Beruflich Dokumente
Kultur Dokumente
doi: 10.1093/ndt/gfq031
Advance Access publication 26 February 2010
1
Nephrology Division, Assaf Harofeh Medical Center, Zerifin, Affiliated to Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv,
Israel and 2Nutrition department, Assaf Harofeh Medical Center, Zerifin, Affiliated to Sackler Faculty of Medicine, Tel Aviv
University, Tel Aviv, Israel
Correspondence and offprint requests to: Ilia Beberashvili; E-mail: iliab@asaf.health.gov.il
BMI, body mass index; Wt, weight; Ht, height; TSF, triceps skinfold thickness; MAC, mid-arm circumference.
a
For haemodialysis patients also treated with statins, a cholesterol level <130 mg/dL (instead of 150 mg/dL) is scored as 0; cholesterol 130–180 mg/dL
is scored as 3 in such patients and cholesterol level above 180 mg/dL is scored as 6.
27 indicates moderate nutritional risk, and a score of <22 indicates an Hospitalization was defined as any hospital admission that included at
unsatisfactory nutritional status of haemodialysis patients. least one overnight stay in the hospital. The admission day was counted
as one full hospitalization day, but the discharge day was not. The sum of
all hospitalization days of a given patient during the study period was
Anthropometric measurements defined as the hospitalization days. The hospitalization frequency was
All measurements were made after dialysis when the patient was at dry the total number of hospital admissions during the same period irrespect-
weight (the right upper arm was used whenever possible, with exceptions ive of the length of each admission. Any access-related hospitalizations
made for patients in whom dialysis access placement, injury or stroke were not included in hospitalization data.
precluded measurement). The same dietitian performed all anthropomet-
ric measurements. BMI, TSF, MAC and calculated mid-arm muscle cir- Laboratory evaluation
cumference (MAMC) were measured as anthropometric variables. The
BMI was calculated as dry weight in kilograms divided by the square Blood samples were taken before starting a dialysis session. CBC, creatin-
of height in meters. TSF was measured with a conventional skinfold cali- ine, urea, albumin, transferrin and total cholesterol were measured by rou-
per using standard techniques. Mid-arm circumference was measured tine laboratory methods. IL-6 was measured in plasma samples using
with a plastic measuring tape. MAMC was estimated as follows: MAMC commercially available enzyme-linked immunosorbent assay (ELISA)
(cm) = mid-arm circumference (cm) − 0.314 × TSF (mm). kits (R&D System, Minneapolis, MN, USA) according to the manufac-
Anthropometric norms for dialysis patients were previously reported turer’s protocol. The mean minimal detectable dose (mean MDD) for IL-6
[6,23,24]. was 0.7 pg/mL.
Demography
Gender (male/female)a 65.4/34.6 66.7/33.3 64.6/35.4 66.7/33.3 NS –
Age (years) 64.3 ± 11.9 64.9 ± 11.3 63.2 ± 11.1 67.2 ± 14.9 NS –
a
Values are expressed as mean ± SD, median and range (for non-normally distributed variables) or percentage.
b
Adjusted for gender, age, DM status and history of cardiovascular disease.
c
Hospitalization days and frequency of hospitalization are 26-month prospective data.
d
Note that P-values are shown here for comparison, despite P above 0.05.
clinical and laboratory parameters were assessed using Pearson correl- tus (score from 28 to 32), moderate nutritional status
ation coefficients or Spearman rank order correlation coefficients, in (score from 23 to 27) and low nutritional status (score
the cases of skewed distribution of data. Alternatively, when distributions
<22). Over half of the patients (54.3%) had diabetes mel-
of variables were skewed, log transformation was undertaken before as-
sessment using Pearson correlation coefficients. Multivariate regression litus (DM), and nearly the same proportion had a history of
analysis was performed to obtain partial (adjusted) correlations (R2) con- cardiovascular disease (51.9%), including myocardial in-
trolled for age, gender, history of cardiovascular (CV) disease and pres- farction, coronary artery procedures such as angioplasty
ence or absence of diabetes mellitus. or surgery, cerebral-vascular accident or peripheral vascu-
Survival analyses were performed using the Kaplan–Meier survival
curve and the Cox proportional hazard model. The univariate and multivari- lar disease. Gender proportion, age, history of cardiovascu-
ate Cox regression analyses are presented as hazard ratio (HR; CI). lar disease and dialysis vintage distribution were similar in
All statistical tests were two-sided, with a value for P < 0.05 defining the three groups. The body mass index was lower across
significance. decreasing OSND categories (P = 0.001).
All statistical analyses were performed using SPSS software, version Hospitalization days and hospitalization frequency in-
16.0 (SPSS Inc, Chicago, IL).
creased across worsening OSND categories (P = 0.014
and P = 0.022, respectively). Anthropometric measure-
Results ments (TSF, MAC and MAMC) as well as body compos-
ition parameters derived by BIA [fat mass (FM), lean body
In this study, we developed and characterized a score for mass (LBM) and FFM] showed statistically significant dif-
the assessment of nutritional status in dialysis patients, ferences between the three OSND groups. Phase angle
based solely on objectively measurable criteria. In our co- (PA)—the BIA prognostic index of morbidity and mortal-
hort, 81 prevalent HD patients (53 men and 28 women) ity [describing the relationship between the two vector
with a mean age 64.3 ± 11.9 years were studied. Table 2 components of impedance (reactance and resistance) of
shows the pertinent demographic, laboratory and clinical the human body to an alternating electric current] was sig-
data of these patients as a whole and for each nutritional nificantly lower in worsening OSND categories (P =
category, as determined by OSND: normal nutritional sta- 0.006). Comparison between different OSND groups ex-
2666 I. Beberashvili et al.
Table 3. Correlation coefficients between Objective Score of Nutrition on
Dialysis (OSND) and nutritional parameters (that not included in the
OSND score), demographic and hospitalization data of study population
In each column, the first set of values includes the unadjusted (bivariate) r
with its Pearson or Spearman (for skewed data) P-value, and the second
set (in parentheses) includes partial correlation based on a multivariate
regression analysis adjusted for case-mix (age, gender, diabetes status
and history of CV disease).
Correlation coefficient values ≥0.25 appear in bold. BMI, body mass
index; TSF, triceps skinfold thickness; MAC, mid-arm circumference;
MAMC, mid-arm muscle circumference calculated; IL-6, interleukin-6;
FFM, fat-free mass.
Variable Units of decrease (↓) or increase (↑) Hazard ratio and 95% CI P-value
OSND, objective score of nutrition on dialysis; MIS, malnutrition-inflammation score; DM, diabetes mellitus; CV, cardiovascular; IL-6, interleukin-6.
model controls for age, gender, diabetes status, IL-6 and lar accident (CVA) or peripheral vascular disease (PVD)
history of cardiovascular disease at baseline to estimate requiring angioplasty, bypass or amputation and diagnosed
relative risks. The OSND showed the strong association after participants entered the study. Unfavorable OSND de-
with prospective mortality. The relative risk for death for monstrated statistically significant hazard ratios for first
each five-unit decrease in OSND was 2.8 (95% CI, 1.3 to CVA (HR = 3.5; 95% CI, 1.1 to 10.8; P = 0.028) and trend
6.2; P = 0.009). In this regard, OSND was comparable with
MIS and phase angle. The relative risk for death for each
five-unit increase in MIS was 2.2 (95% CI, 1.2 to 4.1; P =
0.013) and for each 1-unit decrease in PA was 3.7 (95% CI, Table 5. Case-mix-adjusted (for age, gender, DM status and history of
CV disease) hazard ratios of death for all components of the OSND, as
1.7 to 7.9; P = 0.001). well as the OSND itself, MIS and phase angle
Table 5 lists case-mix-adjusted hazard ratios of death for
each of the seven components of the OSND as well as the OSND components Hazard ratio (95% CI) P-value
composite OSND itself. Each OSND component was en-
tered in the Cox model as a decremental variable, consisting 1. BMI (kg/m2) 1.1 (1.0–1.2)* 0.044
of a number between 1 and 3. The MIS was divided into 2. Weight decrease (%) 1.2 (0.3–3.9) 0.811
three equal increments, and phase angle was divided into 3. TSF (mm) 1.1 (0.7–1.9) 0.669
4. MAC (cm) 1.3 (0.7–2.7) 0.424
three equal decrements as well. Three OSND components 5. Albumin (g/L) 2.2 (1.3–3.8)* 0.005
showed statistically significant hazard ratios for death: 6. Transferrin (mg/dl) 1.1 (0.7–1.9) 0.622
BMI, serum albumin and total cholesterol levels had the 7. Cholesterol (mg/dl) 2.4 (1.3–4.5)* 0.008
strongest association with mortality. However, the OSND OSND 3.6 (1.6–7.9)* 0.002
MIS 4.3 (1.6–11.7)* 0.004
was found to be a more powerful predictor of mortality than Phase angle 4.1 (1.8–9.6)* 0.001
any of its eight components (Table 5). Moreover, in this
model, OSND had as high accuracy for predicting death Each component is scored as a number between 1 and 3. Note that OSND,
(HR = 3.6; 95% CI, 1.6 to 7.9; P = 0.002) as MIS a number between 5 and 32, is divided into three equal groups (32 to 24,
(HR = 4.3; 95% CI, 1.6 to 11.7; P = 0.004) or phase angle 23 to 15 and 14 to 5). MIS, a number between 0 and 30, also is divided
(HR = 4.1; 95% CI, 1.8 to 9.6; P = 0.001). into three equal increments (0 to 10, 11 to 20, and 21 to 30). Phase angle
is divided into three decrements (7.9 to 6.0; 5.9 to 4.0 and 3.9 to 2.0).
Case-mix-adjusted hazard ratios of first cardiovascular BMI, anthropometric indexes, albumin level, transferrin level and choles-
event for OSND, MIS and phase angle for a one-unit in- terol level were each divided into three groups, in a decrement fashion and
crease within the three decrements for OSND and phase expressed as integer numbers between 1 and 3. This enabled comparison
angle and within three increments for MIS (as described of the hazard ratios of the OSND, MIS and phase angle with those of
components of OSND, as well as each with another.
above) were calculated: first cardiovascular event defined OSND, objective score of nutrition on dialysis; MIS, malnutrition-inflam-
as myocardial infarction (MI) requiring coronary artery mation score; TSF, triceps skinfold thickness; MAC, mid-arm circumfer-
procedures such as angioplasty or surgery; cerebral-vascu- ence*P < 0.05.
2668 I. Beberashvili et al.
Table 6. Adjusted multivariate correlation coefficients
Variables Hazard ratio of death Hospitalization frequency Hospitalization days Phase angle
R2 and P for hazard ratios of death, hospitalization frequency, hospitalization days and phase angle derived by BIA, compared to anthropometry (in-
dicated by ‘a’), which consists of the first five components of the OSND, and its modified versions through the stepwise evolution toward the full
version of the OSND by adding three laboratory tests (albumin, transferrin and cholesterol). The 13th row is based on replacing transferrin with
creatinine in the OSND. The case-mix-adjusted correlation coefficients R2 are controlled for age, gender, diabetes status and history of CV disease.
For each multivariate R2, P is listed in parentheses.
BMI, body mass index; TSF, triceps skinfold thickness; MAC, mid-arm circumference; MAMC, mid-arm muscle circumference calculated; ΔW, de-
crease in dry weight (in past 3–6 months).
for first composite cardiovascular event (HR = 2.1; 95% CI, demonstrating a reasonable predictive value for the score.
1.0 to 4.5, P = 0.056). MIS was predictive only for first CVA MIS and phase angle produced very similar AUC [0.678
(HR = 4.8; 95% CI, 1.3 to 17.8; P = 0.019), whereas phase (P = 0.015) and 0.739 (P = 0.001), respectively].
angle had significant associations with first CVA (HR = 4.0;
95% CI, 1.2 to 13.8, P = 0.027), first PVD (HR = 4.0; 95%
CI, 1.2 to 13.4, P = 0.025) and first composite cardiovascular Discussion
event (HR = 3.4; 95% CI, 1.5 to 7.5; P = 0.003).
Table 6 demonstrates the stepwise enhancement of the The purpose of this study was to develop and evaluate a
OSND model based on the addition of serum albumin, scoring system consisting of routine objective measure-
cholesterol and transferrin levels to the anthropometric ments for identifying nutritional risk in HD patients. Such
measures. The first column represents pseudo-R2 based a quantitative and comprehensive scoring system, named
on a Cox proportional hazard model [the case-mix-ad- the Objective Score of Nutrition on Dialysis, was developed
justed (multivariate) correlation coefficients R2 for hazard by combining anthropometric measurements with three la-
ratios for death]. The following columns show multivariate boratory analyses: serum albumin, transferrin and choles-
R2s for prospective hospitalization indices (hospitalization terol levels. As this nutritional score uses only objective
days and hospitalization frequencies) and phase angle information, any subjective assessments and judgments by
based on multivariate linear regression models. The first the examiner can therefore be avoided.
four rows show correlations between hazard ratios for A high proportion of HD patients in this study showed
death, hospitalization indices and phase angle with the signs of malnutrition, which was mild to moderate in the
two anthropometric measures (TSF and MAC), decrease majority of patients. HD patients with malnutrition deter-
in dry weight and BMI, separately or combined together mined by OSND had a significantly lower body mass index,
(fourth row). Although, as single parameters, these mea- serum albumin, total cholesterol and transferrin levels, as
surements did not exhibit correlations with clinical out- well lean body mass, fat mass and FFM determined by
come, they correlated significantly with hazard ratios for BIA. Our report also found a significant difference in an-
death, hospitalization days and phase angle when com- thropometry and MIS scores. The prevalence of malnutri-
bined. The following rows show various combinations to tion within our study (about 77%) population is somewhat
model a scoring system based on the addition of one or higher than that found by other studies (23% to76%) per-
more of the three laboratory tests (albumin, transferrin, formed in patients with ESRD [5,28–30]. A possible ex-
cholesterol) to the combination of anthropometric mea- planation for differences between our results and those
sures. The 12th row is the complete version of the OSND. reported may relate to differences in the baseline demo-
The 13th row represents a scoring model based on re- graphic and some clinical characteristics of HD patients
placing transferrin level by creatinine level. Additional studied: our cohort was older, with longer dialysis vintage
models were also constructed and evaluated, but they did and with higher prevalence of diabetes. It is well known that
not improve correlations with parameters that reflect clin- older age [29], long dialysis vintage [31] and presence of
ical outcome (data not shown). diabetes mellitus [32,33] may influence the nutritional sta-
The ROC curves for predicting mortality are shown in tus of HD patients either by causing a reduced nutritional
Figure 2. The AUC for the OSND is 0.707 (P = 0.004), intake or by promoting catabolism.
Nutritional risk assessment of haemodialysis patients 2669
vated serum albumin and BMI values were independently
associated with decreased mortality. In our study, we found
a strong association between BMI, albumin and choles-
terol with clinical outcome. It is well known that a high
body mass index is paradoxically protective and associated
with improved survival in chronic haemodialysis patients
[35,36]. Underweight itself presents an important clinical
sign of malnutrition that worsens the prognosis of chronic