Early high protein intake is associated with low mortality and energy overfeeding with high mortality in non-septic mechanically ventilated critically ill patients

The septic cohort consisted of 117 patients (14%) admitted with sepsis. Hospital mortality was significantly higher in septic patients than in non-septic patients (48.7% versus 33.9%, P = 0.003; Figure 2). The APACHE II score was higher as well (25.4 versus 22.6, P <0.001) (Table 1). Logistic regression analysis showed that mortality was not related to protein intake, energy overfeeding or APACHE II score in the septic cohort (Table 2).

Patient characteristics and nutritional data of non-septic overfed patients (n = 307) and non-septic non-overfed patients (n = 419) are shown in Table 1. In the non-septic cohort hospital mortality was not significantly higher in the day-4 overfed patients than in the non-overfed group (36.4% versus 32.1%, P = 0.234), the APACHE II scores were similar and energy intake in the non-overfed group was only 71% of measured EE. Figure 3 shows the cumulative energy deficit over the first 4 days of ICU stay (n = 726), with worst hospital mortality outcome in the overfed group (P = 0.053).

In this non-septic cohort (n = 726), logistic regression analysis demonstrated that the day-4 protein intake group (odds ratio (OR) = 0.80, 95% CI 0.67, 0.95, P = 0.011), day 4 overfeeding (OR = 1.89, 95% CI 1.19, 3.02, P = 0.007), and APACHE II score (OR = 1.04, 95% CI 1.01, 1.06, P = 0.001) had significant independent impact on mortality (Table 2). Thus, high day-4 protein intake was related to lower mortality in non-septic patients, while day-4 overfeeding and higher APACHE II score were related to higher mortality. The day-4 protein intake group was not related to mortality in the non-septic overfed group (Table 2).

In patients who were not septic and not overfed (n = 419), the higher protein intake group was associated with lower mortality (Table 3). Hospital mortality was 36.8%, 35.0%, 26.5%, and 19.1% for the <0.8, 0.8 to <1.0, 1.0 to- <1.2, and ≥1.2 g/kg protein-intake groups respectively (P = 0.033). Hospital mortality was 34.5% for day-4 protein intake <1.2 g/kg versus 19.1% for day-4 protein intake ≥1.2 g/kg (P = 0.015; Figure 4). Regression analysis with dummies for protein intake groups showed that the effect of protein was only significant at a protein intake level of ≥1.2 g/kg (OR = 0.42, 95% CI 0.21, 0.83, P = 0.013).

Adjustment for patients with any use of parenteral nutrition did not change the results. BMI was not a significant predictor of mortality either in the whole group or in subgroup analysis.

Up to now, early protein feeding per se has not been evaluated in randomised studies. In the previous analysis of this patient cohort we showed that a protein intake of more than 1.2 g/kg over the whole period of mechanical ventilation was associated with lower mortality [6],[7]. Reaching energy targets alone was not sufficient. These results were confirmed and extended by the observational study of Allingstrup et al., showing a decrease in mortality with increasing protein intake up to 1.5 g/kg [8].

A recent secondary analysis of an observational study including 2,270 septic ventilated medical patients receiving enteral nutrition with a mean of 14.5 kcal/kg and 0.7 g/kg protein per day found that both an increase of 1,000 kcal and of 30 g protein per day, thus a delivery closer to recommended protein intake, was associated with reduced mortality [9]. This study covered the mean energy and protein intake for a maximum of 12 days or until death after ICU discharge, and did not address intake as early as day 4. Of note, mean protein and energy intakes in this septic cohort were lower than in our (septic) cohort, and protein and energy intake were related.

However, results of randomised studies are confusing, because early underfeeding, which implies low protein intake, appears not to increase mortality [1],[4]. Supplemental PN was even associated with an increased infection rate and higher duration of mechanical ventilation and renal replacement therapy in the largest trial [1]. An explanation could be that nutrition inhibits autophagy [23]. Autophagy is considered a housekeeping system to remove dysfunctional and toxic proteins and complete cellular structures [11],[23], and the degradation products subsequently provide nutritional substrate. In a sub-study of the EPanic trial, late PN was associated with reduced muscle weakness and more efficient autophagic control of muscle fibres [24], although the final muscle weakness assessment was not different. Which nutritional component contributes to inhibition of autophagy most is as yet unknown. However, some studies suggest that protein might be more important than glucose [11],[12], although Casaer et al. did not adjust for energy intake [12]. This suggestion seems to contradict our finding that early high-protein intake was associated with lower mortality. However, our study also shows that in the septic cohort early high-protein intake was not associated with lower mortality, which is in line with the patient group of Puthucheary et al. [13] of which half was admitted with sepsis. We hypothesized that septic patients could behave differently because autophagy is also used to degrade intracellular micro-organisms [25]. Thus, autophagy provides a functional role in sepsis by promoting intracellular microbial clearance. Furthermore, in a recent study in critically ill septic patients receiving PN, muscle protein synthesis was normal, but protein breakdown was increased up to 260% compared to healthy controls [26]. Apparently, feeding could not suppress increased protein breakdown in these septic patients. In favour of protein are the randomised trials on supplemental parenteral nutrition that administered a higher amount of protein [2],[27], measured energy expenditure [27], and found a positive impact of supplemental parenteral nutrition on clinical outcomes [2],[27]. Taken together there appears to be a delicate balance in the critically ill patient, and timing and dosing of protein and energy in specific disease groups will have to be addressed in future randomised studies.

Our study also showed that day-4 energy overfeeding was harmful. Overfeeding was defined as an energy intake of more than 110% of measured EE. Forty-one percent of our patients appeared to be overfed on day 4. This means that a standard prescription of estimated energy requirement by Harris-Benedict equation plus 30% is an inaccurate predictor of energy requirements in ICU patients, as has been reported before [5],[28],[29]. Although newer formulas might be more accurate [30], measurement of EE by indirect calorimetry remains the most appropriate tool. However, even when knowing actual EE, it is not known whether energy supply should cover the full equivalent of energy expenditure during the first days of critical illness, because nutrition cannot suppress early endogenous glucose production which may provide more than 50% of energy expenditure [5]. Figure 3 suggests that mild (10 to 20%) underfeeding of energy in the early period of ICU stay might be beneficial. Of note, the increased rate of infections and longer duration of mechanical ventilation and renal replacement therapy found in the EPaNIC trial [1] could partially be explained by a component of overfeeding, because energy needs were not measured but calculated, and rather high energy targets were attained. A smaller study applying calorimetry-tailored nutrition found a trend to lower mortality rate in the supplemental PN group, but also an increased infection rate and longer ventilation and ICU stay [31]. Remarkably, both energy and protein intake were higher in the calorimetry-tailored group. Furthermore, the Swiss trial observed fewer infections in the group receiving supplemental PN from day 4 [27]. Energy intake was tailored by calorimetry and protein intake was higher than in the EPaNIC study. Finally, in the Australian trial, supplemental PN from day 1 provided no effect on mortality or infection rate, but decreased muscle wasting, ventilator duration, and improved quality of life [2]. Thus, in the last three trials, a higher protein and energy intake was beneficial. Whether in favour of targeted feeding or not, these studies were not designed to investigate early protein-targeted feeding and in some of these studies some of the patients were likely overfed [32].

Our study has several limitations, in particular its observational design. A lower early-protein intake may reflect a higher severity of disease. However, protein intake remained a predictor of mortality, independent of APACHE II score, the standard estimate of mortality in critically ill patients. Also, an improved energy intake but with insufficient protein (0.8 to 1.2 g/kg) was not associated with lower mortality. We previously showed that the application of our nutritional algorithm improved adequate protein supply at day 4 from about 30% to almost 60%. However, despite our algorithm, not all patients received adequate protein intake. Furthermore, estimating EE using the old Harris-Benedict equation +30% appears to be associated with significant overfeeding. In addition, the measurement of EE was generally measured after day 4. This means that overfeeding could have been ongoing for a couple of days before it was noticed and corrected. Unfortunately, in clinical practice daily measurement of EE is not feasible. Finally, the number of septic patients was relatively low, but proportional to the admission pattern of the unit.

A strong point of our study is the distinction between patients with and without early energy overfeeding. The negative effect of early overfeeding on patient outcome supports the notion that measuring EE is crucial for optimizing nutrition. Another strong point of this study is that the sample size was large enough to independently assess the effect of protein intake in patients with and without sepsis. Finally, EE was measured while feeding was continued, thereby reflecting real-life total EE.