Indirect Calorimetry Performance Using a Handheld Device Compared to the Metabolic Cart in Outpatients with Cirrhosis

Addressing malnutrition is important to improve health outcomes in outpatients with cirrhosis, yet assessing energy requirements in this population is challenging. Predictive equations of resting energy expenditure (REE) are thought to be unreliable, and traditional indirect calorimetry is expensive and infrequently available for clinical use. The accuracy of REE predictions using a MedGem® handheld indirect calorimeter, the Harris Benedict Equation (HBE), the Mifflin St. Jeor equation (MSJ), and the gold standard Vmax Encore® (Vmax) metabolic cart was compared. The REE of cirrhotic pre-liver transplant outpatients was analyzed using each of the four methods. Agreement between methods was calculated using Bland–Altman analysis. Fourteen patients with cirrhosis participated, and were primarily male (71%) and malnourished (subjective global assessment (SGA) B or C 64%). Lin’s concordance coefficient (ρC) for MedGem® vs. Vmax demonstrated poor levels of precision and accuracy (ρC = 0.80, 95% confidence interval 0.55–0.92) between measures, as did the HBE compared to Vmax (ρC = 0.56, 95% confidence interval 0.19–0.79). Mean REE by MedGem® was similar to that measured by Vmax (−1.5%); however, only 21% of REE measures by MedGem® were within ±5% of Vmax measures. Wide variability limits the use of MedGem® at an individual level; a more accurate and feasible method for determination of REE in patients with cirrhosis and malnutrition is needed.


Introduction
Malnutrition is prevalent among patients with cirrhosis for multiple reasons including inadequate energy intake, impaired gut motility, nutrient malabsorption, increased protein losses, and a metabolic shift to gluconeogenesis and fatty acid oxidation resulting in an accelerated state of starvation [1][2][3]. The loss of skeletal muscle and fat mass is an independent risk factor for increased morbidity and mortality [4,5]. Clinical consequences of malnutrition can include infection, compromised immune function, progression of liver dysfunction, hepatic encephalopathy, and increased mortality both before and after liver transplantation [6][7][8][9]. Optimizing nutrition, using a multi-disciplinary nutrition intervention, can improve patient outcomes and quality of life [10][11][12].
accuracy of the Harris Benedict equation (HBE) and Mifflin St. Jeor (MSJ) equation as compared to the REE calculated by the metabolic cart.

Materials and Methods
Research ethics approval to complete this study was obtained through the University of Calgary Conjoint Health Research Ethics Board. Participants were provided with verbal and written information about the study before providing their written consent.

Patients
The sample population was composed of 15 patients with cirrhosis from the Malnutrition Clinic, a specialty outpatient clinic for pre-liver transplant nutrition assessment and management in Calgary, Alberta. Patients were eligible to participate if they had a current diagnosis of liver cirrhosis and were awaiting liver transplantation assessment. Demographic data collected included age, sex, height, estimated weight, estimated dry BMI, aetiology of cirrhosis, model for end-stage liver disease (MELD) score, presence of ascites and malnutrition status determined using subjective global assessment (SGA) ( Table 1). SGA is a validated tool used to classify patients as mildly, moderately, or severely malnourished [29].
Patients were excluded if they had received a general anesthetic within the previous 24 h, were discharged from hospital within 24 h, were within six hours of hemodialysis, used supplemental oxygen therapy, had claustrophobia, nausea, vomiting, were unable to hold the MedGem ® calorimeter, or were unable to tolerate lying supine for the required time. Patients were required to meet the procedure criteria that included a fasted state for eight hours, avoidance of exercise or caffeine within four hours, and avoidance of nicotine, allergy or cold medications, or herbal supplements within one hour of testing.

Procedure
REE was measured using the Vmax Encore ® metabolic cart (Vyaire Medical Inc., Yorba Linda, CA, USA), which was calibrated according to manufacturer's instructions prior to each use. A clear plastic canopy is placed over the participant's head, attached to the Vmax by disposable plastic tubing. The Vmax Encore ® measures VCO 2 , and VO 2 (litres per minute) and thus calculates the respiratory quotient (RQ) and REE. Vmax testing was carried out for up to 30 min. The first five minutes of testing were discarded in all cases. After the initial, discarded five-minute period, once five minutes of steady state were achieved patient testing was concluded; if achieved, all five minutes of steady state data were included in the analysis. Steady state is defined as a time period in which the mean VO 2 and VCO 2 are changed by less than 10% [30]. For those who did not achieve five minutes of steady state data collection, testing was continued to 30 min; the first five minutes of data were discarded, and the latter 25 minutes of data were averaged for analysis. Data analysis was carried out through a trained user and the Vmax machine. This protocol was in accordance with best practice guidelines for Vmax use [18,21,22,30,31]. The MedGem ® hand-held indirect calorimeter is a self-calibrating device. It measures VO 2 (mL per minute) only and uses a fixed RQ of 0.85 to calculate REE. Patients wore a nose clip and disposable mouthpiece connected to the HHRC, which they supported with one arm for the measurement period. Measurement took approximately five to ten minutes, with the end time automatically determined by the device.
Eligible patients were assessed by the Vmax Encore ® and the MedGem ® indirect calorimeters consecutively. Completion of Vmax and MedGem ® tests were conducted in randomly assigned alternating order to minimize bias. Exceptions to randomly assigned testing order were made in three instances where the Vmax failed to calibrate within a one-hour time frame leading to MedGem ® testing first. In each of these instances the Vmax was calibrated, and testing completed before the end of the appointment.
Patients rested for 20 min prior to the first test, and remained in the recumbent, relaxed position between testing [18,21,22]. The second test was carried out within a 5-15 min period after the first test was completed [17,22]. Appointments were conducted in the morning between 08:00 and 12:00 h; morning hours were chosen considering fasting requirements.
All participants were studied in the supine position, recumbent with torso inclined at approximately 30-45 degree angle, in accordance with best practice guidelines [15]. Slight variations in position were made to assure the comfort of each participant, and to minimize movement during the testing period [21]. When using the MedGem ® device, the arm holding the device was propped in order to minimize the effect of arm activity; this position has been found to be insignificant on final REE [15]. No other changes in positioning were made when switching in between Vmax and MedGem ® .
HBE and MSJ were calculated for each patient based on their weight on the day of the appointment, using their estimated dry weight as evaluated by the trained researcher. HBE values for the purpose of this study do not include a disease cofactor or activity cofactor as they are used to calculate REE versus TEE.

Statistical Analysis
After collection, data from one participant were excluded before analysis due to extreme deviation in measured Vmax REE compared to both REE determined from the other three methods, as well as significant deviation from clinically expected values. We are not aware of any distinguishing characteristics of this patient that might have accounted for this error.
Data are reported as mean ± standard deviation (SD). Analysis of variation with Student's two-sided paired t-test was used to compare mean differences between testing methods, with the statistical significance level set at α = 0.05.
Lin's concordance correlation coefficient ( C) was used to examine the relationship and level of agreement of Vmax to MedGem ® , Vmax to HBE, and Vmax to MSJ [32]. A scale for interpreting relationships between datasets using C values suggests an almost perfect relationship is C >0.99, a moderate relationship is C 0.95-0.99, a fair relationship is C 0.90-0.95, and a poor relationship is < 0.90 [32].
Bland-Altman plots depict the difference between paired measurements to the mean of the pair, and bars to represent limits of agreement depict the agreement between methods for determining REE [33]. The limits of agreement were set ± 1.96 standard deviations from the mean with another line added representing a 5% kcal over-or underestimation, based on literature standard for critical value difference in REE measuring devices [15,26,27].

Results
The mean age of the 14 patients was 53.5 ± 9 years, with a total of 10 (71%) men. Two-thirds of the study patients were malnourished (SGA B or C). Patient demographics, aetiology of cirrhosis, disease characteristics, and nutrition status are reported in Table 1. The mean measured REE using the Vmax was 1472 kcal/day (SD ± 279). Mean REE assessment by MedGem ® underpredicted REE by 1.3% compared to Vmax. HBE underpredicted REE by 4.2%, and MSJ underpredicted REE by 6.6%. The mean REEs are reported in Table 2.
When SGA categories were stratified, MedGem ® appeared to overestimate REE in SGA A patients by 5% and underestimate in SGA B patients (−6%) and SGA C patients (−5%). Only the SGA B category achieved significance (p = 0.04). After HBE values were stratified using SGA, there was a non-significant mean underestimation in all groups, with the highest degree of underestimation in the SGA A group (−7%), followed by SGA B (−2%) and SGA C (−3%). There was a similar, non-significant underestimation of REE by MSJ when stratified to SGA class; SGA A patients were underestimated by 10%, SGA B by 2%, and SGA C by 6%.

Discussion
The present study reports on the effectiveness of two measures of REE compared to the metabolic cart; the reference standard measure of REE. To our knowledge, this is the first study to describe the relationship between a handheld indirect calorimeter in outpatients with cirrhosis. Our findings demonstrate poor relationships between Vmax and MedGem ® ( C = 0.80), Vmax and HBE ( C = 0.56), and Vmax and MSJ ( C = 0.47). Although MedGem ® demonstrated a superior relationship with Vmax than did HBE or MSJ, it was not found to be accurate on an individual patient level. Our chosen level of agreement was ±5% as is standard in similar literature. According to MedGem ® , 21% of patients tested fell in this category, compared to only 14% of HBE MSJ and 7% of HBE REE values. Although MedGem ® demonstrated better agreement, all tools suggest a higher than acceptable level of disagreement for determining REE for the majority of patients in this study.
Mean REE by MedGem most closely approximated Vmax measurement (mean difference −21 kcal/day). MSJ underestimated REE by the largest margin, which was 93 kcal/day on average. Interestingly, MedGem ® indirect calorimetry underpredicted mean REE, but over-and under-predicted in an equal proportion of patients, therefore not supporting the presence of a systematic bias. When MedGem ® indirect calorimetry under-predicted REE, the magnitude of mean difference was nearly 190 kcal/day. This under-prediction is of concern due to magnitude, and largely observed in a malnourished subset of patients (SGA B and C). Similarly, using HBE and MSJ, the magnitude of under-prediction was of clinical relevance, approaching 220 kcal/day and 260 kcal/day respectively. Identifying patient or disease related factors that may predict REE under-prediction is of importance, as under-dosing the nutrition prescription may have deleterious consequences in this population at risk for or with active malnutrition. It is interesting to observe MedGem ® under-predicted mean REE in malnourished patients compared to well-nourished patients, although these results may have been due to chance. Our findings regarding MedGem ® were similar to those reported by Glass et al. (2012) in their validation study. Glass reported a mean difference of −28 kcal/day (95% CI = −150 to 94 kcal/day) between REE measured by MedGem ® and Vmax when tested in a quiet environment that was similar to our testing environment [15]. While the authors concluded the handheld calorimeter is a "valid, feasible" method, their study population may not have represented malnourished cirrhosis patients as they had a mean body mass index of 29.6 ± 7.7 with a range of 18.5-56.1 and 68% had ascites. The authors did not specify if reported BMI was based on current body weight or estimated dry weight. No other measures of nutritional status were reported [15]. Another major difference between studies was the outpatient population in our study, which may be associated with underlying differences in patient health status and activity levels compared to the hospitalized patients. In a similar validation study of MedGem ® in both healthy patients and patients with cancer, Reeves et al. (2005) also used 5% variation in REE as clinically acceptable disagreement in measurement as compared to REE measured by Vmax. The authors found that less than one-third of tested patients fell within the acceptable range, concluding that limits of agreement for MedGem ® versus traditional indirect calorimetry were too wide and provided poor clinical accuracy for the determination of REE at an individual level [27]. Considering that less than one-quarter of our patients fell within this limit, we find this level of disagreement to be higher than acceptable. A major difference between the handheld MedGem ® and traditional indirect calorimetry is the fixed RQ used in calculating the final REE with MedGem ® , which may not accurately reflect metabolic shifts occurring in individuals with advanced liver disease. In comparison, Vmax measures both VCO 2 and VO 2 to calculate a RQ that is specific to each patient, thereby reflecting their current metabolic demands. A low RQ reflects a heavier reliance on fatty acid metabolism and gluconeogenesis, and has been documented in individuals with sarcopenia and cirrhosis, which has been interpreted as evidence of accelerated starvation [34][35][36][37]. Patients with severe malnutrition or advanced cirrhosis could be expected to have a progressively lower RQ, which could result in inaccuracies in REE calculated by MedGem ® [16]. This is generally supported by the trend in subgroup analysis whereby SGA A was overestimated by MedGem ® (+5%), whereas SGA B and C were underestimated (SGA B −6%; SGA C −5%); even though only the SGA B group achieved statistical significance. Our hypothesis has also been endorsed by findings of a MedGem ® validation study among patients with anorexia nervosa, which reported individual RQ varied significantly from the fixed 0.85 used by MedGem ® [28]. Although the use of MedGem ® eliminates the need to calculate dry weight and adjust for disease severity, the fixed RQ of 0.85 may not accurately reflect the metabolic changes associated with advanced cirrhosis and sarcopenia, which is consistent with a lower RQ.

Strengths and Limitations
Strengths of our study include a single operator for both MedGem ® and Vmax measurements to ensure consistency of techniques and process, adherence to testing measures, and a specific pre-liver transplant population.
A major limitation of this study was the low sample size of 14 patients, impacting the power of our study, resulting in wide mean differences and standard deviations, and unequal distribution of gender-a reflection of the broader population with a diagnosis of cirrhosis. Our study was also limited by patient heterogeneity; patients had a range of aetiologies for their liver cirrhosis, MELD scores, and presence or absence of ascites. It is possible that either the disease heterogeneity or low sample size leading to limited power may have contributed to the lack of significant findings. Despite this, our data were still normally distributed and did produce some significant findings. Our low sample size limited the ability to conduct adequately powered subgroup analyses between the SGA malnutrition classes.
We excluded data from one participant due to highly discrepant values using Vmax. We could not identify any features of this participant, including deviation from standard testing protocol, to account for the discrepant findings. It is possible that there was an error in Vmax testing for this patient, as the measured REE by Vmax was significantly higher than other predictions and measures, and was not consistent with the clinical picture. This example highlights potential limitations of the reference standard, Vmax, which carries its own potential for error in REE measurement.
Our protocol followed best practice guidelines for the use of Vmax; however, there is no uniform procedure and analysis protocol available to ensure the accuracy and precision of metabolic carts [22]. This presents a limitation for using Vmax as the gold standard method to assess REE [15,21,22]. Cooper et al. identified Vmax could not be considered "adequately reliable in a research setting" [19]. This challenge, while inherent with all currently commercially available metabolic carts, may limit the quality of our data. Our protocol would have been strengthened by repeat MedGem ® measurements in addition to measurements of the same patients on multiple days to establish intra-device and intra-person reliability of our results.

Future Direction
We hope that our data will serve as a pilot for future testing of the MedGem ® device among patients with cirrhosis. Further testing with larger sample sizes should be carried out specifically focusing on moderately and severely malnourished patients, in patients with and without ascites, and within a more homogenous aetiology of cirrhosis. This will help identify certain patient subsets where MedGem ® indirect calorimetry may have better performance characteristics compared to predictive equations. Future studies should also specifically assess RQ in malnourished cirrhotic patients to further delineate if low RQ affects the reliability of the MedGem ® REE.

Conclusions
To avoid further exacerbating malnutrition or overfeeding adequately nourished cirrhosis patients, reliable, valid, and feasible methods to calculate REE and overall nutrition prescription are needed. Current available methods have poor levels of agreement and should not be considered a definitive guide for prescribing nutrition interventions in these patients. Patients should be followed closely to identify if their nutrition prescription is meeting or exceeding their needs over time.

Conflicts of Interest:
The authors declare no conflict of interest.