Body size, dose of hemodialysis, and mortality

Body Size, Dose of Hemodialysis, and Mortality Robert A. Wolfe, PhD, Valarie B. Ashby, MA, John T. Daugirdas, MD, Lawrence Y.C. Agodoa, MD, Camille A. Jones, MD, and Friedrich K. Port, MD ● This study investigates the role of body size on the mortality risk associated with dialysis dose in chronic hemodialysis patients. A national US random sample from the US Renal Data System was used for this observational longitudinal study of 2-year mortality. Prevalent hemodialysis patients treated between 1990 and 1995 were included (n ⴝ 9,165). A Cox proportional hazards model, adjusting for patient characteristics, was used to calculate the relative risk (RR) for mortality. Both dialysis dose (equilibrated Kt/V [eKt/V]) and body size (body weight, body volume, and body mass index) were independently and significantly (P F 0.01 for each measure) inversely related to mortality when adjusted for age and diabetes. Mortality was less among larger patients and those receiving greater eKt/V. The overall association of mortality risk with eKt/V was negative and significant in all patient subgroups defined by body size and by race-sex categories in the range 0.6 F eKt/V F 1.6. The association was negative in the restricted range 0.9 F eKt/V F 1.6 (although not generally significant) for all body-size subgroups and for three of four race-by-sex subgroups, excepting black men (RR ⴝ 1.003/0.1 eKt/V; P G 0.95). These findings suggest that dose of dialysis and several measures of body size are important and independent correlates of mortality. These results suggest that patient management protocols should attempt to ensure both good patient nutrition and adequate dose of dialysis, in addition to managing coexisting medical conditions. 娀 2000 by the National Kidney Foundation, Inc. INDEX WORDS: Body mass index (BMI); urea reduction ratio (URR); hemodialysis dose; nutritional status; equilibrated Kt/V (eKt/V); total body water (TBW).

S

EVERAL NONRANDOMIZED studies have investigated the association between patient mortality and the reduction of blood urea concentration during hemodialysis among patients treated for end-stage renal disease.1,2 These studies have been motivated by the hypothesis that urea concentration is a marker for the level of small-molecular-weight toxins in the blood and that less removal of these toxins is associated with increased patient mortality. Both urea reduction ratio (URR) and Kt/V are related to the fraction of urea removed during dialysis.3,4 URR measure is the fractional reduction in blood urea concentration (BUN) during dialysis ([pre BUN ⫺ post-BUN]/pre-BUN). During dialysis, the concentration of blood urea shows an approximate exponential decline over time. The rate of that exponential decline per unit of time is K/V, which depends on the clearance rate (K) of the

From the US Renal Data System Coordinating Center, University of Michigan, Ann Arbor, MI; Westside Veterans Administration Hospital, University of Chicago, Chicago, IL; and the National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD. Received February 15, 1999; accepted in revised form July 2, 1999. Address reprint requests to Robert A. Wolfe, PhD, University of Michigan, 315 W Huron, Ste 240, Ann Arbor, MI 48103. E-mail: [email protected]

娀 2000 by the National Kidney Foundation, Inc. 0272-6386/00/3501-0013$3.00/0 80

dialyzer and the urea distribution volume (ie, total-body water [TBW]) to be cleared (V). The equation: (1 ⫺ URR) ⬇ exp (⫺Kt/V) where t is the duration of dialysis and exp(·) is the exponential function, would hold in an idealized patient with no weight change during dialysis, no intradialytic urea generation, and no urea gradient across body compartments (eg, sequestration of urea caused by poor perfusion). This equation has been extended to include approximate adjustments for the effects of intradialytic weight loss and urea generation.5 The National Cooperative Dialysis Study (NCDS) is a prospective study of four dialysis prescriptions.6 A post hoc analysis of data from this study3 showed that mortality rates were less among patients with greater Kt/V. Data from National Medical Care1 have shown that greater URR is associated with lower mortality. The US Renal Data System (USRDS)2 has reported similar results based both on URR and Kt/V, while adjusting for patient comorbidity. The USRDS study showed a strong correlation of URR and Kt/V (R2 ⫽ 0.92). Lower mortality risk has been associated with both greater single-pool Kt/V (Kt/Vsp) and double-pool Kt/V.7 All these reports suggest that, on average, the incremental benefit of increasing dose of dialysis (greater URR or greater Kt/V) tends to become smaller for pa-

American Journal of Kidney Diseases, Vol 35, No 1 (January), 2000: pp 80-88

BODY SIZE, HEMODIALYSIS DOSE, AND MORTALITY

81

tients treated with a greater dose of dialysis (Kt/V ⬎ 1.1 double pool or ⬎ 1.3 single pool), as summarized by Gotch et al.8 Because it has been suggested that the postdialysis rebound in BUN level depends on the efficiency of dialysis relative to body size, we have chosen to report the equilibrated Kt/V (eKt/V) as the main measure of dialysis dose. The same analyses were also performed using Kt/Vsp, as suggested in the Dialysis Outcomes Quality Initiative (DOQI) guidelines9 with very similar results. Nutritionally related factors are among the most important patient characteristics predictive of mortality rates. The USRDS10 has reported that larger body mass index (BMI ⫽ weight divided by height squared) and greater serum albumin level are independent and strong correlates of lower mortality. The USRDS11 and Owen et al1 reported that such nutritional measures as serum albumin level are among the strongest predictors of patient mortality. Some of the variation in these measures is likely caused by malnutrition, which may be subject to clinical intervention. Some researchers12,13 have posited and studied a possible mechanistic link between Kt/V and protein catabolic rate (PCR). Unfortunately, it is difficult to disentangle the true effects of Kt/V and PCR, a measure of protein intake, because the commonly used measures of these quantities are both based on predialysis and postdialysis BUN measures. Measurement errors in BUN are thus propagated to both Kt/V and PCR, which leads to a mathematical link between their measured values. Recently, Owen et al14 reported that the lower mortality benefit associated with greater levels of dialysis dose varied substantially among different patient race-sex subgroups. To further investigate the association of body size and dialysis dose with mortality risk, we analyzed the data from two national random samples of US hemodialysis patients.

and 4 of the USRDS Dialysis Mortality and Morbidity Study [DMMS]). Details of the data collection of both the CMAS and the DMMS have been published elsewhere.15,16 Nationally, the average dose of dialysis increased between the time periods of these two studies16; thus, data from these samples were combined for these analyses to increase the variation in dose of dialysis in the study population. Because other treatment factors also changed over this time period, all analyses were statistically adjusted for the source of the data (DMMS versus CMAS). There was no statistical evidence that the association between mortality and eKt/V differed in the two studies (P ⬎ 0.05 for interaction of study by eKt/V dose). To minimize the influence of unmeasured residual renal function on mortality risk, patients treated for less than 1 year were excluded. Delivered eKt/V was calculated from predialysis and postdialysis BUN and weight measurements according to the Daugirdas formula,5 using data for one to six dialysis sessions over a 6-month period before the start of the study. The average of these values was used in the analysis. The measures listed in Table 1 were collected before patient entry into follow-up for mortality ascertainment and were used as predictors of mortality during a 2-year follow-up period. There were 9,490 patients from the CMAS and DMMS who had started dialysis therapy at least 1 year before the study start, were aged 18 years or older, were matched to the USRDS database for complete mortality ascertainment, received three dialysis sessions weekly, had predialysis and postdialysis BUN measurements, were using a bicarbonate dialysate, had no reported acquired immunodeficiency syndrome, and had data on age at study start, race, prescribed time on dialysis, Kt/V, height, and postdialysis weight. Of these, there were 9,165 patients with an eKt/V between 0.60 and 1.60 (325 patients with eKt/V outside this range were excluded from the analysis). Patient follow-up

METHODS The USRDS Case Mix Adequacy Study (CMAS) enrolled a national sample of 7,096 patients randomly selected from the patients being treated on January 1, 1991, at each of a random selection of 523 facilities in the United States. Another national sample of 16,600 patients was randomly selected from the patients being treated on January 1, 1994, at each of 1,181 facilities in the United States (waves 1, 3,

Table 1. Patient Characteristics at Study Start

Measure/Characteristic

Age at study start (y) Diabetes as ESRD cause (%) History of diabetes (%) Women (%) Race (%) Asian Black Native American White BMI (kg/m2) Weight (kg)* Volume (TBW)† In DMMS study (%) eKt/V (Daugirdas)‡

Mean or Percent

58.8 28 40 50 2 45 1 52 24.9 72.3 40.2 79 1.01

SD

15.5

5.9 21.8 8.9 0.19

Abbreviation: ESRD, end-stage renal disease. *Postdialysis weight. †TBW according to anthropometric equation of Chertow et al.17 ‡Data from.5

82

WOLFE ET AL Table 2. Ranges of Body Size Measures in Approximate Tertiles Low Range

Size Measure

BMI (kg/m2) Postdialysis weight (kg) Volume (L)

Middle Range

Table 4. Mortality RR Percentage of Change per 0.1 Greater eKt/V by Body Size Groups

High Range

9.2-21.8

21.9-26.9

27.0-89.2

25.0-61.4 15.7-35.0

61.5-75.3 35.1-42.8

75.4-210.0 42.9-88.0

for mortality was ascertained for the entire study population and was censored at transplantation or end of study (up to 2 years after entry onto the study). Several measures of body size were used in the various analyses. The prescribed dry weight or postdialysis weight (in kilograms), BMI (weight divided by height squared, in kilograms per square meter), and urea distribution volume (or TBW) according to Chertow et al17 were computed for each patient. The ranges shown in Table 2 divide the available sample approximately into thirds (tertiles) to define groups of low, medium, and high body size for each size measure. The association of mortality to Kt/V was analyzed using several different models, including a model with a linear eKt/V term and a categorical model with the eKt/V ranges shown in Table 3 (1.0 to 1.1 was the reference range) and with a piecewise linear spline function constrained to have a gradient of 0 in the range 0.9 ⬍ eKt/V ⬍ 1.6, all with adjustment for the factors listed in Table 1. The linear association was also analyzed in the eKt/V range greater than 0.9 (Tables 4 and 5), because the previously mentioned categorical models indicated that the association was not linear over the full range of eKt/V for many subgroups. Separate analyses were performed with and without adjustment for the body size measures. Additional analyses were performed for the patients in each of several body size groups to ensure that patients of similar size were being considered when the mortality-Kt/V association was estimated. In other analyses, statistical adjustments for body size were made by including both log (volume) and log (BMI) in the analyses (both statistically significant). Statistical adjustment for body size is an alternative to separate analyses by body-size subgroup, as a way to estimate the Kt/V-mortality association among patients of similar body

0.60-0.79 0.80-0.89 0.90-0.99 1.00-1.09 1.10-1.19 1.20-1.29 1.30-1.59

No. of Patients

1,274 1,524 1,901 1,718 1,281 784 683

BMI (kg/m2)

26.11 25.69 25.30 24.73 24.01 23.61 22.95

Volume (L)

43.97 42.42 41.07 39.86 37.87 36.80 35.37

Weight (kg)

80.10 76.62 74.10 71.26 67.11 65.69 62.94

BMI Low Med High Weight Low Med High Volume Low Med High

0.6 ⬍ eKt/V ⬍ 1.6

0.9 ⬍ eKt/V ⬍ 1.6

⫺3.5%* ⫺7.5%* ⫺7.8%*

⫺2.5% ⫺5.2% ⫺6.2%

⫺2.9% ⫺8.6%* ⫺8.0%*

⫺1.1% ⫺8.0%* ⫺4.9%

⫺4.6%* ⫺5.3%* ⫺6.8%*

⫺3.4% ⫺4.7% ⫺0.1%

*P ⬍ 0.05 compared with 0% (no association).

size. When the mortality-Kt/V association is evaluated with no adjustment for body size, the resulting analysis summarizes the combined association of lower mortality with greater Kt/V and of greater mortality with the smaller body size associated with the greater Kt/V. Additional analyses were performed for the patients in each of several race-sex groups.

RESULTS

The frequency and average measures of patient characteristics are listed in Table 1. Patients with greater eKt/V had smaller body-size measurements, on average (Table 3). The trends for a correlation of greater eKt/V with lower BMI (R ⫽ ⫺0.17; P ⫽ 0.0001), weight (R ⫽ ⫺0.24; P ⫽ 0.0001), and volume (R ⫽ –0.29; P ⫽ 0.0001) were statistically significant. The relative risk (RR) for mortality at different levels of eKt/V is shown in Fig 1 for patients in each of the three body size groups for each of the three ways of measuring size. For the categorical eKt/V ranges of Table 3, the RR values are shown in Fig 1 as isolated points with standard error bars. Separate (log) linear gradients were fit Table 5. Mortality RR Percentage of Change per 0.1 Greater eKt/V by Race-Sex Groups

Table 3. Average Patient Size by Delivered eKt/V Range eKt/V Range

Group

0.6 ⬍ eKt/V ⬍ 1.6

0.9 ⬍ eKt/V ⬍ 1.6

eKt/V

0.73 0.85 0.95 1.05 1.15 1.25 1.40

Group

Unadjusted for Size

Adjusted for Size

Unadjusted for Size

Adjusted for Size

White men White women Black men Black women

⫺1.3% ⫺4.9%* ⫺5.6%* ⫺5.4%*

⫺3.9%* ⫺7.7%* ⫺8.7%* ⫺8.4%*

0.2% ⫺4.2% 2.5% ⫺0.5%

⫺3.7% ⫺7.9%* 0.3% ⫺4.3%

*P ⬍ 0.05 compared with 0% (no association).

Fig 1. Relative mortality risk (RR) by eKt/V dose for low, medium, and high body size by (A) BMI, (B) body weight, and (C) TBW volume, adjusting for age, race, sex, diabetes, and study. Each body size measurement is based on postdialysis weight. eKt/V is based on.3 The sloping lines indicate correlation using eKt/V as a continuous variable, whereas the vertical bars describe standard errors for the categorical estimates by eKt/V interval.

84

WOLFE ET AL

to the data for each size group modeling eKt/V as a continuous linear variable. The results are shown as the solid lines in Fig 1, and the negative gradient shows that the mortality rate tends to be less for patients receiving greater Kt/V in each group of patients. The linear model is a good approximation to the categorical model for many of the groups of patients. The steepness of the association between eKt/V and mortality did not depend substantially on which body-size group was considered. The RR gradients for eKt/V were not significantly different from each other in any of the nine analyses shown (P ⬎ 0.05), which indicates no statistical evidence that the association between eKt/V and mortality differs by body size. However, at any level of eKt/V, patients with larger body size had less mortality risk than those with smaller body size, as evident at the referent eKt/V range of 0.9 to 1.0 for each of the measures. Careful examination of Fig 1 shows that the RR is substantially greater for eKt/V of 0.9 or less compared with eKt/V greater than 0.9 in many size subgroups (note that eKt/V of 0.9 corresponds approximately to an Kt/Vsp of 1.1). Thus, part of the steepness of the linear trends noted in Fig 1 could be caused by the elevated mortality seen for eKt/V of 0.9 or less. Analyses limited to the range of eKt/V greater than 0.9 show that the linear association between RR and eKt/V is less steep in this greater eKt/V range. Table 6 reports the average body size measures and eKt/V for four race-sex groups of patients. Women tended to have greater BMI and eKt/V but less weight and volume than men of the same race on average. Black patients have greater BMI, weight, and volume but lower eKt/V than white patients of the same sex on average. Figure 2 shows the RR for mortality versus eKt/V range for the four major race-sex subgroups. As in Fig 1, the RR is shown both for

discrete ranges of eKt/V and from four separate (log) linear analyses for eKt/V as a continuous variable. The gradient is significantly negative (P ⬍ 0.05) for each of the high, medium, and low body-size groups regardless of the method used to measure body size. The categorical results indicate that the association between RR and eKt/V may be less steep in the eKt/V range greater than 0.9 for several race-sex subgroups. In addition, a spline model is shown, which is constrained to have gradient 0 for eKt/V greater than 0.9. Table 4 lists the association between eKt/V and mortality in several patient body-size subgroups and eKt/V ranges. The table reports the average percentage of reduction in mortality per 0.1 greater eKt/V. The values in the 0.6 ⬍ eKt/V ⬍ 1.6 column of this table correspond to the gradients of the linear associations shown in Fig 1A-D. The values shown for the range 0.9 ⬍ eKt/V ⬍ 1.6 are based on the gradient of the linear associations fit to the four greatest eKt/V categories only. For all nine body-size groups, the percentage of reduction per 0.1 greater eKt/V is greater overall (in the range 0.6 ⬍ eKt/V ⬍ 1.6) than it is in the range 0.9 ⬍ eKt/V ⬍ 1.6. Thus, the linear associations shown in Fig 1A-D show only an average gradient but miss the details of the true association, which is flatter in the greater eKt/V range and steeper in the lower eKt/V range. For eKt/V greater than 0.9, the gradient is negative for all nine body size groupings, indicating less mortality in the greater eKt/V dose ranges, but is significantly different from 0 only in the medium-weight group and is close to 0 in the high-volume group. Table 5 lists the association between eKt/V and mortality in several patient groups by race and sex. This table shows that the gradient is more negative overall than it is in the range of eKt/V greater than 0.9 for most race-sex groups,

Table 6. Average Patient Size and Delivered eKt/V by Race and Sex Subgroup Average Subgroup

No.

BMI (kg/m2)

Volume (L)

Weight (kg)

eKt/V

Percent With eKt/V ⬎0.9

White men White women Black men Black women

2,498 2,237 1,954 2,141

24.34 24.70 24.68 26.02

45.43 33.86 46.65 35.67

76.01 65.51 78.14 71.02

0.99 1.07 0.95 1.01

67 80 59 70

BODY SIZE, HEMODIALYSIS DOSE, AND MORTALITY

85

as it is for most weight groups (Table 4). Table 5 also compares the gradient between mortality and eKt/V with and without adjustment for body size. In all four race-sex groups, the gradient is steeper with adjustment for body size than when unadjusted for body size. That is, because greater eKt/V is delivered to smaller patients on average, the unadjusted association combines the lower mortality associated with greater eKt/V with the greater mortality associated with smaller body size to yield a relatively flatter gradient than seen among patients of a constant body size. Table 7 reports the magnitude of the association of greater eKt/V, greater log (BMI), and greater log (volume) on mortality risk, adjusted for each other and for the other factors in Table 1. To make the values for each measure more similar, the adjusted estimated percentage of changes in mortality risk are reported for a 1-SD increase. The reported percentage of change in RR is therefore interpretable as the change associated with a typical change for each measure. The RR reduction ranges from 10% to 15% for an SD increase in any of these measures (all else held constant). Each RR reduction is statistically significant, although a 1-SD increase in log (BMI) and eKt/V is more significant (has a greater chi-squared statistic) than log (volume). In sensitivity analyses, we adjusted for eight baseline comorbidity measures (cerebrovascular accident, coronary heart disease, congestive heart failure, left ventricle hypertrophy, neoplasm, pulmonary disease, peripheral vascular disease, and smoker) as an additional adjustment for patient characteristics. The results of this sensitivity analysis were similar to that previously reported and did not affect the conclusions. Additionally, results of RR by body-size measures were similar when Kt/Vsp was used instead of eKt/V. Because serum albumin level may be a consequence of dialysis dose,12 the main analysis did not adjust for albumin level. When albumin level was added as a covariate, however, the results remained essentially unchanged. Serum albumin level did not show a significant correlation with eKt/V (r ⫽ –0.00391; P ⫽ 0.72)

come. Larger body size and greater eKt/V are associated with less mortality risk. We believe that these two factors affect mortality by very distinct mechanisms. Although the dialysis dose is calibrated as eKt/V, which appears to be standardized to patient volume, the measure is derived from predialysis and postdialysis BUN levels and not from volume. The dialysis dose is used to measure the removal of uremic toxins from the body as measured by the urea concentration. The urea concentration has been believed to serve as a marker for various uremic toxins derived from protein catabolism and is therefore likely to be an appropriate proxy for their physiological effect on mortality. The consistent finding of a negative association between eKt/V and mortality in every patient group considered supports this interpretation. Analyses from the NCDS,6 the USRDS,2 and National Medical Care1 have all found less mortality among patients treated with greater Kt/V or URR. This study is limited by the use of a baseline measure of eKt/V. It is likely that a study based on serial measures of eKt/V would allow more accurate calibration of the association between the dose of dialysis and mortality. The pathophysiological mechanisms by which body size is linked to mortality are poorly understood. Owen et al1 showed that serum albumin level is a strong correlate of mortality. Leavey et al10 showed that BMI, albumin level, and clinical assessment of nutritional status individually and independently are each strong correlates of mortality. Our present results show that body-size measures (weight, TBW, BMI) with or without adjustment for albumin level were strongly associated with patient mortality. That several bodysize measures are independently (and simultaneously) associated with mortality prevents us from identifying a single biological measure of body size as a gold standard. The log (BMI) was a stronger and more significant predictor of mortality (according to chi-square) than either TBW or weight. Table 7 shows that greater TBW and BMI are both simultaneously related to less mortality. It is not surprising that these measures might have independent effects on mortality because they are conceptually different measures of size; TBW is related to overall fat-free weight, whereas BMI can be interpreted as a measure of

DISCUSSION

Both body size (weight, BMI, or TBW) and dialysis dose (URR, Kt/V, or eKt/V) are independent and significant predictors of patient out-

86

WOLFE ET AL

Fig 2. Relative mortality risk (RR) for patients by race (black and white) and sex adjusting for age, diabetes, study, and body size. (A) White men, (B) white women, (C) black men, and (D) black women are shown. eKt/V is based on.3 The sloping solid lines indicate correlation, using eKt/V as a continuous variable, whereas the vertical bars describe standard errors for the categorical estimates by eKt/V interval. A piecewise linear spline function constrained to have gradient ⴝ 0 for eKt/V greater than 0.9 is shown as a dashed line. X denotes the mean eKt/V.

relative girth and nutrition, including fat. The present cross-sectional study cannot distinguish small body size caused by genetic factors from small body size caused by malnutrition, so the clinical implication of these results are not certain but are consistent with the model that maintenance of nutritional health, especially measured by BMI, is important in the dialysis population. The lack of an association between eKt/V and serum albumin level in this study gives no support for the hypothesis that low eKt/V depresses nutritional intake12 and suggests that other factors may have a greater effect on serum albumin level. The lack of consistently recorded serial

measurements of eKt/V and serum albumin level in this study limits the interpretability of our results regarding this question. A greater reduction in mortality with greater dialysis dose is evident when patients of equal size are considered in the analysis than when patient size varies with the dose of dialysis. A randomized clinical trial comparing mortality for different dialysis dose levels should obtain nearly equal distributions of patient sizes for the different doses through the randomization process. In this observational experiment, which is based on clinical dosing patterns practiced in the United States, the average body size is smaller in the higher dose groups (Table 3). Thus, both body

BODY SIZE, HEMODIALYSIS DOSE, AND MORTALITY

87

Fig 2. (cont’d) (C) and (D).

size and dose of dialysis must be considered in the analyses of these data. Otherwise, it would not be clear whether the differences in mortality among the dose groups were caused by the differences in dose or the differences in body size. The combined effects of these two factors tend to partially cancel each other out because greater Kt/V tends to be administered to smaller patients. Our results indicate that the benefit of increasing Kt/V may be less at greater Kt/V levels than at lower levels, as suggested previously.1,2,3,8 That the mortality RR versus eKt/V association showed a steeper gradient for eKt/V less than 0.9 and a less steep gradient for eKt/V greater than

0.9 in all body size groups (Fig 1) and most race-sex groups (Fig 2) indicates the generality of this result. These results replicate several of the results reported recently by Owen et al.14 However, these results are different in several important Table 7. Reduction in Relative Mortality Risk (RR) According to Kt/V, BMI, and Volume

Measure

Mean

SD

eKt/V 1.01 0.19 Log (BMI) 3.19 0.22 Log (volume) 3.67 0.21

RR % Change per SD Chi-Square

⫺12% ⫺25% ⫺8%

37.7 134.6 6.7

P

0.0001 0.0001 0.0095

88

WOLFE ET AL

regards. First, Owen et al14 reported no association between Kt/V dose and mortality risk among blacks. We show a very strong association for this group, as for all other groups. Our results show a relationship of Kt/V with mortality for almost all size and race-sex groups of patients, except in the range eKt/V greater than 0.9. The results reported by Owen et al14 are limited to patients treated at facilities with similar treatment protocols, and the reported lack of an association of Kt/V with mortality could partially be caused by analyses not adjusted for patient size. We show that such adjustment dramatically clarifies the association of Kt/V to mortality by accounting for the observed significant inverse correlation between body size and Kt/V, ie, smaller patients receiving greater Kt/V. Our results show a negative overall association between eKt/V and death rates for all nine size subgroups and four race-sex groups of patients analyzed. In the range of eKt/V greater than 0.9, the association was less steep but was still negative, except for black men. This study does not have adequate sample size to give detailed estimates of the slope in different eKt/V ranges, but the consistency of the patterns shown here among various subgroups suggests that the gradient is truly steeper in the low eKt/V range than it is in the greater eKt/V range. In summary, this study shows that both Kt/V (or URR) and patient size are important correlates of mortality that should be considered simultaneously in the management of dialysis patients. These findings agree with the general recommendations of the DOQI guidelines.9 Dialysis dose and patient mass may be modified independently, and both appear to be important for survival of patients treated with hemodialysis. Although the dose of hemodialysis can be controlled more directly by health care providers than patient size (or nutritional status), this study supports the hypothesis that both dialysis dose and nutrition are of vital importance to hemodialysis patients. REFERENCES 1. Owen W, Lew N, Liu Y, Lowrie E, Lazarus M: The urea reduction ratio and serum albumin concentration as predictors of mortality in patients undergoing hemodialysis. N Engl J Med 329:1001-1006, 1993

2. Held PJ, Port FK, Wolfe RA, Stannard DC, Carroll CE, Daugirdas JT, Bloembergen WE, Greer JW, Hakim RM: The dose of hemodialysis and patient mortality. Kidney Int 50:550-556, 1996 3. Gotch F, Sargent JA: A mechanistic analysis of the National Cooperative Dialysis Study (NCDS). Kidney Int 28:526-536, 1985 4. Depner TA: Quantifying hemodialysis. Am J Nephrol 16:17-28, 1996 5. Daugirdas JT: Simplified equations for monitoring Kt/V, PCRn, eKt/V, and ePCRn. Adv Ren Replace Ther 2:295-304, 1995 6. Parker TF, Laird NM, Lowrie EG: Comparison of the study groups in the National Cooperative Dialysis Study and a description of morbidity, mortality and patient withdrawal. Kidney Int 23:S42-S49, 1983 (suppl 13) 7. Levin NW, Stannard DC, Gotch F, Port FK: Comparison of mortality risk (RR) by Kt/V single pool (SP) vs double pool (DP) analysis in diabetic and non-diabetic hemodialysis patients. J Am Soc Nephrol 6:606A, 1995 (abstr) 8. Gotch FA, Levin NW, Port FK, Wolfe RA, Uehlinger DE: Clinical outcome relative to the dose of dialysis is not what you think: The fallacy of the mean. Am J Kidney Dis 30:1-15, 1997 9. National Kidney Foundation: DOQI Practice Guidelines for Hemodialysis Adequacy. Am J Kidney Dis 30:S15S64, 1997 (suppl 2) 10. Leavey SF, Strawderman RL, Jones CA, Port FK, Held PJ: Simple nutritional indicators as independent predictors of mortality in hemodialysis patients. Am J Kidney Dis 31:997-1006, 1998 11. US Renal Data System: Comorbid conditions and correlations with mortality risk among 3,399 incident hemodialysis patients. Am J Kidney Dis 20:S32-S38, 1992 (suppl 2) 12. Lindsay RM, Spanner E: A hypothesis: The protein catabolic rate is dependent upon the type and amount of treatment in dialyzed uremic patients. Am J Kidney Dis 13:382-389, 1989 13. Panzetta G: Protein intake does not depend on the dose of dialysis delivered, provided Kt/V is adequate. Nephrol Dial Transplant 10:2286-2289, 1995 14. Owen W Jr, Chertow G, Lazarus JM, Lowrie EG: Mortal risk and URR: Differences among demographic groups. J Am Soc Nephrol 8:207A, 1997 (abstr) 15. US Renal Data System: Researcher’s Guide to the USRDS Database. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Disease, Bethesda, MD, 1997 16. US Renal Data System: USRDS 1997 Annual Data Report. National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases. Bethesda, MD, 1997 17. Chertow GM, Lazarus JM, Lew NL, Ma L, Lowrie EG: Development of a population-specific regression equation to estimate total body water in hemodialysis patients [equation corrected by personal communication from authors]. Kidney Int 51:1578-1582, 1997