Introduction
Both high and low serum phosphate levels in patients on hemodialysis (HD) have been associated with an increased risk on mortality.1,2 This bidirectional or U-shaped relation with mortality is also observed in patients on HD for some other parameters, such as systolic blood pressure.3 Next to this, previous studies showed that there can be a wide variability between serial predialysis serum phosphate levels, even to such a degree that it can influence clinical decision making.4 There are limited data on the relation between variability of serum phosphate and outcome in patients on HD, although two studies showed that, respectively, a higher SD or coefficient of variation (CV) of serum phosphate was related to increased mortality.5,6 Given the relation between changes in serum phosphate and other nutritional markers such as albumin before hospitalization or death,7,8 it is relevant to study the relation between phosphate variability and outcome in the context of other nutritional markers. Different metrics of variability, such as SD, CV, or SD of the residuals (of an individual fitted regression function), were used to study the relationship with outcomes.9,10 However, in the case of serum sodium, we previously showed that the directional range (DR), the difference between minimal and maximal levels during a predefined period, had the strongest relation with outcome.11 The aim of this study was to assess, in a large and diverse cohort of US patients on HD, the relation between variability in serum phosphate and all-cause mortality, in the context of different baseline levels of serum phosphate and serum albumin.
Materials and Methods
Study Design
In this retrospective cohort study, all incident HD patients 18 years or older and treated in US Fresenius Kidney Care (FKC) clinics between January 1, 2010, and October 31, 2018, were included. Patients were required to be treated in FKC clinics within 90 days of their first date on dialysis and survive the baseline period. The study baseline was defined as months 0–6 on HD; the follow-up period was defined as the subsequent 12 months (i.e., months 7–18). Clinical and laboratory parameters were averaged during the baseline period. Demographic and comorbidity data were obtained in the first month of dialysis from medical records. Deaths were identified by Centers for Medicare and Medicaid Services form 2746 and from medical records which our clinicians routinely followed up with the patients or family members. All-cause mortality was documented during follow-up. Patients were censored when lost to follow-up, transferred to non-FKC clinics, they received kidney transplantation, or they recovered kidney function. Only patients with at least two baseline serum phosphate values were included. This study was approved by the Western Institutional Review Board (WIRB number ES-18-010).
Variability markers were computed for serum phosphate and serum albumin during the baseline period. These were calculated with all data available during the baseline period.
DR is Calculated as Following
DR of serum phosphate=serum phosphatemin/max (t2)−serum phosphatemin/max(t1), where t1 and t2 represent the time point when the maximum or minimum level of serum phosphate was measured during baseline (Figure 1). A positive DR means that maximum laboratory value happened after the minimum level; conversely, a negative value indicates that the minimum value happened after the maximum.11
Figure 1: DR calculation. t1 is the first time point and t2 is the second time point where the minimum or maximum value was measured of the laboratory value.
Descriptive Statistic Analysis
Descriptive statistics were computed after the patients were stratified into three groups on the basis of DR of serum phosphate in the adjusted spline model: (1) negative (DR<−2.5 mg/dl), (2) neutral (−2.5≤DR≤3.55 mg/dl), and (3) positive (DR>3.55 mg/dl). These groups were on the basis of upper limits of the 95% confidence intervals (CIs) with the intersection with the line of hazard ratio=1 (Figure 2). For continuous variables, means and SDs were computed. Categorical variables were presented as percentages. All analyses were performed with SAS 9.4 (Cary, NC) and R 3.4.0.5.12
Figure 2: Spline model of all-cause mortality and directional range of average serum phosphate, adjusted. The solid black line shows the estimated hazard ratio, and red dashed lines show the upper and lower limits of the hazard ratio. The model was adjusted for age, sex, race, BMI, DM, CHF, COPD, serum albumin, creatinine, serum calcium, and parathyroid hormone.
Survival Analysis
The survival was assessed on the basis of similar methods used in previous research.11 A smoothing spline ANOVA (tensor product smoothing spline) method13 was used to investigate the joint relationships between baseline serum phosphate, serum albumin, and DR of serum phosphate on all-cause mortality during the follow-up period. The smoothing spline ANOVA method allows to model the joint effects of two or more independent variables without assuming a specific parametric form. The contour plots can be used as a graphic approach to present the joint effect of serum phosphate, serum albumin, and serum phosphate variability on outcome in a three-dimensional format, which is to be interpreted in the same fashion as a topographical map, spreading the risk of death (as the elevation on such a topographical map) as a function of the bivariate distribution of both independent variables. For further interpretation, we computed cross-sectional slices of the graph at four serum phosphate levels of interest (serum phosphate=3.0, 5.0, 7.0, and 9.0 mg/dl) and at four serum albumin levels of interest (serum albumin=3.0, 3.5, 3.8, and 4.0 g/dl) depicting the risk of death depicted on the contour plot with the respective CIs over a continuous axis of the SD and DR of serum phosphate. The contour plots and cross-sectional slices are only shown at regions with enough data. These regions were decided by the SD of the spline estimates. Both unadjusted and adjusted analyses were performed. In the adjusted model, age, sex, race, body mass index (BMI), diabetes mellitus (DM), congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), serum albumin, creatinine, serum corrected calcium, and parathyroid hormone were included in the analysis. As an additional analysis, we also performed a model adjusted for age, male, White, BMI, creatinine, serum corrected calcium, parathyroid hormone, DM, CHF, COPD, number of serum phosphate measurements, initial serum phosphate value, vitamin D, phosphate binder, number of treatment per week, hospitalization days, equilibrated Kt/V, ultrafiltration rate, and catheter. We have included this in the supplementary material (Supplemental Figure 5).
Results
Patient Characteristics and Baseline Data
A total of 302,613 patients were enrolled; a group of 24,274 patients (8%) had missing data in the variables for the adjusted analysis. A total of 278,339 patients were stratified into the negative DR group (<−2.5; N=59,147; 21%), null DR group (−2.5 to +3.55; N=166,858; 60%), and positive DR group (>+3.55 N=52,334; 19%) (Supplementary Figure 1). The average age was 62.4±14.1 years; 60% were male; and 41% were diabetic. Table 1 summarizes patient characteristics. When patients were subdivided according to DR, there were no large differences in characteristics between the groups besides a lower hospitalization rate for the neutral group. A total of 125,433 patients were censored because of the end of the study, 2686 because of transplantation, 643 because of lost to follow-up, 2448 because of recovery of kidney function, 29,829 because of transfer to other non-Fresenius Medical Care clinics, 1119 because of a change of modality, and 89,827 because of other reasons.
Table 1 -
Patient characteristics stratified by
directional range of serum
phosphate
Patient Characteristics |
All |
Negative (<−2.5) |
Neutral (−2.5 to +3.55) |
Positive (≥3.55) |
Missing |
Number of patients |
N=302,613 |
N=59,147 |
N=166,858 |
N=52,334 |
N=24,274 |
Phosphate-related parameters
|
|
|
|
|
|
DR of serum phosphate |
0.6±3.3 |
−3.9±1.2 |
1.2±2.1 |
4.8±1.2 |
0.6±3.3 |
Serum phosphate, mg/dL |
5.1±1.2 |
5.4±1.2 |
4.8±0.9 |
6±1.2 |
5.1±1.2 |
Serum phosphate, SD |
0.9±0.6 |
1.5±0.3 |
0.9±0.3 |
1.8±0.3 |
1.2±0.6 |
Initial serum phosphate, mg/dl |
4.8±1.5 |
6.1±1.7 |
4.5±1.1 |
4.4±1.5 |
4.8±1.4 |
Number of serum phosphate measurements |
6.9±2.4 |
7.5±2.6 |
6.6±2.1 |
7.8±2.7 |
6.2±2.9 |
Demographics
|
|
|
|
|
|
Age, yrs |
62.4±14.1 |
60.6±14.4 |
63.3±13.8 |
59.1±14.7 |
62.1±14.1 |
BMI (kg/m2) |
29.7±11.1 |
29.1±11.1 |
30±24.3 |
29.7±63.9 |
29.7±32.7 |
Sex, male % |
60 |
61 |
58 |
56 |
58 |
Race, White % |
68 |
68 |
68 |
68 |
68 |
DM, % |
41 |
40 |
45 |
40 |
43 |
CHF, % |
23 |
21 |
25 |
22 |
23 |
COPD, % |
10 |
9 |
10 |
9 |
10 |
CVD, % |
66 |
67 |
70 |
68 |
69 |
Drugs/medication
|
|
|
|
|
|
Phosphate binders, % |
44 |
50 |
38 |
52 |
43 |
Calcium-containing |
22 |
24 |
19 |
24 |
21 |
Dialysis parameters
|
|
|
|
|
|
Treatment time, min |
226.8±22.8 |
228.4±21.6 |
227.6±21.9 |
228.4±21.9 |
228±21.9 |
Ultrafiltration, L |
2.1±0.9 |
2.1±0.9 |
2.1±0.9 |
2.4±0.9 |
2.1±0.9 |
IDWG, kg |
2.1±0.9 |
2.1±0.9 |
2.1±0.9 |
2.4±0.9 |
2.1±0.9 |
eKt/V |
1.5±0.6 |
1.5±0.3 |
1.5±3.6 |
1.5±0.9 |
1.5±2.7 |
Laboratory parameters
|
|
|
|
|
|
Serum albumin, g/dL |
3.6±0.3 |
3.6±0.3 |
3.6±0.3 |
3.6±0.3 |
3.6±0.3 |
Serum magnesium, mEq/L |
1.8±0.3 |
1.8±0.3 |
1.8±0.3 |
1.8±0.3 |
1.8±0.3 |
Total serum calcium, mEq/L |
9±0.6 |
9±0.6 |
9±0.6 |
8.7±0.6 |
9±0.6 |
Serum creatinine, mg/dL |
6.3±2.4 |
7.2±2.7 |
6±2.1 |
7.2±2.7 |
6.3±2.4 |
Parathyroid hormone, pg/mL |
385.5±272.4 |
390±268.5 |
357.9±249.9 |
422.1±280.5 |
377.4±261.9 |
NLR |
4.8±13.5 |
4.8±3.9 |
4.5±3.6 |
4.8±4.5 |
4.8±5.4 |
WBC, ×10
9
/L |
82.8±2043.6 |
52.8±1233.9 |
30.6±1145.1 |
43.5±988.2 |
40.2±1207.8 |
Normalized protein catabolic rate, g/kg per day |
0.9±0.3 |
0.9±0.3 |
0.9±0.3 |
0.9±0.3 |
0.9±0.3 |
Outcomes
|
|
|
|
|
|
Hospitalization duration, d |
13.2±15 |
13.2±14.1 |
12±13.8 |
12.3±13.2 |
12.6±13.8 |
Hospitalization rate, per 1000 patient-years |
411.45 |
467.83 |
403.28 |
499.97 |
433.27 |
Mortality, per 1000 patient-years |
131.3 |
120.5 |
117.93 |
118.03 |
119.23 |
Negative, DR <−2.5; neutral, DR −2.5 to +3.55; positive, DR >+3.55; missing, patients with incomplete data for adjusted analyses. DR, directional range; BMI, body mass index; DM, diabetes mellitus; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; IDWG, interdialytic weight gain; eKt/V, equilibrated Kt/V; NLR, neutrophil-to-lymphocytes ratio; WBC, white blood cell.
Interaction Between Serum Phosphate and Variability of Serum Phosphate in Relation to Outcome
Baseline phosphate was 5.1±1.2 mg/dl, and median DR was +0.6±3.3 mg/dl. The medium average duration between minimum and maximum values of phosphate was 70 days (mean 78±43 days). As shown in Figure 2, both positive and negative DRs were related to an increased hazard ratio, whereas a neutral DR seemed associated with a better outcome. As displayed in Figure 3A, in unadjusted analysis, the combination of low average serum phosphate levels and a negative DR was associated with a greatly increased risk of mortality. The effect of DR on the outcome was dependent on the average serum phosphate level during the baseline period. At higher levels of serum phosphate, especially a positive DR was related to an increased hazard ratio. Although less pronounced, comparable trends were observed in adjusted analysis (Figure 3C). Higher phosphate variability, when expressed by SD, was related to mortality at all levels of average serum phosphate in unadjusted analysis (Supplemental Figure 2) and in adjusted analysis (Supplemental Figure 3).
Figure 3: Hazard ratio across levels of DR of serum phosphate versus average serum phosphate in the baseline period. (A) Contour plot of estimated hazard ratio as a joint function of DR of serum phosphate and average serum phosphate, unadjusted. (B) Cross-sections of the contour plot A at different average serum phosphate levels, unadjusted. (C) Contour plot of estimated hazard ratio as a joint function of DR of serum phosphate and average serum phosphate, adjusted. (D) Cross-sections of the contour plot C at different average serum phosphate levels, adjusted. The model was adjusted for age, sex, race, BMI, DM, CHF, COPD, serum albumin, creatinine, serum calcium, and parathyroid hormone. Gray areas are expressing the 95% confidence interval.
Interaction Between Serum Albumin and Variability of Serum Phosphate on Outcome
Baseline serum albumin was 3.6±0.3 g/dl. Figure 4 showed that the effect on mortality of a positive DR was especially pronounced in patients with low averaged albumin levels. In patients with low serum albumin (e.q. 3.0 g/dl), a DR between 0 and +5 was associated with a better outcome as compared with patients with a negative DR, whereas the CIs at levels higher than +5 became wide most likely because of the relatively small number of patients. When variability was expressed as SD (Supplemental Figure 4), the same effect of mortality appeared with a minor effect of variability of serum phosphate.
Figure 4: Hazard ratio across levels of average serum albumin and DR of serum phosphate in the baseline period, unadjusted. (A) Contour plot of estimated hazard ratio as a joint function of DR of serum phosphate and average serum albumin, unadjusted. (B) Cross-sections of the contour plot at different average serum albumin levels, unadjusted. Gray areas are expressing the 95% confidence interval.
Additional Analysis
For the maximally adjusted analysis, 161,910 patients were available. The results regarding the effect of DR on outcome appeared materially (Supplemental Figure 5).
Discussion
This study addressed the variability of serum phosphate with outcomes in relation to averaged serum levels of serum phosphate and albumin. Whereas increased variability, when expressed by SD, was associated with adverse outcomes over the entire range of averaged serum phosphate levels, the analysis of DR showed that at a low serum phosphate level, a decline in the serum phosphate level was associated with an increased mortality, whereas the opposite trend was observed in the case of higher serum phosphate levels. When interaction with serum phosphate was studied, the relation between higher phosphate variability as expressed by SD and outcomes was mainly dependent on serum albumin levels. However, when phosphate variability was expressed by DR, there appeared to be a higher mortality risk in those patients with a decrease in DR at low serum albumin levels. To the best of our knowledge, this is the first study analyzing serum phosphate variability and its interaction with serum albumin in a large cohort of patients on HD. We used the smoothing spline ANOVA technique, which can analyze the relation between two continuous variables with outcome, in which the results are easily interpretable using so-called contour plots.14
This study builds on our earlier results where we showed a bidirectional relation between serum phosphate levels and outcomes1 and adds variability as a prognostic marker for mortality. The results show that a major part of the effect of variability on outcome is explained by the direction of the changes and are not attributable to random variability only. Thus, in the analysis of variability, DR adds pivotal information to SD, as we previously showed for serum sodium.11 We suggest that variability, expressed by DR, reflects the more physiological way of the biomarker that is assessed, in comparison with SD.
In accordance with our study, two earlier studies in patients on HD also showed a relation between change in serum phosphate levels and mortality. Nakazato et al. showed, in a cohort of 125 patients, that a log transformed CV of serum phosphate was associated with increased noncardiovascular mortality (hazard ratio [HR], 6.32 [95% CI, 1.39 to 28.7]), but not with cardiovascular mortality (HR, 0.54 [95% CI, 0.06 to 4.84]).5 Zhu et al. reported that a CV of serum phosphate >0.226 mg/dl was associated with increased all-cause mortality (27.7% versus 19.3%, P=0.028) in a cohort of 502 patients divided into two groups according to median levels of the CV.6
In an unselected cohort of hospitalized patients, not specifically focusing on kidney failure, Thongprayoon et al. showed that, after adjustment for multiple confounders including acute kidney injury, both an increase and a decrease in serum phosphate during admission were associated with increased in-hospital mortality.15 However, these studies did not focus on the direction of change (eq increase or decrease), which can be examined with the DR; therefore, this variability marker seems to be stronger.
Conceptually, the relation between phosphate variability and outcome could be because of a direct effect of the fluctuations in serum phosphate per se or to homeostatic derangements underlying these variations. Increasing levels of serum phosphate could, next to vascular calcification, increase oxidative stress and inhibit nitric oxide production,15 whereas low levels could affect cardiac, neurological, and respiratory function.15 However, given that the latter effects usually become apparent at very low levels, we deem it less likely that the relation between hypophosphatemia and declines in serum phosphate levels is mainly explained by these direct effects. Nakazato et al. found that increased variability in biomarkers in patients on HD was related to phenotypical frailty.5 In our study, we showed that serum albumin was an important modifier for the effect of phosphate variability on outcome, although especially in patients with low serum albumin levels, the DR of serum phosphate still provided prognostic value. In addition, Zitt et al. showed in their prospective study that time-varying albumin and phosphate significantly interact in association with all-cause mortality.16
The importance of nutritional status on serum phosphate variability was also shown in the study of Nakazato et al., where variability of serum phosphate was weakly but significantly related to variability in blood urea nitrogen, serum creatinine, and potassium.5 Whereas a low level of serum phosphate was associated with low levels of nutritional markers such as serum albumin, serum creatinine, and BMI,17 the results of our previous and present studies show that these do interact with each other. Therefore, we suggest that a spontaneous decline in serum phosphate levels in patients with already low values, or in patients with other signs of malnutrition, may be a marker of malnutrition with additional prognostic value. Alternatively, in patients with already high phosphate levels, a further increase is also associated with increased risk of mortality. Thus, changes in serum phosphate should be interpreted not only in the context of previous levels but also in the context of nutritional status and previous levels. Whereas a decline in serum phosphate can generally be interpreted as a positive sign, certainly when based on focused interventions, awareness for malnutrition must be raised when occurring spontaneously, especially in combination with changes in other nutritional parameters that may be subtle.7 Attention should be paid to dietary interventions that could also cause a fall in serum albumin levels.16 The use of a so-called nutritional competence score that combines five nutritional parameters including serum phosphate may aid in the interpretation of these changes.7 In patients in whom indicators of the nutritional score decrease or are already low, phosphate variability might prompt more aggressive nutritional counseling and interventions. The same is for the malnutrition-inflammation score; however, our database did not contain all the variables for the patients, and therefore, we could not apply this in our study.18
Strong points of this study are the presence of a large well-characterized cohort, the use of change in variables in correlation with mortality, and the use of statistical methods to combine several risk factors. The drawbacks are its retrospective nature; residual confounding; and the absence of data on renal function, adherence to phosphate binder medication, or nutritional intake. In addition, because of the design, no causal relationship could be proven between exposure and mortality.
In conclusion, an increase in serum phosphate variability is an independent risk factor of mortality. When analyzing DRs, an increase in serum phosphate is a risk factor in patients with high mean serum phosphate levels, whereas a decline is associated with adverse outcomes in patients with low mean values. Although the relation between mortality with a high variability and decline in serum phosphate is especially pronounced in patients with signs of malnutrition, it can be regarded as a marker with additive prognostic value. An increase in serum phosphate variability may be a relevant prognostic marker, but it should be interpreted to nutritional status and direction of the ranges.
Disclosures
The data were provided by and the analysis was supported by Fresenius Medical Care, which included employee salaries and company infrastructure. P. Kotanko and X. Ye are employees of the Renal Research Institute, a wholly owned subsidiary of Fresenius Medical Care (FMC). P. Kotanko, F.W. Maddux, and L.A. Usvyat report having share options/ownership in Fresenius Medical Care and report being an inventor on patent(s) in the field of dialysis. P. Kotanko receives author honoraria from UpToDate and HS Talks. P. Kotanko is on the Editorial Board of Blood Purification and Kidney and Blood Pressure Research. F.W. Maddux and L.A. Usvyat report being employees of Fresenius Medical Care Medical. F.W. Maddux reports directorships in Fresenius Medical Care Management Board, Goldfinch Bio, and Vifor Fresenius Medical Care Renal Pharma. K.J. ter Meulen received a speaker's fee from Fresenius Medical Care, the Netherlands. The remaining authors have nothing to disclose.
Funding
None.
Acknowledgments
We thank the FKC teams for collecting the data from patients treated in North America.
The results presented in this paper have not been published in whole or part, except had been selected as free communications at the ERA-EDTA congress 2021 in Berlin, Germany, and as poster presentation at the ASN Kidney Week 2021 in San Diego. All authors critically revised the manuscript for intellectual content and approved the final version of the manuscript.
Author Contributions
Conceptualization: Jeroen P. Kooman, Peter Kotanko, Franklin W. Maddux, Karlien J. ter Meulen, Frank M. van der Sande, Xiaoling Ye.
Data curation: Xiaoling Ye.
Formal analysis: Xiaoling Ye.
Methodology: Jeroen P. Kooman, Peter Kotanko, Karlien J. ter Meulen, Len A. Usvyat, Yuedong Wang, Xiaoling Ye.
Project administration: Jeroen P. Kooman, Peter Kotanko, Franklin W. Maddux.
Resources: Peter Kotanko, Franklin W. Maddux.
Software: Yuedong Wang, Xiaoling Ye.
Supervision: Constantijn J. Konings, Jeroen P. Kooman, Peter Kotanko, Franklin W. Maddux, Frank M. van der Sande.
Validation: Karlien J. ter Meulen, Xiaoling Ye.
Visualization: Yuedong Wang, Xiaoling Ye.
Writing – original draft: Jeroen P. Kooman, Peter Kotanko, Karlien J. ter Meulen, Frank M. van der Sande, Xiaoling Ye.
Writing – review & editing: Constantijn J. Konings, Franklin W. Maddux, Len A. Usvyat, Yuedong Wang.
Data Sharing Statement
Partial restrictions to the data and/or materials apply. We can share anonymized data subsets in a safe environment on reasonable request.
Supplemental Materials
This article contains the following supplemental material online at https://links.lww.com/KN9/A311.
Supplemental Figure 1. Histogram of directional range of serum phosphate.
Supplemental Figure 2. Hazard ratio across levels of average serum phosphate and standard deviation of serum phosphate in baseline period, unadjusted.
Supplemental Figure 3. Hazard ratio across levels of average serum phosphate and standard deviation of serum phosphate in baseline period, adjusted.
Supplemental Figure 4. Hazard ratio across levels of average serum albumin and standard deviation of serum phosphate in baseline period, unadjusted.
Supplemental Figure 5. Spline model of all-cause mortality and directional range of serum phosphate, maximal adjusted.
References
1. Ye X, Kooman JP, van der Sande FM, et al. Relationship between serum
phosphate levels and survival in chronic
hemodialysis patients: interactions with age, malnutrition and inflammation. Clin Kidney J. 2019;14(1):348–357. doi:
10.1093/ckj/sfz143
2. Taniguchi M, Fukagawa M, Fujii N, et al. Committee of renal data registry of the Japanese Society for Dialysis T: serum
phosphate and calcium should be primarily and consistently controlled in prevalent
hemodialysis patients. Ther Apher Dial. 2013;17(2):221–228. doi:
10.1111/1744-9987.12030
3. Zager PG, Nikolic J, Brown RH, et al. “U” curve association of blood pressure and
mortality in
hemodialysis patients. Kidney Int. 1998;54(2):561–569. doi:
10.1046/j.1523-1755.1998.00005.x
4. Levitt H, Smith KG, Rosner MH.
Variability in calcium, phosphorus, and parathyroid hormone in patients on
hemodialysis.
Hemodialysis Int. 2009;13(4):518–525. doi:
10.1111/j.1542-4758.2009.00393.x
5. Nakazato Y, Kurane R, Hirose S, Watanabe A, Shimoyama H.
Variability of laboratory parameters is associated with frailty markers and predicts non-cardiac
mortality in
hemodialysis patients. Clin Exp Nephrol. 2015;19(6):1165–1178. doi:
10.1007/s10157-015-1108-0
6. Zhu M, Dou L, Zhu M, et al.
Variability of serum phosphorus and its association with
mortality among
hemodialysis patients. Clin Nephrol. 2018;90(2):79–86. doi:
10.5414/CN109265
7. Ye X, Dekker MJE, Maddux FW, et al. Dynamics of nutritional competence in the last year before death in a large cohort of US
hemodialysis patients. J Ren Nutr. 2017;27(6):412–420. doi:
10.1053/j.jrn.2017.06.006
8. Thijssen S, Wong MM, Usvyat LA, Xiao Q, Kotanko P, Maddux FW. Nutritional competence and resilience among
hemodialysis patients in the setting of dialysis initiation and hospitalization. Clin J Am Soc Nephrol. 2015;10(9):1593–1601. doi:
10.2215/CJN.08430814
9. Mallamaci F, Tripepi G, D'Arrigo G, et al. Blood pressure
variability,
mortality, and cardiovascular outcomes in CKD patients. Clin J Am Soc Nephrol. 2019;14(2):233–240. doi:
10.2215/CJN.04030318
10. Brunelli SM, Lynch KE, Ankers ED, et al. Association of hemoglobin
variability and
mortality among contemporary incident
hemodialysis patients. Clin J Am Soc Nephrol. 2008;3(6):1733–1740. doi:
10.2215/CJN.02390508
11. Ye X, Kooman JP, van der Sande FM, et al. Increased
mortality associated with higher pre-dialysis serum sodium
variability: results of the international MONitoring dialysis outcome initiative. Am J Nephrol. 2019;49(1):1–10. doi:
10.1159/000495354
12. Team RC. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing; 2018.
13. Wang Y. Smoothing Splines: Methods and Applications. Chapman & Hall/CRC; 2011.
14. Liu A, Tong TJ, Wang YD. Smoothing spline estimation of variance functions. J Comput Graphical Stat. 2007;16(2):312–329. doi:
10.1198/106186007x204528
15. Thongprayoon C, Cheungpasitporn W, Hansrivijit P, et al. Impact of serum
phosphate changes on in-hospital
mortality. BMC Nephrol. 2020;21(1):427. doi:
10.1186/s12882-020-02090-3
16. Zitt E, Lamina C, Sturm G, et al. Interaction of time-varying albumin and phosphorus on
mortality in incident dialysis patients. Clin J Am Soc Nephrol. 2011;6(11):2650–2656. doi:
10.2215/CJN.03780411
17. Lee JE, Lim JH, Jang HM, et al. Low serum
phosphate as an independent predictor of increased infection-related
mortality in dialysis patients: a prospective multicenter cohort study. PLoS One. 2017;12(10):e0185853. doi:
10.1371/journal.pone.0185853
18. Kalantar-Zadeh K, Kopple JD, Block G, Humphreys MH. A malnutrition-inflammation score is correlated with morbidity and
mortality in maintenance
hemodialysis patients. Am J Kidney Dis. 2001;38(6):1251–1263. doi:
10.1053/ajkd.2001.29222