ABSTRACT Predictingenergy needs in criticalillness can bedifficult because of uncertainties about the influence of multiplefactors on energy expenditure. Understanding these components isimportant to avoid limiting optimal outcome by underfeeding andto avoid complications of overfeeding. Prediction strategies oftenuse a patient's weight to estimate needs. For overweight patients,there is controversy as to whether actual or modified weight shouldbe used in predictions. This study was designed to evaluate aproposed technique to improve the accuracy of predicting energyneeds in critically ill, overweight subjects. Subjects' energy needswere predicted [with Harris-Benedict equation (HBE) and kilojoules per kilogram (KPK) strategies] by using both actual weightand an adjusted weight developed to attempt to more accuratelyreflect lean mass. Results were compared with measured energyexpenditure determined by indirect calorimetry. Results indicatedthat use of actual weights in predictions for overweight subjectsmay lead to overfeeding. Use of adjusted weights led to moreaccurate energy predictions with the KPK than with the HBEstrategy. Adjusted-weight strategies could explain > 45% of thevariability of resting energy expenditure in subjects 130159%ofideal body weight. Results of this study suggest that using adjustedweights with the KPK prediction strategy may be preferable forthis population, particularly for patients 130% of ideal bodyweight. This study also indicated that multiple diagnoses may notlead to increased energy requirements. Am J Clin Nutr1997;66: 12506.

KEY WORDS Energy expenditure, obesity, indirect cabrimetry, critical illness, body composition, body weight

iNTRODUCTION

Predicting energy needs in critical illness is made difficult byuncertainties regarding the effect on expenditure of differentdisease states, the stage of the patient in the normal cycle ofstress response to illness, and individual variation (1). Although accurate prediction is desirable to avoid potential complications from both under- and overfeeding (2-4), researchershave found that critically ill patients are often fed amounts ofenergy significantly different from their measured resting energy expenditure (REE) as determined by indirect calorimetry(5, 6). Underfeeding may contribute to malnutrition and limitsuccessful recovery, whereas overfeeding may contribute toelevated blood glucose concentrations, increased risk of infec

tion, elevated carbon dioxide production, and respiratory cornplications (24). Prediction strategies often use a patient'sweight as a factor in predicting needs. When a patient isoverweight because of adipose tissue or fluid retention associated with critical illness, there is controversy about whatweight should be used for predictionideal or actual? Someresearchers have advocated use of a preillness weight (7),others an adjusted weight (8), whereas others have found thatactual weight correlates better with measured energy expenditure (9). Weissrnan and Kemper (10) advocate correcting forresuscitation fluid before using weight for prediction. Vicari eta (11) found that actual, ideal, and adjusted weights wereequally poor predictors (r = 0.510.54)of measured energyexpenditure but the percentage error was less when usingadjusted weights.

The purpose of this study was to assess the ability of aproposed technique to improve the accuracy of predictingenergy needs in ventilator-dependent, critically ill, overweightsubjects. The proposed technique involved using neither thepatient's actual weight nor their ideal body weight (IBW) inprediction equations but using an adjusted weight. The objective of this weight adjustment was to correct for excess fluid orfat and to derive a weight more reflective of lean body mass,which should more accurately reflect energy expenditure (1215). The success of this technique was evaluated by comparingpredictions using actual weight and predictions using adjustedweight with actual energy expenditure measured by indirectcalorimetry.

SUBJECTS AND METHODS

SubjectsSubjects were ventilator-dependent adult patients in the med

ical or surgical intensive care units of three hospitals. Thesewere the Harry S Truman Memorial Veterans Hospital, Co

i From the Departments of Food Science and Human Nutrition and

Agriculture, University of Missouri, Columbia, and the Department ofInternal Medicine, Nutrition Support Service, Truman Memorial VeteransHospital and University of Missouri, Columbia.

2 Address reprint requests to TE Edes, Truman Memorial VA Hospital.

Columbia, MO 65201. E-mail: mcutts@mthl.coin.missouri.edu.Received August 7, 1996.Accepted for publication May 16, 1997.

1250 Am J Clin Nutr 1997;66:1250-6. Printed in USA. 1997 American Society for Clinical Nutrition

Predictingenergyneeds in ventilator-dependentcriticallyillpatients:effect of adjustingweightfor edema or

Mary E Cutts, Richard P Dowdy, Mark R Ellersieck@ and Thomas E Edes

by guest on July 23, 2015ajcn.nutrition.org

Dow

nloaded from

Percentage IBWMean relative measuredREEk//kg 159

Percentage ideal body weight (%)FIGURE1. Percentageidealbodyweighton dayof calorimetry.

subject was both overfat and overhydrated, normally hydratedweight was determined first and then this was used to determine obesity-adjusted weight.

Energy needs were predicted for all subjects by using boththe Harris-Benedict equation (HBE; I 8) as follows and thekilojoulesper kilogram strategy (KPK; I9).HBE, males: Energy = 66 + (13.7 X W) + (5 X H)

(6.7xA) (4)HBE, females: Energy = 655 + (9.5 X W) + (I .8 X H)

(4.6XA) (5)

where W is weight in kilograms, H is height in centimeters, andA is age in years. For both calculations, needs were predictedby using both actual weight on the day of calorimetry andadjusted weight (obesity-adjusted or normally hydrated weight,as appropriate for that subject).

For the purposes of applying stress factors (20) and predicting energy needs, subjects were categorized as being eithermoderately or highly stressed. Moderately stressed subjectswere those with one or two organ system failures, evidence ofmild-to-moderate infection, or both. Their energy predictionsused stress factors of 1050%above baseline HBE values andof 105146Id/kg. Highly stressedsubjectswere thosewithmultiple organ system failure, sepsis, or both. Their energypredictions used stress factors of 3070%above baseline HBEvalues and of 125167 kJ/kg. Energy predictions in eithercategory were lowered if subjects were extremely obese or atrisk for refeeding syndrome due to malnutrition. Presence ofdisease states such as cancer and chronic obstructive pulmonary disease were considered in setting goals. Energy predictions were intended to reflect REE because they were to be

TABLE 1Relative energy needs grouped by percentage ideal body weight (IBW)'

by guest on July 23, 2015ajcn.nutrition.org

Dow

nloaded from

1252 ci@mrs ET AL

11

100

80

60

40

20

0101129 130159 > 159

Percentage ideal body weight (%)FIGURE 2. Source of overweight: fat, edema, or both.

compared with REE measured determined by indirect calorimetry. For this reason, predictions were not set to promoteweight gain. Chart reviews were performed retrospectively andthese energy predictions were used for research purposes only.They were not used by the physicians and dietitians caring forthe patients. REE predictions using both strategies (HBE andKPK) and using both actual and adjusted weights were cornpared with measured REE.

Identification of organ failure

1) Cardiac failure was identified through documentation ofperipheral edema plus either positive hepatojugulovenousreflux, elevated jugulovenous pressure, pulmonaryedema, or left ventricular ejection fraction < 40%.

2) Hepatic failure was identified by a bilirubin concentration > 30 mg/L or alkaline phosphatase and transaminaseactivities more than twice normal.

3) Renal failure was identified by a doubling of serumcreatinine from the subject's baseline value and urineoutput < 50 mL/h for 24 h.

4) Respiratory failure was identified by the use of mechanical ventilation.

5) Infection with a systemic response was identified by asuspected infection with a white blood count 12 X 106cellslL, or temperature > 38 C,or sedimentation rate> 30 mm/h.

6) Sepsis was identified by the determination of infection,

120

100806040200

with both positive blood cultures and a drop in systolicblood pressure > 20 mm Hg.

Statistics

Data were collected from 110 subjects: 82 were heavier thantheir IBW and 28 were at lighter than their IBW. Each subjecthad one indirect calorimetry measurement (their first or only)entered in the study. For purposes of statistical analysis, subjects were grouped according to their percentage of IBW:< 101%, 101129%,130159%,and > 159%. A split-plotanalysis of variance (ANOVA) was conducted to look fordifferences between combinations of methods for predicting ormeasuring energy needs and groups of overweight subjects.The main plot was represented by groups of overweight subjects and the subplot included methods of determining REE andthe interaction of weight group X REE. (The six differentmethods of determining REE used were HBE using actualweight, HBE using adjusted weight, KPK using actual weight,KPK using adjusted weight, indirect calorimetry, and the energy prediction of the clinical dietitian caring for the patient.)Mean differences were determined by using Fisher' s leastsignificant-difference (LSD) method. Further ANOVA wasconducted by using a one-way ANOVA examining relativeenergy needs by percentage IBW group and diagnosis. Simpleregression analysis was performed for each method versus REEfor each percentage IBW group. The SAS computer softwareprogram (release 6. 11 of the SAS SYSTEM FOR WINDOWS;

C-0

< 101 101129 130159 > 159

Percentage ideal body weight (%)FIGURE 3. Summary of the similarity of Harris-Benedict (HBE)predictedresting energy expenditure (REE) to measured REE. HBE-Act, HBE

prediction using actual weight; HBE-Adj, HBE prediction using adjusted weight.

. Fat0 EdemaU Both

. HBE-ActU HBE-Adj

by guest on July 23, 2015ajcn.nutrition.org

Dow

nloaded from

Percentage IBWHB R2E-Adj SlopeKPK R2-AdjSlope101129%(n35)0.21370.57250.17250.3980130159%(n 31)0.47030.081760.45960.6776

PREDICTING ENERGY NEEDS IN CRITICAL ILLNESS 1253

C-0

U 200 UKPK-Act@@\-, 1@ E 0 C@ KPK-Adj

{@ @II@15O@j{Jt@ I

< 101 101129130159 > 159

Percentage ideal body weight (%)FIGURE 4. Summary of the similarity of kilojoules per kilogram (KPK)predictedresting energy expenditure (REE) to measured REE. KPK-Act, KPK

prediction using actual weight; KPK-Adj, KPK prediction using adjusted weight.

SAS Institute, Inc, Cary, NC) was used for all statisticalanalyses.

RESULTS

Characteristics of subjectsThere were 35 female and 75 male adult subjects. Their ages

ranged from 18 to 87 y. On the day of calorimetry, all subjectshad respiratory failure requiring mechanical ventilation. Thethree next most frequent diagnoses were, respectively, chronicobstructive pulmonary disease, infection, and diabetes mellitus(all subjects with diabetes mellitus were treated with eitherinsulin, oral hypoglycemic agents, or both). The percentageIBW of the subjects on the day of calorimetry is shown inFigure 1. The largest number of subjects was in the 10 1129%IBW group. The relative measured energy needs, in kilojoulesper kilogram body weight for the different weight groups, areshown in Table 1. Relative measured energy need is defined asthe measured energy need (determined by indirect calorimetry)divided by the actual body weight of the subject. Relativeenergy needs were progressively lower for greater percentageIBW. For the overweight subjects, the source of their excessweight is shown in Figure 2. The groups with higher percentage IBW had a greater proportion of their excess weight as bothfat and edema, rather than as fat or edema alone.

Comparison of predicted with measured resting energyneeds

The accuracy of the four prediction strategies was evaluatedas follows: 1) by the HBE using actual weight (HBE-Act), 2)

TABLE 2R2valuesandslopesfrom regressionanalysisof measuredrestingenergy expenditure (REE) and predicted REE'

by the HBE using adjusted weight (HBE-Adj), 3) by KPKusing actual weight (KPK-Act), and 4) by KPK using adjustedweight (KPK-Adj). These four strategies were evaluated bycomparing their predicted REE with measured REE as determined by indirect calorimetry (Figure 3 and Figure 4). Theaccuracy of the prediction strategy was evaluated by comparingpredicted needs with measured needs and expressing the resultas a percentage.Resultsof 100%would indicate accurateenergy goals, results > 100% would lead to overfeeding, andresults < 100% would lead to underfeeding. Predictions usingactual weight tended to result in energy goals that would leadto overfeeding in the more overweight groups, particularlywhen using the KPK prediction strategy. For overweight subjects, predictions using adjusted weight with the KPK prediction strategy were more accurate, whereas adjusted weight withthe HBE strategy led to underfeeding.

The R2 values and slope of the HBE-Adj and KPK-Adjprediction strategies from regression analysis of measured REEand predicted REE are shown in Table 2. This indicated thatboth strategies were able to explain > 45% of the variance inREE in the 130159%IBW group. The strength of the prediction strategy varied depending on the percentage IBW group towhich it was applied. The results of evaluating the similarity ofthe predicted REEs with the measured REEs, and looking for astatistical match, are shown in Table 3. Actual weight predictions matched for underweight subjects when using both prediction strategies. HBE-Act yielded a match for subjects 101129% IBW, as did KPK-Adj for subjects 130% IBW.

Effect of diagnosisIt was possible to evaluate the effect on energy needs of only

three diagnoses because of the limited numbers of subjects.The numbers of other diagnoses were below those needed forstatistical significance. All subjects had respiratory failure andthe three additional diagnoses evaluated were chronic obstructive pulmonary disease, infection, and diabetes mellitus (alldiabetic patients were treated with insulin, oral hypoglycemicagents, or both). The differences in measured REE on the basisof diagnosis are shown in Figure 5. No significant differenceswere observed in REE on the basis of the presence or absenceof all possible combinations of these diagnoses (P > 0.05). Thepresence of all studied disease states did not increase totalenergy needs above those of subjects having only respiratory

I IBW, ideal body weight; HBE-Adj, Harris-Benedict equationprediction using adjusted weight; KPK, kilojoules per kilogramprediction usingadjusted weight.

by guest on July 23, 2015ajcn.nutrition.org

Dow

nloaded from

PredictionstrategyWeightgroup

159%of IBW

(n =15)HBE-Act

HBE-AdjKPK-ActKPK-AdjYes

YesYes

NoNoNoNo

NoNoYesNo

NoNoYes

1254 Curl's ET AL

isting alterations in metabolism (23), or alterations in themetabolic response to injury that may be specific to obesepersons (24). However, the data in Figure 4 suggest that theproposed weight adjustment strategy, used with the KPK prediction strategy, may offer an improvement over use of actualweight, and help guard against the dangers of overfeeding inthe groups with higher percentages IBW. HBE-Adj led tosignificant underfeeding in the groups with the highest percentage IBW. Dickerson et al (25) found that obese, mildly tomoderately stressed patients achieved net protein anabolismand tissue healing when fed high-protein (> 2 g/kg IBW onaverage), low-energy (5 1.5% of measured REE on average)diets. This suggests that meeting REE may not always benecessary in obese, critically ill subjects.

Ireton-Jones and Turner (9) found that actual weight predicted energy expenditure more accurately than did IBW. Several factors may have contributed to the apparent conflictbetween the results of Ireton-Jones and Turner and those of thepresent study. The two studies used different methods of determining IBW: Ireton-Jones and Turner used 1959 Metropolitan Life Insurance Company standards to determine IBW,whereas the present study used adjusted weights, which arehigher than ideal weights. Additionally, Ireton-Jones and Turner's subjects were > 130% of IBW, whereas ours were allthose > 100% of IBW. Finally, a basic premise of the currentstudy is that arriving at a better estimate of lean mass willimprove the accuracy of prediction. This idea has been challenged recently (26, 27). These authors concluded that therelation between fat-free mass and resting metabolic rate is notfixed and thus cannot be used for prediction.

The adjusted-weight strategies were able to explain > 45%of the variability of REEs in the 130159%IBW group (Table2). This is a reasonably strong predictor, considering the manyother factors affecting energy needs in this population.

The similarity of the four prediction strategies to the measured REE is summarized in Table 3. This analysis was a

140120

@.100@ 80

! 60@ 40@ 20

1 I I I I I I

N C D I DC IC ID IDC

DiagnosisFIGURE6. Relativeenergyneeds(measured)basedondiagnosis.i

SEM. All subjects had respiratory failure. Group N differs from group D,and from group IDC, P < 0.05. REE, resting energy expenditure; N,neither chronic obstructive pulmonary disease, diabetes mellitus, nor infection, n 7; C, chronic obstructive pulmonary disease, n = 16; D,diabetes mellitus, n = 3; I, infection, n = 23; DC, diabetes mellitus andchronic obstructive pulmonary disease, n = 8; IC, infection and chronicobstructive pulmonary disease, n = 19; ID, infection and diabetes mellitus,n = 16; DC, infection, diabetes mellitus, and chronic obstructive pulmonary disease, n 18.

TABLE3Similarity of predicted resting energy expenditure (REE) to measuredREE'

I Fisher's least significant difference = 614 KJ. HBE-Act, HarrisBenedict equationprediction using actual body weight; HBE-Adj, HBEprediction using adjusted body weight; KPK-Act, kilojoule per kilogramprediction using actual body weight; KPK-Adj, KPK prediction usingadjusted weight.

failure. This analysis does not account for the effect of differences in body size on energy requirements. The effect ofdiagnosis on relative energy needs expressed as U/kg actualweight, as opposed to total energy needs, is shown in Figure 6.Here, two groups were found to differ from one group (P 0.05. N, neither chronic obstructive pulmonary disease, diabetesmellitus, nor infection, n = 7; C, chronic obstructive pulmonary disease,n = 16; D, diabetes mellitus, n = 3; I, infection, n = 23; DC, diabetesmellitus and chronic obstructive pulmonary disease, n = 8; IC, infectionand chronic obstructive pulmonary disease, n = 19; ID, infection anddiabetes mellitus, n = 16; DC, infection, diabetes mellitus, and chronicobstructive pulmonary disease, n = 18.

obstructive pulmonary disease).

-I-1@I

@-@

L

I

111Ii@1

L I I.@ I I I

by guest on July 23, 2015ajcn.nutrition.org

Dow

nloaded from

PREDICTING ENERGY NEEDS IN CRITICAL ILLNESS 1255

stricter test, asking if the predicted value matchedthe measured value. This analysis suggests that actual weight, witheither prediction strategy, is acceptable for underweight subjects, HBE-Act is preferable for subjects 101129%18W, andKPK-Adj is preferable for subjects 130% IBW. It is somewhat surprising that any strategy could match in the mostoverweightgroups,whoseenergyneedsmay be moredifficultto predict. beton-Jones and Francis (28) recommended use ofindirect calorimetry for the massively obese (> 170 kg).

Somewhat surprisingly, diagnosis did not appear to play amajor role in either total or relative energy needs (Figures 4 and5). In fact, having multiple diagnoses led to a decrease inrelative energy needs in two separate groups. This is somewhatcounterintuitive, and there is a tendency to think that the moresomeone has wrong with them, the greater their energy needsshould be. These results could be due to the inability to track allmedical conditions statistically. Critically ill patients are aheterogeneous group with multiple combinations of chronicand acute problems not easily segregated in adequate numbersfor statistical analysis. Whether a subject was in the early acutephase of their illness or the late convalescent phase could alsohave been a factor, because the phase of illness may affect themetabolic response. Individual variation is another possibilitybut it was not possible to evaluate the REEs of these subjectsbefore their illness. For the groups that did differ (Figure 6),those with lower relative energy needs had diabetes mellitus.Conversely, two groups with diabetes mellitus did not havesignificantly lower relative energy expenditures. This raises thequestion of whether diabetes mellitus may reduce energyneeds, reduce actual energy expenditure through reduced carbohydrate utilization, or reduce the computed energy expenditure as determined through indirect calorimetry, perhapsthrough some alteration in metabolism affecting carbon dioxidemetabolism. A comparison of indirect calorimetry with directcalorimetry results in diabetic patients could look at thisquestion.

Despite our results indicating that diagnosis did not have amajor effect on energy needs, clinicians may use a subject'sdiagnosis in applying stress factors for predicting energy needs.Liggett and Renfro (29) found that a group of critically illnonsurgical patients did not require the addition of stress factors to the results of the HBE. They refuted the idea that thesicker the patient, the more calories required. Some researchers have suggested that improvements in medical care mayattenuate the hypermetabolism of critical illness (30, 31).Cones and Nelson (32) also concluded that clinicians' estimation of stress may lead to overfeeding. Conversely, Brown et al(33) found a relation between increasingmeasuredenergyexpenditure and increasing severity of disease as measured bythe Acute Physiology and Chronic Health Evaluation(APACHE H) score.

Some authors have created new prediction equations basedon research with critically ill, mechanically ventilated subjects(9, 22). Use of these may be preferable to adjusting previouslyused formulas such as the HBE for this population or usingruleof thumbprediction strategies such as KPK.

The present study has provided insight but has not definitively answered the question of which weight should be used inpredicting energy needs in critical illness. These results alsosuggest that an increasing number of medical diagnoses maynot translate into a greatly increased REE. In summary, a

patient's weight status is of significance when predicting REEin critical illness. Clinicians may improve the accuracy of theirpredictions by noting factors such as hydration status, presenceof edema, and presence of adipose tissue when selecting aweight to use in prediction equations. The method we presented in this study generates an adjusted weight based on thesefactors. Adjusted weights may be preferable to actual weightsfor overweight patients to avoid the hazards associated withoverfeeding. Multiple diagnoses may not result in greatly increased energy needs. It is important to remember that aprediction is an educated guess, but should not replace continued monitoring of other clinical and metabolic variables. mdirect calorimetry, when available, remains the standard forassessing and reassessing needs in those patients for whomenergy balance is crucial.

REFERENCES

1. McClave SA, Snider HL. Understanding the metabolic response tocritical illness: factors that cause patients to deviate from the expectedpattern of hypermetabolism. New Horiz 1994;2:l3946.

2. Pinard B, Geller E. Nutritional support during pulmonary failure. CritCare Clin 1995;ll:70515.

3. Schlichtig R, SargentSC. Nutritional supportof the mechanicallyventilated patient. Crit Care Clin 1990;6:76784.

4. EdesTE. Nutrition supportof critically ill patients:guidelinesforoptimal management. Postgrad Med 199l;89:l93-.200.

5. Guenst JM, Nelson LD. Predictors of total parenteral nutrition-inducedlipogenesis. Chest l994;105:5539.

6. Makk UK, McClave SA, Creech PW, et al. Clinical application of themetabolic cart to the delivery of total parenteral nutrition. Crit CareMed l990;l8:l3207.

7. Wolfe RR, Herndon DN, Jaboor F, Miyoshi H, Wolfe M. Effect ofsevere burn injury on substrate cycling by glucose and fatty acids.N Engl J Med l987;3l7:4038.

8. FrankenfleldDC, OmertLA, BadellinoMM, Wiles CE, BagleySM,Goodarzi S. Correlation between measured energy expenditure andclinically obtained variables in trauma and sepsis patients. JPEN JParenter Enteral Nutr 1994;18:398403.

9. Ireton-JonesCS, Turner WW. Actual or ideal body weight: whichshould be used to predict energy expenditure? J Am Diet Assoc199l;91:1935.

10. Weissman C, Kemper M. Assessing hypermetabolism and hypometabolism in the postoperative critically ill patient. Chest 1992;l02:156671.

11. Vicari J, Watson P. Fagan D, Jasnowski J, Lanspa SJ. Measuredresting energy expenditure: relationship to APACHE II score andobesity. Nutr Res 1995:15:77784.

12. Astrup A, Thorbek G, Lind J, Isaksson B. Prediction of 24-h energyexpenditure and its components from physical characteristics and bodycomposition in normal-weight humans. Am J Clin Nutr 1990;52:77783.

13. Zurlo F, Larson K, Bogardus C, Ravussin E. Skeletal muscle metabolism is a major determinant of resting energy expenditure. J ClinInvest 1990;86:14237.

14. Ravussin E, Bogardus C. Relationship of genetics, age, and physicalfitness to daily energy expenditure and fuel utilization. Am J Clin Nutr1989;49:96875.

15. Bernstein RS, Thornton JC, Yang MU, et al. Prediction of the restingmetabolic rate in obese patients. Am J Clin Nutr l983;37:595602.

16. Davidson JK. Symposium: controlling diabetes mellitus with diettherapy. Postgrad Med l976;59:1 1422.

17. Wilkens K. Adjustment for obesity. ADA Renal Practice Group Newsletterl984;3(winter):6.

by guest on July 23, 2015ajcn.nutrition.org

Dow

nloaded from

1256 CUTFS ET AL

18. Harris JA, Benedict FG. A biometric study of basal metabolism inman. Washington, DC: Carnegie Institute, 1919. (Publication no. 279.)

19. Hopkins B. Assessment of nutritional status. In: Gottschlich M,Matarese L, Shronts E, eds. Nutrition support dietetics core curriculum. 2nd ed. Silver Spring, MD: A.S.P.E.N. Publications, 1993:1570.

20. Van Lanschot J, Feenstra B, Vermeij C, Braining H. Calculationversus measurement of total energy expenditure. Cnt Care Med 1986;14:11,9815.

21. Foster GD, Knox IS, Dempsey DT, Mullen IL. Caloric requirementsin total parenteral nutrition. J Am Col Nutr 1987;6:23153.

22. Swinamer DL, Grace MG, Hamilton SM, Jones RL, Roberts P. KingEG. Predictive equation for assessing energy expenditure in mechanically ventilated critically ill patients. Crit Care Med 1990;18:65761.

23. Leibel RL, Rosenbaum M, Hirsch J. Changes in energy expenditureresulting from altered body weight. N EngI J Med 1995;332:62l8.

24. Jeevanandam M, Young DH, Schiller WR. Obesity and the metabolicresponse to severe multiple trauma in man. J Chin Invest1991;87:2629.

25. Dickerson RN, Rosato EF, Mullen IL. Net protein anabolism withhypocaloric parenteral nutrition in obese stressed patients. Am J ClinNutr l986;44:74755.

26. Carpenter WH, Poehiman ET, O'Connell M, Goran MI. Influence of

body composition and resting metabolic rate on variation in totalenergy expenditure: a mets-analysis. Am J ChinNutr 1995;6l:4l0.

27. Weinsier RL, Schutz Y, Bracco D. Reexamination of the relationshipof resting metabolic rate to fat-free mass and to the metabolicallyactive components of fat-free mass in humans. Am J Clin Nutr1992;55:790-4.

28. Ireton-Jones CS, Francis C. Obesity: nutrition support practice andapplication to critical care. Nutr Clin Pract l995;l0:1449.

29. Liggett SB, Renfro AD. Energy expenditures of mechanically ventilated nonsurgical patients. Chest 1990;98:6826.

30. Elia M. Changing concepts of nutrient requirements in disease: implications for artificial nutritional support. Lancet 1995;345:127984.

31. Boulanger BR, Nayman R, McLean RE, Phillips E, Rizoli SB. Whatare the clinical determinants of early energy expenditure in criticallyinjured adults? J Trauma l994;37:96974.

32. Cortes V, Nelson LD. Errors in estimating energy expenditure incritically ill surgical patients. Arch Surg 1989;l24:28790.

33. Brown PE, McClave SA, Hoy NW, Short AF, Sexton LK, Meyer KL.The acute physiology and chronic health evaluation II classificationsystem is a valid marker for physiologic stress in the critically illpatient. Crit Care Med 1993;2l :3637.

by guest on July 23, 2015ajcn.nutrition.org

Dow

nloaded from