Providing adequate nutrition in critically ill patients is important for improving clinical outcomes. Underfeeding can lead to infections and muscle weakness, while overfeeding also may result in complications such as infections, hyperglycemia, and immune deficiencies [1,2]. Therefore, several guidelines recommend using indirect calorimetry to determine the exact energy requirements of critically ill patients [3–5]. Additionally, when indirect calorimetry is not feasible, it is recommended to use the published predictive equations or a simplistic weight-based equation [4,6]. However, several studies have reported that the predictive equations are less accurate compared to indirect calorimetry [7–9]. Especially, in septic patients who require resuscitation, the previously used predictive equations are not accurate, and to determine the exact energy requirements it is necessary to use indirect calorimetry whenever possible [8,10–12].

Sepsis is the main cause of admission to an intensive care unit (ICU). In patients with sepsis, the metabolic response varies depending on the clinical course, so it is necessary to understand changes at each time point and provide appropriate nutritional support. The European Society for Clinical Nutrition and Metabolism (ESPEN) guidelines recommend reflecting the changes in metabolism at the current point in time. The first 1 to 2 days post-ICU admission are defined as the early period of the acute phase, Days 3 to 7 as the late period of the acute phase, and beyond Day 7 as the late or post-acute phase [13]. The ESPEN guidelines recommend hypocaloric nutrition (not exceeding 70% of energy expenditure) during the early period of the acute phase. It recommends that there is a gradual increase in energy delivery to 80%–100% of the measured energy expenditure, 3 days post-ICU admission [4,13]. Although several studies have analyzed energy requirements at various time points [10,14], there are few studies that have prospectively analyzed changes in the REE and respiratory quotient (RQ) as calories are supplied according to current guidelines. Therefore, this study investigated alterations in energy requirements based on the clinical course of sepsis in patients admitted to the surgical ICU using indirect calorimetry.

A total of 33 sepsis patients were enrolled in this study (Supplementary Material 1). The average age of the enrolled patients was 70.3 ± 10.6 years, accompanied by hypertension (69.7%), and malignancy (72.7%). The most common cause of sepsis was intra-abdominal infection (78.7%). Upon admission to the ICU, 18 patients (54.5%) were assessed as having a normal nutritional status, while 15 patients (45.5%) were assessed to be malnourished. Mortality rates were 21.2% (7 patients; Table 2).

The measured REE was highest on Day 1 (24.29 ± 3.72 kcal/ kg/day), and a statistically significant decrease was observed on Day 3, (22.42 ± 3.72 kcal/kg/day; p = 0.045). On the 7^th day, REE slightly increased to 23.26 ± 5.78 kcal/kg/day, but this change was not statistically significant (Figure 2A).

The RQ was measured on Day 1, (0.69 ± 0.06) and decreased to its lowest value (0.67 ± 0.05) on Day 3. On Day 7, it increased statistically significantly to 0.71 ± 0.06 (p = 0.005; Figure 2B). During the 1^st and 2^nd day post-ICU admission, an average of 396.5 kcal was provided, and between the 3^rd and 6^th days, the average energy intake was increased to 945.3 kcal.

The calories actually supplied to the patients from Day 1 to Day 7 post-ICU admission were analyzed (Figure 3). On Day 1, 336.9 kcal were supplied, and the amount was gradually increased to 1331.5 kcal by Day 7. As there were many patients who had undergone major abdominal surgery, parenteral nutrition was provided more often than enteral nutrition.

The REE values measured by indirect calorimetry on Days 1, 3, and 7 and the energy expenditure values obtained by the 4 predictive REE equations were compared using ICC analysis (Table 3) and the Bland-Altman plot (Figures 4–6). On Day 1, the REE results measured by the weight-based equation (29.4 ± 295.5 kcal) and the Penn State equation (−44.5 ± 216.9 kcal) showed little difference from the REE values measured by indirect calorimetry. The Penn State equation showed the highest correlation with an ICC value of 0.71 (p < 0.001). Even on Day 3, the Penn State equation showed the closest results (3.1 ± 231.9 kcal), with the ICC showing the highest agreement at 0.65 (p < 0.001). When comparing the values measured on Day 7, the Penn State equation was the closest (−37.9 ± 336.6 kcal) and the ICC also showed the highest agreement at 0.53 (p = 0.002). The predictive REE equation analyzed on Day 1 had the highest overall agreement and the ICC values decreased over time. At all 3 time points, the Penn State equation showed the closest results between predictive and measured REE.

In this study, changes in the REE and RQ values obtained using indirect calorimetry at each time point of sepsis, specifically during the early and late periods of the acute phase, and the post-acute phase, were analyzed. The ESPEN guidelines recommend evaluating and providing appropriate energy targets using indirect calorimetry, whenever possible, after the acute phase of critical illness [4,13]. It has been reported that full nutritional support increases the risk of overfeeding because endogenous energy production in the early stages of sepsis meets the caloric requirements needed during this phase [19,20]. Consequently, it is recommended that caloric intake is gradually increased from the 3^rd day post-ICU admission [4,19]. In our patients, the REE measured by indirect calorimetry was highest on Day 1 post-admission and remained similar after decreasing on Day 3. This pattern can be attributed to catabolism which peaks initially, leading to the highest REE values, which then stabilize as the patient enters a plateau period. In the study by Israfilov et al [8], measurements were made only at 2 points: within the acute phase (between 24 and 48 hours) and within the recovery phase (between 72 and 120 hours). Similar to our findings, their measurements also appeared to be high initially and then decreased, indicating a trend in the metabolic response to sepsis.

The RQ exhibits a similar trend on Day 1 and Day 3, but increases significantly on Day 7 as nutrition is provided. This indicated that the RQ reflects the number of calories supplied, demonstrating its use in assessing the metabolic response to nutritional intake over time. Expert recommendations on personalized nutrition propose that measurements should not be used too early post-ICU admission, before patients are adequately resuscitated (post-ICU Day 3) [21,22]. The highest REE value measured on Day 1 in this current study reflects these recommendations.

Several studies have compared the REE values measured by indirect calorimetry and predictive equations. However, most studies have been retrospective, and many studies did not use specific protocols for comparing periods of treatment post-admission to the ICU [6,9,19,23]. Many studies have recommended the use of indirect calorimetry because significant discrepancies were observed in the REE values between indirect calorimetry and those obtained by predictive equations [4,13,19,24]. In this current study, a markedly poor correlation was observed between the REE values measured using indirect calorimetry and those estimated using predictive equations. Among the 4 predictive equations, the Penn State 2003 equation, which incorporates dynamic parameters reflecting the patient’s current physiological state, consistently demonstrated the greatest concordance with measured REE at each time point. The Harris-Benedict and Ireton-Jones equations, which utilize static parameters such as the patient’s weight and height, showed a lower concordance compared with the Penn State equation. Notably, all 4 predictive equations exhibited the highest concordance immediately at ICU admission, with overall ICC declining over time. Lee et al [25] compared the REE values measured using indirect calorimetry within 36 hours, immediately after liver transplantation surgery, with the REE values obtained using predictive equations. Consistent with our findings, the Penn State 2003 equation demonstrated the highest concordance. A retrospective study by Waele et al [7] compared indirect REE values obtained from using calorimetry and predictive equations in critically ill patients and observed that the Penn State 2010 equation showed the highest agreement. These results suggest that predictive equations that incorporate dynamic parameters such as body temperature or tidal volume, which reflect the patient’s current condition, may be accurate and thus preferable. Panitchote et al [14] prospectively measured REE values using indirect calorimetry at 24, 48, and 72 hours in patients with sepsis and compared the results with those obtained using predictive equations such as Harris-Benedict, Ireton-Jones, and weight-based equations. Despite the time difference compared with the current study, this study also reported the highest concordance when measurements were taken within the first 24 hours and this decreased over time, with an overall low ICC < 0.5. The present study showed similar results, with the highest agreement within the first 24 hours. This is attributed to the significant changes in body weight caused by the large volumes of fluid administered during the resuscitation period which increasingly affects the measurements over time. Although predictive equations may be useful at the beginning of ICU admission, their accuracy diminishes over time thus, whenever possible, indirect calorimetry is necessary.

This study has several limitations. Firstly, this study was powered but the small sample size may limit generalizability of the findings. The small cohort also restricted the capacity to evaluate the effect of nutritional support on clinical outcomes. Secondly, the study primarily involved surgical critically ill patients with a high incidence of abdominal sepsis and related postoperative complications such as anastomosis leakage or high-output stoma, which affect administration of appropriate enteral nutrition. These factors may not accurately represent the broader population of patients with sepsis. Thirdly, continuous renal replacement therapy was administered to 13 (39.4%) patients. Although recent guidelines suggest that continuous renal replacement therapy has minimal effects on the reliability of indirect calorimetry measurements, concerns remain about its potential effects on CO₂ removal and nutrient loss which could introduce additional biases into the study’s findings [26,27].

In conclusion, consistent with current guidelines on indirect calorimetry, this study revealed significant discrepancies between the results obtained from indirect calorimetry and those derived from predictive equations. This finding emphasizes the need for further evaluations of energy requirements using indirect calorimetry to ensure accurate nutritional assessments. If indirect calorimetry is not feasible, predictive equations that incorporate dynamic parameters, such as the Penn State equation, are recommended. However, these predictive equations show the highest concordance at the beginning of ICU admission, and this agreement decreases over time.