In this multicenter, controlled, prospective, randomized clinical study, we found that our modified regional citrate anticoagulation protocol, namely, RCA-two in a Ca2+-containing dialysate, was superior compared to either saline flushing or traditional RCA-one in HD patients with a high bleeding risk. Patients on HD have an increased risk of bleeding caused by uremia-associated platelet dysfunction, other defects of hemostasis, and anticoagulation with heparin [20-22]. Approximately 8-36% of dialysis patients are treated with oral anticoagulants or antiplatelet agents for cardio-cerebrovascular complications [23,24] with 0.05–0.22 events/year of major bleeding rates worldwide [25,26]. In addition, perioperative management of HD patients for surgery, e.g., for parathyroidectomy, vascular access surgery, trauma or renal transplantation, has become more and more routine with a systemic anticoagulation contraindication [27]. For the medical insurance policy in China, such a large population may not achieve an adequate dialysis dose after interruption by frequent serious clotting.
Prior clinical studies have explored RCA under different conditions. An open-label single-center prospective clinical trial including 33 patients at risk of bleeding demonstrated that RCA using a Ca2+-free dialysate (RCA-Ca0) revealed fewer clotting events than RCA-one (RCA-Ca3), or anticoagulant-free hemodialysis, while Ca2+ supplementation is necessary for the RCA-Ca0 group [9]. However, the cumbersome and laborious procedure during RCA-Ca0 had limited its clinical application. Subsequent studies detected an RCA effect in other forms with simpler techniques, for example, a Ca2+-containing dialysate or a hemodialysis filtration (HDF) procedure. Evenepoel et al. found that RCA-one using a Ca2+ dialysate for HD was shown to be safe and relatively effective but still exhibited moderate blood clotting compared with heparin-free hemodialysis (17% vs. 45%, P < 0.05) [9]. Later studies found that RCA-one resulted in significant clotting (up to 40%) in the venous bubble trap [28, 29]. However, these observational studies had methodological limitations to some extent. Ponikvar only observed RCA efficacy in HDF, without any contrast or necessary safety assessment. There were no adjustments for gender, age, weight, Ca2+, or blood platelets. Generally, RCA-one was inferior in the venous bubble trap, probably caused by an elimination of TSC by the dialyzer and restoration of Ca2+ by dialysate.
To solve this problem, we modified the RCA-one technique, that one-quarter of the total TSC was imported from the artery line to the venous bubble trap. The strengths of this study included that it was a multicenter, randomized, prospective clinical trial. For the efficiency analysis, we compared RCA-two with saline flushing and RCA-one, and the anticoagulation superiority was significant in RCA-two group, comparing with the saline and RCA-one group. The anticoagulation effect in the RCA-one group in this study was worse than in previous studies [11, 28]. One possible explanation for the disparity between the results observed in these trials were the methodological differences. Previous studies focused more on dialyzer clotting, but we considered the dialyzer, artery bubble, and venous bubble in this trial because any of the above aspects could interrupt the regular therapy. Another factor might be the different Ca2+ concentrations in the dialysate, wherein it was 1.5 mmol/L Ca2+ in our study, compared with 1.25 mmol/L Ca2+ in previous studies.
Dialysis session length is one of the essential factors associated with all-cause mortality among HD patients [2, 30], and anticoagulation efficiency would theoretically account for the dialysis session length. In this study, results of the ECC survival time indicated that RCA-two would provide the best guarantee of dialysis sufficiency. Frequent ECC high-pressure alarms and circuit replacement would enlarge nurse workloads and economic costs. The efficacy of hemodialysis was also assessed by calculating Kt/V and URR [31]. The Kt/V and URR results in this study demonstrated the best efficacy of the dialysis dose in RCA-two, although the Kt/V and URR were affected by other factors, such as the dialyzer type, blood flow, and access type.
Another strength of this study was that major confounders were analyzed based on the Cox regression models, affirming the independent anticoagulation role of RCA-two (Table 2). The previous trials were weak in this respect. Compared with the RCA-one group, adjusted HR in the RCA-two group was slightly increased from 0.211 to 0.224 in the Cox proportional hazards models. Compared with the saline group, adjusted HR was lower, from 0.222 to 0.184, suggesting that the anticoagulation of RCA-two was more effective after adjusting other factors (Table 2).
Generally, statistical differences in the safety population is not expected given the size sample. So a larger clinical cohort was needed to provide more robust data about the safety population. One-quarter of the TSC was imported back to the systemic circulation in the RCA-two procedure. Is it safe enough, especially in patients with a high risk of bleeding? The TSC is mainly metabolized in the liver and skeletal muscle in HD patients, liberating the Ca2+ and rapidly producing bicarbonate [32]. Theoretically, it would not accumulate in patients with normal liver and muscle metabolism. We had excluded patients with a serum bilirubin higher than 60 μmol/L as liver failure. There were six cases in RCA-two (8%) and two cases (5%) in RCA-one with a T/I Ca2+ concentration higher than 2.5 at 4 h. We also tested serum citrate concentration using a citrate assay kit, and the average serum citrate concentration of patients with a T/I Ca2+ concentration higher than 2.5 was 0.02 mmol/L (Range: 0.01-0.07 mmol/L), within the safety range (<1.0 mmol/L) according to previous recommendations. The correlation of serum citrate and T/I Ca2+ was 0.537 (0.123-0.791), P = 0.015 (data not shown). Although several patients were at risk of citrate overdose by the end of therapy, the T/I Ca2+ recovery occurred in all of them by the beginning of the next HD session 2 or 3 days later. The exact recovery time is unknown, which is also a limitation of this study.
Because TSC functions by binding Ca2+, hypocalcemia is a major complication during RCA, which will subsequently cause hypotension and convulsions and even serious arrhythmia [33]. In the modified RCA-two process, we aimed to simplify the procedure so that Ca2+ need not be monitored or supplemented into the systemic circulation. We determined whether the small portion of TSC imported directly back to the body in RCA-two would impact the serum Ca2+? As shown in Figure 4C and Figure 4D, the average concentration of iCa2+ was significantly lower in the RCA-two group, compared with the RCA-one and saline group, but the average concentration was above 0.9 mmol/L throughout the trial. Several patients developed asymptomatic hypocalcemia in RCA-two and RCA-one group demonstrated that patients were endurable with such complication. Hypomagnesemia, hypernatremia, and metabolic alkalosis were rare throughout the trial. Of note is that hyponatremia was observed in several participants, but we theorized that it was partly iatrogenic because we had reduced the concentration of sodium to 137 mmol/L in the dialysate. Other complications (e.g., hypotension, convulsions and hypoglycemia) did not differ among the groups. The major complication was hypotension, and this may be partly related to the larger ultrafiltration per hour with the infusion of TSC.
There are multiple limitations of this study that need to be mentioned. First, a larger clinical cohort is needed to provide more robust data about the safety population mentioned previously. Secondly, due to data limitations, we could not exclude all of the oral and hemostatic anticoagulants that participants were taking as a potential confounder and bias affecting the results [34, 35]. Other additional markers for coagulation activity such as D-dimer and fibrinogen [36, 37], which would be residual confounders, were not tested in this study.
Third, there were several inevitable biases in this trial. All the hospitals participating in this trial were high-grade hospitals in the urban area. The basic homogeneity of the included participants was relatively stable, and the main safety assessment indexes may be better than those in the real world. In this study, participants could voluntarily take part in one to three sessions during the trial, and therefore, for the open label study design, a choice bias exists to some extent; for example, participants with serious clotting may not be willing to continue for another session. To minimize the bias, additional assessment in separated data including only the first session of all participants showed that serious clotting events in the RCA-two group were still significantly lower than in either the RCA-one or saline group. (P < 0.05; Supplementary data, Figure S4).
Fourth, to minimize heterogeneity, we did not carry out the research in the real world, limiting its clinical application. For example, we did not observe the effectiveness of the hemodialysis filtration (HDF) or hemofiltration (HF) models, which are also major patterns of blood purification in our country.
Finally, our study design did not choice the 1:1:1 randomized trial. Because the sample size needed 699 dialysis sessions in three groups. In the furture, the crossover study design was an suitable alternative strategy to reduce the sample size.