Poor Agreement between Predictive Equations of Energy Expenditure and Measured Energy Expenditure in Critically Ill Acute Kidney Injury Patients

Abstract

Background: There are multiple equations for predicting resting energy expenditure (REE), but how accurate they are in severe acute kidney injury (AKI) patients is not clear. Our aim was to determine if predictive equations for estimated REE accurately reflect the requirements of AKI patients. Methods: We included in this prospective and observational study AKI patients AKIN-3 assessed by indirect calorimetry (IC). Bland-Altman, intraclass correlation coefficient and precision (percentagem of predicted values within 10% of measured values) were performed to compare REE by equations with REE measured by IC. Results: IC was applied in 125 AKI patients. The mean age was 62.5 ± 16.6 and 65.6% were male. Mean REE measured was 2,029.11 ± 760.4 kcal/day. There were low precision, and poor agreement between measured and predicted REE by the Harris-Benedict (HB), Mifflin, Ireton-Jones, Penn state, American College of Chest Physicians, and Faisy equations. HB without using injury factor was the least precise (18% of precision). Modified Penn state equation had the best precision, although the precision rate was only 41%. For all equations, the limits of agreement range were large leading to the potential under or overfeeding of individual patients. Conclusion: None of these equations accurately estimated measured REE in severe AKI patients and most of them underestimated energy needs.

Proposal of a new equation for estimating resting energy expenditure for acute kidney injury patients on dialysis. A Machine Learning Approach.

“…Our study found that the REE estimated by the Harris and Benedict formula [19] was signi cantly lower than that measured by IC. This nding corroborates the indication not to use this formula in critically-ill patients and in patients with AKI [11,14,20,21] and the need to propose a new equation for AKI on dialysis. This study aimed to develop and validate predictive equations for REE in severe AKI patients using a machine learning approach.…”

“…Modi ed Penn state equation had the best precision, although the precision rate was only 41%. As a conclusion, none of these equations accurately estimated measured REE in severe AKI patients on dialysis and most of them underestimated energy needs [14]. Recently more sophisticated models that used machine learning were applied in clinical practices resulting and better predict models [15].…”

Daily variability of resting energy expenditure in acute kidney injury patients on dialysis

“…This finding corroborates the indication not to use this formula in critically-ill patients, 7,[16][17][18] and in patients with AKI. 19 The best way to assess the energy needs of critically-ill patients with AKI is by means of IC. 19 In the present study, REE increased as measured by IC on the fifth day of follow-up, compared to days 2 and 3.…”

“…19 The best way to assess the energy needs of critically-ill patients with AKI is by means of IC. 19 In the present study, REE increased as measured by IC on the fifth day of follow-up, compared to days 2 and 3. However, this Significant increase was seen only in REE without normalization for weight.…”

Proposal of a new equation for estimating resting energy expenditure for acute kidney injury patients on dialysis. A Machine Learning Approach.

“…Therefore, many predictive equations exist for predicting resting energy expenditure (REE), but the accuracy of these equations for estimating caloric requirements of critically ill patients is unclear [7][8][9][10][11][12][13]. Goes et al evaluated if nine different standard predictive equations for energy expenditure could accurately re ect the energy requirements of critically ill, mechanically ventilated AKI patients [14]. There was low precision and poor agreement between measured and predicted REE by the Harris-Benedict (HB), Mi in, Ireton-Jones, Penn State, American College of Chest Physicians and Faisy equations.…”