Relationship between dialytic parameters and reviewer confirmed arrhythmias in hemodialysis patients in the monitoring in dialysis study

The design, objectives, and primary outcomes of the Monitoring in Dialysis (MiD) study have been reported previously [13, 14]. MiD (NCT01779856) was a prospective, multi-center study conducted in the US and India between January 2013 and September 2015 designed to characterize arrhythmias occurring in three times-weekly hemodialysis dialysis patients. An implantable cardiac monitoring device (Medtronic Reveal® XT or LINQ) was inserted to record heart rate and rhythm. Major inclusion criteria included the following: a) age ≥ 21 years and b) receiving in-center hemodialysis 3 times/week or with estimated glomerular filtration < 15 mL/min/1.73m² and expected to start dialysis within 2 months (no pre-dialysis patients were enrolled). Individuals with permanent pacemakers or implantable defibrillators and those not expected to remain on in-center dialysis for at least 6 months due to poor prognosis, expected transplant, or change in modality were excluded. All patients provided written informed consent, and institutional review board or ethics committee approval was obtained at each participating center. The trial was terminated when the last enrolled patient had completed the 6-month primary observation period.

ILR were downloaded and vital signs, dialysis prescription, and dialysis flow-sheet parameters were recorded at each dialysis session for 6 months. In addition, downloads were performed after any dialysis session with a blood draw for a total of up to 5 downloads each week. Programming parameters used for automated detection included the following: a) Fast ventricular tachycardia—180 beats per minute (BP) for 30/40 beats; b) ventricular tachycardia—130 BPM for 5 beats; c) bradycardia—40 BPM for 4 beats; d) asystole—duration of 3 s; e) atrial fibrillation detection on with no minimal duration.

Serum chemistries were tested before and after dialysis twice weekly for 4 weeks and then weekly through 6 months. ILR tracings that were marked by patients as symptomatic and those with potential arrhythmias were reviewed by the study sponsor by at least one trained individual with experience in cardiac signal interpretation and adjudication. Those potentially meeting the primary study endpoint, were subsequently adjudicated by a core lab.

The primary study endpoint of clinically significant arrythmia (CSA) has been previously reported [14] and included ventricular tachycardia > 115 beats/minute (modified by protocol amendment to improve detection specificity to ≥130 beats per minute) lasting ≥30 s, bradycardia with rate ≤ 40 beats per minute lasting ≥6 s, asystole for ≥3 s, and patient-marked (symptomatic) events where ECG review showed an arrhythmia considered clinically relevant. We report here, the secondary endpoint of reviewer confirmed arrhythmia (RCA). RCA was defined as an ILR identified or patient marked event in which a manual review of the stored ECG tracing confirmed the presence of abnormal rhythm RCA included outcomes meeting the primary endpoint definition, atrial fibrillation, supraventricular tachycardia, and sinus tachycardia with rate exceeding 130 beats per minute. In addition, ECG-confirmed ventricular tachycardia, asystole, and bradycardia of insufficient rate or duration to meet the primary endpoint definition were included. As for the primary endpoint of CSA, the RCA endpoint was chosen to define a set of events representing clinically relevant electrical instability and chronotropic dysfunction likely to share common physiologies and thus meriting joint analysis. ECG reviewers were not aware of patients characteristics or dialysis prescriptions. .

Baseline demographic, dialysis parameters, and laboratory characteristics are presented as mean ± standard deviation (SD), and median interquartile range (IQR) for continuous variables and percent (n/N) for categorical variables. Distributions in baseline characteristics were analyzed according to quartiles of the observed number of RCA events during follow-up in the primary analysis and according the dichotomous presence or absence of RCA during follow-up in a secondary analysis. Time averaged serum electrolyte concentrations or dialysis prescription parameters are presented as the mean or median of all sessions depending on normality. Characteristics between subjects with and without at least one RCA were compared using an unpaired t-test or Wilcoxon rank sum test for continuous variables, and Fisher’s exact tests for categorical variables. Where 3 or more groups were assessed ANOVA and Kruskal Wallis tests were used for continuous variables and Fisher’s exact test was used for categorical variables.

Negative binomial mixed-effect regression was used to analyze RCA rate over the dialytic week with division into 3 intra-dialytic periods and 14 inter-dialytic blocks of 12 h each. Repeated measures within subjects were accounted for using a random intercept that included an offset to indicate the time within each period. We excluded any off-schedule dialysis sessions as well as week in which dialysis was performed more or less than 3 times because the primary aim was to assess associations with the thrice weekly dialysis schedule. We used the lowest frequency period as the reference period for pairwise comparisons.

Given that dialytic parameters are most likely to directly induce arrythmia during dialysis or within a few hours after the conclusion of dialysis (and less likely to be responsible the further out one gets from dialysis), we investigated differences between sessions with RCA during the interval including dialysis or the 8 h immediately after dialysis using mixed effect models for continuous measures, or logistic regression (binary or multinomial) for categorical measures. The random effect of subject was included in both models. The interval represented the peak period of intra/post-dialysis arrythmia and was therefore the time period during which RCA were most likely to be biologically influenced by the dialysis procedure.

Negative binomial mixed effect regression was used to analyze associations of electrolytes, dialysis prescription, and intradialytic changes in fluid or electrolyte (flux, the difference between pre and post dialysis values) with RCA during the interval beginning with the start of each dialysis session through 8 h after dialysis. A random intercept was included to account for within-subjects repeated measures. Multivariable analyses were adjusted for age, sex, race, dialysis vintage, and vascular access and incidence rate ratios (IRR) were calculated. To avoid over-specification of the models, electrolyte and dialytic parameters of interest were added individually to this base model. First order interactions between volume removal (expressed as intra-dialytic weight change) and dialysate electrolyte concentration were analyzed to assess for effect modification by the extent of ultrafiltration given an extensive literature connecting ultrafiltration and risk of death [15]. All analyses were completed using SAS v 9.4 (Cary, NC) with P < 0.05 considered significant.

Overall, RCA were observed in 64 (97%) subjects during following-up (Table 1). Among individuals with ≤10, 11–49, 50–239 and ≥ 240 RCA events during follow-up, mean age was 61.1 ± 9.1, 56.3 ± 12.0, 54.1 ± 13.0 and 54.1 ± 13.7 years (P = 0.31). Other characteristics at baseline, such as race, sex, systolic and diastolic blood pressures, cause of ESRD, ESRD vintage, history of diabetes, and history of heart failure also did not differ significantly across quartiles of observed number of RCA events (P ≥ 0.14 for all comparisons). Similarly, there were no significant differences in baseline characteristics among those with (n = 64) and without (n = 2) RCA (Additional file 1: Table S1).

Among multiple baseline pre-dialysis laboratory tests examined, few differed significantly across quartiles of observed RCA (Table 2). However, use of high flux dialyzer was more frequent among those with a higher number of RCA (31.3, 56.3, 76.5, and 88.2% among those with ≤10, 11–49, 50–239, and ≥ 240 RCA, P = 0.004). Dialysate chemistries at baseline were similar across quartiles of observed RCA, but there was a non-significant trend towards increased use of higher dialysate potassium concentrations among those with more RCA (P = 0.05). Findings were similar when analyzed according to the dichotomous presence or absence of RCA during follow-up. with exception that the dialysate calcium concentration which was significantly higher (2.5 [IQR: 2.5, 2.5] vs. 1.6 [IQR: 1.6, 1.6], P = 0.04) in those with compared to those without RCA (Additional file 1: Table S2).

Given the high proportion of patients with RCA during follow-up, we analyzed the difference between sessions with and without RCA. There were 4154 sessions without RCA and 605 sessions with RCA during dialysis or during the 8 h after dialysis. As shown in Table 3, dialysis sessions without RCA were characterized by lower ultrafiltration rates (9.9 ± 4.9 vs. 10.1 ± 4.6 ml/kg/hour, P = 0.01) and intra-dialytic decrease in weight (2.6 ± 1.3 vs. 3.0 ± 1.3 kg P = 0.002) as well being closer to the target dry weight. There were no significant differences in pre-dialysis electrolyte concentrations or intra-dialytic change in electrolytes in sessions with or without RCA. However, dialysate temperature was lower in sessions without RCA (P < 0.001) and use of dialysate with 2.5 mEq/L of calcium was more common in sessions with RCA. Similar associations were present for dialysate temperature but not for indices of fluid accumulation or ultrafiltration rate when analyzing sessions with reviewer confirmed bradyarrhythmias (Additional file 1: Table S3). Differences in dialysate calcium were qualitatively similar but did not achieve significance. In contrast, both indices of fluid gain and removal as well dialysate calcium and temperature were different in sessions without and with reviewer confirmed tachyarrhythmias (Additional file 1: Table S4).

Time averaged laboratory and dialysis prescription parameters were analyzed to assess their distribution according to the number of RCA during follow up. (Additional file 1: Tables S5 & S6). Serum chemistries were notable for lower serum magnesium concentration over time in those with more RCA with concentrations of 2.6, 2.3, 2.3 and 2.2 mg/dL among those with ≤10, 11–49, 50–239, and ≥ 240 RCA, P = 0.01). Among dialysis parameters analyzed none were significantly different across categories of the numbers of observed RCA during follow-up. There were no other significant differences between serum chemistries or dialysis prescription parameters among individuals with and without RCA during follow-up.

As reported previously there 12,480 RCA detected in 64 (97%) patients at an overall rate of 33.7 (95% CI: 23.4, 48.7) per patient/month. Multiple episodes demonstrated more than a single type of arrythmia on the tracing, and the total included 7488 atrial arrythmias, 913 ventricular arrythmias, 1770 bradycardia events, 31 asystole, and 6065 sinus tachycardias. The atrial arrythmias included 4419 atrial fibrillation events [16]. There were clear temporal patterns with a nadir occurring 12–24 h after dialysis and an increase in rate during the last 12 h of the inter-dialytic interval through 12 h after dialysis (Fig. 1a). Additional, post-hoc exploration demonstrated that the majority of post-dialysis RCA occur within 8 h after dialysis, and that RCA rate during the next 4 h is not significantly different from that during the nadir period at 12–24 h after 3rd weekly session (Fig. 1b).

Given this distribution and noting that biologically dialytic parameters are most likely to directly induce arrythmia during dialysis or within a few hours after the conclusion of dialysis (and less likely to be responsible the further out one gets from dialysis), we hypothesized the effects of pre-dialysis electrolytes, the dialysis prescription, and the dialysis procedure would be most apparent during dialysis or the 8 h immediately (the period with maximal arrythmia incidence rate). Crude and adjusted associations with RCA rate during dialysis and the 8 h immediately afterwards are shown in Table 4. Only 4 parameters were associated with RCA rate in adjusted analyses: Higher pre-dialysis magnesium concentrations were associated with lower risks of RCA (IRR 0.49, 95% CI: 0.25, 0.94). Conversely, dialysate calcium concentrations below 2.0 mEq/L (IRR 0.13, 95% CI: 0.03, 0.57) and above 2.5 mEq/L were associated with lower risks of RCA compared to 2.5 mEq/L dialysate (IRR 0.52, 95% CI: 0.39, 0.70). Dialysate sodium concentrations of 135–139 mEq/L were also associated with a reduced rate compared to dialysate sodium of 140 mEq/L (IRR 0.40, 95% CI: 0.17, 0.95) as were dialysate bicarbonate concentrations > 35 mEq/L compared to concentrations of 35 mEq/L (IRR 0.51, 95% CI: 0.27, 0.97). Other factors were not significantly associated with RCA rate. Lastly, there was evidence for significant interaction between intradialytic change in potassium and change in weight (P_interaction = 0.01)—whereas RCA rate changed minimally with potassium flux when ≤2 kg were removed, there were sharp increases in RCA rate with greater potassium flux with ≥3 kg of ultrafiltration (Fig. 2).

Results were qualitatively similar for reviewer confirmed bradyarrhythmias and tachyarrhythmias when they were analyzed separately with 3 exceptions: Higher pre-dialysis phosphorous was associated with lower rates of tachyarrhythmia (IRR 0.45, 95% CI: 0.23, 0.91), higher pre-dialysis bicarbonate was associated with lower rates of bradycardia (IRR 0.73, 95% CI: 0.55, 0.98), and bradycardia was less frequent with large intradialytic weight change (IRR 0.58, 95% CI: 0.40, 0.83).