Modified regional citrate anticoagulation is optimal for hemodialysis in patients at high risk of bleeding: a prospective randomized study of three anticoagulation strategies

This investigator-initiated, multicenter, placebo-controlled, prospective, randomized clinical trial evaluated the efficacy and safety of RCA-two in a population of intermittent HD patients with a high risk of hemorrhage. Ethical approval of the study protocol was obtained from the institutional review boards of each participating hospital, including Guangdong Provincial People’s Hospital (Project Number: 2017250H (R1)), the First Affiliated Hospital of Guangzhou University of Chinese Medicine (Project Number: ZYYECK [2018]008), the Second Affiliated Hospital of Guangzhou Medicine University (Project Number: 2018-hs-06), and Guangzhou Hospital of Chinese Medicine (Project Number: B2017–189-01). The trial was prospectively registered at the international clinical trial registry system with the identifier GDREC2017250H (https://​www.​clinicaltrials.​gov/​ct2/​show/​NCT03419923?​term=​GDREC2017250H&​rank=​1).

Eligible patients were 18 years old or older and suffered from CKD requiring HD with a high risk of bleeding. The HD procedure should carried out with a dialysis fluid more than 500 ml/min and blood flow volume more than 3.5-4 ml/min.kg. The definition of high risk of bleeding was defined according to the criteria in previous reports [12] as active hemorrhage (within 3 days), pre-invasive operation (within 7 days), and post-operation (within 3 days). Patients were excluded if they had a high risk of citrate accumulation, defined as total bilirubin greater than 60 μmol/L, lactic acid > 3 mmol/L; a usage of the drugs that impact the coagulation function within 7 days; and severe hypocalcemia defined as a serum calcium concentration less than 1.9 mmol/L. Written informed consent was obtained from all subjects prior to enrollment and participation. This trial was complied with the Declaration of Helsink and adhered to the International Conference on Harmonisation Guidelines on Good Clinical Practice.

The primary outcome was measured as therapy interruption based on visible serious circuit clotting or persistent alarms such as venous pressure (> 200 mmHg) or TMP (> 300 mmHg). The secondary outcomes included circuit survival time (the lifetime of the circuit survival was limited to 240 min, because achieved the goals of treatment), the total clotting score of ECC and urea clearance (Kt/V and URR). The total clotting score was the sum of the clotting scores for the venous expansion chamber, arterial expansion chamber and dialyzer (Additional file 6: Table S1). The formula used for whole body urea clearance (Kt/V) was Kt/V = −In (C_post/C_pre - 0.008 * t) + (4–3.5 * C_post/C_pre) * UF/W, where C_post is the post-dialysis blood urea (mmol/liter), C_pre is the pre-dialysis blood urea (mmol/liter), t is the dialysis session length (hours), UF is the total volume of ultrafiltrate (liters), and W is the post-dialysis weight (kg). The formula used for urea excretion ratio (URR) was URR (%) = (C_pre -C_post)/C_pre*100, where C_post is the post-dialysis blood urea (mmol/liter) and C_pre is the pre-dialysis blood urea (mmol/liter).

Based on previous trials, the clinical risk of circuit clotting with RCA-one was approximately 36% during dialysis treatment compared with 45–50% with saline flushes [14, 15]. Our clinical preliminary experiment showed that the incidence of circuit clotting with RCA-two was 10%. According to the above data, we aimed to include 48 dialysis sessions in part one and 78 sessions in part two at 80% power and an error of 5%, using the PASS software (NCSS, version 11.0.7, LLC.).

Statistical analyses followed the protocol and the primary analysis was intention-to-treat and involved all patients who were randomly assigned. Categorical variables were described as frequencies (n) or percentages (%) and analyzed with Pearson’s chi-square or Fisher’s exact test. The Kolmogorov-Smirnov test was used to check the normal distribution of all continuous data. Parametric continuous parameters are expressed as mean ± standard deviation and analyzed with unpaired Student’s t-tests; nonparametric continuous parameters are expressed as medians (interquartile range, IQR) and analyzed with the Wilcoxon test.

A Kaplan-Meier survival curve was drawn to compare the circuit survival time using the log-rank test. The sensitivity analyses for the primary outcome (effect of RCA-two on the time to clotting) were performed used Cox proportional hazards models to adjust potential confounding variables. The main purpose of this analysis was to address potential differential competing risks from other causes of clotting between groups. To reduce the possibility that the same associated variable entered the multivariate model simultaneously, the correlation analysis between the factors screened using univariate analysis was performed. If the Spearman’s correlation coefficient between variables exceeds 0.60, only the variables considering being more important on a clinical basis were entered into the multivariate model. Variables with similar functions were chosen using the likelihood ratio test, which favored those variables with higher statistical values. If there were nonlinear association between continuous variables and the circuit survival, then the continuous variables were divided into clinically appropriate categories based on the lower quartile (Q_L) and upper quartile (Q_U). The main purpose of this analysis was to avoid the chances that the selected continuous variables violate linearity assumption. The Kaplan-Meier survival curves were drawn for each potential variables and the proportional hazards assumption on the basis of Schoenfeld residuals was tested. We accepted departure from the proportional hazards assumption as long as log-rank curves did not cross for the first 4 h.

The mixed models for repeated measures with an unstructured residual covariance matrix were used to analyze the change of concentrations of serum-ionized calcium from baseline to endpoint by restricted maximum likelihood [16]. The model included treatment of anticoagulation (group effect) and time (repeated measure for time effect) as fixed factors and baseline serum-ionized calcium as covariate, with interactions (interaction effect) between treatment and time included. Adjusted P-values for multiple Comparisons using modified Bonferroni method to analyze the differences between anticoagulation groups at each time points by unpaired Student’s t test.

Statistical significance was assumed if the 2-sided P-value was less than 0.05. The SAS software program was used for statistical analysis and data cleaning (SAS Institute, version 9.3, Cary, NC). The survival curves and other graphs were made using GraphPad Prism (GraphPad Software, version 6.1.0 for Windows, San Diego, CA, USA).

Between Sep 10, 2017, and June 6, 2018, a total of 120 patients were enrolled in the trial, and 23 patients were ineligible for the following reasons: six were unwilling to receive regional citrate anticoagulation, six were discharged from the hospital, three did not agree to give a blood sample, and eight patients met exclusion criteria. The remaining 97 patients were randomized into the trial. In part one of the experiment, 77 HD procedures in 52 patients were randomly assigned to the RCA-one or RCA-two group from Sep 10, 2017 to Mar 2, 2018; in part two of the experiment, 64 HD procedures in 45 patients were randomly assigned to the saline or RCA-two group from Nov 25, 2017 to Jun 6, 2018. Three patients unexpectedly terminated therapy for adverse events (AEs) of hypoglycemia and medical treatment (Fig. 2).

Baseline characteristics and baseline laboratory analyses were generally similar among the groups (Table 1; Additional file 7: Table S2). Compared with the RCA-two group, there were higher amounts of ultrafiltration in the saline group (P = 0.006) and more patients with catheters in the RCA-one group (P = 0.011). In the internal environment evaluation, the hemoglobin and pH value were not balanced among groups (P < 0.05) (Table 1).

Therapy-interrupted events based on serious clotting in the RCA-two group were significantly lower than in the RCA-one and saline groups (3/38 (7.89%) vs. 12/39 (30.77%), P = 0.011; 1/33 (3.03%) vs. 17/31 (54.84%), P < 0.001, respectively) (Fig. 3a). The percentage of sessions with persistent venous pressure (> 200 mmHg), TMP (> 300 mmHg) and visible serious circuit clotting in each group were presented in Additional file 1: Figure S1A and B. The median circuit survival time was 240 min (IQR 240 to 240) in the RCA-two group, which was significantly longer than 230 min (IQR 155 to 240, P < 0.001) in the RCA-one group and 210 min (IQR 135 to 240, P = 0.003) in the saline group (Fig. 3b).