Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein

Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein

Clinical Nutrition ESPEN xxx (2016) 1e5 Contents lists available at ScienceDirect Clinical Nutrition ESPEN journal homepage: http://www.clinicalnutr...

612KB Sizes 0 Downloads 20 Views

Clinical Nutrition ESPEN xxx (2016) 1e5

Contents lists available at ScienceDirect

Clinical Nutrition ESPEN journal homepage: http://www.clinicalnutritionespen.com

Original article

Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein Mara Barbosa*, Joana Magalhaes, Carla Marinho, Jose Cotter ~es e Portugal, Rua dos Cutileiros, 4835 044 Guimara ~es, Portugal Gastroenterology Department, Centro Hospitalar do Alto Ave e Guimara

a r t i c l e i n f o

s u m m a r y

Article history: Received 15 January 2016 Accepted 23 April 2016

Introduction and aims: Percutaneous endoscopic gastrostomy (PEG) is considered one of the preferred routes for long-term enteral feeding. However, early mortality after PEG placement is high. We aimed at analyzing overall and early mortality in patients who underwent PEG insertion and at identifying risk factors of increased mortality after the procedure. Methods: Retrospective study which included patients who had a PEG insertion at our department between May 2007 and January 2013. Variables analyzed: demographic, Charlson's co-morbidity index, past aspiration pneumonia, indication for PEG, hemogram, ionogram, urea, creatinine, albumin and Creactive protein (CRP). Outcome: death. KaplaneMeier survival analysis was used to calculate mortality after PEG placement. Predictive factors of overall mortality were identified by univariate and multivariate analysis and of 30-day, 90-day and 180-day mortality by logistic regression. The AUROC analysis for CRP levels was performed. Results: Inclusion of 135 patients: 51.9% female, mean age of 73 ± 17 years, 90.4% with neurological dysphagia and 9.6% with tumors compromising oral intake. The median survival time was 272 days. The 30-day, 90-day and 180-day mortality was 0.14 ± 0.06 (95% CI 0.08e0.20), 0.29 ± 0.08 (95% CI 0.21e0.37) and 0.43 ± 0.08 (95% CI 0.35e0.51), respectively. Patients with higher levels CPR (hazard ratio (HR) 1.009 95% CI 1.002e1.160, p ¼ 0.012) and higher levels of urea (HR 1.009 95% CI 1.002e1.160, p ¼ 0.012) had worse outcome and those with higher sodium levels (HR 0.945 95% CI 0.908e0.983, p ¼ 0.005) had better prognosis. Higher CRP levels was the only independent predictive factor for 30-day mortality (odds ratio (OR) 1.008 95% CI 1.001e1.014, p ¼ 0.029), and was also a risk factor for 90-day and 180-day mortality (OR 1.013 95% CI 1.005e1.021, p ¼ 0.002 and OR 1.009 95% CI 1.001e1.018, p ¼ 0.026, respectively). CRP levels 35.9 mg/dL could predict death at 30 days with a sensitivity of 0.810 and a specificity of 0.614. Conclusions: The early mortality after PEG placement is high. CRP, an indicator of acute illness, is a useful parameter at identifying patients with increased probability of dying after PEG insertion. High CRP levels should be considered in the decision making process. © 2016 European Society for Clinical Nutrition and Metabolism. Published by Elsevier Ltd. All rights reserved.

Keywords: Percutaneous endoscopic gastrostomy Early mortality C-reactive protein

1. Introduction Percutaneous endoscopic gastrostomy (PEG) was first reported by Gauderer et al. in 1980 in pediatric patients with dysphagia but has rapidly become an established procedure in adults [1]. Nowadays, it is considered one of the preferred routes for long-term enteral

* Corresponding author. Tel.: þ351 93 311 26 32. E-mail addresses: [email protected] (M. Barbosa), [email protected] (J. Magalhaes), [email protected] (C. Marinho), [email protected] (J. Cotter).

feeding in patients who are unwilling or unable to maintain oral intake for a period exceeding 4 weeks [2e6]. The most common conditions responsible for impaired swallowing are cerebrovascular disease, oropharyngeal malignancies and esophageal cancer. Although it is a safe and simple procedure [7], PEG placement is associated with a high early mortality [8,9]. Therefore, it is of utmost importance to find criteria that better select the patients who benefit most from PEG insertion. Some studies have focused on this issue, but published data is still scarce. The aims of this study are: i) to analyze the 30-day, 90-day, 180day and overall mortality in patients who underwent PEG

http://dx.doi.org/10.1016/j.clnesp.2016.04.029 2405-4577/© 2016 European Society for Clinical Nutrition and Metabolism. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Barbosa M, et al., Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein, Clinical Nutrition ESPEN (2016), http://dx.doi.org/10.1016/j.clnesp.2016.04.029

2

M. Barbosa et al. / Clinical Nutrition ESPEN xxx (2016) 1e5

placement and ii) to identify clinical and analytical risk factors of increased mortality after PEG insertion. 2. Patients and methods 2.1. Study population A retrospective study was conducted. Patients who underwent PEG placement between May 2007 and January 2013 were included. Data regarding gender, age, Charlson's co-morbidity index to assess baseline illness severity, previous occurrence of aspiration pneumonia and indication for PEG placement was collected from medical records. Analytic variables such as hemogram, ionogram, urea, creatinine, albumin and C-reactive protein (CRP) were also analyzed. Exclusion criteria: absence of follow-up and PEG re-insertions.

compared by means of the log-rank test (discrete variables). Univariate cox proportional hazard regression was used to investigate the effect of continuous variables on survival. Then, a multivariate cox proportional hazard regression was performed to adjust for confounding factors. In order to identify independent risk factors for mortality at each time-frame (30-day, 90-day and 180-day period), a logistic regression analysis was applied. The receiver operating characteristic (ROC) curve analysis for the variable CRP was performed for each time-frame and also for overall follow-up time. Two-tailed p values of less than 0.05 were considered statistically significant.

3. Results

2.2. Outcome

3.1. Study population

Death of the patient was recorded. The overall and 30-, 90- and 180-day mortality after PEG insertion was calculated.

A total of 190 patients were evaluated with 135 being included (the main exclusion criteria was absence of follow-up): 70 (51.9%) female, mean age of 73 ± 17 years (range 17e97); an average Charlson's index score of 4.7 ± 2.0; 98 patients (72.6%) with previous aspiration pneumonia. The baseline characteristics of patients who underwent PEG placement are listed in Table 1. The indications for PEG placement were neurological dysphagia in 122 (90.4%) patients and tumors compromising oral feeding in 13 patients (9.6%). In those with neurological dysphagia, the underlying diseases were: cerebrovascular disease (n ¼ 68, 55.7%), dementia (n ¼ 24, 19.7%), demyelinating disease (n ¼ 8, 6.6%), anoxic brain injury (n ¼ 8, 6.6%), traumatic brain injury (n ¼ 7, 5.7%) and Parkinson's disease (n ¼ 7, 5.7%). In those regarding tumors compromising oral feeding, the majority of patients had esophageal tumors (n ¼ 7, 53.8%), followed by head and neck tumors (n ¼ 5, 38.5%), with only one patient having a pulmonary tumor (7.7%). No deaths could be directly related to PEG insertion and 96% of them were due to underlying co-morbidities. The mean follow-up period was 321 ± 347 days (range 1e1448). During follow-up, 84 patients (62.2%) died. Two patients (1.5%) had their PEG tube removed due to resumption of oral intake. The baseline characteristics of patients who underwent PEG insertion stratified by survival are listed in Table 1. The patients who died during follow-up had PEG inserted more frequently because of a dysphagia due to a tumor (p < 0.001), had aspiration pneumonia more commonly (p < 0.001) and had lower levels of hemoglobin (p < 0.001), higher values of leucocytes (p ¼ 0.005), lower sodium levels (p ¼ 0.010) and higher CRP values (p ¼ 0.008).

2.3. Procedure The referral for PEG placement was made by the patient's physician (gastroenterologist, neurologist, otorhinolaryngologist, internal medicine or primary care physician). A critical review of the indication was done in every patient. All procedures were performed by the same endoscopic team, under sedation with propofol administered by an anesthesiologist. The “pull” technique described by Gauderer et al. [1] was used to place PEG tubes (COOK®, U.S. Endoscopy®). All patients received antibiotic prophylaxis with cefotaxime (1 g 30 min before and 8 h and 16 h after the procedure) and were started on a proton pump inhibitor. Feeding through the tube was initiated 24 h after PEG placement if no complications occurred. Patients were evaluated by the nutrition team and detailed recommendations were given. Outpatient follow-up was done by a multidisciplinary team (endoscopic team, nutritional team and trained nurse) at one month after the procedure and then at 3-months intervals or less, as needed. 2.4. Statistical analysis SPSS 17.0 (SPSS Inc., Chicago, Illinois, USA) was used for data analysis. Continuous variables were expressed as mean ± standard deviation. Univariate analysis was performed using Student's t-test. Survival curves were plotted as KaplaneMeier estimates and

Table 1 Baseline characteristics of patients who underwent PEG placement (all patients and stratified by survival).

Gender, female/male (n) Age (mean ± SD), years Indication, neurological disease vs tumor (n) Past aspiration pneumonia (n) Charlson's index score (mean ± SD) Hemoglobin (mean ± SD), g/dL Leucocytes (mean ± SD) Platelets (mean ± SD) Urea (mean ± SD), mg/dL Creatinine (mean ± SD), mg/dL Sodium (mean ± SD), mEq/L Potassium (mean ± SD), mEq/L Albumin (mean ± SD), g/dL C-reactive protein (mean ± SD), mg/dL

All patients

Outcome

p value

Dead (n ¼ 84)

Alive ( ¼ 51)

70/65 73.0 ± 17.0 122/13 98 4.7 ± 2.0 11.3 ± 1.6 9.7 ± 3.8 306.000 ± 126.000 35.8 ± 25.5 0.7 ± 0.5 135.1 ± 6.0 4.4 ± 0.6 2.9 ± 0.8 52.3 ± 62.2

41/43 75.0 ± 16.0 73/11 68 4.8 ± 2.0 10.9 ± 1.6 10.3 ± 4.3 305.000 ± 122.000 38.8 ± 27.4 0.7 ± 0.3 134.1 ± 6.5 4.4 ± 0.6 2.9 ± 0.8 71.3 ± 17.1

29/22 70.0 ± 18.0 29/2 30 4.4 ± 1.9 11.9 ± 1.4 8.6 ± 2.7 308.000 ± 134.000 30.9 ± 21.5 0.8 ± 0.6 136.7 ± 4.8 4.5 ± 0.6 2.8 ± 0.8 36.4 ± 38.9

0.235 0.071 <0.001 <0.001 0.531 <0.001 0.005 0.885 0.084 0.513 0.010 0.636 0.389 0.008

SD standard deviation.

Please cite this article in press as: Barbosa M, et al., Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein, Clinical Nutrition ESPEN (2016), http://dx.doi.org/10.1016/j.clnesp.2016.04.029

Fig. 1. KaplaneMeier survival curves after percutaneous endoscopic gastrostomy (PEG) placement.

3.2. Survival analysis The KaplaneMeier survival curve of the 135 patients is shown in Fig. 1. The median survival time was 272.0 days. The 30-day, 90-day and 180-day calculated mortality after PEG placement was 0.14 ± 0.06 (95% CI 0.08e0.20), 0.29 ± 0.08 (95% CI 0.21e0.37) and 0.43 ± 0.08 (95% CI 0.35e0.51), respectively. The median survival time in the group of patients who died was 98.5 days. The survival of patients who had neurological dysphagia was significantly better than those who had a PEG inserted due to malignant disease (p < 0.0001, log-rank test). Gender did not influence survival (p ¼ 0.233, log-rank test). Although the previous occurrence of aspiration pneumonia did not significantly affect the overall survival of patients who underwent PEG placement (p ¼ 0.152, log-rank), the visual analysis of the KaplaneMeier curves (close together at the beginning, separating afterward) suggests that, in our population, the past occurrence of aspiration pneumonia has no impact in the very early mortality but could be relevant when considering longer follow-up times. The KaplaneMeier survival curves according to gender, indication for PEG placement and past occurrence of aspiration pneumonia are shown in Fig. 2. In univariate cox proportional hazard regression analysis, patients with lower hemoglobin levels, higher leucocytes levels, lower sodium levels and higher CRP value were significantly associated with mortality (Table 2). The cox proportional hazard regression models were fitted for multivariate analysis, also including variables statistically significant in the univariate analysis. Only higher CRP levels, lower sodium levels and higher urea levels were independently associated with mortality (Table 2).

3.3. Logistic regression (30-day, 90-day and 180-day period) To perform logistic regression analysis for the 30-day, 90-day and 180-day periods, in order to identify risk factors for mortality

Fig. 2. KaplaneMeier survival curves according to gender, indication for percutaneous endoscopic gastrostomy (PEG) and past aspiration pneumonia after PEG placement.

Please cite this article in press as: Barbosa M, et al., Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein, Clinical Nutrition ESPEN (2016), http://dx.doi.org/10.1016/j.clnesp.2016.04.029

4

M. Barbosa et al. / Clinical Nutrition ESPEN xxx (2016) 1e5

Table 2 Cox proportional hazard regression models (univariate and multivariate analysis). Univariate analysis

Female Age, years Neurological disease Past aspiration pneumonia Charlson's index score Hemoglobin, g/dL Leucocytes Platelets Urea, mg/dL Creatinine, mg/dL Sodium, mEq/L Potassium, mEq/L Albumin, g/dL C-reactive protein, mg/dL

Multivariate analysis

Crude HR

95% CI

p value

0.771 1012 0.326 1486 1044 0.795 1076 0.999 1010 1037 0.960 1001 0.936 1007

0.502e1.185 0.997e1.027 0.169e0.629 0.861e2.566 0.931e1.17 0.68e0.928 1.023e1.131 0.998e1.001 1.003e1.017 0.672e1.599 0.924e0.997 0.706e1.42 0.707e1.239 1.004e1.010

0.235 0.110 0.001 0.155 0.464 0.004 0.004 0.575 0.003 0.870 0.032 0.994 0.643 <0.001

Adjusted HR

95% CI

1009

1.002e1.16

0.945

0.908e0.983

1007

1.004e1.001

HR hazard ratio. CI confidence interval. Bold values indicate that variable is statistically significant (p < 0.050).

at each time-frame, only patients whose dead/alive status was known were considered: 135, 130 and 127 patients, respectively. For the 30-day period, only higher CRP levels were independently associated with mortality; for the 90-day period, neurological dysphagia predicted a better outcome and patients with higher CRP levels and lower platelets levels had a worse prognosis; for the 180-day period, male gender, higher CRP and urea levels were independently associated with death (Table 3).

Table 4 The sensitivity and specificity of C-reactive protein for predicting mortality. Time

Sensitivity

Specificity

30 days 90 days 180 days overall

0.810 0.667 0.625 0.548

0.614 0.637 0.662 0.706

C-reactive protein level 35.9 mg/dL.

3.4. ROC curve analysis for CRP The value of CRP that maximizes simultaneously sensitivity and specificity for 30-day, 90-day, 180-day and overall follow-up time is 35.9 mg/dL. The sensitivity and specificity of CRP for predicting mortality at each time-frame and overall follow-up time are shown in Table 4. CRP is most sensitive at predicting 30-day mortality, though with a little less specificity comparing with longer followup times. The ROC curve analysis for the variable CRP (30 days) is shown in Fig. 3. 4. Discussion Our study clearly demonstrates that mortality after PEG insertion is high, with 14% of patients being dead within 30 days of the procedure and the majority of them not surviving more than a year. This result is in accordance with data published in the literature, which report 30-day mortality rates ranging from 3.3 to 32.8% [10e12]. As short-term mortality is high, emphasis should to be placed on finding risk factors for mortality. In our study, several clinical and laboratory baseline characteristics predicted mortality after PEG placement.

Fig. 3. The receiver operating characteristic curve analysis for the C-reactive protein (30 days after percutaneous endoscopic placement).

Table 3 Risk factors for 30-day, 90-day and 180-day mortality (logistic regression analysis). 30 days (n ¼ 135) Or Female Neurological disease Platelets Urea, mg/dL C-reactive protein, mg/dL

1008

95% CI

1.001e1.014

90 days (n ¼ 130) p value

0.029

Or

95% CI

180 days (n ¼ 127) p value

0.221 0.994

0.052e0.935 0.990e0.998

0.040 0.003

1013

1.005e1.021

0.002

Or

95% CI

p value

0.382

0.175e0.834

0.016

1022 1009

1.005e1.0.39 1.001e1.018

0.010 0.026

OR odds ratio. CI confidence interval.

Please cite this article in press as: Barbosa M, et al., Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein, Clinical Nutrition ESPEN (2016), http://dx.doi.org/10.1016/j.clnesp.2016.04.029

M. Barbosa et al. / Clinical Nutrition ESPEN xxx (2016) 1e5

CRP was a risk factor for predicting overall and early mortality. Using the value of 35.9 mg/dL as cut-off, it showed high sensitivity at identifying patients with worse prognosis, mainly in the very early period. Furthermore, higher CRP levels were the only independent risk factor for 30-day mortality. The importance of CRP in this setting has been corroborated by other authors [8,13,14]. Other laboratory parameters such as sodium and urea had impact on the overall mortality and platelets and urea were associated with a decreased probability of surviving 90 days and 180 days after the procedure, respectively. All of these analytical risk factors are wellestablished indicators of acute illness. Therefore, PEG tubes should be inserted with a delay from acute diseases [15,16]. Regarding gender, being male was associated with higher mortality, which is also reported in other studies [13,17e19]. Age is a recognized predictor of death after PEG placement [9,13,14,18,20]. The mean age of patients who died during follow-up was higher than those who survived; although it did not reach statistical significance, a tendency was noted. Though the prognosis of patients who have a PEG inserted due to a malignant disease has been reported as being better than patients with neurological disease [21], in our study patients with tumors had a worse outcome, as they were in the terminal stage of their disease, with prognosis being dictated by that fact. Most likely, the advanced tumoral disease at the time of PEG placement influenced survival and patients with head and neck cancer with a short life expectancy should not be offered PEG insertion as it is associated with an exceedingly high mortality, which in accordance with data reported in the literature [15]. The history of pneumonia aspiration has been described as a risk factor for mortality after PEG placement [22]. In our study, though not relevant when considering early mortality, aspiration pneumonia seems to have impact in late mortality. Therefore, it should not be overlooked in the evaluation before PEG insertion. Defining high-risk subgroups of patients based on clinical and analytical parameters among PEG candidates is mandatory. Physicians, patients and their families (as appropriate) should be aware of the high mortality associated with PEG placement when risk factors are present. This would help them through the process of decision making and allow for an individualized approach: for instance, a “cooling-off period” could be adopted when the patient is at higher risk of mortality due to an acute illness; if the patient survives after that period he might be considered for PEG placement [15,16]. Our study highlights the importance of CRP in that context: high CRP levels should alert care providers and patients or patients' family to consider postponing PEG insertion or not placing it at all. Moreover, the decision of not inserting a PEG tube in some patients (negative selection) would protect them from invasive procedures that would expose them to increased risk for early mortality without a clear benefit. In these patients, nutritional support would probably be best managed with less invasive options. The decision making process of PEG placement is complex and poses several medical and ethical concerns. Apart from the impact on outcome that we have aforementioned, other factors should be taken into account when PEG is considered an option for nutritional support, such as: baseline disease and probability of recovery; type

5

and severity of co-morbidities; the patient or family's wishes, social and institutional issues [8]. Indeed, the assessment for PEG insertion should always be holistic and individualized to the patient and family we are providing care for, rather than focused on a specific item. Only when all these factors are appreciated, can we anticipate an actual benefit for the patient. References [1] Gauderer MW, Ponsky JL, Izant Jr RJ. Gastrostomy without laparotomy: a percutaneous endoscopic technique. J Pediatr Surg 1980;15:872e5. [2] Eisen GM, Baron TH, Dominitz JA, et al. Role of endoscopy in enteral feeding. Gastrointest Endosc 2002;55:794e7. [3] Loser C, Aschl G, Hebuterne, et al. ESPEN guidelines on artificial enteral nutrition: percutaneous endoscopic gastrostomy (PEG). Clin Nutr 2005;24: 848e61. [4] Angus F, Burakoff R. The percutaneous endoscopic gastrostomy tube: medical and ethical issues in placement. Am J Gastroenterol 2003;98:272e7. [5] DeLegge MH. Percutaneous endoscopic gastrostomy. Am J Gastroenterol 2007;102:2620e3. [6] Kirby DF, DeLegge MH, Fleming CR. American Gastroenterological Association technical review on tube feeding for enteral nutrition. Gastroenterology 1995;108:1282e301. [7] Verhoef MJ, Van Rosendaal GM. Patient outcomes related to percutaneous endoscopic gastrostomy placement. J Clin Gastroenterol 2011;32:49e53. [8] Figueiredo FA, da Costa MC, Pelosi AD, Martins RN, Machado L, Francioni E. Predicting outcomes and complications of percutaneous endoscopic gastrostomy. Endoscopy 2007;39:333e8. [9] Zopf Y, Maiss J, Konturek P, Rabe C, Hahn EG, Schwab D. Predictive factors of mortality after PEG insertion: guidance for clinical practice. J Parenter Enter Nutr 2011;35:50e5. [10] Takayama T, Takayama K, Inoue N, Funakoshi S, Serizawa H, Watanabe N, et al. Prediction of survival and complications after percutaneous endoscopic gastrostomy in an individual by using clinical factors with an artificial neural network system. Eur J Gastroenterol Hepatol 2009;21:1279e85. [11] Paramsothy S, Papadopoulos G, Mollison LC, Leong. Resumption of oral intake following percutaneous endoscopic gastrostomy. J Gastroenterol Hepatol 2009;24:1098e101. [12] Janes SE, Price CS, Khan S. Percutaneous endoscopic gastrostomy: 30-day mortality trends and risk factors. J Postgrad Med 2005;51:23e9. [13] Suzuki Y, Tamez S, Murakami A, Akihiko T, Mizuhara A, Horiuchi A, et al. Survival of geriatric patients after percutaneous endoscopic gastrostomy in Japan. World J Gastroenterol 2010;16:5084e91. [14] Blomberg J, Lagergren P, Martin L, Mattsson F, Lagergren J. Albumin and Creactive protein levels predict short-term mortality after percutaneous endoscopic gastrostomy in a prospective cohort study. Gastrointest Endosc 2011;73:29e36. [15] Niv Y, Abukasis G. Indications for percutaneous endoscopic gastrostomy insertion: ethical aspects. Dig Dis 2002;20:253e6. [16] Lang A, Bardan E, Chowers Y, Sakhnini Fidder HH, Bar-Meir S, Avidan B. Risk factors for mortality in patients undergoing percutaneous endoscopic gastrostomy. Endoscopy 2004;36:522e6. [17] Higaki F, Yokota O, Ohishi M. Factors predictive of survival after percutaneous endoscopic gastrostomy in the elderly: is dementia really a factor? Am J Gastroenterol 2008;103:1011e6. [18] Rimon E, Kagansky N, Levy S. Percutaneous endoscopic gastrostomy; evidence of different prognosis in various patient subgroups. Age Ageing 2005;34: 353e7. [19] Skelly RH, Kupfer RM, Metcalfe ME, Allison SP, Holt M, Hull MA, et al. Percutaneous endoscopic gastrostomy (PEG): change in practice since 1988. Clin Nutr 2002;21:389e94. [20] Light V, Slezak F, Porter J, Gerson L, McCord G. Predictive factors for early mortality after percutaneous endoscopic gastrostomy. Gastrointest Endosc 1995;42:330e5. [21] Blomberg J, Lagergren J, Martin L, Mattson F, Lagergren P. Complications after percutaneous endoscopic gastrostomy in a prospective study. Scand J Gastroenterol 2012;47:737e42. [22] Light VL, Slezak FA, Porter JA, Gerson LW, McCord G. Aspiration and PEG Gastrointest Endosc 1995;42:330e5.

Please cite this article in press as: Barbosa M, et al., Predictive factors of early mortality after percutaneous endoscopic gastrostomy placement: The importance of C-reactive protein, Clinical Nutrition ESPEN (2016), http://dx.doi.org/10.1016/j.clnesp.2016.04.029