Creatinine index and transthyretin as additive predictors of mortality in haemodialysis patients.

BACKGROUND
Malnutrition and inflammation are recognized as important predictors of poor clinical outcome in haemodialysis (HD). This study was designed to estimate the relative contribution of known biological markers of inflammation, malnutrition and muscle mass in the prognosis of HD patients.


METHODS
A total of 187 HD patients (100 women, 87 men, median age 66.7 years [22.3-93.5]) were followed-up yearly for 5 years. At baseline, pre-dialysis values of C-reactive protein (CRP), albumin, transthyretin, total HDL- and LDL-cholesterol and triacylglycerol were determined. Estimation of creatinine index (CI) as muscle mass marker was determined by creatinine kinetic modelling using pre- and post-dialysis creatinine values.


RESULTS
During the follow-up period, 89 deaths (53 from cardiovascular causes) were observed. After adjustment for age, gender, dialysis vintage, smoking, diabetes mellitus and hypertension, the highest tertile of CRP and lowest tertile of transthyretin and CI were significantly associated with all-cause mortality (relative risk (RR)=1.98 [1.12-3.47], 2.58 [1.48-4.50], 2.71 [1.42-5.19], respectively). In addition, low CI had an additive value to low levels of transthyretin. In contrast, high cholesterol (RR=0.47 [0.27-0.83], P=0.0091) and vitamin E concentrations (RR=0.46 [0.26-0.80], P= 0.006) showed a protective trend for all-cause mortality. In the multivariate analysis, transthyretin appeared as the most predictive biological marker of non-CV mortality (RR=3.78 [1.30-10.96], P=0.014), and CI of CV mortality (RR=2.61 [1.06-6.46], P=0.038), respectively. Discussion. These results confirm that uraemic malnutrition constitutes an important risk factor for mortality in HD. Beyond transthyretin, CI seems to be an additional marker routinely available and monthly determined in HD patients.


Introduction
In spite of a clear improvement in dialysis technologies during the last years, the death rate among uremic patients treated by hemodialysis (HD) has always been rather high compared to the general population. It is mostly due to cardiovascular (CV) diseases [1,2]. Nutritional status has long been recognized as an important predictor of poor clinical outcome in these patients [3,4]. Malnutrition in uraemia appears as a multifactorial process [5] mainly due to energetic metabolism impairment, enhancement of catabolic rate and finally inflammation [2]. Chronic kidney disease patients also spontaneously restrict their dietary protein intake while uraemia by itself is a net catabolic status. Muscle metabolism alteration should be considered as the consequence of the uremic proteino-energetic disorders aggravated by inflammation. In fact, it has been shown that muscle amino acid transport alterations could be prevented by increasing amino acid delivery via extracorporal circuit. Similarly, perdialytic nutritional support could partially prevent HD related malnutrition including muscle mass catabolism [11]. In addition, chronic inflammation state results in an enhancement of muscle catabolism and release of amino acids [8,12]. The 4 HEMO study has recently stressed the importance of muscle mass as determined by anthropometric data in clinical outcome [13]. However, exploring muscle metabolism requires specialized methods involving radioisotopes [7].
Alternatively, it has been shown that creatinine modelling could be a simple non invasive and clinically relevant method to assess muscle mass [14]. In a previous work [15], we reported a clear association between the prevalence of CV disease and end-stage renal disease (ESRD)-associated metabolism disorders such as inflammation and undernutrition, using creatinine-derived estimations of Lean Body Mass (LBM) and other markers such as transthyretin. The present study was designed to assess the prognostic value of these markers on CV and non CV mortality.

Study design
192 stable HD patients recruited in one of the three dialysis facilities of Montpellier (France) (a hospital-based facility (Lapeyronie University Hospital), a public non-profit association (AIDER) and a private dialysis clinic (CHLM)) were evaluated for inclusion from October to November 2001. Informed consent was obtained from all participants.
Patients received either standard HD (n=166) or haemodiafiltration (HDF) treatments with on-line ultra-pure bicarbonate-based dialysate (n= 26). All patients were dialysed with low (n=109) or high-flux (n=83) polysulfone membranes. Patients with symptoms or signs of acute inflammatory or infectious diseases were excluded from the study.
The included patients were then prospectively followed up yearly until January 5th, 2007.

5
Baseline data Medical charts were reviewed for age, gender, weight, height, underlying renal disease, dialysis vintage, history of transplantation, diabetes mellitus, current hypertension, co-morbid conditions, past or current smoking and current medication. Existence of hypertension was defined by pre-dialysis blood pressure higher or equal to 140/90mmHg and/or by a current antihypertensive treatment.
The efficiency of dialysis was estimated by calculation of Kt/V (K, clearance of urea of the dialyser, t, time of dialysis and V, volume of purified urea) [16].

Laboratory analysis and procedures
Pre and post dialysis blood samples were collected before and after a mid week dialysis session according to the best practices applied for dialysis adequacy evaluation [16]. Blood samples were centrifuged, treated, analysed for routine parameters and finally stored at -80°C, in order to perform additional analyses. by HPLC method as previously described [17]. Results were normalized and expressed as VitE/(chol+TG) (µmol/mmol ratio).

Measurement of nutritional indices
Body Mass Index (BMI) was obtained from height and post dialysis body weight according to the formula BMI = Weight (kg)/ Height (m 2 ). The creatinine index (CI) was computed from creatinine kinetic modelling as described previously. Briefly, CI was deducted from creatinine generation rate using a single pool variable volume model [14,18]  Descriptive statistics are presented as percentages for categorical variables, as means with s.e.m. for normally distributed variables and as medians with ranges for non-normally distributed variables.
The Kaplan-Meier method was used to describe survival curves. The Cox proportional hazards model was used to identify predictors of mortality. In this manner, deaths were analysed in the 3 following groups : all causes, CV and non CV mortality. Continuous variables were divided into tertiles. Association of deaths with non traditional risk factors was estimated with Relative Risks (RR) and their 95% confidence intervals (95% CI) adjusted for age, gender, dialysis vintage, diabetes mellitus, smoking and hypertension. According to the literature, these last three factors are traditional risks factors for mortality in HD patients.
Significance was set at p<0.05. Five patients (2.6%) had some missing data (5 values of HDL cholesterol because of triacylglycerols values higher than 5 g/l,), leaving 187 patients for the statistical analysis.
Statistical analyses were performed using the SAS software, version 9.1 (SAS Institute, Cary, NC°, USA). 89 deaths occurred during the five year-follow-up, corresponding to a 9.5% annual mortality rate. 53 (59.6%) of these deaths had a cardiovascular cause and 36 (40.4%) a non cardiovascular cause (14 infections, 7 neoplasm, 7 cachexia and 8 miscellaneous causes).

Conventional markers of nutrition and inflammation
As reported in Table 1 Figure 1.
Finally, neither BMI nor serum albumin were predictive factors of mortality.

Lipid parameters as outcome predictors
Comparing with the first tertile, high total and LDL cholesterol concentrations were significantly associated with reduced all-cause mortality ( Figure 2.
Finally, since CI appears as a predictive marker assessing somatic protein metabolism and malnutrition inflammation complex syndrome severity, we explored the potential additive association between CI and transthyretin. As 10 shown in Figure 3, transthyretin and CI are clearly additive in predicting poor outcome in HD patients in all cause as well as in cardiovascular mortality.

Discussion
This prospective study confirms the high annual mortality rate (9.5 %) observed in HD population [1]. Our results confirm previous observations showing that transthyretin is an independent determinant of mortality in HD patients. In addition, we showed in this study that muscle mass determination by creatinine index derived from creatinine modelling has an additive value to transthyretin in all cause and cardiovascular mortality.
Assessment of protein and energy nutritional status can be achieved by determination of visceral and somatic proteins in addition to measurement of energy balance [19]. Visceral protein stores can be determined easily by measurement of biochemical circulating markers such as serum albumin or transthyretin. Usually, studies define serum albumin as the most reliable indicator of nutritional status [20] despite limitations due to the cytokine induced increase in fractional synthesis rate [19] potentially hiding hypoalbuminemia. On the other hand, hypoalbuminemia could be due to non nutritional causes such as increased losses through gastrointestinal tract or volume perturbations. Previous studies have underlined that transthyretin, a negative acute phase protein known as a complex transporter of thyroxine, retinol binding protein, vitamin A, showed higher correlation coefficients with nutritional indices than albumin and appeared to be better and quickly reflects nutritional status changes [21]. Indeed, transthyretin levels, but not albumin levels were correlated with arm muscle circumference, triceps skin fold [21,22], or LBM [23]. Recently Chertow et al. [24] have reported that transthyretin levels were directly related to nutritional parameters and dietary intake such as body weight, predialysis blood urea nitrogen, creatinine, albumin, phosphorus and bicarbonate. In addition, transthyretin levels, but not albumin, are predictive of cardiovascular hospitalization. According to these data, monthly transthyretin measurement is recommended by both American and European consensus group in the nutritional assessment of dialysis patients [25,26]. In the present study, both univariate and multivariate analyses underline transthyretin as a more powerful predictor of all cause mortality in HD population [27] compared to albumin. In fact, in this long term follow-up study, the lowest tertile value of albumin did not appear as a significant predictor of mortality. This result, in apparent contradiction with the K/DOQI guidelines or DOPPS observational study [28,29] could be due to the relative low size of our population. On the other hand, it should be underlined that in our population the lowest tertile value was 35.48 g/l, a threshold value which is above risk. In comparison, the median value observed for albumin in the HEMO Study was 36.3 g/l and the lowest values of albumin, corresponding to a clearly enhanced mortality risk, were 26 -32 g/l [13]. Low albumin levels are rarely observed in our population (only 27 patients had albumin levels lower than 32 g/l). In addition, in the same study, it was shown that low albumin was a predictive indicator on short term but that the predictive RR diminished with extending follow up period (> 6 months). Taken together, these data suggest that in our population with a high dialysis dose (averaged Kt/V of 1.47+0.02), with less than 1/3 of patient lower than the nutrition K DOQI recommandations, transthyretin appears as a better long term predictor of mortality than albumin. The significant involvement of transthyretin in long term poor outcome of HD patients is confirmed with the multivariate analysis ( Table 2) While serum markers reflect visceral proteins, somatic proteins are a determinant of muscle mass [4]. Therefore, the assessment of somatic protein status appears of crucial clinical relevance and is commonly used for assessment of nutritional 13 status in ESRD patients [19]. Several techniques are used to determine LBM, such as anthropometric measurements, DEXA (Dual Energy X-ray Absorptiometry) or isotopic determination with [ 3 H]-leucine [7], which are expensive and difficult to implement in clinical practice.
By contrast, creatinine kinetic, which is based on the principle that creatinine production is proportional to LBM and represents the sum of creatinine excretion (urinary and dialytic) and metabolic degradation, is a simple and a reliable tool for the assessment of protein nutritional status and muscle mass in HD patients which can be easily coupled to urea kinetics modelling [18,30,31]. Such indices derived from creatinine kinetic (CI, observed LBM, observed/expected LBM) have been identified as prognostic markers of mortality in HD patients [14,32], but none of these studies could establish the link between creatinine metabolism and CV risk. Our study adjusted for traditional risk factors established a clear relationship between CI and all cause, CV and non CV mortality. This observation is in agreement with the previous observation of the HEMO study [13] showing that low predialysis creatinine value is an index of poor prognosis. Indeed, CI is directly linked to creatinine production, thus to muscle mass, and is not influenced by dialysis efficacy. The multivariate analysis confirmed the powerful relationship of CI among biological somatic protein markers with HD mortality, particularly when the cause of death was CV disease (RR=2.61, [1.06-6.46], p=0.038, after full adjustment). Our results highlight two interesting facts.
Firstly, the predictive value of both creatinine index and tranthyretin are maintained over time from two to five years of follow up (Figure 4). Secondly, multivariate analysis shows that CI and transthyretin are independent, and therefore additive risk factors for all cause and also for cardiovascular mortality (Figure 3). Sarcopenia in stable HD patients is another expression of the malnutrition inflammation complex syndrom that should be defected early in HD patients since it is associated with a high mortality risk. The close association 14 between inflammation and malnutrition [10, 12] has been highlighted during the past decade. It has also been suggested that inflammation is an underlying In conclusion, in the present study, mortality in HD patients was associated with traditional risk factors such as age, smoking and hypertension whereas an