**Management of Glucose Control in Noncritically Ill, Hospitalized Patients Receiving Parenteral and**/**or Enteral Nutrition: A Systematic Review**

**Céline Isabelle Laesser 1, Paul Cumming 2,3, Emilie Reber 1, Zeno Stanga 1, Taulant Muka <sup>4</sup> and Lia Bally 1,\***


Received: 30 May 2019; Accepted: 26 June 2019; Published: 28 June 2019

**Abstract:** Hyperglycemia is a common occurrence in hospitalized patients receiving parenteral and/or enteral nutrition. Although there are several approaches to manage hyperglycemia, there is no consensus on the best practice. We systematically searched PubMed, Embase, Cochrane Central, and ClinicalTrials.gov to identify records (published or registered between April 1999 and April 2019) investigating strategies to manage glucose control in adults receiving parenteral and/or enteral nutrition whilst hospitalized in noncritical care units. A total of 15 completed studies comprising 1170 patients were identified, of which 11 were clinical trials and four observational studies. Diabetes management strategies entailed adaptations of nutritional regimens in four studies, while the remainder assessed different insulin regimens and administration routes. Diabetes-specific nutritional regimens that reduced glycemic excursions, as well as algorithm-driven insulin delivery approaches that allowed for flexible glucose-responsive insulin dosing, were both effective in improving glycemic control. However, the assessed studies were, in general, of limited quality, and we see a clear need for future rigorous studies to establish standards of care for patients with hyperglycemia receiving nutrition support.

**Keywords:** glucose control; hyperglycemia; parenteral nutrition; enteral nutrition; nutritional support; insulin

#### **1. Introduction**

Hyperglycemia is frequently encountered during parenteral (PN) and/or enteral (EN) nutrition in hospitalized patients with and without pre-existing diabetes [1,2]. Indeed, it is estimated that more than 50% of patients on PN and 30% of patients on EN experience hyperglycemia whilst in the hospital [3,4]. Hyperglycemia arises in these patients due to one or more of the following factors: (1) diminished insulin sensitivity due to inflammation, stress hormones, and sedentarism [5]; (2) increased carbohydrate provision [6]; and (3) side-effects of medication such as glucocorticoids that interfere with glucose metabolism [7]. In patients totally reliant on PN, these factors are compounded by the loss of the physiological incretin effect on insulin release, as occurs when entirely bypassing the gastrointestinal tract with intravenous nutrient supply [8]. Furthermore, the diminished glucose-stimulated insulin secretion in diabetic patients with some residual beta-cell function increases their requirement for exogenous insulin.

Evidence from several observational studies suggests that emergent hyperglycemia during nutrition support is associated with increased morbidity and mortality [1,9]. There is an apparently linear relationship between the incidence of adverse outcomes and mean glucose levels once glycaemia surpasses a threshold of 6.3 mM [10]. In individuals on PN, the risk of any complication increases by a factor of 1.58 for each 1 mM increase in glycaemia above this threshold [11]. Conversely, treatment of hyperglycemia is shown to improve clinical outcomes [12–15]. However, striving for tight glucose control inherently increases the risk of hypoglycemia, which is similarly associated with adverse clinical outcomes [16,17].

PN and EN nutrition support are provided in a number of ways, ranging from continuous to cyclic regimens, often in combination with unpredictable and variable oral intake or additional intravenous glucose administration. Maintaining glycemic control is even more demanding in patients with unanticipated interruptions of their feeding (e.g., due to emergency surgery), or if nutrition support is suspended due to accidental removal or obstruction of tubing. Guidelines such as those from the American Diabetes Society [18] and the Endocrine Society [19] recommend that random blood glucose levels be maintained below 10.0 mM, provided that this target can be safely achieved. According to these guidelines, the mainstay of hyperglycemia management in the hospital is the administration of insulin, given its high efficacy, flexibility, and lack of interference with most other pharmacotherapies or organ dysfunctions.

Although there are some recommendations for insulin dosing tailored to the needs of patients receiving PN and/or EN, there is a lack of evidence-based support for specific insulin regimens. Insulin can be delivered via intravenous or subcutaneous routes, or in patients receiving PN, insulin may simply be mixed in the nutrition solution [20]. Intravenous insulin infusion at a rate continuously adjusted according to regular capillary blood glucose measurements helps to maintain glucose levels within the recommended limits. However, implementation of intravenous protocols imposes considerable demand on nursing staff, calling for two hours of direct nursing daily per patient [21], which substantially increases the workload of ward staff. Noncritical care nurses have to manage several patients, and staff levels are reduced at nighttime, which does not encourage constant vigilance of blood glucose and manual adjustment of insulin infusion rates. Thus, for practical and safety considerations in noncritical care settings, subcutaneous administration is the favored route for insulin delivery. However, the formulation of PN and EN support can also influence glucose levels in patients on nutrition support. Whereas glucose is the only carbohydrate source in standard PN solutions, the glycemic impact can be modulated by changing the caloric contributions of carbohydrates versus monounsaturated fatty acids (MUFAs), or by using alternate carbohydrates such as fructose. Furthermore, the addition of fibers to EN formulae can delay carbohydrate absorption, thereby attenuating the glycemic impact [22,23]. An overview of the different management strategies is provided in Figure 1.

Diabetes technology has progressed greatly over the past decade, bringing considerable improvements to the care of outpatients with type 1 and type 2 diabetes. Notably, recent development of continuous glucose monitoring (CGM) systems that measure interstitial fluid glucose concentration every few minutes allow to depict glucose profiles in higher resolution, thereby facilitating adjustment of insulin dosages [24]. CGM systems also project trends for glucose levels and feature customizable hypo- and hyperglycemia alerts. There is a growing interest in the use of this technology in hospital settings given the abundance of additional information that can guide therapy adjustments.

**Figure 1.** Management options of hyperglycemia during PN/EN. PN = parenteral nutrition, EN = enteral nutrition.

An important variable in the control of glycaemia is the modulation of insulin to meet with the continuously changing metabolic needs. Due to its inherent flexibility of insulin dose adjustment, the predominant mode of delivery in patients with type 1 diabetes is via a subcutaneous insulin pump, also known as continuous subcutaneous insulin infusion [25]. Here, a portable pump infuses rapid-acting insulin at a rate that can be altered on demand or preset to change at fixed times. This flexible adaptation of the insulin delivery profile makes the approach particularly attractive for patients under cyclic nutrition regimens who may receive large amounts of carbohydrates within predefined time windows. The combination of real-time glucose measurements from a CGM device with a control algorithm that directs insulin delivery via an insulin pump constitutes a closed-loop system, also known as the artificial pancreas [26,27]. Closed-loop systems automatically adjust insulin delivery every 10–12 min according to real-time glucose measurements. The autonomy and glucose feedback-regulation obtained through closed-loop systems hold promise in the particular context of hyperglycemia, arising along with nutritional support, while sparing an excess burden on nursing staff.

Our aim in the present review is to provide an overview of the current status and future outlook of glucose management strategies for noncritical care patients receiving PN and/or EN nutrition support.

#### **2. Materials and Methods**

#### *2.1. Literature Search*

This review was conducted and reported in accordance with the PRISMA and MOOSE guidelines [28,29]. Completed checklists can be found in Supplementary Tables S5 and S6. PubMed and Embase databases were used to identify relevant records over the past 20 years. The Cochrane Central Register of Controlled Trials and ClinicalTrials.gov were searched for published and ongoing studies. The search strategies for the databases are summarized in Supplementary Tables S1–S4. Two reviewers independently evaluated the titles and abstracts according to the selection criteria. For each potentially eligible study, two reviewers assessed the full-text. In cases of disagreement, a decision was made by consensus or, if necessary, a third reviewer was consulted.

#### *2.2. Study Selection Criteria*

Studies were included if they met the following criteria: (i) were published or registered in English language between 10 April 1999 and 10 April 2019 (date last searched); (ii) were clinical studies evaluating new treatment approaches against a comparator; (iii) included adult (age ≥ 18 years) noncritical care inpatients receiving PN and/or EN nutrition; (iv) investigated strategies to manage blood glucose control; and (v) had outcomes reflecting glucose control. Exclusion criteria were pregnancy, breast feeding, case reports, abstracts, guidelines, or literature reviews. The selection process is shown in Figure 2.

**Figure 2.** Flowchart illustrating the study selection process.

#### *2.3. Data Extraction*

Two reviewers extracted data independently using a predesigned form, including study design, sample size, glucose management strategy, primary outcome, and main results. If no primary outcome was specified, we obtained the endpoints deemed most relevant.

#### *2.4. Quality Assessment*

The quality of clinical trials was evaluated by two reviewers based on the "Cochrane Risk of Bias tool" [30]. According to this tool, studies are judged to be of low or high risk for bias based on criteria to evaluate random sequence generation, allocation concealment, blinding of participants/personnel and outcome assessment, incomplete outcome data, and selective reporting. Since retrospective publications have an inherently high risk of bias in most domains, and there is not a standardized assessment tool available, we refrained them from quality evaluation.

#### **3. Results**

#### *3.1. Study Identification and Selection*

We identified in total 745 potentially eligible records. Following screening based on titles and abstracts, 34 citations were selected for detailed full-text evaluation. Of those, 16 articles met the selection criteria and were included in the review (Figure 2). One ongoing study was eligible for inclusion.

#### *3.2. Characteristics of the Included Studies*

The included published publications comprised 1170 participants from 15 clinical studies, of which 11 were clinical trials (9 randomized and 2 nonrandomized) and 4 were retrospective observational studies. One study was an ongoing registered study. Seven studies were conducted in Europe, three in North America, and five in Asia. Only three studies included more than one center. The study population represented medical (20%), surgical (27%), or mixed medical and surgical (53%) noncritical care inpatients. Three studies explicitly stated the inclusion of patients with type 1 diabetes, three studies explicitly excluded patients with type 1 diabetes, five studies exclusively recruited patients with type 2 diabetes, whereas the remainder did not further characterize the diabetic state of study participants or the reason for hyperglycemia (*n* = 4). Nutrition support comprised PN in six studies, EN in seven studies, and PN and/or EN in two studies. Regarding glucose management strategies, 4 studies investigated nutritional approaches, and 11 studies explored insulin interventions. No studies were found that explored noninsulin pharmacological strategies. The study duration ranged from 36 h to 46 days. Study characteristics and findings are summarized in Tables 1 and 2.

Four of the clinical trials were considered of medium quality, indicating a low risk of bias in all domains except for performance and detection bias, which was deemed reasonable with a near impossibility of blinding [31–34]. The other studies were deemed to be of limited quality, with high risk of bias in the domains of lack of primary outcome definition and selective or incomplete reporting (Supplementary Figure S1). Since retrospective publications have an inherently high risk of bias in most domains, we refrained from a quality evaluation of the four such studies.

#### *3.3. Studies Examining Nutritional Strategies*

Four studies examined nutritional strategies to manage glycaemia in patients receiving nutrition support. The use of EN formulae with lower glycemic impacts were contrasted against conventional products in two clinical trials [35,36]. In a nonrandomized crossover study involving inpatients with type 2 diabetes, Tiyapanjanit et al. tested a formula compounded in-house to contain 50% of calories as carbohydrates, thereof 67% as fructose, in comparison with an iso-energetic control formula (53% of calories as carbohydrates and 15% as fructose). Over the study period of 36 h, lower mean glucose was achieved with the high-fructose formula compared to control (6.8 ± 1.5 vs. 8.0 ± 2.1 mM, *p* = 0.022). No insulin or antidiabetic medication was administered in these patients [35]. Similarly, a four-center randomized controlled parallel trial conducted in 104 inpatients with type 2 diabetes found a lower nutrition-induced relative change from baseline glycaemia with the use of lower glycemic impact enteral vs. standard enteral nutrition formula (10% vs. 21%, *p* = 0.006). The investigated EN formula contained a reduced amount of carbohydrates (9.4 vs. 12.5 g/100 mL, *p* = n/a) and higher amounts of MUFAs (3.8 vs. 1.0 g/100 mL, *p* = n/a) [36]. Similarly, an ongoing randomized clinical trial (GlyENStroke, NCT03422900) evaluates the efficacy of a diabetes-specific enteral nutrition formula to reduce hyperglycemia (glucose levels > 8.3 mM) in nondiabetic patients with hyperglycemia on nutrition support after stroke [37].


use)medical

= 13)

> (stroke unit)

**Table 1.** Overview of clinical trials.

#### *J. Clin. Med.* **2019**, *8*, 935


Hagedorn insulin, wo = without, i.v. =

*p*-value available. 3 Overall quality assessment; specific domains can be found in

intravenous, CHO =

monounsaturated

 fatty acids. 1

Open-label if not stated otherwise. 2 No

carbohydrates,

 h = hours, d = day(s) and MUFA =

Supplementary

 Figure S1.

**Table 1.** *Cont.*


**Table 2.** Overview of observational, retrospective studies.

BG= blood glucose, vs. = versus, T1D = type 1 diabetes, T2D = type 2 diabetes, PN = parenteral nutrition, s.c. = subcutaneous, POC = point of care, EN = enteral nutrition, NPH = NeutralProtamineHagedorn,ns=notsignificant,VRII=variablerateintravenousinsulin, PO=primaryoutcome,andNOS=nototherwisespecified.

insulin 3×/day (mixed3, *n*

= 6)

There was only one study examining a modified nutritional formula in abdominal surgery patients that were entirely parenterally fed. In this randomized parallel double-blind single-center study, Valero et al. contrasted a PN formula with glucose as a sole carbohydrate source with an iso-energetic PN formulation containing a 2:1:1 glucose:fructose:xylitol carbohydrate mixture. The study population consisted of patients with type 1 diabetes (21%) or type 2 diabetes (79%). Both formulae provided a similar total amount of carbohydrates (2.9 ± 0.5 vs. 2.9 ± 0.7 g/kg/day), but the test formula under investigation contained less glucose (2.9 ± 0.5 vs. 1.5 ± 0.4 g/kg/day). Protein and fat quantity were similar. The time until attaining target glycaemia (8.3–11.1 mM) did not significantly differ (2.5 ± 1.7 vs. 2.4 ± 2.1 day, *p* = ns), nor did the percentage of patients attaining values below 11.1 mM (75% vs. 85%, *p* = ns) and total daily insulin dose (45 ± 19 vs. 45 ± 26 U/day, *p* = ns). However, when stratified according to the occurrence of sepsis, the nonseptic patients showed lower insulin requirements with the glucose–fructose–xylitol regime (37 ± 17 vs. 44 ± 17 U/day, *p* = 0.026) [41].

Managing glycaemia during PN is more demanding than during EN. In this context, the combination of PN with EN, in addition to its well-established trophic benefits on gastrointestinal function, may also confer glycemic advantages. This was shown by a randomized controlled parallel study performed by Lidder et al. in patients undergoing esophagectomy who received nutrition support for up to five days post-surgery. The study contrasted the coverage of 30% of energy needs by EN and 70% by PN feeds with 100% coverage by PN. Although no effect was seen in mean glucose over the five-day study period, the combined use of EN and PN lead to lower glycaemia three days after surgery (*p* = 0.002) [34].

#### *3.4. Studies Examining Insulin Strategies*

Eleven studies examined insulin-based strategies to manage glucose in patients receiving nutrition support (five with PN, five with EN, and one with combined PN/EN). A randomized controlled single-center parallel study involving 212 type 2 diabetic patients who had undergone gastrectomy for gastric cancer compared protocol-driven intravenous insulin therapy with conventional subcutaneous sliding scale insulin delivery over 8–10 days of continuous EN. Mean glucose levels were lower (5.4 ± 1.2 vs. 9.5 ± 1.8 mM, *p* < 0.001), and mean daily insulin doses were higher (55 ± 15 vs. 32 ± 16 U/day, *p* < 0.001) in patients receiving intravenous insulin compared with subcutaneous insulin. However, in the group receiving intravenous insulin, eight participants experienced episodes of severe hypoglycemia (defined as blood glucose < 2.2 mM), versus only one participant in the subcutaneous group (*p* = 0.010). Additionally, this study reported outcomes extending beyond glycemic control, finding a reduced incidence of surgical site infection in the intravenous insulin group [12].

A second randomized controlled trial (RCT) evaluating different insulin treatments during EN included 50 patients with non-type 1 diabetes who were randomly assigned to receive sliding scale subcutaneous regular insulin either with or without once daily subcutaneous glargine. In the group without glargine administration, subcutaneous neutral protamine Hagedorn (NPH, isophane) was given as a rescue medication when glucose levels exceeded 10.0 mM. Mean glucose as the primary outcome was comparable between groups, as were the number of hypoglycemic events and total daily insulin dose. However, NPH was required in 48% of the control participants, as their glucose levels were not sufficiently controlled with the regular sliding scale subcutaneous insulin alone [40].

A single-arm trial evaluated the efficacy of a computerized variable rate insulin infusion rate protocol in previously insulin-naïve stroke patients on continuous EN over five days. Compared to a historical control population who received enteral bolus feeds accompanied by intravenous insulin coverage, the intervention resulted in a higher percentage of values in the target range of 4.4–6.1 mM (55% vs. 19%, *p* < 0.005) [39].

Additionally, two retrospective observational studies evaluated insulin-based strategies to treat hyperglycemia in hospitalized patients receiving EN. Hijaze et al. found comparable mean glucose values, and similar times with glucose values within the target range (7.8–10 mM), in patients receiving subcutaneous NPH insulin thrice daily vs. basal–bolus insulin therapy with insulin analogues [43]. In the second such study, Hsia et al. retrospectively evaluated glucose control using three different insulin regimens in patients with diabetes of type not otherwise specified during at least three days of continuous EN. Treatment consisted of (1) 70/30 biphasic insulin (NPH/regular) every 8 h (*n* = 6), (2) 70/30 biphasic insulin every 12 h (*n* = 8), and (3) a basal–bolus regimen (glargine and lispro, *n* = 8). The 8 h 70/30 biphasic insulin group had the highest proportion of glucose values falling within the target range (7.8–10.0 mM) (69% vs. 22% vs. 24%, *p* < 0.01). Hypoglycemic events (<3.9 mM) occurred five times in glargine/lispro group, twice in the twice-daily biphasic group, and once in the thrice-daily biphasic group [45]. In patients receiving PN, two randomized controlled parallel trials evaluated strategies that involved the addition of regular insulin to the PN feeding bag. Olveira et al. recently compared the use of 100% coverage of insulin needs by regular insulin added to the PN bag with 50% coverage by regular insulin added to the PN bag and 50% administered subcutaneously as insulin glargine in a total of 161 mixed surgical and medical patients with type 2 diabetes recruited at 26 different sites. The glucose values of participants were evaluated while receiving total PN (maximum of 15 days) and two days after cessation of total PN. Mean glucose levels during total PN did not significantly differ between groups (9.2 ± 2.0 vs. 9.6 ± 2.4 mM, *p* = ns); however, mean glucose two days after cessation of total PN was higher in the group that had received 100% of their exogenous insulin requirements added to the PN feeding bag (8.9 ± 2.5 vs. 7.9 ± 2.4 mM, *p* = 0.024). The authors did not provide any details on the requirement for insulin therapy after cessation of nutrition support [32]. Another randomized controlled clinical trial of PN patients contrasted two different basal insulin regimens. Hakeam et al. compared subcutaneous basal–bolus therapy (insulin glargine + short-acting insulin analogue) with the addition of regular insulin to the PN bag in a total of 67 non-type 1 diabetic patients. Both groups received additional corrective regular subcutaneous insulin according to a sliding scale. Basal insulin delivery dose was titrated based on glucose values by a daily 40%–60% dose increase if glucose values were still above target. The percentage of glucose values within the glycemic target range (7.8–10.0 mM) tended to be higher in the group who received s.c. insulin glargine compared to those who received regular insulin added to the PN bag (52% vs. 48%, *p* = 0.06). Mean glucose levels and number of hypoglycemic events did not differ between treatment groups [33].

In contrast to these prospective findings in PN patients, a retrospective evaluation performed by Truong et al. in 102 patients on PN showed superior glucose control defined as percentage of patients with ≥5 of 6 glucose values <10 mM over 2 days in those who had received 100% of required insulin added to the PN bag (*n* = 78), compared with those treated with subcutaneous insulin glargine (*n* = 35) (72% vs. 49%, *p* = 0.017). Additionally, fewer patients receiving insulin via the PN bag experienced two or more hypoglycemic events compared to those with subcutaneous administration (9% vs. 17.1%, *p* = ns) [42].

As reviewed above in patients receiving EN, a protocol-driven intravenous insulin delivery approach has proven to confer glycemic benefits in PN patients, according to a retrospective analysis performed by Neff et al. A total of 53 surgical and medical patients requiring insulin therapy whilst in the hospital were treated either with protocol-driven variable rate intravenous insulin (*n* = 32), or received basal–bolus subcutaneous insulin therapy (*n* = 21). The insulin infusion group compared to the group receiving subcutaneous basal–bolus insulin therapy showed lower mean glucose levels (9.6 ± 2.1 vs. 11.2 ± 2.6 mM, *p* = 0.009) and a higher percentage of glucose values within the glycemic target of 4.0–10.0 mM (62% vs. 43%, *p* = 0.008), without increased risk of hypoglycemia [44].

Two recently performed randomized controlled parallel design trials evaluated insulin pump therapy, also known as continuous subcutaneous insulin infusion (CSII), in patients receiving nutrition support. The study performed by Li et al. compared the use of CSII (*n* = 50) to basal–bolus therapy using insulin glargine in combination with insulin aspart (*n* = 52) in patients receiving PN. Study treatment was initialized before surgery, and PN began on day 1 after surgery, with comparison of glucose control with CGM from postoperative day 1 to day 5. Glycemic variability was assessed by mean amplitude of glucose excursion (MAGE) as the primary outcome. CSII reduced glycemic variability compared to basal–bolus injection therapy (3.7 ± 2.8 vs. 6.2 ± 3.0 mM, *p* < 0.05). No hypoglycemic events occurred in either treatment group [38].

The second study evaluated CSII as part of a fully automated subcutaneous closed-loop glucose control (*n* = 21) against conventional subcutaneous insulin therapy according to local practice (*n* = 22) in two different hospitals. Randomization was stratified according to BMI, prestudy total daily insulin dose, and type of nutrition support to ensure to demographic balance between groups. The closed-loop system consisted of a subcutaneous insulin pump, a CGM device, and a control algorithm, which adjusted insulin delivery every 12 min based on real-time CGM values. An example of such fully automated closed-loop insulin delivery is illustrated in Figure 3. Participants were recruited from medical and surgical wards and received PN (*n* = 13), EN (*n* = 27), or combined PN/EN (*n* = 3). The primary outcome was the proportion of time when sensor glucose was within the target range (5.6–10.0 mM). Participants were followed for up to 15 days or until hospital discharge. The closed-loop system nearly doubled the proportion of time spent in the glycemic target range compared to control (68% ± 16% vs. 36% ± 27%, *p* < 0.0001). Time spent above target, mean glucose level, and glucose variability were all significantly lower in the closed-loop group. Hypoglycemia was infrequent in both arms, and its incidence did not differ significantly. The substantially better glycemic control in the closed-loop compared to the control group was achieved with a similar total daily insulin dose (53.9 vs. 40.3 U, *p* = ns) [31].

**Figure 3.** Profile of fully automated subcutaneous closed-loop insulin delivery over 24 h in a noncritical care patient [31]. A control algorithm modulates subcutaneous insulin delivery via an insulin pump (denoted in blue) according to interstitial sensor glucose values (denoted in red). (Kindly provided by Professor Roman Hovorka, University of Cambridge, UK).

#### **4. Discussion**

Hyperglycemia is a common occurrence in hospitalized patients receiving PN and/or EN, and its management in noncritical care settings is challenging. The present review summarized the available evidence for strategies to improve glucose control in this vulnerable population. Improved glucose control can be achieved either by lowering the glycemic impact of nutrition supply and/or matching the nutrition-induced glycemic excursions with a tailored pharmacokinetic profile of a given insulin preparation. Both approaches have been evaluated in a limited number of randomized and nonrandomized clinical trials and observational studies over the past 20 years.

Altering the macronutrient distribution (an increase of calories as MUFAs at the expense of carbohydrates) and the use of nonglucose carbohydrate sources along with high fiber content to delay absorption in EN formulae have proven effective in various studies as well as in meta-analyses [22,23]. The use of diabetes-specific enteral formulae is therefore supported by the expert group of the European Society of Clinical Nutrition and Metabolism (ESPEN) for patients with a history of diabetes [46]. With respect to PN formulae, glucose substitutes such as fructose or xylitol to lower the glycemic impact are no longer used in clinical practice. This may relate to previously reported metabolic side effects of parenteral xylitol and fructose such as formation of oxalate crystals in the kidney and lactate accumulation [47–49]. To make matters worse, life-threatening metabolic complications can occur in patients with undeclared hereditary fructose intolerance [50]. There is currently no data on the efficacy

and safety of PN formulations with reduced carbohydrate content (and consequently higher protein or lipid fractions) in noncritically ill patients with hyperglycemia. In critically ill patients, however, lipid-based, compared to iso-energetic glucose-based, PN formulae showed more favorable metabolic effects [51]. Additionally, lowering the overall carbohydrate and energy provision has proven effective in reducing hyperglycemia in critically ill patients [52,53].

As an alternative to conventional routes of insulin administration, algorithm-driven intravenous insulin titration protocols during both PN and EN achieved superior glucose control compared to subcutaneous approaches with either sliding scale or basal–bolus insulin therapy. However, this approach poses logistical challenges for ward staff, given the need for frequent glucose draws and insulin dose adjustments, which are either impractical or simply unfeasible in noncritical care settings with low nurse-to-patient ratios. Regarding subcutaneous insulin regimens, administering basal plus supplemental short-acting insulin analogues showed superior efficacy compared to a sliding scale approach with short-acting insulin. The administration of intermediate or long-acting insulins can thus be recommended for PN/EN patients receiving subcutaneous insulin therapy. No data exist with regard to the use of recently introduced ultra-long acting insulins (e.g., degludec). Of note, the required time to reach steady state insulin levels with ultra-long acting formulations imposes certain constraints on the titration method, and prediposes to dysglycemia. There is a practical consideration that administering long-acting insulin in previously insulin-naïve patients brings a risk of hypoglycemia if feeding tubes are accidentally pulled or obstructed.

The admixture of insulin into the PN feeding bag is a safe and effective alternative to using shortor rapid-acting subcutaneous insulin. Moreover, further advantages lie in the lesser need for nursing time, the concomitant discontinuation of insulin delivery upon PN interruption, and the consequently lower risk of hypoglycemic events. The latter is particularly relevant for patients in whom the transient need for exogenous insulin is primarily a result of their nutrition support. However, the need for strict aseptic conditions may render the procedure impractical or not permissible. Furthermore, reservations exist regarding the diminished or highly variable efficacy of PN insulin due to interference from PN ingredients or bag surface material [54,55].

Irrespective to the chosen approach for insulin administration, a limiting factor in obtaining tight glucose control is the risk of inadvertent hypoglycemia [17]. As is the case for hyperglycemia, iatrogenic hypoglycemia is associated with increased cost and adverse medical outcomes [16]. In this context, there is an increasing interest in the use of noninsulin glucose-lowering agents (without hypoglycemia risk) for the treatment of inpatient hyperglycemia. The recent SITA-HOSPITAL randomized controlled study investigated 279 noncritical care patients with type 2 diabetes and showed that oral sitagliptin plus basal insulin led to similar glycemic control than the more labor-intensive basal–bolus insulin regimen [56]. However, challenging patients such as those with high insulin requirements, renal failure, or use of glucocorticoids were excluded from the study. Currently, there are no data on the inpatient use of noninsulin glucose-lowering treatments for hyperglycemia in noncritical care patients on PN and/or EN. Reservations apply to the common side effects of noninsulin treatments (e.g., incretin-based therapies) on the gastrointestinal tract, which is often the primary pathophysiology calling for nutrition support. Further research is needed to explore the potential risks and benefits of noninsulin pharmacotherapy for managing glucose levels in noncritical care settings with EN/PN.

The high prevalence of hyperglycemia amongst noncritical care patients on PN and/or EN support, in conjunction with the increasing workload burden placed on hospital staff, brings an urgent need for innovative approaches to improve the efficacy, efficiency, and safety of healthcare delivery in this context. The advent of novel technologies such as automated closed-loop systems that titrate insulin delivery based on real-time sensor glucose measurements could potentially address this need, whilst reducing staff workload burden. Uncertainties remain with respect to interference with certain medications and inaccurate glucose readouts related to compromised microcirculation. In addition, we concede that there are short-term costs to purchase, install, and train staff in the use of any novel

technology. Ongoing studies will further document the potential role of this technology and the obstacles to its integration into clinical practice without disrupting the usual workflow.

The present systematic review turned up rather few randomized controlled trials and a limited number of retrospective observational studies addressing hyperglycemia in noncritical care patients on PN and/or EN. The studies were highly heterogeneous in terms of study population, nutrition regimen, study endpoints, and glucose measurement techniques. Patients with pre-existing diabetes already on insulin treatment before their admission to the hospital and those receiving steroids or showing impaired renal function clearly have different optimal insulin delivery profiles compared to patients without pre-existing diabetes and/or with stress-induced hyperglycemia. The particular nutrition regime (i.e., the EN/PN feeding schedule and any additional oral intake) leads to variable carbohydrate exposures with relevant impact on insulin requirements. The few studies that utilized CGM more comprehensively assessed both hyper- and hypoglycemic excursions, whilst studies adopting intermittent (i.e., six-hourly) point-of-care measurements may have missed important events such as postprandial transients. Most studies were conducted in patients on continuous PN and/or EN, which may be less demanding for insulin management compared to the less common bolus or cyclic feeding practices. The scarce evidence and the many factors (e.g., patient comorbidity, staffing level, hospital guidelines, and policies) that determine the ability and capacity to treat hyperglycemia effectively and safely challenge the provision of generalizable treatment recommendations. In Figure 4, we propose a workflow recommendation considering both nutritional and insulin adaptation to manage hyperglycemia in the noncritical care population receiving nutrition support.

**Figure 4.** Approach to the management of hyperglycemia in patients receiving enteral or parenteral nutrition. CHO = carbohydrates, EN = enteral nutrition, BG = blood glucose, IV = intravenous, PN = parenteral nutrition, SC = subcutaneous, and q8h = dosing every 8 h.

The majority of the included studies scored poorly in methodology, with a high or unclear risk of biases according to Cochrane criteria. Undoubtably, it is sometimes difficult or unethical to undertake double-blind RCTs with standardized protocols. Also, some studies did not correct for multiple testing, which may have overestimated outcomes. The included retrospective observational studies were not evaluated for quality given the known risk of outcome overestimation and confounding biases inherent in that design. However, we chose to report these studies in this review to cover the widest possible range of different management possibilities.

Small sample sizes and short study durations of some studies may have led to an underestimation of effect sizes, thus hindering the sensitivity to ascertain the efficacy of potential methods to improve glycemic control or indeed to confirm the impact of superior glucose control on patient outcomes. There is clearly a need for further research in the form of well-designed and adequately powered multicenter trials of sufficient duration aiming to examine effects of glucose management strategies on glucose control, clinical outcome, and also optimization of nutritional status in patients receiving PN and/or EN.

#### **5. Conclusions**

The management of hyperglycemia in patients receiving PN and/or EN presents unique clinical challenges for both diabetic and nondiabetic hospitalized patients with hyperglycemia. Coherent approaches to this problem are important to avoid potential complications. Obtaining a better match between the carbohydrate dose and the insulin supply is likely to improve glucose control. Granting more attention to the glycemic impact of nutrition regimes in conjunction with deploying novel technologies such as CGM and glucose-responsive automation of insulin delivery through closed-loop systems may address these needs without increasing staff workload. Supplemental or alternate use of noninsulin pharmacological approaches may further open up new lines of research. Well-designed and adequately powered randomized controlled trials are necessary to define the optimal management of hyperglycemia and consequent clinical benefits in patients receiving nutrition support.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/8/7/935/s1; Figure S1: Quality assessment of clinical trials according to modified Cochrane Risk of Bias tool [30]. Table S1: PubMed Search Strategy 12.04.2019, Table S2: Embase Search Strategy 10.04.2019, Table S3: Cochrane Central Register of Controlled Trials Search Strategy 10.04.2019, Table S4: ClinicalTrials.gov Search Strategy 10.04.2019, Table S5: PRISMA 2009 checklist [28], Table S6: MOOSE checklist [29].

**Author Contributions:** Conceptualization and methodology, L.B.; design and conduction of the search: C.I.L. and L.B.; writing—original draft preparation; C.I.L. and L.B.; writing—reviewing and editing: C.I.L., P.C., E.R., Z.S., T.M. and L.B; supervision: L.B.

**Acknowledgments:** We would like to thank Heidrun Janka for her help in improving the search strategies and Kai Holtmann for his help in improving the design of the figures.

**Conflicts of Interest:** The authors declare no conflict of interest related to this work.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Micronutrient Deficiencies in Medical and Surgical Inpatients**

#### **Mette M Berger \*, Olivier Pantet, Antoine Schneider and Nawfel Ben-Hamouda**

Service of Adult Intensive Care Medicine and Burns, Lausanne University Hospital (CHUV), BH 08.612, Rue du Bugnon 46, 1011 Lausanne, Switzerland

**\*** Correspondence: Mette.Berger@chuv.ch; Tel.: +41-21-3142095

Received: 30 May 2019; Accepted: 21 June 2019; Published: 28 June 2019

**Abstract:** Inpatients are threatened by global malnutrition, but also by specific micronutrient (i.e., trace element and vitamins) deficiencies that frequently are overseen in the differential diagnosis of major organ dysfunctions. Some of them are related to specific geographic risks (iodine, iron, selenium, zinc, vitamin A), while others are pathology related, and finally many are associated with specific feeding patterns, including low dose enteral feeding. Among the pathologies in which laboratory blood investigations should include a micronutrient outwork, anemia is in the front line, followed by obesity with bariatric surgery, chronic liver disease, kidney disease, inflammatory bowel disease, cardiomyopathies and heart failure. The micronutrients at the highest risk are iron, zinc, thiamine, vitamin B12 and vitamin C. Admission to hospital has been linked with an additional risk of malnutrition—feeding below 1500 kcal/day was frequent and has been associated with a structural additional risk of insufficient micronutrient intake to cover basal needs. Although not evidence based, systematic administration of liberal thiamine doses upon admission, and daily complementation of inpatients' food and enteral feeding solutions with multi-micronutrient tablets might be considered.

**Keywords:** iron; copper; selenium; zinc; thiamine; vitamin B12; obesity; inflammation; enteral nutrition

#### **1. Introduction**

Malnutrition includes a wide spectrum of conditions [1] that may affect energy, substrates and micronutrients (i.e., trace elements and vitamins) to variable degrees. Deficiency is defined as a lack, or shortage of a specific micronutrient that is essential for the proper growth and metabolism of a human: 11 trace elements and 13 vitamins qualify as essential in humans [2,3].

Disease related malnutrition has shown to be frequent upon admission [4]. The proportion of patients that have been admitted to hospital in a poor nutritional status varies between nearly zero in trauma and up to 50% in oncologic pathologies [5,6], and has an important impact on costs [6]. Malnutrition affects all ages from pediatric [7] to geriatric admissions [8]. Elderly subjects constituted an important, increasing part of our population with specific physiopathologic characteristics [9]. In developing countries, socio-economic factors will generate specific deficiencies [10], which will also be present in migrant populations [11]. Further, some pathologies threaten more, particularly the micronutrient status. Finally, the geographic place of living in the world, including some western countries and the local soil characteristics, will have their own impact [12,13]. Iodine and iron deficiencies are the most well-known. The below text will review some of these undermining deficiencies.

The definition of deficiency requires knowing the specific needs of diseased patients—this information is generally not available. Micronutrient requirements have been determined for the healthy population with age specificities [3]. They are called dietary reference intakes (DRI), a concept that has replaced the recommended daily allowance (RDA). There is minimal data that exists about the needs during disease. Therefore, the present review, which is based on searches in PubMed and the Cochrane library, is focused on conditions in adults that have been best documented, such as elderly patients, and some selected conditions. The issue of refeeding syndrome in hospitalized patients is be given particular attention.

#### **2. Undermining Micronutrients Deficiencies**

The causes of deficiency can globally be divided into three: Low availability, reduction of intakes, and malabsorption. Some geographic soil and nutrition specificities have threatened the entire populations' micronutrient status. This knowledge is not new [2], and the below text provides a brief summary of their impact on acute diseases.

Micronutrient deficiencies have been called the hidden hunger [14], as they are determining and aggravating factors for one's health status and quality of life. The close relation between fetal malnutrition and the development of chronic non-communicable diseases later in life has been repeatedly confirmed [10,15]. Vitamin A (in developing countries), iodine, and iron deficiencies (worldwide) are the most important in terms of global public health [16]. The below deficiencies have affected patients before their hospital admission and should be integrated in the differential diagnosis of the acute condition.

#### *2.1. Iodine*

Deficiencies early in life impairs cognition and growth. Iodine status is also a key determinant of thyroid disorders in adults. Iodine deficiency disorders have affected 740 million people. A severe deficiency causes goiter and hypothyroidism because, despite an increase in thyroid activity to maximize iodine uptake and recycling, iodine concentrations are insufficient to enable synthesis of thyroid hormones [17].

#### *2.2. Iron*

Deficiencies are a public health problem worldwide [18]. In the year 2000, it was estimated that iron deficiency anemia affected two billion people, mostly women and children. Anemia concerns roughly a third of the world's population [19], but anemia represents the end stage of an iron deficiency [20] which is largely prevalent in all categories of society [21]. The diagnosis of this deficiency is complex in acute and chronic diseases presenting an inflammatory response that modulates blood iron concentrations [20,22]. Its efficient treatment has often been prevented by beliefs such as the risk of iron administration causing additional oxidative stress or favoring infection. These concerns have not been confirmed [23]. The exploration of the iron status (deficiency < 7.1 μg/L or 12.5 mmol/L) included determining the presence of inflammation reflected by C-reactive protein (CRP) > 10 mg/L [24]. In addition to hemoglobin, and erythrocyte morphology (microcytosis, hypochromia), the following have been required: Serum ferritin (<30 ng/L), transferrin saturation (<15%), total iron-binding capacity (>13.1 μmol/L), soluble transferrin receptor (increased > 28.1 nmol/L) [20], and hepcidin (ranges men 0.6–23.3 nmol/L, women 0.4–19.7 nmol/L) [25].

With the availability of hepcidin as a marker of deficiency, the diagnosis should become easier[20,22,25]. Hepcidin is a hormone synthesized in the liver, secreted into the blood that systemically controls the rate of iron absorption as well as its mobilization from stores [26]. The synthesis of hepcidin is up-regulated by inflammatory cytokines (particularly interleukin-6), irrespective of the total level of iron in the body. This relationship most likely accounts for the development of anemia of chronic disease [27,28] (please see Section 3.7 for specificities in kidney diseases).

#### *2.3. Selenium*

Deficiencies have been an issue in Europe and in other parts of the world as the main source of human exposure is diet, which is related to the soil content [29]. Scientists have claimed that changes in climate and the organic carbon content of soil will lead to overall decreased soil Se concentrations, particularly in agricultural areas [30]. Europe and some parts of Australasia are particularly affected, while North America is spared by elevated soil content. Borderline Se status reduced endogenous antioxidant defenses via the reduction of the activity of the glutathione peroxidase (GPX) family of enzymes [31], which contributed to poor outcomes facing critical illness, and particularly sepsis [32]. In major trauma, the intravenous administration of selenium in combination with other micronutrients at 5 times DRI doses (see Table 1) resulted in normalization of plasma GPX3 activity and shortened hospital stays [33]. Nevertheless, the high dose single selenium trials (1000 μg/iv/day) in septic shock have been negative [34]. Therefore this strategy should not be considered, and has been advised against in the Surviving Sepsis Campaign 2017 recommendations [35].


**Table 1.** Micronutrient strategy in critically ill patients admitted to the Lausanne multidisciplinary ICU, according to disease and nutrition therapy.

\*: High risk conditions include shock (cardiogenic, septic, hypovolemic), pancreatitis, severe hepatopathy, major trauma, organ transplant, and malnutrition.

#### *2.4. Zinc*

Deficiencies were first identified in the 1960s [36], and have continued to plague multiple regions of the planet [37]. It has affected all ages, but particularly the elderly in western regions. Zinc deficiency affects cell-mediated immune dysfunction, susceptibility to infections, and increases oxidative stress. A randomized trial including 50 healthy elderly subjects tested the impact of a zinc-supplement (45 mg elemental Zn/day) orally for 12 months versus placebo [38]. First, the authors showed that compared to younger, healthy subjects, the elderly study subjects had lower plasma zinc, higher ex vivo generation of inflammatory cytokines and interleukin 10. The supplement resulted in a significant reduction of the incidence of infections and ex vivo generation of tumor necrosis factor alpha and plasma oxidative stress markers compared to the placebo group [38].

#### *2.5. Copper*

Deficiencies have also been shown to affect specific geographic areas—drinking water may or may not be rich in copper depending on the pipe composition. The deficiency is most frequently acquired [39], e.g., due to insufficient intakes in vulnerable populations, increased demands (pregnancy, lactation, wound healing), malabsorption (including high phytic content of vegetarian diets), increased losses (e.g., major burns, continuous renal replacement therapy), and from hereditary diseases [40,41].

#### *2.6. Vitamin A*

Deficiency is considered the world's most important cause of preventable blindness [42]. It affected 2.8 million children under five years of age. Tragically, the numbers have only grown over the last decades [20].

#### *2.7. Vitamin D*

Long known only for its role in calcium and bone homeostasis, and the development of osteoporosis, its deficiency seems to be a worldwide problem, and to particularly affect inpatients. The multiple effects of vitamin D are mediated by genomic and non-genomic effects, and include muscle function and metabolism, innate and adaptive immune system, lung epithelial function, cardiac function and numerous other functions as the specific nuclear vitamin D receptor are widely expressed throughout the body [43]. As shown in Table 1, low values can be found in nearly any severe condition. Many observational studies have consistently shown an association between low vitamin D levels and poor clinical outcomes. Nevertheless, high-quality evidence showing the benefits of vitamin D supplementation in inpatients is still lacking [44]. One of the largest studies (VITdAL-ICU) was analyzed retrospectively: 475 critically ill patients with 25(OH)D levels < 20 ng/mL [45]. The deficiency was not associated with persistent critical illness, nor did supplementation with vitamin D3 mitigate the development of persistent critical illness. The actual evidence does not support general vitamin D screening and supplementation for the medical inpatient population in an acute care setting [44]. By contrast, screening of chronic kidney disease patients is probably rational (see Section 3.7).

#### **3. Disease Specific Deficiency**

The below Table 2 summarizes the most frequent deficits encountered in the different pathologies.


**Table 2.** Disease specific vitamins and in trace elements deficiencies.

?: means uncertainty as to deficiency.

#### *3.1. Alcoholism*

Micronutrient deficiencies are commonly encountered in alcoholic patients, not only explained by a decrease of global dietary intake, but also because of maldigestion, malabsorption, impaired hepatic activation and an increased breakdown and excretion. The risk of developing micro and macronutrient deficiencies has been known to increase significantly when alcohol makes up more than 30 percent of total caloric intake [46].

All fat-soluble vitamins (A, D, E and K) are susceptible to be decreased [47], although vitamins A and K deficiencies are more common in the case of overt hepatic disease or chronic pancreatitis. Among water soluble vitamins, the vitamin B1 (thiamine) deficiency has been the most frequently described and feared [48], potentially leading to Wernicke's encephalopathy and its well-known triad (delirium, oculomotor abnormalities and ataxia). If left untreated, it can progress to the amnestic-confabulatory syndrome called Korsakoff. The therapeutic benefit of thiamin has been demonstrated in alcoholic patients even without severe Wernicke-Korsakoff encephalopathy [49]. In the emergency department, patients with acute alcohol intoxication have not all suffered thiamine deficiency [50,51]. A systematic review showed that thiamine and vitamin C were the most frequently identified deficiencies [51,52]. Nevertheless, considering the elevated prevalence of malnutrition in these patients, it is cautious and cheap, in case of hospital admission to provide 100–300 mg thiamine prior to any glucose IV infusion to prevent precipitating Wernicke's encephalopathy. Excessive alcohol consumption has also been linked to zinc and copper deficiency, which seems to be associated with a decreased quality of physical and mental life [53].

There has been one randomized trial testing the administration of 600 mg of benfotiamine in alcohol dependent patients—this thiamine analog was associated with a reduction of alcohol consumption, but no metabolic variables were tested [54]. In the absence of other randomized trials, most of the experts advocated the empiric administration to alcohol dependent patients of multivitamin cocktails, including in particular, thiamine (200–300 mg), folic acid, vitamin B6, and vitamin C upon admission.

#### *3.2. Anemia*

A third of the world's population is affected by anemia, and iron deficiency is involved in 50% of the cases [18]. The most common symptoms are paleness, fatigue, dyspnea and headache. Laboratory tests have shown low blood haemoglobin concentration (Hb < 130 g/L in men, <120 g/L in women and <110g/L in pregnancy) with microcytosis, hypochromia and serum ferritin below 30 μg/L. Iron deficiencies can be physiological (e.g., pregnancy), or pathological in case of blood loss (e.g., surgery, trauma, digestive tract bleeding), malabsorption (e.g., celiac disease, gastrectomy), chronic disease (cancer, chronic heart failure) or in some genetic disorders [19]. In the presence of iron-deficiency anemia, investigations to identify a cause of blood loss or malabsorption are required. Preventive iron supplementation should be prescribed in at-risk patients using the oral route, while treatment of in patients with a positive diagnosis [19], may require the intravenous route due to the frequent gastric intolerance and poor absorption of oral supplements.

Macrocytic anemia can be observed in vitamin B12 or/and vitamin B9 (folate) deficiencies. Vitamin B12 deficiencies occur in cases of severe malabsorption (bariatric surgery, gastrectomy, or autoimmune gastritis), in the abuse of nitrous oxide, and in cases of inherited metabolic disorders. Vitamin B12 supply is recommended after confirmed diagnosis. This treatment will be lifelong when the etiology of the deficiency is irreversible or unknown [55].

Copper deficiencies impair the activity of hephaestin, a copper-dependent ferroxidase responsible for transporting dietary iron from intestinal enterocytes into the circulatory system—its depression leads to iron deficiencies and low hemoglobin. Copper deficiency anemia has been treated with oral or intravenous copper supplementation [56]. Cobalt, a component of hydroxycobalamin, is considered one the most stimulator of erythrocyte production. In clinical practice, Cobalt administration is rare considering the high risk of toxicity of cobalt salts in humans [57].

#### *3.3. Cardiomyopathies and Heart Failure*

According to the European Society of Cardiology (ESC), cardiomyopathy (CM) is defined as a myocardial disorder in which the heart muscle is structurally and functionally abnormal, in the absence of coronary artery disease, hypertension, valvular disease and congenital heart disease [58]. CM can lead to heart failure (HF). Deficiencies in thiamine and selenium are nutritional factors that may be involved in the occurrence of such myocardial disorders [58,59] in malnutrition conditions, malabsorption, or exclusive parenteral nutrition (PN) [60–62]. Selenium deficiency cardiomyopathy is known as the Keshan disease in humans. It was initially described in China in 1935 in selenium-poor soils, with multiple case reports. In addition to the low plasma selenium levels, low glutathione peroxidase-1 activity has been reported in animals [63]. Some cases reported have been reversible by short-term oral or IV administration, and others have been fatal [60–63].

Thiamine plays a fundamental role in cellular metabolism, especially in the carbohydrate pathway. Severe and chronic thiamine deficiencies are known as Beriberi disease. The classic presentation includes neurologic features and encephalopathy. HF symptoms have been less frequent and have been associated to metabolic acidosis and hyperlactatemia (due to an inhibition of the pyruvate

deshydrogenase) [64,65]. A fulminant form has also been described [64]. It has been shown that about 40% of the patients hospitalized for HF presented thiamine deficiencies [66]. Animal models have demonstrated that this deficiency causes cardiac disorders (cardiac hypertrophy, depressed cardiac contractility, and dysrhythmias) in the absence of beriberi. Finally, in human studies, the benefit of thiamine supplements in cases of chronic HF are unclear [66].

Iron deficiencies in HF are common (about 30–50%) in patients with chronic HF, independent of anemia, and have lead to skeletal muscle dysfunction. Recent ESC guidelines recommended screening for iron deficits in HF patients [28,59].

Recently, an association between vitamin D deficiency and HF has been suggested [67]. Vitamin D promoted cardioprotection in animals (anti-inflammatory, anti-apoptotic and anti-fibrotic mechanisms) [67].

#### *3.4. Inflammatory Bowel Disease*

Malnutrition is present in the vast majority of patients with inflammatory bowel disease (IBD), the deficiencies being more prevalent in Crohn's disease compared with ulcerative colitis and more important in active diseases [68]. Micronutrient deficiencies are essentially explained by the reduced dietary intake and the underlying malabsorption. They have been associated with prolonged hospitalization and higher mortality [69].

Fat-soluble vitamins are particularly prone to deficiency. A high prevalence of vitamin A and E deficiencies have been reported [70], but also of vitamin D, which has been suspected to play a role in the pathogenesis of IBD [71]. Vitamin K deficiencies are also frequent and correlated with disease activity [72].

Folate deficiencies are common in IBD and aggravated by treatment, such as sulfasalazine or methotrexate. Vitamin B12 has also been frequently observed, especially in Crohn's disease and after ileal resection of ≥30 cm [73]. Thiamin deficiency has also been reported, especially in IBD patients treated with parenteral nutrition [74].

Among trace elements, selenium deficiencies have been reported and may increase the severity of gut inflammation—the repletion data are conflicting [75]. Zinc deficiencies are also prevalent [76]. Iron deficiencies are very frequent and are the leading cause of anemia in patients with IBD. In the presence of inflammation, this diagnosis can be challenging and relies on the values of serum ferritin values.

#### *3.5. Liver Disease*

The liver plays a crucial role in maintaining systemic Zn homeostasis [77]. Chronic liver disease, such as chronic hepatitis, liver cirrhosis, or fatty liver, impairs Zn metabolism, and has resulted in Zn deficiency, which in turn has caused multiple metabolic abnormalities, including insulin resistance, hepatic steatosis and hepatic encephalopathy. Zn deficiency may also favor carcinogenesis of hepatocellular carcinoma (HCC). In chronic liver disease, low levels of selenium [78] have been generally observed. In comparison, copper levels have often been elevated [79]. It is argued that doses required to achieve an effect in chronic hepatitis are far beyond DRI with 150–200 mg/day [80].

In cases of liver disease secondary to alcoholism, please see 3.1 (same picture). Group B vitamin deficiencies, especially thiamine, are common in cirrhosis [81]. Unlike observations for alcoholic patients without liver disease, vitamin B12 levels have been frequently elevated in viral hepatitis, cirrhosis and hepatocellular carcinoma (HCC) [82]. This increase has been explained by the cytolysis of hepatocytes, and vitamin B12 being mainly stored in the liver.

Fat-soluble vitamin deficiencies have been observed in cases of alcoholism, but also in cholestasis with malabsorption and bile salt deficiency [83]. Levels of vitamins A, D and E should therefore be routinely checked as well as prothrombin time. However, prolonged prothrombin time does not purely reflect vitamin K deficiencies, but also reduced levels of coagulation factor V. Importantly, vitamin D has pleiotropic effects for liver disease, including anti-inflammatory, immune-modulatory and anti-fibrotic properties in addition to its classical skeletal effects. This may impact on disease

progression [84], especially in HCC and non-alcoholic steatohepatitis, although any benefit from its repletion has not been formally proven by prospective studies [85].

#### *3.6. Obesity & Bariatric Surgery*

In patients with grade III obesity (body mass index <sup>≥</sup> 40 kg·m−<sup>2</sup> or <sup>≥</sup> 35 kg·m−<sup>2</sup> with comorbid conditions), bariatric surgery has become common. In 2013, over 450,000 bariatric surgeries were performed worldwide. Roux-en-Y gastric bypass, adjustable gastric band, and sleeve gastrectomy were the most frequent procedures [86,87]. Micronutrient deficiencies were observed before (in obese patients, vitamin C deficiency in about 40% and zinc deficiency is up to 50%) and after the surgery. Due to fat malabsorption and maldigestion, all fat-soluble vitamins are at risk. The bypass of the duodenum and proximal jejunum lead to thiamine deficiencies [86], leading to a risk of clinical Wernicke's encephalopathy [88]. Vitamin B9 is approximately 10%. Vitamin B12 deficiencies have been widely described in the literature because of their neurologic complications, especially an acquired myelopathy with paresthesias, ataxia and muscle weakness [89]. It is recalled that the intrinsic factor produced in the stomach is needed for ileal absorption of vitamin B12. The severity of a vitamin B6 deficiencies vary from peripheral neuropathy to seizures [89]. The common trace elements deficiencies after bariatric surgery include copper (absorbed in the stomach and the duodenum), iron (absorbed in the duodenum), and zinc (absorbed in the jejunum). The prevalence of copper deficiency has been reported to be as high as 90% post- surgery (70% for zinc). Further, systematic supplements of micronutrients are recommended after bariatric surgery by the American Society for Parenteral and Enteral Nutrition (ASPEN) [86] and the American Society for Metabolic and Bariatric Surgery clinical practice guidelines (ASMBS) [90].

#### *3.7. Kidney Disease*

Kidney disease, whether chronic (CKD) or acute, affects micronutrient homeostasis and might lead to either deficiency or toxic excess. Indeed, a decreased glomerular filtration rate might lead to accumulation of molecules normally excreted by the kidney such as selenium. On the other hand, renal replacement therapy when applied might lead to uncompensated losses of other micronutrients (copper). Finally, the loss of renal activation might lead to decreased biological activity (vitamin D). Unfortunately, the authors knowledge remains limited [91].

Anemia is commonly observed in CKD. Beyond the lack of erythropoietin, which is now clearly established and easily administered, it might also be associated with alterations in iron metabolism and inflammation. Oral corrections of iron deficiencies in pre-dialysis CKD patients have been shown to be an efficient option in a recent randomized trial [92]. The role of hepcidin, a key regulator of circulating iron level in CKD associated anemia is increasingly recognized. Indeed, the condition has been associated with elevated hepcidin serum levels as it is typically excreted by the kidney [93]. This leads to reduced iron availability and anemia. Its measurement and potential anti-hepcidin therapies could help managing anemia in CKD [27,94].

Subclinical vitamin K deficiencies have been shown to be clinically relevant as requirements have increased due to the vitamin K-dependent proteins required to inhibit calcification [95]. This vitamin governs the gamma-carboxylation of matrix Gla protein for inhibiting vascular calcification, and the vitamin D binding protein receptor is related to vitamin K gene expression [96]. Deficiency may favor vascular calcification.

Chronic dialysis is typically associated with elevated oxidative stress leading to low levels of zinc, selenium and GPX. A French team showed three decades ago that weekly administration of selenium with zinc was able to restore GPX activity and reduced thiobarbituric acid reactants (TBARs) plasma concentrations [97]. In a large cohort of 1278 patients on incident hemodialysis, it was observed that lower selenium and zinc concentrations were strongly and independently associated with death and all-cause hospitalization [98]. Trimestrial monitoring of selenium and zinc may thus be justified, with repletion in case of low values.

In acute kidney failure requiring continuous renal replacement therapy, other micronutrients such as thiamine and copper will be lost in the effluent fluid [39,40,99]. Copper losses causing low blood levels can be associated with severe arrhythmias and wound healing complications. Very low plasma levels (<8 mmol/L) might require active intravenous repletion with doses 6–10 mg/day, i.e., 5–10 times the usual DRI [40].

#### *3.8. Migrant Populations*

Being a migrant is not a disease, but a difficult social condition frequently associated with malnutrition for multiple reasons ranging from insufficient food intake, to exposure to unusual food or unbalanced diet due to incapacity to find the traditional foods. Migrants represent a growing category of inpatients. Micronutrients of concern have shown to be retinol, vitamin D, magnesium, potassium, copper, and selenium [11], in addition to iron.

#### *3.9. Laboratory Investigations*

Laboratory investigations of micronutrient deficiencies have often not been systematic except for 2 conditions: (1) Anemia outwork, which generally included determination of blood Vitamin B6, B12, iron, ferritin and transferrin; (2) screening before and the follow-up after bariatric surgery included thiamine, vitamin B12, folic acid, iron, zinc, copper, calcium and the liposoluble vitamins D, A, E and K were recommended [90]. In other conditions, the diagnosis lacked standardization. A pragmatic approach upon admission or during hospitalization could be to draw an additional blood sample for further diagnostic outwork, and to empirically administer multi-micronutrients without delay.

Inflammation causes a redistribution of micronutrients between the various compartments and generally reduces circulating levels. The intensity of inflammation has been reflected by CRP levels [24]. The example of vitamin D has been emblematic. While a CRP > 80 mg/L has been associated with a reduction of its blood concentration by 40% below reference ranges [24], CRP has never been mentioned in the vitamin D trials. CRP should belong to any micronutrients outwork as low levels do not necessarily indicate a deficiency.

#### **4. Micronutrient Unavailability as Cause of Deficiency**

#### *4.1. Nutritional Sources*

Hospital related malnutrition is a well know entity that is observed worldwide and is related to being bedridden [100]. A large proportion of patients only consume a third or half of the proposed meals [4]. In our hospital, a standard daily serving provides 1700–1800 kcal. If only half is consumed, the daily micronutrients cannot be covered.

The sickest patients are fed with enteral nutrition. Due to regulatory constraints, the industry must respect micronutrient recommendations intended for the general healthy population, the previously mentioned DRI. The concentrations of the products are calculated for feeding doses varying between 1500 and 2500 kcal/day. However, worldwide it has been shown that many patients receive no more than 1000 kcal/day by this route [101]. In addition, the most recent nutrition guidelines for critical care patients recommended ramping up the feeding over several days [102]. The consequences might be an even further reduction of nutrition delivery. By design, the enteral feeding solutions will not be able to cover needs as long as quantities below 1500 kcal per day are provided. Table 3 shows the detailed micronutrient provision for 1000 kcal/day provided by the 10 most frequently used enteral feeding solutions provided by 4 international companies on the Swiss market. While intakes below DRI of fluor and iodine may be less important during an acute phase, low intakes of iron and of the B vitamin group are a concern considering their essential role in energy (ATP synthesis) and carbohydrate metabolism. Moreover, as absorption is unreliable and as needs might be higher, several micronutrients such as vitamin C are just in the reference range, which may be insufficient.


*J. Clin. Med.* **2019**, *8*, 931

Regarding these low micronutrient doses, it might be justified to deliver standard multivitamin and trace element products daily providing DRI doses to the majority of inpatients. This strategy has been applied for many years in the Lausanne university hospital's ICU, as critically ill patients do have higher needs. In the sickest patients, multi-micronutrients cocktails have been delivered IV for the first 5 days [33], resulting in a shortening of the hospital stay, particularly in major trauma patients.

#### *4.2. Geriatric Population*

The qualification, elderly, encompasses patients aged 60 to over 100 years. While the younger seniors often are fit until the seventies, some physiological changes already occur that become exacerbated with growing age. The elderly often present with anorexia which is considered a complex geriatric syndrome and a risk factor for frailty [9]. These changes are associated with lower weight, and lower energy expenditure. The body undergoes specific changes—the gastric mucosa tends to atrophy, reducing vitamin B12 absorption. Indeed, the decline in vitamin B12 is independent of nutrition but caused by a decline of both the intestinal uptake and the renal reabsorption system for vitamin B12 [103]. The elderly also require higher doses of vitamin B6 and D to maintain health, which has been integrated into DRI recommendations for older subjects. Deficiencies can be overcome by supplementation, as shown by a large randomized controlled trial including 652 geriatric patients—daily oral nutrition supplements enriched with proteins, hydroxy-methyl-butyrate, vitamin D and other micronutrients reduced mortality [104].

#### *4.3. Partial or Complete Starving upon Hospital Admission and Refeeding*

The incidence and importance of the refeeding has often been underestimated [105]. The absence of a uniform definition participates in its underestimation [105]. Refeeding syndrome consists of metabolic changes that occur on the reintroduction of food or simply a glucose infusion. A few days of feeding grossly below needs will be sufficient to create the metabolic crisis which is characterized by sudden shifts in the electrolytes that are needed for energy and substrate (mainly glucose) metabolism. The NutritionDay survey, an initiative that analysed the relation between nutritional intakes and outcomes of a wide range of institutionalized and hospitalized patients worldwide, has shown that more than half of the patients admitted to hospital were eating less than half of their normal food intake before admission [106]. This places the majority of hospitalized patients at risk of a refeeding syndrome.

Some categories of patients, such as chronic, alcohol consumers, which are largely prevalent in western countries, being present in nearly 30% of hospital admissions, are at higher risk of thiamine deficiencies—these patients are at particularly high risk of refeeding syndrome and its worst neurologic complication, the Wernicke encephalopathy. A recent review of the literature confirmed the importance of administering intravenous thiamine to these patients in order to prevent severe sequelae [107]. The recommended doses ranged from 50–100 mg/day to 250–500 mg 3 times a day. The IV route was recommended due to the frequent presence of gastritis in these patients, which reduced absorption.

#### *4.4. Economic Considerations*

A Canadian prospective cohort study showed that approximately 40% of the 956 patients admitted to hospital were moderately to severely malnourished. These patients had longer hospital stays, and as a result, cost more than the well-nourished patients [108]. A European narrative review showed that malnutrition increased the length of hospital stays by 2.4 to 7.2 days [5]. Malnutrition led to an additional individual cost ranging between 1640 € and 5829 €.

Clinical evidences are lacking for empirical multi-micronutrient supplements. The cost of malnutrition attributable to micronutrient deficiencies have not been assessed in adult inpatients and several studies have failed to demonstrate significant beneficial effects of various micronutrient supplements in the general population [109]. Nevertheless, there are data suggesting a benefit, at least in the sickest patients [33,110].

However, data exists for children. A study focusing on estimates of disability-adjusted life years and their monetization showed that short-term economic costs of micronutrient malnutrition in India amounted to 0.8% to 2.5% of the gross domestic product [14]. The health and cost consequences of iodine, iron, vitamin A, and zinc deficiencies were assessed in Pakistani children: Societal costs amounted to 1.44% of gross domestic product and 4.45% of disability-adjusted life-years in Pakistan in 2013, which hindered the country's development [111].

When deciding about an empirical administration of micronutrients, the analytical costs of deficiency diagnosis must be considered. The European Society for Clinical Nutrition and Metabolism (ESPEN) monitoring recommendations indicate that some vitamin and trace element analysis (inductively coupled plasma mass spectrometry: ICP-MS) are expensive [112]. They are actually more expensive than the empirical administration. A semi-automatic weekly determination of blood selenium levels in our ICU resulted in major costs that could be contained by the decision to let blood sampling be prescribed only by the ICU nutritionists and dieticians in patients at risk [113].

A multi-micronutrient tablet costs 0.80 € (the IV dose in Table 1 costs 25 €), which is negligible compared to the cost of one day in hospital, or worse, in the ICU. This prescription is likely to be beneficial if the administration is standardized, and limited to the first week of hospitalization, and to patients on enteral feeding.

#### **5. Conclusions**

Micronutrient deficiencies and borderline status are more frequent than generally acknowledged. The most important potential acute deficiency that may compromise outcome is thiamine deficiency. Other deficiencies will impact on immune defenses and anabolic capacity. Therefore, an empirical and cheap complementation strategy, based on daily oral multi-micronutrient products providing DRI, may be justified for hospital inpatients for one week. It is important to state that the evidence from trials is still missing.

**Author Contributions:** Conceptualization M.M.B.; Search methodology M.M.B., N.B.-H., O.P.; validation, M.M.B., O.P., A.S. and N.B.-H.; writing—original draft preparation, M.M.B., O.P., N.B.-H.; writing—review and editing, N.B.-H., O.P., A.S., and M.M.B.

**Conflicts of Interest:** None of the authors have any conflicts of interest to declare.

#### **Abbreviations**


RDA recommended daily allowance

#### **References**

1. Cederholm, T.; Jensen, G.L.; Correia, M.I.T.D.; Gonzalez, M.C.; Fukushima, R.; Higashiguchi, T.; Baptista, G.; Barazzoni, R.; Blaauw, R.; Coats, A.; et al. GLIM criteria for the diagnosis of malnutrition—A consensus report from the global clinical nutrition community. *Clin. Nutr.* **2019**, *38*, 1–9. [CrossRef] [PubMed]


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Early Supplemental Parenteral Nutrition in Critically Ill Children: An Update**

#### **An Jacobs, Ines Verlinden, Ilse Vanhorebeek and Greet Van den Berghe \***

Clinical Division and Laboratory of Intensive Care Medicine, Department of Cellular and Molecular Medicine, KU Leuven University Hospital, 3000 Leuven, Belgium; an.jacobs@kuleuven.be (A.J.);

ines.verlinden@kuleuven.be (I.V.); ilse.vanhorebeek@kuleuven.be (I.V.)

**\*** Correspondence: greet.vandenberghe@kuleuven.be; Tel.: +32-16-344021

Received: 13 May 2019; Accepted: 5 June 2019; Published: 11 June 2019

**Abstract:** In critically ill children admitted to pediatric intensive care units (PICUs), enteral nutrition (EN) is often delayed due to gastrointestinal dysfunction or interrupted. Since a macronutrient deficit in these patients has been associated with adverse outcomes in observational studies, supplemental parenteral nutrition (PN) in PICUs has long been widely advised to meeting nutritional requirements. However, uncertainty of timing of initiation, optimal dose and composition of PN has led to a wide variation in previous guidelines and current clinical practices. The PEPaNIC (Early versus Late Parenteral Nutrition in the Pediatric ICU) randomized controlled trial recently showed that withholding PN in the first week in PICUs reduced incidence of new infections and accelerated recovery as compared with providing supplemental PN early (within 24 h after PICU admission), irrespective of diagnosis, severity of illness, risk of malnutrition or age. The early withholding of amino acids in particular, which are powerful suppressors of intracellular quality control by autophagy, statistically explained this outcome benefit. Importantly, two years after PICU admission, not providing supplemental PN early in PICUs did not negatively affect mortality, growth or health status, and significantly improved neurocognitive development. These findings have an important impact on the recently issued guidelines for PN administration to critically ill children. In this review, we summarize the most recent literature that provides evidence on the implications for clinical practice with regard to the use of early supplemental PN in critically ill children.

**Keywords:** Pediatric Intensive Care Unit; enteral nutrition; early parenteral nutrition; critical illness

#### **1. Introduction**

Optimal nutritional support is considered of paramount importance for critically ill children admitted to the pediatric intensive care unit (PICU), since malnutrition and inadequate nutrient delivery have been associated with worse clinical outcome [1,2]. Moreover, critically ill children have limited macronutrient stores and relatively higher energy requirements than adults admitted to the intensive care unit (ICU), which can lead to substantial caloric and macronutrient deficits [2–4]. The feeding is thought to attenuate the metabolic stress response, prevent oxidative cellular injury and modulate immune responses, and has led to a shift from nutritional support as adjunctive care to actual therapy of the critically ill child [5]. The enteral route is preferred for providing nutrition [5]. However, critically ill children are often too ill to be fed normally by mouth, and nasogastric or nasoduodenal tube feeding is often not tolerated because of gastric dysmotility or ileus. Interruption of enteral feeding also occurs frequently for various reasons, like medical or surgical contraindications, or radiology, bedside or surgical procedures [6]. Therefore, parenteral nutrition (PN) is often initiated to supplement the insufficient enteral intake. Nonetheless, official guidelines on timing and thresholds of initiation, composition and doses of supplemental PN vary widely [5,7–9]. Moreover, concerns about overfeeding have led to even more uncertainty [9]. A recent survey showed significant differences

in nutritional practices in PICUs worldwide, in terms of macronutrient goals, estimation of energy requirements, timing of nutrient delivery and thresholds for starting supplemental PN [10]. In this review, we summarize the most recent literature findings affecting evidence and clinical practices with regard to the use of early PN in critically ill children. We searched PubMed up to April 2019, without language restrictions, using different combinations of the search terms "parenteral nutrition", "PICU", "early" and "pediatric critical illness". We focused on publications of the last eight years, discussed in the context of earlier work.

#### **2. Timing of PN Initiation**

Several observational studies have shown that malnutrition is associated with worse clinical outcome [1,2,11]. A macronutrient deficit has been associated with infections, weakness, prolonged mechanical ventilation and delayed recovery. For that reason, guidelines used to recommend that when provision of enteral nutrition (EN) is insufficient, impossible or contraindicated, supplemental PN should be initiated [7,12,13]. However, observational studies cannot assign causality to an association. Hence, the association between inadequate nutrition and worse clinical outcome might merely exist because of a non-optimal nutritional support for the sickest children, which are at the highest risk of adverse outcome. Although it seems intuitive that providing early nutrition will be beneficial, it does not necessarily mean that nutritional support in the early phase of critical illness will improve clinical outcome [9]. In critically ill adults, the large multicenter EPaNIC (Early versus Late Parenteral Nutrition in ICU, *n* = 4640) randomized controlled trial (RCT) showed that withholding supplemental PN until day eight of an ICU stay (late PN), and thus accepting a substantial macronutrient deficit, was associated with fewer ICU infections, a shorter duration of mechanical ventilation and renal-replacement therapy and a shorter ICU and total hospital stay as compared with initiating supplemental PN early (within 48 h after ICU admission) [14]. Data generated by the broad international yearly survey of clinical nutrition practices "nutritionDay" revealed an important change in the pattern of PN prescription after publication of the EPaNIC results (Personal communication kindly shared by Prof. Dr. M. Hiesmayr, nutritionDay Project Leader). As compared with adults, critically ill children have limited stores of energy, fat and protein, as well as relatively higher energy requirements [2,3]. Since this makes them more vulnerable to a substantial caloric and macronutrient deficit, the effect of withholding supplemental PN in critically ill children could be different than in adults. Therefore, a multicenter PEPaNIC RCT (Early versus Late Parenteral Nutrition in the Pediatric ICU) was conducted [15] that investigated the same intervention in 1440 critically ill children aged 0–17 years in three PICUs in Belgium, the Netherlands and Canada. Withholding supplemental PN during the first week in critically ill children resulted in fewer new infections, a shorter dependency on mechanical ventilation and general intensive care and a shorter hospital stay as compared with providing PN early (within 24 h after PICU admission; see Figure 1). The clinical superiority of late PN was more pronounced in children than it was in adults, and was shown irrespective of diagnosis, severity of illness, risk of malnutrition or age of the child [15]. This last finding was surprising, since neonates are more susceptible to macronutrient deficits than older children [7], raising concerns by experts [16–18]. To address these concerns, a secondary analysis of the PEPaNIC trial was performed to investigate the effects of withholding PN for one week in 209 critically ill neonates who did not (or could hardly) tolerate EN [19]. Analyses were performed for term neonates aged up to four weeks, up to one week and younger than one day. Late PN resulted in fewer nosocomial infections in neonates aged up to one week and younger than one day, and in shorter dependency on intensive care and mechanical ventilation for all studied age groups of neonates. Hence, term neonates also benefited from withholding PN during the first week in the PICU, which is in agreement with findings for older children and adults [14,15]. Moreover, there was a more pronounced benefit of late PN in the youngest children, as shown in Figure 1. Since a macronutrient deficiency is presumed to be more detrimental during acute illness in undernourished children [5], a second subanalysis of the PEPaNIC RCT was performed, investigating the effects of withholding supplemental PN during the first PICU week in a subgroup of critically ill children who were undernourished upon

admission to the PICU [20]. Undernourishment was defined as a weight-for-age z score lower than −2 in children younger than one year, and a body mass index-for-age z score lower than −2 in children one year or older. This identified 289 of 1440 PEPaNIC patients (20%) with undernourishment upon PICU admission. Among the undernourished patients, late PN reduced the absolute risk of new infections and shortened the duration of PICU stay. These effect sizes of late PN were even larger than in the main trial cohort of the PEPaNIC RCT. Late PN did not affect the safety outcomes of mortality, incidence of hypoglycemia or weight deterioration during PICU stay in the undernourished patients. A larger longitudinal study of all PEPaNIC patients with weight z scores available on admission and on the last day in the PICU showed that weight deterioration during PICU stay was associated with worse clinical outcomes, but that withholding supplemental PN during the first week did not alter weight z score deterioration during the PICU stay [21].

**Figure 1.** Bars represent incidence of (**A**) the risk of acquiring a new infection in PICUs (percentage), (**B**) duration (days) of PICU stay, (**C**) duration (days) of mechanical ventilation, and (**D**) duration (days) of total hospital stay. Whiskers indicate standard errors of the mean. Asterisks represent *p* values < 0.05 obtained with multivariable analysis adjusting for baseline risk factors (treatment center, age, risk of malnutrition (STRONGkids score), diagnosis upon admission and severity of illness (Pediatric Logistic Organ Dysfunction (PeLOD) score, and Pediatric Risk of Mortality 2 (PIM2) score) for all patients; treatment center, risk of malnutrition, diagnosis upon admission and severity of illness for children and infants; and treatment center, type of illness upon admission (medical, surgical cardiac, surgical other), severity of illness and weight for age z score for neonates).

The benefits of withholding supplemental PN during the first week in the PICU appeared not only present from a clinical point of view, but also from a health–economic perspective. A cost-effectiveness study indeed showed that the total direct medical costs were considerably lower with late PN as compared with early supplemental PN initiation [22]. This cost saving was beyond the expected lower costs for the use of PN itself, since avoidance of new infections by late PN yielded the largest cost reduction.

A possible limitation of the PEPaNIC RCT is the use of standard equations for the estimation of energy requirements instead of indirect calorimetry [23]. However, the use of indirect calorimetry for estimating energy expenditure does not seem to be accurate [24], or feasible [25], and is not frequently used in daily practice [10,26].

Apart from the PEPaNIC RCT, no other randomized controlled trial investigating the use or timing of supplemental PN in critically ill children has been published in the last eight years. A limited number of observational studies on the use of supplemental PN and over- and underfeeding in PICUs showed different results [2,24,27]. A retrospective single center study showed that late initiation of supplemental PN was associated with a higher nosocomial infection rate as compared with early initiation of supplemental PN [27]. In contrast, an observational study in 31 PICUs showed that the use of PN in general was associated with higher mortality [2]. Another retrospective study determining the incidence of over- and underfeeding in 139 children admitted to a tertiary PICU showed that underfeeding was associated with shorter durations of PICU and hospital stays, as well as with fewer ventilation days, as compared with appropriately fed and overfed patients [24]. However, the observational design of these studies holds a risk of bias by confounding variables, especially in nutritional research [28]. Therefore, comparison with the results of the PEPaNIC RCT is challenging. Further randomized controlled trials are warranted to determine the ideal time point for initiation of supplemental PN in PICUs.

#### **3. Early PN Composition and the Role of Macronutrients**

Although extensive guidelines on the composition of PN in critically ill children are available [5], a recent survey on nutritional practices in PICUs worldwide showed a wide variation in parenterally administered doses of protein, lipids and glucose [10]. Protein targets in particular seem to be a point of discussion. Several studies have assessed the association between protein delivery to and clinical outcome of critically ill children [2,29,30]. In an observational international cohort study that included 500 critically ill children, mortality at 60 days was higher in patients who received PN independently of the amount of energy or protein intake [2]. However, an important severity of illness bias has to be taken into account, since patients who are less sick are more likely to better tolerate EN. The study adjusted for severity of illness using admission scores, but data for calculating this severity of illness score were missing in 31% of the included patients, and the choice of severity of illness score differed between the participating centers [2]. Another large multicenter observational study of the same group showed an association between higher enteral protein intake and lower odds of mortality in more than 1200 mechanically ventilated critically ill children [29]. The effect was dose-dependent, and independent of energy intake. Again, the incomplete datasets and the lack of uniform usage for severity of illness scores, and the substantial number of patients who received EN, could potentially bias these observations. The authors reasoned that an increased demand in amino acids in catabolic disease, such as critical illness, could contribute to increased higher protein degradation from muscle to ensure bodily functions [29], which is associated with poor outcome [31]. By providing proteins, the synthesis of muscle proteins might be boosted, and thereby muscle loss could be prevented, possibly limiting the severity of intensive care unit-acquired weakness [32]. Nevertheless, a preplanned secondary analysis of the adult EPaNIC study did not support this concept, as increased macronutrient intake with early PN, including more amino acids, did not counteract muscle atrophy and actually increased the risk of developing clinically relevant muscle weakness in the ICU [33]. Interestingly, in a preplanned secondary analysis of the PEPaNIC RCT, the dose of amino acids was actually associated with more infections and a longer dependency on mechanical ventilation and other intensive medical care in children admitted to the PICU [30]. This risk of harm associated with early amino acid administration was elevated even at low doses of administered amino acids. A possible explanation for the difference between these results and the ones from the previously cited observational studies is the randomized design of the PEPaNIC trial, in which the doses of macronutrients differed from patient to patient and ranged widely [30]. In critically ill adults, three RCTs could not show benefits from early amino

acid supplementation [34–37], but clinical trials on the effects of amino acid administration on clinical outcome in critically ill children in a randomized manner are lacking [38]. In contrast with the harm of amino acid administration, the secondary analysis of the PEPaNIC RCT suggested a benefit of glucose and lipid administration. Indeed, administering more glucose during the first three days of PICU stay was independently associated with fewer infections, and administering more lipids was independently associated with earlier PICU discharge [30]. Clearly, large-scale prospective RCTs in critically ill children are needed to identify the optimal composition of supplemental PN [9,38].

#### **4. Impact of Early PN on Long-Term Outcome**

Because of new insights in diagnostic and therapeutic measures in the field of pediatric critical care medicine [39], centralization of care [40] and specialized staff training and education [41,42], there has been an important decline in the mortality rate in PICUs over the last decades [43,44]. However, this improved survival has led towards a shift to considerable long-term morbidity, years after discharge [45,46]. This has been most thoroughly documented in regards to impairment of neurocognitive development, but it also includes growth retardation and may comprise poor physical functioning and reduced quality of life [47–50]. The fact that children are treated in the PICU during crucial developmental phases likely plays a role. Interestingly, it appears that, to a certain extent, neurocognitive outcome is modifiable, as shown by the attenuation of neurocognitive impairment with the prevention of hyperglycemia during intensive care [51]. Treatments in PICUs that have been shown to cause neurodevelopmental harm, such as anesthetic and analgesic agents [52,53] and toxicants such as phthalates that leach from indwelling medical devices [54], may also be targets for research into safer alternatives. Concerning nutrition in PICUs, in relation to long-term outcome, experts were concerned about the safety of withholding early supplemental PN in neonates in view of the more frequent episodes of hypoglycemia observed in the late PN arm of the PEPaNIC RCT [15,17]. However, in a previous large randomized controlled trial investigating the effect of tight glycemic control on morbidity and mortality in PICUs and on long-term neurocognitive development, a high incidence of brief hypoglycemia with tight glycemic control was not associated with harm to neurocognitive development, as documented four years later [51]. The proportion of neonates included in the PEPaNIC RCT was similar to that in the tight glycemic control trial [15,55]. In a preplanned two-year follow-up study, in which all patients included in the PEPaNIC RCT were approached for possible assessment of physical and neurocognitive development, exposure to hypoglycemia also did not associate with the investigated long-term outcomes [53]. Moreover, the main results of this follow-up study showed no adverse effects of withholding supplemental PN during the first week in PICUs on survival, anthropometrics, health status and neurocognitive development. In fact, omitting early supplemental PN in PICUs improved parent-reported executive functioning (inhibition, working memory, metacognition and overall executive functioning), externalizing behavioral problems and visual-motor integration two years later, as compared with early supplemental PN. In particular, a better inhibitory control was observed (Figure 2). Since poor inhibitory control in children contributes to impulsive and destructive behaviors that upset or harm others [56], delaying supplemental PN can have important consequences on daily life and social environments later in life. The long-term effects of late versus early supplemental PN were more pronounced in patients who were younger than one year of age at the time of PICU admission as compared with older children. This age-dependent vulnerability supports the hypothesis that the harm induced by early supplemental PN might be caused by a direct metabolic insult to the developing brain, since it was not statistically explained by the acute effects of the intervention itself, such as the increased incidence of new infections or delayed recovery. However, further research is warranted to unravel the underlying mechanisms that would provide support for this hypothesis. Although long-term outcomes and quality-of-life years after PICU discharge have gained great importance in research [45,46], investigating these outcomes is logistically challenging, expensive and time-consuming. To investigate whether these long-term effects persist or

change over time, a four-year follow-up study of the PEPaNIC RCT is currently ongoing, of which the results are expected by 2020.

**Figure 2.** Density estimates for inhibitory function as reported by parents or caregivers. Densities, which correspond to the proportions of children with a certain score (equivalent to a smoothed histogram), are shown separately for healthy control children and for PEPaNIC participants who were randomly assigned to receive late PN or early PN. Higher scores indicate worse functioning. Each line corresponds to one of 31 imputed datasets. Figure reprinted from [53] with permission from Elsevier.

#### **5. Potential Mechanisms Underlying Harm by Early PN**

Several mechanisms may contribute to an adverse clinical impact of administering PN in the early phase of critical illness in children. As early PN has affected both short-term and long-term outcomes of patients, carryover effects persisting in the long term must play a role.

#### *5.1. Potential Mechanisms Underlying the Adverse Impact of Early PN on Short-Term Outcome*

As discussed, it was the administration of more amino acids that was associated with adverse short-term outcomes evoked by early PN in the PEPaNIC study [30]. The mechanisms underlying the potential harm of amino acids in this context remain to be unraveled, but it is possible to speculate. First, amino acids are powerful suppressors of autophagy [57], a pathway shown to be crucial in critically ill humans and animals for innate immunity and the removal of cellular damage [58,59]. Autophagy activation appeared suppressed in muscle from randomized adult patients exposed to early PN as compared with late PN in the EPaNIC trial, and amino acids in particular suppressed autophagy activation in a rabbit model of prolonged critical illness [33,58,59]. Second, macronutrients and amino acids in particular have long been assumed to counteract the hypercatabolic response to critical illness, which mobilizes amino acids by (mainly muscle) protein breakdown to guarantee substrate delivery to vital tissues [60]. More specifically, administration of exogenous amino acids was thought to circumvent the need for endogenous amino acid release and to stimulate muscle protein synthesis. However, once amino acid doses exceed anabolic capacity, they no longer stimulate muscle protein synthesis but are shuttled towards the liver for production of urea [61]. Such increased plasma urea concentrations were observed in early PN patients in the PEPaNIC RCT as compared with late PN patients [15]. In critically ill adults who received early PN, the administered amino acids did not counteract muscle wasting [33], but increased plasma urea and urinary nitrogen excretion (with a net waste of 63% of the extra nitrogen intake from early PN in the first two weeks) [62], which can cause

harm to both liver and kidney. These results are in line with the EAT-ICU trial (Early goal-directed nutrition in ICU patients) results, in which increased urea production—but no beneficial effect on outcome—was seen with early goal-directed nutrition versus standard nutritional care in critically ill adults [63]. Altogether, these findings suggest that muscle wasting during the acute phase of critical illness may be considered an adaptive response designed to provide substrates for gluconeogenesis in order to meet the energy requirements of vital organs [32].

#### *5.2. Potential Mechanisms Underlying the Adverse Impact of Early PN on Long-Term Outcome*

Underlying mechanisms of the long-term harm caused by early PN in the context of critical illness remain largely unraveled. The clinical benefits of late PN observed far beyond the intervention window suggests that early PN induces carry-over "memory" effects with a negative impact on long-term outcome. Poor long-term outcomes in other conditions have been related to accelerated telomere shortening and the induction of aberrant so-called "epigenetic" changes. Importantly, inadequate nutrition may cause both adverse effects. Hence, these processes may also play a role in the developmental impairment of critically ill children and the adverse impact of early PN on neurocognitive development.

Telomeres are nucleoprotein complexes at the end of human chromosomes that shorten with each cell cycle. Telomere shortening can be accelerated by environmental and lifestyle factors [64,65], including excessive food consumption and/or unhealthy nutrition [66,67]. It has been demonstrated that critically ill children enter the PICU with significantly shorter leukocyte telomeres than matched healthy children [68]. More importantly, early PN had a telomere-shortening effect as compared with late PN in critically ill children between PICU admission and discharge, independent of baseline risk factors and post-randomization factors. Whether this accelerated telomere shortening contributes to long-term developmental impairment, and particularly the neurocognitive impairment caused by early PN, remains to be investigated.

The involvement of aberrant epigenetic changes in long-term consequences after acute events in life appears plausible. Epigenetics refers to the study of heritable changes in gene expression that do not involve changes in the underlying DNA sequence. Epigenetic changes play an important role in physical and neurocognitive development [69–72]. The most stable epigenetic change is the methylation or de-methylation of DNA. This is the attachment to or removal of a methyl group from a nucleotide, which occurs almost exclusively at the 5' carbon in the cytosine residue of a CpG dinucleotide [69,70]. Alterations in DNA methylation have been implicated in the adverse effects of various environmental stressors, such as inadequate nutrition (both undernutrition and overfeeding), that have been shown to impact long-term health and disease [73]. Particularly during early life, DNA methylation changes may bring about long-term effects [71,72,74]. Data are needed on whether nutritional management in the early phase of critical illness induces aberrant changes in DNA methylation, which could explain the adverse impact of early PN on neurocognitive development.

#### **6. Conclusions and Newest Guidelines**

Nutritional support is considered to be an important aspect of patient care in PICUs [26], with large differences in PICUs worldwide regarding practices of supplemental PN for patients in whom EN as the preferred route is insufficient or impossible [5,10]. A large multicenter RCT showed that withholding supplemental PN throughout the first week in the PICU was clinically superior for short-term outcome as compared with initiating supplemental PN within 24 h after admission [19]. This was the case independent of age [19] or nutritional status [20]. The administered amino acids in particular appeared to explain the adverse impact of early provision of PN, possibly via suppression of required autophagy activation and the shuttling of amino acids to urea production with harmful effects to the liver and kidney. In the long-term, omitting PN during the first week of pediatric critical illness did not show harm, and actually improved executive functioning, behavioral problems and visual-motor integration, as compared with administering PN early. These findings had an important impact

on recent ESPGHAN (European Society for Pediatric Gastroenterology Hepatology and Nutrition) pediatric PN guidelines [60], in which it is advised to consider withholding PN, including amino acids, for one week in critically ill infants, children and adolescents. However, the lack of other RCTs in this specific field makes it challenging to compare these findings with other available recent studies. Further research, in the form of multicenter RCTs, is warranted to determine the optimal composition and ideal timing of initiation of supplemental PN in critically ill children.

**Author Contributions:** A.J., I.V. (Ilse Vanhorebeek) and G.V.d.B. were responsible for the design of the review article. A.J. performed comprehensive literature search. A.J., I.V. (Ines Verlinden) and I.V. (Ilse Vanhorebeek) wrote the article. G.V.d.B. revised it carefully and gave a significant scientific contribution to its content.

**Funding:** This work was supported by ERC Advanced Grants from the Ideas programme of the European Union Seventh Framework Programme (AdvG-2012-321670) and Horizon 2020 Programme (AdvG-2017-785809) to G.V.d.B.; by the Methusalem programme of the Flemish Government (through the University of Leuven to G.V.d.B. and I.V. (Ilse Vanhorebeek), METH14/06); and by the Agency for Innovation through Science and Technology, Flanders, Belgium (through the University of Leuven to G.V.d.B., IWT/070695/TBM.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


the paediatric intensive care unit: A 2-year follow-up of the PEPaNIC international, randomised, controlled trial. *Lancet Respir Med.* **2018**. [CrossRef]


#### *J. Clin. Med.* **2019**, *8*, 830


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Nutritional Laboratory Markers in Malnutrition**

#### **Ulrich Keller**

FMH Endocrinology-Diabetology, Fichtlirain 33, CH-4105 Biel-Benken, Basel, Switzerland; ulrich.keller@unibas.ch

Received: 10 May 2019; Accepted: 29 May 2019; Published: 31 May 2019

**Abstract:** Serum visceral proteins such as albumin and prealbumin have traditionally been used as markers of the nutritional status of patients. Prealbumin is nowadays often preferred over albumin due to its shorter half live, reflecting more rapid changes of the nutritional state. However, recent focus has been on an appropriate nutrition-focused physical examination and on the patient's history for diagnosing malnutrition, and the role of inflammation as a risk factor for malnutrition has been more and more recognized. Inflammatory signals are potent inhibitors of visceral protein synthesis, and the use of these proteins as biomarkers of the nutritional status has been debated since they are strongly influenced by inflammation and less so by protein energy stores. The current consensus is that laboratory markers could be used as a complement to a thorough physical examination. Other markers of the nutritional status such as urinary creatinine or 3-methylhistidine as indicators of muscle protein breakdown have not found widespread use. Serum IGF-1 is less influenced by inflammation and falls during malnutrition. However, its concentration changes are not sufficiently specific to be useful clinically as a marker of malnutrition, and serum IGF-1 has less been used in clinical trials. Nevertheless, biomarkers of malnutrition such as prealbumin may be of interest as easily measurable predictors of the prognosis for surgical outcomes and of mortality in severe illnesses.

**Keywords:** malnutrition; inflammation; nutritional assessment; biomarkers; albumin; prealbumin; IGF-1; elderly; prognostic marker

#### **1. Introduction**

Malnutrition has a substantial clinical and socioeconomic significance; it increases rates of complications in hospitalized patients and healthcare-associated costs. Its prevalence has been estimated in hospitals of Western countries to be 30–50% and in long-term care facilities up to 85% depending on the definition and the type of population studied [1–3].

One of the problems of diagnosing malnutrition is the lack of a unified definition and of standard methods for screening and diagnosis.

Malnutrition results from a mismatch of nutritional requirements with intake. In many malnourished patients, there is an associated disease-related inflammation, resulting in a complex interplay between the two. Inflammation influences both, requirements and intake. It promotes malnutrition and adverse outcomes by provoking anorexia and by altering metabolism with elevation of resting energy expenditure and by increasing muscle catabolism [4].

The appreciation that inflammation plays a major role in the pathophysiology of malnutrition is often lacking, and many clinicians assume that weight loss is the single most important criterion for a malnourished state. This has led to frequent underrecognition of the condition [5].

The purpose of this review is to review the published literature on the role of laboratory biomarkers as a tool to diagnose malnutrition, to assess nutritional risk and to monitor nutritional intervention.

#### **2. Role of Biomarkers in Published Screening Tools to Assess the Risk of Malnutrition**

About half of all published risk scores of malnutrition use serum laboratory markers such as visceral proteins, and others do not (Table 1).

**Table 1.** Anthropometric parameters and biomarkers in various nutritional assessment and screening tools (adapted from [3] with an update, in chronological order of publication.


Results of these screening tests vary considerably, as shown in a Greek study in elderly subjects [2]. These authors found the highest validity coefficient for MUST but a lower specificity for NRS 2000 which has frequently been used.

A consensus committee of the European and US nutritional societies (ESPEN and ASPEN) proposed three sub-definitions of malnutrition. "Starvation-related malnutrition" is present when there is chronic starvation without inflammation; "chronic disease-related malnutrition" is defined as a condition when inflammation is chronic and of mild to moderate degree, and "acute disease or injury-related malnutrition" occurs when inflammation is acute and of severe degree [21]. The criteria for diagnosis of these sub-definitions were energy intake, weight loss, body fat, muscle mass, fluid accumulation and grip strength but no biomarkers [22]. Among ancillary parameters, serum visceral proteins were mentioned, but since they would rather reflect the state of inflammation it was proposed to use them with caution to diagnose malnutrition.

A meta-regression published recently [23] assessed the role of biomarkers in describing the severity of malnutrition according to established and validated nutritional assessment tools. A total of 111 studies (observational and cohort studies; randomized controlled trials were not available) were included, representing 52,911 participants from various clinical settings. The BMI (*p* < 0.001) and concentrations of albumin (*p* < 0.001), hemoglobin (*p* < 0.001), total cholesterol (*p* < 0.001), prealbumin (*p* < 0.001) and total protein (*p* < 0.05) among subjects at high risk of malnutrition assessed by MNA were significantly lower than those with low risk. Similar results were observed for malnutrition identified by SGA and NRS 2002. When patients with acute illnesses were included, the predictive value of albumin and prealbumin was distinctly reduced, confirming the conclusion that they are more markers of inflammation than of malnutrition. The authors concluded that BMI, hemoglobin, and total cholesterol were useful markers of malnutrition in older adults.

#### **3. Serum Visceral Proteins as Biomarkers of the Nutritional Status**

Visceral proteins are mostly synthesized in the liver. Poor protein and energy intake, impaired liver synthetic function as well as inflammatory status result in low circulating levels of visceral proteins. During inflammatory states and increased production of acute-phase proteins the liver reprioritizes protein synthesis, and lowers as a mirror visceral protein synthesis to a degree which correlates with the severity of the injury.

#### *3.1. Serum Albumin*

Albumin is the most abundant protein in human serum. It has been used for decades as an indicator of malnutrition in patients in clinically stable conditions (review and meta-analysis [24]). Serum albumin concentrations decrease with increasing age by approx. by 0.1 g/L per year; however, age itself is not a cause of distinct hypoalbuminemia.

There is a clear relationship between serum albumin concentrations and all-cause mortality in elderly subjects [25]. In patients with a hip fracture, albumin levels below 35 g/L were associated with higher rates of post-operative complications such as sepsis and higher overall mortality. Significant loss of muscle mass has been observed in elderly people with low albumin levels. Inflammatory states and in particular, high concentrations of the cytokines IL-6 and TNF-alpha, were two of the main factors causing low levels of serum albumin [24]. Systemic inflammation not only reduces albumin synthesis but increases its degradation and promotes its transcapillary leakage.

Other studies also found this protein to be a good predictor of surgical outcome [26,27]. Compared to nine other risk variables, serum albumin was the strongest predictor. These findings were confirmed in a later study [26], but whether hypoalbuminemia was due to undernutrition or advanced disease was not clarified in these trials.

When serum albumin as a biomarker for the differential diagnosis of unexplained weight loss (involuntary weight loss of more than 5 kg in the previous 6 months) was included in a study from Spain in 306 referred patients, a little more than one-third were ultimately diagnosed with a malignancy. Multivariate analysis found the strongest predictors of a neoplasm were age >80 years, white blood cell count > 12,000/mm<sup>3</sup> and serum albumin < 3.5 g/dL [28].

Albumin has been criticized as a player in nutritional assessment due to its lack of specificity and long half-life (approximately 20 days) [29]. Serum albumin concentrations not only decrease during decreased synthesis due to inflammatory cytokines as mentioned above or to hepatic insufficiency, they may also decrease following renal losses in nephrotic syndrome and to losses via the GI tract in protein-losing enteropathies [30].

#### *3.2. Serum Prealbumin*

Prealbumin, also named transthyretin, is a transport protein for thyroid hormone and is synthesized by the liver and partly catabolized by the kidneys. Serum prealbumin concentrations less than 10 mg/dL are associated with malnutrition [31].

The use of prealbumin has been advocated as a nutritional marker, particularly during refeeding and in the elderly [32]. The main advantage of prealbumin compared to albumin is its shorter half-life (two to three days) (Table 2), making it a more favorable marker of acute changes of the nutritional state. In addition, prealbumin was not influenced by intestinal protein losses in patients with protein-losing enteropathy [30].


**Table 2.** Characteristics of serum visceral proteins used as nutritional markers.

Table adapted from Spiekerman AM [33].

Prealbumin levels may be increased in the setting of renal dysfunction, corticosteroid therapy or dehydration, whereas they can be decreased during physiological stress, infection, liver dysfunction, and over-hydration [34].

An algorithm that uses prealbumin has recently been proposed as a practical guide to help the clinician to stratify general medical and intensive care patients by risk of complications and outcome [34]. Prealbumin screening should only performed when an acute inflammatory state (CRP > 15 mg/L) was excluded. A prealbumin level of < 0.11 g/L was associated with increased mortality and length of stay, and an increase by less than 0.04 g/L per week indicated failure of nutritional therapy.

An increase in the C-reactive protein/prealbumin ratio in medical intensive care unit patients has been associated with mortality [35], and a low C-reactive protein/prealbumin ratio in surgical patients predicted the successful closure of gastrointestinal fistulas [36]. Routine measurement of prealbumin has been advocated to be a useful nutritional and prognostic indicator in non-ICU patients without inflammation [34].

Several publications reported a role for prealbumin in predicting prognosis (mostly survival) in various clinical conditions such as gastric cancer [37], lung cancer [38] and cardiovascular diseases [39].

#### *3.3. Albumin and Prealbumin in Starved and Otherwise Healthy Malnourished Subjects*

A systematic review assessed the role of albumin and prealbumin in otherwise healthy subjects who were severely nutrient- deprived due to poor access to food or unwillingness to eat, mostly due to anorexia nervosa [40]. The study showed that serum albumin and prealbumin levels were maintained even in the presence of distinct weight loss, and they were lowered only during extreme starvation (BMI < 11 kg/m2). The authors concluded that serum visceral proteins are not predictive of nutritional deprivation and should not be used to guide nutritional therapy in this group of patients.

#### *3.4. Transferrin*

This acute-phase reactant is a transport protein for iron. It has a relatively long half-life (approx. 10 days), and has also been used as a marker of the nutritional status [41] It is influenced by other factors including iron status, liver disease and inflammatory state. Like prealbumin, transferrin levels increase with renal failure [42]. Some authors found transferrin measurements useful for nutritional assessment [43], other did not [44].

During iron-deficiency the levels of transferrin are elevated whereas they are decreased in iron-overload states. Serum transferrin increased in parallel to prealbumin during nutritional intervention in critically ill children [45]. Serum levels decrease in the setting of severe malnutrition, but this marker has been found to be unreliable in the assessment of mild malnutrition and of fat-free mass in a group of elderly Italian patients [46].

#### *3.5. Retinol-Binding Protein (RBP)*

This is a low molecular weight protein with the physiological role to transport retinol from the liver to target organs. It represents the visceral protein with the shortest half-life (approx. 12 h) [33]. According to a review [47] it provides similar responses to energy intake to prealbumin, but it is more difficult to measure than the latter and it is influenced by the vitamin A status. For these reasons there RBP measurements have not found widespread application.

#### **4. Laboratory Markers of Malnutrition Other Than Visceral Proteins**

#### *4.1. Urinary Creatinine*

Creatinine is the end product of creatine which consists of 3 amino acids and is mainly present in muscle. Provided that renal function is intact its excretion reflects creatinine production which in turn is a mirror of skeletal muscle turnover. Each mmol of creatinine in urine is derived from 1.9 kg skeletal muscle [47]. The disadvantages are that it is slowly responding to changes of the nutritional status and that it depends on renal function and requires 24 h urine collections.

#### *4.2. Urinary 3-Methylhistidine*

3-methylhistidine is a component of muscle fibers and is not reutilized by the body. Its urinary excretion reflects the amount of fat-free mass and it can be used as a measure of the rate of muscle protein breakdown. It is less dependent on renal function than creatinine; it is often expressed per mmol of urinary creatinine [47]. Both assay of urinary creatinine and 3-methylhistidine have not found widespread use mainly due to the fact that urinary collections are often cumbersome, their excretion may increase after meat intake and they show a relatively poor sensitivity to monitor changes of body protein stores.

#### *4.3. Serum Cholesterol*

As can be seen in Table 1, some nutritional screening tools used total serum cholesterol as parameter of malnutrition. Serum cholesterol concentrations show a U-shaped relationship with mortality, and low levels have been associated with increased mortality [48]. However, sensitivity and specificity to monitor malnutrition are low.

#### *4.4. Delayed Hypersensitivity and Blood Lymphocyte Count*

The local inflammatory response to a s.c. injection of an antigen is impaired during severe malnutrition. At the same time, maturation of lymphocytes may be reduced in malnourished patients so that total circulating lymphocyte concentration falls to less than 1500/mm3 (reference range 2000–3500) [47].

These abnormalities can be taken as supporting evidence for protein-energy malnutrition, however, they are not specific and insensitive, and concomitant diseases and a severe stress reaction may also have an effect. Both markers respond slowly to correction of the nutritional status. These reasons limit their use as diagnostic tools for malnutrition.

#### *4.5. Serum Insulin-Like Growth Factor 1 (IGF-1)*

IGF-1 (formerly called Somatomedin C) is a ubiquitous growth factor, and the circulating form is mainly produced by the liver. Pituitary growth hormone stimulates its release. Its serum half-life is short (approx. 24 h), and it is largely bound in plasma to binding proteins (mainly IGF BP 3). Fasting lowers plasma IGF-1 levels more than 4 fold and IGF-1 concentrations increases during nutritional repletion. A correlation between energy intake (and less so of protein intake) and plasma IGF-1 concentrations

has been reported [49]. IGF-1 levels were a reliable index of protein-energy undernutrition in elderly patients in the recovery period after surgery for hip fracture; however, according to this trial, this marker was also influenced by inflammation [50]. In contrast, IGF-1 levels were not clearly influenced by inflammation in other groups of surgical patients [51,52]. IGF-1 concentrations are altered by liver disease, renal impairment and severe trauma such as burns [47]. Nevertheless, IGF-1 performed better during nutritional rehabilitation to monitor protein and energy status than albumin or transferrin [53]. Drawback of IGF-1 measurements is the fact that their serum concentrations are influenced by other factors such as the acute-phase response. More recently, there has been interest in free IGF-I which holds promise as a nutritional marker (review in [54]. In spite of the earlier positive reports, IFG-1 measurements have not been advocated in more recent publications.

#### *4.6. Serum Leptin*

Decreased serum leptin concentrations combined with elevated prothrombin time has been reported in malnourished hospitalized patients with end-stage liver disease [55].

#### *4.7. Serum Nesfatin-1*

Nesfatin-1 is an anorexigenic molecule and seems to play a role in appetite regulation and energy homeostasis. Serum nesfatin-1 concentrations have been shown to be increased in chronically malnourished but otherwise healthy children [56].

#### *4.8. Serum Zinc*

Zinc is the most abundant trace element in man beside iron; it is present in all body tissues and fluids and is an essential component of many enzymes. Zinc deficiency has been associated with impaired taste and smell, reduced immunity and increased risk of pneumonia [57]. In cases of severe zinc deficiency, skin lesions, anemia, diarrhea, anorexia, decreased lymphocyte function, impaired visual function and mental retardation may be observed. Several psychological functions were impaired in elderly subjects with zinc deficiency [58].

Zinc deficiency is due to low intake of zinc-containing foods such as meat and to decreased absorption caused by intestinal malabsorption [57]. According to a large sample of the TromsØ study, the risk of zinc deficiency was increased 3 fold in subjects at high risk of malnutrition, particularly in men [59]. Assessment of the zinc status carries the problems that only a small fraction of body zinc is circulating, and most serum zinc is bound to albumin. Therefore, albumin deficiency makes interpretation of serum zinc levels difficult. In spite of the widespread functions of zinc in the body and the potential importance replacing zinc in subjects with zinc deficiency, there is little high-quality evidence of the therapeutic benefit of zinc replacement in adult subjects. A randomized controlled trial in children with protein-energy malnutrition and zinc deficiency showed benefits of zinc replacement [60]. It is likely that zinc deficiency in subjects at risk of malnutrition remains often unrecognized.

#### *4.9. Other Essential Micronutrients (Trace Elements and Vitamins)*

Laboratory assessment of other trace elements such as iron is not specifically mentioned as part of current nutritional screening tools. This does not mean that in cases with clinical suspicion of micronutrient deficiency this should not be performed. The same can be stated for laboratory screening for vitamin deficiencies, in particular, those of vitamins A, B1, B6, B12, D, and folate.

#### **5. Biomarkers of Nutritional Risk in Some Specific Clinical Conditions**

#### *5.1. Geriatric Patients*

**Dementia:** Eating and swallowing problems increase the risk of malnutrition. According to a study from a memory clinic in the Netherlands, about 14% of community-dwelling subjects with newly diagnosed dementia were at risk of malnutrition [61]. These authors pointed out that it is important to detect malnutrition in dementia as early as possible.

In patients with dementia, the nutritional history may only be obtainable by asking family caregivers about appetite and weight change of patients. However, it is unknown whether nutritional information provided by family caregivers are reliable, and therefore biomarkers of malnutrition would be of particular interest.

A recent study compared biochemical blood markers among patients with Alzheimer's disease (AD), dementia with Lewy bodies (DLB), and frontotemporal lobar degeneration (FTLD) [62]. A total of 339 dementia outpatients and their family caregivers participated. Low serum albumin was 7.2 times more prevalent among patients with DLB and 10.1 times more prevalent among those with FTLD than among those with AD, with adjustment for age. The levels of biochemical markers were not significantly correlated with cognitive function. These authors proposed that a multidimensional approach including serological biomarkers such as albumin are needed to assess malnutrition in patients with dementia.

**Sarcopenia in the elderly:** A clinical investigation performed in elderly persons supported the view that prealbumin levels are useful surrogate indicators of lean body mass (LBM). Compared to serum albumin and RBP, prealbumin showed the highest positive correlation with LBM [46]. In order to improve its predictive potency for sarcopenia the reference values for prealbumin should be adapted to the corresponding age and sex [32]. Ingenbleek proposed that routine screening for protein malnutrition using prealbumin should be performed in elderly subjects [63].

#### *5.2. Chronic Kidney Disease*

A position paper from the International Society of Renal Nutrition and Metabolism (ISRNM) stated that serum biomarkers played a particular role in diagnosing malnutrition in patients with kidney failure [64]. Protein-energy wasting can be observed in chronic and in acute kidney disease, and protein-energy wasting is diagnosed according to this publication when low serum biomarkers (albumin, prealbumin, or cholesterol), reduced body mass and reduced muscle mass are present.

Serum prealbumin levels were positively correlated with body cell mass in pre-dialytic kidney patients [65]. Serum biomarkers were part of a new nutritional risk index for predicting mortality in Japanese hemodialysis patients [66]. Cox proportional hazard models indicated that in addition to low BMI, low albumin, low creatinine, and low serum cholesterol predicted independently and significantly mortality within one year.

#### **6. Conclusions**

The role played by serum biomarkers in diagnosing or monitoring malnutrition is controversial, particularly in more recent reports. This is explained by their relatively low specificity and by the fact that underlying diseases such as inflammation exert a major influence, particularly on serum visceral proteins. In addition, the role of biomarkers to guide nutritional therapy has not been studied in large randomized controlled trials. A recent randomized controlled multicenter trial in hospitalized patients with malnutrition might fill this gap in the near future [67]. Nevertheless, biomarkers such as prealbumin are valid prognostic indicators of disease outcome and of mortality in patients at risk of malnutrition.

**Acknowledgments:** The page charges for this article were covered by the Swiss Society for Clinical Nutrition. **Conflicts of Interest:** The author declares no conflict of interest.

#### **References**

1. Kondrup, J.; Rasmussen, H.H.; Hamberg, O.; Stanga, Z. Nutritional risk screening (NRS 2002): A new method based on an analysis of controlled clinical trials. *Clin. Nutr.* **2003**, *22*, 321–336. [CrossRef]


© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Journal of Clinical Medicine* Editorial Office E-mail: jcm@mdpi.com www.mdpi.com/journal/jcm

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18