Predicted vs Measured Energy Expenditure in Critically Ill, Underweight Patients 
* Montana State University, Department of Health
and Human Development, Bozeman, Montana; Correspondence: Christina Gayer Campbell, PhD, RD, Montana State University, 20 Herrick Hall, Department of Health and Human Development, Bozeman, MT 59717-3540. Electronic mail may be sent to ccampbel{at}montana.edu. A retrospective analysis was conducted to compare 4 energy-prediction equations against measured resting energy expenditure (MREE) determined via indirect calorimetry. Data from a heterogeneous group of 42 critically ill, severely underweight (59.50 ± 17.30 kg; 77.1 ± 9.7% ideal body weight [IBW]) male patients were assessed. The Hamwi formula was used to determine IBW. The Harris-Benedict (HB) equation was calculated for patients <90% IBW using both current body weight (CBW) and IBW. Energy needs were also estimated with an Ireton-Jones formula for all mechanically ventilated patients (n = 37). For patients <85% IBW (n = 31), an adjusted body weight was determined ([CBW + IBW]/2) and used in the HB formula. The HB formula using the IBW, CBW, and adjusted body weight was significantly different (p < .05) than MREE. The Ireton-Jones equation was not significantly different (p > .05) from MREE but tended to overestimate energy needs (109.3% ± 16.8% MREE). Conversely, using the CBW or IBW in the HB underestimated the patient's energy needs; 77.0% ± 11.6% MREE and 90.9 ± 16.1% MREE, respectively. For patients <85% IBW, use of the adjusted body weight in the HB represented 84.2% ± 13.9% MREE. The average caloric need was 31.2 ± 6.0 kcal/kg CBW. Indirect calorimetry remains the best method of determining a patient's energy needs. Until a large prospective trial is conducted, a combination of prediction equations tempered with clinical judgment and monitoring the appropriateness of the nutrition prescription remains the best approach to quality patient care. Malnourished, hospitalized patients have higher rates of morbidity and mortality than do well-nourished patients.1 The hypercatabolism associated with critical illness accelerates the loss of lean body mass and can quickly lead to malnutrition in previously well-nourished patients. Underweight (defined as <90% of ideal body weight [IBW])2 patients have fewer energy reserves in the form of lean body and fat masses than normal-weight patients or may be malnourished, placing them at even greater risk for the complications associated with malnutrition.1,3 Standard equations used to predict energy expenditure such as the Harris-Benedict (HB) equation with the addition of an injury-correction factor can overpredict energy needs.4,5 Overfeeding of the critically ill patient can result in hyperglycemia, increasing the risk of nosocomial infection6; hypertriglyceridemia, which may suppress immune function by affecting the reticuloendothelial system; or increased carbon dioxide production, which may exacerbate respiratory failure or delay ventilator weaning.7,8 Determining an appropriate method for predicting energy needs has been an area of research for many years, dating back to the early 1900s with the creation of the HB equation. Determining energy requirements in the critically ill patient via indirect calorimetry (IC) has long been considered the gold standard.9,10 Limitations for using IC include time constraints and cost.11 Despite recent advances in portable metabolic devices (MEDGEM; Healthetech Inc, Golden, CO) that make measuring IC more realistic,11 the fact remains that many hospitals do not yet have this technology and may never prioritize purchasing this equipment. Furthermore patient conditions and treatments such as renal replacement therapies, leaking chest tubes, acidbase disorders, or an inability to achieve steady state in unstable patients12 may prohibit the use of IC in those patients who need it most. The purpose of the present study was to determine which commonly used prediction equations for patients of normal weight were the best estimates of energy expenditure when compared with measured resting energy expenditure (MREE) in a heterogeneous group of critically ill, underweight, male patients. It is hypothesized that prediction equations typically used in the clinical setting underestimate energy needs for the critically ill, underweight population compared with IC results. Prediction equations will continue to be a necessary component of the nutrition assessment for the critically ill patient in many clinical settings, despite the extensive list of inherent problems11 associated with predicting energy needs. Attention has been given to determining an appropriate prediction equation for critically ill, overweight, and obese patients,13–16 and children17–19 and; however, limited research has specifically focused on the underweight population.20
Study Design A retrospective analysis was performed using IC results on 42 critically ill, male patients admitted to a tertiary-care hospital in the Pacific Northwest from 1992 to 2003. Only men were included in this analysis because of a limited number of data for women (n = 3). As part of the patient's care as recommended by the intensive care unit dietitian or the physician, MREE was determined for each patient using IC to determine appropriate energy goals for nutrition support. All data used in this analysis were obtained from nutrition care plans generated by the attending registered dietitian. To conduct the retrospective analysis, data from patients <90% IBW as calculated using the Hamwi formula (IBW = 106 lbs for 5 ft tall, and 6 lbs/inch for each inch over 5 ft)21 were entered into an Excel spreadsheet. Patients with a body weight >90% IBW were excluded from the analysis. Categorizing individuals as underweight (<90% IBW), average weight for height (90%–110% IBW), overweight (110%–129% IBW), and obese (>130% IBW) is a commonly used method in the clinical setting as a component of the dietitian's diagnosis of malnutrition.2 Patients were admitted to the ICU for a variety of disease states. In reviewing the nutrition-specific medical records maintained by the patient's dietitian, patients had some or all of the following diagnoses: sepsis, respiratory failure, acute renal failure, acute respiratory distress syndrome, multiple trauma, pancreatitis, chronic obstructive pulmonary disease, or diabetes mellitus. Descriptive characteristics for all patients are provided in Table 1. This analysis was approved by the hospital institutional review board.
IC
Predicted Resting Energy Expenditure Equations
Statistical Analysis
A summary of the raw data for MREE and the prediction equations for all patients are presented in Table 3. The Ireton-Jones equation overestimated energy needs with 109.3% ± 16.8% MREE. Conversely, using the CBW in the HB underestimated the patient's energy needs (77.0% ± 11.6% MREE). The prediction equation using the IBW in the HB formula represented 90.9% ± 16.1% MREE. For patients <85% IBW, the HB equation with the adjusted body weight was 84.2% ± 13.9% MREE.
Results of the ANOVA revealed that all prediction equations except the Ireton-Jones were significantly different from IC results. For patients <90% IBW, the Ireton-Jones equation was not significantly different from MREE (Table 3). The mean prediction error for all subjects was 110.1, –440.7, and –204.7 kcal/d for the Ireton-Jones and HB formulas with the CBW and IBW, respectively. All 3 prediction equations were significantly correlated (p < .05) with MREE (r = .61, Ireton-Jones vs MREE; r = .67, HB with CBW vs MREE; r = .53, HB with IBW vs MREE). For patients <85% IBW, the mean prediction error was –314.8 kcal/d for the HB equation with the adjusted body weight. This equation was significantly correlated with MREE (r = .57; p < .05). To assess clinical relevance vs statistical significance, data were divided into 3 categories to represent an overestimation (>110% MREE), an appropriate estimation (90%–110% MREE), and an underestimation (<90% MREE) of energy needs. This method of interpretation has also been used by Alberda et al15 and Hardy et al18 and represents ±10% error. Figure 1 illustrates the percentage of patient's energy needs that were underestimated, accurately estimated, or overestimated as a percent of the MREE (patient's prediction equation result/MREE x 100 = %MREE). The HB (CBW) and HB (IBW) equations underestimated 83.3% and 52.4%, respectively, whereas the Ireton-Jones equation overestimated energy needs in 43.2% of patients. The HB (adjusted body weight) formula underestimated 77.4%, accurately estimated 19.4%, and overestimated 3.2% of the patient's energy needs.
To determine an appropriate empirical formula, the average caloric need was obtained from the MREE results and divided by the average body weight of the patients from each group. The average caloric need according to a calorie-per-kilogram basis was 31.2 ± 6.0 kcal/kg CBW.
The overall incidence of malnutrition in the ICU may be as high as 50%.26 The loss of lean body mass associated with malnutrition can result in impaired cardiac function (decreased heart rate and stroke volume), respiratory function (loss of mass from the muscles required for breathing), immune function, and wound healing.3 Loss of plasma proteins can affect fluid balance (loss of intravascular fluids into extravascular space), further exacerbating cardiac and other organ dysfunction. The population studied in the current analysis represents an already malnourished group according to percent IBW. Under-feeding a critically ill population similar to the one presented in this investigation could further perpetuate malnutrition and increase patient mortality and morbidity. In the present study, the appropriateness of individual prediction equations predominates over other equations. The Ireton-Jones equation is not significantly different from the IC results and is on average only 110 kcal different; however, using the Ireton-Jones represented 109% of the MREE for all patients. When assessed on an individual basis, an equal percentage of patients was appropriately estimated (43.2%) and overestimated (43.2%). Use of the equation had a tendency to provide an excess of the energy requirements as determined by MREE. Conversely, for all patients, according to group averages, the HB with the IBW represented 91% of the MREE, suggesting this prediction equation to underestimate energy needs. On an individual basis, the HB with the IBW underestimated the majority (52.4%) of patients. Using the IBW in the HB formula without an added stress factor would have a tendency to underestimate the appropriate amount of needed calories. The mean prediction error is defined as the MREE minus the estimated energy needs. Ultimately, the mean error will be close to zero. Evaluating the mean prediction error indicates whether the predicted energy needs are either over- or under-estimated compared with the MREE. It was determined that the mean prediction error substantiated the use of the Ireton-Jones as the most appropriate prediction equation. For all prediction equations, the correlation coefficient (r value) was significant. The r value is a measure of the degree to which the uncertainty in predicting energy needs is reduced by using a specific equation.27 The significant r values provide support for the use of the Ireton-Jones and the HB formulas with both the IBW and CBW. Despite the significant r value, the HB with the CBW was found to underestimate energy needs (>400 kcal on average) for the majority of patients, and other formulas represented better choices. In clinical practice, a stress factor is typically added to the HB results. In the current analysis, only the HB data were compared with MREE, not HB multiplied by a stress factor. However, using a stress factor of 1.1 with the HB (IBW) would increase the estimated number of calories by 10%, resulting in a shift of the percentage of patients appropriately estimated from 38.1% (HB [IBW] x 1.0) to 50.0% (HB [IBW] x 1.1), with 28.6% and 21.4% of patients under- and overestimated, respectively. In comparison with this recommendation are recent suggestions by Barak et al14 that energy needs for patients with a body mass index <18.5 kg/m2 should be calculated with HB (CBW) x 1.25. Using the CBW in the HB and then multiplied by 1.25 in the current analysis would increase the number of appropriately estimated patients to 42.9%, and a similar percentage of patients (35.7%) would be underestimated, whereas 21.4% of patients would be overestimated. Ahmad et al20 also reported that body weight is a principal determinant (r2 = 0.558, p < .005) in the HB formula rather than other components (age, sex, height). As a patient's lean body mass may change during the course of the illness, the use of the HB with the IBW may be more accurate as compared with the CBW because the actual body weight is potentially changing with the disease process. This concept substantiates the use of the patient's IBW rather than the CBW. Ahmad et al20 conducted a prospective trial on a heterogeneous group of 14 underweight (40.9 ± 5.1 kg, 70.7 ± 7.8% IBW), critically ill patients. It was determined with IC that 30–32 kcal/kg was an appropriate empirical formula to use for this population. The present analysis is in agreement with the findings of Ahmad et al,20 with an empirical formula of 31 kcal/kg (mean). Additionally, Alberda et al15 conducted a retrospective analysis with a subcohort of 14 critically ill patients (body mass index <20 kg/m2) and determined the energy needs for this patient population are most appropriately predicted with an empirical formula of 37 kcal/kg. A limitation to the present analysis is the relatively small number of patients. According to Ireton-Jones et al,25 a minimum of 50 patients is preferable. Furthermore, the patients in this study represent a heterogeneous population because of a variety of disease states. The energy needs of critically ill patients with the same disease states can vary significantly, as can the energy needs of an individual patient in different stages of the disease process. The common theme between the patients in the present analysis is their body weight (<90% IBW). Despite the diversity of disease states, the recommended empirical formula from this analysis (31 kcal/kg) is supported by research with similar suggestions.15,20 A larger prospective trial that is diagnosis specific should be conducted in the future to determine an accurate method of predicting energy needs in this potentially malnourished patient population. An additional limitation is the fact that the data were collected over an 11-year time span. Several different respiratory therapists conducted the IC tests according to the hospital procedures, and 3–4 different metabolic carts were used for data collection. There was no standardization as to when the IC test was conducted, except that the FIO2 was <70%24 and the patients were hemodynamically stable. Furthermore, the data included in the analysis were obtained from the nutrition-specific medical records, which do not indicate how long the patient had been intubated before the IC test or the length of time the patient had received enteral or parenteral (all as mixed fuels) before the measurement. Because this analysis is retrospective, these are factors that cannot be controlled yet should be accounted for when designing a larger prospective study.
No one prediction equation can be universally applied to all underweight patients without over- or underestimating energy needs for some or all of the patients. A combination of prediction equations tempered with clinical judgment and the judicious use of IC, when feasible, remains the best method of predicting energy needs at this time. Although these results were derived from a heterogeneous group of patients, the findings support the need to conduct a tightly controlled prospective study to develop an appropriate prediction equation for critically ill, underweight patients. All patients should be continuously monitored to assess the appropriateness of the nutrition prescription, including caloric intake, and changes made when indicated by the patient's medical status or metabolic tolerance to the nutrition therapy.
Nutrition in Clinical Practice, Vol. 20, No. 2,
276-280 (2005)
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
|
 |
|
Top
Methods
Results