Accuracy of predictive equations for resting energy expenditure estimation in mechanically ventilated Thai patients

Thirty-one measured REEs were obtained from 25 ventilated Thai patients. One patient was excluded due to air leakage during IC measurement (Figure 1). The eligible patients had a median age of 64.5 [43.5, 75.5] years, a median height of 160 [151, 167] cm, and a median body mass index (BMI) of 22.95 [18, 34.1] kg/m². The baseline characteristics of 24 eligible patients are summarized in Table 2.

Figure 1.

The patients had a median measured REE of 1553 [1126.75, 2065.75] kcal/d and an average measured REE of 23.5 kcal/kg actual body weight/d or 29.5 kcal/kg ideal body weight (IBW)/d. The median estimated REE values, calculated by 14 equations, ranged from 1180.5 kcal/d to 1840.4 kcal/d. We compared the accuracies of 14 equations (Table 3). The accuracies of these equations varied from 6.7% to 48.1%, with the mean differences between the calculated and measured REEs of −320 kcal/d to 440 kcal/d and the mean bias percentages ranging from −13% to 31.8%. The total numbers analyzed in each equation differed noticeably. Some parameters used to calculate REE in some equations might not have been recorded, and some periods might not have been incorporated in the estimations for some equations. To minimize these confounders, we performed sensitivity analysis, where all available data were used (Table 4). In the sensitivity analysis, the accuracy of the equations varied from 9.1% to 40.9%, with the mean differences between the calculated and measured REEs of −272.4 kcal/d to 353.5 kcal/d and the mean bias percentages ranging from −10.2% to 32.4% (Table 4). Notably, the results of the sensitivity analysis and the total results were similar. Penn State 2010 remained the most precise equation, while Ireton-Jones 1992 remained the least accurate. The ACCP, Harris–Benedict equation (HBE), Mifflin–St. Jeor (MSJ), and Penn State 2003 (HBE) also tended to underestimate calorie requirements. Besides, the 1.2×HBE, 1.5×HBE, Ireton-Jones 1992 and 2002, Penn State 2003 (MSJ), Penn State 2010, Swinamer 1990, Faisy, Brandi 1999, and 25 kcal/kg equations tended to overestimate REEs.

In subgroup analysis, the Penn State 2010 equation provided the most accurate estimation of calorie need in obese patients (BMI ≥ 30 kg/m²), while the HBE provided the most reliable prediction of REE in non-obese patients (Figure 2).

Figure 2.

Accordingly, on Bland–Altman plots (Figure 3), there was a poor agreement between the calculated and measured values. Contrastingly, the calculated values had moderately to strongly positive correlations with measured values (r = 0.445–0.778, Table 5). Additionally, in the ICC analysis, 1.2×HBE, MSJ, Ireton-Jones 2002, Penn State 2003 (MSJ), and Penn State 2010 had good reliability, while HBE, 1.5×HBE, Ireton-Jones 1992, Penn State 2003 (HBE), Swinamer 1990, Faisy equation, Brandi 1999, and the 25 kcal/kg actual body weight had moderate reliability. Only ACCP had poor reliability (Table 6).

Figure 3.

Our study demonstrated that no predictive equations accurately estimated calorie demand in ventilated Thai patients, and none had clinically acceptable accuracy, despite the fact that calculated values had moderately to strongly positive correlations with measured values and moderate to good reliability in the ICC analysis for inter-rater agreement measurement.

However, HBE more reliably predicted REE in under-weight (BMI < 18.5 kg/m²) and non-obese (BMI < 30 kg/m²) ventilated patients with the accuracies of 66.7% and 44.4%, respectively, while the Penn State 2010 equation more accurately determined energy need in obese patients with the accuracy of 54.5%. Additionally, the 1.2×HBE, 1.5×HBE, Ireton-Jones 1992 and 2002, Penn State 2003 (MSJ), Penn State 2010, Swinamer 1990, Faisy, Brandi 1999, and 25 kcal/kg equations tended to overestimate REE. Contrastingly, ACCP, HBE, MSJ, and Penn State 2003 (HBE) tended to underestimate energy requirements. The results were consistent with those in a systematic literature review [17]. Tatucu-Babet et al. [17] showed the variability in the accuracy of each equation. The Ireton-Jones, Swinamer, Faisy, and Brandi equations frequently overestimated energy need, while the HBE and Penn State equations were likely to underestimate REE.

Accordingly, the discrepancy between measured and predictive REEs could be explained by the following reasons. First, most equations use body weight (BW) as the basis in their estimation of calorie requirement. Practically, it is difficult to obtain the precise actual weight in mechanically ventilated patients. Commonly, excessive total body water from fluid resuscitation or decreased muscular mass confounds the real values. Second, dynamic changes in clinical conditions result in variable metabolic demands, which depend on phases of critical illness. Furthermore, most equations are validated for healthy volunteers, spontaneously breathing patients, and non-Asian populations, who possibly have metabolic differences from Thais. Finally, there are several factors pertaining to the ICU that might be expected to confound measured values, for example, recent adjustment of the ventilator setting before IC measurement [18], or measurement of calories at different points of time or under the prevalence of different feeding conditions, including fasting, intermittent feeding, or continuous feeding.

Although the predictive equations were initially developed in the western populations, these equations did not precisely estimate REE in mechanically ventilated western patients. Similarly, all of the equations had poor predictive precision in mechanically ventilated Thai patients. Therefore, REE should be assessed using IC.

However, there were some limitations in this study. First, feeding conditions at the time of IC measurement were not recorded. Additionally, the REE was measured during a stable condition, which did not represent energy need in an active phase of illness. Furthermore, there was a selective bias resulting from the fact that physicians usually performed IC in patients with extreme BMI or uncertain actual BW. Finally, the study lacked statistical power due to a small sample size.

The accuracy of the predictive equations was poor, and thus this method of calorie-need estimation was unable to replace IC in the determination of calorie requirement in mechanically ventilated Thai patients.