Ultrafiltration Rate Levels in Hemodialysis Patients Associated with Weight-Specific Mortality Risks

Visual Abstract

Abstract

Background

We hypothesized that the association of ultrafiltration rate with mortality in hemodialysis patients was differentially affected by weight and sex and sought to derive a sex- and weight-indexed ultrafiltration rate measure that captures the differential effects of these parameters on the association of ultrafiltration rate with mortality.

Methods

Data were analyzed from the US Fresenius Kidney Care (FKC) database for 1 year after patient entry into a FKC dialysis unit (baseline) and over 2 years of follow-up for patients receiving thrice-weekly in-center hemodialysis. To investigate the joint effect of baseline-year ultrafiltration rate and postdialysis weight on survival, we fit Cox proportional hazards models using bivariate tensor product spline functions and constructed contour plots of weight-specific mortality hazard ratios over the entire range of ultrafiltration rate values and postdialysis weights (W).

Results

In the studied 396,358 patients, the average ultrafiltration rate in ml/h was related to postdialysis weight (W) in kg: 3W+330. Ultrafiltration rates associated with 20% or 40% higher weight-specific mortality risk were 3W+500 and 3W+630 ml/h, respectively, and were 70 ml/h higher in men than in women. Nineteen percent or 7.5% of patients exceeded ultrafiltration rates associated with a 20% or 40% higher mortality risk, respectively. Low ultrafiltration rates were associated with subsequent weight loss. Ultrafiltration rates associated with a given mortality risk were lower in high–body weight older patients and higher in patients on dialysis for more than 3 years.

Conclusions

Ultrafiltration rates associated with various levels of higher mortality risk depend on body weight, but not in a 1:1 ratio, and are different in men versus women, in high–body weight older patients, and in high-vintage patients.

Introduction

It is now well established that higher ultrafiltration rates during a hemodialysis session are associated with a higher risk of death.^1–5 One currently implemented ultrafiltration rate “warning” level is scaled to body weight as 13 ml/h per kg. This may disadvantage larger patients because the mortality risk of exceeding that particular level is higher in heavier patients.⁶ In a cohort of approximately 2500 incident patients, Raimann et al. examined associations between the weight-specific mortality hazard ratio and the ultrafiltration rate using a variety of body size scalings and found that the mortality hazard ratio associated with the ultrafiltration rate scaled to body weight in a 1:1 ratio (ultrafiltration rate/kg) was markedly higher in heavier patients.⁷ For example, mortality hazard ratio when exceeding an ultrafiltration rate threshold of 13 ml/kg per hour was 1.20, 1.45, or >2.0 for a 60-, 80-, or 100-kg patient, respectively.

In our previous analysis,⁷ we did not propose a substitute ultrafiltration rate level for the currently accepted 13 ml/h per kg quality assurance metric. In the present analysis, we repeated the analytic strategy of our previous report in a much larger patient dataset to explore further the associations of ultrafiltration rate with mortality in hemodialysis patients and to identify potential new candidate ultrafiltration rate warning levels.

Methods

Study Design

This retrospective, observational cohort study was conducted in patients commencing thrice-weekly maintenance hemodialysis in clinics of the Fresenius Medical Care system between January 1, 2015, and May 30, 2020. Data from electronic health records that captured all clinical and treatment-related data from patients receiving treatment were extracted in a deidentified format. Patients older than 18 years, receiving thrice-weekly hemodialysis, and who survived at least 12 months were included in this analysis. From the extracted dataset, we defined the first 12 months as the baseline period and the following 2 years as the follow-up period. The Western Institutional Review Board deemed the study exempt from human subject consent requirements and continuing review.

Measurements

For the primary analysis, baseline parameters were determined as the mean of all available entries during the first 12 months of dialysis in a Fresenius Medical Care dialysis unit. Missing data were excluded from the computation of the mean and the SD. The ultrafiltration rate scaled to weight was calculated as the ultrafiltration rate divided by the postdialysis body weight. Interdialytic weight gain was calculated as the mean of every predialysis weight minus the preceding postdialysis weight (W). All parameters were routinely collected in the electronic health records and extracted for the purpose of analysis. Race data were based on patient self-identification as recorded in the Fresenius North America and Renal Research Institute patient database.

Statistical Analysis

Data are reported as mean±SD unless stated otherwise. Over the 2-year follow-up period, all-cause mortality was assessed from the electronic health record. We fit Cox proportional hazard models with unscaled or scaled ultrafiltration rate and postdialysis weight as continuous independent variables, adjusting for sex, race, age, diabetic status, albumin, predialysis systolic BP, and phosphate.

We included a bivariate tensor product spline function of ultrafiltration rate and postdialysis weight in the Cox proportional hazard model⁸ to explore potential nonlinear effects and their interaction. The bivariate function explores the main effects of and interaction between ultrafiltration rate and postdialysis weight without assuming any specific form. We used the following steps to construct a continuous version of relative mortality risk: (1) fit a linear regression model with unscaled or scaled ultrafiltration rate as the response and postdialysis weight as the predictor; (2) construct a contour plot of mortality hazard ratios—the contour lines represent values of postdialysis weight and ultrafiltration rate for which the mortality hazard ratio equals a constant; and (3) for each fixed postdialysis weight, compute mortality hazard ratios at different ultrafiltration rates with the expected ultrafiltration rate at this postdialysis weight based on the linear model in the first step as the reference point.

The statistical software R 4.1.3, codename “One Push-Up” with packages dplyr, tidyr, survival, mgcv, survminer, mltools, tableone, and doBy, was used for all analyses.⁹ A P value <5% was considered significant.

Results

The demographics and basic clinical characteristics of the 396,358 patients studied are shown in Table 1. The data presented in Table 1 are averages of all values present in the electronic medical record during the baseline year 1. A histogram of year‐1 average postdialysis weights is shown in Supplemental Figure 1. A histogram of patient vintage at time of entry is shown in Supplemental Figure 2.

Table 1.

Patient characteristics

Number of patients	396,358
Male, count (%)	228,346 (58)
Age	62±15
Diabetes (%)	162,879 (41)
Race (self-identified), count (%)
Other	1 (0)
White	193,570 (49)
Black	100,642 (25)
Unknown	87,463 (22)
Asian	9016 (2.3)
Native Hawaiian/Other Pacific Islander	2911 (0.7)
American Indian or Alaskan Native	2755 (0.7)
Treatment and laboratory values (mean±SD)
Ultrafiltration rate, ml/h	582±226
Scaled ultrafiltration rate, ml/h per kg	7.29±2.97
Session length, min	224±24
Kt/V	1.61±0.27
Urea reduction ratio (%)	74±6
Predialysis systolic BP, mm Hg	147± 18
Predialysis serum albumin, g/dl	3.7±0.4
Predialysis serum phosphate, mg/dl	5.2±1.2
Interdialytic weight gain, kg	2.1±1.1
Predialysis weight, kg	85.2±24
Postdialysis weight, kg	83.2±23
Body mass index, kg/m2	29.1±7.5

Treatment, BP, and laboratory values listed are based on patient average values calculated from all data available in the electronic medical record during baseline year 1.

During the 2-year follow-up period (totaling 281,915 during the first and 146,638 patient-years during the second year of follow-up), 68,323 patients (39,237 during the first and 28,940 during the second year of follow-up) died. The resulting mortality rates for the first and the second year of follow-up were 13.9 and 19.7 deaths per 100 patient-years, respectively.

Figure 1 shows contour plots of mortality hazard ratios as a function of postdialysis weight on the vertical axis and unscaled (ultrafiltration rate) on the horizontal axis. In the contour plots, weight-specific mortality hazard ratios are shown as colored bands. A mortality hazard ratio of 1.0 represents average mortality risk for a particular weight stratum. From these contour plots, one can extract the ultrafiltration rate value associated with a 20% or 40% higher weight-specific mortality hazard ratio as shown. Supplemental Figure 3 shows similarly configured contour plots for men and women separately. Mortality was adjusted for race, age, diabetes, sex (in the combined plot), and baseline year average values for predialysis systolic BP, predialysis serum phosphate, and predialysis serum albumin.

Figure 1.

Contour plots of mortality hazard ratios as a function of postdialysis weight on the vertical axis and unscaled ultrafiltration rate on the horizontal axis. The color scheme indicates discrete levels of weight-specific mortality hazard ratio. The analysis was adjusted for age, sex, diabetes, race, and baseline year average values of predialysis serum albumin, phosphate, and systolic BP. The isopleths (lines of equal risk) corresponding to mortality hazard ratios of 1.0 (average risk), 1.2 (+20% risk), and 1.4 (+40% risk) are highlighted in red. MHR, mortality hazard ratio; UFR, ultrafiltration rate.

The regression equations for the contour plot isopleths (lines of equal risk) corresponding to weight-specific mortality hazard ratios of 1.0, 1.2, and 1.4 are shown in Table 2. The average ultrafiltration rate values (mortality hazard ratio=1.0) for both sexes combined could be estimated as 3.14W+320 ml/h, or more simply as 3W+330. The ultrafiltration rate values corresponding to mortality hazard ratios of 1.2 could be estimated as 3W+540 ml/h for men and 3W+470 ml/h for women. The ultrafiltration rate values corresponding to mortality hazard ratios of 1.4 could be estimated as 3W+670 for men and 3W+600 for women.

Table 2.

Ultrafiltration rates (in ml/h) corresponding to mortality hazard ratios of 1.0, 1.2, and 1.4

	Mortality Hazard Ratio=1.0	Mortality Hazard Ratio=1.2	Mortality Hazard Ratio=1.4
All	3.14W+320	3.09W+500	3.23W+630
Men	3.07W+340	2.91W+540	3.08W+670
Women	2.79W+320	2.96W+470	3.03W+600

W=average postdialysis weight in kg. Regression equations based on contour plot analyses shown in Figure 1 for all patients and Supplemental Figure 3, A and B for men and women separately.

Supplemental Figure 4 shows contour plots of weight-specific mortality hazard ratio by ultrafiltration rate scaled to body weight in a 1:1 ratio (as ultrafiltration rate/kg) for different body weights. Data for men and women were combined. The vertical line shows the current quality assurance “warning” value of 13 ml/h per kg.

Figure 2 shows the percent of patients by weight decile who would exceed ultrafiltration rate levels associated with mortality hazard ratios of 1.2 (blue bars) or 1.4 (green bars) or the value of 13 ml/h per kg (red bars). Across all weight deciles, approximately 19% of patients exceeded an ultrafiltration rate associated with a mortality hazard ratio of 1.2, while approximately 7.5% exceeded on average an ultrafiltration rate associated with a mortality hazard ratio of 1.4.

Figure 2.

Percent of patients (by sex and decile of postdialysis weight) who exceed ultrafiltration rate levels associated with a mortality hazard ratio of 1.2 (blue bars) and 1.4 (green bars). The percent exceeding a warning value of 13 ml/h per kg (red bars) is also shown. Weight deciles for men and women were calculated separately. For women, the first, fifth, and 10th decile median body weights were 47.6, 70.6, and 120.0 kg, respectively. For men, the first, fifth, and 10th decile median body weights were 58.1, 81.3, and 130.6 kg, respectively. MHR, mortality hazard ratio, weight (in the legend equatons) is in kg; UFR, unscaled ultrafiltration rate in ml/h.

In Table 3, for those patients exceeding each of the two ultrafiltration rate values, we show the average increase in dialysis session length that would be required to lower the ultrafiltration rates to values associated with mortality hazard ratios of 1.2 or 1.4. The increases in dialysis time needed did not vary substantially by weight nor by sex and averaged 37–40 minutes for the 19% or so patients who exceeded the 1.2 mortality hazard ratio ultrafiltration rate and 27–30 minutes for the 7.5% or so patients who exceeded the 1.4 mortality hazard ratio ultrafiltration rate. Table rows by sex are subdivided according to decile of postdialysis weight. Values for weight decile 1 (lightest group), decile 5, 6 (average weights), and decile 10 (heaviest group) are shown.

Table 3.

Time needed to no longer exceed (sex-adjusted) ultrafiltration rate values associated with mortality hazard ratio of 1.2 or 1.4, by sex- and weight-selected decile

Men by Weight Decile	Average Ultrafiltration Rate (ml/h)	Average Weight (kg)	Ultrafiltration Rate Warning Level=3W+540 ml/h (Mortality Hazard Ratio=1.20)			Ultrafiltration Rate Warning Level=3W+670 ml/h (Mortality Hazard Ratio=1.40)
			Average Dialysis Session Length (min) in Patients Not Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) in Patients Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) Increase Needed to No Longer Exceed Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) in Patients Not Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) in Patients Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) Increase Needed to No Longer Exceed Ultrafiltration Rate Warning Level
1	517	57	215	208	37	215	204	29
5	592	81	228	223	37	227	221	30
6	605	86	230	227	37	230	225	29
10	759	136	251	247	37	251	245	28

Women by Weight Decile	Average Ultrafiltration Rate (ml/h)	Average Weight (kg)	Ultrafiltration Rate Warning Level=3W+470 ml/h (Mortality Hazard Ratio=1.20)			Ultrafiltration Rate Warning Level=3W+600 ml/h (Mortality Hazard Ratio=1.40)
			Average Dialysis Session Length (min) in Patients Not Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) in Patients Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) Increase Needed to No Longer Exceed Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) in Patients Not Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) in Patients Exceeding Ultrafiltration Rate Warning Level	Average Dialysis Session Length (min) Increase Needed to No Longer Exceed Ultrafiltration Rate Warning Level
1	447	47	203	195	40	202	191	30
5	521	71	217	211	39	216	207	30
6	536	76	219	213	38	219	210	29
10	672	125	236	233	37	236	231	27

W, postdialysis weight.

Table 4 shows the observed average ultrafiltration rates for patients of differing body weight and ultrafiltration rate rates associated with the two proposed “warning” values. Values of maximum allowed fluid removal over a 4-hour dialysis session associated with each “warning” value are shown, as well as the estimated weight-specific mortality hazard ratios.

Table 4.

Average ultrafiltration rate, maximum allowed ultrafiltration rate, and maximum allowed weight removal over a 4-hour treatment, as well as different mortality hazard ratios associated with different candidate ultrafiltration rate warning levels: 3W+500 ml/h (+20% mortality risk), 3W+630 ml/h (+40% mortality risk), and 13 ml/h per kg (weight-dependent mortality risk)

Weight	Observed Average Ultrafiltration Rate (ml/h)	Ultrafiltration Rate 3W+500 (ml/h)		Ultrafiltration Rate 3W+630 (ml/h)		Ultrafiltration Rate 13 ml/h per kg		Ultrafiltration Rate 3W+500 (ml/h)	Ultrafiltration Rate 3W+630 (ml/h)	Ultrafiltration Rate 13 ml/h per kg
	ml/h	Ultrafiltration Rate (ml/h)	4-h Weight Removal	Ultrafiltration Rate (ml/h)	4-h Weight Removal	Ultrafiltration Rate (ml/h)	4-h Weight Removal	Weight-Specific Mortality Hazard Ratio
50	480	650	2.6	785	3.1	650	2.6	1.2	1.4	1.1
60	500	680	2.7	815	3.3	780	3.1	1.2	1.4	1.4
80	570	740	3.0	875	3.5	1040	4.2	1.2	1.4	1.6
100	630	800	3.2	935	3.7	1300	5.2	1.2	1.4	1.9
120	700	860	3.4	995	4.0	1560	6.2	1.2	1.4	>2.0
140	760	920	3.7	1055	4.2	1820	7.3	1.2	1.4	>2.0

W, postdialysis weight.

We also constructed contour plots for patients grouped by age ranges: 19–40, 41–70, and 71+ years. The contour plots are shown in Supplemental Figure 5. The ultrafiltration rates associated with a 1.2 mortality hazard ratio in these three age-specific contour plots are shown in Figure 3. The ultrafiltration rates associated with a 1.2 mortality hazard ratio were similar in the two younger age groups but were substantially lower for high–body weight patients in the 71+ age group.

Figure 3.

Ultrafiltration rates associated with a 20% higher weight-specific mortality hazard ratio (mortality hazard ratio) by age range. The regression equations for each group in the 50–160 kg weight range are shown. The 3W+500 ultrafiltration rate level for all patients is depicted by the thick gray line. W, postdialysis weight.

We performed a similar contour plot analysis for patients grouped by vintage at time of entry into the Fresenius Kidney Care system: <12 months, 12–36 months, and >36 months. The contour plots are shown in Supplemental Figure 6. The ultrafiltration rates associated with a 1.2 mortality hazard ratio in these three age-specific contour plots are shown in Figure 4. The ultrafiltration rates associated with a 1.2 mortality hazard ratio were similar in the two shorter vintage subgroups but were substantially higher for patients with vintage >36 months.

Figure 4.

Ultrafiltration rates associated with a 20% higher weight-specific mortality hazard ratio by vintage. The regression equations for each group in the 50–160 kg weight range are shown. The 3W+500 ultrafiltration rate level is depicted by the thick gray line.

To evaluate whether a low ultrafiltration rate was associated with subsequent weight loss, we averaged body weight during months 9–12 of the baseline period (year 1) and then compared weights during 3-month periods 1 and 2 years later (months 9–12 of follow-up years 1 and 2, respectively). We then examined the impact of average ultrafiltration rate during baseline months 9–12 on the subsequent weight change by calculating weight change by ultrafiltration rate decile. Finally, we examined this relationship between ultrafiltration rate decile and subsequent weight loss in various tranches of postdialysis body weight, ranging from 40 to 50 kg through >110 kg. The results are shown in Supplemental Figure 7. For all weight categories, weight change was less positive when the ultrafiltration rate was 700 ml/h or less.

We performed additional sensitivity analyses adjusting the mortality contour plots for systolic BP (predialysis, postdialysis, and delta) and for height. These adjustments did not noticeably change the contour plots (Supplemental Figures 8–12).

Discussion

Our results confirm our previous findings in a much smaller patient cohort describing the relationship between unscaled ultrafiltration rate (ultrafiltration rate) and weight-specific mortality hazard ratio (mortality hazard ratio) in hemodialysis patients⁷: Average ultrafiltration rate values were associated with body weight, but not in a 1:1 ratio. Ultrafiltration rate values associated with a 20% or 40% higher weight-specific mortality risk could be estimated by simple equations of the form: 3W+x, where values for x associated with average mortality risk, +20% higher risk, and +40% higher risk were 330, 500, and 630 ml/h, respectively. There was a differential effect of sex: Ultrafiltration rate values associated with a 20% or 40% higher risk in men were on average 70 ml/h higher than those in women (Table 2). In patients older than 70 years, the ultrafiltration rates associated with mortality hazard ratio of 1.2 were lower than in the two younger cohorts but only for the heaviest patients. Our results suggest that in older patients weighing 110–160 kg, the weight-specific mortality hazard ratios of 1.2 are exceeded when the ultrafiltration rate is above 800–850 ml/h.

In our previous data, we showed exceeding a given ultrafiltration rate/kg value is associated with markedly higher mortality risk in patients of higher body weight compared with patients of lower body weight,⁷ and this nonuniformity of risk was confirmed by the present analysis in a much larger patient cohort. In addition, the percent of patients in our cohort exceeding an average ultrafiltration rate value of 13 ml/h per kg was quite small in all but the lowest deciles of body weight.

Approximately 19% of patients had average ultrafiltration rate values that were associated with a 20% or higher mortality risk, and approximately 7.5% had average ultrafiltration rate values that were associated with a 40% greater mortality risk. Assuming no change in interdialytic weight gain, the average increases in dialysis session length that would be required (in 19% of patients) to no longer exceed a +20% morality risk was approximately 40 min. The average increase in dialysis session length that would be required (in 7.5% of patients) to no longer exceed a +40% mortality risk was approximately 30 minutes.

In our previous analysis of unscaled ultrafiltration rate versus weight-specific mortality hazard ratio factored by body size, it seemed like the risk of higher ultrafiltration rate values was attenuated in smaller patients (40–60 kg). The reasons for this were unclear, and we speculated that ultrafiltration rate, which associates very closely on average with interdialytic weight gain,⁷ when low, might be a marker for poor dietary intake, and that this might be of greatest importance in lighter patients, who have a low body mass index.⁷ In the present analysis, this attenuation of risk between ultrafiltration rate and mortality hazard ratio at low body weights was present but less apparent. To evaluate the hypothesis that low ultrafiltration rate might mark inadequate nutritional intake, we examined weight change during years 1 and 2 of follow-up compared with baseline values as a function of ultrafiltration rate in patients grouped by body weight. Our results (see Supplemental Figure 6) suggested that average ultrafiltration rate value is associated with future weight change, with lower ultrafiltration rate values in the lightest patients being predictive of future weight loss. Thus, a low ultrafiltration rate might be associated with a benefit in cardiovascular health but may mark inadequate nutrition, which in turn might be associated with a competing risk of death. This association merits further study.

Our ultrafiltration rate versus mortality hazard ratio analysis by vintage showed that in patients with a vintage >36 months, the ultrafiltration rate associated with a 1.2 weight-specific mortality hazard ratio was substantially higher than in patients with vintage <36 months. We believe that the explanation for this finding is straightforward. High-vintage patients have a higher ultrafiltration rate on average compared with their lower vintage counterparts, as shown in the contour plots in Supplemental Figure 6. If the average ultrafiltration rate in the high-vintage subgroup is higher, the ultrafiltration rate associated with a 20% higher mortality hazard ratio will also be greater.

Our study has several strengths and some limitations. We averaged ultrafiltration rate and body weight over a 12-month period. It is possible that additional risk might be related to variability of the ultrafiltration rate. Episodically, high values of ultrafiltration rate have been reported to be associated with worsened outcome.¹⁰

The exclusion of patients who died within the first year potentially may have introduced bias and reduced generalizability. The patients in our study were all being dialyzed in the United States. Usvyat et al. have shown that average interdialytic weight gains and session lengths differ among patients being dialyzed in the Asia Pacific region and Europe compared with those dialyzed in North America.¹¹ Accordingly, it is possible that the relative mortality risks associated with specific ultrafiltration rates in Europe and the Asia Pacific countries differ from relative risk in patients dialyzed in North America. Furthermore, average ultrafiltration rates in various parts of the world have been changing such that they are now markedly lower compared with values measured 20 or more years ago. For example, in the Hemodialysis (HEMO) Study,³ the mean ultrafiltration rate was 12.1 ml/h per kg, and in a study by Movilli et al. from Europe published in 2007, the mean ultrafiltration rate was 12.7 ml/h per kg.² The current quality assurance ultrafiltration rate guideline focusing on a rate of 13 ml/h per kg was partly based on these two studies. Contrast this with the average ultrafiltration rate of only 7.3 ml/h per kg (Table 1) in the present US Fresenius Kidney Care dataset.

As with all observational studies, residual confounding may be a factor. High ultrafiltration rates have been associated with occurrence of intradialytic hypotension, and the adverse effects of high ultrafiltration rates might be limited to patients with frequent intradialytic hypotension. We did not study interactions between these two risk factors.

Another limitation of our study was that residual urine output was not routinely measured. The presence of residual kidney function contributes importantly to survival in kidney failure patients.¹² Residual urine output is associated with lower ultrafiltration volume.¹³ We attempted to get some insight in this area by comparing ultrafiltration rate versus mortality hazard ratio curves in patients of varying dialysis vintage (vintage=total length of time on dialysis). In patients on dialysis for more than 3 years, when the proportion and amount of residual kidney function would be expected to be reduced, average ultrafiltration rates were longer compared with those in lower vintage patients, and the ultrafiltration rate estimate associated with a 20% or 40% higher weight-specific mortality hazard ratio (Figure 4) was proportionally higher. Thus, some of the “ultrafiltration risk” suggested by mortality curves uncontrolled for urine output may not be due to volume-related factors at all; rather, a higher ultrafiltration rate may be a marker of low or absent residual kidney function.

Our data suggest (Table 3) that the dialysis session lengths in patients exceeding the two candidate ultrafiltration rate warning levels were only 5–10 minutes shorter than the session lengths in patients whose ultrafiltration rates were below the warning values. A substantial amount of additional dialysis time (30–40 minutes) would be required on average to reduce the ultrafiltration rates below these warning values. Thus, in the absence of a successful attempt to reduce interdialytic weight gain, substantial increases in dialysis session length would be required (Table 3). The difficulty in effecting even a modest (e.g., 30 minutes) increase in dialysis session length in the United States while staying within a 3/week dialysis schedule was illustrated by the difficulties experienced in the Time to Reduce Mortality in ESRD (TiME) trial.¹⁴

One approach to provide the required increase in dialysis time would be to add a fourth session of dialysis. Such an approach would not only decrease the average weekly ultrafiltration rate value but also it would reduce the ultrafiltration rate after the long weekend interval, when mortality rates are especially high.^15,16 Another approach would rely on moving patients with high interdialytic weight gains to home hemodialysis, where the logistics of increasing dialysis session length or adding additional weekly treatments should be easier to manage.

The most important limitation of our study is its observational design. There are no large randomized trials to date that have evaluated the risk-benefit ratio of reducing ultrafiltration rate. Attempts to reduce ultrafiltration rate may lead to worsening of fluid overload¹⁷ with inability to achieve dry weight,¹⁸ and fluid overload has been shown per se to be an important risk factor for mortality.¹⁹ Achievement of dry weight may be more important to lowering cardiovascular risk than lowering ultrafiltration rate per se. In addition, focusing on not exceeding a particular ultrafiltration rate “warning” level may be overly simplistic. There does not seem to be an inflection point in the risk curve depicting the association between ultrafiltration rate and mortality risk. Rather, the relationship seems to be continuous and does not seem to plateau, even at relatively low ultrafiltration rate values. Thus, one could make an argument to discard any particular single warning level of ultrafiltration rate and instead to simply try to minimize the ultrafiltration rate. On the other hand, our data did suggest that the lowest ultrafiltration rates are associated with weight loss in the lightest patients, who have low body mass index and may be nutritionally vulnerable to further weight loss.

In summary, our data suggest that one currently popular quality assurance metric for ultrafiltration rate, namely not to exceed 13 ml/h per kg, is not optimal. Our study did suggest new candidate warning levels that are associated with similar levels of mortality risk across patients with a broad spectrum of body weights and which currently are being exceeded by similar percentages of patients of different body weights. The magnitude of risk directly attributable to ultrafiltration rate may be overstated because of potential confounding between high ultrafiltration rate and absent or reduced residual kidney function, and the risk-benefit ratio of attempting to limit ultrafiltration rates needs to be studied using interventional, randomized controlled designs.

Supplementary Material

cjasn-18-767-s001.pdf ^{(1.4MB, pdf)}

Disclosures

J.T. Daugirdas reports consultancy for Fresenius Medical Care (paid grant reviewer for RRIs extramural grant program) and Unicycive (makers of a lanthanum-containing phosphate binder); stock ownership in Amazon, C3Ai, JD, Mercado Libre, and SE; patent regarding holographic control module for hemodialysis machine (by self); and serving as a journal editor for International Society for Hemodialysis. P. Kotanko reports employment with Renal Research Institute, a wholly owned subsidiary of Fresenius Medical Care; stock in Fresenius Medical Care; research funding from Fresenius Medical Care, NIH, and KidneyX; author honoraria from Henry Stewart Talks; multiple patents in the kidney space; and advisory or leadership roles on the Editorial Boards of Blood Purification, Frontiers in Nephrology, and Kidney and Blood Pressure Research. A. Mermelstein reports employment with Renal Research Institute, a wholly owned subsidiary of Fresenius Medical Care. J.G. Raimann reports employment with Renal Research Institute, a wholly owned subsidiary of Fresenius Medical Care, and owns shares of stock in Fresenius Medical Care. He serves as a member of the Board of Directors for “Easy Water for Everyone” (501c3). The remaining author has nothing to disclose.

Funding

None.

Author Contributions

Conceptualization: John T. Daugirdas, Peter Kotanko.

Data curation: Peter Kotanko, Ariella Mermelstein, Jochen G. Raimann.

Formal analysis: Ariella Mermelstein, Jochen G. Raimann, Yuedong Wang.

Methodology: Ariella Mermelstein, Jochen G. Raimann, Yuedong Wang.

Project administration: Peter Kotanko, Jochen G. Raimann.

Supervision: Peter Kotanko.

Writing – original draft: John T. Daugirdas.

Writing – review & editing: John T. Daugirdas, Peter Kotanko, Ariella Mermelstein, Jochen G. Raimann.

Data Sharing Statement

Anonymized data can be shared on writing to Dr. Peter Kotanko from the Renal Research Institute.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/B659.

Supplemental Figure 1. Histogram of average postdialysis weights during the 1-year baseline period.

Supplemental Figure 2. Histogram of dialysis vintage at time the studied patients entered into the dialysis organization data system.

Supplemental Figure 3. (A) Unscaled ultrafiltration rates (UFR) versus weight-specific mortality hazard ratios (MHR) for men: contour plots of mortality hazard ratios as a function of postdialysis weight on the vertical axis and unscaled ultrafiltration rate on the horizontal axis. The color scheme indicates discrete levels of weight-specific mortality hazard ratio. The analysis was adjusted for age, sex, diabetes, race, and baseline year average values of predialysis serum albumin, phosphate, and systolic BP. (B) Unscaled ultrafiltration rates (UFR) versus weight-specific mortality hazard ratio (MHR), women only. Same adjustments as for Figure 3A.

Supplemental Figure 4. Ultrafiltration rate scaled to body weight in a 1:1 ratio (ultrafiltration rate/kg) versus weight-specific mortality hazard ratio (MHR) for all patients. One suggested quality assurance warning level of 13 ml/hr per kg is depicted by the vertical line.

Supplemental Figure 5. Unscaled ultrafiltration rates (UFR) versus weight-specific mortality hazard ratio (MHR) for men and women combined, for three age ranges: 19–40, 41–70, and 71+ years. Contour plots of mortality hazard ratios as a function of postdialysis weight on the vertical axis and unscaled ultrafiltration rate on the horizontal axis. The color scheme indicates discrete levels of mortality hazard ratios. The analysis was adjusted for age, sex, diabetes, race, predialysis serum albumin, phosphate, and systolic BP.

Supplemental Figure 6. Unscaled ultrafiltration rates (UFR) versus weight-specific mortality hazard ratio (MHR) for men and women combined, for three vintage ranges: <12 months, 12–36 months, and 36+ months. Contour plots of mortality hazard ratios as a function of postdialysis weight on the vertical axis and unscaled ultrafiltration rate on the horizontal axis. The color scheme indicates discrete levels of mortality hazard ratios. The analysis was adjusted for age, sex, diabetes, race, predialysis serum albumin, phosphate, and systolic BP.

Supplemental Figure 7. Weight change (months 9–12 of year 3 [follow-up year 2] or months 9–12 of year 2 [follow-up year 1] versus months 9–12 of baseline [year 1]) by ultrafiltration rate (UFR) decile. Data analyzed separately for various ranges of postdialysis weight.

Supplemental Figure 8. Unscaled ultrafiltration rates (UFR) versus weight-specific mortality hazard ratios for men and women combined, with no adjustment for systolic BP (8A) or with adjustment for predialysis systolic BP (8B), postdialysis systolic BP (8C), and delta systolic BP (8D).

Supplemental Figure 9. Unscaled ultrafiltration rates (UFR) versus weight-specific mortality hazard ratios (MHR) for men and women combined, limited to patients in the lowest tertile of systolic BP.

Supplemental Figure 10. Scatterplot of body mass index (kg/m²) versus postdialysis body weight.

Supplemental Figure 11. Scatterplot of body height versus postdialysis body weight.

Supplemental Figure 12. Unscaled ultrafiltration rates (UFR) versus weight-specific mortality hazard ratios (MHR) for men and women combined, adjusted for height.

STROBE Statement.