Normalized protein nitrogen appearance is correlated with hospitalization and mortality in hemodialysis patients with Kt/V greater than 1.20☆☆☆★

Methods 

Results 

Basic descriptive statistics for pertinent variables are presented in Tables 1 and 2. Table 1 summarizes pertinent continuous variables. Age of the patients averaged 53.4 ± 14 years, and the mean vintage (duration of chronic intermittent dialysis therapy) was 39.9 ± 41.7 months. The average dose of delivered Kt/V was 1.77 ± 0.34, with a range between 1.23 and 2.71. The mean measured nPNA was 1.13 ± 0.29 g/kg/d. The patients were hospitalized for 10.3 ± 26.6 days during the 12-month period of follow-up. Male and female patients were equal in number. Forty-five percent of patients were hospitalized at least once during the 12 months. The annual death rate was 9.8%.

Table 1. Demographic and laboratory data in 122 patients at the start of the cohort*

	Variable	Mean	SD	Minimum	Maximum
	Age (y)	53.4	14	23	89
	Vintage (dialysis mo)	39.9	41.7	1	213.8
	Time to death (mo)	11.1	2.4	1.3	12
	Hospitalization frequency (annual)	1.453	2.773	0	18.461
	Hospitalization days (annual)	10.3	26.6	0	188
	Time to first hospitalization (mo)	8.3	4.4	0.1	12
	Post-HD weight (kg)	68.6	19.3	33.4	152.7
	BMI (kg/m2)	25.6	5.2	16.3	46.1
	nPNA (g/kg/d)	1.13	0.29	0.5	2.15
	Kt/Vsp	1.77	0.34	1.23	2.71
	Urea reduction ratio (%)	76.6	5.8	64	93
	Serum Albumin (g/dL)	3.76	0.37	1.9	4.6
	Total Iron Binding Capacity (mg/dL)	200.2	45.4	81	346
	Hematocrit (%)	34.2	3.7	24	42.3
	Hemoglobin (g/L)	11.4	1.2	8.1	14.6
	Ferritin (μg/mL)	514	307	12	1,757
	Iron (μg/mL)	63.1	28.9	13	150
	Calcium (mg/dL)	9.46	0.75	7.1	11.3
	Phosphorus (mg/dL)	5.53	1.68	1.6	9.7
	Intact PTH (μ/mL)	269.7	243.4	6	1,233
	Creatinine (mg/dL)	9.63	2.83	3.3	16.9
	Urea Nitrogen (mg/dL)	57.6	16.1	21	100
	CO2 (meq/L)	25.9	3.6	17	35
	*Blood specimens were obtained immediately before a midweek hemodialysis treatment.

Note. Laboratory data for each patient are the average of up to 3 values during consecutive months (March through May, 2000).

Table 2. Nonparametric data of the cohort of 122 MHD patients

	Variable	%	n
	Gender (female)	50.0%	61
	Diabetes mellitus	45.1%	55
	Hospitalized at least once	45.1%	55
	Death rate	9.8%	12
	Renal transplantation rate	4.1%	5
	Left the study alive prior to the twelfth month	4.9%	6

Table 2 shows some non-parametric data of the cohort. Approximately 45% of the patients were diagnosed with diabetic nephropathy as their underlying kidney disease. During the 12-month follow-up, 55 patients were hospitalized overnight at least once, 12 patients died, 5 patients underwent renal transplantation, and 6 patients left the study because they changed their dialysis unit locations.

Table 3 shows the case-mix (multivariate) adjusted correlations controlled for age, race, gender, and diabetes. Among pertinent variables, nPNA and serum albumin and total iron binding capacity (TIBC) were the only values with statistically significant correlations with both hospitalization days and frequency. Correlation coefficients for albumin were generally stronger than for nPNA. There was no significant correlation between nPNA and Kt/Vsp (r = 0.09, Fig. 1).

View Large Image Download to PowerPoint

Fig. 1. Exploring the association between nPNA and Kt/Vsp in 122 MHD patients with a Kt/Vsp value > 1.20. No significant correlation existed between these 2 urea kinetic indices despite their known mathematical association (r = .09, P > .20). However, by dividing the patients into 2 distinct subgroups based on a Kt/Vsp cutoff of 1.50, there was a strong, significant correlation between Kt/V and nPNA for the lower Kt/Vsp values < 1.50 (r = .54, P < .001), whereas there was essentially no correlation at higher Kt/Vsp values (r = .03, P > .20).

Table 3. Case-mix (multivariate) adjusted correlation coefficients (controlled for age, race, gender and diabetes)

		Hospitalization days (HD)		Hospitalization frequency (HF)
		Correlation Coefficient (r)	P-value	Correlation Coefficient (r)	P-value
	nPNA	−0.18	.045	−0.26	.005
	Albumin	−0.29	.002	−0.43	<.001
	TIBC	−0.18	.044	−0.20	.028
	Creatinine	−0.14	.112	−0.20	.025
	Urea	−0.10	.263	−0.14	.128
	Hemoglobin	−0.09	.341	−0.17	.055
	BMI	−0.03	.721	−0.03	.747

Table 4 summarizes hospitalization rate ratios based on Poisson regression for all variables shown previously in Table 3.

Table 4. Hospitalization rate ratios (for HD and HF) by means of poisson regression analysis, adjusted for age, gender, race and diabetes, based on 1 decrement amounting to the standard deviation of the variable in question

		Hospitalization Frequency (HF)			Hospitalization Days (HD)
		Rate Ratio (95% CI)	P	R2	Rate Ratio (95% CI)	P	R2
	nPNA	1.74 (1.63-1.86)	<.001	0.1610	1.74 (1.47-2.05)	<.001	0.1183
	Albumin	1.71 (1.64-1.78)	<.001	0.2144	1.83 (1.63-2.05)	<.001	0.2010
	TIBC	1.75 (1.64-1.86)	<.001	0.1835	1.57 (1.33-1.86)	<.001	0.0968
	Creatinine	1.84 (1.73-1.97)	<.001	0.1636	2.08 (1.74-2.48)	<.001	0.1530
	BUN	1.32 (1.24-1.40)	<.001	0.1077	1.40 (1.19-1.64)	<.001	0.0699
	Hemoglobin	1.22 (1.16-1.29)	<.001	0.0994	1.34 (1.16-1.55)	<.001	0.0673

Table 5 describes the hazard ratios and 95% CIs of the time to HT with Cox proportional hazard models based on the initial values at the start of the prospective cohort study. The model controls for age, gender, race, and diabetes status to estimate the relative risks.

Table 5. Relative risk (RR) of time to first hospitalization (HT) for a standard deviation decrement of each variable, based on cox proportional hazard regression analysis, adjusted for age, gender, race and diabetes

		RR (95% CI)	P-Value	Pseudo-R2
	nPNA (for each I of one SD)	1.43 (1.06-1.93)	.019	0.0384
	Albumin	1.54 (1.16-2.04)	.003	0.0418
	Creatinine	1.37 (0.97-1.93)	.075	0.0326
	Urea	1.29 (0.96-1.73)	.089	0.0323

Table 6 lists the hazard ratios and 95% CI of death with Cox proportional hazard models in a similar approach as described previously for HT.

Table 6. Relative risk of death (D) for a standard deviation decrement of each variable, based on cox proportional hazard regression analysis, adjusted for age, gender, race and diabetes

		RR (95% CI)	P-Value	Pseudo-R2
	nPNA	3.29 (1.57-6.91)	.002	0.1550
	Albumin	2.47 (1.62-3.76)	<.001	0.1988
	TIBC	5.65 (2.44-13.05)	<.001	0.3083
	Creatinine	3.02 (1.54-5.92)	.001	0.1471
	Urea	2.59 (1.31-5.12)	.006	0.1267
	Hemoglobin	2.22 (1.29-3.81)	.004	0.1275

In summary, nPNA and serum albumin were the only two variables with statistically significant correlations with both mortality and all 3 measures of hospitalization, although serum TIBC and creatinine had some significant associations, which were occasionally even stronger than those for nPNA. However, serum TIBC and creatinine did not have a significant association with the time to first hospitalization (HT), and serum creatinine did not correlate significantly with the number of hospitalization days.

Discussion 

In this prospective study, we showed that the amount of daily protein intake, as measured by a urea kinetic index and expressed as nPNA, also known as nPCR, has a bearing on hospitalization and mortality in MHD patients whose dialysis dose was considered to be at least adequate. Moreover, we have confirmed previous studies19, 20, 21, 22, 23, 24, 25 indicating that serum albumin was a strong predictor of outcome, and that serum TIBC and creatinine were significantly correlated with mortality and most measures of hospitalization in this cohort of 122 MHD patients where Kt/V was greater than 1.20.

Among the measures of protein intake in MHD patients, net urea generation measurements, as determined by the urea nitrogen appearance or nPNA are ones that are most often used because they are easily measurable and mathematically accurate. The nPNA conveniently can be used to study populations of MHD patients in almost all dialysis units in the United States, because it is one of the monthly screening measures.

There are limitations to the preciseness with which the nPNA indicates the daily protein intake. First, the nPNA is dependent on the manufacturer's estimates of the dialyzer permeability characteristics and also on the accuracy of the measured blood and dialysate flow rates. Second, there are fluctuations of nPNA from day to day caused by changes in daily protein intake.26 Third, when daily protein intake is lower than 1.0 g/kg body weight (bw), it may be overestimated by nPNA.27 Conversely, when daily protein intake is higher than 1.0 g/kg bw, it may be underestimated by nPNA.27 Fourth, for the nPNA to accurately estimate the protein intake, the patient's protein metabolism should be at equilibrium or nearly so at the time of measurement.28 Fifth, the volume of distribution of urea, a measurement necessary for the calculation of nPNA, may be difficult to estimate accurately, particularly in obese, malnourished, or edematous patients.29 Finally, delayed equilibrium with subsequent urea rebound after dialysis, which can vary according to the patient and the characteristics of the dialysis procedure, may lead to an overestimate of the nPNA.30, 31

One of the reported limitations of nPNA is its mathematical coupling with Kt/V, which is another urea kinetic measure that is used to measure the dose of dialysis. One study32 showed that in a group of MHD patients when the Kt/V was increased from 0.82 to 1.33, there was an increase in nPNA from 0.83 to 1.00 g/kg/d, a rise in plasma albumin concentration from 3.5 to 3.9 g/dL, and a reduction in the gross annual mortality rate by over 50%. However, an alternative explanation is that the change in the measured nPNA may have been a consequence of a better dialysis dose and subsequently improved appetite with greater nutrient intake.33, 34, 35, 36, 37, 38, 39, 40 This important feature of nPNA could be ascribed to a reduced uremic toxicity and a mathematic coupling with Kt/V. Nevertheless, not all studies found a strong association between nPNA and Kt/V.41, 42, 43 In an attempt to minimize the possible effect of Kt/V or uremic toxicity on nPNA and, hence, calculated protein intake, we limited the patients in the present study to those with a Kt/V of greater than 1.20. We found no significant statistic association between nPNA and Kt/V in our study (r = 0.09) (Fig. 1). When we split our sample into 2 subgroups based on a Kt/V cutoff of 1.50, we found a strong and significant correlation between Kt/V and nPNA in the lower Kt/V group, but essentially none in the higher Kt/V group (Fig. 1). Therefore, it appears that for higher Kt/V values, the mathematic association between nPNA and Kt/V essentially disappears.

The nPNA has been shown to correlate with morbidity and mortality in MHD patients. In the National Cooperative Dialysis Study (NCDS), an nPNA greater than 1.0 g/kg/d and a timed average urea concentration of approximately 50 mg/dL were associated with low morbidity.42 Several studies indicate that nPNA predicted mortality or hospitalization.2, 44, 45, 46, 47 Our results are consistent with the above-mentioned findings, because we found that nPNA, along with serum albumin, was one of the only variables that had a significant association with both mortality and all 3 measures of hospitalization. However, it should be noted that the metabolic status of any given patient cannot be estimated by means of nPNA. It is not clear as to how many of our patients were in neutral nitrogen balance, catabolic, or anabolic. Other values such as simultaneous direct measurement of dietary nitrogen intake along with nPNA and their comparison would be required to evaluate the true metabolic status of an individual MHD patient.51 On the other hand, MHD patients are rarely in more than 1.0 to 1.5 g of negative or, particularly, positive nitrogen balance for more than a very few days.51 Thus, except for those exceptional times when an MHD patient is in strongly negative or positive nitrogen balance, the nPNA should rather closely reflect the dietary protein intake (eg, within ± 10%-15%).

A unique feature of the present study was that only rather well-dialyzed MHD patients were studied (ie, those with Kt/V >1.20 during 2 to 3 consecutive months). As discussed above, some data provide evidence that protein intake in MHD patients may rise with an increase in the delivered dose of dialysis.31, 48 Therefore, by studying only well dialyzed patients, we mitigated the confounding effect of the dose of dialysis or uremic status on the possible changes in appetite and nutritional intake.31 Moreover, we not only used the Kt/V value obtained during the initial month of this cohort study, but also ensured that the Kt/V during the month before and the month after the initiation of the study was greater than 1.20. Hence, all 122 patients in our cohort had indeed received an adequate dose of dialysis during the first 2 to 3 months of the beginning of the cohort study according to published clinical practice (K/DOQI) guidelines.13 Accordingly, we showed that even among those patients who received more than what is often believed to be an adequate dose of hemodialysis, the daily protein intake still had a strong and significant bearing on important clinical outcome measures.

One of the potential limitations of our study is the lack of follow-up of the initial values throughout the cohort, with the exception of the hospitalization and mortality data. However, cohort studies are not infrequently based on such an approach, because they examine the impact of baseline values (exposure) on the prospective outcome measures. Moreover, our nPNA data were based on measurements in 3 consecutive months. Another limitation is that no other measure of protein intake was utilized in addition to nPNA. Nevertheless, other previous studies already have shown the validation of nPNA as a measure of protein intake (see previously in this article).12, 49, 50, 51 Also, in our study, we did not use any specific measures of inflammation, such as C-reactive protein or cytokines, although it is generally believed that serum albumin and TIBC are negative acute phase reactants, as well as indicators of nutritional status. It is generally accepted that one of the mechanisms by which inflammation engenders adverse clinical events is by inducing anorexia and, hence, protein-energy malnutrition. Thus, the nPNA may itself be an indicator of inflammation. However, epidemiologic research as well as metabolic studies indicate that low protein intakes may, independently of the presence or absence of inflammation, be associated with poor clinical outcome.52, 53 Ling and Bistrian have observed that otherwise normal rats that were randomized to receive a diet low in protein and energy may experience a rise in serum acute phase proteins and cytokines which is not seen in control fed rats (Pei-Ra Ling, MD, and Bruce Bistrian, MD, personal communication, July, 2002).

Our study population has certain characteristics that may restrict the generalization of our findings to the general MHD population. First, our patients are almost 10 years younger than the MHD population in the United States according to the USRDS data of 2000.54 Second, over 30% of our patients are Hispanic compared to only 10.2% in the United States.54 Third, our dialysis time is relatively high (ie, 4 to 5 hours), with Kt/V values > 1.20 in over 94% of our patients (123 out of 135 MHD patients). Taken altogether, it is not surprising to observe significantly lower mortality and hospitalization rates in our study population as compared with the USRDS. Such characteristics generally weaken the existing associations between risk factors (such as nPNA) and the outcome of interest. Therefore, our observed associations might have been even stronger if we had had older and sicker patients with higher mortality and hospitalization rates.

A secondary finding in our study was the significant association between both serum TIBC and serum creatinine and most, but not all, of the clinical outcome measures. As indicated above, serum TIBC is a both nutritional marker55, 56 and acute phase reactant,57, 58, 59, 60, 61, 62 and some previous studies have shown that it is an outcome predictor in ESRD patients.63 In MHD patients, the predialysis serum creatinine reflects both their skeletal muscle mass and skeletal muscle (meat) intake as well, of course, as their dialysis treatment. A higher predialysis serum creatinine also has been shown to be paradoxically associated with a better outcome, a phenomenon that is referred to as reverse epidemiology of cardiovascular risk factors in ESRD patients.64

In summary, the results reported in the present study provide evidence that even in MHD patients who receive adequate to high dialysis doses (Kt/V > 1.20), indicators of protein nutrition (nPNA and serum albumin) predict prospective hospitalization and mortality. Serum creatinine, TIBC (transferrin), urea nitrogen, and hemoglobin are also significant predictors of mortality, but they were not shown to have statistically significant correlations with all 3 of the hospitalization indices studied. The results of this study are consistent with the notion that nutritional intake and nutritional status influence morbidity and mortality in well-dialyzed MHD patients. As nPNA is only a surrogate measure of protein intake, it would be helpful if our findings were confirmed in studies where protein intake is measured more directly (eg, by means of dietary records or food frequency questionnaires).65 Nevertheless, previous studies have shown the validity of the nPNA as a measure of protein intake in MHD patients, even though the validity of nPNA at high doses of dialysis still needs to be investigated more thoroughly. A definitive answer to the question of whether nutritional intake or malnutrition influences clinical outcome in MHD patients may be obtained only by studying MHD individuals who are randomly assigned to nutritional interventions. Until randomized clinical trials are performed, such epidemiologic studies as ours may help elucidate the associations between the measures of nutrition and outcome in the rapidly growing ESRD population.