as determined by IC.

Mean mREE in both BMI categories fell between the mean lower- and upper-end of the range calculated using the ASPEN-SCCM predictive equations. Significantly, however, in only eight cases did an individual’s mREE fall within the personalized range prescribed by these equations. Notably, the lower end of this range correlated more closely with actual mREE in the BMI <30 groups, while the upper end of this range correlated more closely with actual mREE in the BMI between 30 and 50 groups.

Discussion

This prospective pilot study demonstrates that the metabolic requirements of older adult trauma patients are consistently and significantly underestimated by predictive equations frequently used in the ICU compared to IC. As the gold standard for REE estimation, IC was found to yield REE values that were 32.1% and 25.0% higher than those estimated by the commonly used MSJ and HB equations, respectively. The degree to which predictive equations underestimate metabolic requirements may also be exacerbated depending on patient BMI, and this study demonstrates that older adult patients with lower BMI recovering from traumatic injury may be more prone to these effects.

The ASPEN-SCCM guidelines propose a range for estimating metabolic requirements that might more closely resemble the actual mREE than the MSJ or HB equations. However, this study has demonstrated that even these ranges are not consistently reliable, as mREE fell within the calculated range for only a minority of patients who underwent IC. This finding casts doubt on the sufficiency of the ASPEN-SCCM guidelines to accurately guide nutritional interventions in the clinical setting. Furthering the narrative established by previous research, this study reinforces the notion that predictive formulas, once standard practice for calculating energy expenditure, fail to deliver clinically relevant estimates across different patient populations, including those affected by stroke, COVID-19, and critical illnesses. Such inaccuracies underline the critical necessity for individualized metabolic measurements to inform precise nutritional support and improve outcomes in these groups.^18–20

IC is the most precise, noninvasive method to measure patients’ energy expenditures, with universal guideline recommendations calling for the use of IC to determine energy requirements and facilitate nutrition delivery in the ICU. While IC was once untenable to incorporate into routine clinical practice, a new generation of indirect calorimeters permits accurate, reliable, user-friendly energy expenditure measurements.^21,22 Importantly, the Q-NRG Metabolic Monitor utilizes comfortable and lightweight equipment while allowing patients to remain in their own hospital rooms, which is critical for older adults recovering from traumatic injuries who may have difficulty with mobility and physical reserve. Indeed, 20 out of 23 patients offered IC successfully tolerated the measurement, while two declined due to pre-existing nausea, and one was ineligible due to inadequate oxygenation.

In the field of trauma care, significant progress has been made in reducing mortality rates within ICU settings. Consequently, major national ICU/trauma trials groups are now advocating for a shift in focus toward functional quality of life as a primary endpoint for future trials.¹³ One of the key factors affecting recovery is the onset of a catabolic state due to trauma and critical illness, which leads to substantial muscle mass loss and impaired muscle function, particularly during the hospital stay.^23–26 This acute catabolic response significantly increases the risk of malnutrition and once malnutrition sets in, the likelihood of complications rises markedly, resulting in poorer outcomes such as increased mortality, further complications, and chronic illnesses—all of which contribute to escalating health-care costs.^27–30

This issue is exacerbated in older adults (the term preferred by the National Institute on Aging over “geriatric” or “advanced age”) who are particularly vulnerable to trauma, with a high percentage presenting with or at risk for malnutrition upon admission to the hospital.^31,32 For these patients, malnutrition is often worsened by routine classifications of “nothing by mouth” due to physiological complications or the need for repeated procedures. Subsequently, their ICU stays typically involve significant underfeeding with patients receiving about 50% of predicted needs, and post-ICU care often fails to recognize a hypermetabolic state, leading to even poorer nutrition delivery. The consequences of malnutrition in older trauma patients are severe, including a six-fold increase in mortality, greater morbidity, and a reduced likelihood of being discharged home.^33–35

Despite the well-documented benefits of effective, goal-directed nutrition in improving outcomes and resilience in ICU patients, the nutritional needs of older adult trauma patients, particularly in the post-ICU phase, remain poorly addressed.⁵ This period often lasts longer than the initial ICU stay, with caloric requirements significantly increasing as the body attempts to recover.^36,37 Yet, there is an alarming trend of patients consistently failing to meet these increased nutritional needs at a time when adequate nutrition is crucial for regaining physical function.³⁸

Our pilot study brings attention to the often-underestimated metabolic needs of the older trauma population. The predictive energy equations (pREE) currently used tend to contribute to consistent underfeeding. Hence, we recommend avoiding pREE when determining caloric needs for the older, traumatically injured population. In the absence of IC, adopting the lower end of the ASPEN-SCCM range for patients with a BMI <30 and the upper range for those with a BMI between 30 and 50 has shown a better correlation with IC-derived mREE in this patient group.

Accurately determining metabolic demands is critical, especially considering the complex nutritional needs arising from baseline individual variability and the added stress of acute traumatic injuries. This is particularly relevant for older adults prone to increased frailty, susceptibility to refeeding syndrome, and decreased physiologic reserve, underscoring the urgency for optimal nutrition delivery tailored to their specific requirements.

Limitations

As an initial pilot randomized controlled trial, this study inherently faces certain limitations, primarily related to its exploratory nature. The most significant constraint is the small sample size, which, while sufficient for preliminary observations, may not fully capture the variability and complexities of nutritional needs in the traumatically injured, older adult population. Consequently, the results should be interpreted cautiously, as they may not provide the statistical power necessary to detect subtle differences or generalize the findings to a larger, more diverse population.

A strength of this study is its comparison of nutrition delivery guided by predictive resting energy equations to the gold standard of IC. While the use of pREE is commonplace due to its ease and practicality, especially in settings where IC is not readily available, it may not accurately reflect the true metabolic demands of severely injured patients. Therefore, the inability to correlate pREE with actual IC measurements may lead to potential deficiencies in nutritional delivery and could impact clinical outcomes. These potential discrepancies highlight the need for larger, more definitive trials to establish a solid correlation between the methods of nutritional assessment and their impact on recovery and long-term outcomes.

Conclusions

This pilot study has provided valuable insights into the metabolic requirements of older adults following trauma, highlighting the shortcomings of commonly used predictive energy equations in accurately estimating the caloric needs of this population. The study has demonstrated that these equations frequently underestimate older adults’ nutritional needs, particularly those with a BMI less than 30. Our research strongly suggests the adoption of individualized metabolic assessments, such as IC, to determine the true caloric demands is necessary for optimal recovery. The evident gap between predicted and measured REE underscores the imperative to revise current nutritional care practices in geriatric trauma patients to prevent undernutrition and support resilience. These findings serve as an essential step toward advancing nutritional care strategies, aiming to foster improved recovery processes and quality of life in older trauma survivors.