A Low, Rather than a High, Total Plasma Homocysteine Is an Indicator of Poor Outcome in Hemodialysis Patients

. An increased level of total plasma homocysteine (tHcy) is a risk factor for poor cardiovascular outcome in the general population. However, a decreased, rather than an increased, tHcy concentration may predict poor outcome in maintenance hemodialysis (MHD) patients, a phenomenon referred to as reverse epidemiology. Associations were examined between tHcy level and markers of malnutrition-inflammation complex syndrome and 12-mo prospective hospitalization and mortality in 367 MHD patients, aged 54.5 (cid:1) 14.7 (mean (cid:1) SD) years, who included 199 men and 55% individuals with diabetes. tHcy was 24.4 (cid:1) 11.8 (cid:1) mol/L, and 94% of the patients had hyperhomocysteinemia (tHcy (cid:2) 13.5 (cid:1) mol/L). tHcy had weak to moderate but statistically significant bivariate and multivariate correlations with some laboratory markers of nutrition (serum albumin, prealbumin, creatinine, and urea nitrogen) but no significant correlation with serum C-reactive protein or two proinflammatory cytokines (IL-6 and TNF-(cid:2) ).

(tHcy) is a risk factor for poor cardiovascular outcome in the general population.However, a decreased, rather than an increased, tHcy concentration may predict poor outcome in maintenance hemodialysis (MHD) patients, a phenomenon referred to as reverse epidemiology.Associations were examined between tHcy level and markers of malnutrition-inflammation complex syndrome and 12-mo prospective hospitalization and mortality in 367 MHD patients, aged 54.5 Ϯ 14.7 (mean Ϯ SD) years, who included 199 men and 55% individuals with diabetes.tHcy was 24.4 Ϯ 11.8 mol/L, and 94% of the patients had hyperhomocysteinemia (tHcy Ͼ13.5 mol/L).tHcy had weak to moderate but statistically significant bivariate and multivariate correlations with some laboratory markers of nutrition (serum albumin, prealbumin, creatinine, and urea nitrogen) but no significant correlation with serum C-reactive protein or two proinflammatory cytokines (IL-6 and TNF-␣).
During 12 mo of follow-up, 191 MHD patients were hospitalized, 37 died, nine underwent renal transplantation, and 38 transferred out.Hospitalization rates were significantly higher in patients with lower tHcy levels.Mortality rate in the lowest tHcy quartile (17.4%) was significantly higher compared with other three quartiles (6.5 to 9.8%; Kaplan-Meier P ϭ 0.04).Relative risk of death for the lowest tHcy quartile, even after adjustment for case-mix and serum albumin, was 2.27 (95% confidence interval, 1.14 to 4.53; P ϭ 0.02).Hence, tHcy may be a more exclusive nutritional marker in MHD patients with no association with inflammatory measures.Despite a very high prevalence of hyperhomocysteinemia in MHD patients, lower values of tHcy are paradoxically associated with increased hospitalization and mortality.The lowest tHcy quartile confers a twofold increase in risk of death independent of hypoalbuminemia.The nutritional feature of tHcy in MHD patients may explain its reverse association with outcome.
In the general population, an increased level of total plasma homocysteine (tHcy) is a risk factor for increased cardiovascular events and mortality (1)(2)(3)(4).In maintenance hemodialysis (MHD) outpatients, however, the association between tHcy and clinical outcome is inconsistent and even paradoxical.Some studies have shown a poor outcome in MHD patients with hyperhomocysteinemia (5)(6)(7)(8)(9).However, a few recent studies have suggested that a decreased, not an increased, tHcy concentration is related to a higher prevalence of cardiovascular disease and poor outcome in these individuals (10 -12).This seems to be due to the empirical association between a low tHcy and protein-energy malnutrition, which is per se a known risk factor for poor clinical outcome in dialysis patients (11,13,14).This recently described paradoxical association between tHcy and clinical outcome in dialysis patients has now been referred to as a possible component of the reversal of the cardiovascular risks (15)(16)(17).It is believed that both inflammation and protein-energy malnutrition, each independently or together as the "malnutrition-inflammation complex syndrome" (15,18,19), are the main contributors of the abovementioned risk factor reversal phenomenon.Nevertheless, it is not clear whether there is any significant association between tHcy and inflammation in dialysis patients.We studied the hypotheses that a low level of tHcy in dialysis patients indicates protein-energy malnutrition, with or without inflammation, and is associated with poor clinical outcome in maintenance dialysis patients.To verify the above hypotheses, we first examined baseline associations between tHcy and some markers of protein-energy malnutrition and inflammation at the start of the cohort of the Nutritional and Inflammatory Evaluation in Dialysis (NIED) Study (20,21).We then explored longitudinal associations between the baseline tHcy and the prospective mortality and hospitalization measures in the same group of patients.

Patients
Participants in the NIED Study originated from a pool of approximately 1200 MHD outpatients in eight DaVita, Inc., dialysis facilities in the South Bay Los Angeles area (see NIED Study web site at www.NIEDStudy.org for more details, as well as previous publications (20,21)).Inclusion criteria were outpatients who had been undergoing MHD for at least 8 wk, were 18 yr or older, and signed a written consent form.Patients with an anticipated life expectancy of Ͻ6 mo (e.g., because of a metastatic malignancy or terminal HIV disease) were excluded.In the initial phase of the NIED Study (October 2001 to March 2002), 385 patients from eight dialysis units signed the written consent form.Subsequently, blood samples were obtained from 367 of these individuals, because 18 patients were not present in the dialysis units at the time of blood drawing.The medical chart of each MHD patient was thoroughly reviewed by a nephrologist (K.K.Z.), and data pertaining to underlying kidney disease, cardiovascular history, and other comorbid conditions were extracted.A modified version of the Charlson comorbidity index, i.e., without the age and kidney disease components, was used to assess the severity of comorbidity (22,23).

Anthropometric Evaluation
Body weight assessment and anthropometric measurements were performed while patients were undergoing hemodialysis treatment or within 5 to 20 min after termination of the treatment.Triceps skinfold (TSF) thicknesses were measured with a conventional skinfold caliper using standard techniques as described elsewhere (24 -26).Mid-arm circumference (MAC) was measured with a plastic tape.Mid-arm muscle circumference (MAMC) was calculated from the formula (24,27) MAMC ϭ MAC Ϫ (3.1416 ϫ TSF).Height was obtained from the patient's chart.

Near Infrared Interactance
To evaluate the percentage of body fat and lean body mass, we performed the near infrared (NIR) interactance (28,29) at the same time as the anthropometric measurements.A commercial NIR interactance sensor (portable Futrex 6100, Gaithersburg, MD) was used.NIR measurements were performed by placing a Futrex sensor on the non-access upper arm for several seconds, after entering the required data (date of birth, gender, weight, and height) from each patient.NIR measurements of body fat are shown to correlate significantly with subjective global assessment of nutrition (SGA) and other nutritional measures in MHD patients (28,29).

Laboratory Evaluation
Predialysis blood samples and postdialysis serum urea nitrogen were obtained on a mid-week day and coincided chronologically with the quarterly blood tests of DaVita facilities.Patients had been instructed to fast for at least 4 h before the predialysis blood draw.The single-pool Kt/V was used to represent the weekly dialysis dose, and the normalized protein equivalent of total nitrogen appearance (nPNA), also known as normalized protein catabolic rate (nPCR), was calculated to estimate the daily protein intake (30).Total iron binding capacity (TIBC) represented serum transferrin (31).All routine laboratory measurements were performed by DaVita Laboratories (Deland, FL) using automated methods, and the average values for each laboratory test within the 13-wk study period were calculated and used for data analyses in this study.
Serum C-reactive protein (CRP) and two proinflammatory cytokines-IL-6 and TNF-␣-were measured as indices of the degree of inflammation.The high-sensitivity CRP was measured by a turbidometric immunoassay in which a serum sample is mixed with latex beads coated with anti-human CRP antibodies forming an insoluble aggregate (WPCI, Osaka, Japan; normal range, Ͻ3.0 mg/L) (32,33).High-sensitivity IL-6 and TNF-␣ immunoassay kits based on a solidphase sandwich ELISA using recombinant human IL-6 and TNF-␣ were used to measure serum proinflammatory cytokines (R&D Systems, Minneapolis, MN; normal range, IL-6 Ͻ9.9 pg/ml, TNF-␣ Ͻ4.7 pg/ml) (34 -36).CRP and cytokines were measured in the General Clinical Research Center Laboratories of Harbor-UCLA Medical Center.Serum prealbumin was analyzed by an antigen-antibody complex assay, and total plasma homocysteine (tHcy) concentrations were determined by HPLC at Harbor-UCLA Clinical Laboratories.Normal values of tHcy by this method are 2.8 to 13.5 mol/L in Harbor-UCLA Core Laboratories.

Hospitalization
Hospitalization was defined as any hospital admission that included at least one overnight stay in the hospital.The admission day was counted as one full hospitalization day, but the discharge day was not.All types of hospitalizations, including those related to dialysis access, were included.Because the vast majority of dialysis access-related hospitalizations did not require overnight admission, essentially only those that were associated with other complications and comorbidities were included.For patients who died and those who left the cohort during the prospective follow-up, the hospitalization rates during the survival time were standardized by using the factor 12/survival-time (in months).The annual hospitalization days and annual hospitalization frequency were calculated as described elsewhere (24,27).

Statistical Analyses
Conventional t test and ANOVA were used to detect significant differences among continuous variables in two or more groups, respectively, when applicable. 2 and Kruskal-Wallis rank tests were used for categorical variables.We used Pearson correlation coefficient r for analyses of associations between continuous variables and Spearman rank test for categorical variables.Multivariate regression analysis was performed to obtain partial (adjusted) correlations controlled for case-mix features and relevant covariates, including age, gender, race (Black versus other), ethnicity (Hispanic versus other), insurance status (Medicaid versus other), diabetes, Charlson comorbidity score, dialysis center, and dialysis vintage.Hospitalization rate ratios (RR) were calculated using Poisson regression models.We used Cox proportional hazard regression to calculate hazard ratios (HR) of death after controlling for the above-mentioned covariates.Plots of log [Ϫlog (survival rate)] against log (survival time) were performed to establish the validity of the proportionality assumption as described elsewhere (27).Kaplan-Meier survival analysis examined the differences in prospective mortality among four quartiles of tHcy.Fiducial limits are given as mean Ϯ SD.Both natural (untransformed) and logarithmic values of tHcy were examined in all models.P Ͻ 0.05 or a 95% confidence interval (CI) that did not span 1.00 was considered to be statistically significant.P value between 0.05 and 0.10 was considered borderline significant.Descriptive and multivariate statistics were carried out with the statistical software Stata version 7.0 (Stata Corp., College Station, TX).
Table 1 demonstrates pertinent baseline data at the beginning of the cohort in 367 MHD outpatients.The first column includes the statistical mean and SD for all MHD patients.The second and third columns compare the same values in surviving and deceased patients, respectively.P values based on t test (fourth column) reflect the degree of statistical significance for the differences between the last two categories.Men composed 54.2% of the study population, which was heavily dominated by Hispanics (46.6%) and blacks (29.4%).More than one half (54.7%) of all patients had diabetes, and diabetic prevalence was even higher among deceased patients (73.5%).Exactly half of all patients had a history of cardiovascular disease, including a myocardial infarction, coronary artery procedures such as angioplasty or surgery, congestive heart failure, and peripheral vascular disease including amputation as documented in their charts and/or obtained by questionnaires.Almost 72% of the patients were taking daily multivitamin supplementation designed specifically for patients with renal failure, which usually includes 1000 g of folic acid.These  (10), and other brands (46).Moreover, 86 patients were taking folic acid supplement, 1000 g/d, including 67 who were also taking above-mentioned multivitamins.There was no statistically significant difference in multivitamin or folic acid intake between surviving and deceased patients (Table 1).Table 1 shows that Charlson comorbidity score was significantly higher among deceased MHD outpatients, who were also on average 7 yr older than their surviving counterparts were.Patients had undergone hemodialysis for 36.6 Ϯ 34.2 mo (vintage), and their baseline nPNA was 1.05 Ϯ 0.22.The baseline tHcy was significantly lower in those who died (20.44 Ϯ 7.75 mol/dl) compared with those who survived or censored (24.86 Ϯ 12.04 mol/dl; P ϭ 0.03).Among laboratory tests, baseline values of blood hemoglobin and serum concentrations of albumin, prealbumin, and creatinine were significantly lower and CRP values were higher in deceased patients when compared with surviving ones.There were no statistically significant differences between deceased and surviving patients for body mass index (BMI), total body fat via NIR, TSF, biceps skinfold, and MAMC. Figure 1 shows the distribution of tHcy, which ranged between 5.8 and 113.9 mol/L (24.4 Ϯ 11.8 mol/L; median, 22.3 mol/L) and was skewed to the right.Only 24 (6.5%) patents had a tHcy in the normal range of 2.8 to 13.5 mol/L, and the rest (93.5%) had hyperhomocysteinemia.Of note, tHcy in diabetic patients (23.4 Ϯ 10.3 mol/L) was slightly but significantly lower than in nondiabetic patients (25.9 Ϯ 13.4 mol/L; P ϭ 0.04), a finding consistent with previous reports by other investigators (37,38).
Table 2 shows correlations between tHcy and selected values.Both untransformed and logarithmic values of tHcy were examined.Bivariate (unadjusted) Pearson correlation coefficients were compared with multivariate correlations that controlled for age, gender, race, ethnicity, diabetes, Charlson comorbidity score, dialysis center, and dialysis vintage and are reported.There was a weak but negative correlation between tHcy and the Charlson comorbidity score, indicating that multimorbid patients tended to have a lower tHcy value, but the correlations were not statistically significant after multivariate adjustment.There was a positive correlation between loga- rithm of tHcy and nPNA (nPCR), suggesting a higher protein intake in MHD patients whose tHcy was high.Similar but even stronger correlations were found between both untransformed and logarithmic values of tHcy and serum concentrations of such nutritional markers as albumin, prealbumin, creatinine, and urea nitrogen, with serum creatinine showing the strongest  correlations (0.22 to 0.29).Figures 2 and 3 show scatter diagrams for correlations between tHcy and serum albumin and creatinine concentrations, respectively.No significant correlations were found between tHcy and other nutritional markers (TIBC and cholesterol) or markers of inflammation (CRP, IL-6, and TNF-␣).Among anthropometric measures, only a weak, multivariate adjusted correlation was observed between logarithm of tHcy and TSF and between untransformed tHcy and MAMC, indicating a tendency toward higher tHcy concentrations in patients whose TSF or whose MAMC were larger.Spearman rank correlations were closely similar to Pearson correlations (data not shown).Table 3 compares four quartiles of tHcy.All-cause mortality rate was significantly higher in the lowest tHcy quartile group (17.4%) compared with the other three groups (6.5 to 9.8%; P ϭ 0.04).Cardiovascular death was also higher in the lowest tHcy group (10.9% versus 5.4 to 6.5%), but these differences did not achieve statistical significance.Serum concentrations of albumin, creatinine, and urea nitrogen were significantly lower in the lowest tHcy quartile and higher in the highest tHcy quartile, indicating, once again, a positive association between these nutritional markers and tHcy.The values of other nutritional markers or indicators of inflammation were not significantly different across tHcy quartiles.Figure 4 compares the cumulative proportion of the surviving patients in each tHcy quartiles via the Kaplan-Meier method.Once again, the lowest tHcy quartile exhibited the lowest survival rate as compared with the other three quartiles (Kaplan-Meier P ϭ 0.04).
Tables 4 and 5 show associations between tHcy and RR of hospitalization frequency and total days of hospitalization within 12 mo of prospective follow-up, respectively.In general, there was a consistent trend toward higher hospitalization rates with lower, not higher, tHcy values.Mostly unadjusted, rather than multivariate adjusted, hospitalization frequencies were higher in MHD patients with lower tHcy concentrations (Table 4).However, for total days of hospitalization, all multivariate adjusted RR, which did not include serum albumin, were statistically significant, indicating longer hospital stays for MHD patients with lower tHcy levels (Table 5).For the lowest tHcy quartile, this association was even independent of serum albumin concentration, in that hospitalization RR was 1.15 (95% CI, 1.15 to 1.25; P ϭ 0.001) after adjustment for case-mix and hypoalbuminemia.Table 6 shows HR of death for decreasing levels of tHcy as well as across its down-going quartiles.Again, the lowest tHcy quartile, when compared with higher values, was associated with a significantly higher death risk, i.e., a 2.27 time increase in mortality, independent of case-mix variables or hypoalbuminemia (95% CI, 1.14 to 14.53; P ϭ 0.02).No major change in the magnitude or direction of the associations was noticed after the inclusion of additional terms for intake of folic acid and multivitamin supplements and history of cardiovascular disease in multivariate models.Finally, mortality models were also examined for cardiovascular death exclusively (26 of 37 deaths).Although the same inverse trends were noticed, the associations did not reach statistical significance.For example, for each 1 mol/dl decrease in tHcy, the HR of cardiovascular death was 1.06 (95% CI, 1.00 to 1.12; P ϭ 0.058).

Discussion
In this study, we found a high prevalence of hyperhomocysteinemia in a cohort of 367 MHD patients.The baseline values of tHcy exhibited significant correlations with serum creatinine, albumin, prealbumin, and urea nitrogen but did not correlate with inflammatory markers.We also found that lower values of tHcy were associated with increased hospitalization and mortality.Lower tHcy quartile was associated with a twofold increase in death independent of hypoalbuminemia.These findings all are in sharp contrast to the findings in the general population, in whom higher, rather than lower, values of tHcy are associated with poor outcome, including higher risks of cardiovascular events and death.Our findings, however, are consistent with the recently described phenomenon of reverse epidemiology in dialysis patients (15).
Hyperhomocysteinemia is present in the vast majority of maintenance dialysis patients according to almost all descriptive and prevalent studies (5)(6)(7)(8)(9)(10)(11)(12).In most of such studies, including our current study, Ͼ90% of maintenance dialysis patients have hyperhomocysteinemia, and the mean and median values of tHcy are between 20 and 25 mol/L.This range is substantially higher than the normal range reported for the general population (Ͻ13.5 mol/L; Figure 1).It has been proposed that this high prevalence of hyperhomocysteinemia in dialysis patients is one of the main reasons for the high rate of cardiovascular events and poor clinical outcome in these individuals (5)(6)(7)(8)(9).Although some studies reported positive corre-lations between increased tHcy and higher rate of mortality in dialysis patients (7)(8)(9), other studies failed to show such an association (39,40) or found paradoxically reversed correlations (10 -12).Sirrs et al. (10) examined the association between tHcy and vascular access complications in 96 MHD patients and did not find any correlation between them.However, they found a significantly higher 9-mo mortality in MHD patients whose tHcy was Ͻ27 mol/L.Suliman et al. (11) examined possible relationships among tHcy, nutritional status, and ischemic cardiovascular disease in a cohort of 117 MHD patients and found that malnourished MHD patients had significantly lower levels of serum albumin and tHcy.Similar to our results, these investigators also reported positive and significant correlations between tHcy levels and serum concentrations of albumin and creatinine (11).They also found that nPNA, an indicator of protein intake, was independently associated with tHcy concentrations.Finally, Wrone et al. (12) explored cross-sectional associations between tHcy and history of cardiovascular disease in 459 MHD patients and found that mean tHcy was higher in patients without a history of cardiovascular disease (35.2 mol/L versus 30.4 mol/L; P ϭ 0.02).In multivariate models, the history of cardiovascular disease was still negatively associated with higher tHcy levels (12).
Because hyperhomocysteinemia is common in dialysis patients and because it is a cardiovascular risk factor in the general population, the intuitive inference is that hyperhomocysteinemia is the cause of poor outcome in dialysis population.This possibly flawed reasoning is known in epidemiology as ecological fallacy, in that inferences are based on comparing populations rather than individuals (41).However, a few stud- ies indeed have shown a positive correlation between tHcy and poor outcome (7,9,42), but they are subject to restrictions.For instance, in the study by Mallamaci et al. (7), although tHcy was found to be positively associated with cardiovascular outcome in 175 MHD patients who were followed for an average of 29 mo, it failed to independently predict survival.Moreover, in the aforementioned study, neither serum albumin nor CRP predicted mortality, whereas our study did.Hence, the reason that our current study and three other recent studies (10 -12) found the opposite association as compared with a few others (7,9,42) could be due to differences in race or ethnicity, comorbidities, the prevalence of diabetes, and/or study designs.Moreover, publication bias may have handicapped or delayed reporting such paradoxical findings in dialysis patients, because the investigators' first impression upon encountering results with inverse association between tHcy and mortality may be to consider them erroneous or flawed and hence to avoid reporting them (15,43,44).
The positive correlation between tHcy and markers of protein-energy nutritional status in maintenance dialysis patients may partially, but not fully, explain the paradoxical association between tHcy and mortality in these individuals.Diminished nutritional status, with or without inflammation, is a strong predictor of poor outcome in these individuals (24,(45)(46)(47)(48). Similar to our results, both Suliman et al. (11) and Wrone et al. (12) also reported significant correlations between tHcy and such nutritional markers as nPNA and serum concentrations of albumin, prealbumin, and creatinine.It is important to note that in all three studies, including ours, the association between tHcy and serum creatinine is the strongest as indicted by correlation coefficients, i.e., even stronger than associations with serum albumin concentration (see Figures 1 and 2 and Table 2).Serum creatinine is an indicator not only of dialysis dose and efficiency but also of muscle mass and nutritional state in maintenance dialysis patients (49).Its paradoxical association with improved survival is well known and is described as one of the main components of the reverse epidemiology phenomenon (15, 49 -51).Unlike serum albumin, serum creatinine may be less or not at all associated with inflammation.Serum albumin, conversely, is a combined marker of both protein-energy malnutrition and inflammation (52) and, hence, an indicator of malnutrition-inflammation complex syndrome (18).tHcy seems to be a more exclusive nutritional marker, because its association with serum albumin is weaker than that with serum creatinine.Indeed, consistent with this exclusive nutritional hypothesis, we found no association between tHcy and such inflammatory markers as serum CRP, IL-6, and TNF-␣ concentration (Table 2).The correla-  tion between tHcy and serum creatinine concentration could also be the result of the metabolic association between creatinine and homocysteine.The formation of creatine, the precursor of creatinine, depends on methyl donation by S-adenosylmethionine to become S-adenosylhomocysteine, leading to the formation of homocysteine (53).According to Oishi et al. (38), the demethylation pathway of methionine may be disturbed in diabetic MHD patients.This may be the reason for slightly lower serum tHcy levels in diabetic MHD patients when compared with their nondiabetic counterparts as found in the current and other studies (37,38).
Our current study should be qualified by the possibility of selection bias.During the initial recruitment in eight dialysis units (with Ͼ1000 patients), it is possible that only MHD patients who were generally healthier and more health-conscious and had better nutritional status agreed to participate (385 patients).Hence, the final 367 MHD patients of this study whose tHcy was measured might include disproportionately less sick and better nourished individuals.Indeed, the annual mortality rate among all patients of the study dialysis units was 15%, whereas it was as low as 10% among selected patients for the NIED study.However, a selection bias with such a direction generally would lead to a bias toward the null, so without this bias, our positive results probably would have been even stronger and the associations indeed more prominent.Another possible limitation can be related to the relatively short period of longitudinal follow-up (only 12 mo as compared with 4 yr in the study by Suliman et al. (11)).Moreover, some associations existed only when logarithmic values of tHcy were analyzed.As depicted in Figure 1, there was a right-sided skewness in tHcy values.Hence, logarithmic transformation was appropriate and the proper method.Finally, another limitation of our study is the lack of blood folate measurements.However, studying the effect of folate deficiency on tHcy is beyond the scope of this epidemiologic study, which focuses on the direction of the association between tHcy and clinical outcome in MHD patients.Some strengths of our study include its large sample size with inclusion of many individuals with diabetes, its longitudinal design, its inclusion of hospitalization data, and the use of extensive markers of inflammation and nutrition at the same time.Our study is unique because, unlike previous studies in which only serum CRP was measured as the marker of inflammation and only mortality was examined, we measured a multiple number of inflammatory makers (CRP, IL-6, and TNF-␣) along with a number of anthropometric indicators (TSF and MAMC) and analyzed hospitalization measures.
In our study, we did not find any major association between tHcy and anthropometric values such as BMI or body fat.Wrone et al. (12) showed a bivariate correlation coefficient of 0.19 between tHcy and BMI in MHD patients.Our data showed the same trend without adequate statistical significance.This may be due to the selection bias at recruitment level as described above, especially because Ͼ55% of our patients were diabetic.However, we found a weak but statistically significant, positive correlation between tHcy and such anthropometric values as MAMC and TSF.Moreover, consistent with two previous studies (11,12), we, too, found a statistically significant and positive association between tHcy and nPNA, suggesting a positive role of protein intake in determining the tHcy level in MHD patients.
In summary, our data show that in MHD patients, there is a paradoxically inverse association between poor clinical outcome and levels of tHcy, a known risk factor for cardiovascular disease in the general population.Maintenance dialysis patients are a group of highly selected individuals whose high rate of cardiovascular disease and mortality does not seem to be due to such traditional risk factors as hypertension, obesity, hypercholesterolemia, or hyperhomocysteinemia.Such paradoxical associations, in the setting of the unique phenomenon of risk factor reversal, may have major clinical implications in our patients.To that end, the extrapolation of data of traditional risk factors to dialysis patients without first establishing that these factors constitute significant risks in dialysis patients as well may be flawed.The mortality rate of dialysis patients in the United States has not changed significantly in the past 10 to 15 yr (54) despite ongoing efforts to target such traditional risk factors as hypertension, obesity, hypercholesterolemia, and hyperhomocysteinemia. Increasing dialysis dose or use of high-flux dialysis membranes also failed to show any improvement in survival according to the HEMO Study (55).Thus, it may be time to consider carrying out clinical trials that assess the effect of such nontraditional risk factors as malnutrition on survival in maintenance dialysis patients.

Figure 2 .
Figure 2. Scatter plot reflecting the correlation between serum albumin and total plasma homocysteine in 367 MHD patients.Cases with a total plasma homocysteine Ͼ60 mol/dl are not shown (n ϭ 7).

Figure 3 .
Figure 3. Scatter plot reflecting the correlation between serum creatinine and total plasma homocysteine in 367 MHD patients.Cases with a total plasma homocysteine Ͼ60 mol/dl are not shown (n ϭ 7).

Figure 4 .
Figure 4. Cumulative proportion of surviving patients according to quartiles of total plasma homocysteine in 367 MHD patients.

Table 2 .
Correlation coefficients between total plasma homocysteine (both untransformed values and logarithmic scales) and selected laboratory, comorbidity, and nutritional values a a Bivariate (Pearson, unadjusted) and multivariate regression adjusted r values, also known as partial correlation, controlled for age, gender, race, diabetes, Charlson score, dialysis center, and dialysis vintage are reported.Statistically significant correlations are in bold.tHcy,total plasma homocysteine.

Table 3 .
Relevant demographic, mortality, comorbidity, and nutritional and inflammatory values in quartiles of tHcy a a P value is based on ANOVA or Kruskal-Wallis test for continuous or categorical variables, respectively.

Table 4 .
Association Hispanic versus other), insurance status (Medicaid versus other), diabetes, Charlson comorbidity score, dialysis center, and dialysis vintage.The extended multivariate model also includes serum albumin concentration.All models are based on Poisson regression analyses.Note that RR for tHcy is based on each 1 mol/dl decrease in tHcy or each 0.1 unit decrease in logarithm of tHcy.RR, rate ratio.

Table 5 .
Association between tHcy and total days of hospitalization over the 12-month follow-up period, as reflected by hospitalization RR and 95% confidence intervals a Multivariate RR values are adjusted for age, gender, race (Black versus other), ethnicity (Hispanic versus other), insurance status (Medicaid versus other) diabetes, Charlson comorbidity score, dialysis center, and dialysis vintage.The extended multivariate model also includes serum albumin concentration.All models are based on Poisson regression analyses.Note that RR for tHcy is based on each 1 mol/dl decrease in tHcy or each 0.1 unit decrease in logarithm of tHcy. a