| Sign In to gain access to subscriptions and/or personal tools. |
Surrogate Nutrition Markers, Malnutrition, and Adequacy of Nutrition SupportDepartments of Medicine and Surgery, Beth Israel Medical Center and Albert Einstein College of Medicine, New York, New Yorka Correspondence: Correspondence: David S. Seres, MD, CNSP, 170 Second Avenue, Apartment 2D, New York, NY 10003. Electronic mail may be sent to dseres{at}chpnet.org. Surrogate nutrition markers are used to assess adequacy of nourishment and to define malnutrition despite evidence that fails to link nourishment, surrogate markers, and outcomes. Markers such as serum levels of albumin, prealbumin, transferrin, and IGF-1 and delayed hypersensitivity and total lymphocyte count may be valid to help stratify risk. However, it is not appropriate to consider these as markers of adequacy of nourishment in the sick patient. Although there is little debate that the overall incidence of malnutrition is high among hospitalized patients, there is significant doubt as to whether surrogate nutrition markers reflect the adequacy of nourishment. Surrogate markers are often used to prove the value of a medical therapy because the collection of outcome data is cumbersome and expensive. Accurate records of the incidence of complications, length of hospital stay, mortality, and cost of medical therapy are difficult to obtain. Conversely, obtaining a serum albumin level is easy, rapid, and is most often covered by insurance. This has made serum albumin concentration and other simple tests attractive candidates as surrogate markers for the adequacy of nutrition therapy. But, although changes in surrogate nutrition markers may be associated with changes in clinical outcomes, and increases in protein levels have been observed in patient populations when nutrition support techniques have been applied, these changes are not consistent and are too small to be of value in an individual patient. The rationale for providing nutrition support consists of several assumptions: altered level or metabolism of a nutrient indicates malnutrition, this alteration itself worsens prognosis, a nutrition approach can correct the alteration, and correction will yield a better clinical outcome.1 However, there is a lack of uniformity in the definition of malnutrition (Table 1) and nutrition support is unlikely to overcome alterations in metabolism that result from disease.2 Further, there is evidence that changes in anthropometrics, serum protein concentrations, and markers of immune function do not occur in concordance with changes in outcome.3,4 In a recent statistical review, which will be discussed in more detail later, Koretz4 analyzed 99 studies for concordance between the direction of change in 12 nutrition parameters and true outcomes (mortality, infections, complications, duration of hospitalization). He concluded, "... changes in nutritional markers do not predict clinical outcomes."4
In a classic 1981 study, Harvey et al5 evaluated a population of patients receiving nutrition support. They found that improved delayed hypersensitivity was strongly predictive of survival, whereas an increase in serum albumin level was less predictive. They concluded that the likelihood of death was higher when these markers did not improve. This would suggest, because all were receiving nutrition support, that the patients' underlying conditions, and not nourishment, were responsible for the levels of these markers.5 Despite these conclusions and numerous supporting scientific papers, reviews, and editorials, these markers are still considered reflective of adequacy of intake.6,7 In fact, regulatory agencies use serum protein concentrations as quality indicators of adequate nutrition and have gone so far as to penalize skilled nursing facilities when patients have persistent but unavoidable hypoalbuminemia.8 A narrative review of the validity of surrogate nutrition markers follows. The reader is cautioned that, in general, narrative reviews are flawed and biased by their nature.9 Further caution is warranted as, despite an effort to be objective, this author has long held strong opinions about the subject.10
There are as many definitions for malnutrition as there are sources for them (Table 1). The common thread among most is the inadequate or excess intake of nutrients. Less frequently included in these definitions are the alterations in metabolism or in body or blood composition that are induced by disease or medication. It is unusual, due to technical difficulties and cost, for the state of repletion of a body compartment (ie, total body protein) to be measured directly. Rather, we rely on surrogate markers (serum protein concentrations) to help predict the state of the compartment in question.11
The incidence of malnutrition in hospitalized patients is often quoted as approaching 50%,7 but the incidence reported depends upon how malnutrition is defined. Measures of dietary intake, anthropometric measurements, weight change, serum protein concentrations, vitamin levels, hematocrit, and response to immunologic testing have all been used, in a variety of combinations, to help identify malnourished patients. The prevalence of malnutrition reported in hospitalized patients varies widely and increases proportionate to length of stay.12–14 Within a single study population, the incidence of "malnutrition" varied from 44%–76%, depending on methods used to assess nutrition status.14 A significant decline in nutrition status during hospitalization has been reported in as many as 75% of patients.13 Many studies over many years have shown point-in-time relationships between anthropometric or biochemical markers and prognosis. In at least 1 study, a relationship was not found between anthropometric measurements or weight loss and morbidity or mortality in surgical patients.1 However, in a prospective multicenter study evaluating 4301 patients with a variety of severe medical illnesses, low body mass index was a significant and independent predictor of mortality.15 Others report that anthropometric evidence of malnutrition predicts mortality as well as or better than serum albumin concentration in general medical patients13 and patients undergoing peritoneal dialysis.16 An association has been demonstrated between total body protein depletion and the incidence of pneumonia in surgical patients.11 Rubinson et al found that poor intake (<25% of estimated goal), but not serum albumin level, was associated with a higher incidence of bloodstream infection in medical intensive care unit patients.17 On the other hand, serum albumin concentration has been shown to be 4–21 times more powerful than markers of dialysis adequacy in predicting mortality in patients with endstage renal disease.16,18–20 In a variety of clinical studies, serum levels of albumin, prealbumin, and transferrin, delayed hypersensitivity (a surrogate marker for immune function), and total lymphocyte count have been show to variably predict postoperative complications, length of hospital and intensive care unit stay, rehospitalization, mortality, antibiotic use, and resumption of oral intake.1,16,21–24 These findings must be interpreted in context. Research studies, by combining a number of individuals into a large data set, may enable the detection of small effects that are statistically significant, but not clinically relevant. In one study, serum albumin level only predicted 7.1%–11.6% of the variance in outcomes in a mixed population of patients.25
As mentioned above, Koretz recently performed a statistical review of the predictive value of 12 nutrition markers, including calorie or protein intake; weight gain; nitrogen balance; serum levels of albumin, prealbumin, and transferrin; 3 anthropometric measures; skin antigen testing; and lymphocyte count.4 Results of 99 randomized controlled studies were used. Each marker was compared with each of 4 outcomes: mortality, infections, total complications, and duration of hospitalization (48 comparisons). The analysis found that rates of concordance were <50% in 41 of 48 comparisons, demonstrating that the chosen surrogate markers are not valid for assessing the efficacy of enteral or parenteral nutrition.4
Unless outcomes are measured concurrently, the concordance between surrogates cannot be taken as evidence that the surrogates are sensitive or specific. Further, concordance between markers is lacking from study to study. The following studies demonstrate how many researchers have attempted to show a relationship between the surrogate markers and outcome by comparing 1 surrogate to another. Most of the articles cited in the remainder of this section were chosen because they were used in other articles to support the validity of these markers to reflect nutrition status. In one study designed to describe the prevalence of malnutrition in patients, the incidence of abnormal nutrition markers ranged from 44%–76%. Midarm circumference correlated with serum albumin levels, but triceps skinfold measurements did not.14 Another study found that the direction of change of serum prealbumin concentration correlated with the direction of nitrogen balance in only 39 of 54 patients. None of the other plasma proteins measured (albumin, transferrin, and retinol binding protein) achieved 50% concordance with nitrogen balance.26 In another study, the change in serum transferrin levels correlated with nitrogen balance, but changes in prealbumin did not. Decreases in serum levels of prealbumin and transferrin were also associated with lower intake of calories. Surgery was associated with a decrease in prealbumin concentration and negative nitrogen balance.27 Carlson et al measured serum protein levels (albumin, prealbumin, retinol-binding protein, and transferrin) and nitrogen balance.28 There was no correlation between nitrogen balance and albumin level. Although statistically significant correlations were found between the other markers and nitrogen balance, the correlation was too weak to predict even the direction of nitrogen balance. In a study of 41 patients receiving enteral nutrition after trauma, a rise in serum prealbumin concentration occurred in the face of negative nitrogen balance.29 Keshaviah et al found a positive correlation between lean body mass and serum albumin concentration in hemodialysed patients.30 McCusker et al showed that initial serum albumin level correlated with lean mass, subjective global assessment score (SGA), and survival, but that although SGA and lean mass improved over a 6-month period, serum albumin levels remained the same.31 Total body protein was measured in another study over a period of 3 weeks in patients with sepsis and multiple injuries. Although correlating with each other for most of the study period, serum concentrations of insulin-like growth factor 1 (IGF-1), transferrin, or prealbumin did not predict total body protein changes.32
If nutrition markers are reflective of adequacy of intake, then there should be a direct relationship between weight loss and serum protein levels. In 12 consecutive patients who underwent uncomplicated gastric bypass at Beth Israel Medical Center for the treatment of obesity, for whom data were available, serum albumin levels behaved in the opposite manner. Albumin levels were low just before hospital discharge after surgery, and normalized during the next 3 months despite large amounts of weight loss in the same period. Body mass index (BMI) (Fig. 1) decreased significantly from 53.3 kg/m2 preoperatively to 43.2 kg/m2 at 3 months postoperatively (p < .001). Serum albumin concentration (Fig. 2) dropped significantly after surgery from 4.1 mg/dL to 3.1 mg/dL (p < .001). In the subsequent 3 months, serum albumin concentration normalized to 3.9 mg/dL (p = NS for preoperative vs 3 months postoperatively; p < .001 immediate postoperative vs 3 months postoperative; author's unpublished data). Other researchers have examined the relationship between nutrition markers and intake. Once again, comparing marker to marker rather than comparing marker with outcome is common, and there is a lack of concordance in results. Thomas et al reported that after 1 week, serum transthyretin (prealbumin) level increased with increased intake of calories and protein in preterm neonates.33 There was no similar relationship between intake and serum albumin or transferrin levels. In a study of patients receiving parenteral nutrition, an improvement in immune function testing was demonstrated. No other outcomes were evaluated.34 Shenkin et al compared 2 groups receiving isocaloric parenteral nutrition with or without protein. 35 There were no differences in serum levels of albumin, prealbumin, transferrin, retinol binding protein, C3, C-reactive protein, RNAase, or creatinine, despite a significant difference in nitrogen balance between the 2 groups. Clemmons et al were unable to show a difference in pre- and postnutrition support levels of prealbumin, retinol binding protein, or transferrin in a group of 6 chronically ill and malnourished patients receiving nutrition support for 10–16 days.36 IGF-1 levels increased initially, but then fell to baseline. Somatomedin (IGF-1) was found to correlate with caloric intake in a study of 41 hospitalized patients. The correlation was statistically significant, but the variance of values was extreme. In the same study, there was no correlation between anthropometric measures (triceps skinfold and midarm circumference assessed in only 10 patients, and percent ideal body weight) and serum markers (albumin, transferrin, total lymphocyte count, IGF-1), and no correlation between serum albumin or transferrin levels, total lymphocyte count, and calories consumed. Laboratory values were obtained at baseline in only 6 patients and the period of nutrition intervention was only 3–16 days.37 As stated earlier, a statistically significant result in a study may not be clinically relevant. For a marker to be applicable to individual patients, the variance in the level must be lower than the effect of the intervention. For example, statistically significant rises in serum albumin concentration reported are often below the level of day-to-day variation from fluid shifts. In some reports, the range in serum albumin level changes was 0.15–0.3 g/dL.38,39 Statistically significant changes in serum prealbumin concentration have been reported to be as low as 2.5–3 mg/dL.40 Several other studies deserve mention as they evaluate the relationship between surrogate markers and outcome while illustrating other pitfalls in the studies of these markers. In an oral supplementation (mean 32 days) vs no supplementation study of 59 elderly patients (27 subjects, 32 controls) with hip fracture, a significant decrease in negative outcomes was obtained in the supplemented group. The mean serum albumin concentration was statistically higher in the supplemented group at 6 weeks, but, as mentioned above, the difference was only 0.2 g/dL.39 In a crossover study of 43 patients receiving parenteral nutrition (PN) for at least 7 days while undergoing chemotherapy, PN-treated patients had significantly better nitrogen balance but there was no difference in outcome. Serum prealbumin levels rose during PN but were not statistically different from the untreated group.41 If a marker were important to the pathophysiology of an illness, then one would assume that the improvement of the marker would result in improved outcome. In a study of enterally fed trauma patients, growth hormone administration was shown to increase serum albumin concentration (0.2 mg/dL) and serum transferrin concentration (from 179 to 257 mg/dL) without changing nitrogen balance. There was no effect on serum levels of prealbumin or C-reactive protein. Prealbumin increased and C-reactive protein decreased similarly in both groups. Outcome was reported only in the discussion section where the authors report no deleterious side effects.42 It should be noted that the use of growth hormone in the critically ill has since been discarded as it is deemed more harmful than beneficial. Despite improvements in nutrition surrogate markers, true outcomes have not been significantly improved and mortality is increased with its use.43,44
The corollary to whether nutrition surrogate markers predict nutrition or clinical outcomes is whether these tests improve the sensitivity of nutrition assessment and the identification of malnourished patients. In a multicenter study of 680 patients receiving peritoneal dialysis, serum albumin levels were predictive of death, technique failure, and hospitalization, but so were scores on the Subjective Global Assessment.16 The clinical assessments of a series of 59 surgical patients by 2 independent examiners were compared with each other and with measurements of albumin, transferrin, total lymphocyte count, percent ideal lean weight, percent total ideal weight, percent body fat, total body nitrogen, total body potassium, and delayed hypersensitivity. The 2 observers agreed in 81% of initial assessments. Further, there were significant correlations with all but total lymphocyte count between subjective assessments and surrogate markers. Finally, clinical assessment was predictive of complications and length of stay.23 In a recent study,7 researchers concluded that serum prealbumin levels improve the screening of hospitalized patients, identifying those at risk for malnutrition sooner and better than a previously validated nursing admission screening tool. They found a good correlation between decreased serum prealbumin levels and the probability that a registered dietitian (RD) would identify the patient at high risk for malnutrition. They present this as evidence of validity. There were no such relationships between serum levels of retinol-binding protein or albumin and dietitian assessment. Half of the patients not identified as high risk on the nursing screen were identified as "malnourished by prealbumin criteria." The authors conclude that, because the historical rate of malnutrition in hospitalized patients is 50%, and the nursing screen only identified 33% of patients as "at risk," prealbumin improves the sensitivity of screening. The RD evaluated 57% of the patients. The authors suggest that, because there was a high probability that patients with low serum prealbumin concentrations would also be assessed by the RD as malnourished, this concordance would apply to the population not seen by the RD.7 These extrapolations are not appropriate support for the authors' conclusions, or for dismissing a tool that was previously validated (the nursing screen).
As early as 1982, authors had concluded that "... although these indicators are epidemiologically useful and correlate with morbidity and mortality, no single measurement is of consistent value in individual patients."23 In an 1988 editorial in this journal, Wolfe concluded that although "hypoalbuminemia has been identified as an accurate prognostic factor.... [it is] a nonspecific measure of the degree of illness rather than a specific test of the adequacy of... intake."45 In a 1994 study of the incidence of malnutrition in hospitalized patients, the authors excluded "biochemical assessment of nutritional status because of a lack of a reliable biochemical marker of nutritional status."12 In an editorial in JPEN in 1997, Kirby stated that "serum albumin is often abused as a `marker' of nutrition adequacy."6 In 1997, Souba commented that "hypoalbuminemia is not specific to poor nutritional status" and that "improvements in nutritional markers... have not usually been accompanied by clinical benefits."3 McClave, in 1999, wrote that these parameters are "impractical, poorly reproducible, insensitive, inappropriate, and unreliable..."46 Another narrative review supporting this view was published within the past year.47 Koretz has now provided hard evidence that these markers do not reflect adequacy of intake.4 In conclusion, surrogates such as serum levels of albumin, prealbumin, transferrin, and IGF-1; delayed hypersensitivity; total lymphocyte count; and others should no longer be referred to as "nutrition" markers. Although it is certainly true that there may be associations between decreases in these markers and adverse outcomes, nutrition alone does not correct the reductions, and correction may not improve outcome. The author gratefully acknowledges the following for assistance: the Staff of Seymour J. Phillips Medical Library; Beth Israel Medical Center for aid in obtaining reprints; Laurie Focacci, PA, and Elliot Goodman, MD, for generously providing data; Peter Homel, PhD, for statistical guidance; and Kesiah Scully, PhD, for editorial wizardry.
1 Mullen JL, Gertner MH, Buzby GP, Goodhart GL, Rosato EF.
Implications of malnutrition in the surgical patient. Arch
Surg. 1979;114:121
–125.
Nutrition in Clinical Practice, Vol. 20, No. 3,
308-313 (2005) This article has been cited by other articles:
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
|
 |
|
Sign In to gain access to subscriptions and/or personal tools.
Top
Defining Malnutrition
Incidence and Clinical Impact...
