J Am Soc Nephrol 11: 2337–2343, 2000 Effect of Dialysate Sodium Concentration on Interdialytic Increase of Potassium

LUCA DE NICOLA,* VINCENZO BELLIZZI,‡ ROBERTO MINUTOLO,* MARIO CIOFFI,* PAOLO GIANNATTASIO,* VINCENZO TERRACCIANO,‡ CARMELA IODICE,* FRANCESCO UCCELLO,† BRUNO MEMOLI,† BIAGIO RAFFAELE DI IORIO,‡ and GIUSEPPE CONTE* Chair of Nephrology, School of Medicine, *Second University of Naples and †University Federico II, Naples, and ‡Lauria Hospital, Lauria, Italy.

Ϯ ϩ Ϯ Abstract. To evaluate the role of plasma tonicity in the post- 10% and 50 7% in H-NaD and L-NaD, respectively; P dialysis increment of plasma potassium (p[Kϩ]), the outcome Ͻ 0.05). Of note, in the first 2 h after the end of treatment, of two hemodiafiltration treatments that differed only in the bioimpedance analysis revealed only in H-NaD a significant 11 ϩ Ϯ Na level in dialysate (NaD)—143 mmol/L (high dialysate 3% decrement of phase angle that is compatible with a sodium concentration [H-NaD]) and 138 mmol/L (low dialy- decrease of intracellular fluid volume at the expense of the sate sodium concentration [L-NaD])—were compared in the extracellular volume. Similarly, within the same time frame, in same group of uremic patients from the end of treatment (T0) H-NaD, a significant reduction of mean corpuscular volume of to the subsequent 30 to 120 min and up to 68 h. Kt/V and red cells, associated with a 2 Ϯ 1% decrease of the intracellular intradialytic Kϩ removal were comparable. At T0, plasma [Kϩ], was observed. In contrast, mean corpuscular volume of ϩ Ϯ Ϯ ϩ Ϯ [Na ] was 145 1 and 137 1 mmol/L after H-NaD and red cells did not change and erythrocyte [K ] increased by 6 Ͻ Ͻ L-NaD, respectively (P 0.001). The difference in plasma 1% after L-NaD (P 0.005 versus H-NaD). Thus, hypertonic- tonicity persisted from T0 to T68 h. At T120, p[Kϩ] was ity significantly contributes to the increase of p[Kϩ] through- increased from the T0 value of 3.7 Ϯ 0.2 to 4.7 Ϯ 0.2 mmol/L out the whole interdialytic period by determining intracellular Ͻ (P 0.05) after H-NaD, whereas it was unchanged after fluid volume/extracellular volume redistribution of water and ϩ ϩ ϩ L-NaD. The change of p[K ] was still different after 68 h ( 76 K .

Hyperkalemia is a common and dangerous event in patients tion. Indeed, the information on the underlying pathophysio- who are on maintenance hemodialysis (HD). The most prom- logic mechanism is scarce and is limited to the observation that inent adverse effects are unpredictable and potentially lethal this phenomenon is not related to the intradialytic Kϩ removal arrhythmias, respiratory depression, and enhanced weakness (6). We have demonstrated previously that in patients with and fatigue (1). In these patients, HD is the primary method of moderate renal failure, the infusion of 5% hypertonic NaCl potassium (Kϩ) removal (2,3). However, the reduction of solution is associated with an increment of p[Kϩ] of approxi- plasma Kϩ levels (p[Kϩ]) by HD is often significantly im- mately 0.6 mmol/L despite enhanced renal excretion of Kϩ and paired by the postdialysis rebound of Kϩ, i.e., the increment of independently from acid-base or hormonal mechanisms (8). ϩ p[K ] detectable within the initial few hours after treatment We concluded that plasma hypertonicity was the main deter- ϩ (4). The magnitude of K rebound varies, depending on the minant of hyperkalemia by inducing intra-/extracellular redis- ϩ pre-HD values of p[K ], from approximately 1 mmol/L in the tribution of Kϩ as a result of solvent drag. A similar phenom- absence of marked pre-HD hyperkalemia (5,6) up to levels enon also has been reported in individuals with normal renal sufficiently high to cause fatalities, as described in hyperkale- function (1). mic patients with tumor lysis syndrome (7). These studies Because p[Kϩ] correlates with plasma tonicity, it is reason- ϩ emphasize the importance of a periodic control of p[K ] able to hypothesize that a reduction of extracellular sodium during the interdialytic period in hyperkalemic patients even in concentration may prevent Kϩ rebound in HD patients. The the presence of normal levels at the end of the treatment. present study verifies this hypothesis by evaluating the influ- ϩ Despite the clinical relevance of K rebound, no study has ence of sodium dialysate concentration on the postdialysis focused on the prevention of this potentially lethal complica- p[Kϩ] levels and the concurrent transcellular shift of water and Kϩ. To exclude any disturbing factor in this analysis, we performed the study with patients who had no additional risk of Received February 7, 2000. Accepted April 17, 2000. hyperkalemia. Patients were treated with soft hemodiafiltra- Correspondence to Dr. Giuseppe Conte, Via L. Caldieri, 10, 80127 Naples, Italy. Phone/Fax: ϩ39 81 2549409; E-mail: [email protected] tion; such a dialysis technique, characterized by combined 1046-6673/1112-2337 diffusive and convective fluxes, reinfusion in postdilution of Journal of the American Society of Nephrology bicarbonate solution, and an rate of 25 to 50 ϩ Copyright © 2000 by the American Society of Nephrology ml/min, allows an optimal control of p[K ] by ensuring more 2338 Journal of the American Society of Nephrology J Am Soc Nephrol 11: 2337–2343, 2000

– adequate acid-base balance and stability of systemic hemody- mmol/L HCO3 ). Dialysate temperature was constantly kept at namics (9). We compared the effects of two different sodium 36.5°C. dialysate concentrations, 138 and 143 mmol/L, that fall within Patients randomly received the two different treatments on the last the isotonic range commonly used in clinical practice to avoid session before the long interdialytic interval. The second experimental drawbacks secondary to major changes of plasma osmolality session was performed 2 wk after the first treatment. During this period, patients were treated with soft hemodiafiltration with Na of and volume status (10,11). D 140 mmol/L. Blood samples were obtained before the treatment; at the end of Materials and Methods treatment (T0); and after 30, 60, 90, and 120 min through the dialysis Patients needles left in place and in the following 24, 48, and 68 h by To select patients with stable clinical conditions and with no major venipuncture. During the entire dialysis session and in the subsequent ϩ alterations of acid-base balance and K metabolism, we applied the 2 h, all of the patients remained in bed and did not receive any food following inclusion criteria: adult anuric uremic patients treated from or beverage. Body weight, BP, and pulse rate were recorded hourly at least 6 mo with soft hemodiafiltration and with sodium level in during the treatment and at each experimental time. dialysate at approximately 140 mmol/L; absence of any acute illness At each experimental time, we measured hematocrit and the plasma in the last 3 mo; constancy of the main clinical features during the last levels of Kϩ, sodium, total protein, blood urea nitrogen (BUN), 3 mo (dialysis modalities, diet and nutritional status, dry weight, glucose, phosphate, osmolality, bicarbonate and pH, insulin, and arterial BP, medications, and routine laboratory measurements); . We also assessed the PCR value, normalized for dry body ϩ p[K ] Ͻ 7.0 mmol/L and plasma bicarbonate Ͼ 18 mmol/L after the weight (nPCR), from the BUN appearance during the interdialytic long interdialytic interval in the last 3 mo; absence of diabetes interval from T0 to T68 (13). The predialysis sample was obtained mellitus, heart failure (NYHA classes III to IV), or advanced liver before the infusion of saline or heparin and before starting the blood disease and peripheral or pulmonary edema. pump, whereas the postdialysis sample (T0) was drawn after main- We studied, after obtaining informed consent, 12 patients (8 men, taining for 2 min a low blood flow rate (50 ml/min), in the absence of 4 women) with a mean age of 54 Ϯ 4 yr (range, 31 to 70 yr). The dialysate flow to minimize cardiopulmonary blood recirculation. underlying renal disease was primary glomerulonephritis in seven patients, nephroangiosclerosis in three patients, and polycystic disease in two patients. Before the study, the patients had been dialyzed from Erythrocyte Potassium Ϯ To determine whether changes of NaD cause transcellular shift of an average of 76 10 mo (range, 12 to 120 mo). They had been ϩ ϩ treated regularly for 4 h, three times a week, with a delivered Kt/V K , we examined the intracellular K levels in erythrocytes (e[K ]). dose, calculated according to the method of Daugirdas (12), of at least Non-hemolyzed blood samples were drawn before hemodiafiltration 1.20 in the last 3 mo. All of them were prescribed a diet that contained and after treatment, from T0 to T120. Immediately after the blood approximately 1.0 g/kg body wt of protein and 30 to 35 kcal/kg body sample was obtained, erythrocytes were separated from anticoagu- wt. In the last 3 mo, the protein catabolic rate (PCR), calculated with lated blood by centrifugation and processed within1hasdescribed standard formula (13), averaged 1.24 Ϯ 0.23 g/kg body wt with previously (14). Briefly, 1 ml of erythrocytes was washed with 1 ml changes not exceeding 10%. During the study, they received only the of isoosmotic Tris 10 mM and choline chloride 140 mM solution to following medications: calcium carbonate (4 patients), magnesium remove residual plasma cations. Thereafter, 1 ml of the washed cell hydroxide plus aluminum hydroxide (8 patients), antihypertensive suspension was added to 1 ml of hypotonic Tris 5 mM and ethyl- drugs (3 patients were taking calcium antagonists and 3 were treated enediaminetetraacetic acid 1 mM solution and stored for 24 h at 4°C with clonidine), epoetin (11 patients with a mean weekly dose of 6000 to obtain the complete cell lysis. Then, samples were centrifugated at 3500 rpm for 5 min to remove cell membranes. Potassium concen- U), vitamin D3 (2 patients), and H2 antagonists (7 patients). The dose of medications did not vary in the two experimental steps. tration is expressed as millimoles per liter of erythrocytes. The mean corpuscular volume (MCV) of red cells was also assessed from the Study Design ratio between hematocrit (determined by ALC hematocrit centri- fugette 4203, ALC Int., Milano, Italy) and the number of red cells The study was a randomized, single-blind, crossover trial in the (determined by H2 System analyzer, Technicon, Bayer, Germany). same group of 12 patients of two identical treatments of soft hemo- diafiltration differing only in the sodium level in the dialysate, which was kept constant during the entire dialysis treatment: Bioimpedance Analysis

High dialysate sodium concentration: 143 mmol/L (H-NaD) Bioelectrical impedance analysis (BIA) is a technique used to Low dialysate sodium concentration: 138 mmol/L (L-NaD) evaluate indirectly body composition by injecting through the body a The treatment was delivered in each patient, at both experimental low-amplitude alternating electrical current (15). We used this tech- steps, with the same artificial system equipped with automatic device nique to gain information on the transcellular shift of water under the planning the ultrafiltration rate (Integra, Hospal, Bologna, Italy, or two different experimental conditions. BIA analyzes the electrical System 1000, Drake Willok-Althin, Rome, Italy), membrane (poly- characteristics of tissues; the two components that contribute to im- sulfone 1.8 m2; F8, Fresenius, Palazzo Pignano, Italy, or PMMA 2.0 pedance are resistance (R), the pure opposition of the tissue to the m2; Filtryzer B3-2, Toray-Hoechst, Milan, Italy), blood flow rate (315 flow of electrons, and reactance (Xc), reflecting the capacitance of to 345 ml/min), dialysate flow rate (500 ml/min), and ultrafiltration cell membranes, tissue interfaces, and so forth (16). The BIA-derived rate (25 to 35 ml/min). Also, the dialysate composition did not differ: phase angle (PhA) corresponds to the angular transformation of the bicarbonate, 39 mmol/L; acetate, 4.0 mmol/L; calcium, 1.5 mmol/L; ratio Xc/R (15). BIA measurements (R, Xc, and PhA) were obtained magnesium, 0.5 mmol/L; and glucose, 1.0 g/L; the potassium level before the dialysis session and at each time of the postdialysis period; was 2.0 and 3.0 mmol/L in five and seven patients, respectively, and the electrodes were placed on the side free from the vascular access was kept constant in the two treatments. The composition of replace- and were kept in place throughout the study to avoid replacement ment fluid was identical (145 mmol/L Naϩ, 100 mmol/L Cl–,45 errors. Single-frequency BIA was determined at 50 kHz, with an J Am Soc Nephrol 11: 2337–2343, 2000 Dialysate [Naϩ] and Interdialytic Plasma Kϩ 2339 impedance plethysmograph (model BIA 101 RJL, Akern, Firenze, during and after the dialysis session up to T68 h, as depicted by Italy) according to the standard tetrapolar technique (15,16). the plasma sodium (p[Naϩ]) levels (Figure 1). At T0, plasma Ϯ Ϯ osmolality was 298 2.8 and 284 1.4 mOsm/kg H2O after Analytical Determinations Ͻ H-NaD and L-NaD, respectively (P 0.001). The levels of potassium and sodium in plasma and dialysate were The different tonicity resulted in a diverse profile of p[Kϩ] assessed by flame photometer (Beckman Instruments, Inc., Fullerton, (Figure 2). Despite a comparable intradialytic decrement of CA) in triplicate (the coefficient of variation was always Ͻ1%); BUN, ϩ p[K ](Ϫ36 Ϯ 3% at H-Na and Ϫ34 Ϯ 2% at L-Na , not phosphate, total protein, and glucose were measured using an auto- D D significant), only the H-NaD treatment was associated with a analyzer (Olympus AU 560, Olympus Italia, Segrate-Milano, Italy). ϩ Blood pH and bicarbonate levels were analyzed by an automatic significant increase of p[K ] within the initial 2 h postdialysis. ϩ ϩ hemogas analyzer (ABL 625, Radiometer Copenhagen, De Mori, As for p[Na ], the different pattern of p[K ] changes in the Italy). Plasma osmolality was measured by osmometer (model 250 D, two dialysis sessions persisted up to the end of the observation S Fiske Associates Inc., Uxbridge, MA). Aldosterone and insulin period. Of note, the Kϩ level measured 2 h after the end of levels were assessed by standard RIA (Aldoctk 2 and Insik 5, respec- dialysis were predictive of the p[Kϩ] values detected in the tively, Sorin, Vercelli, Italy). subsequent period of follow-up. In fact, significant correlations The amount of potassium removed (K ) by the treatment was also ϩ r were found between the p[K ] value at T120 and the values determined according to the following formula (6): K ϭ (V K ) Ϫ r C C measured at T24 h (r ϭ 0.80), T48 h (r ϭ 0.71), and T68 h (r (V K ); where V is the volume of collected dialysate measured by D D C ϭ collecting the entire effluent dialysate into a high-capacity box placed 0.66) after H-NaD; similar results were obtained with L-NaD ϩ (T24 h, r ϭ 0.72; T48 h, r ϭ 0.69; T68 h, r ϭ 0.78; P Ͻ 0.05). on a plate balance, KC is the K concentration in collected dialysate, VD is the fresh dialysate volume determined by the artificial system, The interdialytic protein intake, evaluated by nPCR deter- ϩ Ϯ and KD is the K concentration in fresh dialysate verified at the minations in the T0 to T68 h interval, was comparable (1.21 prehemofilter level. Ϯ 0.21 and 1.23 0.25 g/kg body wt per in H-NaD and L-NaD, respectively). Plasma tonicity also influenced the postdialysis Statistical Analyses increment of phosphate (Figure 3A); in contrast, no effect was All of the values are reported as mean Ϯ SEM. Intergroup com- observed on the variation of BUN (Figure 3B). parisons were made by two-tailed t test for unpaired data, whereas Table 2 describes the postdialysis variation of the main intragroup comparisons were made by ANOVA for repeated measures factors involved in the control of p[Kϩ]. During the dialysis followed by the Newman-Keuls as post hoc test. Linear regression treatment, blood pH and bicarbonate, which were comparable analysis was also used where indicated. P Ͻ 0.05 was considered statistically significant. at baseline, similarly increased in H-NaD and L-NaD.Inthe postdialysis period, while the value of blood pH was similar for Results the two modalities, plasma bicarbonate was slightly but signif- The two dialysis sessions did not differ in treatment time icantly higher at H-NaD with respect to L-NaD; such a differ- (H-Na , 235 Ϯ 14 min; L-Na , 230 Ϯ 14 min). As reported in ence persisted at T120 but, as opposed to the changes of D D ϩ ϩ Table 1, the sessions led to similar weight loss (2.9 Ϯ 0.3 kg plasma Na and K , disappeared in the subsequent period and 2.9 Ϯ 0.2 kg) and reduction of blood volume, as evidenced from T24 to T68 h. Plasma glucose level (mg/dl), which was by the same increment of hematocrit (10.8 Ϯ 2 and 10.7 Ϯ 2%) significantly higher at the end of dialysis as compared with the and serum total protein (19.9 Ϯ 3 and 18.1 Ϯ 3%); also, the predialysis values in both modalities (it increased from 90 Ϯ ϩ Ϯ Ϯ Ϯ delivered dialysis dose and K removal were similar. Both 21 to 111 20 and from 90 19 to 114 21 in H-NaD and treatments were characterized by the absence of hypotensive L-NaD, respectively), did not vary in postdialysis and remained Ϯ episode, defined as a decrease in systolic pressure of ୑30 similar in H-NaD and L-NaD (at T120, the values were 107 mmHg or a value Ͻ90 mmHg. 17 and 110 Ϯ 15, respectively). Similarly, the insulin levels ϩ The two levels of NaD led to a different plasma tonicity were unaffected by the two dialysate Na concentrations (Ta-

Table 1. Changes of the indexes of volume state and dialysis efficiency in soft hemodiafiltration at high (143 mmol/L,

H-NaD) and low (138 mmol/L, L-NaD) sodium dialysate concentration Body Weight Mean BP Serum Total (kg) (mmHg) Hematocrit (%) Protein (g/dL) ϩ Kt/V K removal (mmol) Pre Post Pre Post Pre Post Pre Post

a a a a H-NaD 71.3 68.4 103 100 32.9 36.5 7.10 8.50 1.32 75 (3.4) (3.3) (2.2) (4.0) (0.8) (1.3) (0.1) (0.2) (0.05) (12) a a a a L-NaD 71.4 68.5 102 100 34.4 38.0 7.07 8.35 1.40 73 (3.5) (3.4) (3.1) (2.5) (1.0) (1.0) (0.2) (0.3) (0.06) (7)

Values are mean Ϯ SEM. a P Ͻ 0.05 versus predialysis value. 2340 Journal of the American Society of Nephrology J Am Soc Nephrol 11: 2337–2343, 2000

Figure 1. Plasma sodium (p[Naϩ]) before (predialysis), at the end (T0), and 30 min to 68 h after soft hemodiafiltration with dialysate ϩ ϪFϪ Na concentration of 143 mmol/L ( , H-NaD) and 138 mmol/L ϪⅢϪ # Ͻ ( , L-NaD). , P 0.05 versus L-NaD.

Figure 3. Plasma phosphate (A) and blood urea nitrogen (B) before (predialysis), at the end (T0), and 30 min to 68 h after soft hemodi- afiltration with dialysate Naϩ concentration of 143 mmol/L (ϪFϪ, ϪⅢϪ Ͻ # H-NaD) and 138 mmol/L ( , L-NaD). *, P 0.05 versus T0; , P Figure 2. Plasma potassium (p[Kϩ]) before (predialysis), at the end Ͻ 0.05 versus L-NaD. (T0), and 30 min to 68 h after soft hemodiafiltration with dialysate ϩ ϪFϪ Na concentration of 143 mmol/L ( , H-NaD) and 138 mmol/L ϪⅢϪ Ͻ # Ͻ ( , L-NaD). *, P 0.05 versus T0; , P 0.05 versus L-NaD. The pattern of the postdialysis variation of the MCV was comparable to that of PhA (Figure 4B). MCV, which was Ϯ Ϯ ble 2). A slight but significant decrease of aldosterone levels similar in predialysis (89.4 1.8 fl at H-NaD and 90.2 1.9 Ͻ was noted 2 h after H-NaD (P 0.05 versus T0; Table 2); fl at L-NaD), was significantly lower at the end of H-NaD however, no significant intergroup difference was detected at versus L-NaD. Thereafter, MCV progressively decreased after T120. H-NaD, whereas it was unmodified in L-NaD. ϩ BIA measures were comparable in the two treatments at In predialysis, e[K ] did not differ significantly in H-NaD Ϯ Ϯ baseline and significantly increased after both sessions; how- and L-NaD, (92.2 1.8 and 95.8 2.5 mmol/L red cells, ever, a diverse pattern of variation was detected in the postdi- respectively). As depicted in Table 3, after H-NaD treatment, alysis T0 to T120 period. In the presence of a stable value of the progressive reduction of MCV was coupled with a slight ϩ ϩ R, Xc diminished by a greater extent in H-NaD than in L-NaD decrement of e[K ]; in contrast, e[K ] significantly increased, Ϯ Ϯ (at T120, the decrement versus T0 was 8 3% and 1 2%, in the presence of a stable value of MCV, after L-NaD. respectively; P Ͻ 0.01). The combination of these changes resulted in a significant difference of the postdialysis values of Discussion

PhA, which did not vary after L-NaD but progressively de- In individuals with normal renal function, the increase of creased after H-NaD (Figure 4A). extracellular osmolality induced by effective osmoles, such as J Am Soc Nephrol 11: 2337–2343, 2000 Dialysate [Naϩ] and Interdialytic Plasma Kϩ 2341

Ϫ Table 2. Blood pH and plasma levels of HCO3 , insulin, and aldosterone immediately (T0) and 120 min (T120) after the end of soft hemodiafiltration at high (143 mmol/L, H-aD) and low (138 mmol/L, L-NaD) sodium dialysate concentration

Ϫ pH HCO3 (mmol/L) Insulin (mU/L) Aldosterone (pg/mL) T0 T120 T0 T120 T0 T120 T0 T120

b b a H-NaD 7.446 7.451 28.0 29.0 24.4 28.3 193 154 (0.012) (0.008) (0.7) (0.7) (4.8) (6.3) (34) (21)

L-NaD 7.447 7.448 26.3 27.5 26.1 29.2 210 175 (0.008) (0.006) (0.4) (0.5) (6.0) (5.9) (45) (35)

Values are mean Ϯ SEM. a P Ͻ 0.05 versus T0. b Ͻ P 0.05 versus L-NaD.

sodium, glucose, amino acid, mannitol, and radiocontrast me- (10,11). In this study, the small change of NaD resulted in dia, causes a significant increment of plasma potassium (17– p[Naϩ] levels that did not fall out of the normal range. 21). It has been hypothesized that in the presence of higher The imposed difference in plasma tonicity strikingly influ- tonicity of the (ECF), Kϩ moves out the enced the postdialysis Kϩ levels. The increase of p[Kϩ] from intracellular fluid (ICF) by solvent drag (1,22); according to T0 to T120 was significant only after the H-NaD treatment, this hypothesis, as water is forced out of cells by the osmotic whereas the L-NaD session led to a postdialysis increment of a gradient, Kϩ will be entrained in the convective current and small and not significant extent. Furthermore, as observed for will exit cells in proportion to the membrane permeability. p[Naϩ], such a difference persisted during the entire period of Alternatively, the increase in ECF and the decrease in ICF will follow-up. Although patients were not monitored in our clini- ϩ ϩ ϩ ϩ change the K ICF/K ECF and Na ICF/Na ECF ratios, thus mod- cal research center during the interdialytic period, the higher ϩ ifying the transmembrane potential or Na-K-ATPase activity. p[K ] observed after H-NaD in the T24 h to T68 h period was Although the hypertonicity-induced increase of p[Kϩ]is probably unrelated to changes of nutritional intake as patients usually of minor importance when renal function is preserved, were asked not to vary the diet. Indeed, the nPCR value was the clinical impact of this phenomenon certainly grows in comparable after the two sessions in the T0 to T68 h interval. patients with renal impairment. We previously demonstrated We analyzed the effect of tonicity on p[Kϩ] in the absence that in moderate chronic renal failure, plasma tonicity is a of factors that interfere with Kϩ balance. The study was specific determinant of Kϩ levels (8). In that study, a signifi- performed in nondiabetic patients who, besides the proven cant increase of p[Kϩ] was detected in the presence of major compliance to the diet, were not treated with any drug that increments of plasma tonicity, i.e., beyond the normal range of modifies p[Kϩ]; in addition, in both experimental sessions, Kϩ p[Naϩ] concentration. Of note, the phenomenon occurred de- removal was similarly adequate. We can also exclude the spite the enhancement of the urinary excretion of Kϩ and was influence of two factors that profoundly affect the relative independent of the main mechanisms known to regulate the ICF/ECF distribution of Kϩ: insulin and acid-base balance extrarenal homeostasis of this solute, such as acid-base bal- (23). It is interesting that the increment of p[Kϩ] was greater ance, insulin, aldosterone, and adrenergic activity. The present after the H-NaD treatment despite the higher values of plasma – study shows that in anuric patients, even modest changes in bicarbonate. The reason for the different p[HCO3 ]isnot plasma tonicity, of a magnitude commonly encountered in readily apparent; however, we hypothesize that in H-NaD, the clinical practice, can have relevant effects on the postdialysis presence of a higher amount of Naϩ at the level of dialyzer levels of p[Kϩ]. may have determined a higher Naϩ concentration gradient Because sodium is the major determinant of effective osmo- between dialysate and blood with enhanced diffusive transport lality in blood and dialysis fluid, dialysate sodium concentra- of Naϩ into blood and consequent proportional diffusion of – tion is the variable that can be more easily varied to change HCO3 to maintain electroneutrality. plasma tonicity in HD patients. Indeed, modification of NaD is Plasma tonicity was therefore a primary determinant of the a common therapeutic approach in these patients. In the past, postdialysis changes of p[Kϩ]. Aldosterone may have medi- both hypernatric and hyponatric dialysates have been used, the ated this effect at least partially; its plasma levels were in fact former to reduce intradialytic morbidity and the latter to obtain significantly decreased at the T120 control after H-NaD, pos- a better control of and interdialytic weight gain. Recently, sibly because of the higher ECF volume. Nevertheless, the however, the variation of NaD has been restricted within the contribution of this hormone was minimal because a similar isotonic or “physiologic” range to avoid excessive alteration of decline also was observed after L-NaD. We cannot exclude a plasma tonicity and the consequent detrimental effects, either role of the adrenergic tone as we did not measure the blood chronic volume overload or poor tolerance to dialysis treatment levels of catecholamines; however, previous studies in nondia- 2342 Journal of the American Society of Nephrology J Am Soc Nephrol 11: 2337–2343, 2000

lyzed renal patients have found that this system is not involved in the mechanism of the tonicity-related changes of p[Kϩ] (8,22). It is interesting that the greater postdialysis changes of

plasma phosphate after H-NaD strengthen the hypothesis of the primary role of tonicity because the metabolism of this solute is certainly not affected by aldosterone, insulin, or adrenergic tone. The present study indicates that the postdialysis increase of p[Kϩ] and the tonicity-induced hyperkalemia share the same pathophysiologic mechanism, i.e., the development of ICF/ ECF osmotic gradients with the consequent proportional exit from the cells of water and Kϩ. To confirm this hypothesis, we assessed the concurrent transcellular shift of water and Kϩ. This evaluation was attained by measuring bioimpedance pa- rameters, which are stable and reproducible in postdialysis (24), as well as the volume of erythrocytes and the intraeryth- rocyte Kϩ levels. Previous studies have shown that two BIA-derived mea- sures—Xc/R ratio and PhA, i.e., the angular transformation of the ratio Xc/R (16)—are reliable predictors of ICF/ECF water distribution in normal subjects and in nonrenal diseases (16,25). Furthermore, in a large population of HD patients, PhA has been demonstrated to correlate inversely with the extracellular water/total ratio (26). In the first 2 h

after the H-NaD treatment, PhA markedly decreased, suggest- ing expansion of ECF at the expense of ICF; indeed, total body water did not change in postdialysis as patients were anuric and no beverage or food was allowed. The parallel decrease of MCV supports the hypothesis of an intra- to extracellular shift of water. More important, the shrinkage of red cells was associated with a decrement of e[Kϩ], demonstrating a cell depletion of Kϩ secondary to the tonicity-induced exit of water

after H-NaD. At variance with the postdialysis period, p[Kϩ] was not Figure 4. Phase angle (A) and mean corpuscular volume of red cells influenced by tonicity in the course of dialysis treatment. A (B) at the end (T0) and 30 to 120 min after soft hemodiafiltration with possible explanation for such a discrepancy is that the intra- ϩ ϪFϪ ϩ dialysate Na concentration of 143 mmol/L ( , H-NaD) and 138 dialytic p[K ] levels are determined primarily by the diffusive ϪⅢϪ Ͻ # Ͻ ϩ mmol/L ( , L-NaD). *, P 0.05 versus T0; , P 0.05 versus removal of K (6). L-NaD. In conclusion, this study provides the first evidence that even a moderate hypertonicity at the end of dialysis treatment determines a further increase of p[Kϩ] throughout the inter- dialytic period. The underlying mechanism is likely repre- sented by the tonicity-induced redistribution of water and Kϩ Table 3. Intraerythrocyte K concentration (mmol/L red cells) from the intra- to the extracellular compartment. immediately (T0) and 30 to 120 min after the end of soft hemodiafiltration at high (143 mmol/L, Acknowledgments H-NaD) and low (138 mmol/L, L-NaD) sodium dialysate concentration This study was partially supported by a grant from M.U.R.S.T. (Ministero dell’Università e della Ricerca Scientifica e Tecnologica), T0 T30 T60 T90 T120 Rome, Italy, to G. C. in 1998.

b H-NaD 96.1 95.3 94.7 95.0 94.3 (2.2) (1.5) (1.7) (1.6) (1.4) References L-Na 92.5 98.4a 98.3a 98.6a 98.5a 1. Weiner ID, Wingo CS: Hyperkalemia: A potential silent killer. D J Am Soc Nephrol 9: 1535–1543, 1998 (1.0) (1.2) (1.2) (1.3) (1.0) 2. Greenberg A: Hyperkalemia: Treatment options. Semin Nephrol Values are mean Ϯ SEM. 18: 46–57, 1998 a P Ͻ 0.05 versus T0. 3. Blumberg A, Weidman P, Shaw S, Gnadinger M: Effect of b Ͻ P 0.05 versus L-NaD. various therapeutic approaches on plasma potassium and major J Am Soc Nephrol 11: 2337–2343, 2000 Dialysate [Naϩ] and Interdialytic Plasma Kϩ 2343

regulating factors in terminal renal failure. Am J Med 85: 507– Technology Assessment Conference Statement. December 12- 512, 1988 14, 1994. Nutrition 12: 749–759, 1996 4. Morgan AG, Burkinshaw L, Robinson PJA, Rosen SM: Potas- 16. Lukaski HC: Biological indexes considering in the derivation of sium balance and acid-base changes in patients undergoing reg- the bioelectrical impedance analysis. Am J Clin Nutr 64[Suppl ular hemodialysis therapy. Br Med J 1: 779–783, 1970 3]: S397–S404, 1996 5. Feig PU, Shook A, Sterns RH: Effects of potassium removal 17. Moreno M, Murphy C, Goldsmith C, Moran H: Increase in serum during hemodialysis on the plasma potassium concentration. potassium resulting from the administration of hypertonic man- 27: 25–30, 1981 nitol and other solutions. J Lab Clin Med 73: 291–298, 1969 6. Blumberg A, Roser HW, Zehnder C, Muller-Brand J: Plasma 18. Makoff DL, Da Silva JA, Rosenbaum BJ, Levy SE, Maxwell potassium in patients with terminal renal failure during and after MH: Hypertonic expansion: Acid-base and electrolyte changes. hemodialysis: Relationship with dialytic potassium removal and Am J Physiol 218: F1201–F1207, 1970 total body potassium. Nephrol Dial Transplant 12: 1629–1634, 19. Goldfarb S, Strunk B, Singer I, Goldberg M: Paradoxical glu- 1997 cose-induced hyperkalemia: Combined aldosterone-insulin defi- 7. Stapleton FB, Strother DR, Roy S, Wyatt RJ, McKay CP, Mur- ciency. Am J Med 59: 744–750, 1975 phy SB: Acute renal failure at onset of therapy for advanced 20. Viberti GC: Glucose-induced hyperkalemia: A hazard for dia- stage Burkitt lymphoma and B cell acute lymphoblastic lym- betics? Lancet 1: 690–691, 1978 phoma. Pediatrics 82: 863–869, 1988 21. Bratusch-Marrain PR, DeFronzo RA: Impairment of insulin- 8. Conte G, Dal Canton A, Imperatore P, De Nicola L, Gigliotti mediated glucose metabolism by hyperosmolality in man. Dia- G, Pisanti N, Memoli B, Fuiano G, Esposito C, Andreucci VE: betes 32: 1028–1034, 1983 Acute increase in plasma osmolality as a cause of hyperkale- 22. De Nicola L, Cappabianca F, Bellizzi V, Conte G: Plasma mia in patients with renal failure. Kidney Int 38: 301–307, hyperosmolality as cause of hyperkalemia. In: Acid-Base and 1990 Electrolyte Balance: Molecular, Cellular and Clinical Aspects, 9. Zucchelli P, Santoro A, Fusaroli M, Borghi M: Biofiltration in edited by De Santo NG, Capasso G, Naples, Italy, Bios Publi- uremia. Kidney Int Suppl 33[Suppl 24]: S141–S144, 1988 10. Daugirdas JT: Preventing and managing hypotension. Semin Dial cations, 1995, pp 279–284 7: 276–283, 1994 23. Tannen RL: Potassium disorders. In: Fluids and Electrolytes, 11. Levin A, Goldstein MB: The benefits and side effects of ramped 2nd Ed., edited by Kokko JP, Tannen RL, Philadelphia, W.B. hypertonic sodium dialysis. J Am Soc Nephrol 7: 242–246, 1996 Saunders, 1990, pp 195–300 12. Daugirdas JT: Second generation logarithmic estimates of single- 24. Bellizzi V, Terracciano V, Scheibel M, Conte G, Di Iorio B: Is pool variable volume Kt/V: An analysis of error. JAmSoc bioimpedance analysis (BIA) able to evaluate changes in total Nephrol 4: 1205–1213, 1993 body water in HD patients? [Abstract] Am J Kidney Dis 31: A10, 13. Sargent J, Gotch FA, Borah M, Piercy L, Spinozzi N, Schoenfeld 1998 P, Humphreys M: Urea kinetics: A guide to nutritional manage- 25. Kotler DP, Burastero S, Wang J, Pierson RN: Prediction of body ment of renal failure. Am J Clin Nutr 31: 1696–1702, 1978 cell mass, fat-free mass, and total body water with bioelectrical 14. Stefano JL, Norman ME, Morales MC, Goplerud JM, Mishra impedance analysis: Effect of race, sex and disease. Am J Clin OP, Delivoria-Papadopoulos M: Decreased erythrocyte Naϩ,Kϩ- Nutr 64[Suppl 3]: S489–S497, 1996 ATPase activity associated with cellular potassium loss in ex- 26. Chertow GM, Lowrie EG, Wilmore DW, Gonzalez J, Lew NL, tremely low birth infants with nonoliguric hyperkalemia. J Pe- Ling J, Leboff MS, Gottlieb MN, Huang W, Zebrowski B, diatr 122: 276–284, 1993 College J, Lazarus JM: Nutritional assessment with bioelectrical 15. NIH Consensus statement: Bioelectrical impedance analysis in impedance analysis in maintenance hemodialysis patients. JAm body composition measurement: National Institutes of Health Soc Nephrol 6: 75–81, 1995

Access to UpToDate on-line is available for additional clinical information at http://www.jasn.org/