The association between prealbumin, all-cause mortality and response to nutritional treatment in patients at nutritional risk. Secondary analysis of a randomized-controlled trial

Introduction Due to the shorter half-life as compared with albumin, serum prealbumin concentrations have been proposed to be useful nutritional biomarkers for the assessment of patients at nutritional risk. In a post-hoc analysis of patients at nutritional risk from a randomized-controlled nutritional trial, we therefore tested the hypothesis that (a) prealbumin is associated with higher all-cause 180-day mortality rates and that (b) individualized nutritional support compared to usual care nutrition more effectively improves survival at 30 days in patients with low prealbumin levels compared to patients with normal prealbumin levels. Methods We performed a pre-specified cohort study in patients included in the pragmatic, Swiss, multicenter, randomized-controlled EFFORT trial comparing the effects of individualized nutritional support with usual care. We studied low prealbumin concentrations (<0.17 g/l) in a subgroup of 517 patients from one participating centre.


Introduction
2][3][4] There is increasing evidence demonstrating that risks of malnutrition can be lowered, at least partly, by addressing malnutrition through active screening followed by a specific nutritional intervention to achieve improvement of clinical outcomes. 3,5,66][7][8] The largest trial in the medical inpatient setting was the

Effect of Early Nutritional support on Frailty and Functional Outcomes, and Recovery of
Malnourished Medical Inpatients Trial (EFFORT), a pragmatic Swiss multicenter trial that included 2028 medical inpatients at nutritional risk and that compared an individualized nutritional support strategy with usual nutritional care. 5While this study and other trials found significant reductions in the risk for severe complications and mortality in the overall population of medical inpatients, secondary analyses suggested that there are patients that show more or less benefit from the intervention opening the door for a more personalized approach. 3,9[12][13] Other markers such as albumin 14 or different metabolomic parameters were not helpful. 15,16rther knowledge of specific nutritional biomarkers to predict response to nutritional treatment may allow better phenotyping patients and help focus specific interventions on patients who may benefit the most.Among potential nutritional biomarkers, there is particular interest in visceral proteins such as serum albumin that has been used to characterise patients since the 1970's. 17In fact, baseline albumin levels were associated with higher risk for mortality among patients in the EFFORT trial but albumin was not helpful in stratifying patients regarding response to nutritional treatment confirming results of other similar studies. 14However, one major drawback of albumin as a nutritional marker is its long half-life of about 20 days and that its concentration is affected by inflammation and fluid balance.9][20] However, there is currently insufficient data from larger trials investigating the possible benefits of measurement of prealbumin among patients at nutritional risk.Answering this question may help to use prealbumin as a nutritional biomarker in the future to identify patients at high risk of adverse outcome that should receive nutritional care.Herein, in a post-hoc analysis of patients at nutritional risk from a randomized-controlled nutritional trial 5 , we therefore tested the hypothesis that (a) prealbumin is associated with higher all-cause 180-day mortality rates and that (b) individualized nutritional support compared to usual care nutrition more effectively improves survival at 30 days in patients with low prealbumin levels compared to patients with normal prealbumin levels.

Study design and setting
We performed a pre-specified retrospective cohort study in patients included in the pragmatic, Swiss, multicenter, randomized-controlled EFFORT trial comparing the effects of individualized nutritional support with usual care.We studied low prealbumin concentrations (<0.17 g/l) in a subgroup of 517 patients from one participating centre.EFFORT was a prospective, randomized-controlled trial studying the effect of early individual nutritional support compared with usual care nutrition in patients at nutritional risk in eight Swiss hospitals from April 2014 to February 2018. 5The trial was approved by the ethics committee of Northwestern Switzerland (EKNZ; 2014_001) and registered at ClinicalTrials.gov in August 2015.(https://clinicaltrials.gov/ct2/show/NCT02517476)The trial protocol, eligibility features and main results were previously published. 5,21

Patient population
The initial EFFORT trial population included 2028 patients at nutritional risk (defined by a Nutritional Risk Screening (NRS 2002) total score ≥3 points). 22,23The NRS includes the patient's current nutritional status and the severity of the underlying disease. 6Each part scores between 0 to 3 points, plus 1 point for age above 70 years (max.7 points).For this secondary analysis, we only focused on 517 patients (25.5%) from one participating centre (Kantonsspital Aarau), where prealbumin levels were measured routinely in a blinded fashion as a preplanned substudy, without communication of results to physicians.
All adult patients were eligible for the trial when there was an expected length of hospital stay ≥5 days and if they provided informed consent.Exclusion criteria were initial admission to an intensive care unit or a surgical unit, inability for oral ingestion of food, already established nutritional support on admission, terminal illness, prior gastric bypass surgery, anorexia nervosa, acute pancreatitis, acute liver failure, cystic fibrosis, stem cell transplantation or contraindications for nutritional support and previous inclusion in the trial.Patients were randomly assigned (1:1) either to the intervention group (individualized nutritional support), or the control group (standard hospital food).The intervention group received individualized nutritional support within 48 hours of admission to reach protein and energy goals according to a previously published consensus protocol and in accordance with recent international guidelines. 24,25Individualized energy and protein goals were defined for each patient upon hospital admission by a trained registered dietician using weight-adjusted Harris-Benedict equation to estimate energy requirements. 26Daily protein intake goals were set at 1.2-1.5 g/kg body weight per day with lower targets of 0.8 g/kg body weight for patients with renal failure. 27 reach these goals, an individualized nutritional plan was developed based initially on oral nutrition provided by the hospital kitchen and oral nutritional supplements. 28,29A further increase in nutritional support to enteral tube feeding or parenteral feeding was recommended if at least 75% of energy and protein targets could not be reached through oral feeding within 5 days.Patients in the control group received usual care nutrition according to their ability and desire to eat, with no nutritional consultation and no recommendation for additional nutritional support.
We prospectively collected different medical and nutritional information in patients including medical diagnosis according to International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes, sociodemographic and anthropometric data, baseline muscle strength, and functional status (using the Barthel-Index). 30

Patient groups and outcomes
To understand the prognostic potential of prealbumin, we stratified the patient population into two groups based on their serum prealbumin levels at admission with a cut-off of 0.17 g/L as recommended. 31We additionally, stratified patients based on their baseline inflammation status using a CRP cut-off level of 100 mg/L as suggested previously. 10,32e primary endpoint for this analysis was all-cause mortality within 180 days for the prognostic analyses based on data from a long-term follow-up analysis. 33For the analysis regarding response to nutritional treatment we used short-term all-cause mortality within 30 days similar to the initial trial. 5To assess response to nutritional treatment we compared whether the difference in mortality in patients in the intervention group receiving individualized nutritional support and control patients would be different according to low or normal prealbumin levels at the 0.17 g/L cut-off.Other secondary endpoints included the composite endpoint of adverse outcomes (consisting of all-cause mortality, admission to the intensive care, readmission, major complications, functional decline), length of hospital stay (LOS) and non-elective hospital readmission after 30 and 180 days.All endpoints are in line with the original publication and collected through a structured telephone interview at 30 and 180 days after inclusion in the trial. 5

Statistical analyses
All analyses were performed in the intention-to-treat population, including all patients with available serum prealbumin concentrations.Continuous variables were expressed as mean and standard deviations (SD) for normally distributed data or as median and interquartile range (IQR) for skewed data, discrete variables as counts and percentages.We compared frequencies using Pearson's χ² test and continuous variables using a two-sample t-test or a Wilcoxon rank-sum test.We used Cox regression models for time-to-event data reporting hazard ratios (HR), logistic regression for binary outcomes reporting odds ratios (OR) and linear regression for continuous outcomes reporting coefficients (Coef) with corresponding 95% confidence intervals (CI).We performed models for prognostic and predictive analysis adjusted for different predefined confounders (age, sex, main diagnosis, comorbidities, randomization as well as CRP and NRS).To investigate possible subgroup effects (effect modification) with regard to admission prealbumin concentrations, we included interaction terms in the statistical models.We also used the Kaplan-Meier method to visualise outcome data over time by calculating the probability of all-cause mortality within 30 days of randomization.
All statistical analyses were performed with STATA 15.1 (Stata Corp, College Station, TX, USA).A P value <0.05 (for a 2-tailed test) was considered to indicate statistical significance.
All patients were at nutritional risk, with 140 (27.1%),191 (36.9%) and 186 (36.0%) having NRS scores of 3, 4 or 5 and more points.Overall, patients had a high burden of comorbidities and of 511 patients with available CRP measurements, 159 (31.12%) patients had high inflammation with CRP levels ≥100 mg/L.Table 1 shows additional baseline characteristics stratified by prealbumin level.Additional baseline tables stratified by CRP concentrations and randomization can be found in the Supplementary document.

Association of admission prealbumin concentrations and clinical outcomes
In a first step, we investigated the prognostic value by calculating the associations of prealbumin with 30-and 180-day outcomes (Table 2).In terms of long-term mortality after 180 days, patients with low prealbumin levels had an almost doubling in mortality resulting in an unadjusted HR of 1.88 (95%CI 1.34 to 1.2.64;P<0.001).These results remained robust in an adjusted model including CRP (HR of 1.52 (95%CI 1.04 to 2.22; P=0.03)) and in a model also including NRS (HR 1.59 (95%CI 1.11 to 2.28; P=0.011).To further visualize 180-day mortality among different patient subgroups, we calculated Kaplan-Meier survival curves.Figure 2 shows survival curves stratified by the two prealbumin groups (Figure 2A), as well as groups additionally stratified by nutritional risk (i.e., NRS 3-4 points vs. ≥5 points) and inflammation (CRP <100 mg/L vs. ≥ 100 mg/L) (Figure 2B-E).All results were significant except for the subgroup of patients in the high inflammation group (p=0.066).
Regarding other adverse outcomes within 30 days, we found significant associations of low baseline prealbumin concentrations with the composite endpoint adverse outcome and length of hospital stay in unadjusted and the adjusted models.For short-term mortality, associations were again significant in the unadjusted models but not in fully adjusted models.
We also compared the prognostic value of prealbumin with established outcome parameters including albumin and NRS with regard to the area under the curve (Supplementary Table 3).With an AUC of 0.62, prealbumin showed the best discrimination regarding 180-day mortality and significantly improved the NRS score from AUC 0.60 to 0.66 (p<0.001).

Association of prealbumin concentrations and effectiveness of nutritional support
To understand whether treatment response to nutritional treatment would differ according to admission prealbumin levels, we compared effects of the initial randomization (intervention vs. control) in subgroups of patients according to prealbumin (Table 3).These analyses are displayed in a Kaplan-Meier survival function (Figure 3) and a forest plot (Figure 4).Patients with normal vs. low prealbumin levels had a similar response to nutritional treatment with regard to 30-day mortality (OR for nutritional support 0.88 vs. 0.90, P interaction=0.823).The same effect was observed within the different CRP subgroups with no evidence for a subgroup effect.Similarly, no differences were found for the composite endpoint adverse outcome and length of hospital stay for these analyses.

Association of nutritional parameters and prealbumin concentrations
Finally, we investigated the association of other nutritional parameters with prealbumin concentration in a linear regression analysis (Supplemental Table 4).Change in albumin concentration was associated with significant change in prealbumin, with an increase of 0.071 (95%CI 0.06 to 0.082; P<0.001) units prealbumin for every 10 unit increase in albumin levelv.
Other significant associations were found for CRP, NRS, BMI chronic kidney failure, with lower coefficients.These results are also graphically displayed with correlation graphs between prealbumin and albumin as well as NRS in the Supplementary Figures.

Discussion
In this secondary analysis of a large randomized clinical trial, we investigated the prognostic value of serum prealbumin at admission to predict mortality at 180 days and response to nutritional treatment.Our results indicate that among medical inpatients at nutritional risk, prealbumin was a strong prognostic marker for long-term mortality as well as other adverse outcomes with robust results in different statistical models adjusted for other prognostic indicators.Also, prealbumin appeared to be the strongest prognostic indicator when compared with albumin and NRS, and improved the NRS score regarding prognostication.However, the difference in morality in intervention group patients receiving individualized nutritional support compared to control group patients receiving usual care nutrition was similar in subgroups of patients with normal and low prealbumin levels suggesting that despite its prognostic value, admission prealbumin concentrations are not helpful in selecting patients for nutritional treatment in a population with elevated nutrition risk (NRS ≥3).
Our finding that prealbumin is a prognostic factor among nutritionally at-risk patients is largely in line with previous results. 31,32,34,35The strength of this analysis includes the large and wellcharacterized patient cohort with collection of different nutritional parameters and prospectively collected short-as well as long-term outcomes.We were thus able to rigorously adjust our analyses for potential confounders including markers of inflammation.Importantly, these adjusted analyses for the inflammatory marker CRP suggest that prealbumin provides independent information from inflammation -a condition well known to influence plasma proteins and its precursors, as also shown in the correlation analyses (Table 4). 34Still, even though our analysis shows that prealbumin is a strong prognostic factor, it does not offer insight on whether low prealbumin concentrations are part of the pathway to increased mortality or whether it is simply a surrogate marker for more severe illness and therefore correlates with higher mortality.As a limitation to this report, we only had admission prealbumin levels and it thus remains unclear whether prealbumin dynamics over time could help predict clinical outcomes and whether it would be useful to monitor prealbumin concentrations in hospitalized patients to understand the risks for treatment failure.
While nutritional factors likely play an important role regarding the concentration of prealbumin, in the hospital setting there are several other important parameters influencing prealbumin including inflammation and the severity of illness. 31,36Thus, similar to albumin, some researchers argue that prealbumin should be considered as a negative acute phase protein reflecting more the acute situation and systemic inflammation and less the nutritional status. 31,36In addition, prealbumin concentrations are influenced by different diseases including kidney-or liver failure, as prealbumin is primarily produced in the liver and degraded by the kidneys. 18,37Still, some authors see prealbumin as a protein that also reflects the nutritional status of patients. 18,19,38,39In fact, a recent study suggested specific cut-offs of transthyretin to define malnutrition. 32Our data also confirm that prealbumin is highly correlated with the NRS score, but may not substitute a nutritional assessment.Importantly, while most studies have correlated prealbumin with nutritional parameters and clinical outcomes, there is a lack of studies looking at this marker to predict response to nutritional treatment.Herein, this report is to our knowledge the first to investigate the role of prealbumin as a predictor for the effectiveness of nutritional support.In our analysis, response to nutritional treatment did not differ in patients with high or low admission prealbumin concentrations and we found no significant interaction between prealbumin concentrations and effectiveness of nutritional support.However, our data suggest that treatment differs according to the baseline inflammatory status of patients -a result that concurs with a previous analysis from our trial. 10It is also in line with other trials suggesting less pronounced effects of nutritional support in severely ill patients and in patients with a high degree of inflammation. 13,40,41Thus, our results indicate that among medical inpatients at nutritional risk, prealbumin is not an optimal nutritional biomarker to select patients for nutritional support.This conclusion is also supported by a recent consensus paper stating that visceral proteins should not be measured to diagnose malnutrition or guide the indication for nutritional support. 31,42is secondary analysis has several strengths and limitations.To our knowledge, this is the first secondary analysis based on a randomized controlled clinical trial data to investigate whether low prealbumin concentrations are associated with effectiveness of nutritional support.Nonetheless, we only measured prealbumin levels, albumin and CRP but did not measure other biomarkers such as retinol-binding protein. 34,43We only had prealbumin in about one fourth of the initial trial population limiting the power of the analysis.This analysis was not done based on a power calculation, but we used all subjects with available prealbumin levels from one site.There was no evidence for a site effect in the original study, but the smaller sample size of this sub-study may explain why confidence intervals were large and effects were not significant, while there was a significant effect reported in the original EFFORT trial.
There are also limitations regarding the underlying trial including selection bias due to inclusion and exclusion criteria, lack of a control group not at nutritional risk, the pragmatic design with some patients not reaching their nutritional goals among others.Thus, results of this secondary analysis should be rather viewed as hypothesis generating and not definite.For sure, a prospective validation in an independent sample is needed.
In conclusion, this secondary analysis of a randomized clinical trial suggests that among medical inpatients at nutritional risk, low admission prealbumin levels correlate well with different nutritional markers and indicate higher mortality risk, but are not helpful in identifying patients who may or may not respond to nutritional support.Further studies are required to identify nutritional markers that help to further personalize nutritional interventions.

List of Steering Committee members of initial trial:
Philipp Schuetz, Filomena Gomes, Rebecca Fehr, Claus Hoess, Vojtech Pavlicek, Christoph Henzen, Jacques Donzé, Zeno Stanga, Beat Mueller Full list of local investigators of initial Trial:

Definition of outcomes assessed during the initial trial
• The primary composite endpoint consists of adverse clinical outcomes within 30 days defined as follows: (a) Mortality defined as all-cause mortality from inclusion to day 30 (b) ICU admission defined as admission to the intensive care unit from the medical ward from inclusion to day 30 V. gastro-intestinal events (hemorrhage, intestinal perforation, pancreatitis [defined as 2 out of 3 criteria: abdominal pain, 3-fold increase in lipase or pancreas-specific amylase, characteristic imaging findings]) (e) decline in functional status of 10% or more from admission to day 30 measured by the Barthel`s index. 1 This index measures performance in activities of daily living and comprises two groups of items, one related to self-care (feeding, grooming, bathing, dressing, bowel and bladder care, and toilet use), the other related to mobility (ambulation, transfers, and stair climbing).We used the German translation which has a score ranging from 100 to 0 with lower scores indicating more severe disability.
• Secondary endpoints assessed were defined as follows: (a) each single component of the primary endpoint at day 30 (b) Longterm mortality at 180 days (c) daily protein and energy intake as assessed by clinical nurses and trained registered dieticians using food records for each patients` meal (d) total length of hospital stay defined as total inhospital days during the index hospital stay from inclusion to day 30 (e) quality of life measured on admission and at 30-day day using the EuroQol Group 5-Dimension Self-Report Questionnaire. 2 This included the European Quality of Life 5 Dimensions index (values range from 0 to 1, with higher scores indicating better life quality) and the visual-analogue scale (EQ-5D VAS) (scores range from 0 to 100, with higher scores indicating better health status).
• Safety endpoints including side effects from nutritional therapy are daily assessed until hospital discharge and are defined as: a) adverse gastrointestinal effects defined as obstipation, diarrhea, nausea, vomiting, abdominal pain b) complications related to enteral nutrition (defined as any complications associated with tube feeding) or parenteral nutrition (defined as any complications associated with central venous catheter) c) refeeding syndrome defined according to a recent consensus definition

942 completed interview at day 30 913 completed interview at day 30 35 withdrew informed consent 0 lost to follow up 25 withdrew informed consent 0
lost to follow-up 264 included in secondary analysis 253 included in secondary analysis 751 excluded because of missing admission prealbumin levels 760 excluded because of missing admission prealbumin levels

Table 3 . Predictive Value of prealbumin regarding effectiveness of nutritional support
*Adjusted for NRS, Barthel Score