The nutritional energy to clinical outcome relation revisited

In the previous issue of Critical Care, Dr Bellomo and colleagues reported an observational study of the relationship between nutritional intake and survival in the RENAL randomized controlled trial. In summary, the total energy intake in a very large and severely ill patient population was low. Higher average daily caloric energy intake was not associated with improved survival. The study illustrates the complexity of the interaction between disease and nutrition.

functional outcome were unaffected or even worsened by different early feeding interventions. This finding has inspired investigation of nutritional data available from previous non-nutritional RCTs ( Figure 1). Multicenter surveys of nutrition practices predict an impressive mortality reduction of more than 30% per additional 1,000 kcal mean daily caloric intake (DCI) achieved [8,9]. RENAL nutrition data [1] and Glucontrol nutrition data [10], on the contrary, suggest a neutral relation between DCI and survival or even increased mortality with more nutrition (Figure 1). The VISEP (Efficacy of Volume Substitution and Insulin Therapy in Severe Sepsis) trial also showed that patients receiving only enteral nutrition experienced a better survival despite much less nutritional energy administered [11]. Likewise, the lowest energy intake interval was associated with the fastest recovery in an observational analysis of enteral plus parenteral energy in the EPaNIC (Impact of Early Parenteral Nutrition Completing Enteral Nutrition in Adult Critically Ill Patients) trial, even in the "late" arm where patients did not receive any parenteral nutrition during the first week in the ICU [12].
Dr Bellomo and coauthors avoided several common problems with observational studies of nutrition in the ICU. First, as explained above, the analyses were performed in a large unselected group of severely ill patients, reducing the risk of nutritional selection bias. Time bias is a second problem with analyses of the association between clinical outcome and average energy intake, as the provision of nutrition improves during ICU stay [13]. The authors therefore provided a separate analysis restricted to patients staying longer than 3 days in the ICU to tackle this problem. In this substudy, Kaplan-Meier analysis revealed a higher mortality in patients with higher DCI [1]! A third, often unrecognized error is informative censoring in time-to-event analyses [13]. Informative censoring occurs, for example, when patients leaving the ICU earlier are censored in ICU survival analyses on the ICU discharge day. This statistical flaw artificially inflates the observed mortality in patient groups experiencing a shorter ICU stay [13]. Bellomo and colleagues avoided this problem by using landmark 90-day-survival independent of ICU or hospital discharge status [1]. With this undisputable mortality endpoint, moreover, assessment bias is very unlikely. Nevertheless, this endpoint is also a limitation, because long-term functional outcomeperhaps more likely to be affected by nutritionwas not reported.
Finally, it is impossible to distinguish in an observational study whether less severely ill patients are easier to feed or whether better feeding improves outcome. Indeed, at baseline the patients in the lower DCI group had more organ failure and were older. However, the results of unadjusted and the multiple adjusted analyses all point robustly in the same direction, even if two (mean and median) energy variables together in one multivariable model might be too much. Ultimately, a RCT allocating patients to different energy intake levels remains the only definite solution to distinguish between cause and consequence [3].
If future RCTs confirm the impossibility to improve survival or functional outcome through enhanced feeding early in critical illness [3][4][5][6][7], the mechanisms behind this failure need to be unraveled to improve therapy. In the past, hyperglycemia might have been an explanation for the complications with early feedingparticularly parenteral feedingin critical illness. Today, hyperglycemia is less likely to be an issue because almost all ICUs (87.2%) implement a glycemic control protocol (see additional online material with [9]). Alternatively, suppression of autophagy might nullify potential benefits of early and enhanced nutrition [14,15]. Autophagy is an intracellular mechanism eliminating damaged organelles and toxic protein aggregates. Autophagy is crucial in maintaining tissue integrity, it is activated by starvation and cellular stress signals and is inhibited by feeding.
In conclusion, observational studies of prospectively collected nutritional data are valuable to designing new RCTs aimed at establishing safe and effective feeding strategies in the ICU. Such RCTs should preferably assess impact on 90-day landmark survival, acute cellular metabolism and long-term functional outcome.