Sodium Prescription in the Prevention of Intradialytic Hypotension: New Insights into an Old Concept

An excessive sodium load is associated with a high mortality in patients with end-stage renal disease on chronic hemodialysis (HD) [1]. A post hoc analysis from the HEMO study considering 1,800 chronic dialysis patients showed a significantly increased risk of death with a dietary sodium load >2.5 g/die. In addition to dietary sodium intake, sodium loading in dialysis patients is due to the diffusive gradient from dialysate to blood [1]. Keen and Gotch [2] found that a 2-10 mEq rise in sodium gradient from dialysate to blood increases the patient's sodium load due to an increase in sodium gained from the dialysate in comparison to that gained from the daily diet. Flanigan [3] found a positive sodium gradient from dialysate to blood in more than 75% of 120 patients with a predialytic natremia of 137 ± 4 mEq/L and a sodium dialysate concentration of 140 mEq/L. Studies on behalf of the Renal Research Institute undertook a 3-year follow-up of 4,000 HD patients finding a predialytic natremia between 139 and 140 mEq/L related to a positive sodium gradient between dialysate and blood in more than 40% of the patients when dialysate sodium was 140 mEq/L and in more than 20% when the dialysate sodium was 138 mEq/L [4,5]. The extracellular volume may increase during excessive sodium load due to both the water shift from the intercellular space and thirst which leads to high interdialytic weight gain and arterial hypertension [6]. Hypertension, in turn, can have a strong impact on patients' morbidity and mortality [7]. The weight gain raises the ultrafiltration rate during dialysis, which is associated with myocardial stunning and increased mortality [8]. Penne and Sergeyeva [5] showed an increased odds ratio for intradialytic morbidity in patients with positive sodium gradient (>6 mEq/L) from blood to dialysate. Mc Causland et al. [1] enrolled 2200 HD patients during a 30-month follow-up finding that a >140 mEq/L sodium concentration in the dialysate increases mortality in patients with predialytic natremia >140 mEq/L.

It is plain from the previous excursus that a diffusive sodium gradient from dialysate to blood should be avoided. To prevent intradialytic patient hypotension, excessive sodium removal by diffusion should also be avoided using isonatric dialysis in which sodium is removed by convection alone. A large amount of sodium can be removed simply by ultrafiltration, with the great benefit of avoiding any alteration in plasma osmolarity. For example, if 3 L are removed by ultrafiltration during a session, and plasma sodium concentration is 137 mEq/L, the total sodium mass removed is as high as 411 mEq. This quantity corresponds to a sodium intake of up to approximately 205 Eq/day considering the 2 previous interdialytic days.

However, to achieve isonatric dialysis, the dialysate sodium can also be aligned to the patient's sodium “setpoint.” Indeed, individual dialysis patients have their own predialysis sodium values that change very little over time (1.6%). This value is the so-called sodium “setpoint” and is generally computed by calculating the monthly mean of the patient's sodium values [3]. Dialysis patients tend to maintain their own sodium setpoint increasing the water gain to balance the excess in dietary sodium that cannot be decreased by urine output. For example, a sodium intake of 8 g per day needs the intake of 1 L of water to maintain the sodium setpoint [9]. Nonetheless, the use of the sodium setpoint has its drawbacks: the correlation between salt intake and thirst is not a constant in a dialysis patient because osmoreceptors function may be impaired and because the setpoint does not consider the hypotonic hyperhydration due to the osmotic burden unrelated to sodium [10].

Reducing the diffusive sodium gradient of sodium from dialysate to blood and decreasing sodium gain with the diet will diminish thirst and lead to negative sodium and water balance. This favors the normalization of arterial pressure and cardiac remodelling, but triggers intradialytic hypotension, presumably, as blood is dialyzed, plasma osmolarity drops from approximately 310 mOsm/L to approximately 290 mOsm/L. When dialyzed blood is reinfused into the patient, whose plasma osmolarity is 310 mOsm/L, the osmotic gap is dissipated when water moves out of the plasma into the interstitial and intracellular space (ICV) [11]. This process reduces plasma volume and fosters edema even in the absence of ultrafiltration. The different mass transfer coefficients yield the following times for equilibrium to take place between the extracellular space (ECV) and the ICV: 30 s for osmotic equilibrium (because of the high cell wall permeability to water), 20 min for urea equilibrium and over 10 h for equilibrium of the ICV-ECV sodium concentrations. A mathematical model devised by Mann and Stiller [12] calculates the volume shift as a function of eliminated quantities of solutes; in particular, it considers sodium to be the very effective osmole: if the extracellular sodium is decreased by 5 mmol/L and osmotic equilibrium is established, the ECV decreases by 6%. These events cause intradialytic hypotension that involves 20-50% of HD sessions and can significantly increase the risk of death in the short and long term [13].

The European Dialysis and Transplant Association and K-DOQI defines hypotension as either a fall in systolic blood pressure (SBP) ≥20 mm Hg or a 10 mm Hg decrease in mean arterial pressure in combination with clinical symptoms requiring intervention [14,15]. Repeated episodes of intradialytic hypotension have been associated with increased mortality in maintenance HD patients, particularly when an absolute nadir SBP <90 mm Hg was reached [16,17]. During a hypotension episode, the diastolic blood pressure (DBP) falls and as the myocardium is perfused during diastole, cardiac ischemia can result due to the drop in perfusion pressure. Repeated episodes of cardiac hypoperfusion may lead to increased cardiac fibrosis and myocardial stunning [18,19]. The commonest cause of intradialytic hypotension is an ultrafiltration rate exceeding the corresponding plasma refilling rate long enough to reduce the plasma volume to a critical level [18]. A number of advances in dialysis machine technology have reduced the risk of intradialytic hypotension (Table 1) [20,21,22,23]. Nonetheless, sodium prescription is the main modifiable factor that can improve the intradialytic tolerance of chronic dialysis patients because a serum sodium change of 1 mEq/L is the osmotic equivalent of a 6 mg/dL change in blood urea nitrogen (2 mmol urea) or the oncotic gradient produced by 10 g/dL of serum protein [11]. Brummelhuis et al. [24] demonstrated that sodium profiling gradually declining from 150 to 140 mmol/L improves the plasma refill rate, whereas a positive effect of cool dialysate 1°C below core temperature could not be established. During profiled HD, the sodium or conductivity dialysate values must not be maintained at the same high levels throughout the dialysis session. To avoid the risk of an elevated sodium level or high dialysate conductivity causing an increase in thirst and hence a higher interdialytic weight gain, dialysis machines are currently equipped with different physiological controllers. A mathematical model can be employed as a physiological controller to reach any clinical target, namely, end-dialysis sodium, blood volume decrease, arterial pressure decrease, and weight loss. The model needs to receive real-time signals of the patient's clinical status from specific sensors. This ability to measure the patient's status and automatically control the clinical targets is called “biofeedback” [25,26]. As a result, conductivity and ultrafiltration profiling during dialysis need (a) a mathematical model to reach the neutral sodium mass balance automatically; (b) a biofeedback system for real-time modulation of the conductivity and ultrafiltration profiles.

The sodium mass to be removed can be determined from a nomogram in which the mass is the result of the difference between dietary sodium intake and physiological loss [27]. Alternatively, the sodium mass can be calculated using the so-called sodium target, which is the average blood sodium value at the end of the last 12 dialysis sessions when the patient is at his/her dry weight [28,29]. After calculating the sodium mass, the best strategy must be devised for its removal.

Colì et al. [27] and Ursino et al. [30] have developed the “Profiler,” an algorithm that makes use of a mathematical model comprising a 2-compartment description (intracellular and interstitial) of sodium and urea kinetics (Fig. 1). The model is applied to calculate sodium and ultrafiltration profiles as concave downward curves whose apices correspond to 60 min after the start of the dialysis session [31]. These profiles are computed from the patient's initial data: session timing, body weight to be lost, predialysis concentration of sodium and urea, and the end-dialysis sodium target. In particular, model equations are used to simulate the time pattern of sodium concentration, starting from predefined profiles of dialysate sodium and ultrafiltration rate. These profiles are then modified with an iterative procedure to establish the sodium mass removal (or the required sodium target) and the dry weight needed. This strategy has been adopted to achieve a neutral sodium balance (interdialytic dietary intake = dialytic removal) and counteract the drop in plasma osmolarity occurring during the first hour of dialysis following the rapid clearance of low molecular weight solutes. This implies a reduced sodium extraction and greater ultrafiltration during the first half of the dialysis session followed by reduced ultrafiltration and increased sodium removal during the second half. During the first half of the dialysis session, the sodium profile prevents a sharp fall in plasma osmolarity and increases the plasma refilling rate. The excess fluid is drained from the intracellular compartment and can be removed by means of a high ultrafiltration rate, thereby achieving most of the weight loss required. During the second half of the dialysis session, the ultrafiltration rate is automatically reduced together with the sodium profile to allow excess sodium removal [31]. The Profiler was successfully validated in experiments and proved to give accurate descriptions of solute kinetics during the dialysis session, maintaining the strategy to increase the ECV/ICV ratio [31]. Considering that there are osmotic sodium and non-osmotic body sodium stores, the Profiler mathematical model was revised and the intracellular sodium compartment was included in the model with a detailed description of interstitial elastance and interstitial fluid pressure [32,33]. The model forecasts the sodium mass balance during the 4-h dialysis. It is up to the physician to check the patient's clinical status and modify the sodium target to maintain the sodium mass balance in the long term. Plasma osmolarity time pattern prediction will be further improved when the model will be integrated with the glucose kinetics. For the time being, the model takes the intradialytic plasma glucose mass as approximately constant because a glucose-containing dialysate fluid is currently used, and this counteracts the diffusive loss of glucose across the dialyzer membrane.

The Profiler system was then validated in the clinical setting enrolling 20 hypotension-prone dialysis patients (Table 2). Blood volume and cardiac output during Profiler HD showed a lower decrease than on standard HD, a more stable intradialytic mean blood pressure and a lower heart rate increase than the values obtained during standard HD [34]. The Profiler was then used extensively in a multicenter study involving 15 institutions enrolling 55 patients (Table 2). A 6-month follow-up comparing 642 dialysis sessions using the usual technique and 2,376 dialysis sessions implementing the Profiler system showed (a) a significant improvement in SBP, DBP and heart rate during dialysis, (b) a neutral sodium balance (no significant increase in predialysis blood sodium level or interdialytic weight gain), and (c) an improvement in disequilibrium syndrome symptoms [35]. As to hypotension-prone diabetic patients, Colì et al. [35] enrolled 18 patients with diabetes out of 55 (33%), the patients were on strict antidiabetic therapy and differences between diabetic and non-diabetic patients were not reported.

The Profiler system was then adapted to hemodiafiltration (HFR) with reinfusion of the endogenous ultrafiltrate, a technique indicated for the treatment of dialysis patients with inflammatory syndrome and malnutrition [36]. HFR is obtained by means of the HFR double chamber filter (Bellco, Mirandola, Italy). The first part of the filter consists in a polyphenylene high flux hemofilter. The ultrafiltrate is driven from the hemofilter to a 40 g neutral styrene resin with an adsorbing area of 28,000 m². The resin does not adsorb either sodium and electrolytes or urea [37]. After adsorption, the ultrafiltrate is added to the whole blood that then passes into the second part of the filter, a polyphenylene low flux filter where the weight loss and diffusive depuration take place. The mathematical model is applied in the second filter to compute the sodium concentration and ultrafiltration profiles in the dialysate. The mathematical model called Profiler takes into account the Gibbs-Donnan effect as reported by Ursino et al. [36] during the Profiler validation in HFR. Using 9 HFR sessions on 9 chronic HD patients (one for each patient), the time course of plasma solutes and osmolarity measured every 30 min during HFR was compared with those predicted by the model. The average deviations between model and real data (sodium: 1.9 mEq/L; potassium: 0.32 mEq/L; urea: 1.04 mmol/L; osmolarity: 5.02 mosm/L) are of the same order as measurement errors and similar to those obtained using our previous models in standard and profiled HD. Sodium concentration prediction only slightly worsens (from 1.9 to 2.02 mEq/L) if default values are used for the initial value of other solutes in blood (i.e., if the algorithm uses only initial body weight and initial sodium concentration in plasma) [36].

A limitation of the Profiler mathematical model is that the initial plasma sodium concentration is not known or can only be partially inferred from the patient's history. This technique requires the use of a sodium sensor to evaluate plasma sodium concentration during the individual session (Fig. 2). The sodium sensor yields continuous real-time measurements of serum sodium, and passes this information to the mathematical model.

The sodium sensor, known as Natrium®, comprises a probe with 2 conductive elements. Each element has a first contact surface facing the inside of the probe and designed to be in contact with the endogenous ultrafiltrate, and a second contact surface facing the outside of the probe in connection with the device for measuring conductivity [38]. The sensor is applied on the endogenous ultrafiltrate circuit and reads the endogenous ultrafiltrate conductivity before the resin and before the second HFR17 filter. The sensor measures the conductivity on the endogenous ultrafiltrate (or plasma water) obtained by mechanical ultrafiltration. This conductivity measurement was experimentally validated during HFR treatments [39]: the sensor measurements showed a significant correlation with the corresponding serum sodium measurements by indirect potentiometry (Fig. 3). The differences observed between the mean of the sodium values obtained by Natrium and those obtained by the current laboratory methods are 0.5 mEq/L. The accuracy of Natrium is 1.2%. The difference observed between the mean sodium values obtained by Natrium at dialysis end and the values of the sodium target is 1.0 mEq/L [39]. An equation ([Na⁺] = 13.95* C_uf - 53.48) is currently used to correlate endogenous ultrafiltrate conductivity and plasma sodium [40]. In turn, the correlation equation calculates the plasma sodium to be given in real time to the mathematical model [40], which then recalculates the patient's predialysis sodium profile by considering the value at 15 min and at 60 min from dialysis start (Fig. 4, 5). Subsequently, some model parameters are adjusted on the basis of the discrepancy between the initial model predictions and the real-time sodium sensor measurement, and the profiles recalculated to ensure the desired sodium target. This produces biofeedback that permits ongoing adjustment of the model to the individual patient. This HFR system is called HFR Aequilibrium (HFR-Aeq).

HFR-Aeq was compared with traditional HFR in an international multicenter randomized study (the AIMS study) that assessed the efficacy of the HFR-Aeq system on intradialytic hypotension (Table 2). The study compared 923 HFR sessions with 988 HFR-Aeq sessions in 43 patients prone to intradialytic hypotension. Intradialytic hypotension was defined according to Colì et al. [35]: (i) an SBP value ≤90 mm Hg in patients with a predialysis SBP value >100 mm Hg, even if not accompanied by symptoms and therapeutic interventions; (ii) any SBP reduction ≥25 mm Hg compared to the predialysis value in the presence of symptoms and therapeutic maneuvers; (iii) an SBP reduction of at least 10% accompanied by typical symptoms (nausea, vomiting, headache, dizziness) in patients with a predialysis SBP value <90 mm Hg. The results showed the following indications: (a) a significant reduction of dialysis sessions complicated by hypotension using HFR-Aeq; (b) a significant reduction of symptomatic hypotension cases (23 ± 3% in HFR-Aeq vs. 31 ± 4% in HFR, p = 0.03); (c) a reduced onset of intradialytic morbid events requiring nursing interventions (17 ± 3% of sessions with at least one intervention in HFR-Aeq vs. 22 ± 1% of sessions with at least one intervention in HFR, p < 0.01) [40]. SBP significantly increased in HFR-Aeq 60 min after the start of dialysis with respect to values measured in HFR, whereas DBP values rose significantly in the second half of the dialysis session with HFR-Aeq compared to HFR. Dividing patients into quartiles according to the percentage of dialysis sessions complicated by hypotension, a significant reduction in the number of hypotension events was found in patients presenting arterial hypotension in more than 75% of dialysis sessions [40]. In addition, HFR-Aeq treatment maintained hemodynamic stability during dialysis with preservation of patients' sodium balance: no differences were found between HFR and HFR-Aeq serum sodium concentrations predialysis (138.6 ± 0.5 vs. 139.4 ± 0.5 mEq/L) and post-dialysis (139.2 ± 0.2 vs. 139.7 ± 0.52 mEq/L). End-dialysis serum sodium in HFR-Aeq was equal to the sodium target (139.7 ± 0.2 vs. 139.0 ± 0.1 mEq/L) [40]. Nonetheless, Locatelli et al. [40] considered 13 chronic HD patients with diabetes out of 50 (26%); they were on strict anti-diabetic therapy. Differences between diabetic and non-diabetic patients were not reported to be the aims of the trial. The findings of the AIMS study were also confirmed by a multicenter study on HFR-Aeq and intradialytic cardiovascular stability carried out in one Italian region (Lazio), Table 2[41].

Maintaining a correct sodium balance and a controlled reduction of osmolarity during dialysis is challenging but not impossible. Sodium and ultrafiltration profiles combined with a correct sodium balance have a positive effect on intradialytic hemodynamic stability. This strategy can be optimized by monitoring intradialytic systems and adjusting the profiles in real time on the basis of sodium and ultrafiltration kinetics.

Nonetheless, the Profiler HD procedure described in this work involves several possible errors, which should be clearly recognized, and may be the target of future technical and theoretical improvements. These can be classified into errors in the dialysis procedure, errors in data measurements, and errors induced by model limitations.

The first aspect includes possible inaccuracies in the dialysate composition and in the delivered acid-base balance, and errors in the determination of the patient sodium intake. With regard to sodium intake, Maduell and Lambie, considering small caseloads of HD patients, found that the daily salt load varied between 170 and 250 mmol/die [42,43]. These may affect the actual final sodium target, which can differ from that presumed by the procedure.

Data used to validate the model are generally achieved with the potentiometry technique, which is quite inaccurate in terms of sodium concentration: indirect potentiometry (ion selective electron potentiometry, Beckman Coulter®, Rome, Italy) is currently used to determine sodium concentration. The coefficient of variation of laboratory sodium determination is 0.64%, while the standard deviation is 0.84 mEq/L. These errors can be reflected inaccurately in the model validation, especially when some parameters are estimated directly from the data set (we remind that the higher the inaccuracy in the measurement, the higher the variance of parameter estimates).

Finally, all physiological models include some simplifications and limitations, since they represent a compromise between simplicity and completeness. To be useful in a clinical set up, a model should be as simple as possible, in order to favor direct implementation in a dialysis machine, allow real time computation, and permit a more straightforward analysis of results. However, each simplification involves some unavoidable errors compared with the complexity of a real physiological system. Among the main points deserving future study, we can mention the work of active ionic transport pumps (in particular, the Na⁺-K⁺ pump) and the effect of cardiovascular regulation mechanisms on fluid pools [44]. Finally, accounting for the large individual variability requires implementation of ad hoc procedures for personalized parameter estimates, which in turn are influenced by data inaccuracy. All these aspects are challenges for future improvements.

This research was supported by a grant from the University of Bologna. Project entitled: “Ricerca fondamentale orientata.” Principal investigator: Gaetano La Manna. Anne Collins edited the English text.

The authors declare that there are no conflicts of interest to disclose.