Predictive Equations Overestimate Resting Metabolic Rate in Chronic Stroke Survivors

Abstract

Objective:

To measure resting metabolic rate (RMR) in chronic (>3 months prior) stroke survivors (aged 61±7.5 yrs, X±SEM, mean ±standard error of the mean) and to compare to predicted RMR using predictive equations in healthy adults.

Design:

Cross-sectional study

Setting:

Hospital

Participants:

Stroke survivors

Intervention:

Not applicable

Main Outcome Measures:

RMR was measured by indirect calorimetry. Participants underwent a total body dual-energy x-ray absorptiometry (DXA) scan and treadmill test for peak oxygen consumption (VO₂ peak). RMR was calculated using nine established equations.

Results:

RMR measured (1552±319 kcal/day) was significantly lower than nine predicted RMR values (all P<0.001) with the best being McArdle-Katch (1652±233 kcal/day), Livingston (1677±230 kcal/day), and Mifflin (1707±242 kcal/day). The IMNA (2437±386 kcal/day) had the largest discrepancy with measured RMR. Predicted RMR determined with eight of nine equations was between 9% and 18% greater than measured RMR. Appendicular lean mass (r=0.64, P<0.001), total lean mass (r=0.64, P<0.001), and VO₂ peak (r=0.41, P<0.001) were associated with measured RMR.

Conclusion:

RMR predictive equations established in healthy adults are not appropriate for the chronic stroke population, indicating the need to measure RMR until a more accurate predictive equation is developed. This could support modifications to nutritional intake guidelines in patients with conditions of muscle atrophy. If measurement of RMR is not feasible, the Katch-McArdle equation should be used to estimate RMR in a stroke patient, as, on average, it provides the lowest percentage overestimate compared to other equations.

Introduction

Resting metabolic rate (RMR), defined as the energetic need for the maintenance of basic vital processes.¹ RMR is responsible for ~60–80% of 24-hour energy expenditure and can be attributed to body mass, age, gender, and genetics.² It is important to note that RMR declines markedly with advancing age not only in sedentary populations,³ but also declines with age even in highly physically active men; the decline in RMR, adjusted for FFM, in active men is related to reductions in exercise volume and/or energy intake,³ suggesting that changes in physical activity and diet influence age-related changes in RMR.

Chronic stroke survivors have distinct physical changes following stroke. Specifically, significant muscle atrophy^{4, 5} and sarcopenia⁶ have been documented, as well as reduced cardiorespiratory fitness⁷ and mobility.⁸ Stroke survivors typically have lower physical activity levels than non-stroke healthy adults.⁹ It is reasonable to hypothesize that the combination of these biological and behavioral factors could contribute to changes in RMR in this patient population. Similarly, in a systematic review of RMR in patients with chronic spinal cord injury, who likewise undergo physical changes following injury, measured RMR was lower than predicted in seven studies, while higher than predicted in two studies.¹⁰ In a small study of eight hemiparetic patients,¹¹ RMR was higher in hemiparetic patients than controls and higher in younger patients in the sample, which contradicts other RMR and aging studies. Thus, equations predicting RMR may not accurately apply to some patients with chronic injury, in light of the physical changes undergone by these populations.

Although RMR can be determined by direct calorimetry, more often it is measured by indirect calorimetry through gas collection measurements in a thermo-neutral environment.¹⁰ Because of the need for specialized equipment and testing parameters to accurately measure RMR, equations were developed to predict RMR for nutritional purposes and field research. Indeed, prediction equations for RMR include the Harris-Benedict formula,¹² first developed over a 100 years ago and then later modified.¹³ Numerous other prediction equations are available, most of which utilize body weight, height, and age^{2, 12–17} and some of which utilize lean body mass^{18, 19} in the calculations. We²⁰ and Finestone et al.²¹ presented findings that RMR in chronic stroke patients was lower than predicted from equations established in studies with healthy adults. However, no studies in a chronic stroke population have examined this potential discrepancy utilizing predictive equations that were established after 1990, or that used variables other than weight, height, and age. Therefore, the objectives of this paper were to determine resting metabolic rate in a larger sample of chronic stroke survivors and compare measured RMR to established equations that utilize relevant variables, including sedentary behaviors and lean body mass. We hypothesize that RMR will be overestimated in chronic stroke patients using prediction equations, given reduced cardiorespiratory fitness,⁷ higher fat content and lower muscle mass,^{5, 22} and reduced ambulatory fitness in chronic stroke populations⁹ as these measures typically contribute to reduced RMR values.

Methods

Participant Selection

Stroke survivors from the Baltimore/Washington DC area (n= 71, 56 males, 15 females) participated in the study from 2011–2020. Inclusion criteria included age >30 and <85 years, any body mass index, (BMI), having a stroke more than three months prior, completion of all conventional inpatient and outpatient physical therapy, adequate language and neurocognitive function to participate in informed consent and testing, under the care of a primary care medical provider, lived independently in the community, and able to walk six minutes without assistance or with their usual assistive device(s) including single point cane or quad cane without human assistance. All participants underwent a medical history and physical examination, ECG, blood chemistries and counts, and were screened to ensure adequate informed consent. Exclusion criteria included those who performed regular structured resistive exercise of ~1 hr. more than two times per week, alcohol consumption .3 oz. liquor, 3 × 4 oz glasses of wine, or 3 × 12 oz beers/day by self-report, or had a neurological history of dementia or aphasia operationally defined as incapacity to follow 2-point commands were excluded, as were those who had recent hospitalization (less than 3 months prior to study entry) for severe medical disease, orthopedic or chronic pain condition restricting exercise, a medical history of pulmonary or renal failure, active cancer, untreated poorly controlled hypertension measured on at least 2 occasions (greater than 190/100), untreated and/or poorly controlled diabetes with fasting blood glucose of >170 and HbA1c > 10.0, cardiac history of unstable angina, recent (less than 3 months prior to study entry) myocardial infarction, congestive heart failure (NYHA category II-IV), peripheral arterial occlusive disease with vascular claudication, or orthopedic or chronic pain conditions. All volunteers signed the Institutional Review Board on record and the Research & Development Committee approved informed consent forms.

Body Composition

Standing height (cm) was measured on a metric stadiometer to the nearest cm. Body weight (kg) was determined on a calibrated scale with the participant wearing light clothing and without shoes. Participants underwent a whole body dual-energy x-ray absorptiometry (DXA) scan (iDXA LUNAR GE Radiation Corp, Madison, WI) performed by one technologist who is certified by the American Registry of Radiologic Technologist in Bone Densitometry (AART) in the Baltimore VA Radiology Department as described by LUNAR. The DXA was calibrated each day prior to scans according to the manufacturer. Participants were positioned on the table lying on his/her back with arms by their sides with palms facing downwards but not touching their thighs and with a small gap between the feet. They were asked to keep as still as possible and breathe normally. A series of transverse scans was made from head to toe. The scans were used to determine fat mass, lean tissue mass, appendicular lean mass (ALM) (sum of arm and leg lean mass), and percent body fat.

Energy Expenditure

Resting metabolic rate was measured via indirect calorimetry using COSMED (Quark PFT Ergo) with the measurement of breath by breath gas exchange measurements of oxygen consumption (VO₂) and carbon dioxide production (VCO₂) with other ventilatory parameters. Participants reported to the lab after a 12 hour fast, including refraining from caffeine and nicotine. They rested ~10 minutes before the measurement was conducted. Measurements were performed for 30 minutes as participants reclined in a supine position beneath a flow-dilution canopy hood in a comfortable, dimly lit, quiet thermo-neutral environment. The Quark RMR metabolic cart was calibrated before each test and data was filtered every 30 seconds. Energy expenditure was calculated by the Weir equation²³ and expressed per 24 hours. These conditions utilized for measurement of RMR have been associated with the most accurate measurements of RMR.²⁴

Of a multitude of predictive equations in the literature to estimate RMR in adults, nine predictive equations were chosen based on the sample sizes greater than 100 and variables used in their equations including age, weight, height, lean body mass, and the diversity of population studied in terms of gender and race (Table 1). Equations were not included if they were developed in small samples (n<100) of special populations (e.g. athletes). The majority of predictive equations were based on the inclusion of healthy adults; however, two predictive equations included obese adults either enrolled in a weight loss program²or in a cross-sectional evaluation of obesity with cardiovascular disease.¹⁶ Of these nine equations, two required measurement of lean body mass and one required an estimate of physical activity levels. Participants in our study were selected based on a ‘sedentary’ lifestyle and therefore, the physical activity coefficient (PAL) of the Institute of Medicine of the National Academies (IMNA) equation¹⁷ was considered as 1 or “sedentary.”

Table 1.

Selective RMR Predictive Equations with the Population and Sample Size

Equation Name	Population	Sample Size	Equations
Harris & Benedict (1919)12	good health 21–70 years old	136 men, 103 women	Males: 66.47+[13.75×weight (kg)]+[5×height (cm)]−[6.76×age (years)] Females: 655.1+[9.563×weight (kg)]+[1.850×height (cm)]−[4.676×age (years)]
Revised Harris & Benedict (Roza & Shizgal, 1984)13	good health wide age range	168 men, 169 women	Males: 88.362+[13.397×weight (kg)]+[4.799×height (cm)]−[5.677×age (years)] Females: 447.593+[9.247×weight (kg)]+[3.098×height (cm)]−[4.33×age (years)]
Schofield (1985)14	data from 114 published studies	3500+ men, 1200+ women	M 30–59 y.o.: 11.472×weight(kg)+873.1 M≥60 y.o.: 11.711×weight(kg)+587.7 F 30–59 y.o.: 8.126×weight(kg)+845.6 F≥60 y.o.: 9.082×weight(kg)+658.5
Owen et al. (1986)15	good health lean and obese	60 men, 44 women	Males: 879 + [10.2×weight (kg)] Females: 795 + [7.18×weight (kg)]
Mifflin-St, Jeor et al. (1990)16	ages 20+ and different weights, enrolled in a 5-year investigation of the relationship of energy expenditure and obesity to cardiovascular disease risk	498 participants	Males: [9.99×weight (kg)]+[6.25×height (cm)]-[4.92×age (years)]+5 Females: [9.99×weight (kg)]+[6.25×height (cm)]−[4.92×age (years)]−161
Cunningham (1991)18	Good health, omitting trained athletes	120 men, 103 women	500 + [22×Lean Body Mass (kg)]
Katch-McArdle19	same sample as in Cunningham equation	200+ participants	370+[21.6×Lean Body Mass (kg)]
Institute of Medicine of the National Academies (IMNA) (2002)17	≥ 19 years old	169 men, 238 women	Males: 662−[9.53×age (years)]+PA×[15.91×weight (kg)]+[539.6×height (m)] Females: 354 − [6.91×age(years)]+PA×[9.36×weight (kg) + 726×height (m)]
Livingston et al. (2005)2	same sample as original Harris-Benedict and Owen databases, + patients in weight loss	327 participants	Males: 293×weight0.4330 −(5.92×age) Females: 248 × weight0.4356 − (5.09 × age)

PA = physical activity coefficient.

Fitness and Functional Tests

Exercise testing with open circuit spirometry was conducted to measure VO₂ peak using a graded treadmill test which was supervised by a clinician.²⁵ Participants walked on a treadmill for the initial two minutes without an incline, followed by two minutes at 4% incline, with the incline advanced 2% every two minutes thereafter. Heart rate and rhythm was monitored throughout the test by electrocardiogram. Blood pressure was measured at each stage during the test. Handrail support, a belt safety harness and/or a spotter behind the participant was used to decrease the risk of falling. Tests were terminated at the participants’ request or if cardiovascular decompensation was observed. Participants underwent a six-minute walk test (6MWD), which measures ambulation distance and represents what may be required for community-based activities of daily living (ADL) and is a measure of gait endurance. Participants were instructed to “cover as much distance as possible” over a flat 100-foot surface marked by traffic cones during a 6-min time period. Assistive devices (e.g. walker, cane) were used by participants when necessary. Standard commands and instructions were used during these walking assessments. Due to conflicts in scheduling, or other non-health related factors, there is missing data for VO₂peak (n=2) and 6MWD (n=19).

Statistical Analyses

Descriptive statistics including mean, standard deviation, and range were analyzed using SPSS (PASW Statistics, Version 22, Chicago, IL). To visually assess the agreement between two measures on an individual level (measured versus predicted RMR), Bland-Altman analyses were conducted and the difference between each pair of measurements was plotted against the mean of the two measurements. Paired t-tests were conducted to test differences between measured and predicted RMR and independent t-tests for differences by race. For each of the nine calculated RMRs, the minimum number of subjects needed to determine whether the calculated and measured RMRs were significantly different was calculated using the paired differences between the RMRs for each subject. The paired differences were analyzed in a one-sample, two-tailed sample size calculation performed using the R function power.t.test (R stats package, version 4.1.0, R Core Team (2021), Vienna, Austria, assuming a significance 0.05 and power of 0.80. Pearson correlational tests were performed between VO₂peak, RMR, ALM and total lean mass.

Results

Seventy-one adults participated in the study, of whom 43 participants were African American (61%), 27 were Caucasian (38%), and one was self-reported as ‘other.’ Participant characteristics are reported in Table 2. DXA data is missing for one participant due to scheduling conflict. Measured RMR did not differ between African American and Caucasian participants (1525 ± 284 kcal/day vs. 1624 ± 367 kcal/day, p=0.16); therefore, the data was examined in the total group.

Table 2.

Participant Characteristics

	N	X ± SD	Range
Age (yrs)	71	61 ± 8	44 – 76
Weight (kg)	71	94.5 ± 19.7	52.4 – 133.8
BMI (kg/m2)	71	31.3 ± 6.7	19.0 – 48.2
RMR (kcal/day)	71	1552 ± 319	918 – 2771
Arm Lean Mass (kg)	70	7.2 ± 2.0	3.4 – 12.1
Leg Lean Mass (kg)	70	19.5 ± 4.2	9.8 – 28.7
Appendicular Lean Mass (kg)	70	26.7 ± 6.0	13.4 – 39.5
Total Lean Mass (kg)	70	56.3 ± 10.3	33.7 – 79.2
VO2peak(L/min)	69	1.7 ± 0.5	0.9 – 3.4
VO2peak (mL/kg/min)	69	18.5 ± 4.6	9.3 – 32.6
6 Minute Walk Distance (m)	52	301 ± 111	133 – 1540
Time since stroke (months)	66	57 ± 64	4 – 372

BMI: body mass index, RMR: resting metabolic rate, VO₂peak: peak oxygen consumption

Sample size calculations indicate that the minimum number of subjects were met for each predictive equation (Cunningham n= 10, Katch-McArdle n=47, Owen n=14, Schofield n =16, Livingston n =34, Harris-Benedict n=12, Revised Harris-Benedict n=11, Mifflin n =22, and IMNA n=4). Estimated RMR was higher than measured RMR (p<0.001) derived from all nine predictive equations (Table 3). Predicted RMR with eight of the nine established equations was between 9% and 20% greater than measured RMR with one equation overestimating RMR by 60% (Figure 1). The Harris-Benedict equation¹² and revised Harris-Benedict equation¹³ results were comparable with respect to prediction accuracy; the original equation¹² overestimated RMR by an average of 232 kcal/day (17%) and the revised equation¹³, overestimated RMR by an average of 246 kcal/day (18%).

Table 3.

Resting Metabolic Rate

	Mean±SD (kcal/day)	95% Confidence Interval (kcal/day)	p-value	Average kcal % overestimation (mean±SEM)
Measured RMR (n=71)	1552±319
RMR Harris & Benedict (n=71)	1784±299	1723 – 1844	7.59×10−11	17.3±2.4
RMR Revised Harris & Benedict (n=71)	1799±295	1738 – 1858	1.00×10−11	18.3±2.4
RMR Schofield (n=71)	1751±274	1688 – 1813	2.13×10−8	15.6±2.5
RMR Owen (n=71)	1769±257	1706 – 1831	1.50×10−9	16.9±2.5
RMR Mifflin (n=71)	1707±242	1649 – 1765	1.11×10−6	12.7±2.3
RMR Cunningham (n=70)	1806±237	1751 – 1870	1.10×10−12	19.8±2.5
RMR Katch-McArdle (n=70)	1652±233	1597 – 1716	0.0007	9.4±2.3
RMR IMNA (n=71)	2437±386	2362 – 2512	4.37 ×10−35	60.8±2.4
RMR Livingston (n=71)	1677±230	1618 – 1736	7.31×10−5	10.8±2.3

RMR: resting metabolic rate, IMNA: Institute of Medicine of the National Academies

Student t-tests between measured RMR and predicted RMR.

Figure 1.

Percent RMR overestimate for each prediction equation

Correlations between measured RMR and physical and functional variables are presented in Table 4. Measured RMR was associated with appendicular lean mass (Figure 2), total lean mass, and VO₂ peak (l/min) (Figure 3) (all p<0.001). In contrast, measured RMR was not significantly associated with 6MWD. RMR predicted by the Livingston equation² was graphed against ALM (Figure 4) since it was the most accurate in estimating RMR in chronic stroke survivors without considering lean body mass as a predictive value.

Table 4.

Correlations between Measured Resting Metabolic Rate, Body Composition and Physical Function

Variable	Pearson Correlation Coefficient
Appendicular Lean Mass (kg)	0.637**
Total Lean Mass (kg)	0.643**
VO2 Peak (L/min)	0.407**
VO2 Peak (ml/kg/min)	−0.128
6 Minute Walk Distance (m)	−0.080

VO₂peak: peak oxygen consumption.

^**P< 0.01 level (2-tailed).

Figure 2.

Relationship between VO₂peak and Measured RMR

Figure 3.

Relationship between Appendicular Lean Mass and Measured RMR

Figure 4.

Relationship between Appendicular Lean Mass, Measured RMR and Predicted RMR (Livingston Equation)

Discussion

RMR is often the most significant component of an individual’s daily energy expenditure and may play a greater role in energy expenditure for disabled chronic stroke survivors. Our study sought to compare measured RMR by indirect calorimetry to predicted RMR using a wider variety of predictive equations with a larger, more diverse sample of chronic stroke survivors than previously reported. Our results indicate that nine tested predictive equations overestimated RMR with eight of the nine equations overestimating on average by 9 – 20% with one equation having a 60% overestimation in chronic stroke patients.

Although all tested equations overestimate RMR, by population averages, Katch-McArdle¹⁹ (76 kcal/day) was the most accurate, followed by Livingston² (125 kcal/day) and then Mifflin-St Jeor¹⁶ (155 kcal/day). In order to explain differences in accuracy across equations, we compare variables included in each equation. The most rigorous equation, Katch-McArdle¹⁹, utilized a single variable, lean body mass. However, the Cunningham equation¹⁸ also only incorporated lean body mass but was less accurate. The inclusion of lean body mass in the prediction equation accounts for changes in muscle mass and muscle strength, which typically ensue following stroke. Common variables used to estimate RMR, such as height, weight and age, may not completely encompass energetic needs for stroke patients, given the physical and muscular changes they undergo following stroke. On the other hand, our data suggest that more precise measurements of muscle composition would best estimate RMR in this chronic stroke population. Further supporting this statement is depicted in Figure 4 which demonstrates the predicted RMR is higher than measured RMR at any given ALM, suggesting that the addition of ALM to a prediction equation² may improve its accuracy, likely functioning as a control for muscle atrophy. Moreover, since our study sample included a large range of ALM, roughly from 13 to 40 kg, it supports the addition of a muscle component to a prediction equation in stroke survivors.

Appendicular lean mass was strongly associated with RMR in stroke survivors, which supports the findings of our previous research where total body, total leg and paretic and non-paretic lean mass predicted RMR.²⁰ Because measured RMR correlated with VO₂peak (L/min) and ALM in our study, but not with functional measures such as 6 meter walk distance, this indicates that muscle atrophy may reduce RMR and fitness more than function in chronic stroke survivors. This is supported by a report which indicated that the 6MWD measurement is not an accurate assessment of cardiorespiratory fitness for stroke patients.²⁶ Moreover, this again supports our recommendation that equations such as the IMNA equation,¹⁷ which utilizes physical activity level, may be less accurate and less appropriate for chronic stroke patients.

We confirm our previous findings²⁰ that the Mifflin-St. Jeor equation¹⁶ overestimated RMR in stroke patients by 14%. Our present data illustrates that this equation overestimates RMR by approximately 13%. The IMNA equation,¹⁷ which was the least accurate equation in predicting RMR, overestimating RMR by more than 60% (~885 kcal/day), utilized age, weight, height and physical activity. The 95% confidence interval suggests a wide range for this mean estimation as well. The lowest estimation (min=1610 kcal/day) predicted by IMNA¹⁷ was higher than the average measured RMR for the subject population. The overestimation is not likely due to the use of common variables such as age or weight but rather the physical activity variable which was not utilized by any other eight equations. The physical activity measure could overestimate energetic needs of chronic stroke patients with hemiparesis or gait deficits that may result in reduced mobility. Conversely, in female athletes, though not male athletes, this equation¹⁷ was the most accurate predictive equation.²⁷ Thus, in healthy and active populations, including factors such as physical activity become more appropriate, but in inactive and sedentary populations, this equation may be less applicable. Future studies could assess the accuracy of the IMNA equation¹⁷ in a more active stroke survivor population.

Given that nine tested predictive equations were inaccurate in this stroke population, this suggests that the chronic stroke population has undergone physical changes which alter their RMR in comparison to healthy adults. This is especially compelling when considering that many of these equations have been validated, suggesting comprehensiveness and verified accuracy; for instance, Mifflin-St. Jeor¹⁶ has undergone at least 10 validation studies, Harris-Benedict¹² 25 validation studies, and Owen¹⁵13 validation studies. Potentially, some of this error could be attributed to the sample, which was on average overweight or obese, as predictive RMR equations tend to be less accurate for this population.²⁴ Yet, this cannot explain the entire variation, given that others^{2, 15, 16} were developed in obese individuals. Our data suggests that RMR is not different between African American and Caucasian stroke survivors, although previous studies of larger sample size suggest that RMR is lower in African American than in Caucasian adults.^{28, 29} It is likely that differences in biological composition ensue following stroke, such as changes in muscle architecture, as well as changes in physical fitness/aptitude, all which may alter RMR values from predicted values. At the same time, there may be other variables, such as sleep patterns, muscle morphology, or other potential factors not yet understood, contributing to varied RMR in chronic stroke patients.

Our findings are also interesting when considered in conjunction with the findings of Zampino et al.³⁰ who reported that a decline in RMR over time was associated with multiple chronic diseases, including type 2 diabetes mellitus, osteoarthritis, cancer, chronic obstructive pulmonary disease, chronic kidney disease, and chronic heart failure, but not with stroke. Because chronic stroke incidence has not been found to result in a reduction in RMR independently over time,³⁰ this leads us to believe that the initial or acute physical changes resulting from stroke, including loss of skeletal muscle may be responsible for RMR being lower than predicted. This may explain why the Katch-McCardle equation,¹⁹ which included lean mass, was the most accurate predictive equation. Future studies could measure RMR in acute stroke patients to test the hypothesis that RMR may increase immediately following disease³⁰ onset due to the increased energetic cost to maintain homeostasis. Our study only included chronic stroke patients, who after three months following stroke incidence, may be assumed to have surpassed the initial homeostatic maintenance phase associated with increased RMR following initial incidence of chronic disease.

Although obesity and stroke have been widely studied, post-stroke weight gain and future obesity status has not been adequately examined. Our analysis could shed light on calorie intake guidelines for stroke survivors in order to prevent overestimation in the number of calories this population burns per day. This overestimation may inform preventative health measures with respect to weight gain for stroke patients. This is especially significant when accounting for obesity observed in patients with stroke; in one study 46% of patients in the rehabilitation wards for stroke were classified as obese.³¹ Based on these considerations, predicted RMR for stroke patients, and their daily caloric intake guidelines, should be adjusted to reduce risk of obesity and correspondingly, to increase the prospect for optimum post-stroke health. Furthermore, the hemiparetic stroke population tends to be sedentary due to limited post-stroke mobility, so RMR becomes an even greater relative component of daily energy expenditure. In this way, it becomes more imperative that a chronic stroke patient’s daily caloric intake is not established based solely upon their predicted RMR as this may be a gross overestimate and has the potential to result in weight gain. As such, these findings with regard to RMR overestimation could be used to advise chronic stroke survivors in long term weight loss for health maintenance.

The correlation between appendicular lean mass and RMR indicates a need to investigate whether weight loss interventions in chronic stroke patients can be aided by strength training to maintain muscle mass, in conjunction with a calorie deficit diet. Muscle mass maintenance is especially crucial in the chronic stroke survivor population because of widespread muscle wasting in this group due to physical disability and aging.³² Furthermore, it would be helpful to examine whether the chronic stroke patients’ intake of particular food groups, such as protein, or other lifestyle habits, such as sleep, can influence RMR, given that RMR decreases after sleep reduction, in healthy adults.³³

Study Limitations

Although our sample size was larger than any other study of RMR in stroke and had more than required for significant power to detect differences between measured and predicted by established equations, an even larger sample size may be necessary to derive a new prediction equation. Our results are limited by potential for recruitment bias and generalizability with more males than females, strict exclusion criteria (aphasia, common comorbidities, not independent ambulators, etc.), and missing data for the six minute walk test. However, the study was racially diverse, including both men and women. Although over half of the participants were African American, we were unable to assess whether the overestimation of RMR predictive equations vary by race. Assessments were developed on a population scale and may be difficult to extrapolate on an individual level.

Conclusion

By comparing measured RMR in chronic stroke patients with predicted RMR from multiple equations established in studies with healthy adults, our study demonstrates the need for population-specific RMR predictions for chronic stroke patients. If measurement of RMR is not feasible, the Katch-McArdle equation should be used to estimate RMR in a stroke patient, as, on average, it provides the lowest percentage overestimate compared to other equations. Since it is difficult to calculate RMR directly and indirectly due to availability of specialized equipment, it is especially important to develop new equations to accurately predict RMR in chronic stroke patients. Since the established equations overestimate RMR, reduction of caloric intake from predicted may need to be considered. Because this population demonstrated a wide variability in many of the measures, as aforementioned, these population level assessments should be interpreted with caution when extrapolated to individuals. Future research may also endeavor to understand variables other than lean mass that explain discrepancies in RMR in chronic stroke as well as interventions to prevent or delay a reduction in RMR after stroke.

Abbreviations:

RMR

resting metabolic rate

REE

resting energy expenditure

PAL

physical activity coefficient