Some researchers have demonstrated that differences in nutritional risk appear to be linked to varying the responsiveness of subjects to nutritional support. Sim J et al.¹² found that among patients with high malnutrition risk, as defined by the modified Nutrition Risk in Critically ill score of ≥ 5 or BMI of < 18.5 kg/m², the “Early” group (who received PN within 48 h) had higher caloric adequacy, and had lower rates of 30-day mortality and in-hospital mortality than the “Usual” group. Whereas, there was no significant difference in mortality between the two groups among patients with low malnutrition risk. Along the same lines, reaching caloric goal was observed to has a beneficial effect on mortality risk in patients with lower respiratory tract infection at NRS 2002 score ≥ 3¹³. In this trial, we only included septic patients with an NRS 2002 score ≥ 3. After adjusting for key potential confounders, the Cox regression models revealed that an increase in the intakes of energy during the late period of acute phase (days 3–7) was associated with improved survival in septic patients, and this finding also held true in the subgroup of patients with septic shock. However, the association between higher energy intake and mortality was not observable in patients with non-shock. Although Kaplan-Meier survival analysis revealed a dose-response trend in mortality rates across energy intake groups, this trend did not achieve statistical significance (p = 0.422). This discrepancy between the observed clinical trend and statistical significance may be explained by several factors. The substantially lower mortality in non-shock versus shock patients (12.0% vs. 42.7%, P < 0.001) resulted in limited statistical power due to both the small number of events and a potential ceiling effect, where the already low baseline mortality in this population may have left little opportunity for nutritional interventions to exhibit additional survival benefits. Pathophysiologically, septic shock patients exhibit profound metabolic dysfunction with accelerated catabolism, rendering them vulnerable to inadequate nutrition¹⁴. Conversely, non-shock patients generally maintain preserved metabolic homeostasis with adequate endogenous reserves, thereby reducing their reliance on exogenous nutritional supplementation for survival outcomes. Additionally, the protective effects of energy intake may manifest over longer time periods in non-shock patients, beyond the 28-day observation window of this study. Further investigation is warranted to elucidate the potential mechanisms driving the aforementioned divergence and to clarify whether the observed trends in non-shock patients represent clinically meaningful effects.

Admittedly, the recent NUTRIREA−3 trial in ventilated adults with shock has shown no benefit of early full nutritional support, and early standard calorie intake (25 kcal/kg/day) did not decrease mortality compared with calorie restriction (6 kcal/kg/day). Yet the truth is that this study did not stratify patients by baseline nutritional status at ICU admission or adjusted for it post hoc¹⁵. An international multicenter observational study of nutritional intake in critically ill patients¹⁶ demonstrated a benefit from aggressive provision of calorie intake solely for patients with low BMI (< 25 kg/m²) and high BMI (≥ 35 kg/m²). Compared to our study, patients enrolled in the NUTRIREA−3 trial generally had higher BMI levels, with a mean BMI of 26.7 kg/m² (23.0, 31.1) and 27.0 kg/m² (23.0, 31.5) in the low and standard groups respectively; thus fewer individuals may have derived benefits from increased feeding regimens. Another randomized trial evaluated initial trophic feeding (about 400 kcal/day) versus full enteral feeding (about 1300 kcal/day) for the first 6 days in 1000 patients with acute lung injury, revealing that full-feeding was not significantly associated with lower 60-day mortality. However, close inspection of the study population found that patients had an overly healthy BMI (about 30.0 kg/m²)¹⁷. In a large multicenter RCT comparing permissive underfeeding (40–60% of calculated target calories) with standard enteral feeding (70%−100% of calculated target calories), there were no differences in ICU and 90-day mortality between the two groups. And similarly, these patients had a mean BMI of approximately 29.0 kg/m²¹⁸. This discrepancy may be related to the varying conclusions drawn by these studies in contrast to ours regarding the associations between enhanced energy intake and patient outcomes.

Another multicenter, double-blind, randomized trial of 3957 patients in 46 ICUs over 3 months confirmed that energy-dense EN (1.5 kcal per milliliter) showed no harms but also no benefits compared with routine EN (1.0 kcal per milliliter), as the first 28 days high-energy feeding did not affect 90-day mortality or adverse events. Subgroup analysis revealed that the association between the dose of nutrition support and 90-day mortality was not modified by their level of BMI¹⁹. But in fact, the energy-dense group included many patients receiving “overaggressive feeding” and their average energy supply on day 2 has exceeded 30 kcal/kg/day. A similar situation was observed in the NUTRIREA−3 trial standard group, where patients were given about 25 kcal/kg/day for energy delivery starting on the second day¹⁵.

Indeed, this nutritional management goes against the physiology needs of critically ill patient. Extreme physical stress triggers a catabolic state, gluconeogenesis, and endogenous substrate production, with the aim of maintaining a continuous glucose supply to vital organs. Premature full feeding may result in overfeeding, because as long as infammation persists, the endogenous calories production remains non-inhibitable even if feeding is resumed in the first several days after the onset of critical illness, and particularly during the initial 72 h²⁰. Overfeeding has been proven to be detrimental in a retrospective cohort study based on indirect calorimetry²¹. In addition, inappropriate dosing and timing of nutrition support in critically ill patients may lead to mitochondrial damage, increased reactive oxygen species production, and eventually cell death via apoptosis or necrosis²².

Current consensus recommends a cautious and progressive ramping-up feeding approach during the first week, increasing in calorie target by 25% daily for the first 3 days and then gradually progressed to target by day 4²³. Therefore, these two studies do not prove that early or high caloric feeding is harmful, but rather confirm that too much energy should not be delivered too early to critically ill patients. Cha JK et al. implemented a stepwise feeding approach, avoiding “full feeding” within the first 4 days, in 834 patients with sepsis and found that higher energy intake was associated with lower 30-day mortality (adjusted HR, 0.94; p = 0.003)²⁴. Similarly, we took care to gradually increase caloric intake in alignment with the physiological needs of critically ill patients, reaching caloric target by day 5. Thus, it is evident that a progressive feeding approach is more likely to show positive effects of greater energy intake on outcomes.

The benefits of early EN has been confirmed more than once. In a large observational study that enrolled 1174 patients who required at least one vasopressor to support blood pressure, early EN (< 48 h, n = 707) was linked to reduced hospital mortality (34% vs. 44%, P < 0.001) and ICU mortality (22.5% vs. 28.3%, P = 0.030) compared to late EN²⁵. Existing guidelines suggest that EN should be initiated early in the first 24 to 48 h after ICU admission in sepsis, except for shock is uncontrolled^6,11. In this study, EN was the first choice of nutritional support and more than half of the patients received early EN, nevertheless early EN showed no correlation with a survival advantage. Moreover, as tolerance to EN is not always well due to gastrointestinal intolerance, especially in patients with septic shock. Accordingly, we do not exclude early individualized supplemental PN to avoid a massive energy deficit in cases of insufficient EN.

Previously, PN was regarded as harmful to critically ill patients, especially in terms of infectious complications^26,27. Nevertheless, the scientificity or clinical logic of this conception has been questioned over the past decade. Three recent large RCTs in adults showed no difference in the mortality and infectious complications when PN provided at isocaloric doses as EN, suggesting that harms of PN are more likely to be explained by the relative overfeeding, rather than by the central line^28,29,30. In our study, overfeeding prevalence (> 110% of 25 kcal/kg/day target) was consistently low throughout the study period. Daily rates in septic shock patients ranged from 1.4% (day 1) to 15.7% (day 7), and in non-shock sepsis patients, from 1.6% (day 1) to 20.1% (day 6). As overfeeding was largely avoided, we observed that the use of EN, PN, or a combination thereof as the primary feeding regimen in the first week showed no correlation with 28-day mortality, but supplemental PN and early EN were helpful to achieve target energy intake. These results suggest that the total dose of caloric intake rather than the route of nutrition delivery correlates with mortality in patients with sepsis, and PN could be considered as a viable alternative treatment when contraindications or poor tolerance to EN are present.

Two previous studies have proved that early energy insufficiency in patients receiving prolonged mechanical ventilation in the ICU heighten the risk of Staphylococcus aureus ventilation-associated pneumonia infections and methicillin-resistant Staphylococcus aureus bloodstream infections^31,32. In a multicenter, prospective cohort study involving patients awaiting abdominal surgery, out of 120 patients with an NRS 2002 score ≥ 5, those who received sufficient preoperative nutrition therapy (at least 10 kcal/kg/day) for a duration of 7 days showed a reduced incidence of infectious complications compared to controls who received inadequate nutrition therapy (16.3% vs. 33.8%, P = 0.040)³³. In our study, the absence of association between energy intake and secondary infections in shock patients may be attributed to the high early mortality rate, with 60 shock patients dying within 12 days. Given that the median time to secondary infection was 12.0 days (9.0–15.0), many shock patients succumbed before reaching the typical window for secondary infection onset. This competing risk may have masked the potential protective effects of adequate nutrition. Conversely, non-shock patients had longer survival durations, allowing for the development of secondary infections. We observed that achieving at least 50% of the caloric target during the late period of acute phase was associated with a reduced risk of fungal and intra-abdominal infections in non-shock patients. However, overall, higher energy intake did not demonstrate a clear benefit in reducing infection-related complications. This finding was in discordance with prior researches. Two primary underlying reasons may be linked to this phenomenon. First, nosocomial infections were predominantly caused by Gram-negative bacteria, while Gram-positive bacterial infections were less prevalent in this study. Second, the caloric intake remained inadequate, even in the highest energy intake group. The strongest evidence supporting this comes from a RCT assessing whether achieving 100% of the energy target from days 4 to 8 after ICU admission could optimize clinical outcomes. The trial showed that the supplemental PN group had a higher mean energy delivery of 28 kcal/kg/day compared to 20 kcal/kg/day in the EN group, and fewer patients developed nosocomial infections in the supplemental PN group between days 9 and 28 (27% vs. 38%, P = 0.0338)³⁴. In conclusion, the clinical and experimental evidence corroborating the association between caloric deficit and infection remains limited. This latter study warrants repetition.

Our study revealed a notably high baseline fungal infection rate (97/602, 16.1%), surpassing previously reported rates in China (10.5%)³⁵ and across Asia (13.3%)³⁶. This discrepancy can largely be attributed to our inclusion criteria, which selected sepsis patients requiring ICU stays of at least seven days, combined with a majority of patients being transferred from general wards with extended pre-enrollment hospitalization. These characteristics suggest that our cohort represents a more severely ill subset of sepsis patients, potentially limiting the generalizability of our findings to broader sepsis populations. Beyond these inherent population traits, several additional study limitations merit clarification. First, as this is a retrospective observational study, our findings may be subject to selection bias and can only indicate associations rather than causal relationships. These results should therefore be validated through RCTs. Second, we used calculated rather than measured calorimetry to estimate daily caloric requirements. Third, although survivors exhibited higher protein intake, multivariate logistic regression indicated that protein dose was not an independent predictor of 28-day mortality. Lastly, monitoring nutritional intake solely within the first seven days may be insufficient for drawing conclusions about 28-day mortality, given the multifactorial nature of prognosis in sepsis patients. Additionally, our records did not distinguish between oral and tube feeding proportions.