Intradialytic acid-base changes and organic anion production during high versus low bicarbonate hemodialysis

Abstract

The use of high dialysate bicarbonate for hemodialysis in end-stage renal disease is associated with increased mortality, but potential physiological mediators are poorly understood. Alkalinization due to high dialysate bicarbonate may stimulate organic acid generation, which could lead to poor outcomes. Using measurements of β-hydroxybutyrate (BHB) and lactate, we quantified organic anion (OA) balance in two single-arm studies comparing high and low bicarbonate prescriptions. In study 1 (n = 10), patients became alkalemic using 37 meq/L dialysate bicarbonate; in contrast, with the use of 27 meq/L dialysate, net bicarbonate loss occurred and blood bicarbonate decreased. Total OA losses were not higher with 37 meq/L dialysate bicarbonate (50.9 vs. 49.1 meq using 27 meq/L, P = 0.66); serum BHB increased in both treatments similarly (P = 0.27); and blood lactate was only slightly higher with the use of 37 meq/L dialysate (P = 0.048), differing by 0.2 meq/L at the end of hemodialysis. In study 2 (n = 7), patients achieved steady state on two bicarbonate prescriptions: they were significantly more acidemic when dialyzed against a 30 meq/L bicarbonate dialysate compared with 35 meq/L and, as in study 1, became alkalemic when dialyzed against the higher bicarbonate dialysate. OA losses were similar to those in study 1 and again did not differ between treatments (38.9 vs. 43.5 meq, P = 0.42). Finally, free fatty acid levels increased throughout hemodialysis and correlated with the change in serum BHB (r = 0.81, P < 0.001), implicating upregulation of lipolysis as the mechanism for increased ketone production. In conclusion, lowering dialysate bicarbonate does not meaningfully reduce organic acid generation during hemodialysis or modify organic anion losses into dialysate.

INTRODUCTION

Despite improvements in the care of patients with end-stage renal disease (ESRD) over the last several decades, mortality remains unacceptably high. All-cause mortality rates for patients receiving dialysis are 6–8 times higher than for individuals in the general population, and only 51% of these patients are still alive after 3 yr of ESRD therapy (33a). To improve these outcomes, it is crucial to improve our understanding of modifiable factors. One such factor may be the use of high dialysate bicarbonate, which in a large international cohort study was implicated as a possible contributor to higher mortality rates: over a median followup of only 16 mo, the use of dialysate containing ≥38 meq/L bicarbonate, compared with ≤32 meq/L, was associated with 19% higher risk of death (33). However, the mechanism underlying this finding is still unknown. Given that the mean dialysate bicarbonate used in the United States is 37 meq/L, and of the >400,000 patients receiving hemodialysis in the United States over 40% receive dialysis using ≥38 meq/L dialysate bicarbonate (33), this is an urgent question to resolve.

To address this knowledge gap, it is important to understand the metabolic perturbations caused by high dialysate bicarbonate. One hypothesized effect is the stimulation of organic acid generation (8, 27), which is supported by a recent simulation model (28). During hemodialysis (HD), use of a high bicarbonate concentration in the dialysate creates a large dialysate-to-blood bicarbonate gradient, leading to rapid alkalinization that can result in alkalemia during and/or after HD (25, 38), which may cause deleterious effects. Alkalinization stimulates organic acid production in healthy individuals, specifically augmenting ketone generation (14); by contributing to net acid production during HD, this could create a negative feedback loop that modulates the overall rate of alkalinization (8). Other sequelae of increased organic acid generation are not yet fully understood; however, this may be a catabolic process that could contribute to the poor outcomes of patients receiving HD (28).

If organic acid generation substantially impacts acid-base homeostasis during HD, and if it is modulated by either the rate of alkalinization or the maximum achieved pH, then the prescription of lower bicarbonate dialysate levels might achieve similar acid-base balance while causing less metabolic perturbation. Despite longstanding concerns, the effects of alkalinization and the dialysate bicarbonate prescription on organic acid metabolism have never been systematically investigated. We hypothesized that lowering dialysate bicarbonate would reduce organic acid production and organic anion losses into dialysate. Therefore, we enrolled patients with ESRD receiving HD in two single-arm interventional studies comparing high and low bicarbonate prescriptions. In study 1, we systematically quantified each of the major sources of alkali gain and organic anion loss during HD and determined the effect of the bicarbonate dialysate prescription on organic anion generation. Next, to further examine the effects of alkalinization on organic anion production and on ketone generation in particular, we studied patients after they had achieved steady state on both a high and low dialysate bicarbonate prescription (study 2). Furthermore, because study 1 used a dialysate containing acetate, study 2 used citrate as the major organic anion in the dialysate, which enabled us to determine whether our findings in study 1 were due to the use of acetate per se or to a more generalizable process occurring during HD. As previous studies have shown that pH modifies lipolysis in humans (14, 16) and because increased lipolysis could stimulate ketone generation (21), we also examined free fatty acid (FFA) levels in relation to acid-base changes. Here, we integrate intradialytic acid-base changes with effects on organic anion metabolism during high and low dialysate bicarbonate conditions.

MATERIALS AND METHODS

Study Population

In study 1, we investigated 10 patients who had been receiving chronic thrice-weekly conventional HD for at least 3 mo and were at least 18 yr old. Exclusion criteria included the use of oral alkali within the previous month, hospitalization within the previous month, and pregnancy.

For study 2, we recruited 7 participants who had been receiving chronic thrice-weekly conventional HD for at least 6 mo and were at least 18 yr old. Exclusion criteria included the use of oral alkali within the previous month, hospitalization within the previous month, plans for kidney transplantation within the next 2 mo, planned change in phosphate binder therapy within the next month, and pregnancy or planned pregnancy within the next 2 mo. All participants provided written informed consent. Both studies were approved by the Institutional Review Board of the Albert Einstein College of Medicine and registered on ClinicalTrials.gov (NCT01777178, NCT02098356).

Study Design

In study 1, each participant was receiving outpatient HD with a standard prescription including 35 meq/L dialysate bicarbonate and 8 meq/L acetate; as part of the study, each participant received two HD sessions 1 wk apart in our hospital dialysis unit with serial blood and dialysate sampling and continued their standard outpatient HD in between the two study sessions. Participants were dialyzed against Fresenius Medical Care Optiflux polysulfone high-flux membranes during their outpatient HD sessions and the sessions in the hospital dialysis unit. The two HD sessions were performed after the long interdialytic interval using dialysate containing 2 meq/L K⁺, 2.25 meq/L Ca²⁺, 1.0 meq/L Mg²⁺, 4 meq/L acetate, 100 mg/dL dextrose, 137 meq/L Na⁺, and 100 meq/L Cl⁻. During the first session, HD was performed using a high bicarbonate dialysate (37 meq/L bicarbonate). One week later, during the second HD session, HD was performed using a lower bicarbonate dialysate prescription (27 meq/L bicarbonate). Other dialysis parameters were held constant, including treatment time, ultrafiltration rates, and blood and dialysate flow rates.

Study 2 was conducted at a local outpatient dialysis unit and also involved two HD sessions with serial blood and dialysate sampling. These participants were receiving outpatient HD against Gambro Revaclear polyarylethersulfone/polyvinylpyrrolidone high-flux membranes using a dialysate containing 35 meq/L bicarbonate, 2 meq/L K⁺, 2.5 meq/L Ca²⁺, 1.0 meq/L Mg²⁺, 0.3 meq/L acetate, 2.4 meq/L citrate, 100 mg/dL dextrose, 138 meq/L Na⁺, and 100 meq/L Cl⁻. After the first serial sampling session, their prescription was switched to a low bicarbonate dialysate prescription (30 meq/L) for 3 wk before the subsequent serial sampling visit. Both serial sampling sessions were performed following the long interdialytic interval.

In both studies, blood samples were drawn from the vascular access after cannulation but before the initiation of HD and then at 15, 45, 90, 135, and 180 min and at the completion of HD from both the arterial and venous lines connecting the patient to the dialyzer in study 1 and from the arterial lines only in study 2. These samples were used for arterial blood gas analysis and measurement of serum bicarbonate and β-hydroxybutyrate (BHB) and blood lactate. In study 2, serum FFAs were also measured at each time point. Dialysate outflow samples were drawn at the same time points. The dialysate inflow was sampled once at the initiation of HD. To assess postdialytic changes in pH and bicarbonate levels, blood samples were also taken 30 min and 60 min after completion of HD. Participants were permitted to eat during dialysis as per their usual practice. Only 3 of 17 participants ate during their dialysis sessions; because exclusion of these participants from our analyses did not change our findings, results are presented including these participants. Hemodynamic status was closely monitored. Blood pressure remained stable in all treatments with the exception of one treatment in one patient in study 1 (1 of the 3 participants who ate); as previously noted, exclusion of this participant did not affect our results.

Outcomes

We compared several outcomes between the two dialysate bicarbonate prescriptions. These included the changes in pH and bicarbonate during HD as well as bicarbonate and acetate gain and BHB and lactate losses via the dialyzer. We measured BHB and lactate specifically because these are the major organic anions produced during HD in prior reports (9, 35).

Calculations of the bicarbonate and acetate delivered and organic anions removed were performed following the method of Uribarri et al. (34). For each anion, we calculated a transfer rate across the dialyzer. This was measured as the difference between the content going into the dialyzer (measured in the arterial line) and content leaving the dialyzer (measured in the venous line). For each anion X, content into the dialyzer was calculated by the equation (Q_pi × [X]_i) + {Q_ri × ([X]_RBC/[X]_p) × 0.72[X]_i}, where Q_pi = inlet plasma flow = total blood flow – red blood cell flow, [X]_i = inlet plasma anion concentration (meq/L), Q_ri = inlet red cell blood flow = total blood flow × hematocrit, [X]_RBC/[X]_p is the ratio of red blood cell concentration to that in plasma, and 0.72 is the red blood cell water fraction (9). [X]_RBC/[X]_p = 0.69 for bicarbonate (9), [X]_RBC/[X]_p = 0.389 for BHB (3), and [X]_RBC/[X]_p = 0.5 for lactate (10). We assumed [X]_RBC/[X]_p = 1.0 for acetate because we were unable to find published data addressing this. We performed sensitivity analyses using 0.5 and 2.0 for the ratio of acetate to determine the effect of this assumption on our results.

Content leaving the dialyzer was calculated similarly using analogous quantities for the dialyzer outlet, with the addition of a correction factor to the bicarbonate content leaving the dialyzer to account for buffering of CO₂ by blood. Uribarri et al. (34) determined this using an in vitro titration curve; as our inlet hematocrit and inlet and outlet bicarbonate concentrations were similar to their values, we used their correction factor, specifically subtracting 2.4 meq/L from the outlet bicarbonate concentration.

The total gain or loss of each anion over each HD session was calculated by integrating its specific transfer rate over the duration of HD using the trapezoidal rule. Additionally, we quantified BHB and lactate losses using concentrations measured from the dialysate effluent. Similar to the method for the calculation of the transfer rate above, for each anion X we calculated the difference in content between the outlet and inlet, X_Do – X_Di. For each anion X, content into the dialyzer, X_Di, was zero. Content leaving the dialyzer was calculated by the equation [X]_Do × Q_Dtotal, where [X]_Do is the dialysate effluent anion concentration (meq/L) and Q_Dtotal is the sum of the dialysate flow rate and ultrafiltration rate. The difference X_Do – X_Di was integrated over the duration of HD, using the trapezoidal rule, to calculate the total loss of anion X.

Measurement of acetate was determined using the EnzyChrom acetate assay kit (BioAssay Systems), which uses enzyme-coupled reactions to form a colored, fluorescent product. The colorimetric protocol was chosen to better capture the acetate range found in plasma, serum, and dialysate. The 96-well plates were read by an Enspire reader set to 570 nm wavelength.

FFA and BHB were measured on a Beckman Coulter AU480 Chemistry Analyzer. FFA was measured using an enzymatic colorimetric method assay (Fujifilm Wako Diagnostics). Blood samples for FFA were collected in tubes containing the lipase inhibitor diethyl p-nitrophenyl phosphate (Sigma-Aldrich, St. Louis, MO) to prevent in vitro lipolysis. BHB was measured using the β-hydroxybutyrate LiquiColor Reagent System (Stanbio Laboratory). Serum IL-6 Chemiluminescence (R&D Systems, Minneapolis, MN) was measured in duplicate by ELISA.

Statistical Analysis

Data for continuous variables are presented as means ± SD. Total gains and losses of each anion, as well as net alkali gain, were compared between high and low bicarbonate dialysis sessions using paired t tests. For other parameters, changes over time between high and low bicarbonate dialysis were compared using random effects models in which time was modeled as a quadratic function. Effect modification of the association between dialysate bicarbonate and organic anion losses was tested using repeated-measures ANOVA with terms for the potential modifier, dialysate assignment, and modifier × dialysate interaction. Correlations were examined by calculating Pearson or Spearman correlation coefficients as appropriate.

RESULTS

A total of 17 participants completed the two studies, and their characteristics are presented in Table 1. Participants were longstanding HD patients and comorbidities were common, including diabetes mellitus, hypertension, coronary artery disease, and congestive heart failure. All were dialyzed via an arteriovenous fistula or graft. No changes to phosphate binder type or dose occurred during either study.

Table 1.

Participant characteristics

	Study 1	Study 2
n	10	7
Age, yr	49.8 ± 17.0	69.6 ± 14.0
Female/male, n (%)	6/4 (60/40)	3/4 (43/57)
Race/ethnicity, n (%)
Hispanic/Latino	4 (40)	2 (29)
Black	6 (60)	2 (29)
White	0 (0)	3 (43)
ESRD duration, mo	53.9 ± 83.2	41.7 ± 21.3
Diabetes, n (%)	6 (60)	6 (86)
Hypertension, n (%)	10 (100)	6 (86)
Coronary artery disease, n (%)	4 (40)	4 (57)
Congestive heart failure, n (%)	4 (40)	3 (43)
Peripheral vascular disease, n (%)	2 (20)	2 (29)
Smoker, n (%)	0 (0)	2 (29)
Body mass index, kg/m2	31.3 ± 6.7	30.5 ± 9.0
Obese (body mass index ≥30 kg/m2), n (%)	6 (60)	3 (43)
Dialysis characteristics
Duration, min	223.5 ± 21.7	205.7 ± 22.4
Blood flow rate, mL/min	385 ± 24.2	400 ± 76.4
Dialysate flow rate, mL/min	700 ± 0.0	671.4 ± 125.4
Total ultrafiltration rate, mL/h	856.5 ± 184.5	882.9 ± 163.9
Total ultrafiltration, L	2.8 ± 0.7	2.9 ± 0.7
Type of dialysis access, n (%)
AV fistula	9 (90)	5 (71)
AV graft	1 (10)	2 (29)
Phosphate binder, n (%)
Sevelamer carbonate	4 (40)	1 (14)
Sevelamer hydrochloride	2 (20)	0 (0)
Calcium acetate	1 (10)	3 (43)
None	3 (30)	3 (43)
Albumin, g/dL	4.1 ± 0.2	4.0 ± 0.3
Hemoglobin, g/dL	11.4 ± 1.4	11.7 ± 1.1
nPCR	not collected	1.2 ± 0.1
spKt/V	not collected	1.6 ± 0.3
IL-6, pg/mL	5.0 (3.6–5.7)	6.8 (4.7–14.4)

Data are presented as means ± SD or medians (interquartile ranges) for continuous variables; n = no. of participants. AV, arteriovenous; ESRD, end-stage renal disease; IL, interleukin; nPCR, normalized protein catabolic rate; spKt/V, single pool Kt/V.

Study 1

Acid-base status and organic anion levels.

The changes in acid-base parameters and acetate, BHB, and lactate levels over the course of HD are shown in Fig. 1. During the 37 meq/L session, pH rose from 7.39 ± 0.05 predialysis to 7.51 ± 0.03 at the end of HD. ${\text{HCO}}_3^{-}$ rose from 23.1 ± 2.8 meq/L to 29.5 ± 1.6 meq/L; the majority of this change occurred during the first 135 min of HD, with little change thereafter (note that for clarity, we chose to denote values measured in blood using ${\text{HCO}}_3^{-}$ and the dialysate concentration using the word “bicarbonate”). During the 27 meq/L session, pH was 7.41 ± 0.04 predialysis and 7.42 ± 0.03 at the end of HD, and ${\text{HCO}}_3^{-}$ declined slightly from 23.9 ± 2.7 meq/L pre-HD to 22.4 ± 0.8 meq/L at the end of the treatment; the discrepancy between the changes in pH and ${\text{HCO}}_3^{-}$ was explained by the intradialytic changes in Pco₂. Serum acetate did not change significantly during either treatment: 0.29 ± 0.39 meq/L pre-HD and 0.36 ± 0.35 meq/L at the end of treatment with 37 meq/L bicarbonate and 0.20 ± 0.39 meq/L pre-HD and 0.3 ± 0.33 meq/L at the end of treatment with 27 meq/L bicarbonate. BHB levels were similar during the two treatments until diverging at the end of HD; at 30 min post-HD, serum BHB with the 37 meq/L and 27 meq/L treatments was 0.3 ± 0.3 meq/L and 0.2 ± 0.2 meq/L, respectively. In 7 of 20 HD sessions, BHB was 5 meq/L or greater at the end of HD; these all occurred with 37 meq/L bicarbonate dialysate. Lactate levels differed over time between the treatments and appeared to diverge slightly during the latter half of HD; at 30-min post-HD, lactate levels with the 37 meq/L and 27 meq/L treatments were 1.3 ± 0.7 meq/L and 1.1 ± 0.7 meq/L, respectively. The area under the curve for lactate did not differ between treatments (217.7 ± 74.3 min × meq/L with 37 meq/L bicarbonate vs. 204.4 ± 68.5 min × meq/L with 27 meq/L bicarbonate, P = 0.54).

Fig. 1.

Systemic acid-base parameters and organic anion concentrations during hemodialysis (HD). Data are plotted for HD using 37 meq/L and 27 meq/L dialysate bicarbonate and represented as means ± SE (n = 10 participants). Postdialysis data are plotted separately for the subgroup of 7 participants who had blood samples collected at 30 min and 60 min after the end of dialysis. BHB, β-hydroxybutyrate.

Bicarbonate and organic anion balance.

Figure 2 shows the transfer rates of ${\text{HCO}}_3^{-}$, acetate, BHB, and lactate. For ${\text{HCO}}_3^{-}$, the rates using 37 meq/L and 27 meq/L bicarbonate dialysate, respectively, were initially 1.7 ± 0.7 meq/min and −0.6 ± 0.8 meq/min before gradually converging to 0.07 ± 0.4 meq/min at the end of HD. Acetate transfer rates were initially 0.29 ± 0.22 meq/min and 0.34 ± 0.17 meq/min with the 37 and 27 meq/L bicarbonate prescriptions, respectively; these remained fairly constant throughout HD and did not differ between treatments. The rate of BHB losses increased throughout HD but did not differ between treatments. The rate of lactate losses differed between 37 meq/L and 27 meq/L treatments and changed over time but not in a clear direction.

Fig. 2.

Transfer rates of bicarbonate, acetate, β-hydroxybutyrate (BHB), and lactate during hemodialysis (HD). Rates were calculated using measurements from the arterial and venous limbs of the HD circuit. Positive values represent gain into the systemic circulation; negative values represent loss into the dialysate. Data are plotted for HD using 37 meq/L and 27 meq/L dialysate bicarbonate and presented as means ± SE.

We next quantified cumulative gains and losses of ${\text{HCO}}_3^{-}$, acetate, BHB, and lactate during HD (Fig. 3, A and B). During the 37 meq/L bicarbonate HD session, mean ${\text{HCO}}_3^{-}$ and acetate gain were 94.5 meq and 60.7 meq, and mean BHB and lactate losses were 4.8 meq and 46.0 meq. During the 27 meq/L bicarbonate session, ${\text{HCO}}_3^{-}$ loss was 65.8 meq (P < 0.001 vs. 37 meq/L bicarbonate), acetate gain was 76.1 meq (P = 0.31), and BHB and lactate losses were 4.8 meq (P = 0.97) and 44.3 meq (P = 0.71), respectively. Net alkali gain was 104.3 meq using 37 meq/L bicarbonate dialysate compared with net loss of 38.9 meq with 27 meq/L dialysate (P < 0.001). Lactate accounted for 90% of organic anion losses in the two treatments.

Fig. 3.

Alkali and organic anion balance during hemodialysis (HD). A: cumulative bicarbonate, acetate, β-hydroxybutyrate (BHB), and lactate gains and losses over time. Data are means ± SE. B: net alkali balance plotted alongside total bicarbonate, acetate, BHB, and lactate gains and losses. C: net alkali and bicarbonate gains during the first 135 min of HD vs. the remainder of HD using 37 meq/L bicarbonate dialysate. D: smoothed locally weighted scatterplot smoothing (LOWESS) curves fitted to observed and predicted values of blood bicarbonate during HD with dialysate containing 37 meq/L bicarbonate (top 2 curves) and 27 meq/L bicarbonate (bottom 2 curves). Predicted values calculated from a random effects model incorporating the predialysis bicarbonate level and cumulative gain or loss of bicarbonate, acetate, BHB, and lactate at each timepoint. *P < 0.001.

The majority of alkali gain occurred during the first 135 min of HD (Fig. 3C): mean ${\text{HCO}}_3^{-}$gain in the first 135 min was 82.1 meq versus 12.4 meq in the remainder of the session (P < 0.001); corresponding net alkali gain in the two time periods was 89.3 meq and 15 meq, respectively (P < 0.001). Importantly, ${\text{HCO}}_3^{-}$, acetate, BHB, and lactate balance throughout HD, combined with the predialysis bicarbonate level, explained 92% of the variability in blood bicarbonate during the two HD treatments (Fig. 3D), suggesting that all major contributors to net alkali balance were accounted for.

Different values of [X]_RBC/[X]_p for acetate would have resulted in similar differences between the high and low bicarbonate treatments: if [X]_RBC/[X]_p = 0.5, cumulative acetate and net alkali gains would have been 52.1 meq and 95.7 meq during the 37 meq/L treatment and 65.8 meq and −49.1 meq during the 27 meq/L treatment; if [X]_RBC/[X]_p = 2.0, cumulative acetate and net alkali gains would have been 77.8 meq and 121.4 meq using 37 meq/L bicarbonate and 96.6 meq and −18.3 meq using 27 meq/L bicarbonate.

Study 2

Acid-base status and organic anion levels.

Participants achieved stable acid-base balance before both the 35 meq/L and 30 meq/L bicarbonate sessions: over the prior three HD treatments, the mean within-person standard deviations for predialysis pH and ${\text{HCO}}_3^{-}$ were 0.007 and 0.5 meq/L, respectively, with 35 meq/L bicarbonate dialysate and 0.014 and 0.7 meq/L, respectively, with 30 meq/L bicarbonate dialysate. Participants were significantly more acidemic before the 30 meq/L session (pH 7.39 ± 0.05 with 35 meq/L vs. 7.32 ± 0.03 with 30 meq/L dialysate, P < 0.001; ${\text{HCO}}_3^{-}$ 21.5 ± 2.4 vs. 18.0 ± 1.4 meq/L, P < 0.001) (Fig. 4). They were significantly more alkalemic at the end of the 35 meq/L session (pH 7.48 ± 0.05 vs. 7.44 ± 0.05, P = 0.003; ${\text{HCO}}_3^{-}$ 26.1 ± 1.9 vs. 21.4 ± 1.3 meq/L, P < 0.001). However, pH increased at similar rates during the two treatments (P = 0.37), and the change over time in ${\text{HCO}}_3^{-}$ also did not differ (P = 0.23). BHB levels were numerically higher before the 30 meq/L bicarbonate session and remained so throughout HD but did not differ significantly from the 35 meq/L treatment. Lactate levels were similar between treatments for most of HD but diverged at the end: 0.6 ± 0.1 meq/L versus 1.0 ± 0.7 meq/L for the 35 meq/L and 30 meq/L treatments, respectively.

Fig. 4.

Systemic acid-base parameters and organic anion concentrations during hemodialysis (HD) in patients with stable acid-base balance treated with high and low bicarbonate HD. Data are plotted for HD using 35 meq/L and 30 meq/L dialysate bicarbonate and presented as means ± SE (n = 7 participants). Postdialysis data are plotted separately for the subgroup of 5 participants who had blood samples collected at 30 min and 60 min after the end of dialysis. BHB, β-hydroxybutyrate.

Bicarbonate and organic anion balance.

To calculate BHB and lactate losses, we first examined data from study 1 to determine whether dialysate measurements provided a valid measure of organic anion losses. Comparison of organic anion losses calculated using blood sampling and dialysate sampling revealed that the latter yielded accurate results. Blood and dialysate measurements were highly correlated for both BHB (r = 0.96, P < 0.001) and lactate (r = 0.98, P < 0.001) (Fig. 5). Dialysate sampling results appeared to systematically underestimate lactate losses; blood lactate loss could be predicted from dialysate sampling by the following equation: blood lactate loss (meq) = 1.40 × dialysate lactate loss (meq) + 1.88 (R² = 0.96). Having thus validated the use of dialysate sampling, we calculated organic anion transfer rates in study 2 (Fig. 6A). The rate of BHB removal remained constant during HD (P = 0.69) and did not differ between treatment groups. Similarly, lactate removal rates remained stable throughout HD (P = 0.19) and did not differ between treatments (P = 0.35).

Fig. 5.

Comparison of blood and dialysate sampling for calculation of β-hydroxybutyrate (BHB) and lactate loss. Scatterplots, fitted regression lines, and Pearson correlation coefficients are from study 1 data. The line of unity is displayed on each plot.

Fig. 6.

β-Hydroxybutyrate (BHB) and lactate losses during hemodialysis (HD) using 35 meq/L and 30 meq/L dialysate bicarbonate. A: transfer rates of BHB and lactate. B: cumulative BHB and lactate losses. Rates were calculated using dialysate measurements. Data are presented as means ± SE.

As in study 1, total organic anion losses were not reduced by the change in dialysate bicarbonate (Fig. 6B). Net BHB losses in the dialysate were 4.0 ± 2.6 meq using 35 meq/L bicarbonate dialysate and 7.7 ± 9.2 meq using 30 meq/L bicarbonate (P = 0.25). Lactate losses were corrected using the equation above and were also similar between prescriptions: 34.9 ± 17.4 meq and 35.8 ± 23.9 meq (P = 0.79) with 35 meq/L and 30 meq/L bicarbonate dialysate, respectively.

Associations of organic anion levels with potential causative factors.

We hypothesized that ketone generation during HD was due to increased lipolysis. FFA levels increased from 0.25 ± 0.16 mmol/L at baseline to 0.62 ± 0.47 mmol/L at the end of HD (P = 0.03) using 35 meq/L dialysate bicarbonate and from 0.34 ± 0.25 mmol/L to 0.60 ± 0.45 mmol/L (P = 0.06) using 30 meq/L dialysate bicarbonate. FFA levels did not differ between treatments (P = 0.90). The change (Δ) in FFA levels during HD was highly correlated with ΔBHB (r = 0.81, P < 0.001) (Fig. 7), which was consistent with increased substrate availability contributing to increased ketogenesis. In contrast, ΔFFA was not correlated with Δlactate during HD (r = −0.18, P = 0.05). To examine whether intradialytic inflammation might contribute to the change in BHB and lactate during HD, we tested associations with serum IL-6 using pooled data from both studies. The median change in IL-6 during HD (post − pre) was 0.3 pg/mL (interquartile range: −0.6 to 3.2). ΔIL-6 was not correlated with ΔBHB (Spearman ρ = 0.17, P = 0.35) or Δlactate (Spearman ρ = −0.24, P = 0.17). Additionally, Δlactate was not correlated with ultrafiltration rate (r = 0.09, P = 0.61).

Fig. 7.

Correlations of the change in free fatty acid (FFA) levels with changes in serum β-hydroxybutyrate (BHB) and blood lactate during hemodialysis (HD). Data were derived from study 2. Correlations were examined by calculating Pearson correlation coefficients.

Next, using pooled data from the two studies, we examined effect modification by diabetes or obesity (body mass index ≥ 30 kg/m²). During HD, levels of serum BHB differed by both diabetes and obesity status, independent of dialysate bicarbonate; blood lactate differed by diabetes status (Fig. 8). Serum BHB levels were higher during HD among participants without diabetes and among those who were not obese (Fig. 8, top). In contrast, participants with diabetes had higher lactate levels throughout HD (Fig. 8, bottom); they did not have higher ultrafiltration rates than nondiabetics (9.1 ± 2.7 mL·kg⁻¹·h⁻¹ vs. 10.6 ± 3.5 mL·kg⁻¹·h⁻¹, respectively, P = 0.33). Differences in organic anion losses by diabetes and obesity mirrored the changes in blood levels (e.g., lactate losses were numerically higher among participants with diabetes) but were not statistically significant (Table 2). Neither diabetes nor obesity modified the effect of the dialysate bicarbonate prescription on serum BHB (P = 0.06 and 0.89, respectively) or blood lactate (P = 0.12 and 0.59, respectively). Furthermore, neither modified the effect of dialysate bicarbonate on total organic anion losses (P = 0.78 and P = 0.34 for diabetes and obesity, respectively); there was similarly no significant effect modification by diabetes or obesity when BHB losses (P = 0.31 and 0.05, respectively) and lactate losses (P = 0.06 and 0.15, respectively) were analyzed separately.

Fig. 8.

Systemic organic anion concentrations during hemodialysis (HD) by diabetes and obesity status. Data are plotted for HD from 2 studies using high (37 or 35 meq/L) and low (30 or 27 meq/L) dialysate bicarbonate and presented as means ± SE (n = 17 participants). Postdialysis data are plotted separately for the subgroup of 12 participants who had blood samples collected at 30 min and 60 min after the end of dialysis. Obesity was defined as body mass index (BMI) ≥ 30 kg/m²; n = 12 participants with diabetes and n = 9 obese participants.

Table 2.

OA losses into dialysate by diabetes and obesity status

	No Diabetes	Diabetes	P Value	Not Obese	Obese	P Value
n	5	12		8	9
Total OA loss, meq	30.1 ± 6.3	45.9 ± 20.4	0.09	36.3 ± 14.9	45.7 ± 21.1	0.29
BHB loss, meq	6.1 ± 4.4	4.1 ± 4.9	0.33	6.5 ± 6.0	3.0 ± 2.6	0.05
Lactate loss, meq	24.0 ± 7.4	41.8 ± 20.4	0.06	29.8 ± 14.5	42.7 ± 21.3	0.15

Analyses use pooled data from study 1 and study 2; n = no. of participants. Obese defined as body mass index ≥ 30 kg/m². P values were calculated using repeated-measures ANOVA and represent statistical significance of terms for diabetes and obesity, respectively. See text for details. BHB, β-hydroxybutyrate; OA, organic anion.

DISCUSSION

In two independent studies comparing high and low bicarbonate dialysis prescriptions, we found that lowering dialysate bicarbonate did not meaningfully alter organic anion metabolism during HD. In study 1, we examined the acute changes of modifying the dialysate bicarbonate prescription and found that even low bicarbonate dialysis that induced net bicarbonate loss was not sufficient to suppress organic anion production. In study 2, the effects of high and low bicarbonate dialysis were examined in patients who had achieved steady-state acid-base balance on each prescription. There as well, we found no differences in organic anion metabolism. Overall, our results do not support the hypothesis that alkalemia induced during HD causes significant organic acid production or speculation that lowering dialysate bicarbonate may reduce organic acid generation (28).

We hypothesized that alkalinization during HD would be a major source of organic acid generation because alkalinization stimulates lipolysis (13), thereby increasing FFA availability, which regulates ketone production (21). In vitro studies have suggested additional effects of alkalinization: increasing pH enhances fatty acid oxidation by reducing malonyl CoA inhibition of carnitine palmitoyltransferase 1 (22, 31), and higher extracellular pH and bicarbonate levels promote ketone body efflux from liver cells (7). In healthy adults, Hood et al. (13) reported that alkalinization with sodium bicarbonate to a pH of 7.46 produced rapid and robust increases in lipolysis and ketogenesis within 60 min, which is a time course and degree of alkalinization similar to that achieved during high bicarbonate HD in our study.

We did indeed find evidence of net organic acid generation during dialysis. BHB and lactate levels in the circulation remained stable or increased despite continuous losses of these organic anions into the dialysate. The strong agreement between the arteriovenous difference of these anions across the dialyzer and measured dialysate losses suggests that quantitatively important transcellular shift did not occur; therefore, maintenance of plasma levels would require ongoing generation. Furthermore, in the report by Hood et al. (13), the rise in circulating ketone levels mirrored increased BHB production measured by plasma appearance rates. Thus, it is highly likely that dialysis induced organic anion generation.

However, contrary to our hypothesis, organic anion generation was not due to the alkalinization induced by high bicarbonate dialysate, as modification of the bicarbonate prescription did not affect dialysate losses and had only a minor effect on circulating levels of BHB and lactate. Furthermore, lactate (rather than the ketone BHB) accounted for the vast majority of organic anion losses into dialysate. Although we cannot comment on the full set of organic anions that may be affected by HD, our estimates are consistent with prior studies: Gotch et al. (9) reported 31 meq losses over 4 h using a 36 meq/L bicarbonate bath concentration and slower blood and dialysate flow rates of 200 and 400 mL/min, respectively; similarly, 27 meq losses were estimated from 4.5-h treatments also using 200 mL/min blood flow rates (28, 35). Additionally, the consistency of results between the two studies presented here, one of which used a citrate-based dialysate, rules out the use of acetate as the etiology and points to a more generalizable process occurring during HD.

Our data suggest that organic anion metabolism during HD should not be viewed as a single metabolic process. BHB production appears to be the end product of intradialytic lipolysis: FFA levels increased during dialysis, indicating upregulation of lipolysis; the mean increase of 0.37 mmol/L during HD with 35 meq/L dialysate bicarbonate was similar in magnitude to the 0.2 mmol/L increase induced by alkalinization in healthy adults (13); and the increase in FFA strongly correlated with increasing BHB levels, consistent with increased substrate availability stimulating ketogenesis. Several factors could contribute to upregulation of lipolysis during HD. Fasting activates lipolysis, but this is unlikely to be a major cause because FFA levels after ~4 h of HD were similar to those reported in healthy humans after 12–15 h of fasting (17, 36). Activation of the sympathetic nervous system could also cause lipolysis, as could heat loss during HD (4, 26, 29). Additionally, heparin has been reported to increase plasma FFAs (19); however, when samples are collected with a lipase inhibitor, as in our study, there may be no heparin-related increase in FFA concentration when used as an anticoagulant during HD (6). Finally, it is worth noting that the intradialytic increase in serum BHB was greatest among nondiabetic and nonobese participants. This suggests that ketogenesis was more pronounced among participants with less metabolic dysregulation. We cannot determine, using the data collected, whether this is due to differences in lipolysis or rather the utilization of FFAs or whether greater autonomic dysfunction among patients with diabetes or the presence of exogenous insulin may explain this finding.

Lactate generation requires an alternative explanation, as the change in lactate during HD was not correlated with the change in FFA. In contrast to BHB, lactate levels were higher among participants with diabetes, which suggests underlying metabolic dysregulation may predispose to lactate generation in response to the stressors introduced by HD. Lactate production could be a marker of tissue ischemia resulting from rapid ultrafiltration, such as that reported in the myocardium and brain during HD (24, 37), but the change in blood lactate did not correlate with ultrafiltration rate. Our studies do not permit us to distinguish between increased generation of lactate and reduced metabolic clearance, which is also a possibility; this is deserving of further study.

It is not clear whether these metabolic perturbations impact clinical outcomes, but data suggest they could: FFA availability affects myocardial substrate utilization and may have detrimental effects in the setting of ischemia (12, 30), which occurs during routine HD treatments (5); high ketone levels may reflect increased sympathetic nervous system activity (18), and blood ketones predict mortality in HD patients (15, 23).

Our findings have additional implications for the prescription of bicarbonate dialysis. There has been debate about whether to consider the concentration of acetate or citrate a part of “total buffer” and to reduce the bicarbonate prescription accordingly. The data presented here suggest this is inappropriate. Net alkali gain during HD was primarily determined by bicarbonate flux across the dialyzer, which was driven by the blood-dialysate bicarbonate gradient. Although in study 1 acetate contributed a substantial proportion of the total alkali gained, the intradialytic increase in blood bicarbonate slowed dramatically as the dialysate-blood bicarbonate gradient shrank. Thus, the addition of acetate or another organic anion likely increases the rate of alkalinization, but the bicarbonate gradient diminishes more rapidly as a result. Any circulating acetate or citrate remaining at the end of dialysis will be metabolized to generate bicarbonate, but as our data show, the increment in bicarbonate is minimal. Therefore, it is the bicarbonate concentration in the dialysate, rather than total buffer, that determines the postdialysis bicarbonate level.

Several additional findings related to intradialytic acid-base changes are worth noting. In study 1, although the dialysate bicarbonate was higher than the blood bicarbonate level during both treatments, the dialysate-blood bicarbonate gradient approached zero during high bicarbonate HD, and during the low bicarbonate treatment there was net bicarbonate loss. This is likely due to three factors that reduce net bicarbonate diffusion from dialysate into plasma: 1) because proteins and lipids account for ~7% of plasma volume, the bicarbonate concentration in plasma water is higher than the measured value, thereby reducing the concentration gradient for diffusion (11); 2) the Gibbs-Donnan effect reduces movement of anions into the plasma space (11); and 3) convective loss of bicarbonate occurs due to ultrafiltration. Diffusion of bicarbonate into plasma may be further reduced by the need to maintain electroneutrality as cations such as K⁺ diffuse from plasma into dialysate.

The effect of respiratory changes during HD is also worth noting, as these changes contribute to the magnitude of alkalemia during high bicarbonate HD. Pco₂ initially increased during HD but declined by the end of dialysis. This is consistent with prior work demonstrating a progressive increase in minute ventilation and CO₂ excretion during the first hour of HD, which was sustained throughout the dialysis treatment and into the early post-HD period (32), and another report of rising Pco₂ during the first hour of HD that was followed by a decline throughout the remainder of the treatment (1). It is not entirely clear why Pco₂ fell below baseline during low bicarbonate HD in study 1. In part, this could be due to a respiratory response to loss of bicarbonate. Additionally, mild hypoxemia during HD may have stimulated ventilation, which would have further driven down Pco₂ (20). Greater delivered bicarbonate during high bicarbonate HD would account for the larger increase in Pco₂ during the first hour of HD and the numerically higher levels that were sustained throughout dialysis.

Several important limitations of our studies should also be noted. First, we did not directly measure organic anion generation rates, but careful quantification of organic anion and bicarbonate balance across the dialyzer and time-course studies of blood levels throughout HD and immediately thereafter enabled us to thoroughly evaluate any clinically meaningful effects on organic acid metabolism. Furthermore, we measured only the acute effects of the HD treatment and cannot comment on whether effects of different dialysate prescriptions on organic anion production may have been seen during the interdialytic interval. We only studied patients receiving HD via an arteriovenous access, but we did not assess access recirculation; however, it is unlikely this is an important concern given both the agreement between arteriovenous and dialysate balance measurements and the excellent performance of our regression model in terms of explaining intradialytic changes in circulating bicarbonate levels. Furthermore, bicarbonate gain during high bicarbonate HD (37 meq/L) in study 1 was 94.5 meq, similar to the gain of 96 meq calculated by Uribarri et al. (34) using a 36 meq/L bicarbonate prescription. We were unable to find published data on acetate partitioning between red blood cells and plasma but tested the effect of our correction factor, and this did not affect comparisons of BHB and lactate metabolism between the two arms in each study; furthermore, using our assumed value for acetate distribution, our regression model accounted for a large proportion of bicarbonate variability during HD. Finally, we performed these studies in small numbers of individuals; as such, they may not be generalizable to all dialysis patients, and we cannot rule out the possibility that there is a subset of patients in whom the use of high dialysate bicarbonate per se spurs substantial organic acid generation. However, our two independent studies were conducted in separate sets of patients from different dialysis units; they also used complementary study designs (e.g., acetate vs. citrate-based dialysate, acute effects of changing dialysate bicarbonate vs. achievement of steady state), which enabled us to thoroughly interrogate the effects of intradialytic alkalinization.

In conclusion, lowering the dialysate bicarbonate prescription does not meaningfully reduce organic acid generation during hemodialysis or modify organic anion losses into dialysate. Whether there are other adverse effects of high bicarbonate dialysate, and whether lowering the bicarbonate prescription produces clinical benefit, is deserving of further study. Hemodialysis does stimulate organic acid production, in part due to upregulation of lipolysis. Investigation of these metabolic perturbations and their clinical significance is warranted.

GRANTS

This work was supported by an American Society of Nephrology Carl W. Gottschalk Research Scholar Grant, by Clinical and Translational Science Award Grants UL1-TR-001073 and UL1-TR-002556 from the National Center for Research Resources, a component of the National Institutes of Health, and by the Einstein-Mt. Sinai Diabetes Research Center (5-P30-DK-020541-41).

DISCLAIMERS

The contents of this study are solely the responsibility of the authors and do not necessarily represent the official views or policies of the National Center for Research Resources or National Institutes of Health.

DISCLOSURES

M. K. Abramowitz and T. H. Hostetter have consulted for Tricida, Inc. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

W.P., T.H.H., and M.K.A. conceived and designed research; W.P., M.C., N.G., D.S., A.B., R.I.L., S.M.R., and M.K.A. performed experiments; S.P. and M.K.A. analyzed data; S.P. and M.K.A. interpreted results of experiments; S.P. and M.K.A. prepared figures; S.P., W.P., and M.K.A. drafted manuscript; S.P., W.P., M.C., N.G., D.S., A.B., R.I.L., S.M.R., T.H.H., and M.K.A. edited and revised manuscript; S.P., W.P., M.C., N.G., D.S., A.B., R.I.L., S.M.R., T.H.H., and M.K.A. approved final version of manuscript.