CJASN ePress. Published on July 16, 2015 as doi: 10.2215/CJN.08430814
Article
Nutritional Competence and Resilience among Hemodialysis Patients in the Setting of Dialysis Initiation and Hospitalization Stephan Thijssen,* Michelle M.Y. Wong,* Len A. Usvyat,† Qingqing Xiao,* Peter Kotanko,*‡ and Franklin W. Maddux†
Abstract Background and objectives Dialysis patients have a high risk for inadequate nutrition. Their nutritional status is particularly susceptible to deterioration when faced with intercurrent events such as hospitalization. This study was conducted to improve the understanding of the temporal evolution of nutritional parameters as a foundation for rational and proactive nutritional intervention. Design, setting, participants, & measurements A retrospective cohort study was performed to investigate the temporal evolution of nutritional parameters (serum albumin, serum phosphate, serum creatinine, equilibrated normalized protein catabolic rate, and interdialytic weight gain) and a composite nutritional score derived from these parameters, in two populations: (1) incident hemodialysis (HD) patients who started HD between January 2006 and December 2011 and were followed for up to 54 months (median 16.3), and (2) prevalent patients with HD vintage $2.5 years who were hospitalized between January 2006 and December 2011 and followed from 6 months before to 6 months after hospitalization. Results In incident patients (n=126,964), each of the nutritional parameters improved after HD initiation, with a mean composite nutritional score at the 24th percentile at the start of HD and reaching a plateau at the 57th percentile toward the end of the second year on dialysis. Nutritional parameters increased more rapidly and reached higher values among patients who survived longer. In hospitalized patients (n=14,193), the nutritional parameters and the composite score began to decline 1–2 months before hospitalization, reached their lowest level in the month after hospitalization, and then partially recovered in the subsequent 5 months. The degree of recovery of the nutritional score was inversely related to the number of rehospitalizations.
*Renal Research Institute, New York, New York; †Fresenius Medical Care North America, Waltham, Massachusetts; and ‡ Icahn School of Medicine at Mount Sinai Hospital, New York, New York Correspondence: Dr. Stephan Thijssen, Renal Research Institute, 315 East 62nd Street, Fourth Floor, New York, NY 10065. Email: [email protected]
Conclusions This study increases the understanding of nutritional resilience and its determinants in HD patients. Application of the nutritional score, pending further validation, may facilitate targeted and timely interventions to avert the negative consequences of inadequate nutrition in chronic HD patients. Clin J Am Soc Nephrol 10: ccc–ccc, 2015. doi: 10.2215/CJN.08430814
Introduction Inadequate nutrition is frequent in chronic hemodialysis (HD) patients, contributing to malnutrition and protein-energy wasting, which are major contributors to morbidity and mortality in this population (1). Assessing nutritional status in a HD patient is complex and often requires a multidisciplinary team. Clinical practice guidelines recommend that nutritional status be assessed using several tools, including dietary assessment, body mass index, subjective global assessment, anthropometry, protein nitrogen appearance, serum albumin, prealbumin, cholesterol, and technology such as bioimpedance (2,3). We use the term “nutritional competence” to refer to a patient’s nutritional status at the time of evaluation. A patient who is nutritionally competent possesses an adequate nutritional status at that particular time. Therefore, this is a cross-sectional assessment. The term “nutritional resilience” refers to a www.cjasn.org Vol 10 September, 2015
patient’s capacity to maintain nutritional competence in the face of a challenge (4,5). Thus, nutritional resilience is a dynamic measure that indicates a patient’s response to nutritional challenges, such as hospitalization. Eligibility for intradialytic nutritional supplements in Fresenius Medical Care North America (FMCNA) clinics is currently based on serum albumin levels. However, the use of serum albumin as a nutritional marker is problematic, because inflammation and fluid status can also affect its levels (6,7). The limitations of using albumin as the sole criterion for nutritional intervention call for improved strategies to assess nutritional competence and resilience. Most existing nutritional scores developed in the ESRD population are derived from cross-sectional data and do not consider dynamics of nutritional parameters (8–10). Therefore, a retrospective cohort study was performed in HD patients with the goals Copyright © 2015 by the American Society of Nephrology
1
2
Clinical Journal of the American Society of Nephrology
of (1) characterizing temporal patterns of nutritional parameters and (2) generating a composite nutritional score from readily available nutritional parameters and assessing its dynamics in incident HD patients over the first 2 years on dialysis and in prevalent patients around hospitalization.
Materials and Methods The study was approved by the New England Institutional Review Board (NEIRB number 14-109). We retrospectively studied HD patients aged $18 years treated in FMCNA clinics between January 1, 2006, and December 31, 2011. Two main analyses were conducted: The first assessed nutritional parameters in incident patients over the first 24, 48, and 54 months on HD, and the second assessed the association of hospitalization with temporal patterns of nutritional parameters in prevalent HD patients. Patients’ monthly averages of serum albumin, creatinine, phosphate, equilibrated normalized protein catabolic rate (enPCR), and interdialytic weight gain (IDWG) were obtained during the analysis periods indicated below. Analyses were performed using all records with complete data for the above five parameters for any given month. A sensitivity analysis was performed in patients with complete data for all months of interest. A composite nutritional score was calculated using serum albumin, creatinine, phosphate, enPCR, and IDWG as follows: For each of these nutritional parameters, the values were standardized by subtracting the mean of all values (for the respective parameter) across all patients and all months of interest from each value and then dividing by the SD of the centered values across all patients and months of interest. For each patient and month, the five standardized values were added and then again standardized as above across all months and patients, resulting in a distribution of composite nutritional scores with a mean of 0 and a SD of 1. These standardized composite score values were then converted to percentiles according to the probability density function of the standard normal distribution. Readers may apply the methodology to their own patients using the data from Table 1, which also gives an example. To graph the rate of change of the composite nutritional score, cubic B-spline basis functions were used to approximate the expectation of percentile of nutritional score. The function of the first derivative was approximated by finite differences. The composite nutritional score was deliberately created based on widely available variables, thus avoiding data acquisition beyond routine practice. Although IDWG is recorded at every dialysis session, FMCNA protocols require at least monthly measurements of the other parameters. Incident HD Patient Analyses Incident patients who started HD between January 1, 2006, and December 31, 2011 were included. Monthly mean serum albumin, creatinine, phosphate, enPCR, and IDWG values were graphed on radar plots at months 1, 3, 6, 9, 15, and 24. The radar plots were scaled such that the innermost point on each spoke represents the minimum average value observed for the respective parameter across all months,
Table 1. Mean and SDs for nutritional parameters and additive nutritional score in our data set
Parameter
Mean
SD
Serum albumin (g/dl) Serum creatinine (mg/dl) Serum phosphate (mg/dl) enPCR (g/kg per day) IDWG (liters) Additive score
3.8 7.3 5.2 0.89 2.57 0
0.4 2.9 1.5 0.27 1.11 3.14
The reader may use these data to calculate nutritional scores in his or her patients. For example, patient values are as follows: albumin 4.1 g/dl, creatinine 7.4 mg/dl, phosphate 5.1 mg/dl, enPCR 0.9 g/kg per day, and IDWG 2.8 L. The additive nutritional score is calculated as follows: (4.1 2 3.774)/0.433 + (7.4 2 7.254)/2.869 + (5.1 2 5.231)/1.5 + (0.9 2 0.885)/0.273 + (2.8 2 2.574)/1.107 = 0.976. The standardized additive score is calculated as follows: (0.976–0)/3.136 = 0.311, which converts to a percentile of 62.2. enPCR, equilibrated normalized protein catabolic rate; IDWG, interdialytic weight gain.
and the outermost point on each spoke represents the maximum average value. The composite nutritional score was tracked monthly over 2 years. A subgroup analysis with stratification according to survival time was performed in patients who either survived $48 months or died within the first 48 months. We conducted a subgroup analysis in patients without residual urea clearance at HD initiation. Hospitalization Analyses We included patients who were hospitalized for 7–14 days between January 1, 2006, and December 31, 2011 (regardless of when their first date of dialysis was) and who had been on HD for at least 2.5 years at the time of hospitalization. We further required patients to have data from 6 months before hospitalization to 6 months after discharge and no other hospitalizations in the 6 months before this index hospitalization. Only data from the first eligible hospitalization per patient were used. This hospitalization cohort and the incident patient cohort described above are separate cohorts with separate criteria (i.e., there are patients who contribute to one but not the other cohort); however, there is some degree of overlap between them. Serum albumin, creatinine, phosphate, enPCR, IDWG, and the composite nutritional score were plotted over 6 months before and after hospitalization. Subgroup analyses were performed with stratification for (1) the number of rehospitalizations over the 6 months after the initial hospitalization, and (2) the three most common categories of reason for hospitalization. For each month after the first hospitalization, statistical comparison across nutritional score groups was performed by the Kruskal–Wallis test, and we further tested for linear trend. Confidence intervals were obtained via bootstrapping with 1000 resampling iterations. Analyses were carried out using R software (version 3.0.2) (11). Radar graphs were created with the fmsb package (12); all other plots were created with ggplot2 (13).
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
Nutritional Patterns in Dialysis, Thijssen et al.
Results Baseline characteristics of the study populations are shown in Table 2. Averages and SDs of the nutritional parameters and additive score are presented in Table 1. Incident Patient Analyses There were 126,964 patients who had complete records for at least 1 month for all nutritional parameters during the first 24 months on HD. All five parameters increased over the first 2 years on HD (Figure 1A). The composite nutritional score improved from the 24th percentile at the start of HD to approximately the 57th percentile at 2 years (Figure 1B). The rate of improvement in the composite nutritional score was rapid after HD initiation and then quickly decelerated over the first 4 months, followed by a slower deceleration over the subsequent several months, with the score gradually approaching a plateau toward the end of year 2 (Figure 1C). Excluding patients who died within the first 6 months on dialysis yielded materially identical results (data not shown). For follow-up over 54 months, see Supplemental Figure 1. A subgroup analysis in patients with no residual urea clearance at HD initiation (59% of the initial cohort) showed materially identical results (Supplemental Figure 2). The nutritional score trajectories in incident patients stratified by survival time showed that both the initial rate of improvement of the nutritional score as well as its maximum level were directly related to patients’ survival times (Figure 2; for number of included records, see Supplemental Figure 3).
Table 2. Characteristics of the study populations
Characteristic
Number of patients Age (yr) Men Race Black Asian Other Diabetes Cardiovascular disease Cancer (excluding skin cancers) Hyperparathyroidism Chronic obstructive pulmonary disease Drug or alcohol dependence Gastrointestinal bleeding Infections
Incident Hemodialysis Patients
Hemodialysis Patients Who Were Hospitalized
145,847 63.1615.0 57.6
14,724 56.4614.7 51.7
30.5 2.3 67.2 61.6 55.6
47.3 1.5 51.2 61.0 67.7
6.9
6.3
4.3 10.9
9.8 12.1
3.0
3.6
1.2
2.6
10.2
20.4
Data are given as the mean6SD or percentage, unless otherwise indicated.
3
Hospitalization Analyses These analyses included 14,193 patients who had complete records for at least 1 month for all nutritional parameters. The nutritional parameters began to decline 1–2 months before hospitalization, reached their nadirs in the month after hospitalization, and then recovered to varying degrees in the 6 months after this initial hospitalization (Figure 3). This pattern is also reflected in the composite nutritional score (Figure 4A). The rate of decline in the nutritional score rapidly accelerated in the 2 months before hospitalization, and the score rebounded over the 3 months after hospitalization, albeit not to the original level (Figure 4B). Of all hospitalized patients, 8431 (59.4%) were rehospitalized within 6 months of the initial hospitalization. Analysis broken down by number of rehospitalizations during these 6 months revealed that the degree of recovery of the nutritional score showed a clear, inverse relationship to the number of rehospitalizations, with patients who were not rehospitalized experiencing a full recovery of their nutritional score (Figure 5A). At each of the 6 months after the initial hospitalization, comparison of mean percentiles across the five rehospitalization categories yielded highly significant results (P,0.001 for every month). Likewise, there was a highly significant “dose-response” relationship between number of rehospitalizations and percentile of the nutritional score at each month after the initial hospitalization (all P,0.001). The most common reasons for hospitalization were infections, circulatory system causes (including cardiovascular disease and stroke), and gastrointestinal system causes. Nutritional parameters showed a similar pattern of decline and recovery regardless of cause of hospitalization (Figure 5B). However, of the three groups, the gastrointestinal diseases group consistently had the lowest nutritional score throughout the 12-month period.
Discussion This large retrospective study, comprising close to 100,000 incident HD patients and .14,000 hospitalized patients, describes the dynamics of nutritional parameters after HD initiation and around hospitalizations. Although the five parameters examined are considered nutritional indicators, there are also caveats regarding their use as nutritional markers, which emphasizes the need to consider multiple parameters in nutritional assessment. Serum albumin is influenced by inflammation (6,7). IDWG is a surrogate for dietary intake but may also reflect fluid and sodium balance. The enPCR is a marker of steadystate protein intake (3). Whereas low enPCR reflects decreased dietary protein intake, very high enPCR could indicate tissue catabolism. Predialysis serum creatinine, an indicator of muscle mass (14), is also influenced by meat intake, residual renal function, and dialysis adequacy (15,16). The score calculated in this study, reflecting serum albumin, creatinine, phosphate, enPCR, and IDWG, indicates a relatively rapid improvement in these nutritional domains over the first few months on HD, followed by a gradual leveling off and eventually leading into a plateau toward the end of the second year. A subgroup
4
Clinical Journal of the American Society of Nephrology
Figure 1. | Evolution of nutritional parameters in incident hemodialysis patients. (A) Radar plots at months 1, 3, 6, 9, 15, and 24 from initiation of hemodialysis. Only patients with a complete set of data for all five shown variables for any given month were included (n=126,964). Each radar plot displays the five variables on axes (“spokes”) arranged radially. The mean value of each variable is plotted along the respective spoke; the innermost ring represents the minimum and the outermost ring represents the maximum value of that variable across all data points. A line connects the data values on each spoke. (B) Evolution of the composite nutritional score in patients starting hemodialysis. The mean percentile of nutritional score over 24 months after initiation of dialysis is shown. Error bars denote 95% confidence intervals. The number of available data records for each month is shown above the x axis along the bottom of the plot. (C) Rate of change in percentile of composite nutritional score in incident hemodialysis patients over the first 2 years on hemodialysis. Alb, serum albumin; Crea, serum creatinine; enPCR, equilibrated normalized protein catabolic rate; IDWG, interdialytic weight gain; P, serum phosphate.
analysis in patients without residual renal function at dialysis initiation showed materially identical results, indicating that the described trajectories are robust findings. Although all five analyzed parameters increased after initiation of HD, their rate of increase differed, with phosphate showing the fastest rise, followed by enPCR, IDWG, albumin, and creatinine. This seems biologically plausible, with phosphate being a rapidly responsive marker reflecting increased dietary intake. This is followed by enPCR, which might be expected to react somewhat slower because it may initially reflect tissue catabolism to some extent, which gradually decreases over time and counteracts the underlying increase in enPCR related to protein intake. It is plausible that albumin would take longer to recover than the other markers mentioned above, and that creatinine, being a reflection of muscle mass, would be the slowest parameter to recover. Improvement in nutritional parameters after initiating dialysis has been observed in previous studies in incident patients (17–19).
As shown in Figure 2, the nutritional score showed a decline in the months before death, a finding that has been demonstrated for several parameters in previous studies, including IDWG, serum phosphate, and serum albumin, which are components of our nutritional score (19,20). We found nutritional parameters to decline 1–2 months before hospitalization, reach their nadir right after hospitalization, and then improve again over the subsequent 5 months, although not to their original levels. Nutritional resilience refers to the degree to which a patient’s nutritional status is affected by intercurrent events (in this case, hospitalization). The nadir as well as the subsequent rate and degree of recovery of each individual parameter and the composite score characterize each patient’s nutritional resilience. Some patients are more susceptible to nutritional setbacks and protracted recovery than others (i.e., they possess a lower nutritional resilience). Of note, not all hospitalizations are comparable stressors to nutritional status. Similar to recovery after initiation of HD,
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
Nutritional Patterns in Dialysis, Thijssen et al.
5
Figure 2. | Time course of composite nutritional score after initiation of hemodialysis, stratified by patient survival time. The mean percentile of nutritional score is shown, with shaded areas representing the 95% confidence band of the fit.
recovery after hospitalization was not homogenous among the five parameters. Most notably, serum phosphate almost recovered to prehospitalization levels, whereas serum creatinine recovered poorly. Loss of lean tissue mass from acute inactivity with hospitalization can occur from decreased protein synthesis rate (so-called “anabolic resistance”) and elevated cortisol levels in the setting of inflammation (21–23). The incomplete recovery of creatinine may reflect the delay in regaining muscle mass, but protein intake likely also plays a role. In hospitalized HD patients, the composite nutritional score demonstrated the greatest magnitude of change in the 2 months before and after hospitalization, but even in patients with no rehospitalizations, the posthospitalization recovery was protracted over 4 months. Among patients who were rehospitalized more often, the nadir of the score was lower and recovery was less complete. The most common hospitalization causes demonstrated similar nutritional trajectories. Compared with other causes of hospitalization, the gastrointestinal diseases group had a lower nutritional score throughout the prehospitalization and posthospitalization periods, and this likely reflects increased susceptibility to dysfunctional nutrient intake, absorption, and/ or metabolism. It is tempting to speculate that the prolonged nutritional recovery time observed in rehospitalized patients is caused by the repeated detrimental effect of hospitalization on nutritional status. An additional potential explanation may be found in the behavior of complex dynamical systems (24): Such systems often exhibit characteristic properties that may serve as early warning signs for an impending critical transition. Dynamical systems theory describes a property called “critical slowing down,” whereby recovery from perturbations becomes increasingly slow. In this context, the recovery time in our nutritional score after
perturbation of the system by hospitalization may also be an indication of how close the patient’s biologic system is to a critical transition. As such, a patient’s nutritional resilience may be a direct measure of his or her health and impending complications. One might expect that an individual patient’s nutritional score recovery time after hospitalization (even without rehospitalization in the subsequent 6 months) may be indicative of downstream complications or mortality. Our study was not designed to analyze this aspect, but it would be a worthwhile investigation. We deliberately chose the 2.5-year vintage cutoff for our hospitalization analyses because the parameters of interest undergo, on average, marked changes in the 2 years after HD initiation. This approach limits the interference between these trajectories and our hospitalization analyses by allowing patients to stabilize after HD initiation before entering the hospitalization analyses. The effect of hospitalization on nutrition might vary in magnitude among patients of different dialysis vintage; however, our focus was on elucidating the general pattern and time course of the response, rather than characterizing the relationship between vintage and the effect of hospitalization or ensuring that the depicted response was the average of patients of all vintages. In addition, a shifting baseline before hospitalization would interfere with the interpretation of the response. Therefore, we felt that the benefit of reduced noise unrelated to hospitalization was the greater value for this particular analysis. A limitation of this study is its retrospective and observational nature. This design renders the study subject to potential bias owing to nonrandom data missingness. For example, in the incident patient analyses, patients who missed laboratory draws may have had worse nutritional
6
Clinical Journal of the American Society of Nephrology
Figure 3. | Time course of individual nutritional parameters around hospitalization. The bottom right plot shows the number of available records for each month. Dots denote the means. Error bars denote 95% confidence intervals. enPCR, equilibrated normalized protein catabolic rate; IDWG, interdialytic weight gain.
status than those who had data available. We addressed this concern by conducting sensitivity analyses that included only those patients who had complete data throughout the entire time period of interest. These analyses, in contrast with the shown analyses, likely included particularly healthy participants and therefore were subject to bias in the opposite direction. Yet these analyses showed materially identical results, which suggests that the presented time courses of the nutritional score are fairly robust across the HD population. For the analysis of hospitalized HD patients, no data were available during hospitalization; therefore, our analyses were limited to outpatient records before and after hospitalization. Therapy and changes in diet during hospitalization may have affected the parameters analyzed in this study. Another limitation is unavailability of data regarding dietary intake, body composition, socioeconomic status, and depression, preventing more in-depth analyses. In the incident patient analysis, changes in the mean composite nutritional score could have been affected by attrition from mortality. Patients with lower nutritional scores who died early in follow-up, resulting in a higher mean score among survivors, could have contributed to the increase in the nutritional score over time. However, sensitivity analysis excluding patients who died within the first 6 months on dialysis showed similar results to the original analysis. The composite nutritional score utilized in this study is amenable to automation and, particularly when expressed as a percentile in relation to a reference population, may have practical utility in nutritional competence assessment. However, before a recommendation can be made for the score to
be applied for nutritional interventions, it will be necessary to validate it further. In particular, it would be desirable to show that the score correlates with nutritional status as determined by reference methods (e.g., bioimpedance), and preferably also that it is responsive to dietary interventions. The presented score and similar methods are expected to afford opportunities to better segregate malnutrition from inflammation and be more proactive in addressing nutritional deficiency. The aggregation of several nutritional domains into a single score provides several potential advantages. First, when automated, a composite score is easier to track in clinical practice than multiple individual parameters. Second, similar to how combining medications with a common desired effect but different side effects can concentrate the desired effect while distributing the side effects, combining several parameters that have a common nutritional component but more dispersed other determinants into a composite score can increase the specificity. Third, the various components of the score represent different aspects of nutrition. For instance, a marker such as serum creatinine can be considered more reflective of the body’s nutritional “inventory,” whereas serum phosphate is arguably more reflective of current intake and appetite. As such, a composite score may provide a more comprehensive picture of nutritional status. A better understanding and more systematic and routine assessment of nutritional competence and resilience in chronic HD patients may improve outcomes by facilitating determination of rational, evidence-based criteria for nutritional supplement prescription and other nutritional interventions. Of note, we do not advocate the use of the presented nutritional score in its current form for prediction of mortality
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
Nutritional Patterns in Dialysis, Thijssen et al.
7
Figure 4. | Evolution of composite nutritional score around time of hospitalization. (A) Mean percentile of composite nutritional score over the 6 months before and after hospitalization. Error bars denote 95% confidence intervals. (B) Rate of change in percentile of composite nutritional score around time of hospitalization.
(or other outcomes). Some of the parameters used for our score may have U-shaped or other nonmonotonic associations with such outcomes, and simple addition with equal
weighting may underestimate the relationship between the individual components and the outcomes of interest. For visualization of trajectories of nutritional status, however, as
8
Clinical Journal of the American Society of Nephrology
Figure 5. | Evolution of composite nutritional score around time of hospitalization, stratified by number of rehospitalizations and by cause of hospital admission. (A) Mean percentile of composite nutritional score over the 6 months before and after initial hospitalization, stratified by number of rehospitalizations. (B) Mean percentile of composite nutritional score over the 6 months before and after hospitalization, stratified by the three most common causes of hospital admission (infectious disorders, cardiovascular/circulatory system disorders, and digestive system disorders). Only patients with no rehospitalizations in the 6 months after the initial hospitalization are included in this analysis. Error bars denote 95% confidence intervals in both plots.
desired for this article, our chosen method of aggregating the components appears justified, because nutritional status does arguably relate to these individual components in a monotonic fashion. In future studies, combining the dynamics of nutritional parameters with cross-sectional parameters can be applied toward development of risk prediction models to identify patients at high risk of hospitalization and mortality, as well as patients with low nutritional resilience who may benefit the most from timely and more aggressive nutritional intervention. However, the same validation requirements as mentioned above would apply. The effect of oral nutritional supplementation on nutritional patterns and associated
clinical outcomes also needs to be assessed in prospective trials. Acknowledgments This work was funded by the Renal Research Institute. Parts of this work were presented at the Annual Meeting of the American Society of Nephrology, held November 5–10, 2013, in Atlanta, Georgia. Disclosures S.T. and P.K. are employees of the Renal Research Institute, which is affiliated with Fresenius Medical Care North America. M.M.Y.W.
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
and Q.X. are former employees of the Renal Research Institute, and M.M.Y.W. is now a research consultant with the Renal Research Institute. L.A.U. and F.W.M. are employees of Fresenius Medical Care North America. P.K. and F.W.M. hold stock in Fresenius Medical Care North America. References 1. Heng AE,Cano NJM: Nutritional problems in adult patients with stage 5 chronic kidney disease on dialysis (both haemodialysis and peritoneal dialysis). NDT Plus 3: 109–117, 2010 2. Fouque D, Vennegoor M, ter Wee P, Wanner C, Basci A, Canaud B, Haage P, Konner K, Kooman J, Martin-Malo A, Pedrini L, Pizzarelli F, Tattersall J, Tordoir J, Vanholder R: EBPG guideline on nutrition. Nephrol Dial Transplant 22[Suppl 2]: ii45–ii87, 2007 3. Kidney Disease Improving Global Outcomes: Clinical practice guidelines for nutrition in chronic renal failure. K/DOQI, National Kidney Foundation. Am J Kidney Dis 35[Suppl 2]: S1–S140, 2000 4. Millikan WJ Jr, Henderson JM, Warren WD, Riepe SP, Davis RC, Hersh T, Wright-Bacon L, Long N, Kutner MH: Maintenance of nutritional competence after gastric partitioning for morbid obesity. Am J Surg 146: 619–625, 1983 5. Massad SG, Nieto FJ, Palta M, Smith M, Clark R, Thabet AA: Nutritional status of Palestinian preschoolers in the Gaza Strip: A cross-sectional study. BMC Public Health 12: 27, 2012 6. Kaysen GA: Serum albumin concentration in dialysis patients: Why does it remain resistant to therapy? Kidney Int Suppl 87: S92–S98, 2003 7. Gama-Axelsson T, Heimbu¨rger O, Stenvinkel P, Ba´ra´ny P, Lindholm B, Qureshi AR: Serum albumin as predictor of nutritional status in patients with ESRD. Clin J Am Soc Nephrol 7: 1446–1453, 2012 8. Mazairac AH, de Wit GA, Grooteman MP, Penne EL, van der Weerd NC, van den Dorpel MA, Nube´ MJ, Le´vesque R, Ter Wee PM, Bots ML, Blankestijn PJ; CONTRAST Investigators: A composite score of protein-energy nutritional status predicts mortality in haemodialysis patients no better than its individual components. Nephrol Dial Transplant 26: 1962–1967, 2011 9. Beberashvili I, Azar A, Sinuani I, Yasur H, Feldman L, Averbukh Z, Weissgarten J: Objective Score of Nutrition on Dialysis (OSND) as an alternative for the malnutrition-inflammation score in assessment of nutritional risk of haemodialysis patients. Nephrol Dial Transplant 25: 2662–2671, 2010 10. Kalantar-Zadeh K, Kopple JD, Block G, Humphreys MH: A malnutrition-inflammation score is correlated with morbidity and mortality in maintenance hemodialysis patients. Am J Kidney Dis 38: 1251–1263, 2001 11. R Core Team: R: A Language and Environment for Statistical Computing, Vienna, Austria, R Foundation for Statistical Computing, 2013 12. Nakazawa M: fmsb: Functions for Medical Statistics Book with Some Demographic Data. R Package Version 0.4.1., Vienna, Austria, R Foundation for Statistical Computing, 2013 13. Wickham H: ggplot2: Elegant Graphics for Data Analysis, New York, Springer, 2009
Nutritional Patterns in Dialysis, Thijssen et al.
9
14. Walther CP, Carter CW, Low CL, Williams P, Rifkin DE, Steiner RW, Ix JH: Interdialytic creatinine change versus predialysis creatinine as indicators of nutritional status in maintenance hemodialysis. Nephrol Dial Transplant 27: 771–776, 2012 15. Fouque D, Kalantar-Zadeh K, Kopple J, Cano N, Chauveau P, Cuppari L, Franch H, Guarnieri G, Ikizler TA, Kaysen G, Lindholm B, Massy Z, Mitch W, Pineda E, Stenvinkel P, Trevi~ noBecerra A, Wanner C: A proposed nomenclature and diagnostic criteria for protein-energy wasting in acute and chronic kidney disease. Kidney Int 73: 391–398, 2008 16. Patel SS, Molnar MZ, Tayek JA, Ix JH, Noori N, Benner D, Heymsfield S, Kopple JD, Kovesdy CP, Kalantar-Zadeh K: Serum creatinine as a marker of muscle mass in chronic kidney disease: results of a cross-sectional study and review of literature. J Cachexia Sarcopenia Muscle 4: 19–29, 2013 17. Mehrotra R, Berman N, Alistwani A, Kopple JD: Improvement of nutritional status after initiation of maintenance hemodialysis. Am J Kidney Dis 40: 133–142, 2002 18. Pupim LB, Kent P, Caglar K, Shyr Y, Hakim RM, Ikizler TA: Improvement in nutritional parameters after initiation of chronic hemodialysis. Am J Kidney Dis 40: 143–151, 2002 19. Usvyat LA, Haviv YS, Etter M, Kooman J, Marcelli D, Marelli C, Power A, Toffelmire T, Wang Y, Kotanko P: The MONitoring Dialysis Outcomes (MONDO) initiative. Blood Purif 35: 37–48, 2013 20. Usvyat LA, Barth C, Bayh I, Etter M, von Gersdorff GD, Grassmann A, Guinsburg AM, Lam M, Marcelli D, Marelli C, Scatizzi L, Schaller M, Tashman A, Toffelmire T, Thijssen S, Kooman JP, van der Sande FM, Levin NW, Wang Y, Kotanko P: Interdialytic weight gain, systolic blood pressure, serum albumin, and C-reactive protein levels change in chronic dialysis patients prior to death. Kidney Int 84: 149–157, 2013 21. English KL, Paddon-Jones D: Protecting muscle mass and function in older adults during bed rest. Curr Opin Clin Nutr Metab Care 13: 34–39, 2010 22. Kortebein P, Ferrando A, Lombeida J, Wolfe R, Evans WJ: Effect of 10 days of bed rest on skeletal muscle in healthy older adults. JAMA 297: 1772–1774, 2007 23. Phillips SM, Glover EI, Rennie MJ: Alterations of protein turnover underlying disuse atrophy in human skeletal muscle. J Appl Physiol (1985) 107: 645–654, 2009 24. Scheffer M, Bascompte J, Brock WA, Brovkin V, Carpenter SR, Dakos V, Held H, van Nes EH, Rietkerk M, Sugihara G: Earlywarning signals for critical transitions. Nature 461: 53–59, 2009 Received: August 25, 2014 Accepted: June 4, 2015 S.T. and M.M.Y.W. contributed equally to this work. Published online ahead of print. Publication date available at www. cjasn.org. This article contains supplemental material online at http://cjasn. asnjournals.org/lookup/suppl/doi:10.2215/CJN.08430814/-/ DCSupplemental.
Article
Nutritional Competence and Resilience among Hemodialysis Patients in the Setting of Dialysis Initiation and Hospitalization Stephan Thijssen,* Michelle M.Y. Wong,* Len A. Usvyat,† Qingqing Xiao,* Peter Kotanko,*‡ and Franklin W. Maddux†
Abstract Background and objectives Dialysis patients have a high risk for inadequate nutrition. Their nutritional status is particularly susceptible to deterioration when faced with intercurrent events such as hospitalization. This study was conducted to improve the understanding of the temporal evolution of nutritional parameters as a foundation for rational and proactive nutritional intervention. Design, setting, participants, & measurements A retrospective cohort study was performed to investigate the temporal evolution of nutritional parameters (serum albumin, serum phosphate, serum creatinine, equilibrated normalized protein catabolic rate, and interdialytic weight gain) and a composite nutritional score derived from these parameters, in two populations: (1) incident hemodialysis (HD) patients who started HD between January 2006 and December 2011 and were followed for up to 54 months (median 16.3), and (2) prevalent patients with HD vintage $2.5 years who were hospitalized between January 2006 and December 2011 and followed from 6 months before to 6 months after hospitalization. Results In incident patients (n=126,964), each of the nutritional parameters improved after HD initiation, with a mean composite nutritional score at the 24th percentile at the start of HD and reaching a plateau at the 57th percentile toward the end of the second year on dialysis. Nutritional parameters increased more rapidly and reached higher values among patients who survived longer. In hospitalized patients (n=14,193), the nutritional parameters and the composite score began to decline 1–2 months before hospitalization, reached their lowest level in the month after hospitalization, and then partially recovered in the subsequent 5 months. The degree of recovery of the nutritional score was inversely related to the number of rehospitalizations.
*Renal Research Institute, New York, New York; †Fresenius Medical Care North America, Waltham, Massachusetts; and ‡ Icahn School of Medicine at Mount Sinai Hospital, New York, New York Correspondence: Dr. Stephan Thijssen, Renal Research Institute, 315 East 62nd Street, Fourth Floor, New York, NY 10065. Email: [email protected]
Conclusions This study increases the understanding of nutritional resilience and its determinants in HD patients. Application of the nutritional score, pending further validation, may facilitate targeted and timely interventions to avert the negative consequences of inadequate nutrition in chronic HD patients. Clin J Am Soc Nephrol 10: ccc–ccc, 2015. doi: 10.2215/CJN.08430814
Introduction Inadequate nutrition is frequent in chronic hemodialysis (HD) patients, contributing to malnutrition and protein-energy wasting, which are major contributors to morbidity and mortality in this population (1). Assessing nutritional status in a HD patient is complex and often requires a multidisciplinary team. Clinical practice guidelines recommend that nutritional status be assessed using several tools, including dietary assessment, body mass index, subjective global assessment, anthropometry, protein nitrogen appearance, serum albumin, prealbumin, cholesterol, and technology such as bioimpedance (2,3). We use the term “nutritional competence” to refer to a patient’s nutritional status at the time of evaluation. A patient who is nutritionally competent possesses an adequate nutritional status at that particular time. Therefore, this is a cross-sectional assessment. The term “nutritional resilience” refers to a www.cjasn.org Vol 10 September, 2015
patient’s capacity to maintain nutritional competence in the face of a challenge (4,5). Thus, nutritional resilience is a dynamic measure that indicates a patient’s response to nutritional challenges, such as hospitalization. Eligibility for intradialytic nutritional supplements in Fresenius Medical Care North America (FMCNA) clinics is currently based on serum albumin levels. However, the use of serum albumin as a nutritional marker is problematic, because inflammation and fluid status can also affect its levels (6,7). The limitations of using albumin as the sole criterion for nutritional intervention call for improved strategies to assess nutritional competence and resilience. Most existing nutritional scores developed in the ESRD population are derived from cross-sectional data and do not consider dynamics of nutritional parameters (8–10). Therefore, a retrospective cohort study was performed in HD patients with the goals Copyright © 2015 by the American Society of Nephrology
1
2
Clinical Journal of the American Society of Nephrology
of (1) characterizing temporal patterns of nutritional parameters and (2) generating a composite nutritional score from readily available nutritional parameters and assessing its dynamics in incident HD patients over the first 2 years on dialysis and in prevalent patients around hospitalization.
Materials and Methods The study was approved by the New England Institutional Review Board (NEIRB number 14-109). We retrospectively studied HD patients aged $18 years treated in FMCNA clinics between January 1, 2006, and December 31, 2011. Two main analyses were conducted: The first assessed nutritional parameters in incident patients over the first 24, 48, and 54 months on HD, and the second assessed the association of hospitalization with temporal patterns of nutritional parameters in prevalent HD patients. Patients’ monthly averages of serum albumin, creatinine, phosphate, equilibrated normalized protein catabolic rate (enPCR), and interdialytic weight gain (IDWG) were obtained during the analysis periods indicated below. Analyses were performed using all records with complete data for the above five parameters for any given month. A sensitivity analysis was performed in patients with complete data for all months of interest. A composite nutritional score was calculated using serum albumin, creatinine, phosphate, enPCR, and IDWG as follows: For each of these nutritional parameters, the values were standardized by subtracting the mean of all values (for the respective parameter) across all patients and all months of interest from each value and then dividing by the SD of the centered values across all patients and months of interest. For each patient and month, the five standardized values were added and then again standardized as above across all months and patients, resulting in a distribution of composite nutritional scores with a mean of 0 and a SD of 1. These standardized composite score values were then converted to percentiles according to the probability density function of the standard normal distribution. Readers may apply the methodology to their own patients using the data from Table 1, which also gives an example. To graph the rate of change of the composite nutritional score, cubic B-spline basis functions were used to approximate the expectation of percentile of nutritional score. The function of the first derivative was approximated by finite differences. The composite nutritional score was deliberately created based on widely available variables, thus avoiding data acquisition beyond routine practice. Although IDWG is recorded at every dialysis session, FMCNA protocols require at least monthly measurements of the other parameters. Incident HD Patient Analyses Incident patients who started HD between January 1, 2006, and December 31, 2011 were included. Monthly mean serum albumin, creatinine, phosphate, enPCR, and IDWG values were graphed on radar plots at months 1, 3, 6, 9, 15, and 24. The radar plots were scaled such that the innermost point on each spoke represents the minimum average value observed for the respective parameter across all months,
Table 1. Mean and SDs for nutritional parameters and additive nutritional score in our data set
Parameter
Mean
SD
Serum albumin (g/dl) Serum creatinine (mg/dl) Serum phosphate (mg/dl) enPCR (g/kg per day) IDWG (liters) Additive score
3.8 7.3 5.2 0.89 2.57 0
0.4 2.9 1.5 0.27 1.11 3.14
The reader may use these data to calculate nutritional scores in his or her patients. For example, patient values are as follows: albumin 4.1 g/dl, creatinine 7.4 mg/dl, phosphate 5.1 mg/dl, enPCR 0.9 g/kg per day, and IDWG 2.8 L. The additive nutritional score is calculated as follows: (4.1 2 3.774)/0.433 + (7.4 2 7.254)/2.869 + (5.1 2 5.231)/1.5 + (0.9 2 0.885)/0.273 + (2.8 2 2.574)/1.107 = 0.976. The standardized additive score is calculated as follows: (0.976–0)/3.136 = 0.311, which converts to a percentile of 62.2. enPCR, equilibrated normalized protein catabolic rate; IDWG, interdialytic weight gain.
and the outermost point on each spoke represents the maximum average value. The composite nutritional score was tracked monthly over 2 years. A subgroup analysis with stratification according to survival time was performed in patients who either survived $48 months or died within the first 48 months. We conducted a subgroup analysis in patients without residual urea clearance at HD initiation. Hospitalization Analyses We included patients who were hospitalized for 7–14 days between January 1, 2006, and December 31, 2011 (regardless of when their first date of dialysis was) and who had been on HD for at least 2.5 years at the time of hospitalization. We further required patients to have data from 6 months before hospitalization to 6 months after discharge and no other hospitalizations in the 6 months before this index hospitalization. Only data from the first eligible hospitalization per patient were used. This hospitalization cohort and the incident patient cohort described above are separate cohorts with separate criteria (i.e., there are patients who contribute to one but not the other cohort); however, there is some degree of overlap between them. Serum albumin, creatinine, phosphate, enPCR, IDWG, and the composite nutritional score were plotted over 6 months before and after hospitalization. Subgroup analyses were performed with stratification for (1) the number of rehospitalizations over the 6 months after the initial hospitalization, and (2) the three most common categories of reason for hospitalization. For each month after the first hospitalization, statistical comparison across nutritional score groups was performed by the Kruskal–Wallis test, and we further tested for linear trend. Confidence intervals were obtained via bootstrapping with 1000 resampling iterations. Analyses were carried out using R software (version 3.0.2) (11). Radar graphs were created with the fmsb package (12); all other plots were created with ggplot2 (13).
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
Nutritional Patterns in Dialysis, Thijssen et al.
Results Baseline characteristics of the study populations are shown in Table 2. Averages and SDs of the nutritional parameters and additive score are presented in Table 1. Incident Patient Analyses There were 126,964 patients who had complete records for at least 1 month for all nutritional parameters during the first 24 months on HD. All five parameters increased over the first 2 years on HD (Figure 1A). The composite nutritional score improved from the 24th percentile at the start of HD to approximately the 57th percentile at 2 years (Figure 1B). The rate of improvement in the composite nutritional score was rapid after HD initiation and then quickly decelerated over the first 4 months, followed by a slower deceleration over the subsequent several months, with the score gradually approaching a plateau toward the end of year 2 (Figure 1C). Excluding patients who died within the first 6 months on dialysis yielded materially identical results (data not shown). For follow-up over 54 months, see Supplemental Figure 1. A subgroup analysis in patients with no residual urea clearance at HD initiation (59% of the initial cohort) showed materially identical results (Supplemental Figure 2). The nutritional score trajectories in incident patients stratified by survival time showed that both the initial rate of improvement of the nutritional score as well as its maximum level were directly related to patients’ survival times (Figure 2; for number of included records, see Supplemental Figure 3).
Table 2. Characteristics of the study populations
Characteristic
Number of patients Age (yr) Men Race Black Asian Other Diabetes Cardiovascular disease Cancer (excluding skin cancers) Hyperparathyroidism Chronic obstructive pulmonary disease Drug or alcohol dependence Gastrointestinal bleeding Infections
Incident Hemodialysis Patients
Hemodialysis Patients Who Were Hospitalized
145,847 63.1615.0 57.6
14,724 56.4614.7 51.7
30.5 2.3 67.2 61.6 55.6
47.3 1.5 51.2 61.0 67.7
6.9
6.3
4.3 10.9
9.8 12.1
3.0
3.6
1.2
2.6
10.2
20.4
Data are given as the mean6SD or percentage, unless otherwise indicated.
3
Hospitalization Analyses These analyses included 14,193 patients who had complete records for at least 1 month for all nutritional parameters. The nutritional parameters began to decline 1–2 months before hospitalization, reached their nadirs in the month after hospitalization, and then recovered to varying degrees in the 6 months after this initial hospitalization (Figure 3). This pattern is also reflected in the composite nutritional score (Figure 4A). The rate of decline in the nutritional score rapidly accelerated in the 2 months before hospitalization, and the score rebounded over the 3 months after hospitalization, albeit not to the original level (Figure 4B). Of all hospitalized patients, 8431 (59.4%) were rehospitalized within 6 months of the initial hospitalization. Analysis broken down by number of rehospitalizations during these 6 months revealed that the degree of recovery of the nutritional score showed a clear, inverse relationship to the number of rehospitalizations, with patients who were not rehospitalized experiencing a full recovery of their nutritional score (Figure 5A). At each of the 6 months after the initial hospitalization, comparison of mean percentiles across the five rehospitalization categories yielded highly significant results (P,0.001 for every month). Likewise, there was a highly significant “dose-response” relationship between number of rehospitalizations and percentile of the nutritional score at each month after the initial hospitalization (all P,0.001). The most common reasons for hospitalization were infections, circulatory system causes (including cardiovascular disease and stroke), and gastrointestinal system causes. Nutritional parameters showed a similar pattern of decline and recovery regardless of cause of hospitalization (Figure 5B). However, of the three groups, the gastrointestinal diseases group consistently had the lowest nutritional score throughout the 12-month period.
Discussion This large retrospective study, comprising close to 100,000 incident HD patients and .14,000 hospitalized patients, describes the dynamics of nutritional parameters after HD initiation and around hospitalizations. Although the five parameters examined are considered nutritional indicators, there are also caveats regarding their use as nutritional markers, which emphasizes the need to consider multiple parameters in nutritional assessment. Serum albumin is influenced by inflammation (6,7). IDWG is a surrogate for dietary intake but may also reflect fluid and sodium balance. The enPCR is a marker of steadystate protein intake (3). Whereas low enPCR reflects decreased dietary protein intake, very high enPCR could indicate tissue catabolism. Predialysis serum creatinine, an indicator of muscle mass (14), is also influenced by meat intake, residual renal function, and dialysis adequacy (15,16). The score calculated in this study, reflecting serum albumin, creatinine, phosphate, enPCR, and IDWG, indicates a relatively rapid improvement in these nutritional domains over the first few months on HD, followed by a gradual leveling off and eventually leading into a plateau toward the end of the second year. A subgroup
4
Clinical Journal of the American Society of Nephrology
Figure 1. | Evolution of nutritional parameters in incident hemodialysis patients. (A) Radar plots at months 1, 3, 6, 9, 15, and 24 from initiation of hemodialysis. Only patients with a complete set of data for all five shown variables for any given month were included (n=126,964). Each radar plot displays the five variables on axes (“spokes”) arranged radially. The mean value of each variable is plotted along the respective spoke; the innermost ring represents the minimum and the outermost ring represents the maximum value of that variable across all data points. A line connects the data values on each spoke. (B) Evolution of the composite nutritional score in patients starting hemodialysis. The mean percentile of nutritional score over 24 months after initiation of dialysis is shown. Error bars denote 95% confidence intervals. The number of available data records for each month is shown above the x axis along the bottom of the plot. (C) Rate of change in percentile of composite nutritional score in incident hemodialysis patients over the first 2 years on hemodialysis. Alb, serum albumin; Crea, serum creatinine; enPCR, equilibrated normalized protein catabolic rate; IDWG, interdialytic weight gain; P, serum phosphate.
analysis in patients without residual renal function at dialysis initiation showed materially identical results, indicating that the described trajectories are robust findings. Although all five analyzed parameters increased after initiation of HD, their rate of increase differed, with phosphate showing the fastest rise, followed by enPCR, IDWG, albumin, and creatinine. This seems biologically plausible, with phosphate being a rapidly responsive marker reflecting increased dietary intake. This is followed by enPCR, which might be expected to react somewhat slower because it may initially reflect tissue catabolism to some extent, which gradually decreases over time and counteracts the underlying increase in enPCR related to protein intake. It is plausible that albumin would take longer to recover than the other markers mentioned above, and that creatinine, being a reflection of muscle mass, would be the slowest parameter to recover. Improvement in nutritional parameters after initiating dialysis has been observed in previous studies in incident patients (17–19).
As shown in Figure 2, the nutritional score showed a decline in the months before death, a finding that has been demonstrated for several parameters in previous studies, including IDWG, serum phosphate, and serum albumin, which are components of our nutritional score (19,20). We found nutritional parameters to decline 1–2 months before hospitalization, reach their nadir right after hospitalization, and then improve again over the subsequent 5 months, although not to their original levels. Nutritional resilience refers to the degree to which a patient’s nutritional status is affected by intercurrent events (in this case, hospitalization). The nadir as well as the subsequent rate and degree of recovery of each individual parameter and the composite score characterize each patient’s nutritional resilience. Some patients are more susceptible to nutritional setbacks and protracted recovery than others (i.e., they possess a lower nutritional resilience). Of note, not all hospitalizations are comparable stressors to nutritional status. Similar to recovery after initiation of HD,
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
Nutritional Patterns in Dialysis, Thijssen et al.
5
Figure 2. | Time course of composite nutritional score after initiation of hemodialysis, stratified by patient survival time. The mean percentile of nutritional score is shown, with shaded areas representing the 95% confidence band of the fit.
recovery after hospitalization was not homogenous among the five parameters. Most notably, serum phosphate almost recovered to prehospitalization levels, whereas serum creatinine recovered poorly. Loss of lean tissue mass from acute inactivity with hospitalization can occur from decreased protein synthesis rate (so-called “anabolic resistance”) and elevated cortisol levels in the setting of inflammation (21–23). The incomplete recovery of creatinine may reflect the delay in regaining muscle mass, but protein intake likely also plays a role. In hospitalized HD patients, the composite nutritional score demonstrated the greatest magnitude of change in the 2 months before and after hospitalization, but even in patients with no rehospitalizations, the posthospitalization recovery was protracted over 4 months. Among patients who were rehospitalized more often, the nadir of the score was lower and recovery was less complete. The most common hospitalization causes demonstrated similar nutritional trajectories. Compared with other causes of hospitalization, the gastrointestinal diseases group had a lower nutritional score throughout the prehospitalization and posthospitalization periods, and this likely reflects increased susceptibility to dysfunctional nutrient intake, absorption, and/ or metabolism. It is tempting to speculate that the prolonged nutritional recovery time observed in rehospitalized patients is caused by the repeated detrimental effect of hospitalization on nutritional status. An additional potential explanation may be found in the behavior of complex dynamical systems (24): Such systems often exhibit characteristic properties that may serve as early warning signs for an impending critical transition. Dynamical systems theory describes a property called “critical slowing down,” whereby recovery from perturbations becomes increasingly slow. In this context, the recovery time in our nutritional score after
perturbation of the system by hospitalization may also be an indication of how close the patient’s biologic system is to a critical transition. As such, a patient’s nutritional resilience may be a direct measure of his or her health and impending complications. One might expect that an individual patient’s nutritional score recovery time after hospitalization (even without rehospitalization in the subsequent 6 months) may be indicative of downstream complications or mortality. Our study was not designed to analyze this aspect, but it would be a worthwhile investigation. We deliberately chose the 2.5-year vintage cutoff for our hospitalization analyses because the parameters of interest undergo, on average, marked changes in the 2 years after HD initiation. This approach limits the interference between these trajectories and our hospitalization analyses by allowing patients to stabilize after HD initiation before entering the hospitalization analyses. The effect of hospitalization on nutrition might vary in magnitude among patients of different dialysis vintage; however, our focus was on elucidating the general pattern and time course of the response, rather than characterizing the relationship between vintage and the effect of hospitalization or ensuring that the depicted response was the average of patients of all vintages. In addition, a shifting baseline before hospitalization would interfere with the interpretation of the response. Therefore, we felt that the benefit of reduced noise unrelated to hospitalization was the greater value for this particular analysis. A limitation of this study is its retrospective and observational nature. This design renders the study subject to potential bias owing to nonrandom data missingness. For example, in the incident patient analyses, patients who missed laboratory draws may have had worse nutritional
6
Clinical Journal of the American Society of Nephrology
Figure 3. | Time course of individual nutritional parameters around hospitalization. The bottom right plot shows the number of available records for each month. Dots denote the means. Error bars denote 95% confidence intervals. enPCR, equilibrated normalized protein catabolic rate; IDWG, interdialytic weight gain.
status than those who had data available. We addressed this concern by conducting sensitivity analyses that included only those patients who had complete data throughout the entire time period of interest. These analyses, in contrast with the shown analyses, likely included particularly healthy participants and therefore were subject to bias in the opposite direction. Yet these analyses showed materially identical results, which suggests that the presented time courses of the nutritional score are fairly robust across the HD population. For the analysis of hospitalized HD patients, no data were available during hospitalization; therefore, our analyses were limited to outpatient records before and after hospitalization. Therapy and changes in diet during hospitalization may have affected the parameters analyzed in this study. Another limitation is unavailability of data regarding dietary intake, body composition, socioeconomic status, and depression, preventing more in-depth analyses. In the incident patient analysis, changes in the mean composite nutritional score could have been affected by attrition from mortality. Patients with lower nutritional scores who died early in follow-up, resulting in a higher mean score among survivors, could have contributed to the increase in the nutritional score over time. However, sensitivity analysis excluding patients who died within the first 6 months on dialysis showed similar results to the original analysis. The composite nutritional score utilized in this study is amenable to automation and, particularly when expressed as a percentile in relation to a reference population, may have practical utility in nutritional competence assessment. However, before a recommendation can be made for the score to
be applied for nutritional interventions, it will be necessary to validate it further. In particular, it would be desirable to show that the score correlates with nutritional status as determined by reference methods (e.g., bioimpedance), and preferably also that it is responsive to dietary interventions. The presented score and similar methods are expected to afford opportunities to better segregate malnutrition from inflammation and be more proactive in addressing nutritional deficiency. The aggregation of several nutritional domains into a single score provides several potential advantages. First, when automated, a composite score is easier to track in clinical practice than multiple individual parameters. Second, similar to how combining medications with a common desired effect but different side effects can concentrate the desired effect while distributing the side effects, combining several parameters that have a common nutritional component but more dispersed other determinants into a composite score can increase the specificity. Third, the various components of the score represent different aspects of nutrition. For instance, a marker such as serum creatinine can be considered more reflective of the body’s nutritional “inventory,” whereas serum phosphate is arguably more reflective of current intake and appetite. As such, a composite score may provide a more comprehensive picture of nutritional status. A better understanding and more systematic and routine assessment of nutritional competence and resilience in chronic HD patients may improve outcomes by facilitating determination of rational, evidence-based criteria for nutritional supplement prescription and other nutritional interventions. Of note, we do not advocate the use of the presented nutritional score in its current form for prediction of mortality
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
Nutritional Patterns in Dialysis, Thijssen et al.
7
Figure 4. | Evolution of composite nutritional score around time of hospitalization. (A) Mean percentile of composite nutritional score over the 6 months before and after hospitalization. Error bars denote 95% confidence intervals. (B) Rate of change in percentile of composite nutritional score around time of hospitalization.
(or other outcomes). Some of the parameters used for our score may have U-shaped or other nonmonotonic associations with such outcomes, and simple addition with equal
weighting may underestimate the relationship between the individual components and the outcomes of interest. For visualization of trajectories of nutritional status, however, as
8
Clinical Journal of the American Society of Nephrology
Figure 5. | Evolution of composite nutritional score around time of hospitalization, stratified by number of rehospitalizations and by cause of hospital admission. (A) Mean percentile of composite nutritional score over the 6 months before and after initial hospitalization, stratified by number of rehospitalizations. (B) Mean percentile of composite nutritional score over the 6 months before and after hospitalization, stratified by the three most common causes of hospital admission (infectious disorders, cardiovascular/circulatory system disorders, and digestive system disorders). Only patients with no rehospitalizations in the 6 months after the initial hospitalization are included in this analysis. Error bars denote 95% confidence intervals in both plots.
desired for this article, our chosen method of aggregating the components appears justified, because nutritional status does arguably relate to these individual components in a monotonic fashion. In future studies, combining the dynamics of nutritional parameters with cross-sectional parameters can be applied toward development of risk prediction models to identify patients at high risk of hospitalization and mortality, as well as patients with low nutritional resilience who may benefit the most from timely and more aggressive nutritional intervention. However, the same validation requirements as mentioned above would apply. The effect of oral nutritional supplementation on nutritional patterns and associated
clinical outcomes also needs to be assessed in prospective trials. Acknowledgments This work was funded by the Renal Research Institute. Parts of this work were presented at the Annual Meeting of the American Society of Nephrology, held November 5–10, 2013, in Atlanta, Georgia. Disclosures S.T. and P.K. are employees of the Renal Research Institute, which is affiliated with Fresenius Medical Care North America. M.M.Y.W.
Clin J Am Soc Nephrol 10: ccc–ccc, September, 2015
and Q.X. are former employees of the Renal Research Institute, and M.M.Y.W. is now a research consultant with the Renal Research Institute. L.A.U. and F.W.M. are employees of Fresenius Medical Care North America. P.K. and F.W.M. hold stock in Fresenius Medical Care North America. References 1. Heng AE,Cano NJM: Nutritional problems in adult patients with stage 5 chronic kidney disease on dialysis (both haemodialysis and peritoneal dialysis). NDT Plus 3: 109–117, 2010 2. Fouque D, Vennegoor M, ter Wee P, Wanner C, Basci A, Canaud B, Haage P, Konner K, Kooman J, Martin-Malo A, Pedrini L, Pizzarelli F, Tattersall J, Tordoir J, Vanholder R: EBPG guideline on nutrition. Nephrol Dial Transplant 22[Suppl 2]: ii45–ii87, 2007 3. Kidney Disease Improving Global Outcomes: Clinical practice guidelines for nutrition in chronic renal failure. K/DOQI, National Kidney Foundation. Am J Kidney Dis 35[Suppl 2]: S1–S140, 2000 4. Millikan WJ Jr, Henderson JM, Warren WD, Riepe SP, Davis RC, Hersh T, Wright-Bacon L, Long N, Kutner MH: Maintenance of nutritional competence after gastric partitioning for morbid obesity. Am J Surg 146: 619–625, 1983 5. Massad SG, Nieto FJ, Palta M, Smith M, Clark R, Thabet AA: Nutritional status of Palestinian preschoolers in the Gaza Strip: A cross-sectional study. BMC Public Health 12: 27, 2012 6. Kaysen GA: Serum albumin concentration in dialysis patients: Why does it remain resistant to therapy? Kidney Int Suppl 87: S92–S98, 2003 7. Gama-Axelsson T, Heimbu¨rger O, Stenvinkel P, Ba´ra´ny P, Lindholm B, Qureshi AR: Serum albumin as predictor of nutritional status in patients with ESRD. Clin J Am Soc Nephrol 7: 1446–1453, 2012 8. Mazairac AH, de Wit GA, Grooteman MP, Penne EL, van der Weerd NC, van den Dorpel MA, Nube´ MJ, Le´vesque R, Ter Wee PM, Bots ML, Blankestijn PJ; CONTRAST Investigators: A composite score of protein-energy nutritional status predicts mortality in haemodialysis patients no better than its individual components. Nephrol Dial Transplant 26: 1962–1967, 2011 9. Beberashvili I, Azar A, Sinuani I, Yasur H, Feldman L, Averbukh Z, Weissgarten J: Objective Score of Nutrition on Dialysis (OSND) as an alternative for the malnutrition-inflammation score in assessment of nutritional risk of haemodialysis patients. Nephrol Dial Transplant 25: 2662–2671, 2010 10. Kalantar-Zadeh K, Kopple JD, Block G, Humphreys MH: A malnutrition-inflammation score is correlated with morbidity and mortality in maintenance hemodialysis patients. Am J Kidney Dis 38: 1251–1263, 2001 11. R Core Team: R: A Language and Environment for Statistical Computing, Vienna, Austria, R Foundation for Statistical Computing, 2013 12. Nakazawa M: fmsb: Functions for Medical Statistics Book with Some Demographic Data. R Package Version 0.4.1., Vienna, Austria, R Foundation for Statistical Computing, 2013 13. Wickham H: ggplot2: Elegant Graphics for Data Analysis, New York, Springer, 2009
Nutritional Patterns in Dialysis, Thijssen et al.
9
14. Walther CP, Carter CW, Low CL, Williams P, Rifkin DE, Steiner RW, Ix JH: Interdialytic creatinine change versus predialysis creatinine as indicators of nutritional status in maintenance hemodialysis. Nephrol Dial Transplant 27: 771–776, 2012 15. Fouque D, Kalantar-Zadeh K, Kopple J, Cano N, Chauveau P, Cuppari L, Franch H, Guarnieri G, Ikizler TA, Kaysen G, Lindholm B, Massy Z, Mitch W, Pineda E, Stenvinkel P, Trevi~ noBecerra A, Wanner C: A proposed nomenclature and diagnostic criteria for protein-energy wasting in acute and chronic kidney disease. Kidney Int 73: 391–398, 2008 16. Patel SS, Molnar MZ, Tayek JA, Ix JH, Noori N, Benner D, Heymsfield S, Kopple JD, Kovesdy CP, Kalantar-Zadeh K: Serum creatinine as a marker of muscle mass in chronic kidney disease: results of a cross-sectional study and review of literature. J Cachexia Sarcopenia Muscle 4: 19–29, 2013 17. Mehrotra R, Berman N, Alistwani A, Kopple JD: Improvement of nutritional status after initiation of maintenance hemodialysis. Am J Kidney Dis 40: 133–142, 2002 18. Pupim LB, Kent P, Caglar K, Shyr Y, Hakim RM, Ikizler TA: Improvement in nutritional parameters after initiation of chronic hemodialysis. Am J Kidney Dis 40: 143–151, 2002 19. Usvyat LA, Haviv YS, Etter M, Kooman J, Marcelli D, Marelli C, Power A, Toffelmire T, Wang Y, Kotanko P: The MONitoring Dialysis Outcomes (MONDO) initiative. Blood Purif 35: 37–48, 2013 20. Usvyat LA, Barth C, Bayh I, Etter M, von Gersdorff GD, Grassmann A, Guinsburg AM, Lam M, Marcelli D, Marelli C, Scatizzi L, Schaller M, Tashman A, Toffelmire T, Thijssen S, Kooman JP, van der Sande FM, Levin NW, Wang Y, Kotanko P: Interdialytic weight gain, systolic blood pressure, serum albumin, and C-reactive protein levels change in chronic dialysis patients prior to death. Kidney Int 84: 149–157, 2013 21. English KL, Paddon-Jones D: Protecting muscle mass and function in older adults during bed rest. Curr Opin Clin Nutr Metab Care 13: 34–39, 2010 22. Kortebein P, Ferrando A, Lombeida J, Wolfe R, Evans WJ: Effect of 10 days of bed rest on skeletal muscle in healthy older adults. JAMA 297: 1772–1774, 2007 23. Phillips SM, Glover EI, Rennie MJ: Alterations of protein turnover underlying disuse atrophy in human skeletal muscle. J Appl Physiol (1985) 107: 645–654, 2009 24. Scheffer M, Bascompte J, Brock WA, Brovkin V, Carpenter SR, Dakos V, Held H, van Nes EH, Rietkerk M, Sugihara G: Earlywarning signals for critical transitions. Nature 461: 53–59, 2009 Received: August 25, 2014 Accepted: June 4, 2015 S.T. and M.M.Y.W. contributed equally to this work. Published online ahead of print. Publication date available at www. cjasn.org. This article contains supplemental material online at http://cjasn. asnjournals.org/lookup/suppl/doi:10.2215/CJN.08430814/-/ DCSupplemental.
